Author response:

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

Although the reviewers found our work interesting, they raised several important concerns about our study. To address these concerns, mostly we performed new experiments. The most important changes are highlighted in the summary paragraphs.

First, in response to Reviewer 1’s suggestions, we have conducted the SFN experiments systematically, e.g., we further confirmed the mechanism of SFN-activated TFEB in HeLa NPC1 cells with new experiments including: the effect of BAPTA-AM (a calcium chelator), FK506+CsA (calcineurin inhibitors) and NAC (ROS scavenger) on SFN-induced TFEB-nuclear translocation in HeLa NPC1 cells (New Fig. S3). The effect of SFN on NPC1 expression (New Fig. S5). Particularly, we examined the colocalization of DiO (a PM marker) staining and surface LAMP1 staining in HeLa NPC1 cells under SFN treatment to confirm the PM exocytosis. In main text and figure legends, accuracy of sentence is thoroughly checked and defined. Hence, we have significantly improved the presentation and clarity in the revision.

Second, in response to Reviewer 2’s suggestions, we have performed additional experiments to demonstrate that the role of TFEB in SFN-evoked the lysosomal exocytosis by using TFEB-KO cells (New Fig. S7B). In TFEB KO cells, this increase of surface LAMP1 signal by SFN treatment was significantly reduced, suggestive of SFN-induced exocytosis in a TFEB-dependent manner. We also investigated the effect of U18666A on CF555-dextran endocytosis. By examining the localization of CF-dex and Lamp1, we found that CF555 is present in the lysosome with U18666A treatment (Fig for reviewers only A,B), suggesting that NPC1 deficiency/U18666A treatment has no effect on CF-dex endocytosis.

Third, in response to Reviewer 3’s suggestions, we have performed experiments in addition to response to other reviewers’ suggestion ie. the cytotoxicity of the concentration of SFN used in this study in various cell lines (New Fig.S10).

In addition, according to the reviewers’ suggestions, we made clarifications and corrections wherever appropriate in the manuscript.

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 raising this critical point! The reviewer is right that in this manuscript we did not talk too much about the molecular mechanism of SFN-evoked TFEB activation. Because in our previous study (Li, Shao et al. 2021), we explored the mechanism of SFN-induced TFEB activation. We show that SFN-evoked TFEB activation via a ROS-Ca²⁺-calcineurin dependent but MTOR -independent pathway (Li, Shao et al. 2021). In the current manuscript, we cited this paper, but did not talk the details of the mechanism, which obviously confused the reviewers. Therefore, in the revision manuscript we added more details of the molecular mechanism of SFN-activated TFEB. Also, we further confirmed this mechanism in HeLa NPC1 cells with new experiments including: the effect of BAPTA-AM (a calcium chelator), FK506+CsA (calcineurin inhibitors) and NAC (ROS scavenger) on SFN-induced TFEB-nuclear translocation in NPC cells (New Fig.S3).

(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 this constructive comment. From our data, in NPC1 cells SFN reduced the cholesterol levels by inducing lysosomal exocytosis and increasing lysosomal biogenesis. We understand the reviewer’s point that it would be really helpful to differentiate the exact three states of original number of lysosomes, number of new lysosomes, and number of lysosomes being exocytosis. Unfortunately, due to the technique limitation, so far seems there is no appropriate method that could clearly differentiate the lysosomes exactly come from which state. In the future, hopefully we will have technique to explore this mechanism.

(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 removed this sentence. Thanks!

Reviewer #2 (Public review):

(4) 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.

Thank you for this constructive suggestion. We examined the effect of two doses of SFN30 and 50mg/kg on NPC mice. 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 the future study, we will continue optimizing the dosage form and amount of SFN and do a dose-responsive analysis.

(5) Additional data supporting the activation of TFEB by SFN for cholesterol clearance in vivo would strengthen the overall impact of the study.

Thank the reviewer for this constructive comment. We have detected a significant decrease of pS211-TFEB protein in brain tissues of NPC mice upon SFN treatment compared to vehicle, suggesting that SFN activates TFEB in brain tissue for the first time. It is worth to further examine the lysosomal cholesterol levels in brain tissues to show the direct effect of SFN. However, in our hands and in the literatures Filipin seems not suitable for detecting lysosomal cholesterol accumulation in brain tissue. So far there isn’t a good method to directly measure lysosomal cholesterol in tissue.

(6) 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.

Great suggestion! We investigated the role of TFEB in SFN-evoked the lysosomal exocytosis by using TFEB-KO cells. As shown in New Suppl. Fig. 7B, in TFEB KO cells, this increase of surface LAMP1 signal by SFN (15 μM, 12 h) treatment was significantly reduced, suggestive of SFN induced exocytosis in a TFEB-dependent manner.

(7) 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 as stated in the methods and main text, we used pHrodo™ Green-Dextran (P35368, Invitrogen), rather than pHrodo Red-dextran. According to the product information from Invitrogen, pHrodo Green-dex conjugates are non-fluorescent at neural pH, but fluorescence bright green at acidic pH around 4, such as those in endosomes and lysosomes. Therefore, pHrodo Green-dex is suitable to monitor the acidity of lysosome (Hu, Li et al. 2022). We also used LysoTracker Red DND-99 (Thermo Scien fic, L7528) to measure lysosomal pH (Fig. 4G, H), which is consistent with results from pHrodo Green/CF measurement.

The reviewer mentioned that previous studies have reported an increase in lysosomal pH in U18666Atreated cells. We understood this concern. But in our hands, from our data with two lysosomal pH sensors, we have not detected lysosomal pH change in U18666A-treated NPC1 cell models.

(7) 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.

Thank you for raising this important point and suggestion was taken! We investigated the effect of NPC1 deficiency on CF555-dextran trafficking into lysosome by examining the localization of CF-dex and Lamp1. To clearly define whether CF555-dex is present in the lysosome, we first used apilimod to enlarge lysosomes and then examined the relative posi on of CF555-dex and lamp1. As shown in Author response image 1A,B, in HeLa cells treated with U18666A, CF555 signals (red) clearly present inside lysosome (LAMP1 labelled lysosomal membrane, green signal), suggesting that CF555dex endocytosis is not affected by NPC1 deficiency (U18666A treatment).

Author response image 1.

The effect of NPC1 deficiency on CF555 endocytosis. HeLa cells were transiently transfected with LAMP1-GFP plasmid for 24 h. Cells were then treated with apilimod (100 nM) for 2 h to enlarge the lysosomes, and followed by co- treatment of U18666A (2.5 μM, 24 h) and CF555 (12 h). (A)Each panel shows fluorescence images taken by confocal microscopes. (B) Each panel shows the fluorescence intensity of a line scan (white line) through the double labeled object indicated by the white arrow. Scale bar, 20 μm or 2 μm (for zoom-in images).

(9) 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 comments. Please see response to point #5.

Reviewer #3 (Public review):

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

Sorry for the defects in the writing and grammar. We had thoroughly checked grammar and polished the English to improve the manuscript.

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

We will address the technical issues mentioned in the following ques ons.

(12) 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.

Thank you for raising this point. The reviewer is right that not all the mCherry alone cells with the same level of filipin signal and not all mCherry-TFEBS211 transfected cells show completely no filipin signal. The statistical results were from randomly selected cells from 3 independent experiments. To avoid the confusion, we have included more cells in the statistical analysis to cover all the conditions as shown in the new Fig. 1B. Hopefully this helps to clarify the confusion.

(13) 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.

The reason that we use this concentration of SFN is based on our previous study (Li, Shao et al. 2021). We had shown that SFN (10–15 μM, 2–9 h) induces robust TFEB nuclear translocation in a dose- and time-dependent manner in HeLa cells as well as in other human cell lines without cytotoxicity (Li, Shao et al. 2021). Also, tissue concentrations of SFN can reach 3–30 μM upon broccoli consumption (Hu, Khor et al. 2006), so we used low micromolar concentrations of SFN (15 μM) in our study. Moreover, we further confirmed that SFN (15 μM) induces TFEB nuclear translocation in HeLa NPC1 cells (Fig. 1F, G Fig. 2B, G) and this concentration of SFN has no cytotoxicity (New Fig.S10).

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The following comments are designed to improve and focus the authors' work.

(14) Related to data in Figure 1. The mechanism through which TFEB can reduce Filipin in U18 conditions is unclear. Inhibi on of NPC1 results in hyperactivation of mTOR through cholesterol transport at ER-Lysosome contacts (see Zoncu group publications). If mTORC is hyperac ve in NPC disease models, TFEB would be expected to remain cytoplasmic and not enter the nucleus as the representative image in Figure 1A demonstrates.

In our previous study (Li, Shao et al. 2021), we have shown that SFN induces TFEB nuclear translocation in a mTOR-independent manner (Li, Shao et al. 2021). Consistent with this result, in this study we confirmed that SFN-induced TFEB nuclear translocation is mTor-independent in NPC1 cells (Now Fig. S4A, B). Thus, SFN induced TFEB nuclear translocation in various NPC cells (Fig. 1F, G, Fig. 2B, G). Please also see the discussion about the mechanism of SFN in response to point #1.

(15) Therefore, how does overexpression of TFEB, which remains in the cytoplasm, result in a decreased filipin signal? Similar ques ons relate to Figure 1C-H.

Medina et. al (Medina, Fraldi et al. 2011) show that TFEB overexpression (not activation, so overexpressed TFEB is in the cytoplasm) increases the pool of lysosomes in the proximity of the plasma membrane and promotes their fusion with PM by raising intracellular Ca²⁺ levels through lysosomal Ca²⁺ channel MCOLN1, leading to increased lysosomal exocytosis. Hence, TFEB overexpression only (TFEB is not activated) could reduce filipin signal via increasing lysosomal exocytosis. And with TFEB agonist treatment such as TFEB could further boost this increase.

(16) It would seem appropriate to measure the NPC1 and NPC2 proteins using western blot to ensure that SFN-dependent clearance of cholesterol is not due to enhanced expression of the native protein in U18-treated cells or enhanced folding of the protein in patient fibroblasts.

Thank you for this constructive comment! Because NPC1 gene mutation takes about 95% of NPC cases and NPC2 mutation takes about 5% of NPC cases. And in this study we focused on NPC1 deficiency cases. Thus, we measured the effect of SFN on the expression of NPC1 in human NPC1-patient fibroblasts. Western blot analysis showed that SFN (15 μM, 24 h) treatment did not affect NPC1 expression in human NPC1-patient fibroblasts (new Fig. S5).

(17) Related to data in Figures 1C-E. Controls are missing related to the effect SFN has on steady-state cholesterol levels. This may be insightful in providing information on the mode of action of this compound.

Suggestion was taken! We have supplemented the control- SFN only in new Fig. 1C-E.

(18) The mechanism that links SFN to TFEB-dependent translocation is suggested to involve calcineur independent dephosphorylation of TFEB. However, no data is provided. It would seem important to iden fy the mechanism(s) through which SFN positively regulates TFEB location. This would shift the manuscript and its model from correlations to causation. Experiments involving calcineurin inhibitors, or agonists of TRPML1 that have been reported as being a key source of Ca²⁺ for calcineurin activation, may provide molecular insight.

Please see the paragraph in response to point #1.

(19) Related to Figure 4. Using a plasma membrane counterstain to quantify plasma membrane LAMP1 would increase the rigor of the analysis.

Great idea! We examined the colocalization of DiO (a PM marker) staining and LAMP1 staining in HeLa NPC1 cells under SFN treatment. As shown in new Fig.4A, surface LAMP1 signal(red) colocalized with DiO (green), a PM marker.

(20) Related to Figure 5. How do the authors explain the kinetic disparity between SFN treatment for 24 vs 72 hrs? IF TFEB is activated and promoting lysosomal biogenesis and increased lysosomal flux across the PM, why does cholesterol accumulation lag? Perhaps related to this point. Are other cholesterol metabolizing enzymes that may have altered activity in NPC sensitive to SFN? A similar comment applies to the Sterol regulatory element binding protein pathway, which has been shown to be activated in models of NPC disease.

We understand the reviewer’s point. As shown in Fig. 5C, D, in NPC1^-/- MEF cells, SFN treatment for 24 h showed relative weaker cholesterol clearance compared to the effects in human cells (Fig.1C, D, Fig.2.E, I). Thus, we explored a longer treatment of SFN for 72 h (fresh SFN in medium was added every 24 h), and 72h treatment of SFN exhibited substantial cholesterol reduction (Fig. 5C, D). This different effect could be attributed to the continuous action of SFN, which could prolong the exocytosis, leading to more effective cholesterol clearance. As shown in the DMSO-treated MEF cells, the cholesterol levels are similar in both 24 and 72 h, thus 24 h U18666A treatment has reached the upper limit of the accumulated cholesterol, longer treatment me would not change the cholesterol levels. Thus, cholesterol accumulation has no lag.

We did not investigate whether SFN regulates other cholesterol metabolizing enzymes or sterol regulatory element binding proteins although we cannot rule out this possibility. In this study we mainly focus on the cholesterol clearance effect by SFN via TFEB-mediated pathways. From our data, TFEB KO could significantly diminish SFN-evoked cholesterol clearance. Hence, the effect of other cholesterol metabolizing enzymes or sterol regulatory element binding proteins maybe not as important as TFEB, thus out of scope of this study. In the future, we may explore the involvement of possible other pathways on SFN’s effects.

(21) Related to Figure 7. The western blots for pS211-TFEB are poor. It's suggested that whole blots are shown to increase rigor.

Thank you for the comments. We have represented the blots with more spare space to increase the rigor.

(22) Data demonstrating the ability of SFN to improve Purkinje cell survival are exci ng and pair well with the weight analysis, however, to address the overall goal of determining if "SFN could be a good therapeutic candidate for neuropathology in NPC disease" survival analysis should be tested as well.

Please see the paragraph in response to point #3.

Minor

(23) Throughout the manuscript many different Fonts and font sizes are used. This is very jarring to readers. It is suggested that a more uniform approach is taken to presenting these nice datasets.

We are so sorry and apologize for these oversights. We have thoroughly checked all the manuscript to make sure that Fonts and sizes of font are synchronized.

(24) Related to data presentation. In general, there is a lack of alignment and organization of the figures.

So sorry about this. We have reorganized the figures to get them better aligned.

(25) Line 149, SFN is missing.

Corrected!

Reviewer #3 (Recommendations for the authors):

(26) In Figure 3 the authors should use multiple single siRNAs or perform a functional rescue to determine specificity.

We understand the reviewer’s point. We did design several siRNAs and the efficiency of these siRNAs were validated. Finally, we decide use this siRNA whose knockdown efficiency is best in the study and the specificity of the siTFEB has been validated by Western blot as shown in Fig. 3A. Furthermore, we used TFEB knockout cells constructed by CRISPR/Cas9 to further examine the role of TFEB in SFN-induced cholesterol clearance (Fig. 3D). Consistently with the results in the siTFEB-transfected HeLa NPC1 cells (Fig. 3B, C), SFN failed to diminish cholesterol in HeLa TFEB KO cells. The result from TFEB KO cells is even convincing than siRNA experiment. We also performed a functional rescue of re-expressing TFEB in TFEB KO cells, in which SFN-induced cholesterol clearance was restored (Fig. 3E, F). Collectively, these data indicate that TFEB is required for lysosomal cholesterol reduction upon SFN treatment. Thus, we did not repeat this rescue experiment in the siTFEB-transfected HeLa NPC1 cells.

(27) The label for 3D is missing.

Corrected! Thanks!

(28) Figure 4, although the authors use an an body against the luminal domain of LAMP1 there could s ll be some permeabilization. A marker of the plasma membrane would be helpful.

Please see the response to point #19.

(29) Figure 4, cholesterol in the media because of lysosome exocytosis. This is where the high concentration of SFN is of concern. Is there any cell death that could explain the result? The authors should test for cell death with the SFN treatment.

Thank you for raising this important point! We have measured the cytotoxicity of SFN of the concentrations used in this study in various cell lines (New Fig.S10). Please also see the paragraph in response to point #13.

(30) The blot in Figure 6A is unclear. It is very hard to see any change in pS211-TFEB levels, and, the blurry signal is the detection of phospho-TFEB is uncertain.

Please see the summary paragraph in response to point #21.

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-+.

Hu, R., T. O. Khor, G. Shen, W. S. Jeong, V. Hebbar, C. Chen, C. Xu, B. Reddy, K. Chada and A. N. Kong (2006). "Cancer chemoprevention of intestinal polyposis in ApcMin/+ mice by sulforaphane, a natural product derived from cruciferous vegetable." Carcinogenesis 27(10): 2038-2046.

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.

Medina, D. L., A. Fraldi, V. Bouche, F. Annunziata, G. Mansueto, C. Spampanato, C. Puri, A. Pignata, J. A. Martina, M. Sardiello, M. Palmieri, R. Polishchuk, R. Puertollano and A. Ballabio (2011). "Transcriptional activation of lysosomal exocytosis promotes cellular clearance." Dev Cell 21(3): 421-430.

eLife Assessment

This revision of important work is a versatile addition to the chemical protein modifications and bioconjugation toolbox in synthetic biology. The technology developed cleverly uses Connectase to irreversibly fuse proteins of interest together so they can be studied in their native context, with compelling well-controlled data showing the technique works for various protein partners. This work will help multiple fields to explore multi-function constructs in basic synthetic biology. This work will also be of interest to those studying fusion oncoproteins commonly expressed in various human pathologies.

Reviewer #1 (Public review):

Fuchs describes a novel method of enzymatic protein-protein conjugation using the enzyme Connectase. The author is able to make this process irreversible by screening different Connectase recognition sites to find an alternative sequence that is also accepted by the enzyme. They are then able to selectively render the byproduct of the reaction inactive, preventing the reverse reaction, and add the desired conjugate with the alternative recognition sequence to achieve near-complete conversion. I agree with the authors that this novel enzymatic protein fusion method has several applications in the field of bioconjugation, ranging from biophysical assay conduction to therapeutic development. Previously the author has published on the discovery of the Connectase enzymes and has shown its utility in tagging proteins and detecting them by in-gel fluorescence. They now extend their work to include the application of Connectase in creating protein-protein fusions, antibody-protein conjugates, and cyclic/polymerized proteins. As mentioned by the author, enzymatic protein conjugation methods can provide several benefits over other non-specific and click chemistry labeling methods. Connectase specifically can provide some benefits over the more widely used Sortase, depending on the nature of the species that is desired to be conjugated. Overall, this method provides a novel, reproducible way to enzymatically create protein-protein conjugates.

The manuscript is well-written and will be of interest to those who are specifically working on chemical protein modifications and bioconjugation.

Comments on revisions:

The authors have improved the manuscript significantly by clarifying the questions raised adding new text, providing additional references and/or adding additional data. The thorough study and efficiency of the method for enzymatic protein-protein conjugation using the enzyme Connectase warrants publication of this manuscript in its current form.

Reviewer #2 (Public review):

Summary:

Unlike previous traditional protein fusion protocols, the author claims their proposed new method is fast, simple, specific, reversible, and results in a complete 1:1 fusion. A multi-disciplinary approach from cloning and purification, biochemical analyses, and proteomic mass spec confirmation revealed fusion products were achieved.

Strengths:

The author provides convincing evidence that an alternative to traditional protein fusion synthesis is more efficient with 100% yields using connectase. The author optimized the protocol's efficiency with assays replacing a single amino acid and identification of a proline aminopeptidase, Bacilius coagulans (BcPAP), as a usable enzyme to use in the fusion reaction. Multiple examples including Ubiquitin, GST, and antibody fusion/conjugations reveal how this method can be applied to a diverse range of biological processes.

Weaknesses:

Though the ~100% ligation efficiency is an advancement, the long recognition linker may be the biggest drawback. For large native proteins that are challenging/cannot be synthesized and require multiple connectase ligation reactions to yield a complete continuous product, the multiple interruptions with long linkers will likely interfere with protein folding, resulting in non-native protein structures. This method will be a good alternative to traditional approaches as the author mentioned but limited to generating epitope/peptide/protein tagged proteins, and not for synthetic protein biology aimed at examining native/endogenous protein function in vitro.

Author response:

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

Reviewer #1 (Public review):

Fuchs describes a novel method of enzymatic protein-protein conjugation using the enzyme Connectase. The author is able to make this process irreversible by screening different Connectase recognition sites to find an alternative sequence that is also accepted by the enzyme. They are then able to selectively render the byproduct of the reaction inactive, preventing the reverse reaction, and add the desired conjugate with the alternative recognition sequence to achieve near-complete conversion. I agree with the authors that this novel enzymatic protein fusion method has several applications in the field of bioconjugation, ranging from biophysical assay conduction to therapeutic development. Previously the author has published on the discovery of the Connectase enzymes and has shown its utility in tagging proteins and detecting them by in-gel fluorescence. They now extend their work to include the application of Connectase in creating protein-protein fusions, antibody-protein conjugates, and cyclic/polymerized proteins. As mentioned by the author, enzymatic protein conjugation methods can provide several benefits over other non-specific and click chemistry labeling methods. Connectase specifically can provide some benefits over the more widely used Sortase, depending on the nature of the species that is desired to be conjugated. However, due to a similar lengthy sequence between conjugation partners, the method described in this paper does not provide clear benefits over the existing SpyTag-SpyCatcher conjugation system.  Additionally, specific disadvantages of the method described are not thoroughly investigated, such as difficulty in purifying and separating the desired product from the multiple proteins used. Overall, this method provides a novel, reproducible way to enzymatically create protein-protein conjugates.

The manuscript is well-written and will be of interest to those who are specifically working on chemical protein modifications and bioconjugation.

I'd like to comment on two points.

(1) The benefits over the SpyTag-SpyCatcher system. Here, the conjugation partners are fused via the 12.3 kDa SpyCatcher protein, which is considerably larger than the Connectase fusion sequence (19 aa). This is mentioned in the introduction (p. 1 ln 24-26). Furthermore, SpyTag-SpyCatcher fusions are truly irreversible, while Connectase/BcPAP fusions may be reversed (p. 8, ln 265-273). For example, target proteins (e.g., AGAFDADPLVVEI-Protein) may be covalently fused to functionalized magnetic beads (e.g., Bead-ELASKDPGAFDADPLVVEI) in order to perform a pulldown assay. After the assay, the target protein and any bound interactors could be released from the beads by the addition of a Connectase / peptide (AGAFDAPLVVEI) mixture.

In a related technology, the SpyTag-SpyCatcher system was split into three components, SpyLigase, SpyTag and KTag  (Fierer et al., PNAS 2014). The resulting method introduces a sequence between the fusion partners (SpyTag (13aa) + KTag (10aa)), which is similar in length to the Connectase fusion sequence (p. 8, ln 297 - 298). Compared to the original method, however, this approach seems to require longer incubation times, while yielding less fusion product (Fierer et al., Figure 2).

(2) Purification of the fusion product. The method is actually advantageous in this respect, as described in the discussion (p. 8, ln 258-264). Examples are now provided in Figure 6.

Reviewer #2 (Public review):

Summary:

Unlike previous traditional protein fusion protocols, the author claims their proposed new method is fast, simple, specific, reversible, and results in a complete 1:1 fusion. A multi-disciplinary approach from cloning and purification, biochemical analyses, and proteomic mass spec confirmation revealed fusion products were achieved.

Strengths:

The author provides convincing evidence that an alternative to traditional protein fusion synthesis is more efficient with 100% yields using connectase. The author optimized the protocol's efficiency with assays replacing a single amino acid and identification of a proline aminopeptidase, Bacilius coagulans (BcPAP), as a usable enzyme to use in the fusion reaction. Multiple examples including Ubiquitin, GST, and antibody fusion/conjugations reveal how this method can be applied to a diverse range of biological processes.

Weaknesses:

Though the ~100% ligation efficiency is an advancement, the long recognition linker may be the biggest drawback. For large native proteins that are challenging/cannot be synthesized and require multiple connectase ligation reactions to yield a complete continuous product, the multiple interruptions with long linkers will likely interfere with protein folding, resulting in non-native protein structures. This method will be a good alternative to traditional approaches as the author mentioned but limited to generating epitope/peptide/protein tagged proteins, and not for synthetic protein biology aimed at examining native/endogenous protein function in vitro.

The assessment is fair, and I have no further comments to add.

Reviewer #1 (Recommendations for the authors):

Major/Experimental Suggestions:

(1) Throughout the paper only one reaction shown via gels had 100% conversion to desired product (Figure 3C). It is misleading to title a paper with absolutes such as "100% product yield", when the majority of reactions show >95% product yield, without any purification. Please change the title of the manuscript to something along the lines of "Novel Irreversible Enzymatic Protein Fusions with Near-Complete Product Yield".

The conjugation reaction is thermodynamically favored. It is driven by the hydrolysis of a peptide bond (P|GADFDADPLVVEI), which typically releases 8 - 16 kJ/mol energy. This should result in a >99.99% complete reaction (DG° = -RT ln (Product/Educt)). In line with this, 99% - 100% of the less abundant educts (LysS, Figure 3A; MBP, Figure 3B; Ub-Strep, Figure 3C) are converted in the time courses (Figure 3D-F show different reaction conditions, which slow down conjugate formation). 100% conversion are also shown in Figure 5, Figure 6, and Figure S4. Likewise, 99.6% relative fusion product signal intensity in an LCMS analysis (Figure S2) after 4h reaction time (0.13% and 0.25% educts). In this experiment, the proline had been removed from 99.8% of the peptide byproducts (P|GADFDADPLVVEI). It is clear that this reaction is still ongoing and that >99.99% of the prolines will be removed from the peptides in time. These findings suggest that the conjugation reaction gradually slows down the less educt is available, but eventually reaches completion.

For some experiments, lower product yields (e.g. 97% in Figure 3B) are reported in the paper. These were calculated with Yield = 100% x Product / (Educt1 + Educt 2 + Product). With this formula, 100% conjugation can only be achieved with exactly equimolar educt quantities, because both educt 1 and educt 2 need to be converted entirely. If one educt 1 is available in excess, for example because of protein concentration measurement inaccuracies or pipetting errors, some of it will be left without fusion partner. In case of Figure 3B, 3% more GST seemed to have been in the mixture. These are methodological inaccuracies.

(2) Please provide at least one example of a purified desired product, and mention the difficulties involved as a disadvantage to this particular method. Separating BcPAP, Connectase, and the desired protein-protein conjugate may prove to be quite difficult, especially when Connectase cleaves off affinity tags.

Examples are now provided in Figure 6. As described in the discussion (p. 8, ln 258-264), the simple product purification is one of the advantages of the method.

(3) For the antibody conjugate, please provide an example of conjugating an edduct that would prove to be more useful in the context of antibodies. For example, as you mention in the introduction, conjugation of fluorophores, immobilization tags such as biotin, and small molecule linker/drugs are useful bioconjugates to antibodies.

Antibody-biotinylation is now shown in Figure S6; Antibody-fluorophore conjugates are part of Figures S5 and S7.

(4) Please assess the stability of these protein-protein conjugates under various conditions (temperature, pH, time) to ensure that the ligation via Connectase is stable over a broad array of conditions. In particular, a relevant antibody-conjugate stability assay should be done over the period of 1-week in both buffer and plasma to show applicability for potential therapeutics.

The stability of an antibody-biotin conjugate in blood plasma over 7 days at different temperatures is now shown in Figure S7.

Generally, Connectase introduces a regular peptide bond (Asp-Ala) with a high chemical and physical stability (e.g. 10 min incubation at 95°C in SDS-PAGE loading buffer; H2O-formic acid / acetonitrile gradients for LC-MS). The sequence may be susceptible to proteases, although this is not the case in HEK293 cells (antibody expression), E. coli, or blood plasma (Figure S7).

(5) Please conduct functional assays with the antibody-protein/peptide conjugates to show that the antibody retains binding capabilities to the HER-2 antigen and the modification was site-selective, not interfering with the binding paratope or binding ability of the antibody in any way. This can be done through bio-layer interferometry, surface plasmon resonance, ELISA, etc.

We plan the immobilization of the HER2 antibody on microplates and its use in an ELISA. However, this experiment requires significant testing and optimizations. It will be part of a future paper on the use of Connectase for protein immobilization.

For now, the mass spectrometry data provide clear evidence of a single site-selective conjugation, as the C-terminal ELASKDPGAFDADPLVVEI-Strep sequence is replaced by ELASKDAGAFDADPLVVEI(-Ub). Given that the conjugation sites at the C-termini are far from the antigen binding sites, and have already been used in a number of other approaches (e.g., SpyTag, SnapTag, Sortase), it appears unlikely that these conjugations interfere with antigen binding.

(6) Please include gels of all proteins used in ligation reactions after purification steps in the SI to show that each species was pure.

The pure proteins are now shown in Figure S9.

(7) Please provide the figures (not just tables) of LC/MS deconvoluted mass spectra graphs for all conjugates, either in the main text or the SI.

Please specify which spectra you are missing. I believe all relevant spectra are shown in Figures 4, 5, and S3. The primary data can be found in Dataset S2.

(8) Please provide more information in the methods section on exactly how the densitometry quantification of gel bands was performed with ImageJ.

Details on the quantification with Image Studio Lite 5.2 were added in the method section (p. 17, ln 461-463).

Minor Suggestions:

(1) Page 1, line 19: can include one sentence on what assays these particular bioconjugations are usefule for (e.g. internalization cell studies, binding assays, etc.)

I prefer not to provide additional details here to keep the text concise and focused.

(2) Page 1, line 22: "three to ten equivalents" instead of 3x-10x.

Done.

(3) Page 1, line 23: While NHS labeling is widely considered non-specific, maleimide conjugation to free cysteines is generally considered specific for engineered free cysteine residues, since native proteins often do not have free cysteine residues available for conjugation. If you are referring to the potential of maleimides to label lysines as well, that should be specifically stated.

I modified the sentence, now stating that these methods are "can be" unspecific.

As pointed out, it is possible to achieve specificity by eliminating all other free cysteines and/or engineering a cysteine in an appropriate position. In many other cases, however (e.g., natural antibodies), several cysteines are available, or the sample contains other proteins/peptides. I did not want to go into more detail here and refer to the cited review.

(4) Page 1, line 31: "and an oligoglycine G(1-5)-B"

Done.

(5) Page 1, line 34: It is not clear where in the source these specific Km values are coming from, considering these are variable based on specific conditions/substrates and tend to be reaction-specific.

I cited another review, which lists the same values, along with a few other measurements (Jacobitz et al., Adv Protein Chem Struct Biol 2017, Table 2). It is clear that each of these measurements differs somewhat, but they are generally comparable (K_M(LPETG) = 5500 - 8760 µM; K_M(GGGGG) = 140 - 196 µM). I chose the cited study (Frankel et al., Biochemistry 2005), because it also investigated hydrolysis rates. In this study, the measurements are derived from the plots in Figure 2.

(6) Page 1, line 47: the comparison to western blots feels a little like apples to oranges, even though this comparison was made in previous literature. Engineering an expressed protein to have this tag and then using the tag to detect and quantify it, feels more akin to a tagging/pull down assay than a western blot in which unmodified proteins are easily detected.

It is akin to a frequently used type of western blots with tag-specific antiboies, e.g. Anti-His₆, -Streptavidin, -His₆, -HA ,-cMyc, -Flag. I modified the sentence to clarify this.

(7) Page 2, line 51: "Connectase cleaves between the first D and P amino acids in the recognition sequence, resulting in an N-terminal A-ELASKD-Connectase intermediate and a C-terminal PGAFDADPLVVEI peptide."

I prefer the current sentence, because we assume that a bond between the aspartate and Connectase is formed before PGAFDADPLVVEI is cleaved off.

(8) Page 3, line 94: "Exact determination is not possible due to reversibility of the reaction", the way it is stated now sounds like it is a flaw in the methods. Also, update Figure 2 to read "Estimated relative ligation rate".

Done.

(9) Page 3, lines 101-107: This is worded in a confusing way. It can either be X₁ or X₂ that is inactivated depending on if the altered amino acid is on the original protein sequence or on the desired edduct to conjugate. You first give examples of how to render other amino acids inactive, but then ultimately state that proline made inactive, so separate the two distinct possibilities a bit more clearly.

The reaction requires the inactivation of X₁, without affecting X₂ (ln 100 - 102). This is true, no matter whether it is X₁ = A, C, S, or P that is inactivated. I added a sentence to clarify this (ln 102 – 103).

(10) Page 4, line 118: Give a one-sentence justification for why these proteins were chosen to work with (easy to express, stable, etc).

Done.

(11) Page 5, line 167: "payload molecules".

Done.

(12) Page 5, lines 170-173: Word this more clearly- "full conversion with many of these methods is difficult on antibodies due to each heavy and light chain being modified separately, resulting in only a total yield of 66% DAR4 even when 90% of each chain is conjugated."

I rephrased the section.

(13) Page 8, line 290: Discuss other disadvantages of this method including difficulties purifying and in incorporating such a long sequence into proteins of interest.

Product purification is shown in the new Figure 6. As stated above, I consider the simple purification process an advantage of the method.  The genetic incorporation of the sequence into proteins is a routine process and should not make any difficulties. The disadvantages of long linker sequences between fusion partners are now discussed (p.8 – 9, ln 300-302).

(14) Page 10, line 341: 'The experiment is described and discussed in detail in a previously published paper.31"

Done.

Reviewer #2 (Recommendations for the authors):

Minor Points:

(1) It's unclear how the author derived 100% ligation rate with X = Proline in Figure 2 when there is still residual unligated UB-Strep at 96h. Please provide an expanded explanation for those not familiar with the protocol. Is the assumption made that there will be no UB-Strep if the assay was carried out beyond 96h?

I clarified the figure legend. The assay shows the formation of an equilibrium between educts and products. Therefore, only ~50% Ub-Strep is used with X = Proline (see p. 2, ln 79 - 81). The "relative ligation rate" refers to the relative speed with which this equilibrium is established. The highest rate is seen with X = Proline, and it is set to 100%. The other rates are given relative to the product formation with X = Proline.

(2) Though the qualitative depiction of the data in Figure 3 is appreciated, an accompanying graphical representation of the data in the same figure will greatly enhance reception and better comprehension of several of the author's conclusions.

Graphs are now shown in Figure S1.

(3) Figure 3 panel E is misaligned. Please align it with panel B above it.

Done, thank you.

(4) The author refers to 'The resulting circular assemblies (37% UB2...)' in the text but identifies it as UB-C2 in Figure 5B. Is this a mistake or does UB2 refer to another assembly not mentioned in the Figures? Please check for inconsistencies.

All circular assemblies are now labeled Ub-C _1-6.

(5) Finishing with a graphical schematic that depicts the entire protocol in a simple image would be much appreciated and well-received by readers. Including the scheme with A and B proteins, the recognition linkers, the addition of connectase and BcPAP, etc. to the final resulting protein with connected linker.

A graphical summary of the reaction is now included in Figure 6.

eLife Assessment

This manuscript addresses a mechanism by which dopamine (DA) regulates synaptic plasticity. The authors build upon their previous finding that DA applied after a timing pattern that ordinarily induces long-term depression (LTD) now induces long-term potentiation (LTP). The new findings that this "DA-dependent LTP" involves de novo protein synthesis, a cyclicAMP signalling pathway, and calcium-permeable AMPA receptors (CP-AMPARs) are of valuable significance. The conclusions are convincing and largely supported by the evidence provided.

Reviewer #1 (Public review):

Summary:

In this manuscript, Fuchsberger et al. demonstrate a set of experiments which ultimately identifies the de novo synthesis of GluA1-, but not GluA2-containing Ca2+ permeable AMPA receptors as a key driver of dopamine-dependent LTP (DA-LTP) during conventional post-before-pre spike-timing dependent (t-LTD) induction. The authors further identify adenylate cyclase 1/8, cAMP, and PKA as the crucial mitigators of these actions. While some comments have been identified below, the experiments presented are thorough and address the aims of the manuscript, figures are presented clearly (with minor comments), and experimental samples sizes and statistical analyses are suitable. Suitable controls have been utilized to confirm the role of Ca2+ permeable AMPAR. This work provides a valuable step forward built on convincing data towards understanding the underlying mechanisms of spike-timing dependent plasticity and dopamine.

Strengths:

Appropriate controls were used.

The flow of data presented is logical and easy to follow.

The quality of the data is solid.

Weaknesses:

Our concerns raised within the first round of review have been appropriately addressed by the authors.

Reviewer #2 (Public review):

Summary:

The aim was to identify the mechanisms that underlie a form of long-term potentiation (LTP) that requires activation of dopamine (DA).

Strengths:

The authors have provided multiple lines of evidence that supports their conclusions; namely that this pathway involves activation of a cAMP / PKA pathway that leads to the insertion of calcium permeable AMPA receptors.

Weaknesses:

Some of the experiments could have been conducted in a more convincing manner.

Reviewer #3 (Public review):

The manuscript of Fuchsberger et al. investigates the cellular mechanisms underlying dopamine-dependent long-term potentiation (DA-LTP) in mouse hippocampal CA1 neurons. The authors conducted a series of experiments to measure the effect of dopamine on the protein synthesis rate in hippocampal neurons and its role in enabling DA-LTP. The key results indicate that protein synthesis is increased in response to dopamine and neuronal activity in the pyramidal neurons of the CA1 hippocampal area, mediated via the activation of adenylate cyclases subtypes 1 and 8 (AC1/8) and the cAMP-dependent protein kinase (PKA) pathway. Additionally, the authors show that postsynaptic DA-induced increases in protein synthesis are required to express DA-LTP, while not required for conventional t-LTP.

The increased expression of the newly synthesized GluA1 receptor subunit in response to DA supports the formation of homomeric calcium-permeable AMPA receptors (CP-AMPARs). This evidence aligns well with data showing that DA-LTP expression requires the GluA1 AMPA subunit and CP-AMPARs, as DA-LTP is absent in the hippocampus of a GluA1 genetic knock-out mouse model.

Comments on revisions:

The authors addressed adequately all my comments.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Fuchsberger et al. demonstrate a set of experiments that ultimately identifies the de novo synthesis of GluA1-, but not GluA2-containing Ca2+ permeable AMPA receptors as a key driver of dopamine-dependent LTP (DA-LTP) during conventional post-before-pre spike-timing dependent (t-LTD) induction. The authors further identify adenylate cyclase 1/8, cAMP, and PKA as the crucial mitigators of these actions. While some comments have been identified below, the experiments presented are thorough and address the aims of the manuscript, figures are presented clearly (with minor comments), and experimental sample sizes and statistical analyses are suitable. Suitable controls have been utilized to confirm the role of Ca2+ permeable AMPAR. This work provides a valuable step forward built on convincing data toward understanding the underlying mechanisms of spike-timing-dependent plasticity and dopamine.

Strengths:

Appropriate controls were used.

The flow of data presented is logical and easy to follow.

The quality of the data, except for a few minor issues, is solid.

Weaknesses:

The drug treatment duration of anisomycin is longer than the standard 30-45 minute duration (as is the 500uM vs 40uM concentration) typically used in the field. Given the toxicity of these kinds of drugs long term it's unclear why the authors used such a long and intense drug treatment.

In an initial set of control experiments (Figure S 1C-D) we wanted to ensure that protein synthesis was definitely blocked and therefore used a relatively high concentration of anisomycin and a relatively long pre-incubation period. We agree with the Reviewer that we cannot exclude the possibility that this treatment could compromise cell health in addition to the protein synthesis block. Therefore, we carried out an additional experiment with an alternative protein synthesis inhibitor cycloheximide at a lower standard concentration (10 µM) which confirmed a significant reduction in the puromycin signal (Figure S 1A-B). Together these results support the conclusion that puromycin signal is specific to protein synthesis in our labelling assay.

Furthermore, in the electrophysiology experiments, we used 500 μM anisomycin in the patch pipette solution. Under these conditions, we recorded a stable EPSP baseline for 60 minutes, indicating that the treatment did not cause toxic effects to the cell (Figure S1F). This high concentration would ensure an effective block of local translation at dendritic sites. Nevertheless, we also carried out this experiment with cycloheximide at a lower standard concentration (10 µM) and observed a similar result with both protein synthesis inhibitors (Figure 1F).

With some of the normalizations (such as those in S1) there are dramatic differences in the baseline "untreated" puromycin intensities - raising some questions about the overall health of slices used in the experiments.

We agree with the Reviewer that there is a large variability in the normalised puromycin signal which might be due to variability in the health of slices. However, we assume that the same variability would be present in the treated slices, which showed, despite the variability, a significant inhibition of protein synthesis. To avoid any bias by excluding slices with low puromycin signal in the control condition, we present the full dataset.

The large set of electrophysiology experiments carried out in our study (all recorded cells were evaluated for healthy resting membrane potential, action potential firing, and synaptic responses) confirmed that, generally, the vast majority of our slices were indeed healthy. 

Reviewer #2 (Public Review):

Summary:

The aim was to identify the mechanisms that underlie a form of long-term potentiation (LTP) that requires the activation of dopamine (DA).

Strengths:

The authors have provided multiple lines of evidence that support their conclusions; namely that this pathway involves the activation of a cAMP / PKA pathway that leads to the insertion of calcium-permeable AMPA receptors.

Weaknesses:

Some of the experiments could have been conducted in a more convincing manner.

We carried out additional control experiments and analyses to address the specific points that were raised.

Reviewer #3 (Public Review):

The manuscript of Fuchsberger et al. investigates the cellular mechanisms underlying dopamine-dependent long-term potentiation (DA-LTP) in mouse hippocampal CA1 neurons. The authors conducted a series of experiments to measure the effect of dopamine on the protein synthesis rate in hippocampal neurons and its role in enabling DA-LTP. The key results indicate that protein synthesis is increased in response to dopamine and neuronal activity in the pyramidal neurons of the CA1 hippocampal area, mediated via the activation of adenylate cyclases subtypes 1 and 8 (AC1/8) and the cAMP-dependent protein kinase (PKA) pathway. Additionally, the authors show that postsynaptic DA-induced increases in protein synthesis are required to express DA-LTP, while not required for conventional t-LTP.

The increased expression of the newly synthesized GluA1 receptor subunit in response to DA supports the formation of homomeric calcium-permeable AMPA receptors (CP-AMPARs). This evidence aligns well with data showing that DA-LTP expression requires the GluA1 AMPA subunit and CP-AMPARs, as DA-LTP is absent in the hippocampus of a GluA1 genetic knock-out mouse model. Overall, the study is solid, and the evidence provided is compelling. The authors clearly and concisely explain the research objectives, methodologies, and findings. The study is scientifically robust, and the writing is engaging. The authors' conclusions and interpretation of the results are insightful and align well with the literature. The discussion effectively places the findings in a meaningful context, highlighting a possible mechanism for dopamine's role in the modulation of protein-synthesis-dependent hippocampal synaptic plasticity and its implications for the field. Although the study expands on previous works from the same laboratory, the findings are novel and provide valuable insights into the dynamics governing hippocampal synaptic plasticity.

The claim that GluA1 homomeric CP-AMPA receptors mediate the expression of DA-LTP is fascinating, and although the electrophysiology data on GluA1 knock-out mice are convincing, more evidence is needed to support this hypothesis. Western blotting provides useful information on the expression level of GluA1, which is not necessarily associated with cell surface expression of GluA1 and therefore CP-AMPARs. Validating this hypothesis by localizing the protein using immunofluorescence and confocal microscopy detection could strengthen the claim. The authors should briefly discuss the limitations of the study.

Although it would be possible to quantify the surface expression of GluA1 using immunofluorescence, it would not be possible to distinguish  between GluA1 homomers and GluA1-containing heteromers. It would therefore not be informative as to whether these are indeed CP-AMPARs. This is an interesting problem, which we have briefly discussed in the Discussion section.

Additional comments to address:

(1) In Figure 2A, the representative image with PMY alone shows a very weak PMY signal. Consequently, the image with TTX alone seems to potentiate the PMY signal, suggesting a counterintuitive increase in protein synthesis.

We agree with the Reviewer that the original image was not representative and have replaced it with a more representative image.

(2) In Figures 3A-B, the Western blotting representative images have poor quality, especially regarding GluA1 and α-actin in Figure 3A. The quantification graph (Figure 3B) raises some concerns about a potential outlier in both the DA alone and DA+CHX groups. The authors should consider running a statistical test to detect outlier data. Full blot images, including ladder lines, should be added to the supplementary data.

We have replaced the western blot image in Figure 3A and have also presented full blot images including ladder lines in supplementary Figure S3.

Using the ROUT method (Q=1%) we identified one outlier in the DA+CHX group of the western blot quantification. The quantification for this blot was then removed from the dataset and the experiment was repeated to ensure a sufficient number of repeats.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) How the authors perform these experiments with puromycin, these are puromycilation experiments - not SuNSET. The SuNSET protocol (surface sensing of translation) specifically refers to the detection of newly synthesized proteins externally at the plasma membrane. I'd advise to update the terminology used.

We thank the Reviewer for pointing this out. We have updated this to ‘puromycin-based labelling assay’.

(2) The legend presented in Figure 2F suggests WT is green and ACKO is orange, however, in Figure 2G the WT LTP trace is orange, consider changing this to green for consistency.

We thank the Reviewer for this suggestion and agree that a matching colour scheme makes the Figure clearer. This has been updated.

(3) In the results section, it is recommended to include units for the values presented at the first instance and only again when the units change thereafter.

The units of the electrophysiology data were [%], this is included in the Results section. Results of western blots and IHC images were presented as [a.u.]. While we included this in the Figures, we have not specifically added this to the text of individual results. 

(4) Two hours pre-treatment with anisomycin vs 30 minutes pretreatment with cycloheximide seems hard to directly compare - as the pharmokinetics of translational inhibition should be similar for both drugs. What was the rationale for the extremely long anisomycin pretreatment? What controls were taken to assess slice health either prior to or following fixation? This is relevant to the below point (5).

In an initial set of control experiments (Figure S 1C-D) we wanted to ensure that protein synthesis was definitely blocked and therefore used a relatively high concentration of anisomycin and a relatively long pre-incubation period. We agree with the Reviewer that we cannot exclude the possibility that this treatment could compromise cell health in addition to the protein synthesis block. Therefore, we carried out an additional experiment with an alternative protein synthesis inhibitor cycloheximide at a lower standard concentration (10 µM) which confirmed a significant reduction in the puromycin signal (Figure S1A-B). Together these results support the conclusion that puromycin signal is specific to protein synthesis in our labelling assay.

IHC slices were visually assessed for health. The large set of electrophysiology experiments carried out in our study (all recorded cells were evaluated for healthy resting membrane potential, action potential firing, and synaptic responses) also confirmed that, generally, the vast majority of our slices were indeed healthy. 

(5) In Supplementary Figure 1, there is a dramatic difference in the a.u. intensities across CHX (B) and AM (D), please explain the reason for this. It is understood these are normalised values to nuclear staining, please clarify if this is a nuclear area.

We agree with the Reviewer that there is a large variability in normalised puromycin signal which may be due to variability in the health of the slices. However, we assume that the same variability would be present in the treated slices, which showed, despite the variability, a significant effect of protein synthesis inhibition. To prevent introducing bias by excluding slices with low puromycin signal in the control condition, we present the full dataset.

The CA1 region of the hippocampus contains of a dense layer of neuronal somata (pyramidal cell layer). We normalized against the nuclear area as it provides a reliable estimate of the number of neurons present in the image. This approach minimizes bias by accounting for variation in the number of neurons within the visual field, ensuring consistency and accuracy in our analysis.

(6) Please clarify the decision to average both the last 5 minutes of baseline recordings and the last 5 minutes of the recording for the normalisation of EPSP slopes.

The baseline usually stabilises after a few minutes of recording, thus the last 5 minutes were used for baseline measurement, which are the most relevant datapoints to compare synaptic weight change to. After induction of STDP, potentiation or depression of synaptic weights develops gradually. Based on previous results, evaluating the EPSP slopes at 30-40 minutes after the induction protocol gives a reliable estimate of the amount of plasticity.

Reviewer #2 (Recommendations For The Authors):

The concentration of anisomycin used (0.5 mM) is very high.

As described above, in an initial set of control experiments (Figure S 1C-D) we wanted to ensure that protein synthesis was definitely blocked and therefore used a relatively high concentration of anisomycin and a relatively long pre-incubation period. We agree with the Reviewer that this is higher than the standard concentration used for this drug and we cannot exclude the possibility that this treatment could compromise cell health in addition to the protein synthesis block. Therefore, we carried out an additional experiment with an alternative protein synthesis inhibitor cycloheximide at a lower standard concentration (10 µM) which confirmed a significant reduction in the puromycin signal (Figure S1A-B). Together these results support the conclusion that puromycin signal is specific to protein synthesis in our labelling assay.

Furthermore, in the electrophysiology experiments, we also used 500 µM anisomycin in the patch pipette solution. Under these conditions, we recorded a stable EPSP baseline for 60 minutes, indicating that the treatment did not cause toxic effects to the cell (Figure S1F). This high concentration would ensure an effective block of local translation at dendritic sites. Nevertheless, we also carried out this experiment with cycloheximide at a lower standard concentration (10 µM) and observed a similar result with both protein synthesis inhibitors (Figure 1F).

The authors conclude that the effect of DA is mediated via D1/5 receptors, which based on previous work seems likely. But they cannot conclude this from their current study which used a combination of a D1/D5 and a D2 antagonist.

We thank the Reviewer for pointing this out. We agree and have updated this in the Discussion section to ‘dopamine receptors’, without specifying subtypes.

There is no mention that I can see that the KO experiments were conducted in a blinded manner (which I believe should be standard practice). Did they verify the KOs using Westerns?

Only a subset of the experiments was conducted in a blinded manner. However, the results were collected by two independent experimenters, who both observed significant effects in KO mice compared to WTs (TF and ZB).

We received the DKO mice from a former collaborator, who verified expression levels of the KO mice (Wang et al., 2003). We verified DKO upon arrival in our facility using genotyping.

Maybe I'm misunderstanding but it appears to me that in Figure 1F there is LTP prior to the addition of DA. (The first point after pairing is already elevated). I think the control of pairing without DA should be added.

We thank the Reviewer for pointing this out. Based on previous results (Brzosko et al., 2015) we would expect potentiation to develop over time once DA is added after pairing, however, it indeed appears in the Figure here as if there was an immediate increase in synaptic weights after pairing. It should be noted, however, that when comparing the first 5 minutes after pairing to the baseline, this increase was not significant (t(9)=1.810, p =0.1037). Nevertheless, we rechecked our data and noticed that this initial potentiation was biased by one cell with an increasing baseline, which had both the test and control pathway strongly elevated. We had mistakenly included this cell in the dataset, despite the unstable conditions (as stated in the Methods section, the unpaired control pathway served as a stability control). We apologise for the error and this has now been corrected (Figure 1F). In addition, we present the control pathway in Figure S1G and I.

We have also now included the control for post-before-pre pairing (Δt = -20 ms) without dopamine in a supplemental figure (Figure S1E and F).

The Westerns (Figure 3A) are fairly messy. Also, it is better to quantify with total protein. Surface biotinylation of GluA1 and GluA2 would be more informative.

We carried out more repeats of Western blots and have exchanged blots in Figure 3A.

We observed that DA increases protein synthesis, we therefore cannot exclude the possibility that application of DA could also affect total protein levels. Thus quantifying with total protein may not be the best choice here. Quantification with actin is standard practice.

While we agree with the Reviewer that surface biotinylation of GluA1 and GluA2 could in principle be more informative, we do not think it would work well in our experimental setup using acute slice preparation, as it strictly requires intact cells. Slicing generates damaged cells, which would take up the surface biotin reagents. This would cause unspecific biotinylation of the damaged cells, leading to a strong background signal in the assay.

In Figure 4 panels D and E the baselines are increasing substantially prior to induction. I appreciate that long stable baselines with timing-dependent plasticity may not be possible but it's hard to conclude what happened tens of minutes later when the baseline only appears stable for a minute or two. Panels A and B show that relatively stable baselines are achievable.

We agree with the Reviewer that the baselines are increasing, however, when looking at the baseline for 5 minutes prior to induction (5 last datapoints of the baseline), which is what we used for quantification, the baselines appeared stable. Unfortunately, longer baselines are not suitable for timing-dependent plasticity. In addition, all experiments were carried out with a control pathway which showed stable conditions throughout the recording.

In general, the discussion could be better integrated with the current literature. Their experiments are in line with a substantial body of literature that has identified two forms of LTP, based on these signalling cascades, using more conventional induction patterns.

We thank the Reviewer for this suggestion and have added more discussion of the two forms of LTP in the Discussion section.

It would be helpful to include the drug concentrations when first described in the results.

Drug concentration have now been included in the Results section.

It is now more common to include absolute t values (not just <0.05 etc).

While we indicate significance in Figures using asterisks when p values are below the indicated significance levels, we report absolute values of p and t values in the Results section.

Similarly full blots should be added to an appendix / made available.

We have now included full blot images in Supplementary Figure S3.

A 30% tolerance for series resistance seems generous to me. (10-20% would be more typical).

We thank the Reviewer for their suggestion, and will keep this in mind for future studies. However, the error introduced by the higher tolerance level is likely to be small and would not influence any of the qualitative conclusions of the manuscript.

Whereas series resistance is of course extremely important in voltage-clamp experiments, changes in series resistance would be less of a concern in current-clamp recordings of synaptic events. We use the amplifier as a voltage follower, and there are two problems with changes in the electrode, or access, resistance. First, there is the voltage drop across the electrode resistance. Clearly this error is zero if no current is injected and is also negligible for the currents we use in our experiments to maintain the membrane voltage at -70 mV. For example, the voltage drop would be 0.2 mV for 20 pA current through a typical 10 MOhm electrode resistance, and a change in resistance of 30% would give less than 0.1 mV voltage change even if the resistance were not compensated. The second problem is distortion of the EPSP shape due to the low-pass filtering properties of the electrode set up by the pipette capacitance and series resistance (RC). This can be a significant problem for fast events, such as action potentials, but less of a problem for the relatively slow EPSPs recorded in pyramidal cells. Nevertheless, we take on board the advice provided by the Reviewer and will use the conventional tolerance of 20% in future experiments.

Reviewer #3 (Recommendations For The Authors):

In the references, the entry for Burnashev N et al. has a different font size. Please ensure that all references are formatted consistently.

We thank the Reviewer for spotting this and have updated the font size of this reference.

eLife Assessment

Birdsong production depends on precise neural sequences in a vocal motor nucleus HVC. In this useful biophysical model, Daou and colleagues identify specific biophysical parameters that result in sparse neural sequences observed in vivo. While the model is presently incomplete because it is overfit to produce sequences and therefore not robust to real biological variation, the model has the potential to address some outstanding issues in HVC function.

Reviewer #1 (Public review):

Summary:

The paper presents a model for sequence generation in the zebra finch HVC, which adheres to cellular properties measured experimentally. However, the model is fine-tuned and exhibits limited robustness to noise inherent in the inhibitory interneurons within the HVC, as well as to fluctuations in connectivity between neurons. Although the proposed microcircuits are introduced as units for sub-syllabic segments (SSS), the backbone of the network remains a feedforward chain of HVC_RA neurons, similar to previous models.

Strengths:

The model incorporates all three of the major types of HVC neurons. The ion channels used and their kinetics are based on experimental measurements. The connection patterns of the neurons are also constrained by the experiments.

Weaknesses:

The model is described as consisting of micro-circuits corresponding to SSS. This presentation gives the impression that the model's structure is distinct from previous models, which connected HVC_RA neurons in feedforward chain networks (Jin et al 2007, Li & Greenside, 2006; Long et al 2010; Egger et al 2020). However, the authors implement single HVC_RA neurons into chain networks within each micro-circuit and then connect the end of the chain to the start of the chain in the subsequent micro-circuit. Thus, the HVC_RA neuron in their model forms a single-neuron chain. This structure is essentially a simplified version of earlier models.

In the model of the paper, the chain network drives the HVC_I and HVC_X neurons. The role of the micro-circuits is more significant in organizing the connections: specifically, from HVC_RA neurons to HVC_I neurons, and from HVC_I neurons to both HVC_X and HVC_RA neurons.

How useful is this concept of micro-circuits? HVC neurons fire continuously even during the silent gaps. There are no SSS during these silent gaps.

A significant issue of the current model is that the HVC_RA to HVC_RA connections require fine-tuning, with the network functioning only within a narrow range of g_AMPA (Figure 2B). Similarly, the connections from HVC_I neurons to HVC_RA neurons also require fine-tuning. This sensitivity arises because the somatic properties of HVC_RA neurons are insufficient to produce the stereotypical bursts of spikes observed in recordings from singing birds, as demonstrated in previous studies (Jin et al 2007; Long et al 2010). In these previous works, to address this limitation, a dendritic spike mechanism was introduced to generate an intrinsic bursting capability, which is absent in the somatic compartment of HVC_RA neurons. This dendritic mechanism significantly enhances the robustness of the chain network, eliminating the need to fine-tune any synaptic conductances, including those from HVC_I neurons (Long et al 2010).

Why is it important that the model should NOT be sensitive to the connection strengths?

First, the firing of HVC_I neurons is highly noisy and unreliable. HVC_I neurons fire spontaneous, random spikes under baseline conditions. During singing, their spike timing is imprecise and can vary significantly from trial to trial, with spikes appearing or disappearing across different trials. As a result, their inputs to HVC_RA neurons are inherently noisy. If the model relies on precisely tuned inputs from HVC_I neurons, the natural fluctuations in HVC_I firing would render the model non-functional. The authors should incorporate noisy HVC_I neurons into their model to evaluate whether this noise would render the model non-functional.

Second, Kosche et al. (2015) demonstrated that reducing inhibition by suppressing HVC_I neuron activity makes HVC_RA firing less sparse but does not compromise the temporal precision of the bursts. In this experiment, the local application of gabazine should have severely disrupted HVC_I activity. However, it did not affect the timing precision of HVC_RA neuron firing, emphasizing the robustness of the HVC timing circuit. This robustness is inconsistent with the predictions of the current model, which depends on finely tuned inputs and should, therefore, be vulnerable to such disruptions.

Third, the reliance on fine-tuning of HVC_RA connections becomes problematic if the model is scaled up to include groups of HVC_RA neurons forming a chain network, rather than the single HVC_RA neurons used in the current work. With groups of HVC_RA neurons, the summation of presynaptic inputs to each HVC_RA neuron would need to be precisely maintained for the model to function. However, experimental evidence shows that the HVC circuit remains functional despite perturbations, such as a few degrees of cooling, micro-lesions, or turnover of HVC_RA neurons. Such robustness cannot be accounted for by a model that depends on finely tuned connections, as seen in the current implementation.

The authors examined how altering the channel properties of neurons affects the activity in their model. While this approach is valid, many of the observed effects may stem from the delicate balancing required in their model for proper function.

In the current model, HVC_X neurons burst as a result of rebound activity driven by the I_H current. Rebound bursts mediated by the I_H current typically require a highly hyperpolarized membrane potential. However, this mechanism would fail if the reversal potential of inhibition is higher than the required level of hyperpolarization. Furthermore, Mooney (2000) demonstrated that depolarizing the membrane potential of HVC_X neurons did not prevent bursts of these neurons during forward playback of the bird's own song, suggesting that these bursts (at least under anesthesia, which may be a different state altogether) are not necessarily caused by rebound activity. This discrepancy should be addressed or considered in the model.

Some figures contain direct copies of figures from published papers. It is perhaps a better practice to replace them with schematics if possible.

Reviewer #2 (Public review):

Summary:

In this paper, the authors use numerical simulations to try to understand better a major experimental discovery in songbird neuroscience from 2002 by Richard Hahnloser and collaborators. The 2002 paper found that a certain class of projection neurons in the premotor nucleus HVC of adult male zebra finch songbirds, the neurons that project to another premotor nucleus RA, fired sparsely (once per song motif) and precisely (to about 1 ms accuracy) during singing.

The experimental discovery is important to understand since it initially suggested that the sparsely firing RA-projecting neurons acted as a simple clock that was localized to HVC and that controlled all details of the temporal hierarchy of singing: notes, syllables, gaps, and motifs. Later experiments suggested that the initial interpretation might be incomplete: that the temporal structure of adult male zebra finch songs instead emerged in a more complicated and distributed way, still not well understood, from the interaction of HVC with multiple other nuclei, including auditory and brainstem areas. So at least two major questions remain unanswered more than two decades after the 2002 experiment: What is the neurobiological mechanism that produces the sparse precise bursting: is it a local circuit in HVC or is it some combination of external input to HVC and local circuitry? And how is the sparse precise bursting in HVC related to a songbird's vocalizations?

The authors only investigate part of the first question, whether the mechanism for sparse precise bursts is local to HVC. They do so indirectly, by using conductance-based Hodgkin-Huxley-like equations to simulate the spiking dynamics of a simplified network that includes three known major classes of HVC neurons and such that all neurons within a class are assumed to be identical. A strength of the calculations is that the authors include known biophysically deduced details of the different conductances of the three major classes of HVC neurons, and they take into account what is known, based on sparse paired recordings in slices, about how the three classes connect to one another. One weakness of the paper is that the authors make arbitrary and not well-motivated assumptions about the network geometry, and they do not use the flexibility of their simulations to study how their results depend on their network assumptions. A second weakness is that they ignore many known experimental details such as projections into HVC from other nuclei, dendritic computations (the somas and dendrites are treated by the authors as point-like isopotential objects), the role of neuromodulators, and known heterogeneity of the interneurons. These weaknesses make it difficult for readers to know the relevance of the simulations for experiments and for advancing theoretical understanding.

Strengths:

The authors use conductance-based Hodgkin-Huxley-like equations to simulate spiking activity in a network of neurons intended to model more accurately songbird nucleus HVC of adult male zebra finches. Spiking models are much closer to experiments than models based on firing rates or on 2-state neurons.

The authors include information deduced from modeling experimental current-clamp data such as the types and properties of conductances. They also take into account how neurons in one class connect to neurons in other classes via excitatory or inhibitory synapses, based on sparse paired recordings in slices by other researchers.

The authors obtain some new results of modest interest such as how changes in the maximum conductances of four key channels (e.g., A-type K+ currents or Ca-dependent K+ currents) influence the structure and propagation of bursts, while simultaneously being able to mimic accurately current-clamp voltage measurements.

Weaknesses:

One weakness of this paper is the lack of a clearly stated, interesting, and relevant scientific question to try to answer. In the introduction, the authors do not discuss adequately which questions recent experimental and theoretical work have failed to explain adequately, concerning HVC neural dynamics and its role in producing vocalizations. The authors do not discuss adequately why they chose the approach of their paper and how their results address some of these questions.

For example, the authors need to explain in more detail how their calculations relate to the works of Daou et al, J. Neurophys. 2013 (which already fitted spiking models to neuronal data and identified certain conductances), to Jin et al J. Comput. Neurosci. 2007 (which already discussed how to get bursts using some experimental details), and to the rather similar paper by E. Armstrong and H. Abarbanel, J. Neurophys 2016, which already postulated and studied sequences of microcircuits in HVC. This last paper is not even cited by the authors.

The authors' main achievement is to show that simulations of a certain simplified and idealized network of spiking neurons, which includes some experimental details but ignores many others, match some experimental results like current-clamp-derived voltage time series for the three classes of HVC neurons (although this was already reported in earlier work by Daou and collaborators in 2013), and simultaneously the robust propagation of bursts with properties similar to those observed in experiments. The authors also present results about how certain neuronal details and burst propagation change when certain key maximum conductances are varied.

However, these are weak conclusions for two reasons. First, the authors did not do enough calculations to allow the reader to understand how many parameters were needed to obtain these fits and whether simpler circuits, say with fewer parameters and simpler network topology, could do just as well. Second, many previous researchers have demonstrated robust burst propagation in a variety of feed-forward models. So what is new and important about the authors' results compared to the previous computational papers?

Also missing is a discussion, or at least an acknowledgment, of the fact that not all of the fine experimental details of undershoots, latencies, spike structure, spike accommodation, etc may be relevant for understanding vocalization. While it is nice to know that some models can match these experimental details and produce realistic bursts, that does not mean that all of these details are relevant for the function of producing precise vocalizations. Scientific insights in biology often require exploring which of the many observed details can be ignored and especially identifying the few that are essential for answering some questions. As one example, if HVC-X neurons are completely removed from the authors' model, does one still get robust and reasonable burst propagation of HVC-RA neurons? While part of the nucleus HVC acts as a premotor circuit that drives the nucleus RA, part of HVC is also related to learning. It is not clear that HVC-X neurons, which carry out some unknown calculation and transmit information to area X in a learning pathway, are relevant for burst production and propagation of HVC-RA neurons, and so relevant for vocalization. Simulations provide a convenient and direct way to explore questions of this kind.

One key question to answer is whether the bursting of HVC-RA projection neurons is based on a mechanism local to HVC or is some combination of external driving (say from auditory nuclei) and local circuitry. The authors do not contribute to answering this question because they ignore external driving and assume that the mechanism is some kind of intrinsic feed-forward circuit, which they put in by hand in a rather arbitrary and poorly justified way, by assuming the existence of small microcircuits consisting of a few HVC-RA, HVC-X, and HVC-I neurons that somehow correspond to "sub-syllabic segments". To my knowledge, experiments do not suggest the existence of such microcircuits nor does theory suggest the need for such microcircuits.

Another weakness of this paper is an unsatisfactory discussion of how the model was obtained, validated, and simulated. The authors should state as clearly as possible, in one location such as an appendix, what is the total number of independent parameters for the entire network and how parameter values were deduced from data or assigned by hand. With enough parameters and variables, many details can be fit arbitrarily accurately so researchers have to be careful to avoid overfitting. If parameter values were obtained by fitting to data, the authors should state clearly what the fitting algorithm was (some iterative nonlinear method, whose results can depend on the initial choice of parameters), what the error function used for fitting (sum of least squares?) was, and what data were used for the fitting.

The authors should also state clearly the dynamical state of the network, the vector of quantities that evolve over time. (What is the dimension of that vector, which is also the number of ordinary differential equations that have to be integrated?) The authors do not mention what initial state was used to start the numerical integrations, whether transient dynamics were observed and what were their properties, or how the results depended on the choice of the initial state. The authors do not discuss how they determined that their model was programmed correctly (it is difficult to avoid typing errors when writing several pages or more of a code in any language) or how they determined the accuracy of the numerical integration method beyond fitting to experimental data, say by varying the time step size over some range or by comparing two different integration algorithms.

Also disappointing is that the authors do not make any predictions to test, except rather weak ones such as that varying a maximum conductance sufficiently (which might be possible by using dynamic clamps) might cause burst propagation to stop or change its properties. Based on their results, the authors do not make suggestions for further experiments or calculations, but they should.

Author response:

eLife Assessment

Birdsong production depends on precise neural sequences in a vocal motor nucleus HVC. In this useful biophysical model, Daou and colleagues identify specific biophysical parameters that result in sparse neural sequences observed in vivo. While the model is presently incomplete because it is overfit to produce sequences and therefore not robust to real biological variation, the model has the potential to address some outstanding issues in HVC function.

We are grateful for the extensive supportive comments from the reviewers, including broad, strong appreciation of the novel aspects of our manuscript. We believe these will be only strengthened in the next submission.

Public Reviews:

Reviewer #1 (Public review):

Summary:

The paper presents a model for sequence generation in the zebra finch HVC, which adheres to cellular properties measured experimentally. However, the model is fine-tuned and exhibits limited robustness to noise inherent in the inhibitory interneurons within the HVC, as well as to fluctuations in connectivity between neurons. Although the proposed microcircuits are introduced as units for sub-syllabic segments (SSS), the backbone of the network remains a feedforward chain of HVC_RA neurons, similar to previous models.

Strengths:

The model incorporates all three of the major types of HVC neurons. The ion channels used and their kinetics are based on experimental measurements. The connection patterns of the neurons are also constrained by the experiments.

Weaknesses:

The model is described as consisting of micro-circuits corresponding to SSS. This presentation gives the impression that the model's structure is distinct from previous models, which connected HVC_RA neurons in feedforward chain networks (Jin et al 2007, Li & Greenside, 2006; Long et al 2010; Egger et al 2020). However, the authors implement single HVC_RA neurons into chain networks within each micro-circuit and then connect the end of the chain to the start of the chain in the subsequent micro-circuit. Thus, the HVC_RA neuron in their model forms a single-neuron chain. This structure is essentially a simplified version of earlier models.

In the model of the paper, the chain network drives the HVC_I and HVC_X neurons. The role of the micro-circuits is more significant in organizing the connections: specifically, from HVC_RA neurons to HVC_I neurons, and from HVC_I neurons to both HVC_X and HVC_RA neurons.

We thank Reviewer 1 for their thoughtful comments.

While the reviewer is correct about the fact that the propagation of sequential activity in this model is primarily carried by HVC_RA neurons in a feed-forward manner, we need to emphasize that this is true only if there is no intrinsic or synaptic perturbation to the HVC network. For example, we showed in Figures 10 and 12 how altering the intrinsic properties of HVC_X neurons or for interneurons disrupts sequence propagation. In other words, while HVC_RA neurons are the key forces to carry the chain forward, the interplay between excitation and inhibition in our network as well as the intrinsic parameters for all classes of HVC neurons are equally important forces in carrying the chain of activity forward. Thus, the stability of activity propagation necessary for song production depend on a finely balanced network of HVC neurons, with all classes contributing to the overall dynamics. Moreover, all existing models that describe premotor sequence generation in the HVC either assume a distributed model (Elmaleh et al., 2021) that dictates that local HVC circuitry is not sufficient to advance the sequence but rather depends upon momentto-moment feedback through Uva (Hamaguchi et al., 2016), or assume models that rely on intrinsic connections within HVC to propagate sequential activity. In the latter case, some models assume that HVC is composed of multiple discrete subnetworks that encode individual song elements (Glaze & Troyer, 2013; Long & Fee, 2008; Wang et al., 2008), but lacks the local connectivity to link the subnetworks, while other models assume that HVC may have sufficient information in its intrinsic connections to form a single continuous network sequence (Long et al. 2010). The HVC model we present extends the concept of a feedforward network by incorporating additional neuronal classes that influence the propagation of activity (interneurons and HVC_X neurons). We have shown that any disturbance of the intrinsic or synaptic conductances of these latter neurons will disrupt activity in the circuit even when HVC_RA neurons properties are maintained.

In regard to the similarities between our model and earlier models, several aspects of our model distinguish it from prior work. In short, while several models of how sequence is generated within HVC have been proposed (Cannon et al., 2015; Drew & Abbott, 2003; Egger et al., 2020; Elmaleh et al., 2021; Galvis et al., 2018; Gibb et al., 2009a, 2009b; Hamaguchi et al., 2016; Jin, 2009; Long & Fee, 2008; Markowitz et al., 2015), all the models proposed either rely on intrinsic HVC circuitry to propagate sequential activity, rely on extrinsic feedback to advance the sequence or rely on both. These models do not capture the complex details of spike morphology, do not include the right ionic currents, do not incorporate all classes of HVC neurons, or do not generate realistic firing patterns as seen in vivo. Our model is the first biophysically realistic model that incorporates all classes of HVC neurons and their intrinsic properties. We tuned the intrinsic and the synaptic properties bases on the traces collected by Daou et al. (2013) and Mooney and Prather (2005) as shown in Figure 3. The three classes of model neurons incorporated to our network as well as the synaptic currents that connect them are based on HodgkinHuxley formalisms that contain ion channels and synaptic currents which had been pharmacologically identified. This is an advancement over prior models that primarily focused on the role of synaptic interactions or external inputs. The model is based on a feedforward chain of microcircuits that encode for the different sub-syllabic segments and that interact with each other through structured feedback inhibition, defining an ordered sequence of cell firing. Moreover, while several models highlight the critical role of inhibitory interneurons in shaping the timing and propagation of bursts of activity in HVC_RA neurons, our work offers an intricate and comprehensive model that help understand this critical role played by inhibition in shaping song dynamics and ensuring sequence propagation.

How useful is this concept of micro-circuits? HVC neurons fire continuously even during the silent gaps. There are no SSS during these silent gaps.

Regarding the concern about the usefulness of the 'microcircuit' concept in our study, we appreciate the comment and we are glad to clarify its relevance in our network. While we acknowledge that HVC_RA neurons interconnect microcircuits, our model's dynamics are still best described within the framework of microcircuitry particularly due to the firing behavior of HVC_X neurons and interneurons. Here, we are referring to microcircuits in a more functional sense, rather than rigid, isolated spatial divisions (Cannon et al. 2015). A microcircuit in our model reflects the local rules that govern the interaction between all HVC neuron classes within the broader network, and that are essential for proper activity propagation. For example, HVC_INT neurons belonging to any microcircuit burst densely and at times other than the moments when the corresponding encoded SSS is being “sung”. What makes a particular interneuron belong to this microcircuit or the other is merely the fact that it cannot inhibit HVC_RA neurons that are housed in the microcircuit it belongs to. In particular, if HVC_INT inhibits HVC_RA in the same microcircuit, some of the HVC_RA bursts in the microcircuit might be silenced by the dense and strong HVC_INT inhibition breaking the chain of activity again. Similarly, HVC_X neurons were selected to be housed within microcircuits due to the following reason: if an HVC_X neuron belonging to microcircuit i sends excitatory input to an HVC_INT neuron in microcircuit j, and that interneuron happens to select an HVC_RA neuron from microcircuit i, then the propagation of sequential activity will halt, and we’ll be in a scenario similar to what was described earlier for HVC_INT neurons inhibiting HVC_RA neurons in the same microcircuit.

We agree that there are no sub-syllabic segments described during the silent gaps and we thank the reviewer to pointing this out. Although silent gaps are integral to the overall process of song production, we have not elaborated on them in this model due to the lack of a clear, biophysically grounded representation for the gaps themselves at the level of HVC. Our primary focus has been on modeling the active, syllable-producing phases of the song, where the HVC network’s sequential dynamics are critical for song. However, one can think the encoding of silent gaps via similar mechanisms that encode SSSs, where each gap is encoded by similar microcircuits comprised of the three classes of HVC neurons (let’s called them GAP rather than SSS) that are active only during the silent gaps. In this case, the propagation of sequential activity is carried throughout the GAPs from the last SSS of the previous syllable to the first SSS of the subsequent syllable. We’ll make sure to emphasize this mechanism more in the revised version of the manuscript.

A significant issue of the current model is that the HVC_RA to HVC_RA connections require fine-tuning, with the network functioning only within a narrow range of g_AMPA (Figure 2B). Similarly, the connections from HVC_I neurons to HVC_RA neurons also require fine-tuning. This sensitivity arises because the somatic properties of HVC_RA neurons are insufficient to produce the stereotypical bursts of spikes observed in recordings from singing birds, as demonstrated in previous studies (Jin et al 2007; Long et al 2010). In these previous works, to address this limitation, a dendritic spike mechanism was introduced to generate an intrinsic bursting capability, which is absent in the somatic compartment of HVC_RA neurons. This dendritic mechanism significantly enhances the robustness of the chain network, eliminating the need to fine-tune any synaptic conductances, including those from HVC_I neurons (Long et al 2010).

Why is it important that the model should NOT be sensitive to the connection strengths?

We thank the reviewer for the comment. While mathematical models designed for highly complex nonlinear biological processes tangentially touch the biological realism, the current network as is right now is the first biologically realistic-enough network model designed for HVC that explains sequence propagation. We do not include dendritic processes in our network although that increases the realistic dynamics for various reasons. 1) The ion channels we integrated into the somatic compartment are known pharmacologically (Daou et al. 2013), but we don’t know about the dendritic compartment’s intrinsic properties of HVC neurons and the cocktail of ion channels that are expressed there. 2) We are able to generate realistic bursting in HVC_RA neurons despite the single compartment, and the main emphasis in this network is on the interactions between excitation and inhibition, the effects of ion channels in modulating sequence propagation, etc. 3) The network model already incorporates thousands of ODEs that govern the dynamics of each of the HVC neurons, so we did not want to add more complexity to the network especially that we don’t know the biophysical properties of the dendritic compartments.

Therefore, our present focus is on somatic dynamics and the interaction between HVC_RA and HVC_INT neurons, but we acknowledge the importance of these processes in enhancing network resiliency. Although we agree that adding dendritic processes improves robustness, we still think that somatic processes alone can offer insightful information on the sequential dynamics of the HVC network. While the network should be robust across a wide range of parameters, it is also essential that certain parameters are designed to filter out weaker signals, ensuring that only reliable, precise patterns of activity propagate. Hence, we specifically chose to make the HVC_RA-to-HVC_RA excitatory connections more sensitive (narrow range of values) such that only strong, precise and meaningful stimuli can propagate through the network representing the high stereotypy and precision seen in song production.

First, the firing of HVC_I neurons is highly noisy and unreliable. HVC_I neurons fire spontaneous, random spikes under baseline conditions. During singing, their spike timing is imprecise and can vary significantly from trial to trial, with spikes appearing or disappearing across different trials. As a result, their inputs to HVC_RA neurons are inherently noisy. If the model relies on precisely tuned inputs from HVC_I neurons, the natural fluctuations in HVC_I firing would render the model non-functional. The authors should incorporate noisy HVC_I neurons into their model to evaluate whether this noise would render the model non-functional.

We acknowledge that under baseline and singing settings, interneurons fire in an extremely noisy and inaccurate manner, although they exhibit time locked episodes in their activity (Hahnloser et al 2002, Kozhinikov and Fee 2007). In order to mimic the biological variability of these neurons, our model does, in fact, include a stochastic current to reflect the intrinsic noise and random variations in interneuron firing shown in vivo (and we highlight this in the Methods). If necessary and to make sure the network is resilient to this randomness in interneuron firing, we will investigate different approaches to enhance the noise representation even further and check its effect on sequence propagation.

Second, Kosche et al. (2015) demonstrated that reducing inhibition by suppressing HVC_I neuron activity makes HVC_RA firing less sparse but does not compromise the temporal precision of the bursts. In this experiment, the local application of gabazine should have severely disrupted HVC_I activity. However, it did not affect the timing precision of HVC_RA neuron firing, emphasizing the robustness of the HVC timing circuit. This robustness is inconsistent with the predictions of the current model, which depends on finely tuned inputs and should, therefore, be vulnerable to such disruptions.

We thank the reviewer for the comment. The differences between the Kosche et al. (2015) findings and the predictions of our model arise from differences in the aspect of HVC function we are modeling. Our model is more sensitive to inhibition, which is a designed mechanism for achieving precise song patterning. This is a modeling simplification we adopted to capture specific characteristics of HVC function. Hence, Kosche et al. (2015) findings do not invalidate the approach of our model, but highlights that HVC likely operates with several, redundant mechanisms that overall ensure temporal precision.Nevertheless, we will investigate further the effects of the degree of inhibition on song patterning.

Third, the reliance on fine-tuning of HVC_RA connections becomes problematic if the model is scaled up to include groups of HVC_RA neurons forming a chain network, rather than the single HVC_RA neurons used in the current work. With groups of HVC_RA neurons, the summation of presynaptic inputs to each HVC_RA neuron would need to be precisely maintained for the model to function. However, experimental evidence shows that the HVC circuit remains functional despite perturbations, such as a few degrees of cooling, micro-lesions, or turnover of HVC_RA neurons. Such robustness cannot be accounted for by a model that depends on finely tuned connections, as seen in the current implementation.

Our model of individual HVC_RA neurons and as stated previously is reductive model that focuses on understanding the mechanisms that govern sequential neural activity. We agree that scaling the model to include many of HVC_RA neurons poses challenges, specifically concerning the summation of presynaptic inputs. However, our model can still be adapted to a larger network without requiring the level of fine-tuning currently needed. In fact, the current fine-tuning of synaptic connections in the model is a reflection of fundamental network mechanisms rather than a limitation when scaling to a larger network. Besides, one important feature of this neural network is redundancy. Even if some neurons or synaptic connections are impaired, other neurons or pathways can compensate for these changes, allowing the activity propagation to remain intact.

The authors examined how altering the channel properties of neurons affects the activity in their model. While this approach is valid, many of the observed effects may stem from the delicate balancing required in their model for proper function.

In the current model, HVC_X neurons burst as a result of rebound activity driven by the I_H current. Rebound bursts mediated by the I_H current typically require a highly hyperpolarized membrane potential. However, this mechanism would fail if the reversal potential of inhibition is higher than the required level of hyperpolarization. Furthermore, Mooney (2000) demonstrated that depolarizing the membrane potential of HVC_X neurons did not prevent bursts of these neurons during forward playback of the bird's own song, suggesting that these bursts (at least under anesthesia, which may be a different state altogether) are not necessarily caused by rebound activity. This discrepancy should be addressed or considered in the model.

In our HVC network model, one goal with HVC_X neurons is to generate bursts in their underlying neuron population. Since HVC_X neurons in our model receive only inhibitory inputs from interneurons, we rely on inhibition followed by rebound bursts orchestrated by the IH and the I_CaT currents to achieve this goal. The interplay between the T-type Ca⁺⁺ current and the H current in our model is fundamental to generate their corresponding bursts, as they are sufficient for producing the desired behavior in the network. Due to this interplay, we do not need significant inhibition to generate rebound bursts, because the T-type Ca⁺⁺ current’s conductance can be stronger leading to robust rebound bursting even when the degree of inhibition is not very strong. We will highlight this with more clarity in the revised version.

Some figures contain direct copies of figures from published papers. It is perhaps a better practice to replace them with schematics if possible.

We will replace the relevant figures with schematic representations where possible.

Reviewer #2 (Public review):

Summary:

In this paper, the authors use numerical simulations to try to understand better a major experimental discovery in songbird neuroscience from 2002 by Richard Hahnloser and collaborators. The 2002 paper found that a certain class of projection neurons in the premotor nucleus HVC of adult male zebra finch songbirds, the neurons that project to another premotor nucleus RA, fired sparsely (once per song motif) and precisely (to about 1 ms accuracy) during singing.

The experimental discovery is important to understand since it initially suggested that the sparsely firing RA-projecting neurons acted as a simple clock that was localized to HVC and that controlled all details of the temporal hierarchy of singing: notes, syllables, gaps, and motifs. Later experiments suggested that the initial interpretation might be incomplete: that the temporal structure of adult male zebra finch songs instead emerged in a more complicated and distributed way, still not well understood, from the interaction of HVC with multiple other nuclei, including auditory and brainstem areas. So at least two major questions remain unanswered more than two decades after the 2002 experiment: What is the neurobiological mechanism that produces the sparse precise bursting: is it a local circuit in HVC or is it some combination of external input to HVC and local circuitry?

And how is the sparse precise bursting in HVC related to a songbird's vocalizations?

The authors only investigate part of the first question, whether the mechanism for sparse precise bursts is local to HVC. They do so indirectly, by using conductance-based Hodgkin-Huxley-like equations to simulate the spiking dynamics of a simplified network that includes three known major classes of HVC neurons and such that all neurons within a class are assumed to be identical. A strength of the calculations is that the authors include known biophysically deduced details of the different conductances of the three major classes of HVC neurons, and they take into account what is known, based on sparse paired recordings in slices, about how the three classes connect to one another. One weakness of the paper is that the authors make arbitrary and not well-motivated assumptions about the network geometry, and they do not use the flexibility of their simulations to study how their results depend on their network assumptions. A second weakness is that they ignore many known experimental details such as projections into HVC from other nuclei, dendritic computations (the somas and dendrites are treated by the authors as point-like isopotential objects), the role of neuromodulators, and known heterogeneity of the interneurons. These weaknesses make it difficult for readers to know the relevance of the simulations for experiments and for advancing theoretical understanding.

Strengths:

The authors use conductance-based Hodgkin-Huxley-like equations to simulate spiking activity in a network of neurons intended to model more accurately songbird nucleus HVC of adult male zebra finches. Spiking models are much closer to experiments than models based on firing rates or on 2-state neurons.

The authors include information deduced from modeling experimental current-clamp data such as the types and properties of conductances. They also take into account how neurons in one class connect to neurons in other classes via excitatory or inhibitory synapses, based on sparse paired recordings in slices by other researchers.

The authors obtain some new results of modest interest such as how changes in the maximum conductances of four key channels (e.g., A-type K⁺ currents or Ca-dependent K⁺ currents) influence the structure and propagation of bursts, while simultaneously being able to mimic accurately current-clamp voltage measurements.

Weaknesses:

One weakness of this paper is the lack of a clearly stated, interesting, and relevant scientific question to try to answer. In the introduction, the authors do not discuss adequately which questions recent experimental and theoretical work have failed to explain adequately, concerning HVC neural dynamics and its role in producing vocalizations. The authors do not discuss adequately why they chose the approach of their paper and how their results address some of these questions.

For example, the authors need to explain in more detail how their calculations relate to the works of Daou et al, J. Neurophys. 2013 (which already fitted spiking models to neuronal data and identified certain conductances), to Jin et al J. Comput. Neurosci. 2007 (which already discussed how to get bursts using some experimental details), and to the rather similar paper by E. Armstrong and H. Abarbanel, J. Neurophys 2016, which already postulated and studied sequences of microcircuits in HVC. This last paper is not even cited by the authors.

We thank the reviewer for this valuable comment, and we agree that we did not clarify enough throughout the paper the utility of our model or how it advanced our understanding of the HVC dynamics and circuitry. To that end, we will revise several places of the manuscript and make sure to cite and highlight the relevance and relatedness of the mentioned papers.

In short, and as mentioned to Reviewer 1, while several models of how sequence is generated within HVC have been proposed (Cannon et al., 2015; Drew & Abbott, 2003; Egger et al., 2020; Elmaleh et al., 2021; Galvis et al., 2018; Gibb et al., 2009a, 2009b; Hamaguchi et al., 2016; Jin, 2009; Long & Fee, 2008; Markowitz et al., 2015; Jin et al., 2007), all the models proposed either rely on intrinsic HVC circuitry to propagate sequential activity, rely on extrinsic feedback to advance the sequence or rely on both. These models do not capture the complex details of spike morphology, do not include the right ionic currents, do not incorporate all classes of HVC neurons, or do not generate realistic firing patterns as seen in vivo. Our model is the first biophysically realistic model that incorporates all classes of HVC neurons and their intrinsic properties.

No existing hypothesis had been challenged with our model, rather; our model is a distillation of the various models that’s been proposed for the HVC network. We go over this in detail in the Discussion. We believe that the network model we developed provide a step forward in describing the biophysics of HVC circuitry, and may throw a new light on certain dynamics in the mammalian brain, particularly the motor cortex and the hippocampus regions where precisely-timed sequential activity is crucial. We suggest that temporally-precise sequential activity may be a manifestation of neural networks comprised of chain of microcircuits, each containing pools of excitatory and inhibitory neurons, with local interplay among neurons of the same microcircuit and global interplays across the various microcircuits, and with structured inhibition as well as intrinsic properties synchronizing the neuronal pools and stabilizing timing within a firing sequence.

The authors' main achievement is to show that simulations of a certain simplified and idealized network of spiking neurons, which includes some experimental details but ignores many others, match some experimental results like current-clamp-derived voltage time series for the three classes of HVC neurons (although this was already reported in earlier work by Daou and collaborators in 2013), and simultaneously the robust propagation of bursts with properties similar to those observed in experiments. The authors also present results about how certain neuronal details and burst propagation change when certain key maximum conductances are varied.

However, these are weak conclusions for two reasons. First, the authors did not do enough calculations to allow the reader to understand how many parameters were needed to obtain these fits and whether simpler circuits, say with fewer parameters and simpler network topology, could do just as well. Second, many previous researchers have demonstrated robust burst propagation in a variety of feed-forward models. So what is new and important about the authors' results compared to the previous computational papers?

A major novelty of our work is the incorporation of experimental data with detailed network models. While earlier works have established robust burst propagation, our model uses realistic ion channel kinetics and feedback inhibition not only to reproduce experimental neural activity patterns but also to suggest prospective mechanisms for song sequence production in the most biophysical way possible. This aspect that distinguishes our work from other feed-forward models. We go over this in detail in the Discussion. However, the reviewer is right regarding the details of the calculations conducted for the fits, we will make sure to highlight this in the Methods and throughout the manuscript with more details.

We believe that the network model we developed provide a step forward in describing the biophysics of HVC circuitry, and may throw a new light on certain dynamics in the mammalian brain, particularly the motor cortex and the hippocampus regions where precisely-timed sequential activity is crucial. We suggest that temporally-precise sequential activity may be a manifestation of neural networks comprised of chain of microcircuits, each containing pools of excitatory and inhibitory neurons, with local interplay among neurons of the same microcircuit and global interplays across the various microcircuits, and with structured inhibition as well as intrinsic properties synchronizing the neuronal pools and stabilizing timing within a firing sequence.

Also missing is a discussion, or at least an acknowledgment, of the fact that not all of the fine experimental details of undershoots, latencies, spike structure, spike accommodation, etc may be relevant for understanding vocalization. While it is nice to know that some models can match these experimental details and produce realistic bursts, that does not mean that all of these details are relevant for the function of producing precise vocalizations. Scientific insights in biology often require exploring which of the many observed details can be ignored and especially identifying the few that are essential for answering some questions. As one example, if HVC-X neurons are completely removed from the authors' model, does one still get robust and reasonable burst propagation of HVC-RA neurons? While part of the nucleus HVC acts as a premotor circuit that drives the nucleus RA, part of HVC is also related to learning. It is not clear that HVC-X neurons, which carry out some unknown calculation and transmit information to area X in a learning pathway, are relevant for burst production and propagation of HVC_RA neurons, and so relevant for vocalization. Simulations provide a convenient and direct way to explore questions of this kind.

One key question to answer is whether the bursting of HVC-RA projection neurons is based on a mechanism local to HVC or is some combination of external driving (say from auditory nuclei) and local circuitry. The authors do not contribute to answering this question because they ignore external driving and assume that the mechanism is some kind of intrinsic feed-forward circuit, which they put in by hand in a rather arbitrary and poorly justified way, by assuming the existence of small microcircuits consisting of a few HVC-RA, HVC-X, and HVC-I neurons that somehow correspond to "sub-syllabic segments". To my knowledge, experiments do not suggest the existence of such microcircuits nor does theory suggest the need for such microcircuits.

Recent results showed a tight correlation between the intrinsic properties of neurons and features of song (Daou and Margoliash 2020, Medina and Margoliash 2024), where adult birds that exhibit similar songs tend to have similar intrinsic properties. While this is relevant, we acknowledge that not all details may be necessary for every aspect of vocalization, and future models could simplify concentrate on core dynamics and exclude certain features while still providing insights into the primary mechanisms.

The question of whether HVC_X neurons are relevant for burst propagation given that our model includes these neurons as part of the network for completeness, the reviewer is correct, the propagation of sequential activity in this model is primarily carried by HVC_RA neurons in a feed-forward manner, but only if there is no perturbation to the HVC network. For example, we have shown how altering the intrinsic properties of HVC_X neurons or for interneurons disrupts sequence propagation. In other words, while HVC neurons are the key forces to carry the chain forward, the interplay between excitation and inhibition in our network as well as the intrinsic parameters for all classes of HVC neurons are equally important forces in carrying the chain of activity forward. Thus, the stability of activity propagation necessary for song production depend on a finely balanced network of HVC neurons, with all classes contributing to the overall dynamics.

We agree with the reviewer however that a potential drawback of our model is that its sole focus is on local excitatory connectivity within the HVC (Kornfeld et al., 2017; Long et al., 2010), while HVC neurons receive afferent excitatory connections (Akutagawa & Konishi, 2010; Nottebohm et al., 1982) that plays significant roles in their local dynamics. For example, the excitatory inputs that HVC neurons receive from Uvaeformis may be crucial in initiating (Andalman et al., 2011; Danish et al., 2017; Galvis et al., 2018) or sustaining (Hamaguchi et al., 2016) the sequential activity. While we acknowledge this limitation, our main contribution in this work is the biophysical insights onto how the patterning activity in HVC is largely shaped by the intrinsic properties of the individual neurons as well as the synaptic properties where excitation and inhibition play a major role in enabling neurons to generate their characteristic bursts during singing. This is true and holds irrespective of whether an external drive is injected onto the microcircuits or not. We will however elaborate on and investigate this more during the next submission.

Another weakness of this paper is an unsatisfactory discussion of how the model was obtained, validated, and simulated. The authors should state as clearly as possible, in one location such as an appendix, what is the total number of independent parameters for the entire network and how parameter values were deduced from data or assigned by hand. With enough parameters and variables, many details can be fit arbitrarily accurately so researchers have to be careful to avoid overfitting. If parameter values were obtained by fitting to data, the authors should state clearly what the fitting algorithm was (some iterative nonlinear method, whose results can depend on the initial choice of parameters), what the error function used for fitting (sum of least squares?) was, and what data were used for the fitting.

The authors should also state clearly the dynamical state of the network, the vector of quantities that evolve over time. (What is the dimension of that vector, which is also the number of ordinary differential equations that have to be integrated?) The authors do not mention what initial state was used to start the numerical integrations, whether transient dynamics were observed and what were their properties, or how the results depended on the choice of the initial state. The authors do not discuss how they determined that their model was programmed correctly (it is difficult to avoid typing errors when writing several pages or more of a code in any language) or how they determined the accuracy of the numerical integration method beyond fitting to experimental data, say by varying the time step size over some range or by comparing two different integration algorithms.

We thank the reviewer again. The fitting process in our model occurred only at the first stage where the synaptic parameters were fit to the Mooney and Prather as well as the Kosche results. There was no data shared and we merely looked at the figures in those papers and checked the amplitude of the elicited currents, the magnitudes of DC-evoked excitations etc, and we replicated that in our model. While this is suboptimal, it was better for us to start with it rather than simply using equations for synaptic currents from the literature for other types of neurons (that are not even HVC’s or in the songbird) and integrate them into our network model. However, we will certainly highlight the details of this fitting process in the new submission. We will also highlight more technical details in the Methods regarding the exact number of ODEs, the initial conditions to run them, etc.

Also disappointing is that the authors do not make any predictions to test, except rather weak ones such as that varying a maximum conductance sufficiently (which might be possible by using dynamic clamps) might cause burst propagation to stop or change its properties. Based on their results, the authors do not make suggestions for further experiments or calculations, but they should.

We agree that making experimental testable predictions is crucial for the advancement of the model. Our predictions include testing whether eradication of a class of neurons such as HVC_X neurons disrupts activity propagation which can be done through targeted neuron elimination. This also can be done through preventing rebound bursting in HVC_X by pharmacologically blocking the I_h channels. Others include down regulation of certain ion channels (pharmacologically done through ion blockers) and testing which current is fundamental for song production (and there a plenty of test based our results, like the SK current, the T-type Ca⁺⁺ current, the A-type K⁺ current, etc). We will incorporate these into the revised manuscript to better demonstrate the model's applicability and to guide future research directions.

eLife Assessment

This manuscript presents important findings on how structural color can be manipulated through a specific single-gene mutation in the motile bacterium Flavobacterium IR1. It provides a promising model to identify genes and molecular mechanisms supporting this widespread optical phenomenon. The story relies on convincing data with proteomic analysis and well-designed experiments, although it remains rather descriptive. This work will be of interest to biophysicists and microbiologists working on structural colors and Flavobacterium.

Reviewer #1 (Public review):

Summary:

Structural colors (SC) are based on nanostructures reflecting and scattering light and producing optical wave interference. All kinds of living organisms exhibit SC. However, understanding the molecular mechanisms and genes involved may be complicated due to the complexity of these organisms. Hence, bacteria that exhibit SC in colonies, such as Flavobacterium IR1, can be good models.

Based on previous genomic mining and co-occurrence with SC in flavobacterial strains, this article focuses on the role of a specific gene, moeA, in SC of Flavobacterium IR1 strain colonies on an agar plate. moeA is involved in the synthesis of the molybdenum cofactor, which is necessary for the activity of key metabolic enzymes in diverse pathways.

The authors clearly showed that the absence of moeA shifts SC properties in a way that depends on the nutritional conditions. They further bring evidence that this effect was related to several properties of the colony, all impacted by the moeA mutant: cell-cell organization, cell motility and colony spreading, and metabolism of complex carbohydrates. Hence, by linking SC to a single gene in appearance, this work points to cellular organization (as a result of cell-cell arrangement and motility) and metabolism of polysaccharides as key factors for SC in a gliding bacterium. This may prove useful for designing molecular strategies to control SC in bacterial-based biomaterials.

Strengths:

The topic is very interesting from a fundamental viewpoint and has great potential in the field of biomaterials.

The article is easy to read. It builds on previous studies with already established tools to characterize SC at the level of the flavobacterial colony. Experiments are well described and well executed. In addition, the SIBR-Cas method for chromosome engineering in Flavobacteria is the most recent and is a leap forward for future studies in this model, even beyond SC.

Weaknesses:

The paper appears a bit too descriptive and could be better organized. Some of the results, in particular the proteomic comparison, are not well exploited (not explored experimentally). In my opinion, the problem originates from the difficulty in explaining the link between the absence of moeA and the alterations observed at the level of colony spreading and polysaccharide utilization, and the variation in proteomic content.

First, the effect of moeA deletion on molybdenum cofactor synthesis should be addressed.

Second, as I was reading the entire manuscript, I kept asking myself if moeA (and by extension molybdenum cofactor) was really involved in SC or it was an indirect effect. For example, what if the absence of moeA alters the cell envelope because the synthesis of its building blocks is perturbed, then subsequently perturbates all related processes, including gliding motility and protein secretion? It would help to know if the effects on colony spreading and polysaccharide metabolism can be uncoupled. I don't think the authors discussed that clearly.

Reviewer #2 (Public review):

Summary:

The authors constructed an in-frame deletion of moeA gene, which is involved in molybdopterin cofactor (MoCo) biosynthesis, and investigated its role in structural colors in Flavobacterium IR1. The deletion of moeA shifted colony color from green to blue, reduced colony spreading, and increased starch degradation, which was attributed to the upregulation of various proteins in polysaccharide utilization loci. This study lays the ground for developing new colorants by modifying genes involved in structural colors.

Major strengths and weaknesses:

The authors conducted well-designed experiments with appropriate controls and the results in the paper are presented in a logical manner, which supports their conclusions. Using statistical tests to compare the differences between the wild type and moeA mutant, and adding a significance bar in Figure 4B, would strengthen their claims on differences in cell motility regarding differences in cell motility. Additionally, in the result section (Figure 6), the authors suggest that the shift in blue color is "caused by cells which are still highly ordered but narrower", which to my knowledge is not backed up by any experimental evidence.

Overall, this is a well-written paper in which the authors effectively address their research questions through proper experimentation. This work will help us understand the genetic basis of structural colors in Flavobacterium and open new avenues to study the roles of additional genes and proteins in structural colors.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

Structural colors (SC) are based on nanostructures reflecting and scattering light and producing optical wave interference. All kinds of living organisms exhibit SC. However, understanding the molecular mechanisms and genes involved may be complicated due to the complexity of these organisms. Hence, bacteria that exhibit SC in colonies, such as Flavobacterium IR1, can be good models.

Based on previous genomic mining and co-occurrence with SC in flavobacterial strains, this article focuses on the role of a specific gene, moeA, in SC of Flavobacterium IR1 strain colonies on an agar plate. moeA is involved in the synthesis of the molybdenum cofactor, which is necessary for the activity of key metabolic enzymes in diverse pathways.

The authors clearly showed that the absence of moeA shifts SC properties in a way that depends on the nutritional conditions. They further bring evidence that this effect was related to several properties of the colony, all impacted by the moeA mutant: cell-cell organization, cell motility and colony spreading, and metabolism of complex carbohydrates. Hence, by linking SC to a single gene in appearance, this work points to cellular organization (as a result of cell-cell arrangement and motility) and metabolism of polysaccharides as key factors for SC in a gliding bacterium. This may prove useful for designing molecular strategies to control SC in bacterial-based biomaterials.

Strengths:

The topic is very interesting from a fundamental viewpoint and has great potential in the field of biomaterials.

Thank you for your comments.

The article is easy to read. It builds on previous studies with already established tools to characterize SC at the level of the flavobacterial colony. Experiments are well described and well executed. In addition, the SIBR-Cas method for chromosome engineering in Flavobacteria is the most recent and is a leap forward for future studies in this model, even beyond SC.

We appreciate these comments.

Weaknesses:

The paper appears a bit too descriptive and could be better organized. Some of the results, in particular the proteomic comparison, are not well exploited (not explored experimentally). In my opinion, the problem originates from the difficulty in explaining the link between the absence of moeA and the alterations observed at the level of colony spreading and polysaccharide utilization, and the variation in proteomic content.

We will look at the organisation of the manuscript carefully in the coming, detailed revision, as suggested. In terms of the proteomics, there are clearly a large number of proteins affected by the moeA deletion. In terms of experimental exploration, we chose spreading, structural colour formation and starch degradation to test phenotypically, as the most relevant. For example, in L615-617, we discuss the downregulation of GldL (which is known to be involved Flavobacterial gliding motility [Shrivastava et al., 2013]) in the _moe_A KO as a possible explanation for the reduced colony spreading of moeA mutant. Changes in polysaccharide (starch) utilization were seen on solid medium, as well as in the proteomic profile where we observed the upregulation of carbohydrate metabolism proteins linked to PUL (polysaccharide utilisation locus) operons (Terrapon et al., 2015), such as PAM95095-90 (Figure 8), and other carbohydrate metabolism-related proteins, including a pectate lyase (Table S7) which is involved in starch degradation (Aspeborg et al., 2012). And as noted in L555-566 and Figure 9, starch metabolism was tested experimentally.

First, the effect of moeA deletion on molybdenum cofactor synthesis should be addressed.

MoeA is the last enzyme in the MoCo synthesis pathway, thus if only MoeA is absent the cell would accumulate MPT-AMP (molybdopterin-adenosine monophosphatase) (Iobbi-Nivol & Leimkühler, 2013), and the expressed molybdoenzymes would not be functional. In L582-585, we commented how the lack of molybdenum cofactor may affect the synthesis of molybdoenzymes. However, if you meant to analyse the presence of the small molecules, the cofactors, involved in these pathways, that was an assay we were not able to perform. Moreover, in L585-587, we addressed how the deletion of _moe_A affected the proteins encoded by the rest of genes in the operon.

Second, as I was reading the entire manuscript, I kept asking myself if moeA (and by extension molybdenum cofactor) was really involved in SC or it was an indirect effect. For example, what if the absence of moeA alters the cell envelope because the synthesis of its building blocks is perturbed, then subsequently perturbates all related processes, including gliding motility and protein secretion? It would help to know if the effects on colony spreading and polysaccharide metabolism can be uncoupled. I don't think the authors discussed that clearly.

The message of the paper is that the moeA gene, as predicted from a previous genomics analysis, is important in SC. This is based on the representation of the _moe_A gene in genomes of bacteria that display SC. This analysis does not predict the mechanism. When knocked out, a significant change in structural colour occurred, supporting this hypothesis. Whether this effect is direct or indirect is difficult to assess, as this referee rightly suggests. In order to follow up this central result, we performed proteomics (both intra- and extracellular). As we observed, the deletion of a single gene generated many changes in the proteomic profile, thus in the biological processes. Based on the known functions of molybdenum cofactor, we could only hypothesize that pterin metabolism is important for SC, not exactly how.

We intend to discuss the links between gliding/spreading and polysaccharide metabolism more clearly, with reference to the literature, as quite a bit is known here including possible links to SC.

Reviewer #2 (Public review):

Summary:

The authors constructed an in-frame deletion of moeA gene, which is involved in molybdopterin cofactor (MoCo) biosynthesis, and investigated its role in structural colors in Flavobacterium IR1. The deletion of moeA shifted colony color from green to blue, reduced colony spreading, and increased starch degradation, which was attributed to the upregulation of various proteins in polysaccharide utilization loci. This study lays the ground for developing new colorants by modifying genes involved in structural colors.

Major strengths and weaknesses:

The authors conducted well-designed experiments with appropriate controls and the results in the paper are presented in a logical manner, which supports their conclusions.

We appreciate your comment.

Using statistical tests to compare the differences between the wild type and moeA mutant, and adding a significance bar in Figure 4B, would strengthen their claims on differences in cell motility regarding differences in cell motility.

Thank you. Figure 4B contains the significance bars that represent the standard deviation of the mean value of the three replicates, but we will modify it to make them more clear.

Additionally, in the result section (Figure 6), the authors suggest that the shift in blue color is "caused by cells which are still highly ordered but narrower", which to my knowledge is not backed up by any experimental evidence.

Thanks. We mentioned that the mutant cells are narrower than the wild type based on the estimated periodicity resulting from the goniometry analysis (L427-430). We will now say “likely to be narrower based on the estimated periodicity from the optical analysis” rather than just “narrower” in the revision.

Overall, this is a well-written paper in which the authors effectively address their research questions through proper experimentation. This work will help us understand the genetic basis of structural colors in Flavobacterium and open new avenues to study the roles of additional genes and proteins in structural colors.

Much appreciated.

REFERENCES

Aspeborg, Henrik, Pedro M. Coutinho, Yang Wang, Harry Brumer, and Bernard Henrissat. "Evolution, substrate specificity and subfamily classification of glycoside hydrolase family 5 (GH5)." BMC evolutionary biology 12 (2012): 1-16.

lobbi-Nivol, Chantal, and Silke Leimkühler. "Molybdenum enzymes, their maturation and molybdenum cofactor biosynthesis in Escherichia coli." Biochimica et Biophysica Acta (BBA)-Bioenergetics 1827, no. 8-9 (2013): 1086-1101.

Shrivastava, Abhishek, Joseph J. Johnston, Jessica M. Van Baaren, and Mark J. McBride. "Flavobacterium johnsoniae GldK, GldL, GldM, and SprA are required for secretion of the cell surface gliding motility adhesins SprB and RemA." Journal of bacteriology 195, no. 14 (2013): 3201-3212.

Terrapon, Nicolas, Vincent Lombard, Harry J. Gilbert, and Bernard Henrissat. "Automatic prediction of polysaccharide utilization loci in Bacteroidetes species." Bioinformatics 31, no. 5 (2015): 647-655.

eLife Assessment

This fundamental research conducted a molecular comparison between smooth muscle cells and adjacent fibroblast cells within lung blood vessels affected by pulmonary arterial hypertension. The study identified distinct disease-related states in each cell type and provided deeper insights into their interactions and communication. While certain conclusions should be interpreted with caution due to inherent methodological limitations, the study's findings remain convincing and robust. This is supported by the use of advanced and complementary techniques, as well as the rare isolation of diseased lung blood vessel cells from the same donor, enabling direct comparison.

Reviewer #1 (Public review):

Summary:

The authors isolated and cultured pulmonary artery smooth muscle cells (PASMC) and pulmonary artery adventitial fibroblasts (PAAF) of the lung samples derived from the patients with idiopathic pulmonary arterial hypertension (PAH) and the healthy volunteers. They performed RNA-seq and proteomics analyses to detail the cellular communication between PASMC and PAAF, which are the main target cells of pulmonary vascular remodeling during the pathogenesis of PAH. The authors revealed that PASMC and PAAF retained their original cellular identity and acquired different states associated with the pathogenesis of PAH, respectively.

Strengths:

Although previous studies have shown that PASMC and PAAF cells each have an important role in the pathogenesis of PAH, there have been scarce reports focusing on the interactions between PASMC and PAAF. These findings may provide valuable information for elucidating the pathogenesis of pulmonary arterial hypertension.

Comments on revisions:

The authors adequately responded to my concerns and revised their manuscript to elaborate on the new data from new experiments and address my queries. Although some of the issues I initially raised could not be fully resolved, the revised manuscript has been significantly improved. This manuscript provides essential insights into the communications across the PASMCs and PAAFs in PAH. This would greatly interest various researchers in both basic and clinical fields.

Reviewer #2 (Public review):

Summary:

Utilizing a combination of transcriptomic and proteomic profiling as well as cellular phenotyping from source-matched PASMC and PAAFs in IPAH, this<br /> study sought to explore a molecular comparison of these cells in order to track distinct cell fate trajectories and acquisition of their IPAH-associated cellular states. The authors also aimed to identify cell-cell communication axes in order to infer mechanisms by which these two cells interact and depend upon external cues. This study will be of interest to the scientific and clinical communities of those interested in pulmonary vascular biology and disease. It also will appeal to those interested in lung and vascular development as well as multi-omic analytic procedures.

Strengths:

(1) This is one of the first studies using orthogonal sequencing and phenotyping for characterization of source-matched neighoring mesenchymal PASMC and PAAF cells in healthy and diseased IPAH patients. This is a major strength which allows for direct comparison of neighboring cell types and the ability to address an unanswered question regarding the nature of these mesenchymal "mural" cells at a precise molecular level.

(2) Unlike a number of multi-omic sequencing papers that read more as an atlas of findings without structure, the inherent comparative organization of the study and presentation of the data were valuable in aiding the reader in understanding how to discern the distinct IPAH-associated cell states. As a result, the reader not only gleans greater insight into these two interacting cell types in disease but also now can leverage these datasets more easily for future research questions in this space.

(3) There are interesting and surprising findings in the cellular characterizations, including the low proliferative state of IPAH-PASMCs as compared to the hyperproliferative state in IPAH-PAAFs. Furthermore, the cell-cell communication axes involving ECM components and soluble ligands provided by PAAFs that direct cell state dynamics of PASMCs offer some of the first and foundational descriptions of what are likely complex cellular interactions that await discovery.

(4) Technical rigor is quite high in the -omics methodology and in vitro phenotyping tools used.

Weaknesses:

There are some weaknesses in the methodology that should temper the conclusions:

(1) The number of donors sampled for PAAF/PASMCs was relatively small for both healthy controls and IPAH patients. Thus, while the level of detail of -omics profiling was quite deep, the generalizability of their findings to all IPAH patients or Group 1 PAH patients is limited. In the revised manuscript, the authors addressed this concern with important text changes and additional data.

(2) While the study utilized early passage cells, these cells nonetheless were still cultured outside the in vivo milieu prior to analysis. Thus, while there is an assumption that these cells do not change fundamental behavior outside the body, that is not entirely proven for all transcriptional and proteomic signatures. As such, the major alterations that are noted would be more compelling if validated from tissue or cells derived directly from in vivo sources. Without such validation, the major limitation of the impact and conclusions of the paper is that the full extent of the relevance of these findings to human disease is not known. The authors addressed this concern appropriately with significant text changes to clarify these limitations for the reader.

(3) While the presentation of most of the manuscript was quite clear and convincing, the terminology and conclusions regarding "cell fate trajectories" throughout the manuscript did not seem to be fully justified. That is, all of the analyses were derived from cells originating from end-stage IPAH, and otherwise, the authors were not lineage tracing across disease initiation or development (which would be impossible currently in humans). So, while the description of distinct "IPAH-associated states" makes sense, any true cell fate trajectory was not clearly defined. The revised manuscript has removed this terminology and replaced it with more precise language.

Comments on revisions:

The authors were quite responsive to all of my concerns, offering both important revisions to the presentation of the work as well as new data. While some of the limitations were not fully resolved (and the authors provide appropriate justification for this), the revised manuscript is much improved. It will be of great interest to both the scientific and clinical communities.

Author response:

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

eLife assessment

This important study explored a molecular comparison of smooth muscle and neighboring fibroblast cells found in lung blood vessels afflicted by a disease called pulmonary arterial hypertension. In doing so, the authors described distinct disease-associated states of each of these cell types with further insights into the cellular communication and crosstalk between them. The strength of evidence was convincing through the use of complementary and sophisticated tools, accompanied by rare isolation of human diseased lung blood vessel cells that were source-matched to the same donor for direct comparison.

We thank the editors and reviewers in their highly positive and encouraging assessment of our manuscript detailing the cell state changes of arterial smooth muscle cells and fibroblasts in the pulmonary bed. We addressed reviewers’ major comments in the revised manuscript by providing validation of key in vitro findings, such as preserved marker localization and increased GAG deposition in IPAH pulmonary arteries. We additionally provide comparison of transcriptomic profiles spanning fresh, very early and late passage cells. Finally, we present expanded experimental data in support of cellular crosstalk, including testing of additional PAAF ligands on donor PASMC and influence of PTX3/HGF on IPAH PASMC.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors isolated and cultured pulmonary artery smooth muscle cells (PASMC) and pulmonary artery adventitial fibroblasts (PAAF) of the lung samples derived from the patients with idiopathic pulmonary arterial hypertension (PAH) and the healthy volunteers. They performed RNA-seq and proteomics analyses to detail the cellular communication between PASMC and PAAF, which are the main target cells of pulmonary vascular remodeling during the pathogenesis of PAH. The authors revealed that PASMC and PAAF retained their original cellular identity and acquired different states associated with the pathogenesis of PAH, respectively.

Strengths:

Although previous studies have shown that PASMC and PAAF cells each have an important role in the pathogenesis of PAH, there have been scarce reports focusing on the interactions between PASMC and PAAF. These findings may provide valuable information for elucidating the pathogenesis of pulmonary arterial hypertension.

We appreciate the reviewer’s positive view of our study.

Weaknesses:

The results of proteome analysis using primary culture cells in this paper seem a bit insufficient to draw conclusions. In particular, the authors described "We elucidated the involvement of cellular crosstalk in regulating cell state dynamics and identified pentraxin-3 and hepatocyte growth factor as modulators of PASMC phenotypic transition orchestrated by PAAF." However, the presented data are considered limited and insufficient.

We thank the reviewer for drawing our attention to this point and have accordingly modified the conclusion section to read: “We investigated the involvement of cellular crosstalk….” Moreover, we provide further experimental evidence demonstrating the effect of both PTX3 and HGF on cell state marker expression in IPAH-PASMC cells (Figure 7H). In addition, we clarify the selection strategy applied to investigate particular PAAF-secreted ligands and test three additional ligands on donor PASMC (Figure S8), supporting the original focus on PTX3 and HGF.

Reviewer #2 (Public Review):

Summary:

Utilizing a combination of transcriptomic and proteomic profiling as well as cellular phenotyping from source-matched PASMC and PAAFs in IPAH, this study sought to explore a molecular comparison of these cells in order to track distinct cell fate trajectories and acquisition of their IPAH-associated cellular states. The authors also aimed to identify cell-cell communication axes in order to infer mechanisms by which these two cells interact and depend upon external cues. This study will be of interest to the scientific and clinical communities of those interested in pulmonary vascular biology and disease. It also will appeal to those interested in lung and vascular development as well as multi-omic analytic procedures.

We thank the reviewer for overall highly positive assessment of our study.

Strengths:

(1) This is one of the first studies using orthogonal sequencing and phenotyping for the characterization of source-matched neighboring mesenchymal PASMC and PAAF cells in healthy and diseased IPAH patients. This is a major strength that allows for direct comparison of neighboring cell types and the ability to address an unanswered question regarding the nature of these mesenchymal "mural" cells at a precise molecular level.

We value the reviewer’s kind and objective summary of our study.

(2) Unlike a number of multi-omic sequencing papers that read more as an atlas of findings without structure, the inherent comparative organization of the study and presentation of the data were valuable in aiding the reader in understanding how to discern the distinct IPAH-associated cell states. As a result, the reader not only gleans greater insight into these two interacting cell types in disease but also now can leverage these datasets more easily for future research questions in this space.

We thank the reviewer for this highly positive comment.

(3) There are interesting and surprising findings in the cellular characterizations, including the low proliferative state of IPAH-PASMCs as compared to the hyperproliferative state in IPAH-PAAFs. Furthermore, the cell-cell communication axes involving ECM components and soluble ligands provided by PAAFs that direct cell state dynamics of PASMCs offer some of the first and foundational descriptions of what are likely complex cellular interactions that await discovery.

We agree with the reviewer’s assessment that some of the novel data in our study helps to formulate testable hypothesis that can be followed through with more focused follow-up research.

(4) Technical rigor is quite high in the -omics methodology and in vitro phenotyping tools used.

We are grateful for reviewer’s assessment of our work and positive recognition.

Weaknesses:

There are some weaknesses in the methodology that should temper the conclusions:

(1) The number of donors sampled for PAAF/PASMCs was small for both healthy controls and IPAH patients. Thus, while the level of detail of -omics profiling was quite deep, the generalizability of their findings to all IPAH patients or Group 1 PAH patients is limited.

We appreciate the reviewers concerns regarding the generalizability of the findings and have acknowledged this as the study limitation in the discussion: “A low case number and end-stage disease samples used for omics characterization represents a study limitation that has to be taken into account before assuming similar findings would be evident in the entire PAH patient population over the course of the disease development and progression”. We have addressed this issue by performing validation of key in vitro findings using fresh cells or assessment of FFPE lung material from additional independent samples in the revised manuscript (Figures 2D, 3D, 3H, 4H). For transparency, we provide biological sample number in the result section of the modified manuscript.

(2) While the study utilized early passage cells, these cells nonetheless were still cultured outside the in vivo milieu prior to analysis. Thus, while there is an assumption that these cells do not change fundamental behavior outside the body, that is not entirely proven for all transcriptional and proteomic signatures. As such, the major alterations that are noted would be more compelling if validated from tissue or cells derived directly from in vivo sources. Without such validation, the major limitation of the impact and conclusions of the paper is that the full extent of the relevance of these findings to human disease is not known.

We thank the reviewer for this constructive and excellent suggestion. The comparison of fresh and cultured cells revealed a strong and early divergence of differentially regulated pathways for PAAF, while a more gradual transition for PASMC. The results of this analysis are included in the new Figures 2D, 3D, 3H, and 4H. Implications are discussed in the revised manuscript: “However, the same mechanism renders cells susceptible to phenotypic change induced simply by extended vitro culturing, testified by broad expression profile differences between fresh and cultured cells. This common caveat in cell biology research and represents a technical and practical tradeoff that requires cross validation of key findings. Using a combination of archived lung tissue and available single cell RNA sequencing dataset of human pulmonary arteries, we show that some of the key defining phenotypic features of diseased cells, such as altered proliferation rate and ECM production, are preserved and gradually lost upon prolonged culturing”.

(3) While the presentation of most of the manuscript was quite clear and convincing, the terminology and conclusions regarding "cell fate trajectories" throughout the manuscript did not seem to be fully justified. That is, all of the analyses were derived from cells originating from end-stage IPAH, and otherwise, the authors were not lineage tracing across disease initiation or development (which would be impossible currently in humans). So, while the description of distinct "IPAH-associated states" makes sense, any true cell fate trajectory was not clearly defined.

In accordance to reviewer’s comment, we have decided to modify the wording to exclude the “cell fate trajectory” phrase and replace it with “acquisition of disease cell state”.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Major comments:

(1) In Figure 1, PASMC and PAAF were collected from the lungs of healthy donors and analyzed for transcriptomics and proteomics; in Figure 1A, it can be taken as if both cells from IPAH patients were also analyzed, but this is not reflected in the results. In Figure1D, immunostaining of normal lungs confirms the localization of PASMC and PAAF markers found by transcriptomics. The authors describe a strong, but not perfect, correlation between the transcriptomics and proteomics data from Figure S1, but the gene names of each cellular marker they found should also be listed. In addition, the authors have observed the expression of markers characteristic of PASMC and PAAF in pulmonary vessels of healthy subjects by IH, but is there any novelty in these markers? Furthermore, are the expression sites of these markers altered in IPAH patients?

In the revised manuscript we have adjusted the schematic to reflect the fact that only donor cells are compared in Figure 1. We additionally provide a correlation of cell type markers between proteomic and transcriptomic data sets for those molecules that are detected in both datasets (Figure S1B).

We provide clarification on the novelty aspect in the result section: “Some of the molecules were previously associated with predominant SMC, such as RGS5 and CSPR1 (Crnkovic et al., 2022; Snider et al., 2008), or adventitial fibroblast, such as SCARA5, CFD and MGST1 (Crnkovic et al., 2022; Sikkema et al., 2023) expression”. Except for RGS5, expression and localization of other markers in IPAH was previously unknown.

The conservation of expression sites for reported markers was validated in IPAH in the revised manuscript (Figure 2D), with IGFBP5 showing dual localization in both cell types. Moreover, results in Figure 1D, 1E and 2D support the validity of omics findings and preservation of key markers during passaging.

(2) In Figure 2, the authors compare PASMC and PAAF derived from IPAH patients and donors. The results show that transcriptomics and proteomics changes are clearly differentiated by cell type and not by pathological state. In the pathological state, transcriptional changes are more pronounced. The GO analysis of the factors that showed significant changes in each cell type is shown in Figure 2E, but the differences between the GO analysis of the transcriptomics and proteomics results are not clearly shown. The reviewer believes that the advantages of a combined analysis of both should be indicated. Also, in Figure 2G, the GAG content in PA appears to be elevated in only 3 cases, while the other 5 cases appear to be at the same level as the donor; is there a characteristic change in these 3 cases? Figure 2I shows that the phenotype of PAAF changes with cell passages. Since this phenomenon would be interesting and useful to the reader, additional discussion regarding the mechanism would be desired.

We have integrated both data sets in order to achieve stronger and meaningful analysis due to weaker and uncomplete correlation between transcriptomic and protein dataset as indicated in the results section: “Comparative analysis of transcriptomic and proteomic data sets revealed a strong, but not complete level of linear correlation between the gene and protein expression profiles (Figure S1B, C). We therefore decided to use an integrative dataset and analyzed all significantly enriched genes and proteins (-log10(P)>1.3) between both cell types to achieve stronger and more robust analysis”. In general, proteomic profile showed fewer significant differences and extent of change was lesser compared with transcriptomics, likely due to technical limitations of the method and sensitivity, testified by the complete lack of top transcriptomic molecules (RGS5, ADH1C, IGFBP5, CFD, SCARA5) in the protein dataset.

To strengthen the findings of increased GAG in IPAH pulmonary arteries, we have performed compartment-specific, quantitative image analysis of Alcian blue staining on additional donor and patient samples (n=10 for each condition). The new analysis totaling around 40 PA confirmed significantly increased deposition of GAG in IPAH pulmonary arteries.

We have addressed the issue of phenotypic change with prolonged cell culture in the revised manuscript by systematically comparing enrichment for biological processes between fresh (Crnkovic et al., 2022: GSE210248), very early (this study: GSE255669) and later passage cells (Chelladurai et al., 2022: GSE144932; Gorr et al., 2020: GSE144274). We observed cell type differences in the rate of change of phenotypic features, with PAAF showing faster shift early on during culturing that could for some of the features be due to isolation from immunomodulatory environment or presence of hydrocortisone supplement in the PAAF cell media. These points have been described in the revised results section and mentioned in the discussion.

(3) The authors claim that one feature of this paper is the use of "very early passage (p1)" of pulmonary artery smooth muscle cells (PASMC). Since there are other existing (previouly reported) data that are publicly available, such as RNA-seq data using cells with 2-4 cell passages, it may be possible to show that fewer passages are better in primary culture by comparing the data presented in this paper.

Following reviewers’ comments, we have performed systematic comparison (Crnkovic et al., 2022: GSE210248), very early (this study: GSE255669) and later passage cells (Chelladurai et al., 2022: GSE144932; Gorr et al., 2020: GSE144274). in the revised manuscript in order to comprehensively address the issue and define changes occurring as a result of prolonged in vitro conditions (Figure 3H). The results showed that the expression profile of early passage cells retains some of the key phenotypic features displayed by cells in their native environment, with PASMC displaying a more gradual loss of phenotypic characteristics compared to PAAF. Interestingly, PAAF displayed a striking inverse enrichment for inflammatory/NF-kB signaling between fresh and cultured PAAF, which could potentially be caused by the hydrocortisone supplement in the PAAF cell media or due to the isolation from its highly immunomodulatory enviroment. These points have been described in the revised results section and mentioned in the discussion.

(4) The authors describe a study characterized by decreased expression of "cytoskeletal contractile elements" in pulmonary artery smooth muscle cells (PASMC) derived from patients with IPAH. What are the implications of this result, and does it arise from the use of smooth muscle in patients resistant to pulmonary artery smooth muscle dilating agents? A discussion on this issue needs to be made in a way that is easy for the reader to understand.

The reviewer raises an interesting point regarding the loss the contractile markers and response to vasodilating therapy. We would speculate that isolated decrease in contractile machinery, without concomitant change in ECM and other PASMC features, would dampen both the contraction and relaxation properties of the single PASMC, affecting not only its response to dilating agents, but also to vasoconstrictors. Clinical consequences and responsiveness to dilating agents are more difficult to predict, since the vasoactive response would additionally depend on mechanical properties of the pulmonary artery defined by cellular and ECM composition. Nevertheless, we believe that decreased expression of contractile machinery reflects an intrinsic, “programmed” response of SMC to remodeling, rather than vasodilator therapy-induced selection pressure, since similar phenotypic change is observed in SMC from systemic circulation and in various animal models without exposure to PAH medication. These considerations have been included in the revised discussion section.

(5) There are a lot of secreted proteins that increase or decrease in Figure 6G, but there is scant reason to focus on PTX3 and HGF among them. The authors need to elaborate on the above issue.

We regret the lack of clarity and provide improved explanation of the ligand selection strategy in the revised manuscript. In order to prioritize the potential hits, we first used hierarchical clustering to group co-regulated ligands into smaller number of groups. We then prioritized for the ligands that lacked or had limited information with respect to IPAH. Based on these results, we analyzed the effect of three additional ligands on PASMC cell state marker expression (Figure S8). This additional data supported the initial focus on PTX3 and HGF.

Minor comments:

(1) Regarding the number of specimens used in the Result, it would be more helpful to the reader if the number of samples were also mentioned in the text.

We have included the number of used samples in manuscript text.

(2) There is no explanation of what R2Y represents in Figure 2B. This reviewer is not able to understand the statistical analysis of Figure 2H. The detailed results should be explained.

We apologize for the oversight in labeling of Figure 2B and modify the figure legend: “Orthogonal projection to latent structures-discriminant analysis (OPLS-DA) T score plots separating predictive variability (x-axis), attributed to biological grouping, and non-predictive variability (technical/inter-individual, y-axis). Monofactorial OPLS-DA model for separation according to cell type or disease. C) Bifactorial OPLS-DA model considering cell type and disease simultaneously. Ellipse depicting the 95% confidence region, Q2 denoting model’s predictive power (significance: Q2>50%) and R2Y representing proportion of variance in the response variable explained by the model (higher values indicating better fit)”.

We also modified figure legend wording for the analysis in Figure 2H (new Figure 3E) to clarify the independent factors whose interaction was investigated using 3-way ANOVA: “Interaction effects of stimulation, cell type, and disease state on cellular proliferation were analyzed by 3-way ANOVA. Significant interaction effects are indicated as follows: * for stimulation × cell type interactions and # for cell type × disease state interactions (both *, # p<0.05)”.

(3) In Figure 3, the authors examined whether there were molecular abnormalities common to IPAH-PASMC and IPAH-PAAF and found that the number of commonly regulated genes and proteins was limited to 47. Further analysis of these regulators by STRING analysis revealed that factors related to the regulation of apoptosis are commonly altered in both cells. On the other hand, the authors focused on mitochondria, as SOD2 is downregulated, and found an increase in ROS production specific to PASMC, indicating that mitochondrial dysfunction is common to PASMC and PAAF in IPAH, but downstream phenomena are different between cell types. Factors associated with apoptosis regulation have been found to be both upward and downward regulated, but the actual occurrence of apoptosis in both cell types has not been addressed.

We have performed TUNEL staining on FFPE lung tissue from donors and IPAH patients that revealed apoptosis as a rare event in both conditions in PASMC and PAAF. Therefore, no meaningful quantification could be conducted. An example of pulmonary artery where rare positive signal in either PAAF or PASMC could be found is provided in Figure 4H.

Unfortunately, association of a particular gene with a pathway is by default arbitrary and potentially ambiguous. In particular, factors identified as associated in apoptosis are also involved in regulation of inflammatory signaling (BIRC3, DDIT3) and amino acid metabolism (SHMT1). Nevertheless, mitochondria represent a crucial cellular hub for apoptosis regulation and, as shown in the current study, display significant functional alterations in IPAH in both cell types, aligning with reduced mitochondrial superoxide dismutase (SOD2) expression.

(4) The meaning of the gray circle in Figure 3C should be clarified. Similarly, the meaning of the color in Fig. 3D should be clearly explained. In Figure 3E-G, each cell is significantly different from 18-61 cells, and the number of each cell and the reason should be described.

We regret the confusion and provide better explanation of the figure legend: “gray nodes representing their putative upstream regulators”, “with color coding reflecting the IPAH dependent regulation”. In the revised Figure panels 4E-G (old 3E-G) we provide the exact number of cells measured in each condition. Although we tried to have comparable cell confluency at the time of measurement, different proliferation rates between cells from different cell type and condition led to different number of measured cells per donor/patient.

(5) In Figure 4, the authors focus on factors that vary in different directions between cells, revealing fingerprints of molecular changes that differ between cell types, particularly IPAH-PASMC, which acquires a synthetic phenotype with enhanced regulation of chemotaxis elements, whereas IPAH-PAAF, a fast cycling cell characteristics. Next, focusing on the ECM components that were specifically altered in IPAH-PASMC, Nichenet analysis in Figure 5 suggested that ligands from PAAF may act on PASMC, and the authors focused on integrin signaling to examine ECM contact and changes in cell function. The results indicate that adhesion to laminin is poor in PASMC. Although no difference was observed between donor and IPAH PASMCs, a discussion of the reasons for this would be desired and helpful to the readers.

Both donor and IPAH PASMCs respond similarly to laminin. However, our key finding is the downregulation of laminin in IPAH PAAF, which likely leads to a skewed laminin-to-collagen ratio and altered ECM composition in remodeled arteries. This shift in the ECM class results in altered PASMC behavior, affecting both donor and IPAH cells similarly. In the revised manuscript, we demonstrate that PASMC largely retain the expression pattern of integrin subunits that serve as high-affinity collagen and laminin receptors, with higher levels compared to PAAF (Figure 6F, G). Furthermore, we speculate that the distinct cellular phenotypic responses to collagen versus laminin coatings may arise from different downstream signaling pathways activated by the various integrin subunits (Nguyen et al., 2000). These considerations have been included in the revised discussion: “The comparable responses of donor and IPAH PASMC likely result from their shared integrin receptor expression profiles. Meanwhile, ECM class switching engages different high-affinity integrin receptors, which activate alternative downstream signaling pathways (Nguyen et al., 2000) and lead to differential responses to collagen and laminin matrices. We thus propose a model in which laminins and collagens act as PAAF-secreted ligands, regulating PASMC behavior through their ECM-sensing integrin receptors.”

(6) Since Figure 3B and Figure 4A seem to show the same results, why not combine them into one?

Indeed, these figure panels show the same results, but the focus of the investigations in each Figure is different. We therefore opted to keep the panels separate for better clarity and logical link to other panels in the same figure

(7) In Figure 6, the interaction analysis of scRNAseq data with respect to signaling between PASMC and PAAF was performed using Nichenet and CellChat, showing that signaling from PAAF to PASMC is biased toward secreted ligands and that a functionally relevant set of soluble ligands is impaired in the IPAH state. From there, they proceeded with co-culture experiments and showed that co-culture healthy PASMC with PAAF of IPAH patients abolished PASMC markers in the healthy state. Furthermore, the authors attempted to identify ligands that induce functional changes in PASMCs produced from IPAH PAAFs and found that HGF is a factor that downregulates the expression of contractile markers in PASMCs. Further insights may be gained by co-culturing IPAH-derived cells in co-culture experiments. Also, no beneficial effect of pentraxin3 was found in Figure 6H. The authors should examine the effect of pentraxin3 on PASMC cells derived from IPAH patients, rather than healthy donors.

We tested the influence of IPAH-PASMC on donor-PAAF and found no effect on the expression of the selected markers. We thank the reviewer for the suggestion to conduct the experiments on IPAH-PASMC. The new data show that both PTX3 and HGF have a significant effect, but differential effect on IPAH-PASMC as compared to donors-PASMC. Whereas PTX lacks effect on donor PASMC, it leads to downregulation of some of the contractile markers in IPAH PASMC, while HGF upregulates VCAN synthetic marker in IPAH PASMC. These results are now included in Figure 7H.

Reviewer #2 (Recommendations For The Authors):

The authors should double-check for grammar and typos in the manuscript. I caught a few such as "therefor" and others, but there could be more.

We thank the reviewer for the effort and time in reading and evaluating the manuscript. To the best of our knowledge, we have corrected the grammatical errors in the revised manuscript.

eLife Assessment

The paper presents a valuable theoretical treatment of the role of passage of time in optimal decision strategies in pursuit based tasks. The computational evidence and methodologies employed are novel, and the authors offer solid evidence for the majority of the claims.

Reviewer #2 (Public review):

Summary:

This paper from Sutlief et al. focuses on an apparent contradiction observed in experimental data from two related types of pursuit-based decision tasks. In "forgo" decisions, where the subject is asked to choose whether or not to accept a presented pursuit, after which they are placed into a common inter-trial interval, subjects have been shown to be nearly optimal in maximizing their overall rate of reward. However, in "choice" decisions, where the subject is asked which of two mutually-exclusive pursuits they will take, before again entering a common inter-trial interval, subjects exhibit behavior that is believed to be sub-optimal. To investigate this contradiction, the authors derive a consistent reward-maximizing strategy for both tasks using a novel and intuitive geometric approach that treats every phase of a decision (pursuit choice and inter-trial interval) as vectors. From this approach, the authors are able to show that previously-reported examples of sub-optimal behavior in choice decisions are in fact consistent with a reward-maximizing strategy. Additionally, the authors are able to use their framework to deconstruct the different ways the passage of time impacts decisions, demonstrating the time cost contains both an opportunity cost and an apportionment cost, as well as examine how a subject's misestimation of task parameters impacts behavior.

Strengths:

The main strength of the paper lies in the authors' geometric approach to studying the problem. The authors chose to simplify the decision process by removing the highly technical and often cumbersome details of evidence accumulation that is common in most of the decision-making literature. In doing so, the authors were able to utilize a highly accessible approach that is still able to provide interesting insights into decision behavior and the different components of optimal decision strategies.

Weaknesses:

The authors have made great improvements to the strength of their evidence through revision, especially concerning their treatment of apportionment cost. However, I am concerned that the story this paper tells is far from concise, and that this weakness may limit the paper's audience and overall impact. I would strongly suggest making an effort to tighten up the language and structure of the paper to improve its readability and accessibility.

Reviewer #3 (Public review):

Summary:

The goal of the paper is to examine the objective function of total reward rate in an environment to understand behavior of humans and animals in two types of decision-making tasks: 1) stay/forgo decisions and 2) simultaneous choice decisions. The main aims are to reframe the equation of optimizing this normative objective into forms that are used by other models in the literature like subjective value and temporally discounted reward. One important contribution of the paper is the use of this theoretical analysis to explain apparent behavioral inconsistencies between forgo and choice decisions observed in the literature.

Strengths:

The paper provides a nice way to mathematically derive different theories of human and animal behavior from a normative objective of global reward rate optimization. As such, this work has value in trying to provide a unifying framework for seemingly contradictory empirical observations in literature, such as differentially optimal behaviors in stay-forgo v/s choice decision tasks. The section about temporal discounting is particularly well motivated as it serves as another plank in the bridge between ecological and economic theories of decision-making. The derivation of the temporal discounting function from subjective reward rate is much appreciated as it provides further evidence for potential equivalence between reward rate optimization and hyperbolic discounting, which is known to explain a slew of decision-making behaviors in the economics literature.

Weaknesses:

(1) Readability and organization:<br /> While I appreciate the detailed analysis and authors' attempts to provide as many details as possible, the paper would have benefitted from a little selectivity on behalf of the authors so that the main contributions aren't buried by the extensive mathematical detail provided.<br /> For instance, in Figure 5, the authors could have kept the most important figures (A, B and G) to highlight the most relevant terms in the subjective value instead of providing all possible forms of the equation.

Further, in subfigure 5E, is there a reason that the outside reward r_out is shown to be zero? The text referencing 5E is also very unclear: "In so downscaling, the subjective value of a considered pursuit (green) is to the time it would take to traverse the world were the pursuit not taken, 𝑡_out, as its opportunity cost subtracted reward (cyan) is to the time to traverse the world were it to be taken (𝑡_in+ 𝑡_out) (Figure 5E)."

In the abstract, the malapportionment of time is mentioned as a possible explanation for reconciling observed empirical results between simultaneous and sequential decision-making. However, perhaps due to the density of mathematical detail presented, the discussion of the malapportionment hypothesis is pushed all the way to the end of the discussion section.

(2) Apportionment Cost definition and interpretation<br /> This additional cost arises in their analyses from redefining the opportunity cost in terms of just "outside" rewards so that the subjective value of the current pursuit and the opportunity cost are independent of each other. However, in doing so, an additional term arises in defining the subjective value of a pursuit, named here the "apportionment cost". The authors have worked hard to provide a definition to conceptualize the apportionment cost though it remains hard to intuit, especially in comparison to the opportunity cost. The additive form of apportionment cost (Equation 9) doesn't add much in way of intuition or their later analyses for the malapportionment hypothesis. It appears that the most important term is the apportionment scaling term so just focusing on this term will help the reader through the subsequent analyses.

(3) Malapportionment Hypothesis: From where does this malapportionment arise?<br /> The authors identify the range of values for t_in and t_out in Figure 18, the terms comprising the apportionment scaling term, that lead to optimal forgo behaviors despite suboptimally rejecting the larger-later (LL) choice in choice decisions. They therefore conclude that a lower apportionment scale, which arises from overestimating the time required outside the pursuit (t_out) or underestimating the time required at the current pursuit (t_in). What is not discussed though is whether and how the underestimation of t_out and overestimation of t_in can be dissociated, though it is understood that empirical demonstration of this dissociation is outside the scope of this work.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

(1) Although there are many citations acknowledging relevant previous work, there often isn't a very granular attribution of individual previous findings to their sources. In the results section, it's sometimes ambiguous when the paper is recapping established background and when it is breaking new ground. For example, around equation 8 in the results (sv = r - rho*t), it would be good to refer to previous places where versions of this equation have been presented. Offhand, McNamara 1982 (Theoretical Population Biology) is one early instance and Fawcett et al. 2012 (Behavioural Processes) is a later one. Line 922 of the discussion seems to imply this formulation is novel here.

We would like to clarify that original manuscript equation 8, , as we derive, is not new, as it is similarly expressed in prior foundational work by McNamara (1982), and we thank the reviewer for drawing our attention to the extension of this form by Fawcett, McNamara, Houston (2012).

We now so properly acknowledge this foundational work and extension in the results section…

“This global reward-rate equivalent immediate reward (see Figure 4) is the subjective value of a pursuit, svPursuit (or simply, sv, when the referenced pursuit can be inferred), as similarly expressed in prior foundational work (McNamara 1982), and subsequent extensions (see (Fawcett, McNamara, Houston (2012)).”

…and in the Discussion section at the location referenced by the reviewer:

“From it, we re-expressed the pursuit’s worth in terms of its global reward rate-equivalent immediate reward, i.e., its ‘subjective value’, reprising McNamara’s foundational formulation (McNamara 1982).”

(2) The choice environments that are considered in detail in the paper are very simple. The simplicity facilitates concrete examples and visualizations, but it would be worth further consideration of whether and how the conclusions generalize to more complex environments. The paper considers "forgo" scenario in which the agent can choose between sequences of pursuits like A-B-A-B (engaging with option B at all opportunities, which are interleaved with a default pursuit A) and A-A-A-A (forgoing option B). It considers "choice" scenarios where the agent can choose between sequences like A-B-A-B and A-C-A-C (where B and C are larger-later and smaller-sooner rewards, either of which can be interleaved with the default pursuit). Several forms of additional complexity would be valuable to consider. [A] One would be a greater number of unique pursuits, not repeated identically in a predictable sequence, akin to a prey-selection paradigm. It seems to me this would cause t_out and r_out (the time and reward outside of the focal prospect) to be policy-dependent, making the 'apportionment cost' more challenging to ascertain. Another relevant form of complexity would be if there were [B] variance or uncertainty in reward magnitudes or temporal durations or if [C] the agent had the ability to discontinue a pursuit such as in patch-departure scenarios.

A) We would like to note that the section “Deriving Optimal Policy from Forgo Decision-making worlds”, addresses the reviewer’s scenario of n-number of pursuits”, each occurring at their own frequency, as in prey selection, not repeating identically in a predictable sequence. Within our subsection “Parceling the world…”, we introduce the concept of dividing a world (such as that) into the considered pursuit type, and everything outside of it. ‘Outside’ would include any number of other pursuits currently part of any policy, as the reviewer intuits, thus making t^out and r^out policy dependent. Nonetheless, a process of excluding (forgoing) pursuits by comparing the ‘in’ to the ‘out’ reward rate (section “Reward-rate optimizing forgo policy…”) or its equivalent sv (section “The forgo decision can also be made from subjective value), would iteratively lead to the global reward rate maximizing policy. This manner of parceling into ‘in’ and ‘out’ thus simplifies visualization of what can be complex worlds. Simpler cases that resemble common experimental designs are given in the manuscript to enhance intuition.

We thank the reviewer for this keen suggestion. We now include example figures (Supplemental 1 & 2) for multi-pursuit worlds which have the same (Supplemental 1) and different pursuit frequencies (Supplemental 2), which illustrate how this evaluation leads to reward-rate optimization. This addition demonstrates how an iterative policy would lead to reward rate maximization and emphasizes how parcellating a world into ‘in’ and ‘out’ of the pursuit type applies and is a useful device for understanding the worth of any given pursuit in more complex worlds. The policy achieving the greatest global reward rate can be realized through an iterative process where pursuits with lower reward rates than the reward rate obtained from everything other than the considered pursuit type are sequentially removed from the policy.

B) We would also emphasize that the formulation here contends with variance or uncertainty in the reward magnitudes or temporal durations. The ‘in’ pursuit is the average reward and the average time of the considered pursuit type, as is the ‘out’ the average reward and average time outside of the considered pursuit type.

C) In this work, we consider the worth of initiating one-or-another pursuit (from having completed a prior one), and not the issue of continuing within a pursuit (having already engaged it), as in patch/give-up. Handling worlds in which the agent may depart from within a pursuit, which is to say ‘give-up’ (as in patch foraging), is outside the scope of this work.

(3) I had a hard time arriving at a solid conceptual understanding of the 'apportionment cost' around Figure 5. I understand the arithmetic, but it would help if it were possible to formulate a more succinct verbal description of what makes the apportionment cost a useful and meaningful quality to focus on.

We thank the reviewer for pressing for a succinct and intuitive verbal description.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in new paragraphs (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

The above definition of apportionment cost adds to other stated relationships of apportionment cost found throughout the paper (original lines 434,435,447,450).

I think Figure 6C relates to this, but I had difficulty relating the axis labels to the points, lines, and patterned regions in the plot.

We thank the reviewer for pointing out that this figure can be made to be more easily understood.

We have done so by breaking its key features over a greater number of plots so that no single panel is overloaded. We have also changed text in the legend to clarify how apportionment and opportunity costs add to constitute time’s cost, and also correspondingly in the main text.

I also was a bit confused by how the mathematical formulation was presented. As I understood it, the apportionment cost essentially involves scaling the rest of the SV expression by t^out/(tⁱⁿ + t^out).

The reviewer’s understanding is correct: the amount of reward of the pursuit that remains after subtracting the opportunity cost, when so scaled, is equivalent to the subjective value of that pursuit. The amount by which that scaling decreases the rest of the SV expression is equal to the apportionment cost of time.

The way this scaling factor is written in Figure 5C, as 1/(1 + (1/t^out) tⁱⁿ), seems less clear than it could be.

To be sure, we present the formula in original Figure 5C in this manner to emphasize the opportunity cost subtraction as separable from the apportionment rescaling, expressing the opportunity cost subtraction and the apportionment scaling component of the equation as their own terms in parentheses.

But we understand the reviewer to be referring to the manner by which we chose to express the scaling term. We presented it in this way in the original manuscript, (rather than its more elegant form recognized by the reviewer) to make direct connection to temporal discounting literature. In this literature, discounting commonly takes the same mathematical form as our apportionment cost scaling, but whereas the steepness of discounting in this literature is controlled by a free fit parameter, k, we show how for a reward rate maximizing agent, the equivalent k term isn’t a free fit parameter, but rather is the reciprocal of the time spent outside the considered pursuit type.

We take the reviewer’s advice to heart, and now first express subjective value in the format that emphasizes opportunity cost subtraction followed by an apportionment downscaling, identifying the apportionment scaling term, t^out/(t^out + tⁱⁿ), ie the outside weight. Figure 5 now shows the geometric representation of apportionment scaling and apportionment cost. Only subsequently in the discounting function section then do we now in the revised manuscript rearrange this subjective value expression to resemble the standard discounting function form.

Also, the apportionment cost is described in the text as being subtracted from sv rather than as a multiplicative scaling factor.

What we describe in the original text is how apportionment cost is a component of time’s cost, and how sv is the reward less time’s cost. It would be correct to say that apportionment cost and opportunity cost are subtracted from the pursuit’s reward to yield the subjective value of the pursuit. This is what we show in the original Figure 5D graphically. Original Figure 5 and accompanying formulas at its bottom show the equivalence of expressing sv in terms of subtracting time’s cost as calculated from the global reward rate under a policy of accepting the considered pursuit, or, of subtracting opportunity cost and then scaling the opportunity cost subtracted reward by the apportionment scaling term, thereby accounting for the apportionment cost of time.

The revision of original figure 5, its figure legend, and accompanying text now make clear the meaning of apportionment cost, how it can be considered a subtraction from the reward of a pursuit, or, equivalently, how it can be thought of as the result of scaling down of opportunity cost subtracted reward.

It could be written as a subtraction, by subtracting a second copy of the rest of the SV expression scaled by t_in/(t_in + t_out). But that shows the apportionment cost to depend on the opportunity cost, which is odd because the original motivation on line 404 was to resolve the lack of independence between terms in the SV expression.

On line 404 of the original manuscript, we point out that the simple equation―which is a reprisal of McNamara’s insight―is problematic in that its terms on the RHS are not independent: the global reward rate is dependent on the considered pursuit’s reward (see Fig5B). The alternative expression for subjective value that we derive expresses sv in terms that are all independent of one another. We may have unintentionally obscured that fact by having already defined rhoⁱⁿ as rⁱⁿ/ tⁱⁿ and rho^out as r^out/t^out on lines 306 and 307.

Therefore, in the revision, Ap 8 is expressed so to keep clear that it uses terms that are all independent of one another, and only subsequently express this formula with the simplifying substitution, rho^out.

That all said, we understand the reviewer’s point to be that the parenthetical terms relating the opportunity cost and the apportionment rescaling both contain within them the parameter t^out, and in this way these concepts we put forward to understand the alternative equation are non-independent. That is correct, but it isn’t at odds with our objective to express SV in terms that are independent with one another (which we do). Our motivation in introducing these concepts is to provide insight and intuition into the cost of time (especially now with a clear and simple definition of apportionment cost stated). We go to lengths to demonstrate their relationship to each other.

(4) In the analysis of discounting functions (line 664 and beyond), the paper doesn't say much about the fact that many discounting studies take specific measures to distinguish true time preferences from opportunity costs and reward-rate maximization.

We understand the reviewer’s comment to connote that temporal decision-making worlds in which delay time does not preclude reward from outside the current pursuit is a means to distinguish time preference from the impact of opportunity cost. One contribution of this work is to demonstrate that, from a reward-rate maximization framework, an accounting of opportunity cost is not sufficient to understand apparent time preferences as distinguishable from reward-rate maximization. The apportionment cost of time must also be considered to have a full appreciation of the cost of time. For instance, let us consider a temporal decision-making world in which there is no reward received outside the considered pursuit. In such a world, there is no opportunity cost of time, so apparent temporal discounting functions would appear as if purely hyperbolic as a consequence of the apportionment cost of time alone. Time preference, as revealed experimentally by the choices made between a SS and a LL reward, then, seem confounding, as preference can reverse from a SS to a LL option as the displacement of those options (maintaining their difference in time) increases (Green, Fristoe, and Myerson 1994; Kirby and Herrnstein 1995). While this shift, the so-called “Delay effect”, could potentially arise as a consequence of some inherent time preference bias of an agent, we demonstrate that a reward-rate maximal agent exhibits hyperbolic discounting, and therefore it would also exhibit the Delay effect, even though it has no time preference.

In the revision we now make reference to the Delay Effect (in abstract, results new section “The Delay Effect” with new figure 14, and in the discussion), which is taken as evidence of time preference in human and animal literature, and note explicitly how a reward-rate maximizing agent would also exhibit this behavior as a consequence of apparent hyperbolic discounting.

In many of the human studies, delay time doesn't preclude other activities.

Our framework is generalizable to worlds in which being in pursuit does not preclude an agent from receiving reward during that time at the outside reward rate. Original Ap 13 solves for such a condition, and shows that in this context, the opportunity cost of time drops out of the SV equation, leaving only the consequences of the apportionment cost of time. We made reference to this case on lines 1032-1034 of the original manuscript: “In this way, such hyperbolic discounting models [models that do not make an accounting of opportunity cost] are only appropriate in worlds with no “outside” reward, or, where being in a pursuit does not exclude the agent from receiving rewards at the rate that occurs outside of it (Ap. 13).”

The note and reference is fleeting in the original work. We take the reviewer’s suggestion and now add paragraphs in the discussion on the difference between humans and animals in apparent discounting, making specific note of human studies in which delay time doesn’t preclude receiving outside reward while engaged in a pursuit. Relatedly, hyperbolic discounting is oft considered to be less steep in humans than in animals. As the reviewer points out, these assessments are frequently made under conditions in which being in a pursuit does not preclude receiving reward from outside the pursuit. When humans are tested under conditions in which outside rewards are precluded, they exhibit far steeper discounting. We now include citation to that observation (Jimura et al. 2009). We handle such conditions in original AP 13, and show how, in such worlds, the opportunity cost of time drops out of the equation. The consequence of this is that the apparent discounting function would become less steep (the agent would appear as if more patient), consistent with reports.

“Relating to the treatment of opportunity cost, we also note that many investigations into temporal discounting do not make an explicit distinction between situations in which 1) subjects continue to receive the usual rewards from the environment during the delay to a chosen pursuit, and 2) situations in which during a chosen pursuit’s delay no other rewards or opportunities will occur (Kable & Glimcher, 2007; Kirby & Maraković, 1996; McClure, Laibson, Loewenstein, & Cohen, 2004). Commonly, human subjects are asked to answer questions about their preferences between options for amounts they will not actually earn after delays they will not actually have to wait, during which it is unclear whether they are really investing time away from other options or not (Rosati et al., 2007). In contrast, in most animal experiments, subjects actually receive reward after different delays during which they do not receive new options or rewards. By our formulation, when a pursuit does not exclude the agent from receiving rewards at the rate that occurs outside, the opportunity cost of time drops out of the subjective value equation (Ap 12).

Equation 10. The value of initiating a pursuit when pursuit does not exclude receiving rewards at the outside rate (Ap 12)

Therefore, the reward-rate maximizing discounting function in these worlds is functionally equivalent to the situation in which the outside reward rate is zero, and will―lacking an opportunity cost―be less steep. This rationalizes why human discounting functions are often reported to be longer (gentler) than animal discounting functions: they are typically tested in conditions that negate opportunity cost, whereas animals are typically tested in conditions that enforce opportunity costs. Indeed, when humans are made to wait for actually received reward, their observed discounting functions are much steeper (Jimura et al. 2009). “

In animal studies, rate maximization can serve as a baseline against which to measure additional effects of temporal discounting. This is an important caveat to claims about discounting anomalies being rational under rate maximization (e.g., line 1024).

We agree that the purpose of this reward-rate maximizing framework is to serve as a point of comparison in which effects of temporal intervals and rewards that define the environment can be analyzed to better understand the manner in which animals and humans deviate from this ideal behavior. Our interest in this work is in part motivated by a desire to have a deeper understanding of what “true” time preference means. Using the reward-rate maximizing framework here provides a means to speak about time preferences (ie biases) in terms of deviation from optimality. From this perspective, a reward-rate maximal agent doesn’t exhibit time preference: its actions are guided solely by reward-rate optimizing valuation. Therefore, one contribution of this work is to show that purported signs of time preference (hyperbolic discounting, magnitude, sign, and (now) delay effect) can be explained without invoking time preference. What errors from optimality that remain following an proper accounting of reward-rate maximizing behavior should then, and only then, be considered from the lens of time preference (bias).

(5) The paper doesn't feature any very concrete engagement with empirical data sets. This is ok for a theoretical paper, but some of the characterizations of empirical results that the model aims to match seem oversimplified. An example is the contention that real decision-makers are optimal in accept/reject decisions (line 816 and elsewhere). This isn't always true; sometimes there is evidence of overharvesting, for example.

We would like to note that the scope of this paper is limited to examining the value of initiating a pursuit, rather than the value of continuing within a pursuit. The issue of continuing within a pursuit constitutes a third fundamental topology, which could be called give-up or patch-foraging, and is complex and warrants its own paper. In Give-up topologies, which are distinct from Forgo, and Choice topologies, the reviewer is correct in pointing out that the preponderance of evidence demonstrates that animals and humans are as if overpatient, adopting a policy of investing too much time within a pursuit, than is warranted_._ In Forgo instances, however, the evidence supports near optimality.

(6) Related to the point above, it would be helpful to discuss more concretely how some of this paper's theoretical proposals could be empirically evaluated in the future. Regarding the magnitude and sign effects of discounting, there is not a very thorough overview of the several other explanations that have been proposed in the literature. It would be helpful to engage more deeply with previous proposals and consider how the present hypothesis might make unique predictions and could be evaluated against them.

We appreciate the reviewer’s point that there are many existing explanations for these various ‘anomalous’ effects. We hold that the point of this work is to demonstrate that these effects are consistent with a reward-rate maximizing framework so do not require additional assumptions, like separate processes for small and large rewards, or the inclusion of a utility function.

Nonetheless, there is a diversity of explanations for the sign and magnitude effect, and, (now with its explicit inclusion in the revision) the delay effect. Therefore, we now also include reference to additional work which proffers alternative explanations for the sign and magnitude effects, (as reviewed by (Kalenscher and Pennartz 2008; Frederick et al. 2002)), as well as a scalar timing account of non-stationary time preference (Gibbon, 1977).

With respect to making predictions, this framework makes the following in regards to the magnitude, sign, and (now in the revision) delay effect: in Discussion, Magnitude effect subsection: “The Magnitude Effect should be observed, experimentally, to diminish when 1) increasing the outside time while holding the outside reward constant, (thus decreasing the outside reward rate), or when 2) decreasing the outside reward while holding the outside time constant (thus decreasing the outside reward rate). However, 3) the Magnitude Effect would exaggerate as the outside time increased while holding the outside reward rate constant.”, in Sign effect subsection: “…we then also predict that the size of the Sign effect would diminish as the outside reward rate decreases (and as the outside time increases), and in fact would invert should the outside reward rate turn negative (become net punishing), such that punishments would appear to discount more steeply than rewards.” Delay effect subsection: “...a sign of irrationality is that a preference reversal occurs at delays greater than what a reward-rate-maximizing agent would exhibit.”

A similar point applies to the 'malapportionment hypothesis' although in this case there is a very helpful section on comparisons to prior models (line 1163). The idea being proposed here seems to have a lot in common conceptually with Blanchard et al. 2013, so it would be worth saying more about how data could be used to test or reconcile these proposals.

We thank the reviewer for holding that the section of model comparisons to be very helpful. We believe the text previously dedicated to this issue to be sufficient in this regard. We have, however, adding substantively to the Malapportionment Hypothesis section (Discussion) and its accompanying figure, to make explicit a number of predictions from the Malapportionment hypothesis as it relates to Hyperbolic discounting, the Delay Effect, and the Sign and Magnitude Effects.

Reviewer #1 Recommendations

(1) As a general note about the figures, it would be helpful to specify, either graphically or in the caption, what fixed values of reward sizes and time intervals are being assumed for each illustration.

Thank you for the suggestion. We attempted to keep graphs as uncluttered as possible, but agree that for original figures 4,5,16, and 17, which didn’t have numbered axes, that we should provide the amounts in the captions in the revised figures (4,5, and now 17,18). These figures did not have numerics as their shapes and display are to illustrate the form of the relationship between vectors, being general to the values they may take.

We now include in the captions for these figures the parameter amounts used.

(2) Should Equation 2 have t in the denominator instead of r?

Indeed. We thank the reviewer for catching this typographical error.

We have corrected it in the revision.

(3) General recommendation:

My view is that in order for the paper's eLife assessment to improve, it would be necessary to resolve points 1 through 4 listed under "weaknesses" in my public review, which pertain to clarity and acknowledgement of prior work. I think a lot hinges on whether the authors can respond to point #3 by making a more compelling case for the usefulness and generality of the 'apportionment cost' concept, since that idea is central to the paper's contribution.

We believe these critical points (1-4) to improve the paper will now have been addressed to the reviewer’s satisfaction.

Reviewer #2 (Public review):

While the details of the paper are compelling, the authors' presentation of their results is often unclear or incomplete:

(1) The mathematical details of the paper are correct but contain numerous notation errors and are presented as a solid block of subtle equation manipulations. This makes the details of the authors' approach (the main contribution of the paper to the field) highly difficult to understand.

We thank the reviewers for having detected typographical errors regarding three equations. They have been corrected. The first typographical error in the original main text (Line 277) regards equation 2 and will be corrected so that equation 2 appears correctly as

The second typo regards the definition of the considered pursuit’s reward rate which appear in the original main text (line 306), and has been corrected to appear as

The third typographical error occurred in conversion from Google Sheets to Microsoft Word appearing in the original main text (line 703) and regards the subjective value expression when no reward is received in an intertrial interval (ITI). It has been corrected to appear as

(2) One of the main contributions of the paper is the notion that time’s cost in decision-making contains an apportionment cost that reflects the allocation of decision time relative to the world. The authors use this cost to pose a hypothesis as to why subjects exhibit sub-optimal behavior in choice decisions. However, the equation for the apportionment cost is never clearly defined in the paper, which is a significant oversight that hampers the effectiveness of the authors' claims.

We thank the reviewer for pressing on this critical point. Reviewers commonly identified a need to provide a concise and intuitive definition of apportionment cost, and to explicitly solve and provide for its mathematical expression.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in new paragraphs (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

(3) Many of the paper's figures are visually busy and not clearly detailed in the captions (for example, Figures 6-8). Because of the geometric nature of the authors' approach, the figures should be as clean and intuitive as possible, as in their current state, they undercut the utility of a geometric argument.

We endeavored to make our figures as simple as possible. We have made in the revision changes to figures that we believe improve their clarity. These include: 1) breaking some figures into more panels when more than one concept was being introduced (such as in revised Figure 5 , 6, 7, and 8), 2) using the left hand y axis for the outside reward, and the right hand axis for the inside reward when plotting the “in” and “outside” reward, and indicating their respective numerics (which run in opposite directions), 3) adding a legend to the figures themselves where needed (revised figures 10, 11, 12, 14) 4) adding the values used to the figure captions, where needed, and 5) ensuring all symbols are indicated in legends.

(4) The authors motivate their work by focusing on previously-observed behavior in decision experiments and tell the reader that their model is able to qualitatively replicate this data. This claim would be significantly strengthened by the inclusion of experimental data to directly compare to their model's behavior. Given the computational focus of the paper, I do not believe the authors need to conduct their own experiments to obtain this data; reproducing previously accepted data from the papers the authors' reference would be sufficient.

Our objective was not to fit experimentally observed data, as is commonly the goal of implementation/computational models. Rather, as a theory, our objective is to rationalize the broad, curious, and well-established pattern of temporal decision-making behaviors under a deeper understanding of reward-rate maximization, and from that understanding, identify the nature of the error being committed by whatever learning algorithm and representational architecture is actually being used by humans and animals. In doing so, we make a number of important contributions. By identifying and analyzing reward-rate-maximizing equations, we 1) provide insight into what composes time’s cost and how the temporal structure of the world in which it is embedded (its ‘context’) impacts the value of a pursuit, 2) rationalize a diverse assortment of temporal decision-making behaviors (e.g., Hyperbolic discounting, the Magnitude Effect, the Sign Effect, and the Delay effect), explaining them with no assumed free-fit parameter, and then, by analyzing error in parameters enabling reward-rate maximization, 3) identify the likely source of error and propose the Malapportionment Hypothesis. The Malapportionment Hypothesis identifies the underweighting of a considered pursuit’s “outside”, and not error in pursuit’s reward rates, as the source of error committed by humans and animals. It explains why animals and humans can present as suboptimally ‘impatient’ in Choice, but as optimal in Forgo. At the same time, it concords with numerous and diverse observations in decision making regarding whether to initiate a pursuit. The nature of this error also, then, makes numerous predictions. These insights inform future computational and experimental work by providing strong constraints on the nature of the algorithm and representational architecture used to learn and represent the values of pursuits. Rigorous test of the Malapportionment Hypothesis will require wholly new experiments.

In the revision, we also now emphasize and add predictions of the Malapportionment Hypothesis, updated its figure (Figure 21), its legend, and its paragraphs in the discussion.

“We term this reckoning of the source of error committed by animals and humans the Malapportionment Hypothesis, which identifies the underweighting of the time spent outside versus inside a considered pursuit but not the misestimation of pursuit rates, as the source of error committed by animals and humans (Figure 21). This hypothesis therefore captures previously published behavioral observations (Figure 21A) showing that animals can make decisions to take or forgo reward options that optimize reward accumulation (Krebs et al., 1977; Stephens and Krebs, 1986; Blanchard and Hayden, 2014), but make suboptimal decisions when presented with simultaneous and mutually exclusive choices between rewards of different delays (Logue et al., 1985; Blanchard and Hayden, 2015; Carter and Redish, 2016; Kane et al., 2019). The Malapportionment Hypothesis further predicts that apparent discounting functions will present with greater curvature than what a reward-rate-maximizing agent would exhibit (Figure 21B). While experimentally observed temporal discounting would have greater curvature, the Malapportionment Hypothesis also predicts that the Magnitude (Figure 21C) and Sign effect (Figure 21D) would be less pronounced than what a reward-rate-maximizing agent would exhibit, with these effects becoming less pronounced the greater the underweighting. Finally, with regards to the Delay Effect (Figure 21E), the Malapportionment Hypothesis predicts that preference reversal would occur at delays greater than that exhibited by a reward-rate-maximizing agent, with the delay becoming more pronounced the greater the underweighting outside versus inside the considered pursuit by the agent.”

(5) While the authors reference a good portion of the decision-making literature in their paper, they largely ignore the evidence-accumulation portion of the literature, which has been discussing time-based discounting functions for some years. Several papers that are both experimentally-(Cisek et al. 2009, Thurs et al. 2012, Holmes et al. 2016) and theoretically-(Drugowitsch et al. 2012, Tajima et al. 2019, Barendregt et al. 22) driven exist, and I would encourage the authors to discuss how their results relate to those in different areas of the field.

In this manuscript, we consider the worth of initiating one or another pursuit having completed a prior one, and not the issue of continuing within a pursuit having already engaged in it. The worth of continuing a pursuit, as in patch-foraging/give-up tasks, constitutes a third fundamental time decision-making topology which is outside the scope of the current work. It engages a large and important literature, encompassing evidence accumulation, and requires a paper on the value of continuing a pursuit in temporal decision making, in its own right, that can use the concepts and framework developed here. The excellent works suggested by the reviewer will be most relevant to that future work concerning patch-foraging/give-up topologies.

Reviewer #2 Recommendations:

(1) In Equation 1, the term rho_d is referred to as the reward rate of the default pursuit, when it should be the reward of the default pursuit.

Regarding Equation 1, it is formulated to calculate the average reward received and average time spent per unit time spent in the default pursuit. So, f_i is the encounter rate of pursuit i for one unit of time spent in the default pursuit (lines 259-262). Added to the summation in the numerator, we have the average reward obtained in the default pursuit per unit time () and in the denominator we have the time spent in the default pursuit per unit time (1).

We have added clarifying text to assist in meaning of the equation in Ap 1, and thank the reviewer for pointing out this need.

(2) The notation for "in" and "out" of a considered pursuit type begins as being used to describe the contribution from a single pursuit (without inter-trial interval) towards global reward rate and the contribution of all other factors (other possible pursuits and inter-trial interval) towards global reward rate, respectively, but is then used to describe the pursuit's contribution and the inter-trial interval's contribution, respectively, to the global reward rate. This should be cleaned up to be consistent throughout, or at the very least, it should be addressed when this special case is considered the default.

As understood by the reviewer, “in” and “out” of the considered pursuit type describes the general form by which a world can be cleaved into these two parts: the average time and reward received outside of the considered pursuit type for the average time and reward received within that pursuit type. A specific, simple, and common experimental instance would be a world composed of one or another pursuit and an intertrial interval.

We now make clear how such a world composed of a considered pursuit and an inter trial interval would be but one special case. In example cases where t^out represents the special case of an inter-trial interval, this is now stated clearly. For instance, we do so when discussing how a purely hyperbolic discounting function would apply in worlds in which no reward is received in t^out, stating that this is often the case common to experimental designs where t^out represents an intertrial interval with no reward. Importantly, by the new inclusion of illustrated worlds in the revision that have n-number pursuits that could occur from a default pursuit and 1) equal frequency (Supplemental 1), and 2) at differing frequencies (Supplemental 2), we make more clear the generalizability and utility of this t^out/tin concept.

(3) Figure 5 should make clear the decomposition of time's cost both graphically and functionally. As it stands, the figure does not define the apportionment cost.

In the revision of original fig 5, we now further decompose the figure to effectively convey 1) what opportunity cost, and (especially) 2) the apportionment cost is, both graphically and mathematically, 3) how time’s cost is comprised by them, 4) how the apportionment scaling term scales the opportunity-cost-subtracted reward by time’s allocation to equal the subjective value, and 4) the equivalence between the expression of time’s cost using terms that are not independent of one another with the expression of time’s cost using terms that are independent of one another.

(4) Figures 6-8 do not clearly define the dots and annuli used in panels B and C.

We have further decomposed figures 6-8 so that the functional form of opportunity, apportionment, and time’s cost can be more clearly appreciated, and what their interrelationship is with respect to changing outside reward and outside time, and clearly identify symbols used in the corresponding legends.

(5) The meaning of a negative subjective value should be specifically stated. Is it the amount a subject would pay to avoid taking the considered pursuit?

As the reviewer intuits, negative subjective value can be considered the amount an agent ought be willing to pay to avoid taking the considered pursuit.

We now include the following lines in “The forgo decision can also be made from subjective value” section in reference to negative subjective value…

“A negative subjective value thus indicates that a policy of taking the considered pursuit would result in a global reward rate that is less than a policy of forgoing the considered pursuit. Equivalently, a negative subjective value can be considered the amount an agent ought be willing to pay to avoid having to take the considered pursuit.”

(6) Why do you define the discounting function as the normalized subjective value? This choice should be justified, via literature citations or a well-described logical argument.

The reward magnitude normalized subjective value-time function is commonly referred to as the temporal discounting function as it permits comparison of the discount rate isolated from a difference in reward magnitude and/or sign and is deeply rooted in historical precedent. As the reviewer points out, the term is overloaded, however, as investigations in which comparisons between the form of subjective value-time functions is not needed tend to refer to these functions as temporal discounting functions as well.

We make clear in the revised text in the introduction our meaning and use of the term, the justification in doing so, and its historical roots.

“Historically, temporal decision-making has been examined using a temporal discounting function to describe how delays in rewards influence their valuation. Temporal discounting functions describe the subjective value of an offered reward as a function of when the offered reward is realized. To isolate the form of discount rate from any difference in reward magnitude and sign, subjective value is commonly normalized by the reward magnitude when comparing subjective value-time functions (Strotz, 1956, Jimura, 2009). Therefore, we use the convention that temporal discounting functions are the magnitude-normalized subjective value-time function (Strotz, 1956).”

Special addition. In investigating the historical roots of the discounting function prompted by the reviewer, we learned (Grüne-Yanoff 2015) that it was Mazur that simply added the “1+k” in the denominator of the hyperbolic discounting function. Our derivation for the reward-rate optimal agent makes clear why apparent temporal discounting functions ought have this general form.

Therefore, we add the following to the “Hyperbolic Temporal Discounting Function section in the discussion…

“It was Ainslie (Ainslie, 1975) who first understood that the empirically observed “preference reversals” between SS and LL pursuits could be explained if temporal discounting took on a hyperbolic form, which he initially conjectured to arise simply from the ratio of reward to delay (Grüne-Yanoff 2015). This was problematic, however, on two fronts: 1) as the time nears zero, the value curve goes to infinity, and 2) there is no accommodation of differences observed within and between subjects regarding the steepness of discounting. Mazur (Mazur, 1987) addressed these issues by introducing 1 + k into the denominator, providing for the now standard hyperbolic discounting function, . Introduction of “1” solved the first issue, though “it never became fully clear how to interpret this 1” (Grüne-Yanoff 2015; interviewing Ainslie). Introduction of the free-fit parameter, k, accommodated the variability observed across and within subjects by controlling the curvature of temporal discounting, and has become widely interpreted as a psychological trait, such as patience, or willingness to delay gratification (Frederick et al., 2002).”

…continuing later in that section to explain why the reward-rate optimal agent would exhibit this general form…

“Regarding form, our analysis reveals that the apparent discounting function of a reward-rate-maximizing agent is a hyperbolic function…

…which resembles the standard hyperbolic discounting function, , in the denominator, where . Whereas Mazur introduced 1 + k to t in the denominator to 1) force the function to behave as t approaches zero, and 2) provide a means to accommodate differences observed within and between subjects, our derivation gives cause to the terms 1 and k, their relationship to one another, and to t in the denominator. First, from our derivation, “1” actually signifies taking t_out amount of time expressed in units of t_out (t_out/t_out=1) and adding it to t_in  amount of time expressed in units of t_out (ie, the total time to make a full pass through the world expressed in terms of how the agent apportions its time under a policy of accepting the considered pursuit).”

Additional Correction. In revising the section, “Hyperbolic Temporal Discounting Functions” in the discussion, we also detected an error in our description of the meaning of suboptimal bias for SS. In the revision, the sentence now reads…

“More precisely, what is meant by this suboptimal bias for SS is that the switch in preference from LL to SS occurs at an outside reward rate that is lower—and/or an outside time that is greater —than what an optimal agent would exhibit.”

(7) Figure 15B should have negative axes defined for the pursuit's now negative reward.

Yes- excellent point.

To remove ambiguity regarding the valence of inside and outside reward magnitudes, we have changed all such figures so that the left hand y-axis is used to signify the outside reward magnitude and sign, and so that the right hand y-axis is used to signify the inside reward magnitude and sign.

With respect to the revision of original 15B, this change now makes clear that the inside reward label and numerics on the right hand side of the graph run from positive (top) to negative (bottom) values so that it can now be understood that the magnitude of the inside reward is negative in this figure (ie, a punishment). The left hand y-axis labeling the outside reward magnitude has numerics that run in the opposite direction, from negative (top) to positive (bottom). In this figure, the outside reward rate is positive whereas the inside reward rate is negative.

(8) When comparing your discounting function to the TIMERR and Heuristic models, it would be useful to include a schematic plot illustrating the different obtainable behaviors from all models rather than just telling the reader the differences.

We hold that the descriptions and references are sufficient to address these comparisons.

(9) I would strongly suggest cleaning up all appendices for notation…

The typographical errors that have been noted in these reviews have all been corrected. We believe the reviewer to be referring here to the manner that we had cross-referenced Equations in the appendices and main text which can lead to confusion between whether an equation number being referenced is in regard to its occurrence in the main text or its occurrence in the appendices.

In the revision, we eliminate numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are numbered sequentially and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

…and replacing some of the small equation manipulations with written text describing the goal of each derivation.

To increase clarity, we have taken the reviewer’s helpful suggestion, adding helper text in the appendices were needed, and have bolded the equations of importance within the Appendices (rather than removing equation manipulations making clear steps of derivation).

(10) I would suggest moving the table in Appendix 11 to the main text where misestimation is referenced.

So moved. This appendix now appears in the main text as table 1 “Definitions of misestimating global reward rate-enabling parameters”.

Reviewer #3 (Public review):

One broad issue with the paper is readability. Admittedly, this is a complicated analysis involving many equations that are important to grasp to follow the analyses that subsequently build on top of previous analyses.

But, what's missing is intuitive interpretations behind some of the terms introduced, especially the apportionment cost without referencing the equations in the definition so the reader gets a sense of how the decision-maker thinks of this time cost in contrast with the opportunity cost of time.

We thank the reviewer for encouraging us to formulate a succinct and intuitive statement as to the nature of apportionment cost. We thank the reviewer for pressing for a succinct and intuitive verbal description.

We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in a new paragraph (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5). We also expanded original figure 5 and its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

The above definition of apportionment cost adds to other stated relationships of apportionment cost found throughout the paper (original lines 434,435,447,450).

Re-analysis of some existing empirical data through the lens of their presented objective functions, especially later when they describe sources of error in behavior.

Our objective was not to fit experimentally observed data, as is commonly the goal of implementation/computational models. Rather, as a theory, our objective is to rationalize the broad, curious, and well-established pattern of temporal decision-making behaviors under a deeper understanding of reward-rate maximization, and from that understanding, identify the nature of the error being committed by whatever learning algorithm and representational architecture is actually being used by humans and animals. In doing so, we make a number of important contributions. By identifying and analyzing reward-rate-maximizing equations, we 1) provide insight into what composes time’s cost and how the temporal structure of the world in which it is embedded (its ‘context’) impacts the value of a pursuit, 2) rationalize a diverse assortment of temporal decision-making behaviors (e.g., Hyperbolic discounting, the Magnitude Effect, the Sign Effect, and the Delay effect), explaining them with no assumed free-fit parameter, and then, by analyzing error in parameters enabling reward-rate maximization, 3) identify the likely source of error and propose the Malapportionment Hypothesis. The Malapportionment Hypothesis identifies the underweighting of a considered pursuit’s “outside”, and not error in pursuit’s reward rates, as the source of error committed by humans and animals. It explains why animals and humans can present as suboptimally ‘impatient’ in Choice, but as optimal in Forgo. At the same time, it concords with numerous and diverse observations in decision making regarding whether to initiate a pursuit. The nature of this error also, then, makes numerous predictions. These insights inform future computational and experimental work by providing strong constraints on the nature of the algorithm and representational architecture used to learn and represent the values of pursuits. Rigorous test of the Malapportionment Hypothesis will require wholly new experiments.

In the revision, we also now emphasize and add predictions of the Malapportionment Hypothesis, augmenting its figure (Figure 21), its legend, and its paragraphs in the discussion.

“We term this reckoning of the source of error committed by animals and humans the Malapportionment Hypothesis, which identifies the underweighting of the time spent outside versus inside a considered pursuit but not the misestimation of pursuit rates, as the source of error committed by animals and humans (Figure 21). This hypothesis therefore captures previously published behavioral observations (Figure 21A) showing that animals can make decisions to take or forgo reward options that optimize reward accumulation (Krebs et al., 1977; Stephens and Krebs, 1986; Blanchard and Hayden, 2014), but make suboptimal decisions when presented with simultaneous and mutually exclusive choices between rewards of different delays (Logue et al., 1985; Blanchard and Hayden, 2015; Carter and Redish, 2016; Kane et al., 2019). The Malapportionment Hypothesis further predicts that apparent discounting functions will present with greater curvature than what a reward-rate-maximizing agent would exhibit (Figure 21B). While experimentally observed temporal discounting would have greater curvature, the Malapportionment Hypothesis also predicts that the Magnitude (Figure 21C) and Sign effect (Figure 21D) would be less pronounced than what a reward-rate-maximizing agent would exhibit, with these effects becoming less pronounced the greater the underweighting. Finally, with regards to the Delay Effect (Figure 21E), the Malapportionment Hypothesis predicts that preference reversal would occur at delays greater than that exhibited by a reward-rate-maximizing agent, with the delay becoming more pronounced the greater the underweighting outside versus inside the considered pursuit by the agent.”

Reviewer #3 Recommendations:

As mentioned above, the readability of this paper should be improved so that the readers can follow the derivations and your analyses better. To this end, careful numbering of equations, following consistent equation numbering formats, and differentiating between appendix referencing and equation numbering would have gone a long way in improving the readability of this paper. Some specific questions are noted below.

To increase clarity, in the revision we eliminated numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are thus numbered sequentially as they appear and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

(1) In general, it is unclear what the default pursuit is. From the schematic on the left (forgo decision), it appears to be the time spent in between reward-giving pursuits. However, this schematic also allows for smaller rewards to be attained during the default pursuit as do subsequent equations that reference a default reward rate. Here is where an example would have really benefited the authors in getting their point across as to what the default pursuit is in practice in the forgo decisions and how the default reward rate could be modulated.

(1) The description of the default pursuit has been modified in section “Forgo and Choice decision topologies” to now read… “After either the conclusion of the pursuit, if accepted, or immediately after rejection, the agent returns to a pursuit by default (the “default” pursuit). This default pursuit effectively can be a waiting period over which reward could be received, and reoccurs until the next pursuit opportunity becomes available.” (2) Additionally, helper text has been added to Ap1 regarding the meaning of time and reward spent in the default pursuit. Finally, (3) new figures concerning n-pursuits occurring at the same (Supplement 1) or different (Supplement 2) frequencies from a default pursuit is now added, providing examples as suggested by the reviewer.

(2) I want to clarify my understanding of the topologies in Figure 1. In the forgo, do they roam in the "gold" pursuit indefinitely before they are faced with the purple pursuit? In general, comparing the 2 topologies, it seems like in the forgo decision, they can roam indefinitely in the gold topology or choose the purple but must return to the gold.

The reviewer’s understanding of the topology is correct. The agent loops across one unit time in the default gold pursuit indefinitely, though the purple pursuit (or any pursuit that might exist in that world) occurs on exit from gold at its frequency per unit time. The default gold pursuit will then itself have an average duration in units of time spent in gold. As the reviewer states, the agent can re-enter into gold from having exited gold, and can enter gold from having exited purple, but cannot re-enter purple from having exited purple; rather, it must enter into the default pursuit.

…Another point here is that this topology is highly simplified (only one considered pursuit). So it may be helpful to either add a schematic for the full topology with multiple pursuits or alternatively, provide the corresponding equations (at least in appendix 1 and 2) for the simplified topology so you can drive home the intuition behind derived expressions in these equations.

We understand the reviewer to be noting that, while, the illustrated example is of the simple topology, the mathematical formulation handles the case of n-number pursuits, and that illustrating a world in which there are a greater number of pursuits, corresponding to original appendices 1&2, would assist readers in understanding the generality of these equations.

An excellent suggestion. We have now n-pursuit world illustrations where each pursuit occurs at the same (Supplemental Figure 1) and at different frequencies (Supplemental Figure 2) to the manuscript, and have added text to assist in understanding the form of the equation and its relationship to unit time in the default pursuit in the main and in the appendices.

(3) In Equation and Appendix 1, there are a few things that are unclear. Particularly, why is the expected time of the default option E(t_default )= 1/(∑_(i=1)^n f_i )? Similarly, why is the E(r_default )= ρ_d/(∑_(i=1)^n f_i )? Looking at the expression for E(r_default ), it implies that across all pursuits 1 through n, the default option is encountered only once. Ultimately, in Equation 1.4, (and Equation 1), the units of the two terms in the numerator don't seem to match. One is a reward rate (ρ_d) and the other is a reward value. This is the most important equation of the paper since the next several equations build upon this. Therefore, the lack of clarity here makes the reader less likely to follow along with the analysis in rigorous detail. Better explanations of the terms and better formatting will help alleviate some of these issues.

The equation is formulated to calculate the average reward received and average time spent per unit time spent in the default pursuit. So, f_i is the encounter rate of pursuit i for one unit of time spent in the default pursuit. Added to the summation in the numerator we have the average reward obtained in the default pursuit per unit time () and in the denominator we have the time spent in the default pursuit per unit time (1).

Text explaining the above equation has been added to Ap 1.

(4) In equation and appendix 2, I'm trying to relate the expressions for t_out and r_out to the definitions "average time spent outside the considered pursuit". If I understand the expression in Equation 2.4 on the right-hand side, the numerator is the total time spent in all of the pursuits in the environment and the denominator refers to the number of times the considered pursuit is encountered. It is unclear as to why this is the average time spent outside the considered pursuit. In my mind, the expression for average time spent outside the considered pursuit would look something like t_out=1+ ∑_(i≠in)〖p_i t_i 〗= 1+ ∑_(i≠in)〖f_i/(∑_(j=1)^n f_j ) * t_i 〗. It is unclear how these expressions are then equivalent.

Regarding the following equation,

f_i is the probability that pursuit i will be encountered during a single unit of time spent in the default pursuit. The numerator of the expression is the average amount of time spent across all pursuits, excepting the considered pursuit, per unit time spent in the default pursuit. Note that the + 1 in the numerator is accounting for the unit of time spent in the default pursuit and is added outside of the sum. Since f_in is the probability that the considered pursuit will be encountered per unit of time spent in the default pursuit, is the average amount of time spent in the default pursuit between encounters of the considered pursuit. By multiplying the average time spent across all outside pursuits per unit of time in the default pursuit by the average amount of time spent in the default pursuit between encounters of the considered pursuit, we get the average amount of time spent outside the considered pursuit per encounter of the considered pursuit. This is calculated as if the pursuit encounters are mutually exclusive within a single unit of time spent within the default pursuit, as this is the case as the length of our unit time (delta t) approaches zero.

The above text explaining the equation has been added to Ap 2.

(5) In Figure 3, one huge advantage of this separation into in-pursuit and out-of-pursuit patches is that the optimal reward rate maximizing rule becomes one that compares ρ_in and ρ_out. This contrasts with an optimal foraging rule which requires comparing to the global reward rate and therefore a circularity in solution. In practice, however, it is unclear how ρ_out will be estimated by the agent.

How, in practice, a human or animal estimates the reward rates―be they the outside and/or global reward rate under a policy of accepting a pursuit―is the crux of the matter. This work identifies equations that would enable a reward-rate maximizing agent to calculate and execute optimal policies and emphasizes that the effective reward rates and weights of pursuits must be accurately appreciated for global reward rate optimization. In so doing, it makes a reckoning of behaviors commonly but erroneously treated as suboptimal. Then, by examining the consequences of misestimation of these enabling parameters, it identifies mis-weighting pursuits as the nature of the error committed by whatever algorithm and representational architecture is being used by humans and animals (the Malapportionment Hypothesis). This curious pattern identified and analyzed in this work thus provides a clue into the nature of the learning algorithm and means of representing the temporal structure of the environment that is used by humans and animals―the subject of future work.

We note, however, that we do discuss existing models that grapple with how, in practice, how a human or animal may estimate the outside reward rate. Of particular importance is the TIMERR model, which estimates the outside reward rate from its past experience, and can make an accounting of many qualitative features widely observed. However, while appealing, it would mix prior ‘in’ and ‘outside’ experiences within that estimate, and so would fail to perform forgo tasks optimally. Something is still amiss, as this work demonstrates.

(6) The apportionment time cost needs to be explained a little bit more intuitively. For instance, it is clear that the opportunity cost of time is the cost of not spending time in the rest of the environment relative to the current pursuit. But given the definition of apportionment cost here in lines 447- 448 "The apportionment cost relates to time's allocation in the world: the time spent within a pursuit type relative to the time spent outside that pursuit type, appearing in the denominator." The reference to the equation (setting aside the confusion regarding which equation) within the definition makes it a bit harder to form an intuitive interpretation of this cost. Please reference the equation being referred to in lines 447-448, and again, an example may help the authors communicate their point much better

We thank the reviewer for pressing on this critical point.

Action: We added the following succinct verbal description of apportionment cost… “Apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration.” This definition appears in a new paragraph (as below) describing apportionment cost in the results section “Time’s cost: opportunity & apportionment costs determine a pursuit’s subjective value”, and is accompanied by equations for apportionment cost, and a figure giving its geometric depiction (Figure 5).

“What, then, is the amount of reward by which the opportunity cost-subtracted reward is scaled down to equal the sv of the pursuit? This amount is the apportionment cost of time. The apportionment cost of time (height of the brown vertical bar, Figure 5F) is the global reward rate after taking into account the opportunity cost (slope of the magenta-gold dashed line in Figure 5F) times the time of the considered pursuit. Equally, the difference between the inside and outside reward rates, times the time of the pursuit, is the apportionment cost when scaled by the pursuit’s weight, i.e., the fraction that the considered pursuit is to the total time to traverse the world (Equation 9, right hand side). From the perspective of decision-making policies, apportionment cost is the difference in reward that can be expected, on average, between a policy of taking versus a policy of not taking the considered pursuit, over a time equal to its duration (Equation 9 center, Figure 5F).

Equation 9. Apportionment Cost.

While this difference is the apportionment cost of time, the opportunity cost of time is the amount that would be expected from a policy of not taking the considered pursuit over a time equal to the considered pursuit’s duration. Together, they sum to Time’s Cost (Figure 5G). Expressing a pursuit’s worth in terms of the global reward rate obtained under a policy of accepting the pursuit type (Figure 5 left column), or from the perspective of the outside reward and time (Figure 5 right column), are equivalent. However, the latter expresses sv in terms that are independent of one another, conveys the constituents giving rise to global reward rate, and provides the added insight that time’s cost comprises an apportionment as well as an opportunity cost.”

(7) The analyses in Figures 6 and 7 give a nice visual representation of how the time costs are distributed as a function of outside reward and time spent. However, without an expression for apportionment cost it is hard to intuitively understand these visualizations. This also relates to the previous point of requiring a more intuitive explanation of apportionment costs in relation to the opportunity cost of time. Based on my quick math, it seems that an expression for apportionment cost would be as follows: (r_in- ρ_out*t_in)*(t_in⁄t_out )/(t_in⁄t_out +1 ). The condition described in Figure 7 seems like the perfect place to compute the value of just apportionment cost when the opportunity cost is zero. It would be helpful to introduce the equation here.

We designed original figure 7, as the reviewer appreciates, to emphasize that time has a cost even when there is no opportunity cost, being due entirely to the apportionment cost of time.

We now provide the mathematical expression of apportionment cost and apportionment scaling in Figure 5, the point in the main text of its first occurrence.

…and have expanded original figure 5, its legend (so as to illustrate the apportionment scaling factor and the apportionment cost), and its accompanying main text, to further illustrate and clarify apportionment cost, and its relationship to opportunity cost, and time’s cost.

(8) The analysis regarding choice decisions is relatively straightforward, pending the concerns for the main equations listed above for the forgo decisions. Legends certainly would have helped me grasp Figures 10-12 better.

We believe the reviewer is referring to missing labels for the Sooner Smaller pursuit, and the Larger Later Pursuit in these figures? We used the same conventions as in Figure 9, but we see now that adding these labels to these figures would be helpful, and add them in the revision.

We have now added to the figures themselves figure legends indicating the Sooner Small Pursuit and the Larger Later Pursuit. We have also added to the main text to emphasize the points made in these figures regarding the impact of opportunity cost and apportionment cost.

(9) The derivation of the temporal discounting function from subjective reward rate is much appreciated as it provides further evidence for potential equivalence between reward rate optimization and hyperbolic discounting, which is known to explain a slew of decision-making behaviors in the economics literature.

We thank and greatly appreciate the reviewer for this recognition.

In response to the reviewer’s comment, we have added text that further relates reward rate optimization to hyperbolic discounting…

(1) We add discussion of how our normative derivation gives explanation to Mazur’s ad hoc addition of 1 + k to Ainslie’s reward/time hyperbolic discounting conception. See new first paragraph under “Hyperbolic Temporal Discounting Functions” for the historical origins of the standard hyperbolic equation (which are decidedly not normatively derived). And then see our discussion (new second paragraph in sections “The apparent discounting function of global….”) of how our normative derivation gives explanation to “1”, “k”, and their relationship to each other.

(2) We add explicit treatment of the Delay Effect in a new “The Delay Effect” section of the results along with a figure, and in its corresponding Discussion section.

Minor comments:

(1) Typo in equation 2, should be t_i in the denominator within the summation, not r_i .

We thank the reviewer for catching this typo, and have corrected it in the revision.

(2) Before equation 6, typo when defining ρ_in= r_in/(t_in.). Should be t_in in the denominator, not r_out.

We thank the reviewer for catching this typo, and have corrected it in the revision.

(3) Please be consistent with equation numbers, placement of equation references, and the reason for placing appendix numbers. This will improve readability immensely.

To increase clarity, in the revision we eliminated numbering of equations in the appendices except where an equation occurs in an appendix that is referenced within the main text. In the main text, important equations are thus numbered sequentially and note the appendix from which they derive. If an equation in an appendix is referenced in the main text, it is noted within the appendix it derives.

(4) Line 505 - "dominants" should be dominates.

Typo fixed as indicated

(5) Figures 10-12: add legends to the figures.

Now so included.

(6) Lines 701-703: please rewrite the equation separately. It is highly unclear what rt is here.

We thank the reviewer for bringing attention to this error. The error arose in converting from Google Sheets to Microsoft Word.

The equation has now been corrected.

Additional citations noted in reply and appearing in Main text

Ainslie, George. 1975. “Specious Reward: A Behavioral Theory of Impulsiveness and Impulse Control.” Psychological Bulletin 59: 257–72.

Frederick, Shane, George Loewenstein, Ted O. Donoghue, and T. E. D. O. Donoghue. 2002. “Time Discounting and Time Preference : A Critical Review.” Journal of Economic Literature 40: 351–401.

Gibbon, John. 1977. “Scalar Expectancy Theory and Weber’s Law in Animal Timing.” Psychological Review 84: 279–325.

Green, Leonard, Nathanael Fristoe, and Joel Myerson. 1994. “Temporal Discounting and Preference Reversals in Choice between Delayed Outcomes.” Psychonomic Bulletin & Review 1: 383–89.

Grüne-Yanoff, Till. 2015. “Models of Temporal Discounting 1937-2000: An Interdisciplinary Exchange between Economics and Psychology.” Science in Context 28 (4): 675–713.

Jimura, Koji, Joel Myerson, Joseph Hilgard, Todd S. Braver, and Leonard Green. 2009. “Are People Really More Patient than Other Animals? Evidence from Human Discounting of Real Liquid Rewards.” Psychonomic Bulletin & Review 16: 1071–75.

Kalenscher, Tobias, and Cyriel M. A. Pennartz. 2008. “Is a Bird in the Hand Worth Two in the Future? The Neuroeconomics of Intertemporal Decision-Making.” Progress in Neurobiology 84 (3): 284–315.

Kirby, Kris N., and R. J. Herrnstein. 1995. “Preference Reversals Due to Myopic Discounting of Delayed Reward.” Psychological Science 6 (2): 83–89.

Mazur, James E. 1987. “An Adjusting Procedure for Studying Delayed Reinforcement.” In The Effect of Delay and of Intervening Events on Reinforcement Value., 55–73. Quantitative Analyses of Behavior, Vol. 5. Hillsdale, NJ, US: Lawrence Erlbaum Associates, Inc.

McNamara, John. 1982. “Optimal Patch Use in a Stochastic Environment.” Theoretical Population Biology 21 (2): 269–88.

Rosati, Alexandra G., Jeffrey R. Stevens, Brian Hare, and Marc D. Hauser. 2007. “The Evolutionary Origins of Human Patience: Temporal Preferences in Chimpanzees, Bonobos, and Human Adults.” Current Biology: CB 17: 1663–68.

Strotz, R. H. 1956. “Myopia and Inconsistency in Dynamic Utility Maximization.” The Review of Economic Studies 23: 165–80.

eLife Assessment

This useful study by Gao et al identifies Hspa2 as a heterogeneous transcript in the early embryo and proposes a plausible mechanism showing interactions with Carm1. The authors propose that variability in HSPA2 levels among blastomeres at the 4-cell stage skews their relative contribution to the embryonic lineage. Given only 4 other heterogeneous transcripts/non-coding RNA have been proposed to act similarly at or before the 4-cell stage, this would be a key addition to our understanding of how the first cell fate decision is made. While this is a solid study, further data are needed to fully support the conclusions.

Reviewer #1 (Public review):

Summary:

The authors investigate the role of HSPA2 during mouse preimplantation development. Knocking down HSPA2 in zygotes, the authors describe lower chances of developing into blastocysts, which show a reduced number of inner cell mass cells. They find that HSPA2 mRNA and protein levels show some heterogeneity among blastomeres at the 4-cell stage and propose that HSPA2 could contribute to skewing their relative contribution to embryonic lineages. To test this, the authors try to reduce HSPA2 expression in one of the 2-cell stage blastomere and propose that it biases their contribution to towards extra-embryonic lineages. To explain this, the authors propose that HSPA2 would interact with CARM1, which controls chromatin accessibility around genes regulating differentiation into embryonic lineage.

Strengths:

(1) The study offers simple and straightforward experiments with large sample sizes.

(2) Unlike most studies in the field, this research often relies on both mRNA and protein levels to analyse gene expression and differentiation.

Weaknesses:

(1) Image and statistical analyses are not well described.

(2) The functionality of the overexpression construct is not fully validated.

(3) Tracking of KD cells in embryos injected at the 2-cell stage with GFP is unclear.

(4) A key rationale of the study relies on measuring small differences in the levels of mRNA and proteins using semi-quantitative methods to compare blastomeres. As such, it is not possible to know whether those subtle differences are biologically meaningful. For example, the lowest HSPA2 level of the embryo with the highest level is much higher than the top cell from the embryo with the lowest level. What does this level mean then? Does this mean that some blastomeres grafted from strong embryos would systematically outcompete all other blastomeres from weaker embryos? That would be very surprising. I think the authors should be more careful and consider the lack of quantitative power of their approach before reaching firm conclusions. Although to be fair, the authors only follow a long trend of studies with the same intrinsic flaw of this approach.

(5) Some of the analyses on immunostaining do not take into account that this technique only allows for semi-quantitative measurements and comparisons.<br /> a) Some of the microscopy images are shown with an incorrect look-up table.<br /> b) Some of the schematics are incorrect and misleading.

Reviewer #2 (Public review):

Summary:

In this study, Gao et al. use RNA-seq to identify Hspa2 as one of the earliest transcripts heterogeneously distributed between blastomeres. Functional studies are performed using siRNA knockdown showing Hspa2 may bias cells toward the ICM lineage via interaction with the known methyltransferase CARM1.

Strengths:

This study tackles an important question regarding the origins of the first cell fate decision in the preimplantation embryo. It provides novelty in its identification of Hspa2 as a heterogeneous transcript in the early embryo and proposes a plausible mechanism showing interactions with Carm1. Multiple approaches are used to validate their functional studies (FISH, WB, development rates, proteomics). Given only 4 other transcripts/RNA have been identified at or before the 4-cell stage (LincGET, CARM1, PRDM14, HMGA1), this would be an important addition to our understanding of how TE vs ICM fate is established.

Weaknesses:

The RNA-seq results leading the authors to focus on Hspa2 are not included in the manuscript. This dataset would serve as an important resource but is neither included nor discussed. Nor is it mentioned whether Hspa2 was identified in prior RNA-seq embryos studies (for example Deng Science 2014).

Furthermore, the authors show that Hspa2 knockdown at the 1-cell stage lowers total Carm1 levels at the 4-cell stage. However, it is unclear how total abundance within the embryo alters lineage specification within blastomeres. The authors go on to propose a plausible mechanism involving Hspa2 and Carm1 interaction, but do not discuss how expression levels may be involved.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The authors investigate the role of HSPA2 during mouse preimplantation development. Knocking down HSPA2 in zygotes, the authors describe lower chances of developing into blastocysts, which show a reduced number of inner cell mass cells. They find that HSPA2 mRNA and protein levels show some heterogeneity among blastomeres at the 4-cell stage and propose that HSPA2 could contribute to skewing their relative contribution to embryonic lineages. To test this, the authors try to reduce HSPA2 expression in one of the 2-cell stage blastomere and propose that it biases their contribution to towards extra-embryonic lineages. To explain this, the authors propose that HSPA2 would interact with CARM1, which controls chromatin accessibility around genes regulating differentiation into embryonic lineage.

Strengths:

(1) The study offers simple and straightforward experiments with large sample sizes.

Thanks for your kind recognition.

(2) Unlike most studies in the field, this research often relies on both mRNA and protein levels to analyses gene expression and differentiation.

Thanks for your kind recognition.

Weaknesses:

(1) Image and statistical analyses are not well described.

Thanks for your advisable comment. We redescribe the image and statistical analyses in our revised version (line 255-257).

(2) The functionality of the overexpression construct is not validated.

Thanks for your kind suggestion. We validate the functionality of the overexpression construct in our revised version (Figure S3).

(3) Tracking of KD cells in embryos injected at the 2-cell stage with GFP is unclear.

Thanks for your kind suggestion. We randomly co-injected green fluorescent protein (Gfp) mRNA as a linage tracer with either Hspa2-siRNA or NC-FAM into one of the 2 -cell, and then monitored embryo development to the blastocyst stage (line 342-344).

(4) A key rationale of the study relies on measuring small differences in the levels of mRNA and proteins using semi-quantitative methods to compare blastomeres. As such, it is not possible to know whether those subtle differences are biologically meaningful. For example, the lowest HSPA2 level of the embryo with the highest level is much higher than the top cell from the embryo with the lowest level. What does this level mean then? Does this mean that some blastomeres grafted from strong embryos would systematically outcompete all other blastomeres from weaker embryos? That would be very surprising. I think the authors should be more careful and consider the lack of quantitative power of their approach before reaching firm conclusions. Although to be fair, the authors only follow a long trend of studies with the same intrinsic flaw of this approach.

Thanks for your advisable comment. Indeed, despite the approach drew on previous research (Zhou Cell 2018), we were clearly aware that this approach can only reflect relative comparisons. This means that the relative difference among the blastomeres from the same embryo were detected and compared. We did not compare the absolute levels of mRNA between different embryos. We also offered simple and straightforward experiments with large sample sizes to confirm this conclusion.

(5) Some of the analyses on immunostaining do not take into account that this technique only allows for semi-quantitative measurements and comparisons.

a) Some of the microscopy images are shown with an incorrect look-up table.

b) Some of the schematics are incorrect and misleading.

Thanks for your advisable comment. We revised microscopy images and schematics in our revised version.

Reviewer #2 (Public review):

Summary:

In this study, Gao et al. use RNA-seq to identify Hspa2 as one of the earliest transcripts heterogeneously distributed between blastomeres. Functional studies are performed using siRNA knockdown showing Hspa2 may bias cells toward the ICM lineage via interaction with the known methyltransferase CARM1.

Strengths:

This study tackles an important question regarding the origins of the first cell fate decision in the preimplantation embryo. It provides novelty in its identification of Hspa2 as a heterogeneous transcript in the early embryo and proposes a plausible mechanism showing interactions with Carm1. Multiple approaches are used to validate their functional studies (FISH, WB, development rates, proteomics). Given only 4 other transcripts/RNA have been identified at or before the 4-cell stage (LincGET, CARM1, PRDM14, HMGA1), this would be an important addition to our understanding of how TE vs ICM fate is established.

Thanks for your kind recognition.

The RNA-seq results leading the authors to focus on Hspa2 are not included in the manuscript. This dataset would serve as an important resource but is neither included nor discussed. Nor is it mentioned whether Hspa2 was identified in prior RNA-seq embryos studies (for example Deng Science 2014).

Thanks for your advisable comment. To identify genes that show a significantly high variability across blastomeres in the same embryo, we regressed out the embryo effect by established a new method, which will be published and uploaded to the database in the future. Thus, the RNA-seq results leading the we focus on Hspa2 are not included in the manuscript.   

In addition, the functional studies are centered on Hspa2 knockdown at the zygote (1-cell) stage, which would largely target maternal transcript. Given the proposed mechanism relies on Hspa2 heterogeneity post-ZGA (late 2-cell stage), the knockdown studies don't necessarily test this and thus don't provide direct support to the authors' conclusions. The relevance of the study would be improved if the authors could show that zygotic knockdown leads to symmetric Hspa2 levels at the late 2-cell and/or 4-cell stage. It may be possible that zygotic knockdown leads to lower global Hspa2 levels, but that asymmetry is still generated at the 4-cell stage.

Thanks for your advisable comment. We showed that the Hspa2 levels at the late 2-cell and 4cell stage after zygotic knockdown in our revised version (Figure S1 G-H, line 450-452).

Furthermore, the authors show that Hspa2 knockdown at the 1-cell stage lowers total Carm1 levels at the 4-cell stage. However, it is unclear how total abundance within the embryo alters lineage specification within blastomeres. The authors go on to propose a plausible mechanism involving Hspa2 and Carm1 interaction, but do not discuss how expression levels may be involved.

Thanks for your advisable comment. Previous research suggests that heterogeneous activity of the methyltransferase CARM1 results in differential methylation of histone H3R26 to modulate establishment of lineage specification (Zernicka-Goetz Cell 2018). Thus, we didn't discuss the total abundance within the embryo alters lineage specification.

Recommendations for the authors:  

Reviewer #1 (Recommendations for the authors):

(1) Major issue with analyses:

Image analysis needs to be much better explained than simply saying that ImageJ was used. Where are cells measured (at their equatorial plane? What is the size of the ROI?)? Ideally, the ROI and/or raw measurements should be provided.

Thanks for your advisable comment. We redescribe the Image analysis in our revised version (line 187-194). 

What are the objective criteria determining whether a cell is counted as GFP positive, CDX2 positive, or OCT4 positive? This is very unclear and key to the interpretation of many experiments.

Thanks for your advisable comment. We think that the cell containing fluorescence signals above background noise were counted positive.

Statistical analyses mention ANOVA in the methods but the student's t-test in the figure legend. Which is which? Most data are heavily normalized, which would unlikely fit the description for Student's t-test analyses.

Thanks for your advisable comment. We redescribe the statistical analyses in our materials and methods (line 253-260).

Figure 5H describes a relative fluorescence intensity with control at 1. The legend describes a normalization to "DNA" (I guess the authors meant DAPI), which is unlikely to give 1. This suggests that additional normalization was done and is not described. Is that the case? Also, since the authors propose that HSPA2 would control Histone modification and chromatin packing, I do not think that using DAPI is an appropriate way of normalizing the fluorescence signal.

Thanks for your advisable comment. We replaced DNA with DAPI in our revised version. Based on previous studies, we adopted DAPI as a normalized fluorescence signal (Zhou Cell 2018, Zernicka-Goetz Cell 2018).

Figure 1E shows data normalized to the lowest level while Figure 1H is normalized to the highest level. A consistent representation would be welcome.

Thanks for your advisable comment. We revised the Figure 1H in our revised version.

Is Figure 1C showing a t-test between correlations?

Yes, Figure 1C shows the t-test between correlation.

(2) Major issue with the interpretation of semi-quantitative methods and measurements:

qPCR, WB, immunostaining are all semi-quantitative methods that require some kind of normalization due to non-linear bias in the way the molecules are picked up. Such normalization makes it difficult to know whether a detectable difference is meaningful biologically speaking i.e. if a difference of 1 CT between blastomeres can be detected after qPCR, is it meaningful? If that were the case, then embryos with lower CT than others (Figure 1D) would not be able to develop into blastocyst, like siRNA injected embryos, or grafting a blastomere with a high CT onto an embryo with low CT would lead to the systematic differentiation of these strong blastomeres into ICM.

Thanks for your advisable comment. The CT values represent the relative mRNA levels of Hspa2 between blastomeres, and the higher CT value represents the lower expression of Hspa2 at mRNA level. Figure 1D shows the Hspa2 mRNA levels between blastomeres. The blastomere with lowlevel expression of the Hspa2 mRNA is not bias an ICM fates.  

The same goes for fluorescence analyses (Figure 1F). Can the authors also provide the measurements for DAPI as they did for HSPA2? I am sure that with enough measurements, DAPI is variable enough to give a statistical difference among blastomeres with questionable biological meaning.

I think the reasoning used here (unfortunately following the reasoning that has been used in a series of studies by other groups) of ranking blastomeres after semi-quantitative measurement is fundamentally flawed.

Thanks for your advisable comment. The DAPI was determined by the maximal area using a custom Python script. Based on previous studies, we adopted DAPI as a normalized fluorescence signal (Zhou Cell 2018). This approach is to normalize embryo-to-embryo variance from the technical reason.

(3) Major issue with overexpression experiment:

While the siRNA experiment is partially validated by qPCR and WB measurements of HSPA2 after KD, the overexpression experiment is not. Do the authors have any evidence that the construct they use is produced into protein and functional? Can the authors check by WB? Can the authors rescue the siRNA with their overexpression?

Thanks for your advisable comment. We verified the overexpression experiment by WB in in our revised version (Figure S3, line 360-361). Considering that siRNA degrades mRNA and prevents the mRNA translation process, we did not co-inject the siRNA with their overexpression.

The lack of effect of HSPA2 overexpression on blastocyst formation is difficult to reconcile with the interpretation from the authors that levels of HSPA2 bias lineages.

Have the authors tried lower concentrations? Have the authors tried FISH on their half-injected 2cell embryos? Of course, if the antibody against HSPA2 would work with immunostaining, that would be ideal.

Thanks for your advisable comment. We chose the concentrations for our study based on previous research (Zernicka-Goetz Cell 2016). To verified Hspa2 was successfully inject into one blastomere at the 2-cell stage, we observed green fluorescence after co-injected GFP mRNA with either siRNA or NC-FAM into one blastomere of the two-cell embryos. Thus, we didn't try FISH on half-injected 2-cell embryos. We tried to perform immunostaining experiments with various HSPA2 antibodies (Proteintech: 12797-1-AP, Abcam: ab108416) and no good results were achieved.

Author response image 1.

(4) Major issue with tracking of injected cells:

It is unclear what counts as a GFP-positive cell. In Figure 3D, most cells appear to have the same level of GFP.

Thanks for your advisable comment. The cell containing green fluorescence signals above background noise were counted GFP-positive in Figure 3D. Most cells seem to have the same level of GFP because they are daughter cells of the blastomeres injected with GFP.

In the images of GFP-expressing cells used to track the control of KD cells shown in Figure 3A, it seems that the control embryos have mostly GFP cells in the ICM. Is that the case, or just a bad example?

Thanks for your advisable comment. The green fluorescent signals in Figure 3A represented OCT4 protein, an ICM marker.

Can the authors do FISH against HSPA2 and visualize their GFP cells to validate the heterogeneous expression in situ?

Thanks for your advisable comment. We have verified the heterogeneous expression of HSPA2 in Figure1.

(5) Issue with fluorescent images:

Many images are shown with inappropriate look-up tables with saturated DAPI, OCT4, CDX2, and FISH. This raises the doubt that analyses were made on saturated images, which would be incorrect.

The LUT of Figure 5H should be adjusted similarly between the control and siRNA.

Thanks for your advisable comment. We revised some images which showed inappropriate lookup tables in our revised version. The LUT of Figure 5H had been adjusted between the control and siRNA. 

(6) Issue with schematics:

Schematics of blastomere isolation grown into blastocyst-like structures are misleading since the final blastocyst-like structure should not have a zona pellucida and should have fewer cells than regular blastocysts.

Thanks for your advisable comment. We revised schematics of blastomere grown into morula in our revised version (Figure 1A and Figure S1A).

The summary schematics in the final figure should not state HSPA2 -/- since experiments in the study did not use KO but KD.

Thanks for your advisable comment. We revised the summary schematics in our revised version.

The blastocysts are the same sizes as the cleavage stage or morula embryos which implies that cells lose volume to the lumen, which is not the case.

Thanks for your advisable comment. We revised the schematics in our revised version.

(7) Issue with data presentation:

In the tables within the figures, the number of decimals given should be the same for the mean and SE (one decimal should be more than enough).

Thanks for your advisable comment. We revised the figure 2H in our revised version.

The comparison of cell number and distribution within embryos (e.g. Figure 2B) would be best represented by a correlation analysis of TE vs ICM cells.

Thanks for your advisable comment. We add the figure of a correlation analysis of TE vs ICM cells in our revised version (Figure 3B).

The docking simulations are described in the main text as "experiments".

Thanks for your advisable comment. We redescribed the docking simulations in our revised version.

(8) Issue with data interpretation:

The reduced number of ICM cells is interpreted as a slowed-down cell cycle. This could also be explained by failed cytokinesis and the generation of binucleated or polyploid cells. Have the authors checked for that? For example, by looking at their DAPI staining. 

Thanks for your advisable comment. Our RNA-seq results revealed that the differentially expressed genes (DEGs) at blastocyst stage with HSPA2 knocking down are closely related to negative regulation of cell cycle, G1/S transition of mitotic cell cycle, mitotic cell cycle phase transition and regulation of mitotic cell cycle phase transition. Additionally, the previous study demonstrated that knockdown of HSPA2 reduced cell proliferation and led to G1/S phase cell cycle arrest (Hu Ann Transl Med 2019). Additionally, the lower cell number in ICM may also associated with failed cytokinesis and the generation of binucleated or polyploid cells. Thus, we guessed that HSPA2 has a role in ICM lineage establishment, although half of the ICM cells were able to survive with HSPA2 deficiency (line 463-472).

It is unclear to me why reduced ICM should lead to fewer blastocysts. Blastocysts should be able to form as long as their TE is fine. In Figure 2G, embryos seem to be cultured in close proximity, which is fine if they are healthy but not if some of the embryos start dying and releasing toxic compounds (e.g. ROS). Have the authors tried removing the dying KD embryos to see if the development of the remaining embryos would improve?

Thanks for your advisable comment. We think HSPA2 may affect blastocyst development by affecting other signaling pathways. And, the GO enriched terms was closely related to blastocyst development (Figure 2E). There was no significant difference in morula formation rate between Hspa2-KD group and NC group, thus the assumption that the toxic compounds released by some of the embryos that lead to downregulation of blastocyst rate may not be correct. Indeed, the rate of blastocyst formation in Hspa2-KD embryos was reduced significantly lower when few embryos was cultured separately. In addition, we discussed the possibility that the lower cell number in ICM may also associated with failed cytokinesis and the generation of binucleated or polyploid cells.

Author response image 2.

Reviewer #2 (Recommendations for the authors):

One of the significant findings in the paper is the discovery portion where Hspa2 is identified as a heterogeneous transcript. To improve the logic and impact of the manuscript, it may benefit from reorganizing some of the figures and text. For example:

(1) The paragraph in the introduction (Lines 56-68) should be moved to the discussion as the Hspa2 reveal should be in section 3.1, not prior to the RNA-seq results presented in Figure 1.

Thanks for your advisable comment. We think it is more logical that HSPA2 needs to be introduced in the introduction.

(2) Add text at the beginning of Section 3.1 to describe the rationale and results for the RNAseq. It would help the readers if the authors clearly stated why they chose the 4-cell stage.

Thanks for your advisable comment. We explain why we chose the 4-cell stage in our revised version (line 272-273).

(3) As this is the first time Hspa2 is identified, consider moving Figure S1C to the main figure to show expression throughout development.

Thanks for your advisable comment. We moved Figure S1C to the main figure in our revised version (line 286-291).

(4) Figure 1C: the correlation between Hspa2 and ICM markers would be strengthened if additional transcripts were used (Oct4, Sox2, Sox21). The graph in 1C would also be more informative if represented as a scatter plot with correlation coefficients (Nanog log2TPM vs Hspa2 log2TPM), rather than bar graphs.

Thanks for your advisable comment. We chose Nanog as the correlation between Hspa2 and Nanog, a ICM markers, was showing the strongest correlation in result. And, the figure 1C shows the stronger positive correlation between Nanog and Hspa2 in gene expression than random gene pairs (n=100, n means the number of random gene pairs). Thus, the figure 1C with bar graphs is easier to understand.

(5) Figure 1D: how were individual blastomeres grouped into B1-4? Individually run and then pooled based on relative expression?

Thanks for your advisable comment. Blastomeres are named B1 to B4 according to increasing Hspa2 concentration in figure 1E.

(6) Figures 1F, 1I, 5H: the DAPI channel appears to be saturated, but is used to normalize fluorescence intensity and may incorrectly account for light scattering within the embryo. Please clarify by adding more details regarding image analysis. Were partial stacks through the nucleus used for analysis, or max projections? Graph axes should be "relative fluorescence intensity."

Thanks for your advisable comment. We added the details of fluorescence images analysis. The graph axes had revised in our revised version.

(7) Line 278: the results in Figure S1C would benefit from more text regarding expression patterns throughout development. The maternal transcript appears to have a sharp downregulation by the early 2-cell stage, and is then upregulated coinciding with ZGA.

Thanks for your advisable comment. We added more describe of the Figure in main text (LINE 285-290).

(8) For the analyses in Figure 2 I-J and 2K-L, were arrested embryos excluded from analysis? This is an important detail as including arrested embryos would significantly bias the RNA-seq results. 

Thanks for your advisable comment. The arrested embryos were excluded in Figure 2 I-J and 2K-L.

(9) Figures 2G-H would be aided by converting the table in 2H to a bar graph and adding development rates for all stages (2-, 4-, 8-, morula, and blast). This would also show when an arrest occurs.

Thanks for your advisable comment. We converted the table in 2H to a bar graph.

(10) Blast rates are represented with too many significant digits (Figures 2H, 4B). They should only be reported to the closest ones given the unit of measure (number of blasts divided by number of zygotes). For instance, a blast rate of 81.63 {plus minus} 2.000 reflects excessive precision that is not measured in the data, it should rather read 82 {plus minus} 2%. This is also true for % cells (Figures 3E, 4H).

Thanks for your advisable comment. Values were rounded down to the one decimal place (rounded down).

(11) The clarity and impact of Figure 3A and 3D would benefit from 2D slices through the ICM. 

Thanks for your advisable comment. In order to get more comprehensive understanding of the 3D structure of blastocyst of Figure 3A and 3D, we did not choose 2D slices.

(12) To improve clarity and logic, separate the 1-cell and 2-cell knockdown experiments in the text and figures:

a) 1-cell knockdown with RNA-seq results (Fig 2A-F).

b) 1-cell knockdown showing less ICM/pluripotency markers in (combine Figures 2G-M and Figures 3A-B; "new Fig 3").

c) 2-cell knockdown tracing lineage (Figures 2D-E; "new Fig 4").

The new Figures 3 and 4 should mirror one another (i.e. for each knockdown experiment, development rates and cell counts should be included). For the 2-cell knockdown (Figures 2 D-E), what were the developmental rates (8-cell, morula, blast)?

Thanks for your advisable comment. However, in order to the overall logical of the article, we do not separate the 1-cell and 2-cell knockdown experiments in the text and figures. And, we added the developmental rates (8-cell, morula, blast) of 2-cell knockdown group in our revised version (Figure S2).

For the overexpression experiment (Figure 4), why were injections performed at the zygote stage versus the 2-cell stage? Given the significant downregulation of maternal transcript demonstrated in Figure S1C, it seems plausible that the injected RNA was also downregulated.

Thanks for your advisable comment. For the overexpression experiment, we first chose to inject Hspa2 mRNA at the zygote stage and found that the overexpression of Hspa2 does not induce blastomere cells to bias an ICM fate. The qRT-PCR results indicated that the expression level of Hspa2 in overexpression group was significantly increased compared with normal group at 4cell and blastocyst stage (Figure 4C, 4D).  In addition, there is no guarantee that an equal amount of Hspa2 mRNA be injected into each blastomere in 2-cell stage. Thus, we did not microinject Hspa2 mRNA into the 2-cell stage.

The 3.5 subheading overstates the results as the Hspa2-Carm1 interaction is not linked to lineage segregation. For example, a more specific subtitle might be, "Hspa2 interacts with Carm1 and alters H3R26me2 levels."

Thanks for your advisable comment. We revised the subtitle in our revised version (line 376).

Figures 5B-C and 5D-E. The qRT-PCR and WB analysis of knockdown blasts shows a correlation between Hspa2 downregulation and Carm1 downregulation. However, if the proposed mechanism is Hspa2 binding to Carm1 to mediate downstream methylation, why would it be expected to alter transcript levels at the 4-cell or blast stage? Please add further details and discussion in the results and discussion sections.

Thanks for your advisable comment. The reason we chose to work at the 4-cell stage is because previous studies on CARM1 have focused on the 4-cell stage (Zernicka-Goetz Cell 2018,2016). 

In the discussion, the statement in Lines 430-431 is an overinterpretation: "the heterogeneity of HSPA2... acts as an upstream factor to drive [the] first cell-fate decision." The knockdown experiments don't alter heterogeneity per se, but total abundance. Furthermore, the results do not show that heterogeneity drives heterogeneity in H3R26me2 patterns, for example.

Thanks for your advisable comment. We redescribe the relevant statement in the discussion.

More needs to be said regarding the ICM cells that persisted in the 1-cell KD experiment (Fig 3B). Lines 449-450 point out this result, but do not propose any plausible explanations. For instance, ICM cells may still form due to the incomplete knockdown achieved or the possibility that redundant pathways exist.

Thanks for your advisable comment. We redescribe the relevant statement in our revised version (line 468-473).

The 5th paragraph of the discussion seems incomplete. The authors point out a possible link between Hspa2 and Hippo and Wnt signaling pathways, but need to expand their discussion on how this may act as an additional mechanism incorporating Hspa2 with lineage segregation.

Thanks for your advisable comment. We redescribe the 5th paragraph of the discussion (line 483-494).

Statistics: all comparisons with greater than 2 groups should be performed with a one-way ANOVA and multiple comparisons, rather than Student's t-test (Figures 1B, 1D, 1E, 1F).

All figure legends lack statistical test details.

Thanks for your advisable comment. All figure legends added statistical test details in statistical analysis.

Minor comments:

In all graphs, individual blastomere expression levels should be represented as boxwhisker/bar/scatter/violin plots since the comparison is groups rather than time points (i.e. symbols should not be connected with a line in Figures 1B, 1D, 1F-G, 1I, S1D, S1F).

Thanks for your advisable comment. Each colored line represents a single cell, and the dots of the same color represent the blastomere of the same cell. Thus, we use a line representation individual blastomere.

For all fluorescent images, having two representative images may be confusing for the reader. Figures may be improved by just including one representative image for each stage/treatment (Figures 1F, 1I, S1F, 3A, 3D, 4E, 4G).

Thanks for your advisable comment. The figures just including one representative image for each stage in our revised version. In addition, two representative images from each group were shown for each treatment (Figures 3A, 3D, 4E, 4G).

The manuscript would be improved with thorough grammar and typo editing.

For example:

(1) Lines 18, 73, the wording is confusing, consider: "knockdown of Hspa2 in one of the two-cell blastomeres biased its progeny towards the trophectoderm lineage.".

(2) Line 23, overstatement. Consider: "we demonstrated that HSPA2 levels correlate with ICMassociated genes and that it interacts with the CARM1.".

(3) Line 25 confusing wording, "via the execution of commitment and differentiation phases.".

(4) Line 37, replace "that" with "of;" replace "cell-fate decisions" with "cell-fate decision".

(5) Line 40: needs space before (CARM1).

(6) Line 43: the wording is confusing, consider "can result in higher expression levels of".

(7) Line 45: wording, consider "Recent [studies have] further suggested".

(8) Line 70: plurality, consider "analyzed gene expression pattern".

(9) Line 73 typo: "prevents its".

(10) Line 76-77 wording, consider "Hspa2 expression patterns can bias cell fate in the mouse embryo".

(11) Line 276: remove "in whole embryos," since MII eggs are not embryos.

(12) Line 617 "There" should be "Three".

(13) Axis label in Fig 3b "Totle" should be "Total".

(14) Lines 417, 419 missing spaces.

(15) Line 448 missing word, "interfering [with] the cell cycle".

(16) Line 462 incorrect word, "[a]polar cells being specified as ICM".

(17) Line 469 incorrect plural, "cell differentiation".

Thanks for your advisable comment. We revised the whole manuscript carefully according to the reviewers' suggestions.

eLife Assessment

This important work addresses the relationship between the transdiagnostic compulsivity dimension and confidence as well as confidence-related behaviours like reminder setting. The relationship between confidence and compulsive disorders has recently received a lot of attention and has been considered to be a key cognitive change. The authors paired an elegant experimental design and pre-registration to give convincing evidence of the relationship between compulsivity, reminder setting, and confidence. In the revised version they thoroughly addressed the reviewer's comments, in particular adding new analyses clarifying how their findings relate to prediction error based learning as well as presenting additional recovery analyses and psychometric curves further strengthening the manuscript.

Author response:

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

Reviewer #1:

(1) To improve the clarity of the work, I suggest a final note to the authors to say more explicitly that objective accuracy has a finer resolution *due to the number of "special circles" per trial* in their task. This task detail got lost in my read of the manuscript, and confused me with respect to the resolution of each accuracy measure.

We agree with the reviewer that this would be a useful clarification and have therefore added the following statement to the Methods section on p. 20:

“It should be noted that the OIP has a slightly finer resolution due to the number of special circles per trial.”

(2) Similarly for clarification, they could point out that their exclusion criteria removes subjects that have lower OIP than their AIP analysis allows (which is good for comparison between OIP and AIP). Thus, it removes the possibility that very poor performing subjects (OIP) are forced to have a higher than actual AIP due to the range).

We agree this would be a useful statement to add and have included the following sentence in the Supplement on p. 8:

“Such a restriction of the threshold parameter was intended to increase the comparability between AIP and OIP, and hence improved the calculation of the reminder bias.”

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

Reviewer #1:

(1) Upon reading their response to the question I had regarding AIP and OIP, a few more questions came up regarding OIP, AIP, how they're calculations differ, and how the latter was computed in R. I hope these help readers to clarify how to interpret these key measures, and the hypotheses that rely upon them.

Regarding fitting, and in relation to power, is16 queries adequate to estimate an AIP using the R's quickpsy? That is, assuming some noise in the choice process, how recoverable is a true indifference points from 16 trials? If there's a parameter recovery analysis (ie generating choice via the fitting parameters, which will have built-in stochasticity, and seeing how well you recover the parameter) of interest would be helpful. It may help to characterize why the present study might differ from prior studies (maybe a power issue here).

The reviewer is absolutely correct that we should have provided more detail when describing our fitting procedure for the psychometric curves. We have now addressed this by adding the following statements to the Methods section and Supplement:

Page 20 in the main manuscript: “Fitting was done using the quickpsy package in R and more detail is given in the Supplement.”

Pages 8 and 9 in the Supplement: 

“Psychometric curve fitting

We used the quickpsy package in R to fit psychometric curves to each participant’s choice data to derive their actual indifference point (AIP), which was operationalised as the threshold parameter when predicting reminder choices from target values. We restricted the possible parameter ranges from 2 to 9 for the threshold parameter and from 1 to 500 for the slope parameter, based on the task’s properties and pilot data. Apart from those parameter ranges, we used only default settings of the quickpsy() function.

Each participant has only 16 trials (2 for each target value) contribute to the curve fitting. To understand the robustness of the AIP based on such limited data, we conducted a parameter recovery analysis. We simulated 16 trials based on each psychometric function and re-ran the curve fitting based on those simulated choices. There was close correspondence between the actual and recovered threshold parameters (or AIPs) with a correlation of r = 0.97, p < 0.001 (see also Figure S1). In contrast, the slope parameter—which was not central to any of our analyses—exhibited greater variability during the initial fitting. This increased uncertainty likely contributed to its poor recovery in the simulation, as evidenced by a near-zero correlation (r = −0.01, p = 0.82).”

(2) Along these lines, it would be helpful for the reader to actually see the individual psychometric curve, now how quickpsy was used (did you fit left and right asymptotes), etc, to understand how that fitting procedure works and how the assumptions of the fitting procedure compare to what can be gleaned through seeing the choice curves plotted.

As stated above, we used default settings of the quickpsy() function and hence assumed symmetric asymptotes at 0 and 1. However, the reviewer mentions “left and right asymptotes”, so maybe this question is about restricting the possible parameter range for the threshold, which we restricted to values from 2 to 9, as described above.

Regarding the individual curves, we have now include the following statement on page 9 in the Supplement: “Figures S2 to S31 show the individual psychometric curves that were estimated for each participant.” Please refer to the Supplement for the added figures.

(3) A more full explanation of quickpsy, its parameters, and how choice curves look might also generate interesting further questions to think about with respect to biases and compulsivity. Two individuals might have similar indifference points, but an asymptote might reflect a bias to always have some percent chance of for example to take the reminders even at the lowest offer available for them.

We agree that this is an interesting focus which we will keep in mind for future studies.

(4) Regarding comparing OIP to AIP: 

For OIP, as far as I can understand, the resolution of it is decreased compared to AIP.  Accuracies for OIP can only be 0/4,1/4,2/4,3/4, or 4/4. Yet, the resolution for AIP is the full range of offers (2 to 9) with respect to the parameter of interest (the indifference point). Could this bias the estimation of OIP (for instance, someone who scored 25% might actually be much closer to either 50 or 0, but we can't tell due to resolution?

As mentioned in response to comment (1), we restricted the parameter range for the thresholds to 2 to 9 to increase comparability. The reviewer is right to point out that the OIP  still has lower resolution than the AIP, which is one of the downsides of having a shortened paradigm (cf. the longer version in Gilbert et al., 2019), which is optimised for online testing, especially if used in combination with additional questionnaires. We have no reason to believe though that this could have led to any bias, especially none that would contribute to the individual differences which are the main focus of our study.

Gilbert, S. J., Bird, A., Carpenter, J. M., Fleming, S. M., Sachdeva, C., & Tsai, P.-C. (2020). Optimal use of reminders: Metacognition, effort, and cognitive offloading. Journal of Experimental Psychology: General, 149(3), 501–517. https://doi.org/10.1037/xge0000652

(5) Additionally, it seems like the upper and lower bounds of OIP (0 and 10) differ from AIP (2 and 9). Could this also introduce bias (for example, if someone terrible performance, the mean would artificially be higher under AIP than OIP because the smallest indifference point is 2 under AIP, but could be 0 under OIP.

See our response to comment (1), we fixed the range to 2 to 9 (which was the range of target values used in our study).

(6) Finally seeing how CIT actually corresponds to accuracy overall (not a relative measure like AIP compared to OIP) I think would also be helpful as this is related to most points noted above.

We included the suggested test as an exploratory analysis on pages 42-43 in the Supplement: “Third, we were interested in how the transdiagnostic phenotypes would correspond to performance. We therefore fitted a model which predicted internal accuracy (that is, unaided task performance on trials where no reminders could be used) from AD, CIT, and the other covariates (age, education and gender). We found that neither AD, β = -0.02, SE = 0.05, t = 0.44, p = 0.658, nor CIT, β = -0.03, SE = 0.05, t = -0.66, p = 0.510, predicted internal accuracy.

The full results can be found in Table S13 as well as in Figure S32.”

Reviewer #1 (Public review):

Summary:

Boldt et al test several possible relationships between trandiagnostically-defined compulsivity and cognitive offloading in a large online sample. To do so, they develop a new and useful cognitive task to jointly estimate biases in confidence and reminder-setting. In doing so, they find that over-confidence is related to less utilization of reminder-setting, which partially mediates the negative relationship between compulsivity and lower reminder-setting. The paper thus establishes that, contrary to the over-use of checking behaviors in patients with OCD, greater levels of transdiagnostically-defined compulsivity predicts less deployment of cognitive offloading. The authors offer speculative reasons as to why (perhaps it's perfectionism in less clinically-severe presentations that lowers the cost of expending memory resources), and sets an agenda to understand the divergence in cognitive between clinical and nonclinical samples. Because only a partial mediation had robust evidence, multiple effects may be at play, whereby compulsivity impacts cognitive offloading via overconfidence and also by other causal pathways.

Strengths:

The study develops an easy-to-implement task to jointly measure confidence and replicates several major findings on confidence and cognitive offloading. The study uses a useful measure of cognitive offloading - the tendency to set reminders to augment accuracy in the presence of experimentally manipulated costs. Moreover, the utilizes multiple measures of presumed biases -- overall tendency to set reminders, the empirically estimated indifference point at which people engage reminders, and a bias measure that compares optimal indifference points to engage reminders relative to the empirically observed indifference points. That the study observes convergenence along all these measures strengthens the inferences made relating compulsivity to the under-use of reminder-setting. Lastly, the study does find evidence for one of several a priori hypotheses and sets a compelling agenda to try to explain why such a finding diverges from an ostensible opposing finding in clinical OCD samples and the over-use of cognitive offloading.

Weaknesses:

Although I think this design and study are very helpful for the field, I felt that a feature of the design might reduce the tasks's sensitivity to measuring dispositional tendencies to engage cognitive offloading. In particular, the design introduces prediction errors, that could induce learning and interfere with natural tendencies to deploy reminder-setting behavior. These PEs comprise whether a given selected strategy will be or not be allowed to be engaged. We know individuals with compulsivity can learn even when instructed not to learn (e.g., Sharp, Dolan and Eldar, 2021, Psychological Medicine), and that more generally, they have trouble with structure knowledge (eg Seow et al; Fradkin et al), and thus might be sensitive to these PEs. Thus, a dispositional tendency to set reminders might be differentially impacted for those with compulsivity after an NPE, where they want to set a reminder, but aren't allowed to. After such an NPE, they may avoid moreso the tendency to set reminders. Those with compulsivity likely have superstitious beliefs about how checking behaviors lead to a resolution of catastrophes, that might in part originate from inferring structure in the presence of noise or from purely irrelevant sources of information for a given decision problem.<br /> It would be good to know if such learning effects exist, if they're modulated by PE (you can imagine PEs are higher if you are more incentivized - e.g., 9 points as opposed to only 3 points - to use reminders, and you are told you cannot use them), and if this learning effect confounds the relationship between compulsivity and reminder-setting.

A more subtle point, I think this study can be more said to be an exploration than a deductive of test of a particular model -> hypothesis -> experiment. Typically, when we test a hypothesis, we contrast it with competing models. Here, the tests were two-sided because multiple models, with mutually exclusive predictions (over-use or under-use of reminders) were tested. Moreover, it's unclear exactly how to make sense of what is called the direct mechanism, which is supported by the partial (as opposed to complete) mediation.

Comments on revisions:

I have the following final comments for your manuscript revisions:

To improve the clarity of the work, I suggest a final note to the authors to say more explicitly that objective accuracy has a finer resolution *due to the number of "special circles" per trial* in their task. This task detail got lost in my read of the manuscript, and confused me with respect to the resolution of each accuracy measure. Similarly for clarification, they could point out that their exclusion criteria removes subjects that have lower OIP than their AIP analysis allows (which is good for comparison between OIP and AIP). Thus, it removes the possibility that very poor performing subjects (OIP) are forced to have a higher than actual AIP due to the range).

eLife Assessment

You et al. present an important study that applied a high-resolution transposon-based barcoding system to show the clonal contribution of hematopoietic stem and progenitor cells during aging, after 5-FU treatment, and upon transplantation. The results are convincing and show that there are different categories of multipotent progenitors that are either active or indolent, and that long-term fates are dominated by clones that either favor differentiation or self-renewal. This study will be of broad interest to stem-cell biologists and could reach an even wider audience with a clearer and more concise presentation and discussion of the results.

Reviewer #1 (Public review):

Summary:

You, Zhang et al. comprehensively characterize the long-term fates of mouse HSCs in the unperturbed setting using transposon-based lineage tracing for up to 2 years post-labeling. Their analyses reveal a complex heterogeneity of long-term fates, dominated by two behaviors: i) long-lived differentiation-biased clones, and ii) self-renewal & platelet-biased clones. They further identify two categories of multipotent progenitor clones, with one group showing a markedly reduced differentiation activity.

Strengths:

You et al. present a very comprehensive and high-resolution characterization of mouse hematopoietic clonal dynamics, with robust replicates, and technical prowess. The manuscript is beautifully written, with in-depth and clear explanations of the logic behind experimental design choices, and very well-thought-out interpretations of results.

Some of the results integrate well with past observations in the field, whereas many of them are quite unique and novel.

This will surely be a highly impactful study in the field of hematopoiesis and stem cell biology.

Weaknesses:

The authors trace hematopoiesis in situ, in a fully unbiased way for almost 2-years. They compare this time course with the last few years of Cre-LoxP-based tracing studies and they make an assumption that most hematopoiesis will be derived from some type of HSC at that point in time. They then use this assumption to support that what is being measured in their model are the long-term fates of HSCs (or at least cells that were HSC at the point of labeling). While this is a generally valid assumption, the short-lived nature of certain populations (myeloid cells, megakaryocytes) means that these cells are being produced in the context of a relatively aged environment by the time of sampling, which might change the properties of the system. In other words, the "steady-state" is always changing. It is important to read and interpret this manuscript with this in consideration.

Reviewer #2 (Public review):

Summary:

The work from You et al. elucidates the clonal contribution of ageing stem and progenitor cells to both native and perturbed hematopoiesis. The authors use a previously published in vivo lineage tracing system (Patel et al., 2022) that relies on the random integration of a transposon element in the mouse genome. They barcode all mouse cells and then look at lineage relationships between HSPC and mature populations after ~90 weeks.

Strengths:

This work offers very interesting insights into the clonal behaviour of HSPC in the native and perturbed setting during ageing. Experiments are well-planned and well-executed. Understanding the clonal output of HSPCs in aged mice in a native setting, after 5-FU treatment, and upon transplantation are important findings for the field.

Weaknesses:

We found appraising the graphs, interpreting the findings, and understanding those findings in the main text very difficult to follow. While we have made some suggestions below, we encourage the authors to think carefully about what the core messages are, and how best to visualise those, both in terms of data viz and in a schematic to summarise the key findings, and to use plain language in the text.

Author response:

We genuinely appreciate the reviewers' interest and recognition of our work. The comments and suggestions on the results presentation and interpretation are well taken. We plan to revise the manuscript based on the reviewers' recommendations in the following aspects.

(1) We fully agree with the reviewer that the aged environment indeed would affect the myeloid and megakaryocyte differentiation behaviors of HSC. As a result, the clonal behaviors of HSCs presented in the current manuscript could be different from how HSCs differentiate in young mice. This point will be discussed in the revised manuscript.

(2) We agree with the reviewer that the manuscript was not as easy to follow as many other papers in experimental hematology, primarily because the analyses presented in the current manuscript were not frequently used in previous studies. To address this, we will try to revise the manuscript using plain language to describe the results and conclusions. We will also provide graphical summary schematics where appropriate to present the findings better. We will further discuss our results in the context of previous findings to better illustrate the novelty of the current work.

(3) We will provide more technical details of our analysis in the revised manuscript for readers to better understand how results are obtained and data analyses are performed in the current manuscript.

eLife Assessment

This study reports important advances in understanding how pyrazinamide, a first-line antibiotic for tuberculosis treatment, is effective in vivo. The experimental design and data provide solid evidence that the production of reactive oxygen species by host cells contributes to how pyrazinamide is more potent in the host than in culture conditions; however, additional experiments and controls would strengthen these conclusions. This work is of interest to the antibiotic drug development field.

Reviewer #1 (Public review):

Summary

Pyrazinamide (PZA) is a key drug in the anti-TB arsenal, yet despite over 50 years of clinical use, its precise mechanism of action remains unclear. This study offers valuable insights into the in vitro potentiating effect of PZA when used with exogenous oxidative agents. The authors suggest that oxidative stress, specifically thiol oxidation, may be a primary driver of PZA/POA's bactericidal activity. Although the work is substantial, conceptually innovative, and timely, the evidence supporting the authors' conclusions requires further investigation with additional controls and experiments to fully validate the proposed mechanism of action. Once revised, this work will undoubtedly be of significant interest to the TB drug discovery community and researchers focusing on mycobacterial diseases.

Strengths

The authors have long-standing experience in the field of PZA mode of action, with several publications that have been highly relevant to the field. They are particularly well aware of the literature, and this is clearly visible in the introduction of the manuscript which is beautifully articulated. The biological question(s) and their hypotheses are also well-formulated in the introduction section.

The understanding of PZA mode of action is a long-lasting question in the TB community, therefore studies reporting well-conducted research that aims at deciphering the underlying mechanism responsible for PZA peculiar activity is always appreciated. Since PZA/POA are poorly active in conventional 7H9 media, but very potent in cellulo or in vivo; looking at host-mediated stress that can eventually lead to an increased vulnerability is extremely relevant. In that context, most of the work has been focused on host-cell endolysosomal pH but very little information is available on other stress. Thus, investigating the contribution of oxidative stress and ROS as specific host environments that might contribute to PZA/POA activity is overall novel and conceptually very interesting.

To address this question, the authors combine multiple approaches including conventional antimicrobial susceptibility profiling, CFU-based counting, and checkerboard assays to report the potentiating effect of PZA pre-treatment on hydrogen peroxide- and diamide-mediated antibacterial action. The use of multiple reference strains including Mtb H37Ra, Mtb H37Rv, M.bovis BCG, and M.bovis BCG::pncA is a great asset of the manuscript, even though they might have been more appropriately used to get further mechanistical insights on the proposed model of action. The findings are reported in 4 major figures that are clear and in an order that appears logical for the understanding of the story.

Weaknesses

Although the manuscript is conceptually very interesting and contains intriguing results, it sometimes fails to fully convince and some additional controls/experiments might help to better back up the authors' claims and really strengthen the study. Indeed, some conclusions seem premature therefore leading to some molecular assumptions regarding a potential mode of action that is not fully supported by the presented data.<br /> The rationale behind some of the experiments is not always clearly explained which makes difficult to follow the authors ideas, the biological hypothesis/model that they test, and therefore the overall scientific story.

The authors conclude their study by proposing a mechanism by which the active form of the drug POA acts in concert with exogenous ROS to promote cellular oxidative damage. This is tested within two models of macrophage infection where they propose that IFN-γ mediated ROS production is essential for PZA activity. Unfortunately, the in cellulo part presents some weaknesses and inconsistencies that the authors need to carefully address.

Finally, the in vitro experiments performed in this manuscript mainly report that PZA pre-treatment increases H2O2-mediated killing or inhibition. There is no direct evidence that clearly shows that oxidative stress drives the potent bactericidal activity of PZA. In these settings, the oxidative stress is always applied after PZA pre-treatment and is therefore likely displaying the major lethal effect.

Reviewer #2 (Public review):

Summary:

The authors tested how ROS and PZA affected Mycobacterium survival to determine if ROS could have a role in the remarkable in vivo efficacy of PZA.

Strengths:

This is a well-written and clear manuscript convincingly demonstrating the synergy between PZA and reactive oxygen species in the inhibition of growth and survival of Mycobacterium tuberculosis.

Weaknesses:

The manuscript would benefit from a clear statement of the rationale for the protocols used to examine the synergy of PZA with ROS, the possible models their protocols could be testing, and then how their data supports or disproves the models being tested. The manuscript appears to propose, as stated in the title, that "Oxidative stress drives potent bactericidal activity of pyrazinamide...". However their experimental design more likely tests the effect of PZA on ROS sensitivity. Indeed, by the last figure, the authors begin the present their data as PZA sensitizing the bacteria to ROS. More clarity on these possible models and the different interpretations of the data should be considered.

Impact:

The data provide important insight to expand our understanding of the in vivo efficacy of PZA in the treatment of tuberculosis.

Author response:

We thank the reviewers for their thoughtful and constructive assessment of our manuscript. We agree that additional clarity on some key points in the manuscript will be valuable additions to this work. Both reviewers expressed a related concern regarding the basis for design and interpretation of our pyrazinamide ROS synergy experiments. 

Reviewer 1:

The in vitro experiments performed in this manuscript mainly report that PZA pre-treatment increases H2O2-mediated killing or inhibition. There is no direct evidence that clearly shows that oxidative stress drives the potent bactericidal activity of PZA. In these settings the oxidative stress is always applied after PZA pre-treatment and is therefore likely displaying the major lethal effect.

Reviewer 2:

The manuscript would benefit from a clear statement of the rationale for the protocols used to examine the synergy of PZA with ROS, the possible models their protocols could be testing, and then how their data supports or disproves the models being tested. The manuscript appears to propose, as stated in the title, that "Oxidative stress drives potent bactericidal activity of pyrazinamide...". However their experimental design more likely tests the effect of PZA on ROS sensitivity. Indeed, by the last figure, the authors begin the present their data as PZA sensitizing the bacteria to ROS. More clarity on these possible models and the different interpretations of the data should be considered.

We agree that the data presented in the current version of the manuscript is incomplete in supporting our assertion that oxidative stress drives bactericidal activity of pyrazinamide. As both reviewers note, pretreatment of bacilli with pyrazinamide followed by challenge with ROS indicates that pyrazinamide enhances susceptibility to oxidative stress but does not address whether oxidative stress enhances susceptibility to pyrazinamide. Further, we neglected to provide information regarding why we chose to pretreat bacilli with pyrazinamide before ROS exposure. Over the course of our work, we had found that pyrazinoic acid, the active form of pyrazinamide, showed potent synergy with hydrogen peroxide.  In contrast to the time-dependent synergy that we observed between pyrazinamide and peroxide, synergy between pyrazinoic acid and peroxide did not require pretreatment. We will revise our manuscript to include results that address these key issues and we will carefully consider revising our interpretations accordingly.

eLife Assessment

This study demonstrated that the conditional knockout of afadin disrupts retinal laminar organization and reduced number of photoreceptors while preserving some of the structure and light responsiveness of retinal ganglion cells. These findings are solid and useful for understanding afadin's role in retinal cell generation, lamination, and functional organization. However, the study provides limited new insights into the relationship between retinal lamination defects and overall retinal function.

Reviewer #1 (Public review):

Summary:

The question of how central nervous system (CNS) lamination defects affect functional integrity is an interesting topic, though it remains a subject of debate. The authors focused on the retina, which is a relatively simple yet well-laminated tissue, to investigate the impact of afadin - a key component of adherens junctions on retinal structure and function. Their findings show that the loss of afadin leads to significant disruptions in outer retinal lamination, affecting the morphology and localization of photoreceptors and their synapses, as illustrated by high-quality images. Despite these severe changes, the study found that some functions of the retinal circuits, such as the ability to process light stimuli, could still be partially preserved. This research offers new insights into the relationship between retinal lamination and neural circuit function, suggesting that altered retinal morphology does not completely eliminate the capacity for visual information processing.

Strengths:

The retina serves as an excellent model for investigating lamination defects and functional integrity due to its relatively simple yet well-organized structure, along with the ease of analyzing visual function. The images depicting outer retinal lamination, as well as the morphology and localization of photoreceptors and their synapses, are clear and well-described. The paper is logically organized, progressing from structural defects to functional analysis. Additionally, the manuscript includes a comprehensive discussion of the findings and their implications.

Weaknesses:

While this work presents a wealth of descriptive data, it lacks quantification, which would help readers fully understand the findings and compare results with those from other studies. Furthermore, the molecular mechanisms underlying the defects caused by afadin deletion were not explored, leaving the role of afadin and its intracellular signaling pathways in retinal cells unclear. Finally, the study relied solely on electrophysiological recordings to demonstrate RGC function, which may not be robust enough to support the conclusions. Incorporating additional experiments, such as visual behavior tests, would strengthen the overall conclusions.

Reviewer #2 (Public review):

Summary:

Ueno et al. described substantial changes in the afadin knockout retina. These changes include decreased numbers of rods and cones, an increased number of bipolar cells, and disrupted somatic and synaptic organization of the outer limiting membrane, outer nuclear layer, and outer plexiform layer. In contrast, the number and organization of amacrine cells and retinal ganglion cells remain relatively intact. They also observed changes in ERG responses and RGC receptive fields and functions using MEA recordings.

Strengths:

The morphological characterization of retinal cell types and laminations is detailed and relatively comprehensive.

Weaknesses:

(1) The major weakness of this study, perhaps, is that its findings are predominantly descriptive and lack any mechanistic explanation. As afadin is key component of adherent junctions, its role in mediating retinal lamination has been reported previously (see PMCID: PMC6284407). Thus, a more detailed dissection of afadin's role in processes, such as progenitor generation, cell migration, or the formation of retinal lamination would provide greater insight into the defects caused by knocking out afadin.

(2) The authors observed striking changes in the numbers of rods, cones, and BCs, but not in ACs or RGCs. The causes of these distinct changes in specific cell classes remain unclear. Detailed characterizations, such as the expression of afadin in early developing retina, tracing cell numbers across various early developmental time points, and staining of apoptotic markers in developing retinal cells, could help to distinguish between defects in cell generation and survival, providing a better understand of the underlying causes of these phenotypes.

(3) Although the total number of ACs or RGCs remains unchanged, their localizations are somewhat altered (Figures 2E and 4E). Again, the cause of the altered somatic localization in ACs and RGCs is unclear.

(4) One conclusion that the authors emphasise is that the function of RGCs remains detectable despite a major disrupted outer plexiform layer. However, the organization of the inner plexiform layer remains largely intact, and the axonal innervation of BCs remains unchanged. This could explain the function integrity of RGCs. In addition, the resolution of detecting RGCs by MEA is low, as they only detected 5 clusters in heterozygous animals. This represents an incomplete clustering of RGC functional types and does not provide a full picture of how functional RGC types are altered in the afadin knockout.

Minor Comments:

(1) Line 56-67: "Overall, these findings provide the first evidence that retinal circuit function can be partially preserved even when there are significant disruptions in retinal lamination and photoreceptor synapses" There is existing evidence showing substantial adaption in retinal function when retinal lamination or photoreceptor synapses are disrupted, such as PMCID: PMC10133175.

(2) Line 114-115: "we focused on afadin, which is a scaffolding protein for nectin and has no ortholog in mice." The term "Ortholog" is misused here, as the mouse has an afadin gene. Should the intended meaning be that afadin has no other isoforms in mouse?

Author response:

Reviewer #1 (Public review):

Summary:

The question of how central nervous system (CNS) lamination defects affect functional integrity is an interesting topic, though it remains a subject of debate. The authors focused on the retina, which is a relatively simple yet well-laminated tissue, to investigate the impact of afadin - a key component of adherens junctions on retinal structure and function. Their findings show that the loss of afadin leads to significant disruptions in outer retinal lamination, affecting the morphology and localization of photoreceptors and their synapses, as illustrated by high-quality images. Despite these severe changes, the study found that some functions of the retinal circuits, such as the ability to process light stimuli, could still be partially preserved. This research offers new insights into the relationship between retinal lamination and neural circuit function, suggesting that altered retinal morphology does not completely eliminate the capacity for visual information processing.

Strengths:

The retina serves as an excellent model for investigating lamination defects and functional integrity due to its relatively simple yet well-organized structure, along with the ease of analyzing visual function. The images depicting outer retinal lamination, as well as the morphology and localization of photoreceptors and their synapses, are clear and well-described. The paper is logically organized, progressing from structural defects to functional analysis. Additionally, the manuscript includes a comprehensive discussion of the findings and their implications.

Weaknesses:

While this work presents a wealth of descriptive data, it lacks quantification, which would help readers fully understand the findings and compare results with those from other studies. Furthermore, the molecular mechanisms underlying the defects caused by afadin deletion were not explored, leaving the role of afadin and its intracellular signaling pathways in retinal cells unclear. Finally, the study relied solely on electrophysiological recordings to demonstrate RGC function, which may not be robust enough to support the conclusions. Incorporating additional experiments, such as visual behavior tests, would strengthen the overall conclusions.

Thank you very much for taking the time and thoughtful and valuable comments. Following your suggestions, we will quantify some of the histological data and explore the mechanisms underlying the defects of lamination and cell fate determination observed in afadin cKO retina. We will also try to examine the vision of afadin cKO mice by visual behavior tests.

Reviewer #2 (Public review):

Summary:

Ueno et al. described substantial changes in the afadin knockout retina. These changes include decreased numbers of rods and cones, an increased number of bipolar cells, and disrupted somatic and synaptic organization of the outer limiting membrane, outer nuclear layer, and outer plexiform layer. In contrast, the number and organization of amacrine cells and retinal ganglion cells remain relatively intact. They also observed changes in ERG responses and RGC receptive fields and functions using MEA recordings.

Strengths:

The morphological characterization of retinal cell types and laminations is detailed and relatively comprehensive.

Weaknesses:

(1) The major weakness of this study, perhaps, is that its findings are predominantly descriptive and lack any mechanistic explanation. As afadin is key component of adherent junctions, its role in mediating retinal lamination has been reported previously (see PMCID: PMC6284407). Thus, a more detailed dissection of afadin's role in processes, such as progenitor generation, cell migration, or the formation of retinal lamination would provide greater insight into the defects caused by knocking out afadin.

Thank you for taking the time and valuable comments. Following your suggestions, we will perform experiments to evaluate mechanisms of retinal lamination and cell fate determination defects observed in the afadin cKO retina. However, we would like to note that the paper cited in the comment (PMCID: PMC6284407) analyzed the function of afadin in the formation of dendrites of direction selective RGCs in the IPL, and that the word "lamination" refers to the layering of RGC dendrites in the IPL. Here, we analyzed the function of afadin in laminar construction of the retina.

(2) The authors observed striking changes in the numbers of rods, cones, and BCs, but not in ACs or RGCs. The causes of these distinct changes in specific cell classes remain unclear. Detailed characterizations, such as the expression of afadin in early developing retina, tracing cell numbers across various early developmental time points, and staining of apoptotic markers in developing retinal cells, could help to distinguish between defects in cell generation and survival, providing a better understand of the underlying causes of these phenotypes.

Following your suggestion, we will perform the experiments to characterize the causes of distinct changes in the afadin cKO retina.

(3) Although the total number of ACs or RGCs remains unchanged, their localizations are somewhat altered (Figures 2E and 4E). Again, the cause of the altered somatic localization in ACs and RGCs is unclear.

To clarify the reviewer’s point, we will analyze the progenitor and those cell positions in the developing stage of the afadin cKO retina.

(4) One conclusion that the authors emphasise is that the function of RGCs remains detectable despite a major disrupted outer plexiform layer. However, the organization of the inner plexiform layer remains largely intact, and the axonal innervation of BCs remains unchanged. This could explain the function integrity of RGCs. In addition, the resolution of detecting RGCs by MEA is low, as they only detected 5 clusters in heterozygous animals. This represents an incomplete clustering of RGC functional types and does not provide a full picture of how functional RGC types are altered in the afadin knockout.

We appreciate the reviewer’s insightful comments. Although our clustering of RGC subtypes in afadin cHet retinas resulted in only five clusters, the key finding of our study is the preservation of RGC receptive fields in afadin cKO retinas, despite severe photoreceptor loss (reduced to about one-third of normal) and disruption of photoreceptor-bipolar cell synapses in the OPL. This suggests that even with crucial damage to the OPL, the primary photoreceptor-bipolar-RGC pathway can still function as long as the IPL remains intact. Moreover, the presence of rod-driven responses in RGCs indicates that the AII amacrine cell-mediated rod pathway may also continue to function. We agree that our functional clustering in afadin cHet retinas was incomplete. However, we guess that the absence of RGCs with fast temporal responses in afadin cKO retinas may not simply due to the loss of specific RGC subtypes but due to disrupted synaptic connections between photoreceptors and fast-responding bipolar cells. Furthermore, the structural abnormalities in retinal lamination in afadin cKO retinas may alter RGC response properties, making strict functional classification less meaningful. We would like to emphasize the finding that disruption of the retinal lamination in afadin cKO retinas leads to the absence of RGCs with fast temporal response properties, rather than focusing solely on the classification of RGC subtypes.

Minor Comments:

(1) Line 56-67: "Overall, these findings provide the first evidence that retinal circuit function can be partially preserved even when there are significant disruptions in retinal lamination and photoreceptor synapses" There is existing evidence showing substantial adaption in retinal function when retinal lamination or photoreceptor synapses are disrupted, such as PMCID: PMC10133175.

Thank you for your comment. The paper you mentioned is crucial for discussing and considering the results of our study. We will refer the paper and mention in Discussion.  

(2) Line 114-115: "we focused on afadin, which is a scaffolding protein for nectin and has no ortholog in mice." The term "Ortholog" is misused here, as the mouse has an afadin gene. Should the intended meaning be that afadin has no other isoforms in mouse?

Thank you for pointing it out. As we misused "Ortholog" as "Paralog", we will revise it.

eLife Assessment

This useful study integrates experimental methods from materials science with psychophysical methods to investigate how frictional stabilities influence tactile surface discrimination. The authors argue that force fluctuations arising from transitions between frictional sliding conditions facilitate the discrimination of surfaces with similar friction coefficients. However, the reliance on friction data obtained from an artificial finger, together with the ambiguous correlative analyses relating these measurements to human psychophysics, renders the findings incomplete.

Reviewer #1 (Public review):

Summary:

In this paper, Derkaloustian et. al look at the important topic of what affects fine touch perception. The observations that there may be some level of correlation with instabilities are intriguing. They attempted to characterize different materials by counting the frequency (occurrence #, not of vibration) of instabilities at various speeds and forces of a PDMS slab pulled lengthwise over the material. They then had humans make the same vertical motion to discriminate between these samples. They correlated the % correct in discrimination with differences in frequency of steady sliding over the design space as well as other traditional parameters such as friction coefficient and roughness. The authors pose an interesting hypothesis and make an interesting observation about the occurrences of instability regimes in different materials while in contact with PDMS, which is interesting for the community to see in the publication. It should be noted that the finger is complex, however, and there are many factors that may be quite oversimplified with the use of the PDMS finger, and the consideration and discounting of other parameters are not fully discussed in the main text or SI. Most importantly, however, the conclusions as stated do not align with the primary summary of the data in Figure 2.

Strengths:

The strength of this paper is in its intriguing hypothesis and important observation that instabilities may contribute to what humans are detecting as differences in these apparently similar samples.

Weaknesses:

The most important weakness is that the findings do not support the statements of findings made in the abstract. Of specific note in this regard is the primary correlation in Figure 2B between SS (steady sliding) and percent correct discrimination. While the statistical test shows significance (and is interesting!), the R-squared value is 0.38, while the R-squared value for the "Friction Coefficient vs. Percent Correct" plot has an R-squared of 0.6 and a p-value of < 0.01 (including Figure 2B). This suggests that the results do not support the claim in the abstract: "We found that participant accuracy in tactile discrimination was most strongly correlated with formations of steady sliding, and response times were negatively correlated with stiction spikes. Conversely, traditional metrics like surface roughness or average friction coefficient did not predict tactile discriminability." This is the most fundamental weakness of this paper.

Along the same lines, other parameters that were considered such as the "Percent Correct vs. Difference in Sp" and "Percent Correct vs. Difference in SFW" were not plotted for consideration in the SI. It would be helpful to compare these results with the other three metrics in order to fully understand the relationships. Other parameters such as stiction magnitude and differences in friction coefficient over the test space could also be important and interesting.

Beyond this fundamental concern, there is a weakness in the representativeness of the PDMS finger, the vertical motion, and the speed of sliding to real human exploration. The real finger has multiple layers with different moduli. In fact, the stratum corneum cells, which are the outer layer at the interface and determine the friction, have a much higher modulus than PDMS. In addition, the slanted position of the finger can cause non-uniform pressures across the finger. Both can contribute to making the PDMS finger have much more stick-slip than a real finger. In fact, if you look at the regime maps, there is very little space that has steady sliding. This does not represent well human exploration of surfaces. We do not tend to use a force and velocity that will cause extensive stick-slip (frequent regions of 100% stick-slip) and, in fact, the speeds used in the study are on the slow side, which also contributes to more stick-slip. At higher speeds and lower forces, all of the materials had steady sliding regions. Further, on these very smooth surfaces, the friction and stiction are more complex and cannot dismiss considerations such as finger material property change with sweat pore occlusion and sweat capillary forces. Also, the vertical motion of both the PDMS finger and the instructed human subjects is not the motion that humans typically use to discriminate between surfaces. Finally, fingerprints may not affect the shape and size of the contact area, but they certainly do affect the dynamic response and detection of vibrations.

This all leads to the critical question, why are friction, normal force, and velocity not measured during the measured human exploration and in a systematic study using the real human finger? The authors posed an extremely interesting hypothesis that humans may alter their speed to feel the instability transition regions. This is something that could be measured with a real finger but is not likely to be correlated accurately enough to match regime boundaries with such a simplified artificial finger.

Reviewer #2 (Public review):

Summary:

In this paper, the authors want to test the hypothesis that frictional instabilities rather than friction are the main drivers for discriminating flat surfaces of different sub-nanometric roughness profiles.

They first produced flat surfaces with 6 different coatings giving them unique and various properties in terms of roughness (picometer scale), contact angles (from hydrophilic to hydrophobic), friction coefficient (as measured against a mock finger), and Hurst exponent.

Then, they used those surfaces in two different experiments. In the first experiment, they used a mock finger (PDMS of 100kPA molded into a fingertip shape) and slid it over the surfaces at different normal forces and speeds. They categorized the sliding behavior as steady sliding, sticking spikes, and slow frictional waves by visual inspection, and show that the surfaces have different behaviors depending on normal force and speed. In a second experiment, participants (10) were asked to discriminate pairs of those surfaces. It is found that each of those pairs could be reliably discriminated by most participants.

Finally, the participant's discrimination performance is correlated with differences in the physical attributes observed against the mock finger. The authors found a positive correlation between participants' performances and differences in the count of steady sliding against the mock finger and a negative correlation between participants' reaction time and differences in the count of stiction spikes against the mock finger. They interpret those correlations as evidence that participants use those differences to discriminate the surfaces.

Strengths:

The created surfaces are very interesting as they are flat at the nanometer scale, yet have different physical attributes and can be reliably discriminated.

Weaknesses:

In my opinion, the data presented in the paper do not support the conclusions. The conclusions are based on a correlation between results obtained on the mock finger and results obtained with human participants but there is no evidence that the human participants' fingertips will behave similarly to the mock finger during the experiment. Figure 3 gives a hint that the 3 sliding behaviors can be observed in a real finger, but does not prove that the human finger will behave as the mock finger, i.e., there is no evidence that the phase maps in Figure 1C are similar for human fingers and across different people that can have very different stiffness and moisture levels.

I believe that the authors collected the contact forces during the psychophysics experiments, so this shortcoming could be solved if the authors use the actual data, and show that the participant responses can be better predicted by the occurrence of frictional instabilities than by the usual metrics on a trial by trial basis, or at least on a subject by subject basis. I.e. Poor performers should show fewer signs of differences in the sliding behaviors than good performers.

The sample size (10) is very small.

Reviewer #3 (Public review):

Strengths:

The paper describes a new perspective on friction perception, with the hypothesis that humans are sensitive to the instabilities of the surface rather than the coefficient of friction. The paper is very well written and with a comprehensive literature survey.

One of the central tools used by the author to characterize the frictional behavior is the frictional instabilities maps. With these maps, it becomes clear that two different surfaces can have both similar and different behavior depending on the normal force and the speed of exploration. It puts forward that friction is a complicated phenomenon, especially for soft materials.

The psychophysics study is centered around an odd-one-out protocol, which has the advantage of avoiding any external reference to what would mean friction or texture for example. The comparisons are made only based on the texture being similar or not.

The results show a significant relationship between the distance between frictional maps and the success rate in discriminating two kinds of surface.

Weaknesses:

The main weakness of the paper comes from the fact that the frictional maps and the extensive psychophysics study are not made at the same time, nor with the same finger. The frictional maps are produced with an artificial finger made out of PDMS which is a poor substitute for the complex tribological properties of skin.

The evidence would have been much stronger if the measurement of the interaction was done during the psychophysical experiment. In addition, because of the protocol, the correlation is based on aggregates rather than on individual interactions.

The authors compensate with a third experiment where they used a 2AFC protocol and an online force measurement. But the results of this third study, fail to convince the relation.

No map of the real finger interaction is shown, bringing doubt to the validity of the frictional map for something as variable as human fingers.

Author response:

eLife Assessment

This useful study integrates experimental methods from materials science with psychophysical methods to investigate how frictional stabilities influence tactile surface discrimination. The authors argue that force fluctuations arising from transitions between frictional sliding conditions facilitate the discrimination of surfaces with similar friction coefficients. However, the reliance on friction data obtained from an artificial finger, together with the ambiguous correlative analyses relating these measurements to human psychophysics, renders the findings incomplete.

Our main goal with this paper was to show that the most common metric, i.e. average friction coefficient—widely used in tactile perception and device design—is fundamentally unsound, and to offer a secondary parameter that is compatible with the fact that human motion is unconstrained, leading to dynamic interfacial mechanics. In contrast with the summary assessment, we also note that the average friction coefficients in our study were not particularly similar, ranging from differences of 0.4 – 1, a typical range seen in most studies. We believe some of the comments originate from a misinterpretation of our statistically significant, but negative correlation between human results and friction coefficients – which leads to the spurious conclusion that nearly identical objects should be very easy to tell apart, thus supporting our central argument for the need of an alternative. We understand the Reviewers wanting to see that we can demonstrate that humans using instabilities in situ. This is seemingly reasonable, but we explain the significant challenges and fundamental unknowns to those experiments. However, we modified our title to reflect our focus on offering an alternative to the average coefficient of friction.

We do not think it was feasible, at this stage, to demonstrate that humans use friction instabilities through direct manipulation and observation in human participants. In short, there are still several fundamental unknowns: (1) a decision-making model would need to be created, but it is unknown if tactile decision making follows other models, (2) it is further unknown what constitutes “tactile evidence”, though at our manuscript’s conclusion, we propose that friction instabilities are better suited for to be tactile evidence than the averaging of friction coefficients from a narrow range of human exploration (3) in the design of samples, from a friction mechanics and materials perspective, it is not at this point, possible to pre-program surfaces a priori to deliver friction instabilities and instead must be experimentally determined – especially when attempting to achieve this in controlled surfaces that do not create other overriding tactile cues, like macroscopic bumps or large differences in surface roughness. (4) Given that the basis for tactile percepts, like which object feels “rougher” or “smoother” is not sufficiently established and we have seen leads to confusion, it is necessary to use a 3-alternative forced choice task which avoids asking objects along a preset perceptual dimension – a challenge recognized by Reviewer 3. However, this would bring in issues of memory in the decision-making model. (5) The prior points are compounded by the fact that, we believe, tactile exploration must be performed in an unconstrained manner, i.e., without an apparatus generating motion onto a stationary finger. Work by Liu et al. (IEEE ToH, 2024) showed that recreating friction obtained during free exploration onto a stationary finger was uninterpretable by the participants, hinting at the importance of efference copies(1). We believe that each of the above-mentioned issues constitutes a significant advance in knowledge and would require discussion and dissemination with the community. Finally, one of our overarching goals is to create a consistent method to characterize surfaces, and given individual variability in human fingers and motion, a machine-based method that can rapidly, consistently, and sufficiently replicate tactile exploration is needed.

Finally, we also justify our use of a mock finger to provide a method to characterize surfaces in tactile studies that other researchers could reasonably recreate, without creating a standard around individual humans, considering the variability in finger shape and motion during exploration. We do not believe this is an “either-or” argument, but rather that standardized methods to characterize surfaces and devices are greatly needed in the field. From these standardized methods, like surface roughness, some tabulated values of friction coefficient, or surface energy, etc., the current metrics to parameterize results are largely incapable of capturing the dynamic changes in forces expected during human tactile exploration.

Our changes to the manuscript (Page 1 & SI Page 1, Title)

“Alternatives to Friction Coefficient: Role of Frictional Instabilities for Fine Touch Perception”

Reviewer 1 (Public review):

Summary:

In this paper, Derkaloustian et. al look at the important topic of what affects fine touch perception. The observations that there may be some level of correlation with instabilities are intriguing. They attempted to characterize different materials by counting the frequency (occurrence #, not of vibration) of instabilities at various speeds and forces of a PDMS slab pulled lengthwise over the material. They then had humans make the same vertical motion to discriminate between these samples. They correlated the % correct in discrimination with differences in frequency of steady sliding over the design space as well as other traditional parameters such as friction coefficient and roughness. The authors pose an interesting hypothesis and make an interesting observation about the occurrences of instability regimes in different materials while in contact with PDMS, which is interesting for the community to see in the publication. It should be noted that the finger is complex, however, and there are many factors that may be quite oversimplified with the use of the PDMS finger, and the consideration and discounting of other parameters are not fully discussed in the main text or SI. Most importantly, however, the conclusions as stated do not align with the primary summary of the data in Figure 2.

Strengths:

The strength of this paper is in its intriguing hypothesis and important observation that instabilities may contribute to what humans are detecting as differences in these apparently similar samples.

We thank Reviewer 1 for their time on the manuscript, recognizing the approach we took, and offering constructive feedback. We believe that our conclusions, in fact, are supported by the primary summary of the data in Figure 2 but we believe that our use of R² could have led to misinterpretation. The trend with friction coefficient and percent correct was indeed statistically significant but was spurious because the slope was negative. In the revision, we add clarifying comments throughout, change from R² to r as to highlight the negative trend, and adjust the figures to better focus on friction coefficient.

Finally, we added a new section to discuss the tradeoffs between using a real human finger versus a mock finger, and which situations may warrant the use of one or the other. In short, for our goal of characterizing surfaces to be used in tactile experiments, we believe a mock finger is more sustainable and practical than using real humans because human fingers are unique per participant, humans move their fingers at constantly changing pressures and velocities, and friction generated during free exploring human cannot be satisfactorily replicated by moving a sample onto a stationary finger. But, we do not disagree that for other types of experiments, characterizing a human participant directly may be more advantageous.

Weaknesses:

Comment 1 - The most important weakness is that the findings do not support the statements of findings made in the abstract. Of specific note in this regard is the primary correlation in Figure 2B between SS (steady sliding) and percent correct discrimination. Of specific note in this regard is the primary correlation in Figure 2B between SS (steady sliding) and percent correct discrimination. While the statistical test shows significance (and is interesting!), the R-squared value is 0.38, while the R-squared value for the "Friction Coefficient vs. Percent Correct" plot has an R-squared of 0.6 and a p-value of < 0.01 (including Figure 2B). This suggests that the results do not support the claim in the abstract: "We found that participant accuracy in tactile discrimination was most strongly correlated with formations of steady sliding, and response times were negatively correlated with stiction spikes. Conversely, traditional metrics like surface roughness or average friction coefficient did not predict tactile discriminability."

We disagree that the trend with friction coefficient suggests the results do not support the claim because the correlation was found to be negative. However, we could have made the comparison more apparent and expanded on this point, given its novelty.

While the R² value corresponding to the “Friction Coefficient vs. Percent Correct” plot is notably higher, our results show that the slope is negative, which would be statistically spurious. This is because a negative correlation between percent correct (accuracy in discriminating surfaces) and difference in friction coefficient means that the more similar two surfaces are (by friction coefficient), the easier it would be for people to tell them apart. That is, it incorrectly concludes that two identical surfaces would be much easier to tell apart than two surfaces with greatly different friction coefficients.

This is counterintuitive to nearly all existing results, but we believe our samples were well-positioned to uncover this trend by minimizing variability, by controlling multiple physical parameters in the samples, and that the friction coefficient — typically calculated in the field as an average friction coefficient — ignores all the dynamic changes in forces present in elastic systems undergoing mesoscale friction, i.e., human touch, as seen in Fig. 1 in a mock finger and Fig. 3 in a real finger. By demonstrating this statistically spurious trend, we believe this strongly supports our premise that an alternative to friction coefficient is needed in the design of tactile psychophysics and haptic interfaces.

We believe that this could have been misinterpreted, so we took several steps to improve clarity, given the importance of this finding: we separated the panel on friction coefficient to its own panel, we changed from R² to r throughout, and we added clarifying text. We also added a small section focusing on this spurious trend.

Our changes to the manuscript (Page 10)

“To compare the value of looking at frictional instabilities, we also performed GLMM fits on common approaches in the field, like a friction coefficient or material property typically used in tactile discrimination, shown in Fig. 2D-E. Interestingly, in Fig. 2D, we observed a spurious, negative correlation between friction coefficient (typically and often problematically simplified as across all tested conditions) and accuracy (r = -0.64, p < 0.01); that is, the more different the surfaces are by friction coefficient, the less people can tell them apart. This spurious correlation would be the opposite of intuition, and further calls into question the common practice of using friction coefficients in touch-related studies. The alternative, two-term model which includes adhesive contact area for friction coefficient(29) was even less predictive (see Fig. S6A of SI). We believe such a correlation could not have been uncovered previously as our samples are minimal in their physical variations. Yet, the dynamic changes in force even within a single sample are not considered, despite being a key feature of mesoscale friction during human touch.

We investigate different material properties in Fig. 2E. Differences in average roughness R_a (or other parameters, like root mean square roughness R_rms (Fig. S6A of SI) did not show a statistically significant correlation to accuracy. Though roughness is a popular parameter, correlating any roughness parameter to human performance here could be moot: the limit of detecting roughness differences has previously been defined as 13 nm on structured surfaces33 and much higher for randomly rough surfaces,(46) all of which are magnitudes larger than the roughness differences between our surfaces. The differences in contact angle hysteresis – as an approximation of the adhesion contributions(47) – do not present any statistically significant effects on performance.”

Comment 2, Part 1

Along the same lines, other parameters that were considered such as the "Percent Correct vs. Difference in Sp" and "Percent Correct vs. Difference in SFW" were not plotted for consideration in the SI. It would be helpful to compare these results with the other three metrics in order to fully understand the relationships.

We have added these plots to the SI. We note that we had checked these relationships and discussed them briefly, but did not include the plot. The plots show that the type of instability was not as helpful as its presence or absence.

Our changes to the manuscript (Page 9)

“Furthermore, a model accounting for slow frictional waves alone specifically shows a significant, negative effect on performance (p < 0.01, Fig. S5 of SI), suggesting that in these samples and task, the type of instability was not as important.”

Added (SI Page 4)

“and no correlation between accuracy and stiction spikes (Fig. S5).”

Comment 2, Part 2

Other parameters such as stiction magnitude and differences in friction coefficient over the test space could also be important and interesting.

We agree these are interesting and have thought about them. We are aware that others, like Gueorguiev et al., have studied stiction magnitudes, and though there was a correlation, the physical differences in surface roughness (glass versus PMMA) investigated made it unclear if these could be generalized further(2). We are unsure how to proceed here with a satisfactory analysis of stiction magnitude, given that stiction spikes are not always generated. In fact, Fig. 1 shows that for many velocities and pressures, they do not form. However, we offer some speculation on why stiction spikes may be overrepresented in the literature because:

(1) They are prone to being created if the finger was loaded for a long time onto a surface prior to movement, thus creating adhesion by contact aging which is unlike active human exploration. We avoid this by discarding the first pull in our measurements, and is a standard practice in mechanical characterization if contact aging needs to be avoided.

(2) The ranges of velocities and pressures explored were small.

(3) In an effort to generate strong tactile stimuli, highly adhesive or rough surfaces are used.

(4) They are visually distinctive on a plot, but we are unaware of any mechanistic reason that mechanoreceptors would be extremely sensitive to this low frequency event over other signals.

In ongoing work, however, we are always cognizant that if stiction spikes are a dominant factor, then a secondary analysis on their magnitude would be important.

We interpret “difference in friction coefficient over the test space” to be, for a single surface, like C4, to find the highest average friction for a condition of single velocity and mass and subtract that from the lowest average friction for a condition of single velocity and mass. We calculated the difference in friction coefficient in the typical manner of the field, by averaging all data collected at all velocities and masses and assigning a single value for all of a surface, like C4. We had performed this, and have the data, but we are wary of overinterpreting secondary and tertiary metrics because they do not have any fundamental basis in traditional tribology, and this value, if used by humans, would suggest that they rapidly explore a large parameter space to find a “maximum” and “minimum” friction. Furthermore, the range in friction across the test space, after averaging, may in fact, be smaller than the range of friction in a single measurement. For example, in Fig. 1B, the friction coefficient can be calculated by dividing the data by the normal force ([applied mass + 6 g finger] × gravity). The friction coefficient in a single run varies widely, as expected.

Fig. 2D shows a GLMM fit between percent correct responses across our pairs and the differences in friction coefficient for each pair, where we see a spurious negative correlation. As we had the data of all average friction coefficients for each condition for a given material, we also looked at the difference in maximum and minimum friction coefficients. For our tested pairs, these differences also lined up on a statistically significant, negative GLMM fit (r = -0.86, p < 0.005). However, the values for a given surface can vary drastically, with an interquartile range of 1.20 to 2.09 on a single surface. We fit participant accuracy to the differences in these IQRs across pairs. This also led to a negative GLMM fit (r = -0.65, p < 0.05). However, we are hesitant to add this to the manuscript for the reasons stated previously.

Comment 3, Part 1

Beyond this fundamental concern, there is a weakness in the representativeness of the PDMS finger, the vertical motion, and the speed of sliding to real human exploration.

Overall, this is a continuous debate that we think offers two solutions. There is always a tradeoff between using a synthetic model of a finger versus a real human finger, and there is a place for both models. That is, while our mock finger will be more successful the closer it is to a human finger, it is not our goal to fully replace a human finger, rather our goal is to provide a method of characterizing surfaces that is indeed relevant on the length scale of human touch.

The usefulness of the mock finger is in isolating the features of each surface that is independent of human variability, i.e., instabilities that form without changing loading conditions between sliding motions or even within one sliding motion. Of course, with this method, we still require confirmation of these features still forming during human exploration, which we show in Fig. 3.

We believe that this method of characterizing surfaces at the mesoscale will ultimately lead to more successful human studies on tactile perception. Currently, and as shown in the paper, characterizing surfaces through traditional techniques, such as a commercial tribometer (friction coefficient, using a steel or hard metal ball), roughness (via atomic force microscopy or some other metrology), surface energy are less predictive. Thus, we believe this mock finger is stronger than the current state-of-the-art characterizing surfaces (we are also aware of a commercial mock finger company, but we were unable to purchase or obtain an evaluation model).

One of the main – and severe – limitations of using a human finger is that all fingers are different, meaning any study focusing on a particular user may not apply to others or be recreated easily by other researchers. We cannot set a standard for replication around a real human finger as that participant may no longer be available, or willing to travel the world as a “standard”. Furthermore, the method in which changes their pressures and velocities is different. We note that this is a challenge unique to touch perception – how an object is touched changes the friction generated, and thus the tactile stimulus generated, whereas a standardized stimulus is more straightforward for light or sound.

However, we do emphasize that we have strongly considered the balance between feasibility and ecological validity in the design of a mock finger. We have a mock finger, with the three components of stiffness of a human finger (more below). Furthermore, we have also successfully used this mock finger in correlations with human psychophysics in previous work, where findings from our mechanical experiments were predictive of human performance(3-6).

Our changes to the manuscript Added (Page 2-3)

“Mock finger as a characterization tool

In this work, we use a mechanical setup with a PDMS mock finger to derive tactile predictors from controlled friction traces alternative to average friction coefficients. While there is a tradeoff in selecting a synthetic finger over a more accurate, real human finger in modeling touch, our aim to design a method of mesoscale surface characterization for more successful studies on tactile perception cannot be fulfilled using one human participant as a standard. We believe that with sufficient replication of surface and bulk properties as well as contact geometry, and controlled friction measurements collected at loading conditions observed during a tactile discrimination task, we can isolate unique frictional features of a set of surfaces that do not arise from human-to-human variability.

The major component of a human finger, by volume, is soft tissue (~56%)(22), resulting in an effective modulus close to 100 kPa(23,24). In order to achieve this same softness, we crosslink PDMS in a 1×1×5 cm mold at a 30:1 elastomer:crosslinker ratio. However, two more features impart increased stiffness in a human finger. Most of this added rigidity is derived from the bone at the fingertip, the distal phalanx(23–25), which we mimic with an acrylic bone within our PDMS network. The stratum corneum, the stiffer, glassier outer layer of skin(26), is replicated with the surface of the mock finger glassified, or further crosslinked, after 8 hours of UV-Ozone treatment(27). This treatment also modifies the surface properties of the native PDMS to align with those of a human finger more closely. It minimizes the viscoelastic tack at the surface, resulting in a comparable non-sticky surface. At least one day after treatment, the finger surface returns to moderate hydrophilicity (~60º), as is typically observed for a real finger(28).

The initial contact area formed before a friction trace is collected is a rectangle of 1×1 cm. While this shape is not entirely representative of a human finger with curves and ridges, human fingers flatten out enough to reduce the effects of curvature with even very light pressures(28–30). This implies that regardless of finger pressure, the contact area is largely load-independent, which is more accurately replicated with a rectangular mock finger. It is still a challenge to control pressure distribution with this planar interface, but non-uniform pressures are also expected during human exploration.

Lastly, we consider fingerprints vs. flat fingers. A key finding of our previous work is that while fingerprints enhanced frictional dynamics at certain conditions, key features were still maintained with a flat finger.7 Furthermore, for some loading conditions, the more amplified signals could also result in more similar friction traces for different surfaces. We have continued to use flat fingers in our mechanical experiments, and have observed good agreement between these friction traces and human experiments(7,8,21,31).”

(Page 3-4, Materials and Methods)

“Mock Finger Preparation

Friction forces across all six surfaces were measured using a custom apparatus with a polydimethylsiloxane (PDMS, Dow Sylgard 184) mock finger that mimics a human finger’s

mechanical properties and contact mechanics while exploring a surface relatively closely(7,8). PDMS and crosslinker were combined in a 30:1 ratio to achieve a stiffness of 100 kPa comparable to a real finger, then degassed in a vacuum desiccator for 30 minutes. We are aware that the manufacturer recommended crosslinking ratio for Sylgard 184 is 10:1 due to potential uncrosslinked liquid residues(32), but further crosslinking concentrated at the surface prevents this. The prepared PDMS was then poured into a 1×1×5 cm mold also containing an acrylic 3D-printed “bone” to attach applied masses on top of the “fingertip” area contacting a surface during friction testing. After crosslinking in the mold at 60ºC for 1 hour, the finger was treated with UV-Ozone for 8 hours out of the mold to minimize viscoelastic tack.

Mechanical Testing

A custom device using our PDMS mock finger was used to collect macroscopic friction force traces replicating human exploration(7,8). After placing a sample surface on a stage, the finger was lowered at a slight angle such that an initial 1×1 cm rectangle of “fingertip” contact area could be established. We considered a broad range of applied masses (M \= 0, 25, 75, and 100 g) added onto the deadweight of the finger (6 g) observed during a tactile discrimination task. The other side of the sensor was connected to a motorized stage (V-508 PIMag Precision Linear Stage, Physikinstrumente) to control both displacement (4 mm across all conditions) and sliding velocity (v \= 5, 10, 25, and 45 mm s^-1). Forces were measured at all 16 combinations of mass and velocity via a 250 g Futek force sensor (k \= 13.9 kN m^-1) threaded to the bone, and recorded at an average sampling rate of 550 Hz with a Keithley 7510 DMM digitized multimeter. Force traces were collected in sets of 4 slides, discarding the first due to contact aging. Because some mass-velocity combinations were near the boundaries of instability phase transitions, not all force traces at these given conditions exhibited similar profiles.

Thus, three sets were collected on fresh spots for each condition to observe enough occurrences of multiple instabilities, at a total of nine traces per combination for each surface.”

Added References (Page 13)

M. Murai, H.-K. Lau, B. P. Pereira and R. W. H. Pho, J. Hand Surg., 1997, 22, 935–941.

A. Abdouni, M. Djaghloul, C. Thieulin, R. Vargiolu, C. Pailler-Mattei and H. Zahouani, R. Soc. Open Sci., DOI:10.1098/rsos.170321.

P.-H. Cornuault, L. Carpentier, M.-A. Bueno, J.-M. Cote and G. Monteil, J. R. Soc. Interface, DOI:10.1098/rsif.2015.0495.

K. Qian, K. Traylor, S. W. Lee, B. Ellis, J. Weiss and D. Kamper, J. Biomech., 2014, 47, 3094– 3099.

Y. Yuan and R. Verma, Colloids Surf. B Biointerfaces, 2006, 48, 6–12.

Y.-J. Fu, H. Qui, K.-S. Liao, S. J. Lue, C.-C. Hu, K.-R. Lee and J.-Y. Lai, Langmuir, 2010, 26, 4392–4399.

Comment 3, Part 2

“The real finger has multiple layers with different moduli. In fact, the stratum corneum cells, which are the outer layer at the interface and determine the friction, have a much higher modulus than PDMS. The real finger has multiple layers with different moduli. In fact, the stratum corneum cells, which are the outer layer at the interface and determine the friction, have a much higher modulus than PDMS.

We have approximated the softness of the finger with 100 kPa crosslinked PDMS, which is close to what has been reported for the bulk of a human fingertip(8,9). However, as mentioned in the Materials and Methods, there are two additional features of the mock finger that impart greater strength. The PDMS surrounds a rigid, acrylic bone comparable to the distal phalanx, which provides an additional layer of higher modulus(10). Additionally, the 8-hour UV-Ozone treatment decreases the viscoelastic tack of the pristine PDMS by glassifying, or further crosslinking the surface of the finger(11), therefore imparting greater stiffness at the surface similar to the contributions of the stratum corneum, along with a similar surface energy(12). This technique is widely used in wearables(13), soft robotics(14), and microfluidics(15) to induce both these material changes. Additionally, the finger is used at least a day after UV-Ozone treatment is completed in order for the surface to return to moderate hydrophilicity, similar to the outermost layer of human skin(16).

Comment 3, Part 3

In addition, the slanted position of the finger can cause non-uniform pressures across the finger. Both can contribute to making the PDMS finger have much more stick-slip than a real finger.

To ensure that there is minimal contribution from the slanted position of the finger, an initial contact area of 1×1 cm is established before sliding and recording friction measurements. As the PDMS finger is a soft object, the portion in contact with a surface flattens and the contact area remains largely unchanged during sliding. Any additional stick-slip after this alignment step is caused by contact aging at the interface, but the first trace we collect is always discarded to only consider stick-slip events caused by surface chemistry. We recognize that it is difficult to completely control the pressure distribution due to the planar interface, but this is also expected when humans freely explore a surface.

Comment 3, Part 4

In fact, if you look at the regime maps, there is very little space that has steady sliding. This does not represent well human exploration of surfaces. We do not tend to use a force and velocity that will cause extensive stick-slip (frequent regions of 100% stick-slip) and, in fact, the speeds used in the study are on the slow side, which also contributes to more stick-slip. At higher speeds and lower forces, all of the materials had steady sliding regions.

We are not aware of published studies that extensively show that humans avoid stickslip regimes. In fact, we are aware familiar with literature where stiction spike formation is suppressed – a recent paper by AliAbbasi, Basdogan et. al. investigates electroadhesion and friction with NaCl solution-infused interfaces, resulting in significantly steadier forces(17). We also directly showed evidence of instability formation that we observed during human exploration in Fig. 3B-C. These dynamic events are common, despite the lack of control of normal forces and sliding velocities. We also note that Reviewer 1, Comment 2, was suggesting that we further explore possible trends from parameterizing the stiction spike.

We note that many studies have often not gone at the velocities and masses required for stiction spikes – even though these masses and velocities would be routinely seen in free exploration – this is usually due to constraints of equipment(18). Sliding events during human free exploration of surfaces can exceed 100 mm/s for rapid touches. However, for the surfaces investigated here, we observe that large regions of stick-slip can emerge at velocities as low as 5 mm/s depending on the applied load. The incidence of steady sliding appears more dependent on the applied mass, with almost no steady sliding observed at or above 75 g. Indeed, the force categorization along our transition zones is the main point of the paper.

Comment 3, Part 5

Further, on these very smooth surfaces, the friction and stiction are more complex and cannot dismiss considerations such as finger material property change with sweat pore occlusion and sweat capillary forces. Also, the vertical motion of both the PDMS finger and the instructed human subjects is not the motion that humans typically use to discriminate between surfaces.

We did not describe the task sufficiently. Humans were only given the instruction to slide their finger along a single axis from top to bottom of a sample, not vertical as in azimuthal to gravity. We have updated our wording in the manuscript to reflect this.

Our changes to the manuscript (Page 4)

“Participants could touch for as long as they wanted, but were asked to only use their dominant index fingers along a single axis to better mimic the conditions for instability formation during mechanical testing with the mock finger.”

(Page 11)

“The participant was then asked to explore each sample simultaneously, and ran over each surface in strokes along a single axis until the participant could decide which of the two had “more friction”.”

Comment 3, Part 6

Finally, fingerprints may not affect the shape and size of the contact area, but they certainly do affect the dynamic response and detection of vibrations.

We are aware of the nuance. Our previous work on the role of fingerprints on friction experienced by a PDMS mock finger showed enhanced signals with the incorporation of ridges on the finger and used a rate-and-state model of a heterogenous, elastic body to find corresponding trends (though there is no existing model of friction that can accurately model experiments on mesoscale friction)(7). The key conclusion was that a flat finger still preserved key dynamic features, and the presence of stronger or more vibrations could result in more similar forces for different surfaces depending on the sliding conditions.

This is also in the context that we are seeking to provide a reasonable and experimentally accessible method to characterize surfaces, which will always be better as we get closer in replicating a true human finger. But our goal here was to replicate the finger sufficiently for use in human studies. We believe the more appropriate metric of success is if the mock finger is more successful than replacing traditional characterization experiments, like friction coefficient, roughness, surface energy, etc.

Comment 4

This all leads to the critical question, why are friction, normal force, and velocity not measured during the measured human exploration and in a systematic study using the real human finger? The authors posed an extremely interesting hypothesis that humans may alter their speed to feel the instability transition regions. This is something that could be measured with a real finger but is not likely to be correlated accurately enough to match regime boundaries with such a simplified artificial finger.

We are excited that our manuscript offers a tractable manner to test the hypothesis that tactile decision-making models use friction instabilities as evidence. However, we lay out the challenges and barriers, and how the scope of this paper will lead us in that direction. We also clarify that our goals are to provide a method to characterize samples to better design tactile interfaces in haptics or in psychophysical experiments and raise awareness that the common methods of sample characterization in touch by an average friction coefficient or roughness is fundamentally unsound.

In short, in our view, to further support our findings on instabilities would require answering:

(1) Which one, or combination of, of the multiple swipes that people make responsible for a tactile decision? (The need for a decision-making model)

(2) Establish what is, or may be, tactile evidence.

(3) Establish tactile decision-making models are similar or different than existing decision-making models.

(4) Test the hypothesis, in these models, that friction instabilities are evidence, and not some other unknown metric. This requires design samples that vary in the amount of evidence generated, but this evidence cannot be controlled directly. Rather, the samples indirectly vary evidence by how likely it is for a human to generate different types of friction instabilities during standard exploration.

(5) Design a task that does not require the use of subjective tactile descriptors, like “which one feels rougher”, which we see cause confusion in participants, which will likely require accounting for memory effects.

We elaborate these points below:

To successfully perform this experiment, we note that freely exploring humans make multiple strokes on a surface. Therefore, we would need to construct a decision-making model. It has not yet been demonstrated whether tactile decision making follows visual decision making, but perhaps to start, we can assume it does. Then, in the design of our decision-making paradigm, we immediately run into the problem: What is tactile evidence?

From Fig. 3C, we already can see that identifying evidence is challenging. Prior to this manuscript, people may have chosen the average force, or the highest force. Or we may choose the average friction force. Then, after deciding on the evidence, we need to find a method to manipulate the evidence, i.e., create samples or a machine that causes high friction, etc. We show that during the course of human touch, due to the dynamic nature of friction, the average can change a large amount and sample design becomes a central barrier to experiments. Others may suggest immobilizing the finger and applying a known force, but given how much friction changes with human exploration, there is no known method to make a machine recreate temporally and spatially varying friction forces during sliding onto a stationary finger. Finally, perhaps most importantly, in addition to mechanical challenges, a study by Liu, Colgate et al. showed that even if they recorded the friction (2D) of a finger exploring a surface and then replicated the same friction forces onto a finger, the participant could not determine which surface the replayed friction force was supposed to represent.1 This supports that the efference copy is important, that the forces in response to expected motion are important to determine friction. Finally, there is no known method to design instabilities a priori. They must be found through experiments. Especially since if we were to introduce, say a bump or a trough, then we bring in confounding variables to how participants tell surfaces apart.

Furthermore, even if we had some consistent method to create tactile “evidence”, the paradigm also deserves some consideration. In our experience, the 3-AFC task we perform is important because the vocabulary for touch has not been established. That is, in 3-AFC, by asking to determine which one sample is unlike the others, we do not have to ask the participant questions like “which one is rougher” or “which one has less friction”. In contrast, 2-AFC, which is better for decision-making models because it does not include memory, requires the asking of a perceptual question like: “which one is rougher?”. In our ongoing work, taking two silane coatings, we found that participants could easily identify which surface is unlike the others above chance in a 3-AFC, but participants, even within their own trials, could not consistently identify one silane as perceptually “rougher” by 2-AFC. To us, this calls into question the validity of tactile descriptors, but is beyond the scope of this manuscript.

This is not our only goal, but in the context of human exploration, in this manuscript here, we believed it was important to identify a mechanical parameter that was consistent with how humans explore surfaces, but was also a parameter that could characterize to some consistent property of a surface – irrespective of whether a human was touching it. We thought that designing human decision-making models and paradigms around the friction coefficient would not be successful.

Given the scope of these challenges, we do not think it would be possible to establish these conceptual sequences in a single manuscript.

Reviewer 2 (Public review):

Summary:

In this paper, the authors want to test the hypothesis that frictional instabilities rather than friction are the main drivers for discriminating flat surfaces of different sub-nanometric roughness profiles.

They first produced flat surfaces with 6 different coatings giving them unique and various properties in terms of roughness (picometer scale), contact angles (from hydrophilic to hydrophobic), friction coefficient (as measured against a mock finger), and Hurst exponent.

Then, they used those surfaces in two different experiments. In the first experiment, they used a mock finger (PDMS of 100kPA molded into a fingertip shape) and slid it over the surfaces at different normal forces and speeds. They categorized the sliding behavior as steady sliding, sticking spikes, and slow frictional waves by visual inspection, and show that the surfaces have different behaviors depending on normal force and speed. In a second experiment, participants (10) were asked to discriminate pairs of those surfaces. It is found that each of those pairs could be reliably discriminated by most participants.

Finally, the participant's discrimination performance is correlated with differences in the physical attributes observed against the mock finger. The authors found a positive correlation between participants' performances and differences in the count of steady sliding against the mock finger and a negative correlation between participants' reaction time and differences in the count of stiction spikes against the mock finger. They interpret those correlations as evidence that participants use those differences to discriminate the surfaces.

Strengths:

The created surfaces are very interesting as they are flat at the nanometer scale, yet have different physical attributes and can be reliably discriminated.”

We thank Reviewer 2 for their notes on our manuscript. The responses below address the reviewer’s comments and recommendations for revised work.

Weaknesses:

Comment 1

In my opinion, the data presented in the paper do not support the conclusions. The conclusions are based on a correlation between results obtained on the mock finger and results obtained with human participants but there is no evidence that the human participants' fingertips will behave similarly to the mock finger during the experiment. Figure 3 gives a hint that the 3 sliding behaviors can be observed in a real finger, but does not prove that the human finger will behave as the mock finger, i.e., there is no evidence that the phase maps in Figure 1C are similar for human fingers and across different people that can have very different stiffness and moisture levels.

The mechanical characterization conducted with the mock finger seeks to extract significant features of friction traces of a set of surfaces to use as predictors of tactile discriminability. The goal is to find a consistent method to characterize surfaces for use in tactile experiments that can be replicated by others and used prior to any human experiments. However, in the overall response and in a response to a similar comment by Reviewer 1, we also explain why we believe experiments on humans to establish this fact is not yet reasonable.

Comment 2

I believe that the authors collected the contact forces during the psychophysics experiments, so this shortcoming could be solved if the authors use the actual data, and show that the participant responses can be better predicted by the occurrence of frictional instabilities than by the usual metrics on a trial by trial basis, or at least on a subject by subject basis. I.e. Poor performers should show fewer signs of differences in the sliding behaviors than good performers.

To fully implement this, a decision-making model is necessary because, as a counter example, a participant could have generated 10 swipes of SFW and 1 swipe of a Sp, but the Sp may have been the most important event for making a tactile decision. This type of scenario is not compatible with the analysis suggested — and similar counterpoints can be made for other types of seemingly straightforward analysis.

While we are interested and actively working on this, the study here is critical to establish types of evidence for a future decision-making model. We know humans change their friction constantly during real exploration, so it is unclear which of these constantly changing values we should input into the decision making model, and the future challenges we anticipate are explained in Comment 1.

Comment 3

The sample size (10) is very small.

We recognize that, with all factors being equal, this sample size is on the smaller end. However, we emphasize the degree of control of samples is far above typical, with minimal variations in sample properties such as surface roughness, and every sample for every trial was pristine. Furthermore, the sample preparation (> 300 individual wafers were used) and cost became a factor. Although not typically appropriate, and thus not included in the manuscript, a post-hoc power analysis for our 100 trials of our pair that was closest to chance, P4, (53%, closest to chance at 33%) showed a power of 98.2%, suggesting that the study was appropriately powered.

Reviewer 2 (Recommendations for the authors):

Comment 1

Differences in SS and Sp (Table 2) are NOT physical or mechanical differences but are obtained by counting differences in the number of occurrences of each sliding behavior. It is rather a weird choice.

We disagree that differences in SS and Sp are not physical or mechanical, as these are well-established phenomena in the soft matter and tribology literature(19-21). These are known as “mechanical instabilities” and generated due to the effects of two physical phenomena: the elasticity of the finger (which is constant in our mechanical testing) and the friction forces present (which change per sample type). The motivation behind using these different shapes is that the instabilities, in some conditions, can be invariant to external factors like velocity. This would be quite advantageous for human exploration because, unlike friction coefficient, which changes with nearly any factor, including velocity and mass, the instabilities being invariant to velocity would mean that we are accurately characterizing a unique identifier of the surface even though velocity may be variable.

This “weird choice” is the central innovation of this paper. This choice was necessary because we demonstrated that the common usage of friction coefficient is fundamentally flawed: we see that friction coefficient suggests that surface which are more different would feel more similar – indeed the most distinctive surfaces would be two surfaces that are identical, which is clearly spurious. One potential explanation for why we were able to see this is effect is because our surfaces have similar (< 0.6 nm variability) roughness, removing potential confounding factors, and this type of low roughness control has not been used in tactile studies to the best of our knowledge.

Comment 2

Figures 2B-C: why are the x-data different than Table 2?

The x-data in Fig. 2B-C are the absolute differences in the number of occurrences measured for a given instability type or material property out of 144 pulls. Modeling the human participant results in our GLMMs required the independent variables to be in this form rather than percentages. We initially chose to list percent differences in Table 2 to highlight the ranges of differences instead of an absolute value, but have added both for clarity.

Our changes to the manuscript (Page 7)

“To determine if humans can detect these three different instabilities, we selected six pairs of surfaces to create a broad range of potential instabilities present across all three types. These are summarized in Table 2, where the first column for each instability is the difference in occurrence of that instability formed between each pair, and the second is the percent difference.”

Comment 3

"We constructed a set of coated surfaces with physical differences which were imperceptible by touch but created different types of instabilities based on how quickly a finger is slid and how hard a human finger is pressed during sliding." Yet, in your experiment, participants could discriminate them, so this is incoherent.

To clarify the point, macroscopic objects can differ in physical shape and in chemical composition. What we meant was that the physical differences, i.e., roughness, were below a limit (Skedung et al.) that participants, without a coating, would not be able to tell these apart(22). Therefore, the reason people could tell our surfaces apart was due to the chemical composition of the surface, and not any differences in roughness or physical effects like film stiffness (due to the molecular-scale thinness of the surface coatings, they are mechanically negligible). However, we concede that at the molecular scale, the traditional macroscopic distinction between physical and chemical is blurred.

We have made minor revisions to the wording in the abstract. We clarify that the surface coatings had physical differences in roughness that were smaller than 0.6 nm, which based purely on roughness, would not be expected to be distinguishable to participants. Therefore, the reason participants can tell these surfaces apart is due to differences in friction generated by chemical composition, and we were able to minimize contributions from physical differences in the sample our study.

Our changes to the manuscript (Page 1, Abstract)

“We constructed a set of coated surfaces with minimal physical differences that by themselves, are not perceptible to people, but instead, due to modification in surface chemistry, the surfaces created different types of instabilities based on how quickly a finger is slid and how hard a human finger is pressed during sliding.”

Reviewer 3 (Public review):

Strengths:  

The paper describes a new perspective on friction perception, with the hypothesis that humans are sensitive to the instabilities of the surface rather than the coefficient of friction. The paper is very well written and with a comprehensive literature survey.

One of the central tools used by the author to characterize the frictional behavior is the frictional instabilities maps. With these maps, it becomes clear that two different surfaces can have both similar and different behavior depending on the normal force and the speed of exploration. It puts forward that friction is a complicated phenomenon, especially for soft materials.

The psychophysics study is centered around an odd-one-out protocol, which has the advantage of avoiding any external reference to what would mean friction or texture for example. The comparisons are made only based on the texture being similar or not.

The results show a significant relationship between the distance between frictional maps and the success rate in discriminating two kinds of surface.”

We thank Reviewer 3 for their notes and interesting discussion points on our manuscript. Below, we address the reviewer’s feedback and comments on related works.

Weaknesses:

Comment 1

The main weakness of the paper comes from the fact that the frictional maps and the extensive psychophysics study are not made at the same time, nor with the same finger. The frictional maps are produced with an artificial finger made out of PDMS which is a poor substitute for the complex tribological properties of skin.

A similar comment was made by Reviewers 1 and 2 and parts are replicated below. We are not claiming that our PDMS fingers are superior to real fingers, but rather, we cannot establish standards in the field by using real human fingers that vary between subjects and researchers. We believe the mock finger we designed is a reasonable mimic of the human finger by matching surface energy, heterogeneous mechanical structure, and the ability to test multiple physiologically relevant pressures and sliding velocities.

We achieve a heterogeneous mechanical structure with the 3 primary components of stiffness of a human finger. The effective modulus of ~100 kPa, from soft tissue,8,9 is obtained with a 30:1 ratio of PDMS to crosslinker. The PDMS also surrounds a rigid, acrylic bone comparable to the distal phalanx, which provides an additional layer of higher modulus.10 Additionally, the 8-hour UV-Ozone treatment decreases the viscoelastic tack of the pristine PDMS by glassifying, or further crosslinking the surface of the finger,11 therefore imparting greater stiffness at the surface similar to the contributions of the stratum corneum, along with a similar surface energy.12 The finger is used at least a day after UV-Ozone treatment is completed in order for the surface to return to moderate hydrophilicity, similar to the outermost layer of human skin.16 We also discuss the shape of the contact formed. To ensure that there is minimal contribution from the slanted position of the finger, an initial contact area of 1×1 cm is established before sliding and recording friction measurements. As the PDMS finger is a soft object, the portion in contact with a surface flattens and the contact area remains largely unchanged during sliding. We recognize that it is difficult to completely control the pressure distribution due to the planar interface, but this variation is also expected when humans freely explore a surface. Finally, we consider flat vs. fingerprinted fingers. Our previous work on the role of fingerprints on friction experienced by a PDMS mock finger showed enhanced signals with the incorporation of ridges on the finger and used a rate-andstate model of a heterogenous, elastic body to find corresponding trends.7 The key conclusion was that a flat finger still preserved key dynamic features, and the presence of stronger or more vibrations could result in more similar forces for different surfaces depending on the sliding conditions. We note that we have subsequently used the controlled mechanical data collected with this flat mock finger in correlations with human psychophysics in previous work, where findings from our mechanical experiments were predictive of human performance.3–6 Ultimately, we see from our prior work and here that, despite the drawbacks of our mock finger, it outperforms other standard characterization technique in providing information about the mesoscale that correlates to tactile perception. We have added these details to the manuscript.

We also note that an intermediate option, replicating real fingers, even in a mold, may also inadvertently limit trends from characterization to a specific finger. One of the main – and severe – limitations of using a human finger is that all fingers are different, meaning any study focusing on a particular user may not apply to others or be recreated easily by other researchers. We cannot set a standard for replication around a real human finger as that participant may no longer be available, or willing to travel the world as a “standard”. Furthermore, the method in which a single person changes their pressures and velocities as they touch a surface is highly variable. We also note that in the Summary Response, we noted that a study by Colgate et al. (IEEE ToH 2024) demonstrated that efference copies may be important, and thus constraining a human finger and replaying the forces recorded during free exploration will not lead to the participant identifying a surface with any consistency. Thus, it is important to allow humans to freely explore surfaces, but creates nearly limitless variability in friction forces.

This is also against the backdrop that we are seeking to provide a method to characterize surfaces, which will be aided as we get closer in replicate a true human finger. Indeed, the more features we replicate, the more successful the mechanical data will be in correlating to tactile distinguishability. But reasonably, our success would be in replacing traditional characterization experiments, not in recreating the forces of an arbitrary human finger.

Our changes to the manuscript Added (Page 2-3)

“Mock finger as a characterization tool

In this work, we use a mechanical setup with a PDMS mock finger to derive tactile predictors from controlled friction traces alternative to average friction coefficients. While there is a tradeoff in selecting a synthetic finger over a more accurate, real human finger in modeling touch, our aim to design a method of mesoscale surface characterization for more successful studies on tactile perception cannot be fulfilled using one human participant as a standard. We believe that with sufficient replication of surface and bulk properties as well as contact geometry, and controlled friction measurements collected at loading conditions observed during a tactile discrimination task, we can isolate unique frictional features of a set of surfaces that do not arise from human-to-human variability.

The major component of a human finger, by volume, is soft tissue (~56%)(22), resulting in an effective modulus close to 100 kPa(23,24). In order to achieve this same softness, we crosslink PDMS in a 1×1×5 cm mold at a 30:1 elastomer:crosslinker ratio. However, two more features impart increased stiffness in a human finger. Most of this added rigidity is derived from the bone at the fingertip, the distal phalanx(23-25), which we mimic with an acrylic bone within our PDMS network. The stratum corneum, the stiffer, glassier outer layer of skin(26), is replicated with the surface of the mock finger glassified, or further crosslinked, after 8 hours of UV-Ozone treatment(27). This treatment also modifies the surface properties of the native PDMS to align with those of a human finger more closely. It minimizes the viscoelastic tack at the surface, resulting in a comparable non-sticky surface. At least one day after treatment, the finger surface returns to moderate hydrophilicity (~60º), as is typically observed for a real finger(28).

The initial contact area formed before a friction trace is collected is a rectangle of 1×1 cm. While this shape is not entirely representative of a human finger with curves and ridges, human fingers flatten out enough to reduce the effects of curvature with even very light pressures(28-30). This implies that regardless of finger pressure, the contact area is largely load-independent, which is more accurately replicated with a rectangular mock finger. It is still a challenge to control pressure distribution with this planar interface, but non-uniform pressures are also expected during human exploration.

Lastly, we consider fingerprints vs. flat fingers. A key finding of our previous work is that while fingerprints enhanced frictional dynamics at certain conditions, key features were still maintained with a flat finger(7). Furthermore, for some loading conditions, the more amplified signals could also result in more similar friction traces for different surfaces. We have continued to use flat fingers in our mechanical experiments, and have observed good agreement between these friction traces and human experiments(7,8,21,31).”

(Page 3-4, Materials and Methods)

“Mock Finger Preparation

Friction forces across all six surfaces were measured using a custom apparatus with a polydimethylsiloxane (PDMS, Dow Sylgard 184) mock finger that mimics a human finger’s

mechanical properties and contact mechanics while exploring a surface relatively closely(7,8). PDMS and crosslinker were combined in a 30:1 ratio to achieve a stiffness of 100 kPa comparable to a real finger, then degassed in a vacuum desiccator for 30 minutes. We are aware that the manufacturer recommended crosslinking ratio for Sylgard 184 is 10:1 due to potential uncrosslinked liquid residues(32), but further crosslinking concentrated at the surface prevents this. The prepared PDMS was then poured into a 1×1×5 cm mold also containing an acrylic 3D-printed “bone” to attach applied masses on top of the “fingertip” area contacting a surface during friction testing. After crosslinking in the mold at 60ºC for 1 hour, the finger was treated with UV-Ozone for 8 hours out of the mold to minimize viscoelastic tack.  

Mechanical Testing

A custom device using our PDMS mock finger was used to collect macroscopic friction force traces replicating human exploration(7,8). After placing a sample surface on a stage, the finger was lowered at a slight angle such that an initial 1×1 cm rectangle of “fingertip” contact area could be established. We considered a broad range of applied masses (M \= 0, 25, 75, and 100 g) added onto the deadweight of the finger (6 g) observed during a tactile discrimination task. The other side of the sensor was connected to a motorized stage (V-508 PIMag Precision Linear Stage, Physikinstrumente) to control both displacement (4 mm across all conditions) and sliding velocity (v \= 5, 10, 25, and 45 mm s^-1). Forces were measured at all 16 combinations of mass and velocity via a 250 g Futek force sensor (k \= 13.9 kN m^-1) threaded to the bone, and recorded at an average sampling rate of 550 Hz with a Keithley 7510 DMM digitized multimeter. Force traces were collected in sets of 4 slides, discarding the first due to contact aging. Because some mass-velocity combinations were near the boundaries of instability phase transitions, not all force traces at these given conditions exhibited similar profiles. Thus, three sets were collected on fresh spots for each condition to observe enough occurrences of multiple instabilities, at a total of nine traces per combination for each surface.”

Added References (Page 13)

M. Murai, H.-K. Lau, B. P. Pereira and R. W. H. Pho, J. Hand Surg., 1997, 22, 935–941.

A. Abdouni, M. Djaghloul, C. Thieulin, R. Vargiolu, C. Pailler-Mattei and H. Zahouani, R. Soc. Open Sci., DOI:10.1098/rsos.170321.

P.-H. Cornuault, L. Carpentier, M.-A. Bueno, J.-M. Cote and G. Monteil, J. R. Soc. Interface, DOI:10.1098/rsif.2015.0495.

K. Qian, K. Traylor, S. W. Lee, B. Ellis, J. Weiss and D. Kamper, J. Biomech., 2014, 47, 3094– 3099.

Y. Yuan and R. Verma, Colloids Surf. B Biointerfaces, 2006, 48, 6–12.

Y.-J. Fu, H. Qui, K.-S. Liao, S. J. Lue, C.-C. Hu, K.-R. Lee and J.-Y. Lai, Langmuir, 2010, 26, 4392–4399.

Comment 2

The evidence would have been much stronger if the measurement of the interaction was done during the psychophysical experiment. In addition, because of the protocol, the correlation is based on aggregates rather than on individual interactions.

Our Response: We agree that this would have helped further establish our argument, but in the overall statement and in other reviewer responses, we describe the significant challenges to establishing this.

To fully implement this, a decision-making model is necessary because, as a counter example, a participant could have generated 10 swipes of SFW and 1 swipe of a Sp, but the Sp may have been the most important event for making a tactile decision. We also clarify that our goals are to provide a method to characterize samples to better design tactile interfaces in haptics or in psychophysical experiments.

In short, in our view, to develop a decision-making model, the challenges are as follows:

(1) Which one, or combination of, of the multiple swipes that people make responsible for a tactile decision?

(2) Establish what is, or may be, tactile evidence.

(3) Establish tactile decision-making models are similar or different than existing decision-making models.

(4) Test the hypothesis, in these models, that friction instabilities are evidence, and not some other unknown metric.

(5) Design a task that does not require the use of subjective tactile descriptors, like “which one feels rougher”, which we see cause confusion in participants, which will likely require accounting for memory effects.

(6) Design samples that vary in the amount of evidence generated, but this evidence cannot be controlled directly. Rather, the samples indirectly vary evidence by how likely it is for a human to generate different types of friction instabilities during standard exploration.

We elaborate these points below:

To successfully perform this experiment, we note that freely exploring humans make multiple strokes on a surface. Therefore, we would need to construct a decision-making model. It has not yet been demonstrated whether tactile decision making follows visual decision making, but perhaps to start, we can assume it does. Then, in the design of our decision-making paradigm, we immediately run into the problem: What is tactile evidence?

From Fig. 3C, we already can see that identifying evidence is challenging. Prior to this manuscript, people may have chosen the average force, or the highest force. Or we may choose the average friction force. Then, after deciding on the evidence, we need to find a method to manipulate the evidence, i.e., create samples or a machine that causes high friction, etc. We show that during the course of human touch, due to the dynamic nature of friction, the average can change a large amount and sample design becomes a central barrier to experiments. Others may suggest to immobilize the finger and applying a known force, but given how much friction changes with human exploration, there is no known method to make a machine recreate temporally and spatially varying friction forces during sliding onto a stationary finger. Finally, perhaps most importantly, in addition to mechanical challenges, a study by Liu, Colgate et al. showed that even if they recorded the friction (2D) of a finger exploring a surface and then replicated the same friction forces onto a finger, the participant could not determine which surface the replayed friction force was supposed to represent.1 This supports that the efference copy is important, that the forces in response to expected motion are important to determine friction. Finally, there is no known method to design instabilities a priori. They must be found through experiments, especially since if we were to introduce, say a bump or a trough, then we bring in confounding variables to how participants tell surfaces apart.

Furthermore, even if we had some consistent method to create tactile “evidence”, the paradigm also deserves some consideration. In our experience, the 3-AFC task we perform is important because the vocabulary for touch has not been established. That is, in 3-AFC, by asking to determine which one sample is unlike the others, we do not have to ask the participant questions like “which one is rougher” or “which one has less friction”. In contrast, 2-AFC, which is better for decision-making models because it does not include memory, requires the asking of a perceptual question like: “which one is rougher?”. In our ongoing work, taking two silane coatings, we found that participants could easily identify which surface is unlike the others above chance in a 3-AFC, but participants, even within their own trials, could not consistently identify one silane as perceptually “rougher” by 2-AFC. To us, this calls into question the validity of tactile descriptors, but is beyond the scope of the current manuscript.

This is not our only goal, but in the context of human exploration, in this manuscript here, we believed it was important to identify a mechanical parameter that was consistent with how humans explore surfaces, but was also a parameter that could characterize to some consistent property of a surface – irrespective of whether a human was touching it. We thought that designing human decision-making models and paradigms around the friction coefficient would not be successful.

Given the scope of these challenges, we do not think it would be possible to establish this conceptual sequence in a single manuscript.

Comment 3

The authors compensate with a third experiment where they used a 2AFC protocol and an online force measurement. But the results of this third study, fail to convince the relation.

With this experiment, our central goal was to demonstrate that the instabilities we have identified with the PDMS finger also occur with a human finger. Several instances of SS, Sp, and SFW were recorded with this setup as a participant touched surfaces in real time.

Comment 4

No map of the real finger interaction is shown, bringing doubt to the validity of the frictional map for something as variable as human fingers.

Real fingers change constantly during exploration, and friction is state-dependent, meaning that the friction will depend on how the person was moving the moment prior. Therefore, a map is only valid for a single human movement – even if participants all were instructed to take a single swipe and start from zero motion, humans are unable to maintain constant velocities and pressures. Clearly, this is not sustainable for any analysis, and these drawbacks apply to any measured parameter, whether instabilities suggested here, or friction coefficients used throughout. We believe the difficulty of this approach emphasizes why a standard map of characterization of a surface by a mock finger, even with its drawbacks, is a viable path forward.

Reviewer 3 (Recommendations for the authors):

Comment 1

It would be interesting to comment on a potential connection between the frictional instability maps and Schalamack waves

Schallamach waves are a subset of slow frictional waves (SFW). Schallmach waves are very specifically defined. They are a are pockets of air that form between a soft sliding object and rigid surface, and propagate rear-to-front (retrograde waves) as a soft object is slid and buckles due to adhesive pinning. Wrinkles form at the detached portion of the soft material, until the interface reattaches and the process repeats.23 There is typically a high burden of proof to establish a Schallamach wave over a more general slow frictional wave. We note that it would be exceeding difficult to design samples that can reliably create subsets of SFW, but we are aware that this may be an interesting question at a future point in our work.

Comment 2

The force sensors look very compliant, and given the dynamic nature of the signal, it is important to characterize the frequency response of the system to make sure that the fluctuations are not amplified.

Our Response: Thank you for noticing. We mistyped the sensor spring constant as 13.9 N m^-1 instead of kN m^-1. However, below we show how the instabilities are derived from the mechanics at the interface due to the compliance of the finger. The “springs” of the force sensor and PDMS finger are connected in parallel. Since k_sensor = 13.9 kN m^-1, the spring constant of the system overall reflects the compliance of the finger, and highlights the oscillations arising solely from stick-slip. A sample calculation is shown below.

Author response image 1.

Fitting a line to the initial slope of the force trace for C6 gives the equation y = 25.679_x_ – 0.2149. The slope here represents force data over time data, and is divided by the velocity (25 mm/s) to determine 𝐹𝐹 the spring constant of the system . This value is lower than ksensor = 13.9 kN/m, indicating that the “springs” representing the force sensor and PDMS finger are connected in parallel: . The finger is the compliant component of the system, with k_finger = 0.902 N/m, and of course, real human fingers are also compliant so this matches our goals with the design of the mock finger.

Our changes to the manuscript (Page 4)

(k \= 13.9 kN m^-1)

Comment 3

The authors should discuss about the stochastic nature of friction:

Wiertlewski, Hudin, Hayward, IEEE WHC 2011

Greenspon, McLellan, Lieber, Bensmaia, JRSI 2020”

We believe that, given the references, this comment on “stochastic” refers to the macroscopically-observable fluctuations (i.e., the mechanical “noise” which is not due to instrument noise) in friction arising from the discordant network of stick-slip phenomena occurring throughout the contact zone, and not the stochastic nature of nanoscale friction that occurs thermal fluctuations nor due to statistical distributions in bond breaking associated with soft contact.

We first note that our small-scale fluctuations do not arise from a periodic surface texture that dominates in the frequency regime. However, even on our comparatively smooth surfaces, we do expect fluctuations due to nanoscale variation in contact, generation of stick-slip across at microscale length scales that occur either concurrently or discordantly across the contact zone, and the nonlinear dependence of friction to nearly any variation in state and composition(7).

Perhaps the most relevant to the manuscript is that a major advantage of analysis by friction is that it sidesteps these ever-present microscale fluctuations, leading to more clearly defined classifiers or categories during analysis. Wiertlewski et. al. showed repeated measurements in their systems ultimately gave rise to consistent frequencies(24) (we think their system was in a steady sliding regime and the patterning gave rise to underlying macroscopic waves). These consistent frequencies, at least in soft systems and absent obvious macroscopic patterned features, would be expected to arise from the instability categories and we see them throughout.

Comment 4

It is stated that "we observed a spurious, negative correlation between friction coefficient and accuracy”.

What makes you qualify that correlation as spurious?

We mean this as in the statistical definition of “spurious”.

This correlation would indicate that by the metric of friction coefficient, more different surfaces are perceived more similarly. Thus, two very different surfaces, like Teflon and sandpaper, by friction coefficient would be expected to feel very similar. Two nearly identical surfaces would be expected to feel very different – but of course, humans cannot consistently distinguish two identical surfaces. This finding is counterintuitive and refutes that friction coefficient is a reliable classifier of surfaces by touch. We do not think it is productive to determine a mechanism for a spurious correlation, but perhaps one reason we were able to observe this is because our study, to the best of our knowledge, is unique for having samples that are controlled in their physical differences in roughness and surface features.

Our changes to the manuscript (Page 10)

“To compare the value of looking at frictional instabilities, we also performed GLMM fits on common approaches in the field, like a friction coefficient or material property typically used in tactile discrimination, shown in Fig. 2D-E. Interestingly, in Fig. 2D, we observed a spurious, negative correlation between friction coefficient (typically and often problematically simplified as across all tested conditions) and accuracy (r = -0.64, p < 0.01); that is, the more different the surfaces are by friction coefficient, the less people can tell them apart. This spurious correlation would be the opposite of intuition, and further calls into question the common practice of using friction coefficients in touch-related studies. The alternative, two-term model which includes adhesive contact area for friction coefficient(29) was even less predictive (see Fig. S6A of SI). We believe such a correlation could not have been uncovered previously as our samples are minimal in their physical variations. Yet, the dynamic changes in force even within a single sample are not considered, despite being a key feature of mesoscale friction during human touch.

We investigate different material properties in Fig. 2E. Differences in average roughness R_a (or other parameters, like root mean square roughness R_rms (Fig. S6A of SI) did not show a statistically significant correlation to accuracy. Though roughness is a popular parameter, correlating any roughness parameter to human performance here could be moot: the limit of detecting roughness differences has previously been defined as 13 nm on structured surfaces(33) and much higher for randomly rough surfaces(46), all of which are magnitudes larger than the roughness differences between our surfaces. The differences in contact angle hysteresis – as an approximation of the adhesion contributions(47) – do not present any statistically significant effects on performance.”

Comment 5

The authors should comment on the influence of friction on perceptual invariance. Despite inducing radially different frictional behavior for various conditions, these surfaces are stably perceived. Maybe this is a sign that humans extract a different metric?

We agree – we are excited that frictional instabilities may offer a more stable perceptual cue because they are not prone to fluctuations (Recommendations for the authors, Comment 3) and instability formation, in many conditions, is invariant to applied pressures and velocities – thus forming large zones where a human may reasonable encounter a given instability.

Raw friction is highly prone to variation during human exploration (in alignment with Recommendations for the authors, Comment 3), but ongoing work seeks to explain tactile constancy, or the ability to identify objects despite these large changes in force. Very recently published work by Fehlberg et. al. identified the role of modulating finger speed and normal force in amplifying the differences in friction coefficient between materials in order to identify them(25), and we postulate that their work may be streamlined and consistent with the idea of friction instabilities, though we have not had a chance to discuss this in-depth with the authors yet.

We think that the instability maps show a viable path forward to how surfaces are stably perceived, and instabilities themselves show a potential mechanism: mathematically, instabilities for given conditions can be invariant to velocity or mass, creating zones where a certain instability is encountered. This reduces the immense variability of friction to a smaller, more stable classification of surfaces (e.g., a 30% SS surface or a 60% SS surface). A given surface will typically produce the same instability at a specific condition (we found some boundaries are extremely condition sensitive, but many conditions are not), whereas a single friction trace which is highly prone to variation is not a stable metric.

Added References (Page 14)

53 M. Fehlberg, E. Monfort, S. Saikumar, K. Drewing and R. Bennewitz, IEEE Trans. Haptics, 2024, 17, 957–963.

References

Z. Liu, J.-T. Kim, J. A. Rogers, R. L. Klatzky and J. E. Colgate, IEEE Trans. Haptics, 2024, 17, 441– 450.

D. Gueorguiev, S. Bochereau, A. Mouraux, V. Hayward and J.-L. Thonnard, Sci Rep, 2016, 6, 25553.

C. W. Carpenter, C. Dhong, N. B. Root, D. Rodriquez, E. E. Abdo, K. Skelil, M. A. Alkhadra, J. Ramírez, V. S. Ramachandran and D. J. Lipomi, Mater. Horiz., 2018, 5, 70–77.

A. Nolin, A. Licht, K. Pierson, C.-Y. Lo, L. V. Kayser and C. Dhong, Soft Matter, 2021, 17, 5050– 5060.

A. Nolin, K. Pierson, R. Hlibok, C.-Y. Lo, L. V. Kayser and C. Dhong, Soft Matter, 2022, 18, 3928– 3940.

Z. Swain, M. Derkaloustian, K. A. Hepler, A. Nolin, V. S. Damani, P. Bhattacharyya, T. Shrestha, J. Medina, L. Kayser and C. Dhong, J. Mater. Chem. B, DOI:10.1039/D4TB01646G.

C. Dhong, L. V. Kayser, R. Arroyo, A. Shin, M. Finn, A. T. Kleinschmidt and D. J. Lipomi, Soft Matter, 2018, 14, 7483–7491.

A. Abdouni, M. Djaghloul, C. Thieulin, R. Vargiolu, C. Pailler-Mattei and H. Zahouani, Royal Society Open Science, DOI:10.1098/rsos.170321.

P.-H. Cornuault, L. Carpentier, M.-A. Bueno, J.-M. Cote and G. Monteil, Journal of The Royal Society Interface, DOI:10.1098/rsif.2015.0495.

K. Qian, K. Traylor, S. W. Lee, B. Ellis, J. Weiss and D. Kamper, J Biomech, 2014, 47, 3094–3099.

Y.-J. Fu, H. Qui, K.-S. Liao, S. J. Lue, C.-C. Hu, K.-R. Lee and J.-Y. Lai, Langmuir, 2010, 26, 4392– 4399.

Y. Yuan and R. Verma, Colloids Surf B Biointerfaces, 2006, 48, 6–12.

G. Yu, J. Hu, J. Tan, Y. Gao, Y. Lu and F. Xuan, Nanotechnology, 2018, 29, 115502.

L. Zheng, S. Dong, J. Nie, S. Li, Z. Ren, X. Ma, X. Chen, H. Li and Z. L. Wang, ACS Appl. Mater. Interfaces, 2019, 11, 42504–42511.

K. Ma, J. Rivera, G. J. Hirasaki and S. L. Biswal, Journal of Colloid and Interface Science, 2011, 363, 371–378.

A. Mavon, H. Zahouani, D. Redoules, P. Agache, Y. Gall and Ph. Humbert, Colloids and Surfaces B: Biointerfaces, 1997, 8, 147–155.

E. AliAbbasi, M. Muzammil, O. Sirin, P. Lefèvre, Ø. G. Martinsen and C. Basdogan, IEEE Trans. Haptics, 2024, 17, 841–849.

G. Corniani, Z. S. Lee, M. J. Carré, R. Lewis, B. P. Delhaye and H. P. Saal, eLife, DOI:10.7554/eLife.93554.1.

J. N. Israelachvili, Intermolecular and Surface Forces, Academic Press, 2011.

S. Das, N. Cadirov, S. Chary, Y. Kaufman, J. Hogan, K. L. Turner and J. N. Israelachvili, J R Soc Interface, 2015, 12, 20141346.

B. N. J. Persson, O. Albohr, C. Creton and V. Peveri, The Journal of Chemical Physics, 2004, 120, 8779–8793.

L. Skedung, M. Arvidsson, J. Y. Chung, C. M. Stafford, B. Berglund and M. W. Rutland, Sci Rep, 2013, 3, 2617.

K. Viswanathan, N. K. Sundaram and S. Chandrasekar, Soft Matter, 2016, 12, 5265–5275.

M. Wiertlewski, C. Hudin and V. Hayward, in 2011 IEEE World Haptics Conference, 2011, pp. 25– 30.

M. Fehlberg, E. Monfort, S. Saikumar, K. Drewing and R. Bennewitz, IEEE Transactions on Haptics, 2024, 17, 957–963.

eLife Assessment

This valuable study offers insights into the role of Leiomodin-1 (LMOD1) in muscle stem cell biology, advancing our understanding of myogenic differentiation and indicating LMOD1 as a regulator of muscle regeneration, aging, and exercise adaptation. The integration of in vitro and in vivo approaches, complemented by proteomic and imaging methodologies, is solid. However, certain aspects require further attention to improve the clarity, impact, and overall significance of the work, particularly in substantiating the in vivo relevance. This work will provide a starting point that will be of value to medical biologists and biochemists working on LMOD and its variants in muscle biology.

Reviewer #1 (Public review):

This manuscript by Ori and colleagues investigates the role of Lmod1 in muscle stem cell activation and differentiation. The study begins with a time-course mass spectrometry analysis of primary muscle stem cells, identifying Lmod1 as a pro-myogenic candidate (Figure 1). While the initial approach is robust, the subsequent characterization lacks depth and clarity. Although the data suggest that Lmod1 promotes myogenesis, the underlying mechanisms remain vague, and key experiments are missing. Please find my comments below.

(1) The authors mainly rely on coarse and less-established readouts such as myotube length and spherical Myh-positive cells. More comprehensive and standard analyses, such as co-staining for Pax7, MyoD, and Myogenin, would allow quantification of quiescent, activated, and differentiating stem cells in knockdown and overexpression experiments. The exact stage at which Lmod1 functions (stem cell, progenitor, or post-fusion) is unclear due to the limited depth of the analysis. Performing similar experiments on cultured single EDL fibers would add valuable insights.

(2) In supplementary Figure 2E, the distinction between Hoechst-positive cells and total cell counts is unclear. The authors should clarify why Hoechst-positive cells increase and relabel "reserve cells," as the term is confusing without reading the legend.

(3) The specificity of Lmod1 and Sirt1 immunostaining needs validation using siRNA-treated samples, especially as these data form the basis of the mechanistic conclusions.

(4) The authors must test the effect of Lmod1 siRNA on Sirt1 localization, as only overexpression experiments are shown.

(5) In Figure S3, the biotin signal in LMOD2 samples appears weak. The authors need to address whether comparing LMOD1 and LMOD2 is valid given the apparent difference in reaction efficiency. It would also help to highlight where Sirt1 falls on the volcano plot in S3B.

(6) The immunostaining data suggest that Lmod1 remains cytoplasmic throughout differentiation, whereas Sirt1 shows transient cytoplasmic localization at day 1 of differentiation. The authors should explain why Sirt1 is not constantly sequestered if Lmod1's cytoplasmic localization is consistent. It is also unclear whether day 1 is the key time point for Lmod1 function, as its precise role during myogenesis remains ambiguous.

(7) The introduction does not sufficiently establish the motivation or knowledge gap this work aims to address. Instead, it reads like a narration of disparate topics in a single paragraph. The authors should clarify the statement in line 150, "since this protein has been...,".

Overall, while the identification of Lmod1 as a pro-myogenic factor is convincing, the mechanistic insights are insufficient, and the manuscript would benefit from addressing these concerns.

Reviewer #2 (Public review):

Summary:

In this manuscript, the authors identify Leiomodin-1 (LMOD1) as a key regulator of early myogenic differentiation, demonstrating its interaction with SIRT1 to influence SIRT1's cellular localization and gene expression. The authors propose that LMOD1 translocates SIRT1 from the nucleus to the cytoplasm to permit the expression of myogenic differentiating genes such as MYOD or Myogenin.

Strengths:

A major strength of this work lies in the robust temporal resolution achieved through a time-course mass spectrometry analysis of in vitro muscle differentiation. This provides novel insights into the dynamic process of myogenic differentiation, often under-explored in terms of temporal progression. The authors provide a strong mechanistic case for how LMOD1 exerts its role in muscle differentiation which opens avenues to modulate.

Weaknesses:

One limitation of the study is the in vivo data. Although the authors do translate their findings in vivo for LMOD1 localization and expression, the cross-sectional imaging is not highly convincing. Longitudinal cuts or isolated fibers could have been more useful specimens to answer these questions. Moreover, the authors do not assess their in vitro SIRT1 findings in vivo. A few key experiments in regenerating or aged mice would strengthen the mechanistic insight of the findings.

Discussion:

Overall, the study emphasizes the importance of understanding the temporal dynamics of molecular players during myogenic differentiation and provides valuable proteomic data that will benefit the field. Future studies should explore whether LMOD1 modulates the nuclear-cytoplasmic shuttling of other transcription factors during muscle development and how these processes are mechanistically achieved. Investigating whether LMOD1 can be therapeutically targeted to enhance muscle regeneration in contexts such as exercise, aging, and disease will be critical for translational applications. Additionally, elucidating the interplay among LMOD1, LMOD2, and LMOD3 could uncover broader implications for actin cytoskeletal regulation in muscle biology.

Reviewer #3 (Public review):

Summary:

In this manuscript, the investigators identified LMOD1 as one of a subset of cytoskeletal proteins whose levels increase in the early stages of myogenic differentiation. Lmod1 is understudied in striated muscle and in particular in myogenic differentiation. Thus, this is an important study. It is also a very thorough study - with perhaps even too much data presented. Importantly, the investigators observed that LMOD1 appears to be important for skeletal regeneration, and myogenic differentiation and that it interacts with SIRT1. Both primary myoblast differentiation and skeletal muscle regeneration were studied. Rescue experiments confirmed these observations: SIRT1 can rescue perturbations of myogenic differentiation as a result of LMOD1 knockdown.

Strengths:

Particular strengths include: important topic, the use of primary skeletal cultures, the use of both cell culture and in vivo approaches, careful biomarker analysis of primary mouse myoblast differentiation, the use of two methods to probe the function of the Lmod1/SIRT1 pathway via using depletion approaches and inhibitors, and generation of six independent myoblast cultures. Results support their conclusions.

Weaknesses:

(1) Figure 1. Images of cells in Figure 1A are too small to be meaningful (especially in comparison to the other data presented in this figure). Perhaps the authors could make graphs smaller?

(2) Line 148 "We found LMOD2 to be the most abundant Lmod in whole skeletal muscle." This is confusing since most if not all prior studies have shown that Lmod3 is the predominant isoform in skeletal muscle. The two papers that are cited are incorrectly cited. Clarification to resolve this discrepancy is needed.

(3) Figure 2. Immunoflorescence (IF) panels are too small to be meaningful. Perhaps the graphs could be made smaller and more space allocated for the IF panels? This issue is apparent for just about all IF panels - they are simply too small to be meaningful. Additionally, in many of the immunofluorescence figures, the colors that were used make it difficult to discern the stained cellular structures. For example in Figure S1, orange and purple are used - they do not stand out as well as other colors that are more commonly used.

(4) There is huge variability in many experiments presented - as such, more samples appear to be required to allow for meaningful data to be obtained. For example, Figure S2. Many experimental groups, only have 3 samples - this is highly problematic - I would estimate that 5-6 would be the minimum.

(5) Ponceau S staining is often used as a loading control in this manuscript for western blots. The area/molecular weight range actually used should be specified. Not clear why in some experiments GAPDH staining is used, in other experiments Ponceau S staining is used, and in some, both are used. In some experiments, the variability of total protein loaded from lane to lane is disconcerting. For example, in Figure S4C there appears to be more than normal variability. Can the protein assay be redone and samples run again?

(6) Figure S3 - Lmod3 is included in the figure but no mention of it occurs in the title of the figure and/or legend.

(7) Abstract, line 25. "overexpression accelerates and improves the formation of myotubes". This is a confusing sentence. How is it improving the formation? A little more information about how they are different than developing myotubes in normal/healthy muscles would be helpful.

(8) It is impossible from the IF figures presented to determine where Lmod1 localizes in the myocytes. Information on its subcellular localization is important. Does it localize with Lmod2 and Lmod3 at thin filament pointed ends?

eLife Assessment

ProtSSN is a valuable approach that generates protein embeddings by integrating sequence and structural information, demonstrating improved prediction of mutation effects on thermostability compared to competing models. The evidence supporting the authors' claims is compelling, with well-executed comparisons. This work will be of particular interest to researchers in bioinformatics and structural biology, especially those focused on protein function and stability.

Reviewer #1 (Public review):

Summary:

The authors introduce a denoising-style model that incorporates both structure and primary-sequence embeddings to generate richer embeddings of peptides. My understanding is that the authors use ESM for the primary sequence embeddings, take resolved structures (or use structural predictions from AlphaFold when they're not available), then develop an architecture to combine these two with a loss that seems reminiscent of diffusion models or masked language model approaches. The embeddings can be viewed as ensemble-style embedding of the two levels of sequence information, or with AlphaFold, an ensemble of two methods (ESM+AlphaFold). The authors also gather external datasets to evaluate their approach and compare it to previous approaches. The approach seems promising and appears to out-compete previous methods at several tasks. Nonetheless, I have strong concerns about a lack of verbosity as well as exclusion of relevant methods and references.

Advances:

I appreciate the breadth of the analysis and comparisons to other methods. The authors separate tasks, models, and sizes of models in an intuitive, easy-to-read fashion that I find valuable for selecting a method for embedding peptides. Moreover, the authors gather two datasets for evaluating embeddings' utility for predicting thermostability. Overall, the work should be helpful for the field as more groups choose methods/pretraining strategies amenable to their goals, and can do so in an evidence-guided manner.

Considerations:

Primarily, a majority of the results and conclusions (e.g., Table 3) are reached using data and methods from ProteinGym, yet the best-performing methods on ProteinGym are excluded from the paper (e.g., EVE-based models and GEMME). In the ProteinGym database, these methods outperform ProtSSN models. Moreover, these models were published over a year---or even 4 years in the case of GEMME---before ProtSSN, and I do not see justification for their exclusion in the text.

Secondly, related to comparison of other models, there is no section in the methods about how other models were used, or how their scores were computed. When comparing these models, I think it's crucial that there are explicit derivations or explanations for the exact task used for scoring each method. In other words, if the pre-training is indeed the important advance of the paper, the paper needs to show this more explicitly by explaining exactly which components of the model (and previous models) are used for evaluation. Are the authors extracting the final hidden layer representations of the model, treating these as features, then using these features in a regression task to predict fitness/thermostability/DDG etc.? How are the model embeddings of other methods being used, since, for example, many of these methods output a k-dimensional embedding of a given sequence, rather than one single score that can be correlated with some fitness/functional metric. Summarily, I think the text is lacking an explicit mention of how these embeddings are being summarized or used, as well as how this compares to the model presented.

I think the above issues can mainly be addressed by considering and incorporating points from Li et al. 2024[1] and potentially Tang & Koo 2024[2]. Li et al.[1] make extremely explicit the use of pretraining for downstream prediction tasks. Moreover, they benchmark pretraining strategies explicitly on thermostability (one of the main considerations in the submitted manuscript), yet there is no mention of this work nor the dataset used (FLIP (Dallago et al., 2021)) in this current work. I think a reference and discussion of [1] is critical, and I would also like to see comparisons in line with [1], as [1] is very clear about what features from pretraining are used, and how. If the comparisons with previous methods were done in this fashion, this level of detail needs to be included in the text.

To conclude, I think the manuscript would benefit substantially from a more thorough comparison of previous methods. Maybe one way of doing this is following [1] or [2], and using the final embeddings of each method for a variety of regression tasks---to really make clear where these methods are performing relative to one another. I think a more thorough methods section detailing how previous methods did their scoring is also important. Lastly, TranceptEVE (or a model comparable to it) and GEMME should also be mentioned in these results, or at the bare minimum, be given justification for their absence.

[1] Feature Reuse and Scaling: Understanding Transfer Learning with Protein Language Models, Francesca-Zhoufan Li, Ava P. Amini, Yisong Yue, Kevin K. Yang, Alex X. Lu bioRxiv 2024.02.05.578959; doi: https://doi.org/10.1101/2024.02.05.578959<br /> [2] Evaluating the representational power of pre-trained DNA language models for regulatory genomics, Ziqi Tang, Peter K Koo bioRxiv 2024.02.29.582810; doi: https://doi.org/10.1101/2024.02.29.582810

Comments on revisions:

My concerns have been addressed. What seems to remain are some semantical disagreements and I'm not sure that these will be answered here. Do MSAs and other embedding methods lead to some notable type of data leakage? Does this leakage qualify as "x-shot" learning under current definitions?

Reviewer #2 (Public review):

Summary:

To design proteins and predict disease, we want to predict the effects of mutations on the function of a protein. To make these predictions, biologists have long turned to statistical models that learn patterns that are conserved across evolution. There is potential to improve our predictions however by incorporating structure. In this paper the authors build a denoising auto-encoder model that incorporates sequence and structure to predict mutation effects. The model is trained to predict the sequence of a protein given its perturbed sequence and structure. The authors demonstrate that this model is able to predict the effects of mutations better than sequence-only models.

As well, the authors curate a set of assays measuring the effect of mutations on thermostability. They demonstrate their model also predicts the effects of these mutations better than previous models and make this benchmark available for the community.

Strengths:

The authors describe a method that makes accurate mutation effect predictions by informing its predictions with structure.

The authors curate a new dataset of assays measuring thermostability. These can be used to validate and interpret mutation effect prediction methods in the future.

Weaknesses:

In the review period, the authors included a previous method, SaProt, that similarly uses protein structure to predict the effects of mutations, in their evaluations. They see that SaProt performs similarly to their method.

ProteinGym is largely made of deep mutational scans, which measure the effect of every mutation on a protein. These new benchmarks contain on average measurements of less than a percent of all possible point mutations of their respective proteins. It is unclear what sorts of protein regions these mutations are more likely to lie in; therefore it is challenging to make conclusions about what a model has necessarily learned based on its score on this benchmark. For example, several assays in this new benchmark seem to be similar to each other, such as four assays on ubiquitin performed in pH 2.25 to pH 3.0.

Comments on revisions:

I think the rounds of review have improved the paper and I've raised my score.

Author response:

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

Response to Reviewer 1

Thank you for your recognition of our revised work.

Response to Reviewer 2

It would be useful to have a demonstration of where this model outperforms SaProt systematically, and a discussion about what the success of this model teaches us given there is a similar, previously successful model, SaProt.

As two concurrent works, ProtSSN and SaProt employ different methods to incorporate the structure information of proteins. Generally speaking, for two deep learning models that are developed during a close period, it is challenging to conclude that one model is systematically superior to another. Nonetheless, on DTm and DDG (the two low-throughput datasets that we constructed), ProtSSN demonstrates better empirical performance than SaProt.  

Moreover, ProtSSN is more efficient in both training and inference compared to SaProt. In terms of training cost, SaProt uses 40 million protein structures for pretraining (requiring 64 A100 GPUs for three months), whereas ProtSSN requires only about 30,000 crystal structures from the CATH database (trained on a single 3090 GPU for two days). Despite SaProt’s significantly higher training cost, its pretrained version does not exhibit superior performance on low-throughput datasets such as DTm, DDG, and Clinvar. Furthermore, the high training cost limits many users from retraining or fine-tuning the model for specific needs or datasets.

Regarding the inference cost, ProtSSN requires only one embedding computation for a wild-type protein, regardless of the number of mutants (n). In contrast, SaProt computes a separate embedding and score for each mutant. For instance, when evaluating the scoring performance on ProteinGym, ProtSSN only needs 217 inferences, while SaProt needs more than 2M inferences. This inference speed is important in practice, such as high-throughput design and screening.

Please remove the reference to previous methods as "few shot". This typically refers to their being trained on experimental data, not their using MSAs. A "few shot" model would be ProteinNPT.

The definition of "few-shot" we used here is following ESM1v [1]. This concept originates from providing a certain number of examples as input to GPT-3 [2]. In the context of protein deep learning models, MSA serves as the wild-type protein examples.

Also, Reviewer 1 uses the concept in the same way. 

“Readers should note that methods labelled as "few-shot" in comparisons do not make use of experimental labels, but rather use sequences inferred as homologous; these sequences are also often available even if the protein has never been experimentally tested.”

In the main text, we also included this definition as well as the reference of ESM-1v in lines 457-458.

“We extend the evaluation on ProteinGym v0 to include a comparison of our zero-shot ProtSSN with few-shot learning methods that leverage MSA information of proteins (Meier et al., 2021).”

(1) Meier J, Rao R, Verkuil R, et al. Language models enable zero-shot prediction of the effects of mutations on protein function. Advances in Neural Information Processing Systems, 2021.

(2) Brown T, Mann B, Ryder N, et al. Language models are few-shot learners. Advances in Neural Information Processing Systems, 2020.

Furthermore, I don't think it is fair to state that your method is not comparable to these models -- one can run an MSA just as one can predict a structure. A fairer comparison would be to highlight particular assays for which getting an MSA could be challenging -- Transcription did this by showing that they outperform EVE when MSAs are shallow.

We recognize that there are often differences in the definitions and classifications of various methodologies. Here, we follow the definitions provided by ProteinGym. As the most comprehensive and large scale open benchmark in the community, we believe this classification scheme should be widely accepted. All classifications are available on the official website of ProteinGym (https://proteingym.org/benchmarks), which categorizes methods into PLMs, Structure-based models, and Alignment-based models. For example, GEMME is classified as an alignment-based model, and MSA Transformer is considered a hybrid model combining alignment and PLM features.

We believe that methodologies with different inputs and architectures can lead to inherent unfairness. Also, it is generally believed that models including evolutionary relationships tend to outperform end-to-end models due to the extra information and efforts involved during the training phase. Some empirical evidence and discussions are in the ablation studies of retrieval factors in Tranception [3]. Moreover, the choice of MSA search parameters can introduce uncertainty, which could have positive or negative impacts. 

We showcase the impact of MSA depth on model performance with an additional analysis below. Author response image 1 visualizes the Spearman’s correlation between the scores of each model and the number of MSAs on 217 ProteinGym assays, where each point represents one of 217 assays. The summary correlation of each model with respect to all assays are reported in Author response table 1. These results demonstrate no clear correlation between MSA depth and model performance even for MSA-based models.

Author response image 1.

Scatter plots of the number of MSA sequences and spearman’s correlation.

Author response table 1.

Spearmar’s score of the number of MSA sequences and the model’s performance.

(3) Notin P, Dias M, Frazer J, et al. Tranception: protein fitness prediction with autoregressive transformers and inference-time retrieval. International Conference on Machine Learning, 2022.

The authors state that DTm and DDG are conceptually appealing because they come from low-throughput assays with lower experimental noise and are also mutations that are particularly chosen to represent the most interesting regions of the protein. I agree with the conceptual appeal but I don't think these claims have been demonstrated in practice. The cited comparison with Frazer as a particularly noisy source of data I think is particularly unconvincing: ClinVar labels are not only rigorously determined from multiple sources of evidence, Frazer et al demonstrates that these labels are actually more reliable than experiment in some cases. They also state that ProteinGym data doesn't come with environmental conditions, but these can be retrieved from the papers the assays came from. The paper would be strengthened by a demonstration of the conceptual benefit of these new datasets, say a comparison of mutations and signal for a protein that may be in one of these datasets vs ProteinGym.

In the work by Frazer et al. [4], they mentioned that

"However, these technologies do not easily scale to thousands of proteins, especially not to combinations of variants, and depend critically on the availability of assays that are relevant to or at least associated with human disease phenotypes." 

It points out that the results of high-throughput experiments are usually based on the design of specific genes (such as BRCA1 and TP53.) and cannot be easily extended to thousands of other genes. At the same time, due to the complexity of the experiment, there may be problems with reproducibility or deviations from clinical relevance.

This statement aligns with our perspective that high-throughput experiments inherently involve a significant amount of noise and error. It is important to clarify that the noise we discuss here arises from the limitations of high-throughput experiments themselves, instead of from the reliability of the data sources, such as systematic errors in experimental measurements. This latter issue is a complex problem common to all wetlab experiments and falls outside the scope of our study.

Under this premise, low-throughput datasets like DTm and DDG can be considered to have less noise than high-throughput datasets, as they have undergone manual curation. As for your suggestion, while valuable, unfortunately, we were unable to identify datasets in DTM and DDG that align with those in ProteinGym after a careful search. Thus, we are unable to conduct this comparative experiment at this stage.

(4) Frazer J, Notin P, Dias M, et al. Disease variant prediction with deep generative models of evolutionary data. Nature, 2021.

eLife Assessment

This study presents a valuable theoretical exploration on the electrophysiological mechanisms of ionic currents via gap junctions in hippocampal CA1 pyramidal-cell models, and their potential contribution to local field potentials (LFPs) that is different from the contribution of chemical synapses. The biophysical argument regarding electric dipoles appears solid, but the evidence would be stronger if their predictions are tested against experiments. A shortage of model validation and strictly comparable parameters used in the comparisons between chemical vs. junctional inputs makes the modeling approach incomplete; once strengthened, the finding can be of broad interest to electrophysiologists, who often make recordings from regions of neurons interconnected with gap junctions.

Reviewer #1 (Public review):

This manuscript makes a significant contribution to the field by exploring the dichotomy between chemical synaptic and gap junctional contributions to extracellular potentials. While the study is comprehensive in its computational approach, adding experimental validation, network-level simulations, and expanded discussion on implications would elevate its impact further.

Strengths:

Novelty and Scope:<br /> The manuscript provides a detailed investigation into the contrasting extracellular field potential (EFP) signatures arising from chemical synapses and gap junctions, an underexplored area in neuroscience.<br /> It highlights the critical role of active dendritic processes in shaping EFPs, pushing forward our understanding of how electrical and chemical synapses contribute differently to extracellular signals.

Methodological Rigor:<br /> The use of morphologically and biophysically realistic computational models for CA1 pyramidal neurons ensures that the findings are grounded in physiological relevance.<br /> Systematic analysis of various factors, including the presence of sodium, leak, and HCN channels, offers a clear dissection of how transmembrane currents shape EFPs.

Biological Relevance:<br /> The findings emphasize the importance of incorporating gap junctional inputs in analyses of extracellular signals, which have traditionally focused on chemical synapses.<br /> The observed polarity differences and spectral characteristics provide novel insights into how neural computations may differ based on the mode of synaptic input.

Clarity and Depth:<br /> The manuscript is well-structured, with a logical progression from synchronous input analyses to asynchronous and rhythmic inputs, ensuring comprehensive coverage of the topic.

Weaknesses and Areas for Improvement:

Generality and Validation:<br /> The study focuses exclusively on CA1 pyramidal neurons. Expanding the analysis to other cell types, such as interneurons or glial cells, would enhance the generalizability of the findings.<br /> Experimental validation of the computational predictions is entirely absent. Empirical data correlating the modeled EFPs with actual recordings would strengthen the claims.

Role of Active Dendritic Currents:<br /> The paper emphasizes active dendritic currents, particularly the role of HCN channels in generating outward currents under certain conditions. However, further discussion of how this mechanism integrates into broader network dynamics is warranted.

Analysis of Plasticity:<br /> While the manuscript mentions plasticity in the discussion, there are no simulations that account for activity-dependent changes in synaptic or gap junctional properties. Including such analyses could significantly enhance the relevance of the findings.

Frequency-Dependent Effects:<br /> The study demonstrates that gap junctional inputs suppress high-frequency EFP power due to membrane filtering. However, it could delve deeper into the implications of this for different brain rhythms, such as gamma or ripple oscillations.

Visualization:<br /> Figures are dense and could benefit from more intuitive labeling and focused presentations. For example, isolating key differences between chemical and gap junctional inputs in distinct panels would improve clarity.

Contextual Relevance:<br /> The manuscript touches on how these findings relate to known physiological roles of gap junctions (e.g., in gamma rhythms) but does not explore this in depth. Stronger integration of the results into known neural network dynamics would enhance its impact.

Suggestions for Improvement:

Broader Application:<br /> Simulate EFPs in multi-neuron networks to assess how the findings extend to network-level interactions, particularly in regions with mixed synaptic connectivity.

Experimental Correlation:<br /> Collaborate with experimental groups to validate the computational predictions using in vivo or in vitro recordings.

Mechanistic Insights:<br /> Provide a more detailed mechanistic explanation of how specific ionic currents (e.g., HCN, sodium, leak) interact during gap junctional vs. chemical synaptic inputs.

Implications for Neural Coding:<br /> Discuss how the observed differences in EFP signatures might influence neural coding, especially in circuits with heavy gap junctional connectivity.

Reviewer #2 (Public review):

Summary:

This computational work examines whether the inputs that neurons receive through electrical synapses (gap junctions) have different signatures in the extracellular local field potential (LFP) compared to inputs via chemical synapses. The authors present the results of a series of model simulations where either electric or chemical synapses targeting a single hippocampal pyramidal neuron are activated in various spatio-temporal patterns, and the resulting LFP in the vicinity of the cell is calculated and analyzed. The authors find several notable qualitative differences between the LFP patterns evoked by gap junctions vs. chemical synapses. For some of these findings, the authors demonstrate convincingly that the observed differences are explained by the electric vs. chemical nature of the input, and these results likely generalize to other cell types. However, in other cases, it remains plausible (or even likely) that the differences are caused, at least partly, by other factors (such as different intracellular voltage responses due to, e.g., the unequal strengths of the inputs). Furthermore, it was not immediately clear to me how the results could be applied to analyze more realistic situations where neurons receive partially synchronized excitatory and inhibitory inputs via chemical and electric synapses.

Strengths:

The main strength of the paper is that it draws attention to the fact that inputs to a neuron via gap junctions are expected to give rise to a different extracellular electric field compared to inputs via chemical synapses, even if the intracellular effects of the two types of input are similar. This is because, unlike chemical synaptic inputs, inputs via gap junctions are not directly associated with transmembrane currents. This is a general result that holds independent of many details such as the cell types or neurotransmitters involved.

Another strength of the article is that the authors attempt to provide intuitive, non-technical explanations of most of their findings, which should make the paper readable also for non-expert audiences (including experimentalists).

Weaknesses:

The most problematic aspect of the paper relates to the methodology for comparing the effects of electric vs. chemical synaptic inputs on the LFP. The authors seem to suggest that the primary cause of all the differences seen in the various simulation experiments is the different nature of the input, and particularly the difference between the transmembrane current evoked by chemical synapses and the gap junctional current that does not involve the extracellular space. However, this is clearly an oversimplification: since no real attempt is made to quantitatively match the two conditions that are compared (e.g., regarding the strength and temporal profile of the inputs), the differences seen can be due to factors other than the electric vs. chemical nature of synapses. In fact, if inputs were identical in all parameters other than the transmembrane vs. directly injected nature of the current, the intracellular voltage responses and, consequently, the currents through voltage-gated and leak currents would also be the same, and the LFPs would differ exactly by the contribution of the transmembrane current evoked by the chemical synapse. This is evidently not the case for any of the simulated comparisons presented, and the differences in the membrane potential response are rather striking in several cases (e.g., in the case of random inputs, there is only one action potential with gap junctions, but multiple action potentials with chemical synapses). Consequently, it remains unclear which observed differences are fundamental in the sense that they are directly related to the electric vs. chemical nature of the input, and which differences can be attributed to other factors such as differences in the strength and pattern of the inputs (and the resulting difference in the neuronal electric response).

Some of the explanations offered for the effects of cellular manipulations on the LFP appear to be incomplete. More specifically, the authors observed that blocking leak channels significantly changed the shape of the LFP response to synchronous synaptic inputs - but only when electric inputs were used, and when sodium channels were intact. The authors seemed to attribute this phenomenon to a direct effect of leak currents on the extracellular potential - however, this appears unlikely both because it does not explain why blocking the leak conductance had no effect in the other cases, and because the leak current is several orders of magnitude smaller than the spike-generating currents that make the largest contributions to the LFP. An indirect effect mediated by interactions of the leak current with some voltage-gated currents appears to be the most likely explanation, but identifying the exact mechanism would require further simulation experiments and/or a detailed analysis of intracellular currents and the membrane potential in time and space.

In every simulation experiment in this study, inputs through electric synapses are modeled as intracellular current injections of pre-determined amplitude and time course based on the sampled dendritic voltage of potential synaptic partners. This is a major simplification that may have a significant impact on the results. First, the current through gap junctions depends on the voltage difference between the two connected cellular compartments and is thus sensitive to the membrane potential of the cell that is treated as the neuron "receiving" the input in this study (although, strictly speaking, there is no pre- or postsynaptic neuron in interactions mediated by gap junctions). This dependence on the membrane potential of the target neuron is completely missing here. A related second point is that gap junctions also change the apparent membrane resistance of the neurons they connect, effectively acting as additional shunting (or leak) conductance in the relevant compartments. This effect is completely missed by treating gap junctions as pure current sources.

One prominent claim of the article that is emphasized even in the abstract is that HCN channels mediate an outward current in certain cases. Although this statement is technically correct, there are two reasons why I do not consider this a major finding of the paper. First, as the authors acknowledge, this is a trivial consequence of the relatively slow kinetics of HCN channels: when at least some of the channels are open, any input that is sufficiently fast and strong to take the membrane potential across the reversal potential of the channel will lead to the reversal of the polarity of the current. This effect is quite generic and well-known and is by no means specific to gap junctional inputs or even HCN channels. Second, and perhaps more importantly, the functional consequence of this reversed current through HCN channels is likely to be negligible. As clearly shown in Supplementary Figure S3, the HCN current becomes outward only for an extremely short time period during the action potential, which is also a period when several other currents are also active and likely dominant due to their much higher conductances. I also note that several of these relevant facts remain hidden in Figure 3, both because of its focus on peak values, and because of the radically different units on the vertical axes of the current plots.

Finally, I missed an appropriate validation of the neuronal model used, and also the characterization of the effects of the in silico manipulations used on the basic behavior of the model. As far as I understand, the model in its current form has not been used in other studies. If this is the case, it would be important to demonstrate convincingly through (preferably quantitative) comparisons with experimental data using different protocols that the model captures the physiological behavior of at least the relevant compartments (in this case, the dendrites and the soma) of hippocampal pyramidal neurons sufficiently well that the results of the modeling study are relevant to the real biological system. In addition, the correct interpretation of various manipulations of the model would be strongly facilitated by investigating and discussing how the physiological properties of the model neuron are affected by these alterations.

Author response:

eLife Assessment

This study presents a valuable theoretical exploration on the electrophysiological mechanisms of ionic currents via gap junctions in hippocampal CA1 pyramidal-cell models, and their potential contribution to local field potentials (LFPs) that is different from the contribution of chemical synapses. The biophysical argument regarding electric dipoles appears solid, but the evidence can be more convincing if their predictions are tested against experiments. A shortage of model validation and strictly comparable parameters used in the comparisons between chemical vs. junctional inputs makes the modeling approach incomplete; once strengthened, the finding can be of broad interest to electrophysiologists, who often make recordings from regions of neurons interconnected with gap junctions.

We gratefully thank the editors and the reviewers for the time and effort in rigorously assessing our manuscript, for the constructive review process, for their enthusiastic responses to our study, and for the encouraging and thoughtful comments. We especially thank you for deeming our study to be a valuable exploration on the differential contributions of active dendritic gap junctions vs. chemical synapses to local field potentials. We thank you for your appreciation of the quantitative biophysical demonstration on the differences in electric dipoles that appear in extracellular potentials with gap junctions vs. chemical synapses.

However, we are surprised by aspects of the assessment that resulted in deeming the approach incomplete, especially given the following with specific reference to the points raised:

(1) Testing against experiments: With specific reference to gap junctions, quantitative experimental verification becomes extremely difficult because of the well-established nonspecificities associated with gap junctional modulators (Behrens et al., 2011; Rouach et al., 2003). The non-specific actions of gap junctions are tabulated in Table 2 of (Szarka et al., 2021), reproduced below. In addition, genetic knockouts of gap junctional proteins are either lethal or involve functional compensation (Bedner et al., 2012; Lo, 1999), together making causal links to specific gap junctional contributions with currently available techniques infeasible.

In addition, the complex interactions between co-existing chemical synaptic, gap junctional, and active dendritic contributions from several cell-types make the delineation of the contributions of specific components infeasible with experimental approaches. A computational approach is the only quantitative route to specifically delineate the contributions of individual components to extracellular potentials, as seen from studies that have addressed the question of active dendritic contributions to field potentials (Halnes et al., 2024; Ness et al., 2018; Reimann et al., 2013; Sinha & Narayanan, 2015, 2022) or spiking contributions to local field potentials (Buzsaki et al., 2012; Gold et al., 2006; Schomburg et al., 2012). The biophysically and morphologically realistic computational modeling route is therefore invaluable in assessing the impact of individual components to extracellular field potentials (Einevoll et al., 2019; Halnes et al., 2024).

Together, we emphasize that the computational modeling route is currently the only quantitative methodology to delineate the contributions of gap junctions vs. chemical synapses to extracellular potentials.

(2) Model validation: The model used in this study was adopted from a physiologically validated model from our laboratory (Roy & Narayanan, 2021). Please note that the original model was validated against several physiological measurements along the somatodendritic axis. We sincerely regret our oversight in not mentioning clearly that we have used an existing, thoroughly physiologically-validated model from our laboratory in this study.

(3) Comparisons between chemical vs. junctional inputs: We had taken elaborate precautions in our experimental design to match the intracellular electrophysiological signatures with reference to synchronous as well as oscillatory inputs, irrespective of whether inputs arrived through gap junctions or chemical synapses.

In a revised manuscript, we will address all the concerns raised by the reviewers in detail. We have provided point-by-point responses to reviewers’ helpful and constructive comments below. We thank the editors and the reviewers for this constructive review process, which we believe will help us in improving our manuscript with specific reference to emphasizing the novelty of our approach and conclusions.

Reviewer #1 (Public review):

This manuscript makes a significant contribution to the field by exploring the dichotomy between chemical synaptic and gap junctional contributions to extracellular potentials. While the study is comprehensive in its computational approach, adding experimental validation, network-level simulations, and expanded discussion on implications would elevate its impact further.

We gratefully thank you for your time and effort in rigorously assessing our manuscript, for the enthusiastic response, and the encouraging and thoughtful comments on our study. In what follows, we have provided point-by-point responses to the specific comments.

Strengths

Novelty and Scope

The manuscript provides a detailed investigation into the contrasting extracellular field potential (EFP) signatures arising from chemical synapses and gap junctions, an underexplored area in neuroscience. It highlights the critical role of active dendritic processes in shaping EFPs, pushing forward our understanding of how electrical and chemical synapses contribute differently to extracellular signals.

We thank you for the positive comments on the novelty of our approach and how our study addresses an underexplored area in neuroscience. The assumptions about the passive nature of dendritic structures had indeed resulted in an underestimation of the contributions of gap junctions to extracellular potentials. Once the realities of active structures are accounted for, the contributions of gap junctions increases by several orders of magnitude compared to passive structures (Fig. 1D).

Methodological Rigor

The use of morphologically and biophysically realistic computational models for CA1 pyramidal neurons ensures that the findings are grounded in physiological relevance. Systematic analysis of various factors, including the presence of sodium, leak, and HCN channels, offers a clear dissection of how transmembrane currents shape EFPs.

We thank you for your encouraging comments on the experimental design and methodological rigor of our approach.

Biological Relevance

The findings emphasize the importance of incorporating gap junctional inputs in analyses of extracellular signals, which have traditionally focused on chemical synapses. The observed polarity differences and spectral characteristics provide novel insights into how neural computations may differ based on the mode of synaptic input.

We thank you for your positive comments on the biological relevance of our approach. We also gratefully thank you for emphasizing the two striking novelties unveiling the dichotomy between gap junctions and chemical synapses in their contributions to field potentials: polarity differences and spectral characteristics.

Clarity and Depth

The manuscript is well-structured, with a logical progression from synchronous input analyses to asynchronous and rhythmic inputs, ensuring comprehensive coverage of the topic.

We sincerely thank you for the positive comments on the structure and comprehensive coverage of our manuscript encompassing different types of inputs that neurons typically receive.

Weaknesses and Areas for Improvement

Generality and Validation

The study focuses exclusively on CA1 pyramidal neurons. Expanding the analysis to other cell types, such as interneurons or glial cells, would enhance the generalizability of the findings. Experimental validation of the computational predictions is entirely absent. Empirical data correlating the modeled EFPs with actual recordings would strengthen the claims.

We thank you for raising this important point. The prime novelty and the principal conclusion of this study is that gap junctional contributions to extracellular field potentials are orders of magnitude higher when the active nature of cellular compartments are accounted for. The lacuna in the literature has been consequent to the assumption that cellular compartments are passive, resulting in the dogma that gap junctional contributions to field potentials are negligible. Despite knowledge about active dendritic structures for decades now, this assumption has kept studies from understanding or even exploring the contributions of gap junctions to field potentials. The rationale behind the choice of a computational approach to address the lacuna were as follows:

(1) The complex interactions between co-existing chemical synaptic, gap junctional, and active dendritic contributions from several cell-types make the delineation of the contributions of specific components infeasible with experimental approaches. A computational approach is the only quantitative route to specifically delineate the contributions of individual components to extracellular potentials, as seen from studies that have addressed the question of active dendritic contributions to field potentials (Halnes et al., 2024; Ness et al., 2018; Reimann et al., 2013; Sinha & Narayanan, 2015, 2022) or spiking contributions to local field potentials (Buzsaki et al., 2012; Gold et al., 2006; Schomburg et al., 2012). The biophysically and morphologically realistic computational modeling route is therefore invaluable in assessing the impact of individual components to extracellular field potentials (Einevoll et al., 2019; Halnes et al., 2024).

(2) With specific reference to gap junctions, quantitative experimental verification becomes extremely difficult because of the well-established non-specificities associated with gap junctional modulators (Behrens et al., 2011; Rouach et al., 2003). The non-specific actions of gap junctions are tabulated in Table 2 of (Szarka et al., 2021). In addition, genetic knockouts of gap junctional proteins are either lethal or involve functional compensation (Bedner et al., 2012; Lo, 1999), together making causal links to specific gap junctional contributions with currently available techniques infeasible.

We highlight the novelty of our approach and of the conclusions about differences in extracellular signatures associated with active-dendritic chemical synapses and gap junctions, against these experimental difficulties. We emphasize that the computational modeling route is currently the only quantitative methodology to delineate the contributions of gap junctions vs. chemical synapses to extracellular potentials. Our analyses clearly demonstrates that gap junctions do contribute to extracellular potentials if the active nature of the cellular compartments is explicitly accounted for (Fig. 1D). We also show theoretically well-grounded and mechanistically elucidated differences in polarity (Figs. 1–3) as well as in spectral signatures (Figs. 5–8) of extracellular potentials associated with gap junctional vs. chemical synaptic inputs. Together, our fundamental demonstration in this study is the critical need to account for the active nature of cellular compartments in studying gap junctional contributions of extracellular potentials, with CA1 pyramidal neuronal dendrites used as an exemplar.

In a revised version of the manuscript, we will emphasize the motivations for the approach we took, highlighting the specific novelties both in methodological and conceptual aspects, finally emphasizing the need to account for other cell types and gap junctional contributions therein. Importantly, we will emphasize the non-specificities associated with gap-junctional blockers as the reason why experimental delineation of gap junctional vs. chemical synaptic contributions to LFP becomes tedious. We hope that these points will underscore the need for the computational approach that we took to address this important question, apart from the novelties of the manuscript.

Role of Active Dendritic Currents

The paper emphasizes active dendritic currents, particularly the role of HCN channels in generating outward currents under certain conditions. However, further discussion of how this mechanism integrates into broader network dynamics is warranted.

We thank you for this constructive suggestion. We agree that it is important to consider the implications for broader network dynamics of the outward HCN currents that are observed with synchronous inputs. In a revised manuscript, we will elaborate on the implications of the outward HCN current to network dynamics in detail.

Analysis of Plasticity

While the manuscript mentions plasticity in the discussion, there are no simulations that account for activity-dependent changes in synaptic or gap junctional properties. Including such analyses could significantly enhance the relevance of the findings.

We thank you for this constructive suggestion. Please note that we have presented consistent results for both fewer and more gap junctions in our analyses (Figure 1 with 217 gap junctions and Supplementary Figure 1 with 99 gap junctions). Thus, our fundamentally novel result that gap junctions onto active dendrites differentially shape LFPs holds true irrespective of the relative density of gap junctions onto the neuron. Thus, these results demonstrate that the conclusions about their contributions to LFP are invariant to plasticity in their gap junctional numerosity.

We had only briefly mentioned plasticity in the Introduction to highlight the different modes of synaptic transmission and to emphasize that plasticity has been studied in both chemical synapses and gap junctions, playing a role in learning and adaptation. However, if this wording inadvertently suggests that our study includes plasticity simulations, we would remove it from Introduction in the updated manuscript to ensure clarity.

In the ‘Limitations of analyses and future studies’ section in Discussion, we suggested investigating the impact of plasticity mechanisms—specifically, activity-dependent plasticity of ion channels—on synaptic receptors vs. gap junctions and their effects on extracellular field potentials under various input conditions and plasticity combinations across different structures. We fully agree with the reviewer that such studies would offer valuable insights and further enhance the broader relevance of our findings. However, while our study implies this direction, it was not the primary focus of our investigation.

In the revised manuscript, we will expand on intrinsic/synaptic plasticity and how they could contribute to LFPs (Sinha & Narayanan, 2015, 2022), while also pointing to simulations with different numbers of gap junction in this context.

Frequency-Dependent Effects

The study demonstrates that gap junctional inputs suppress highfrequency EFP power due to membrane filtering. However, it could delve deeper into the implications of this for different brain rhythms, such as gamma or ripple oscillations.

We sincerely thank you for these insightful comments that we totally agree with. As it so happens, this manuscript forms the first part of a broader study where we explore the implications of gap junctions to ripple frequency oscillations. The ripple oscillations part of the work was presented as a poster in the Society for Neuroscience (SfN) annual meeting 2024 (Sirmaur & Narayanan, 2024). There, we simulate a neuropil made of hundreds of morphologically realistic neurons to assess the role of different synaptic inputs — excitatory, inhibitory, and gap junctional — and active dendrites to ripple frequency oscillations. We demonstrate there that the conclusions from single-neuron simulations in this current manuscript extend to a neuropil with several neurons, each receiving excitatory, inhibitory and gap-junctional inputs, especially with reference to high-frequency oscillations. Our networkbased analyses unveiled a dominant mediatory role of patterned inhibition in ripple generation, with recurrent excitations through chemical synapses and gap junctions in conjunction with return-current contributions from active dendrites playing regulatory roles in determining ripple characteristics (Sirmaur & Narayanan, 2024).

Our principal goal in this study, therefore, was to lay the single-neuron foundation for network analyses of the impact of gap junctions on LFPs. We are preparing the network part of the study, with a strong focus on ripple-frequency oscillations, for submission for peer review separately.

In a revised manuscript, we will mention the results from our SfN abstract with reference to network simulations and high-frequency oscillations, while also presenting discussions from other studies on the role of gap junctions in synchrony and LFP oscillations.

Visualization

Figures are dense and could benefit from more intuitive labeling and focused presentations. For example, isolating key differences between chemical and gap junctional inputs in distinct panels would improve clarity.

We thank you for this constructive suggestion. In the revised manuscript, we will enhance the visualization of the figures to ensure a clearer and more intuitive distinction between chemical synapses and gap junctions.

Contextual Relevance

The manuscript touches on how these findings relate to known physiological roles of gap junctions (e.g., in gamma rhythms) but does not explore this in depth. Stronger integration of the results into known neural network dynamics would enhance its impact.

We sincerely appreciate your valuable suggestion and acknowledge the importance of integrating our results into established neural network dynamics, particularly their implications for gamma rhythms. We will address this aspect more comprehensively in the revised version of our manuscript.

Reviewer #2 (Public review):

This computational work examines whether the inputs that neurons receive through electrical synapses (gap junctions) have different signatures in the extracellular local field potential (LFP) compared to inputs via chemical synapses. The authors present the results of a series of model simulations where either electric or chemical synapses targeting a single hippocampal pyramidal neuron are activated in various spatio-temporal patterns, and the resulting LFP in the vicinity of the cell is calculated and analyzed. The authors find several notable qualitative differences between the LFP patterns evoked by gap junctions vs. chemical synapses. For some of these findings, the authors demonstrate convincingly that the observed differences are explained by the electric vs. chemical nature of the input, and these results likely generalize to other cell types. However, in other cases, it remains plausible (or even likely) that the differences are caused, at least partly, by other factors (such as different intracellular voltage responses due to, e.g., the unequal strengths of the inputs). Furthermore, it was not immediately clear to me how the results could be applied to analyze more realistic situations where neurons receive partially synchronized excitatory and inhibitory inputs via chemical and electric synapses.

We gratefully thank you for your time and effort in rigorously assessing our manuscript, for the enthusiastic response, and the encouraging and thoughtful comments on our study. In what follows, we have provided point-by-point responses to the specific comments.

Strengths

The main strength of the paper is that it draws attention to the fact that inputs to a neuron via gap junctions are expected to give rise to a different extracellular electric field compared to inputs via chemical synapses, even if the intracellular effects of the two types of input are similar. This is because, unlike chemical synaptic inputs, inputs via gap junctions are not directly associated with transmembrane currents. This is a general result that holds independent of many details such as the cell types or neurotransmitters involved.

We gratefully thank you for the positive comments and the encouraging words about the novel contributions of our study. We are particularly thankful to you for your comment on the generality of our conclusions that hold for different cell types and neurotransmitters involved.

Another strength of the article is that the authors attempt to provide intuitive, non-technical explanations of most of their findings, which should make the paper readable also for non-expert audiences (including experimentalists).

We sincerely thank you for the positive comments about the readability of the paper.

Weaknesses

The most problematic aspect of the paper relates to the methodology for comparing the effects of electric vs. chemical synaptic inputs on the LFP. The authors seem to suggest that the primary cause of all the differences seen in the various simulation experiments is the different nature of the input, and particularly the difference between the transmembrane current evoked by chemical synapses and the gap junctional current that does not involve the extracellular space. However, this is clearly an oversimplification: since no real attempt is made to quantitatively match the two conditions that are compared (e.g., regarding the strength and temporal profile of the inputs), the differences seen can be due to factors other than the electric vs. chemical nature of synapses. In fact, if inputs were identical in all parameters other than the transmembrane vs. directly injected nature of the current, the intracellular voltage responses and, consequently, the currents through voltage-gated and leak currents would also be the same, and the LFPs would differ exactly by the contribution of the transmembrane current evoked by the chemical synapse. This is evidently not the case for any of the simulated comparisons presented, and the differences in the membrane potential response are rather striking in several cases (e.g., in the case of random inputs, there is only one action potential with gap junctions, but multiple action potentials with chemical synapses). Consequently, it remains unclear which observed differences are fundamental in the sense that they are directly related to the electric vs. chemical nature of the input, and which differences can be attributed to other factors such as differences in the strength and pattern of the inputs (and the resulting difference in the neuronal electric response).

We thank you for raising this important point. We would like to emphasize that our experimental design and analyses quantitatively account for the spatial distribution and temporal pattern of specific kinds of inputs that arrive through gap junctions and chemical synapses. We submit that our analyses quantitatively demonstrates that the fundamental difference between the gap junctional and chemical synaptic contributions to extracellular potentials is the absence of the direct transmembrane component from gap junctional inputs. We elucidate these points below:

(1) Spatial distribution: The inputs were distributed randomly across the basal dendrites, irrespective of whether they were through gap junctions or chemical synapses. For both chemical synapses and gap junctions, the inputs were of the same nature: excitatory.

(2) Different numbers of inputs: We have presented consistent results for both fewer and more gap junctions or chemical synapses in our analyses (see Figure 1 with 217 gap junctions or 245 chemical synapses and Supplementary Figure 2 with 99 gap junctions or 30 chemical synapses). Our fundamentally novel result that gap junctions onto active dendrites shape LFPs holds true irrespective of the relative density of gap junctions onto the neuron.

(3) Synchronous inputs (Figs. 1–3): For chemical synapses, the waveforms are in the shape of postsynaptic potentials. For gap junctional inputs, the waveforms are in the shape of postsynaptic potentials or dendritic spikes (to respect the active nature of inputs from the other cell). Here, the electrical response of the postsynaptic cell is identical irrespective of whether inputs arrive through gap junctions or chemical synapses: an action potential. We quantitatively matched the strengths such that the model generated a single action potential in response to synchronous inputs, irrespective of whether they arrived through chemical synaptic and gap junctional inputs. We mechanistically analyze the contributions of different cellular components and show that the direct transmembrane current in chemical synapses is the distinguishing factor that determines the dichotomy between the contributions of gap junctions vs. chemical synapses to extracellular potentials (Figs. 2–3). In a revised manuscript, we will show the intracellular responses to demonstrate that they are electrically matched.

(4) Random inputs (Fig. 4): For random inputs, we did not account for the number of action potentials that arrived, as the only observation we made here was with reference to the biphasic nature of the extracellular potentials with gap junctional inputs in the “No Sodium” scenario. We note that in the “No Sodium” scenario, the time-domain amplitudes were comparable for the field potentials (Fig. 4B, Fig. 4D).

(5) Rhythmic inputs (Fig. 5–8): For rhythmic inputs, please note that the intracellular and extracellular waveforms for every frequency are provided in supplementary figures S5– S11. It may be noted that the intracellular responses are comparable. In simulations for assessing spike-LFP comparison, we tuned the strengths to produce a single spike per cycle, ensuring fair comparison of LFPs with gap junctions vs. chemical synapses.

Taken together, we demonstrate through explicit sets of simulations and analyses that the differences in LFPs were not driven by the strength or patterns of the inputs but rather by the differences in direct transmembrane currents, which are subsequently reflected in the LFPs. In a revised manuscript, we will add a section to emphasize these points apart from providing intracellular traces for cases where they are not provided.

Some of the explanations offered for the effects of cellular manipulations on the LFP appear to be incomplete. More specifically, the authors observed that blocking leak channels significantly changed the shape of the LFP response to synchronous synaptic inputs - but only when electric inputs were used, and when sodium channels were intact. The authors seemed to attribute this phenomenon to a direct effect of leak currents on the extracellular potential - however, this appears unlikely both because it does not explain why blocking the leak conductance had no effect in the other cases, and because the leak current is several orders of magnitude smaller than the spike-generating currents that make the largest contributions to the LFP. An indirect effect mediated by interactions of the leak current with some voltage-gated currents appears to be the most likely explanation, but identifying the exact mechanism would require further simulation experiments and/or a detailed analysis of intracellular currents and the membrane potential in time and space.

We thank you for raising this important question. Leak channels were among the several contributors to the positive deflection observed in LFPs associated with gap junctions. This effect was present not only in gap junctional models with intact sodium conductance but also in the no-sodium model, where the amplitude of the positive deflection was reduced across other models as well (Fig. 2F, I). Furthermore, even in the absence of leak conductance, a small positive deflection was still observed (Fig. 2F), leading us to further investigate other transmembrane currents over time and across spatial locations, from the proximal to the distal dendritic ends relative to the soma (Fig. 3D). We had observed that the dominant contributor in the case of chemical synapses was the inward synaptic current (Fig. 3A), whereas for gap junctions, the primary contributors were leak conductance along with other outward currents, such as potassium and HCN currents (Fig. 3D). Together, the direct transmembrane component of chemical synapses provides a dominant contribution to extracellular potentials. This dominance translates to differences in the relative contributions of indirect currents (including leak currents) to extracellular potentials associated chemical synaptic vs. gap junctional inputs. Our analyses of the exact ionic mechanisms (Fig. 3) demonstrates the involvement of several ion channels contributing to the indirect component in either scenario.

In every simulation experiment in this study, inputs through electric synapses are modeled as intracellular current injections of pre-determined amplitude and time course based on the sampled dendritic voltage of potential synaptic partners. This is a major simplification that may have a significant impact on the results. First, the current through gap junctions depends on the voltage difference between the two connected cellular compartments and is thus sensitive to the membrane potential of the cell that is treated as the neuron "receiving" the input in this study (although, strictly speaking, there is no pre- or postsynaptic neuron in interactions mediated by gap junctions). This dependence on the membrane potential of the target neuron is completely missing here. A related second point is that gap junctions also change the apparent membrane resistance of the neurons they connect, effectively acting as additional shunting (or leak) conductance in the relevant compartments. This effect is completely missed by treating gap junctions as pure current sources.

We thank you for raising this important point. We agree with the analyses presented by the reviewer on the importance of network simulations and bidirectional gap junctions that respect the voltages in both neurons. However, the complexities of LFP modeling precludes modeling of networks of morphologically realistic models with patterns of stimulations occurring across the dendritic tree. LFP modeling studies predominantly uses “post-synaptic” currents to analyze the impact of different patterns of inputs arriving on to a neuron, even when chemical synaptic inputs are considered. Explicitly, individual neurons are separately simulated with different patterns of synaptic inputs, the transmembrane current at different locations recorded, and the extracellular potential is then computed using line source approximation (Buzsaki et al., 2012; Gold et al., 2006; Halnes et al., 2024; Ness et al., 2018; Reimann et al., 2013; Schomburg et al., 2012; Sinha & Narayanan, 2015, 2022). Even in scenarios where a network is analyzed, a hybrid approach involving the outputs of a pointneuron-based network being coupled to an independent morphologically realistic neuronal model is employed (Hagen et al., 2016; Martinez-Canada et al., 2021; Mazzoni et al., 2015). Given the complexities associated with the computation of electrode potentials arising as a distance-weighted summation of several transmembrane currents, these simplifications becomes essential.

Our approach models gap junctional currents in a similar way as the other model incorporate synaptic currents in LFP modeling (Buzsaki et al., 2012; Gold et al., 2006; Halnes et al., 2024; Ness et al., 2018; Reimann et al., 2013; Schomburg et al., 2012; Sinha & Narayanan, 2015, 2022). As gap junctions are typically implemented as resistors from the other neuronal compartment, we accounted for gap-junctional variability in our model by randomizing the scaling-factors and the exact waveforms that arrive through individual gap junctions at specific locations. Thus, the inputs were not pre-determined by “pre” neurons. Instead, the recorded voltages from potential synaptic partner neurons were randomized across locations and scaled using factors at the dendrites before being injected into the target neuron (Supplementary Fig. S1). While incorporating a network of interconnected neurons is indeed important, we utilized biophysical, morphologically realistic CA1 neuron model with different sets of input patterns to model LFPs, which were derived from the total transmembrane currents across all compartments of the multi-compartmental neuron model. Given the complexity of this approach, adding further network-level interactions or pre-post connections would have been computationally demanding.

In a revised manuscript, we will introduce the general methodology used in LFP modeling studies to introduce synaptic currents. We will emphasize that our study extends this approach to modeling gap junctional inputs, while also highlighting randomization of locations and the scaling process in assigning gap junctional synaptic strengths.

One prominent claim of the article that is emphasized even in the abstract is that HCN channels mediate an outward current in certain cases. Although this statement is technically correct, there are two reasons why I do not consider this a major finding of the paper. First, as the authors acknowledge, this is a trivial consequence of the relatively slow kinetics of HCN channels: when at least some of the channels are open, any input that is sufficiently fast and strong to take the membrane potential across the reversal potential of the channel will lead to the reversal of the polarity of the current. This effect is quite generic and well-known and is by no means specific to gap junctional inputs or even HCN channels. Second, and perhaps more importantly, the functional consequence of this reversed current through HCN channels is likely to be negligible. As clearly shown in Supplementary Figure S3, the HCN current becomes outward only for an extremely short time period during the action potential, which is also a period when several other currents are also active and likely dominant due to their much higher conductances. I also note that several of these relevant facts remain hidden in Figure 3, both because of its focus on peak values, and because of the radically different units on the vertical axes of the current plots.

We thank you for raising this point and agree with you on every point. Please note that we do not assert that the outward HCN currents are exclusively associated with gap junctional inputs. Rather, our results show that synchronous inputs generate outward HCN currents in both chemical synapses (Fig. 3B; positive/outward HCN currents, except in the no sodium or leak model) and gap junctions (Fig. 3D; positive/outward HCN currents). We emphasized this in the case of gap junctions because, in the absence of inward synaptic currents, HCN (acting as outward currents with synchronous inputs) contributed to the positive deflection observed in the LFPs. While HCN would also contribute in the case of chemical synapses, its effect was negligible due to the presence of large inward synaptic currents. Since LFPs reflect the collective total transmembrane currents, the dominant contributors differ between these two scenarios, which we aimed to highlight. Since HCN exhibited outward currents in our synchronous input simulations, we have elaborated on this mechanism in the supplementary figure (Fig. S3). Our intention was not to emphasize this effect for only one synaptic mode but rather to highlight HCN's contribution to the positive deflection as one of the contributing factors.

We agree that HCN currents are relatively small in magnitude; therefore, our conclusions were based on HCN being one of the several contributing factors. Leak conductance and other outward conductances, including HCN currents (Fig. 3D), collectively contribute to the positive deflections observed in the case of gap junctional synchronous inputs.

We will ensure that we will account for all the points appropriately in a revised manuscript.

Finally, I missed an appropriate validation of the neuronal model used, and also the characterization of the effects of the in silico manipulations used on the basic behavior of the model. As far as I understand, the model in its current form has not been used in other studies. If this is the case, it would be important to demonstrate convincingly through (preferably quantitative) comparisons with experimental data using different protocols that the model captures the physiological behavior of at least the relevant compartments (in this case, the dendrites and the soma) of hippocampal pyramidal neurons sufficiently well that the results of the modeling study are relevant to the real biological system. In addition, the correct interpretation of various manipulations of the model would be strongly facilitated by investigating and discussing how the physiological properties of the model neuron are affected by these alterations.

We thank you for raising this important point. The CA1 pyramidal neuronal model used in this study is built with ion-channel models derived from biophysical and electrophysiological recordings from these cells. As mentioned in the Methods section “Dynamics and distribution of active channels” and Supplementary Table S1, models for individual channels, their gating kinetics, and channel distributions across the somatodendritic arbor (wherever known) are all derived from their physiological equivalents. Importantly, these values were derived from previously validated models from the laboratory, which contain these very ion channel models and the exact same morphology (Roy & Narayanan, 2021). Please compare Supplementary Table S1 with the Table 1 from (Roy & Narayanan, 2021). Please note that this model was validated against several physiological measurements along the somatodendritic axis (Fig. 1 of (Roy & Narayanan, 2021)).

In a revised manuscript, we will explicitly mention this while also mentioning the different physiological properties that were used for the validation process from (Roy & Narayanan, 2021). We sincerely regret not mentioning these details in the current version of our manuscript.

We will fix these in a revised version of the manuscript.

References

Bedner, P., Steinhauser, C., & Theis, M. (2012). Functional redundancy and compensation among members of gap junction protein families? Biochim Biophys Acta, 1818(8), 1971-1984. https://doi.org/10.1016/j.bbamem.2011.10.016

Behrens, C. J., Ul Haq, R., Liotta, A., Anderson, M. L., & Heinemann, U. (2011). Nonspecific effects of the gap junction blocker mefloquine on fast hippocampal network oscillations in the adult rat in vitro. Neuroscience, 192, 11-19. https://doi.org/10.1016/j.neuroscience.2011.07.015

Buzsaki, G., Anastassiou, C. A., & Koch, C. (2012). The origin of extracellular fields and currents--EEG, ECoG, LFP and spikes. Nat Rev Neurosci, 13(6), 407-420. https://doi.org/10.1038/nrn3241

Einevoll, G. T., Destexhe, A., Diesmann, M., Grun, S., Jirsa, V., de Kamps, M., Migliore, M., Ness, T. V., Plesser, H. E., & Schurmann, F. (2019). The Scientific Case for Brain Simulations. Neuron, 102(4), 735-744. https://doi.org/10.1016/j.neuron.2019.03.027

Gold, C., Henze, D. A., Koch, C., & Buzsaki, G. (2006). On the origin of the extracellular action potential waveform: A modeling study. J Neurophysiol, 95(5), 3113-3128. https://doi.org/10.1152/jn.00979.2005

Hagen, E., Dahmen, D., Stavrinou, M. L., Linden, H., Tetzlaff, T., van Albada, S. J., Grun, S., Diesmann, M., & Einevoll, G. T. (2016). Hybrid Scheme for Modeling Local Field Potentials from Point-Neuron Networks. Cereb Cortex, 26(12), 4461-4496. https://doi.org/10.1093/cercor/bhw237

Halnes, G., Ness, T. V., Næss, S., Hagen, E., Pettersen, K. H., & Einevoll, G. T. (2024). Electric Brain Signals: Foundations and Applications of Biophysical Modeling. Cambridge University Press. https://doi.org/DOI: 10.1017/9781009039826

Lo, C. W. (1999). Genes, gene knockouts, and mutations in the analysis of gap junctions. Dev Genet, 24(1-2), 1-4. https://doi.org/10.1002/(SICI)1520-6408(1999)24:1/2<1::AIDDVG1>3.0.CO;2-U

Martinez-Canada, P., Ness, T. V., Einevoll, G. T., Fellin, T., & Panzeri, S. (2021). Computation of the electroencephalogram (EEG) from network models of point neurons. PLoS Comput Biol, 17(4), e1008893. https://doi.org/10.1371/journal.pcbi.1008893

Mazzoni, A., Linden, H., Cuntz, H., Lansner, A., Panzeri, S., & Einevoll, G. T. (2015). Computing the Local Field Potential (LFP) from Integrate-and-Fire Network Models. PLoS Comput Biol, 11(12), e1004584. https://doi.org/10.1371/journal.pcbi.1004584

Ness, T. V., Remme, M. W. H., & Einevoll, G. T. (2018). h-Type Membrane Current Shapes the Local Field Potential from Populations of Pyramidal Neurons. J Neurosci, 38(26), 6011-6024. https://doi.org/10.1523/jneurosci.3278-17.2018

Reimann, M. W., Anastassiou, C. A., Perin, R., Hill, S. L., Markram, H., & Koch, C. (2013). A biophysically detailed model of neocortical local field potentials predicts the critical role of active membrane currents. Neuron, 79(2), 375-390. https://doi.org/10.1016/j.neuron.2013.05.023

Rouach, N., Segal, M., Koulakoff, A., Giaume, C., & Avignone, E. (2003). Carbenoxolone blockade of neuronal network activity in culture is not mediated by an action on gap junctions. Journal of Physiology, 553(Pt 3), 729-745. https://doi.org/10.1113/jphysiol.2003.053439

Roy, A., & Narayanan, R. (2021). Spatial information transfer in hippocampal place cells depends on trial-to-trial variability, symmetry of place-field firing, and biophysical heterogeneities. Neural Netw, 142, 636-660. https://doi.org/10.1016/j.neunet.2021.07.026

Schomburg, E. W., Anastassiou, C. A., Buzsaki, G., & Koch, C. (2012). The spiking component of oscillatory extracellular potentials in the rat hippocampus. J Neurosci, 32(34), 11798-11811. https://doi.org/10.1523/JNEUROSCI.0656-12.2012

Sinha, M., & Narayanan, R. (2015). HCN channels enhance spike phase coherence and regulate the phase of spikes and LFPs in the theta-frequency range. Proc Natl Acad Sci U S A, 112(17), E2207-2216. https://doi.org/10.1073/pnas.1419017112

Sinha, M., & Narayanan, R. (2022). Active Dendrites and Local Field Potentials: Biophysical Mechanisms and Computational Explorations. Neuroscience, 489, 111-142. https://doi.org/10.1016/j.neuroscience.2021.08.035

Sirmaur, R., & Narayanan, R. (2024). Distinct extracellular signatures of chemical and electrical synapses impinging on active dendrites differentially contribute to ripple-frequency oscillations. Society for Neuroscience annual meeting (https://www.abstractsonline.com/pp8/?_gl=1*1bxo7m*_gcl_au*MTc5MTQ0NjE0NC4xNzI3MDcwOTMw*_ga*MTMxMTE5OTcyMy4xNzI3MDcwOTMx*_ga_T09K 3Q2WDN*MTcyNzA3MDkzMS4xLjEuMTcyNzA3MDkzNy41NC4wLjA.#!/20433/ presentation/13949), Chicago, USA.

Szarka, G., Balogh, M., Tengolics, A. J., Ganczer, A., Volgyi, B., & Kovacs-Oller, T. (2021). The role of gap junctions in cell death and neuromodulation in the retina. Neural Regen Res, 16(10), 1911-1920. https://doi.org/10.4103/1673-5374.308069

eLife Assessment

This study describes the impact of mycobacterial genetic diversity on host-infection phenotypes by assessing the effect of different M. tuberculosis lineages on granulomatous inflammation using a 3D in vitro granuloma model. Despite being descriptive and showing mostly correlative relationships, the findings are useful and provide some solid support regarding the functional impact of M. tuberculosis's natural diversity on host-pathogen interactions. The study will interest researchers working on mycobacteria and motivate future studies to understand how genetic diversity influences virulence and immunity outcomes.

Reviewer #2 (Public review):

Summary:

This manuscript reports a comparison of microbial traits and host response traits in a laboratory model of infected granuloma using Mtb strains from different lineages. The authors report increased bacillary growth and granuloma formation, inversely associated with T cell activation that is characterized by CXCL9, granzyme B and TNF expression. They therefore infer that these T cell responses are likely to be host-protective and that the greater virulence of modern Mtb lineages may be driven by their ability to avoid triggering these responses.

Strengths:

The comparison of multiple Mtb lineages in a granuloma model that enables evaluation of the potential role of multiple host cells in Mtb control, offers a valuable experimental approach to study the biological mechanisms that underpin differential virulence of Mtb lineages that has been previously reported in clinical and epidemiological studies.

Weaknesses:

The study is rather limited to descriptive observations, and lacks experiments to test causal relationships between host and pathogen traits. Some of the presentation of the data are difficult to interpret, and some conclusions are not adequately supported by the data.

Comments on revisions:

The authors have addressed my previous comments with appropriate revisions and explanations.

Reviewer #3 (Public review):

Arbués and colleagues describe the impact of mycobacterial genetic diversity on host-infection phenotypes. The authors evaluate Mtb infection and contextualize host-responses, bacterial growth and metabolic transitioning in vitro using their previously established model of blood-derived, primary-human-cells cultured within a collagen/fibronectin matrix. They seek to demonstrate the effectiveness of the model in determining mycobacterial strain specific granuloma-dependent host-pathogen interactions.

Understanding the way mycobacterial genetic diversity impacts granuloma biology in tuberculosis is an important goal. One of this works strengths is the use of primary human cells and two constituents of pulmonary extracellular matrix to model Mtb infection. The authors and others have previously shown that Mtb infected PBMC aggregates share important characteristics with early pulmonary TB granulomas. Use of multiple genetically distinct strains of Mtb defines this work and further bolsters it potential impact. However, the study is not comprehensive as lineages 6 and 7 are not tested. Experiments are primarily descriptive, and the methodologies are conventional. Correlative relationships are the manuscripts focus and effect sizes are generally small.

The main aim of this work is to extend an in vitro granuloma model to the study of a large collection of well characterized, genetically diverse representatives of the mycobacterium tuberculosis complex (MTBC). I believe that they accomplish that aim. The work does investigate MTBC infection of aggregated PBMCs using three strains each of Mtb lineages 1-5 and H37Rv, which is not a trivial undertaking. The experimental aims are to show that MTBC genetic diversity impacts growth and dormancy of granuloma bound bacteria and, the host responses of granulomatous aggregation as well as macrophage apoptosis, lymphocyte activation and soluble mediator release within granulomas. The methodologies employed are sufficient to test most of these aims. The authors conclusions regarding their results are mostly supported by the data. The conclusion that lineage impacts growth within granulomas is likely true and the data as presented reflect such a relationship. Their conclusions regarding lineage's impact on dormancy are partially supported, as their findings demonstrate that assays for dormancy identify strain-specific metabolic changes in the bacteria consistent with a dormancy-like state but also identify replicating bacteria as being dormant. The data strongly supports the impact of mycobacterial genetic diversity on a spectrum of granulomatous responses in their model system. Those findings are a highlight of the publication. The data further supports the idea that strain diversity impacts macrophage apoptosis but a relationship of apoptosis to the granulomatous response is not effectively evaluated. The association of lymphocyte activation with reduced mycobacterial growth as an aspect of granulomas is well documented in the literature and a negative correlation between T cell activation and growth is supported by the authors results. Their data also support the conclusion that soluble mediator production by PBMCs is different based on the infecting strain of mycobacteria and that IL1b modulates aggregate phenotypes in their model.

The authors contribute some valuable insights, particularly in Figure 3. Their model is higher echelon relative to others in the field, but I don't believe that it possesses all the components necessary to replicate formation of mycobacterial granulomas in vivo. That being said, their identification of donor-dependent aggregation phenotypes by mycobacterial strain has the potential to enable future investigations of human and mycobacterial genetic components that are involved in the formation of TB granulomas.

Author response:

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

Reviewer #2:

The authors indicated that they had added coefficients of variation for within-lineage heterogeneity (line 93), but I can't seem to find this.

The coefficients of variation were indeed included as suggested, and can be found in lines 94-96 of the current revised version of the manuscript. The sentence states: “Nevertheless, substantial intra-lineage heterogeneity could be observed, particularly within L1 and L2 (coefficients of variation 84.4% [L1] and 66.0% [L2] vs. 32.6% [L3], 34.6% [L4] and 31.9% [L5]).”

They were unable to address my question on the impact of T-cell depletion from PBMC on bacterial growth? Their discussion should include that this experimental limitation means that they are unable to test cause and effect for the relationship between T cell proliferation and bacterial growth.

As recommended, this experimental limitation is now included in the discussion in lines 344-346.

Reviewer #3:

EM:

Based on the authors lack of resources, I don't believe that electron microscopy experiments should be required for this publication. However, it should be noted that EM is performed on fixed samples such that implementation of those protocols as it relates to bio-safety is no more demanding than the preparation of samples for other common assays performed outside of the BSL3.

We appreciate your understanding regarding our lack of resources to carry out the EM experiments, although we recognize the possibility of them being performed on BSL3 samples.

Granuloma score:

From the author comments and the manuscript's text, it appears that the "granuloma score" is an attempt at quantitation of PBMC organization. Where every component of the metric [(mean area / mean aspect ratio) / mean n ] is a visual facet of the relative integration of PBMCs into a more organized aggregate. The area and number (n) of aggregates both address regional coalescence of the total number of PBMCs added into the matrix. Whereas the aspect ratio component is an indicator of uniformity of the PBMCs that have been assigned to an individual aggregate. Perhaps another roundness estimation would have been a more precise, but aspect ratio seems fine for their assay. Considering these factors and the author's contention that the aggregates making up (n) are granulomas, the name "granuloma score" is inaccurate and a more appropriate title would be "aggregate organization score" or "aggregate organization index".

Thank you for the suggested alternative terminology, the term “granuloma score” has been substituted with “aggregate formation score” throughout the manuscript.

Dormancy:

In the manuscript, the authors should explicitly reference the validation studies which demonstrate induction of the DosR regulon in the model, lest their previously generated and conducted studies go unappreciated by a broader audience. In the title of that previous work (PMID: 32069329) this group used the designation "dormant-like" to describe the state observed in bacteria within their in vitro granuloma model system, as they also do in LINE 124. This term or a variation of it should be exchanged for dormant/dormancy throughout the manuscript when referring to observations in the model bacteria. It is a more precise description. Further, "dormant-like" allows the latitude to refer to actively growing bacteria in the context of dormancy without running the risk of putting forth confusing or potentially erroneous assertions.

As recommended, the suffix “-like” has been added to the designation “dormant” when referring to the bacterial phenotype induced in the model. In addition, de induction of the DosR regulon in the model is now mentioned in line 116 and the reference to Kapoor’s work that originally demonstrated it by qPCR included.

PBMC aggregation:

I would like to make the authors aware that in well vetted models, cell aggregation as a function of infection does not typically occur in PBMCs on tissue culture plates until day 6 post infection (PMID: 25691598, Fig 2). Further, this group's own published protocol for the model under consideration in this manuscript (PMID: 33659472, Fig1) explicitly states that "Formation of granuloma like structures can be observed after 7-8 days", the implication being that prior to 7 days granuloma like structures cannot be observed reliably. Regardless, it seems evident that the authors will not be conducting additional experiments for this publication, which I find acceptable. However a proper negative control would certainly strengthen evidence for the association of strain specific bacterial and host responses with the granulomatous response in this model.

We had interpreted the reviewer’s previous comment regarding PBMC aggregation as referring to a different experimental model rather than a matter of timing. Since many other studies have previously assessed the impact of strain/lineage variability in macrophage responses, in this work we decided to focus on later time points and we did include uninfected as a negative control. Nonetheless, we agree it would be indeed very interesting to additionally evaluate monocyte/macrophage early responses and we will take it into account for future studies.

Use of antiquated terminology:

I can appreciate the desire to establish continuity between publications by using the same abbreviation for TNF but it will come at a cost. Using outdated terms in general makes people more dismissive of the work. Perhaps something to consider.

Since this seems an important issue to the reviewer, we have replaced the term TNF-a with TNF throughout the manuscript.

eLife Assessment

The study by Chen and Phillips provides evidence for a dynamic switch in the small RNA repertoire of the Argonaute protein NRDE-3 during embryogenesis in C. elegans. The work is supported by convincing experimental data, shedding light on RNA regulation during development. While the functional relevance of this process warrants further investigation, this study provides valuable insights into small RNA pathways with broader implications for developmental biology and gene regulation in other systems.

Reviewer #1 (Public review):

Summary:

Chen and Phillips describe the dynamic appearance of cytoplasmic granules during embryogenesis analogous to SIMR germ granules, and distinct from CSR-1-containing granules, in the C. elegans germline. They show that the nuclear Argonaute NRDE-3, when mutated to abrogate small RNA binding, or in specific genetic mutants, partially colocalizes to these granules along with other RNAi factors, such as SIMR-1, ENRI-2, RDE-3, and RRF-1. Furthermore, NRDE-3 RIP-seq analysis in early vs. late embryos is used to conclude that NRDE-3 binds CSR-1-dependent 22G RNAs in early embryos and ERGO-1-dependent 22G RNAs in late embryos. These data lead to their model that NRDE-3 undergoes small RNA substrate "switching" that occurs in these embryonic SIMR granules and functions to silence two distinct sets of target transcripts - maternal, CSR-1 targeted mRNAs in early embryos and duplicated genes and repeat elements in late embryos.

Strengths:

The identification and function of small RNA-related granules during embryogenesis is a poorly understood area and this study will provide the impetus for future studies on the identification and potential functional compartmentalization of small RNA pathways and machinery during embryogenesis.

Weaknesses:

(1) The authors acknowledge the following issue that loss of SIMR granules have no significant impact on NRDE-3 small RNA loading weakens the functional relevance of these structures. However, this point is clearly discussed and, as they note in their Discussion, it is entirely possible that these embryonic granules may be "incidental condensates."

Reviewer #2 (Public review):

Summary:

NRDE-3 is a nuclear WAGO-clade Argonaute that, in somatic cells, binds small RNAs amplified in response to the ERGO-class 26G RNAs that target repetitive sequences. This manuscript reports that, in the germline and early embryos, NRDE-3 interacts with a different set of small RNAs that target mRNAs. This class of small RNAs were previously shown to bind to a different WAGO-clade Argonaute called CSR-1, which is cytoplasmic unlike nuclear NRDE-3. The switch in NRDE-3 specificity parallels recent findings in Ascaris where the Ascaris NRDE homolog was shown to switch from sRNAs that target repetitive sequences to CSR-class sRNAs that target mRNAs.

The manuscript also correlates the change in NRDE-3 specificity with the appearance in embryos of cytoplasmic condensates that accumulate SIMR-1, a scaffolding protein that the authors previously implicated in sRNA loading for a different nuclear Argonaute HRDE-1. By analogy, and through a set of corelative evidence, the authors argue that SIMR foci arise in embryogenesis to facilitate the change in NRDE-3 small RNA repertoire. The paper presents lots of data that beautifully documents the appearance and composition of the embryonic SIMR-1 foci, including evidence that a mutated NRDE-3 that cannot bind sRNAs accumulate in SIMR-1 foci in SIMR-1-dependent fashion.

Reviewer #3 (Public review):

Summary:

Chen and Phillips present intriguing work that extends our view on the C. elegans small RNA network significantly. While the precise findings are rather C. elegans specific there are also messages for the broader field, most notably the switching of small RNA populations bound to an argonaute, and RNA granules behavior depending on developmental stage. The work also starts to shed more light on the still poorly understood role of the CSR-1 argonaute protein and supports its role in the decay of maternal transcripts. Overall, the work is of excellent quality, and the messages have a significant impact.

Strengths:

Compelling evidence for major shift in activities of an argonaute protein during development, and implications for how small RNAs affect early development. Very balanced and thoughtful discussion.

Weaknesses:

The switch between maternal and zygotic NRDE-3 remains unaddressed

Author response:

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

Reviewer #1 (Public review):

Summary:

Chen and Phillips describe the dynamic appearance of cytoplasmic granules during embryogenesis analogous to SIMR germ granules, and distinct from CSR-1-containing granules, in the C. elegans germline. They show that the nuclear Argonaute NRDE-3, when mutated to abrogate small RNA binding, or in specific genetic mutants, partially colocalizes to these granules along with other RNAi factors, such as SIMR-1, ENRI-2, RDE-3, and RRF-1. Furthermore, NRDE-3 RIP-seq analysis in early vs. late embryos is used to conclude that NRDE-3 binds CSR-1-dependent 22G RNAs in early embryos and ERGO-1dependent 22G RNAs in late embryos. These data lead to their model that NRDE-3 undergoes small RNA substrate "switching" that occurs in these embryonic SIMR granules and functions to silence two distinct sets of target transcripts - maternal, CSR-1 targeted mRNAs in early embryos and duplicated genes and repeat elements in late embryos.

Strengths:

The identification and function of small RNA-related granules during embryogenesis is a poorly understood area and this study will provide the impetus for future studies on the identification and potential functional compartmentalization of small RNA pathways and machinery during embryogenesis.

Weaknesses:

(1) While the authors acknowledge the following issue, their finding that loss of SIMR granules has no apparent impact on NRDE-3 small RNA loading puts the functional relevance of these structures into question. As they note in their Discussion, it is entirely possible that these embryonic granules may be "incidental condensates." It would be very welcomed if the authors could include some evidence that these SIMR granules have some function; for example, does the loss of these SIMR granules have an effect on CSR-1 targets in early embryos and ERGO-1-dependent targets in late embryos?

We appreciate reviewer 1’s concern that we do not provide enough evidence for the function of the SIMR granules. As suggested, we examined the NRDE-3 bound small RNAs more deeply, and we do observe a slight but significant increased CSR-class 22G-RNAs binding to NRDE-3 in late embryos of simr-1 and enri-2 mutants (see below, right). We hypothesize that this result could be due to a slower switch from CSR to ERGO 22G-RNAs in the absence of SIMR granules. We added these data to Figure 6G.

(2) The analysis of small RNA class "switching" requires some clarification. The authors re-define ERGO1-dependent targets in this study to arrive at a very limited set of genes and their justification for doing this is not convincing. What happens if the published set of ERGO-1 targets is used? 

As we mentioned in the manuscript, we initially attempted to use the previously defined ERGO targets. However, the major concern is fewer than half the genes classified as ERGO targets by Manage et al. and Fischer et al. overlap with one another (Figure 6—figure supplement 1D and below). We reason this might because the gene sets were defined as genes that lose small RNAs in various ERGO pathway mutants and because different criteria were used to define the lists as discussed in the manuscript (lines 471-476). As a result, some of the previously defined ERGO target genes may actually be indirect targets of the pathway. Here we focus on genes targeted by small RNAs enriched in an ERGO pathway Argonaute IP, which should be more specific.

In this manuscript, we are interested specifically in the ERGO targets bound by NRDE-3, thus we utilized the IP-small RNA sequencing data from young adult animals (Seroussi et al, 2023), to define a new ERGO list. We are confident about this list because 1) Most of our new ERGO genes overlap with the overlap between ERGO-Manage and ERGO-Fischer list (see Figure 6—figure supplement 1D in our manuscript and below). 2) We observed the most significant decrease of small RNA levels and increase of mRNA levels in the nrde-3 mutants using our newly defined list (see Figure 6—figure supplement 1E-F in our manuscript).

To further address reviewer 1’s concern about whether the data would look significantly different when using the ERGO-Manage and ERGO-Fischer lists, we made new scatter plots shown in Author response image 1 panels A-C below (ERGO-Manage – purple, ERGO-Fischer- yellow, and the overlap - yellow with purple ring). We found that the small switching pattern of NRDE-3 is consistent with our newly defined list, particularly if we look at the overlap of ERGO-Manage and ERGO-Fischer list (Author response image 1 panels D-F below, red).

Author response image 1.

Further, the NRDE-3 RIP-seq data is used to conclude that NRDE-3 predominantly binds CSR-1 class 22G RNAs in early embryos, while ERGO-1-dependent 22G RNAs are enriched in late embryos. a) The relative ratios of each class of small RNAs are given in terms of unique targets. What is the total abundance of sequenced reads of each class in the NRDE-3 IPs? 

To address the reviewer’s question about the total abundance of sequenced reads of each class in the NRDE-3 IPs: Author response image 2 panel A-B below show the total RPM of CSR and ERGO class sRNAs in inputs and IPs at different stages. Focusing on late embryos, the total abundance of ERGO-dependent sRNAs is similar to CSR-class sRNAs in input, while much higher in IP, indicating an enrichment of ERGO-dependent 22G-RNAs in NRDE-3 consistent with our log2FC (IP vs input) in Figure 6B. This data supports our conclusion that NRDE-3 preferentially binds to ERGO targets in late embryos.

Author response image 2.

b) The "switching" model is problematic given that even in late embryos, the majority of 22G RNAs bound by NRDE-3 is the CSR-1 class (Figure 5D). 

It is important to keep in mind the difference in the total number of CSR target genes (3834) and ERGO target genes (119).  The pie charts shown in Figure 6D are looking at the total proportion of the genes enriched in the NRDE-3 IP that are CSR or ERGO targets. For the NRDE-3 IP in late embryos, that would be 70/119 (58.8%) of ERGO targets are enriched, while 172/3834 (4.5%) of CSR targets are enriched. These data are also supported by the RPM graphs shown in Author response image 2 panels A-B above, which show that the majority of the small RNA bound by NRDE-3 in late embryos are ERGO targets. Nonetheless, NRDE-3 still binds to some CSR targets shown as Figure 6D and panel B, which may be because the amount of CSR-class 22G-RNAs is reduced gradually across embryonic development as the maternally-deposited NRDE-3 loaded with CSR-class 22G-RNAs is diluted by newly transcribed NRDE-3 loaded with ERGOdependent 22G-RNAs (lines 857-862). 

c) A major difference between NRDE-3 small RNA binding in eri-1 and simr-1 mutants appears to be that NRDE-3 robustly binds CSR-1 22G RNAs in eri-1 but not in simr-1 in late embryos. This result should be better discussed.

In the eri-1 mutant, we hypothesize that NRDE-3 robustly binds CSR-class 22G-RNAs because ERGOclass 22G-RNAs are not synthesized during mid-embryogenesis, so either NRDE-3 is unloaded (in granule at 100-cell stage in Figure 2A) or mis-loaded with CSR-class 22G-RNAs (in the nucleus at 100cell stage in Figure 2A). We don’t have a robust method to address the proportion of loaded vs. unloaded NRDE-3 so it is difficult to address the degree to which NRDE-3 is misloaded in the eri-1 mutant. In the simr-1 mutant, both classes of small RNAs are present and NRDE-3 is still preferentially loaded with ERGO-dependent 22G-RNAs, though we do see a subtle increase in association with CSR-class 22GRNAs. These data could suggest a less efficient loading of NRDE-3 with ERGO-dependent 22G-RNAs, but we would need more precise methods to address the loading dynamics in the simr-1 mutant.

(3) Ultimately, if the switching is functionally important, then its impact should be observed in the expression of their targets. RNA-seq or RT-qPCR of select CSR-1 and ERGO-1 targets should be assessed in nrde-3 mutants during early vs late embryogenesis.

The function of NRDE-3 at ERGO targets has been well studied (Guang et al, 2008) and is also assessed in our H3K9me3 ChIP-seq analysis in Figure 7E where, in mixed staged embryos, H3K9me3 level on ERGO targets (labeled as ‘NRDE-3 targets in young adults’) is reduced significantly in the nrde-3 mutant.

To understand the function of NRDE-3 binding on CSR targets in early embryos, we attempted to do RTqPCR, smFISH, and anti-H3K9me3 CUT&Tag-seq on early embryos, and we either failed to obtain enough signal or failed to detect any significant difference (data not shown). We additionally tested the possibility that NRDE-3 functions with CSR-class 22G-RNAs in oocytes. We present new data showing that NRDE-3 represses RNA Pol II in oocytes to promote global transcriptional repression at the oocyteto-embryo transition, we now included these data in Figure 8. 

Reviewer #2 (Public review):

Summary:

NRDE-3 is a nuclear WAGO-clade Argonaute that, in somatic cells, binds small RNAs amplified in response to the ERGO-class 26G RNAs that target repetitive sequences. This manuscript reports that, in the germline and early embryos, NRDE-3 interacts with a different set of small RNAs that target mRNAs. This class of small RNAs was previously shown to bind to a different WAGO-clade Argonaute called CSR1, which is cytoplasmic, unlike nuclear NRDE-3. The switch in NRDE-3 specificity parallels recent findings in Ascaris where the Ascaris NRDE homolog was shown to switch from sRNAs that target repetitive sequences to CSR-class sRNAs that target mRNAs.

The manuscript also correlates the change in NRDE-3 specificity with the appearance in embryos of cytoplasmic condensates that accumulate SIMR-1, a scaffolding protein that the authors previously implicated in sRNA loading for a different nuclear Argonaute HRDE-1. By analogy, and through a set of corelative evidence, the authors argue that SIMR foci arise in embryogenesis to facilitate the change in NRDE-3 small RNA repertoire. The paper presents lots of data that beautifully documents the appearance and composition of the embryonic SIMR-1 foci, including evidence that a mutated NRDE-3 that cannot bind sRNAs accumulates in SIMR-1 foci in a SIMR-1-dependent fashion.

Weaknesses:

The genetic evidence, however, does not support a requirement for SIMR-1 foci: the authors detected no defect in NRDE-3 sRNA loading in simr-1 mutants. Although the authors acknowledge this negative result in the discussion, they still argue for a model (Figure 7) that is not supported by genetic data. My main suggestion is that the authors give equal consideration to other models - see below for specifics.

We appreciate reviewer 2’s comments on the genetic evidence for the function of SIMR foci.  A similar concern was also brought up by reviewer 1. By re-examining our sequencing data, we found that there is a modest but significant increase in NRDE-3 association with CSR-class sRNAs in simr-1 and enri-2 mutants in late embryos. We believe that this data supports our model that SIMR-1 and ENRI-2 are required for an efficient switch of NRDE-3 bound small RNAs. Please refer our response to the reviewer 1 - point (1), and Figure 6G in the updated manuscript. 

Reviewer #3 (Public review):

Summary:

Chen and Phillips present intriguing work that extends our view on the C. elegans small RNA network significantly. While the precise findings are rather C. elegans specific there are also messages for the broader field, most notably the switching of small RNA populations bound to an argonaute, and RNA granules behavior depending on developmental stage. The work also starts to shed more light on the still poorly understood role of the CSR-1 argonaute protein and supports its role in the decay of maternal transcripts. Overall, the work is of excellent quality, and the messages have a significant impact.

Strengths:

Compelling evidence for major shift in activities of an argonaute protein during development, and implications for how small RNAs affect early development. Very balanced and thoughtful discussion.

Weaknesses:

Claims on col-localization of specific 'granules' are not well supported by quantitative data

We have now included zoomed images of individual granules to better show the colocalization in Figure 4 and Figure 4—figure supplement 1, and performed Pearson’s colocalization analysis between different sets of proteins in Figure 4B. 

Reviewer #2 (Recommendations for the authors):

- The manuscript is very dense and the gene names are not helpful. For example, the authors mention ERGO-1 without clarifying the type of protein, etc. I suggest the authors include a figure to go with the introduction that describes the different classes of primary and secondary sRNAs, associated Argonautes, and other accessory proteins. Also include a table listing relevant gene names, protein classes, main localizations, and proposed functions for easy reference by the readers.

We agree that the genes names in different small RNA pathways are easily confused. We added a diagram and table in Figure 1—figure supplement 1 depicting the ERGO/NRDE and CSR pathways and added clarification about the ERGO/NRDE-3 pathway in the text from line 126-128.  

- Line 424 - the wording here and elsewhere seems to imply that SIMR-1 and ENRI-2, although not essential, contribute to NRDE-3 sRNA loading. The sequencing data, however, do not support this - the authors should be clearer on this. If the authors believe there are subtle but significant differences, they should show them perhaps by adding a panel in Figure 5 that directly compares the NRDE-3 IPs in wildtype versus simr-1 mutants. Figure 5H however does not support such a requirement.

As brought up by reviewer 1, we do not see difference in binding of ERGO-dependent sRNA in simr-1 mutant in late embryos. We do, however, see a modest, but significant, increase of CSR-sRNAs bound by NRDE-3 in simr-1 and enri-2 mutants, which we hypothesize could be due to a less efficient loading of ERGO-dependent 22G-RNAs by NRDE-3. The updated data are now in Figure 6G. We have also edited the text and model figure to soften these conclusions.

- Condensates of PGL proteins appear at a similar time and place (somatic cells of early embryos) as the embryonic SIMR-1 foci. The PGL foci correspond to autophagy bodies that degrade PGL proteins. Is it possible that SIMR-1 foci also correspond to degradative structures? The possibility that SIMR-1 foci are targeted for autophagy and not functional would fit with the finding that simr-1 mutants do not affect NRDE-3 loading in embryos.

We appreciate reviewer 2’s comments on possibility of SIMR granules acting as sites for degradation of SIMR-1 and NRDE-3. We think this is not the case for the following reasons: 1) if SIMR granules are sites of autophagic degradation, then we would expect that embryonic SIMR granules in somatic cells, like PGL granules, should only be observed in autophagy mutants; however we see them in wild-type embryos 2) we would not expect a functional Tudor domain to be required for granule localization; however in Figure 1—figure supplement 2B, we show that a point mutation in the Tudor domain of SIMR-1 abrogates SIMR granule formation, and 3) if NRDE-3(HK-AA) is recruited to SIMR granules for degradation while wild-type NRDE-3 is cytoplasmic, then NRDE-3(HK-AA) should shows a significantly reduced protein level comparing to wild-type NRDE-3. In the western blot in Figure 2—figure supplement 1B, NRDE-3 and NRDE-3(HK-AA) protein levels are similar, indicating that NRDE-3(HK-AA) is not degraded despite being unloaded. This is in contrast to what we have observed previously for HRDE-1, which is degraded in its unloaded state. If SIMR-1 played a role directly in promoting degradation of NRDE-3(HK-AA), we would similarly expect to see a change in NRDE-3 or NRDE-3(HK-AA) expression in a simr-1 mutant. We performed western blot and did not observe a significant change in protein expression for NRDE-3 (Figure 3—figure supplement 1A). 

Although under wild-type conditions, SIMR granules do not appear to be sites of autophagic degradation, upon treatment with lgg-1 (an autophagy protein) RNAi, we found that SIMR-1, as well as many other germ granule and embryonic granule-localized proteins, increase in abundance in late embryos.  This data demonstrates that ZNFX-1, CSR-1, SIMR-1, MUT-2/RDE-3, RRF-1, and unloaded NRDE-3 are removed by autophagic degradation similar to what have been shown previously for PGL-1 proteins (Zhang et al, 2009, Cell). We added these data to Figure 5. It is important to emphasize, however, that the timing of degradation differs for each granule assayed (Lines 447-450), indicating that there must be multiple waves of autophagy to selectively degrade subsets of proteins when they are no longer needed by the embryo.

- The observation that an NRDE-3 mutant that cannot load sRNAs localizes to SIMR-1 foci does not necessarily imply that wild-type unloaded NRDE-3 would also localize there. Unless the authors have additional data to support this idea, the authors should acknowledge that this hypothesis is speculative. In fact, why does cytoplasmic NRDE-3 not localize to granules in the rde-3;ego-1degron strain shown in Figure 6B?? Is it possible that the NRDE-3 mutant accumulates in SIMR-1 foci because it is unfolded and needs to be degraded?

We believe that wild-type NRDE-3 also localize to SIMR foci when unloaded. This is supported by the localization of wild-type NRDE-3 in eri-1 and rde-3 mutants, where a subset of small RNAs are depleted. Wild-type NRDE-3 localizes to both somatic SIMR-1 granules and the nucleus, depending on embryo stage (Figure 2A, Figure 2—figure supplement 1C). The granule numbers in eri-1 and rde-3 mutants are less than the nrde-3(HK-AA) mutant, consistent with the imaging data that NRDE-3 only partially localize to somatic granule (Figure 2A – 100-cell stage).

In the rde-3; ego-1 double mutant, the embryos have severe developmental defect: they cannot divide properly after 4-8 cell stage and exhibit morphology defects after that stage. In wild-type, SIMR foci does not appear until around 8-28-cell stage (shown in Figure 1C), so we believe that cytoplasmic NRDE-3 does not localize to foci in the double mutant is because of the timing.

- The authors propose that NRDE-3 functions in nuclei to target mRNAs also targeted in the cytoplasm by CSR-1. If so, how do they propose that NRDE-3 might do this since little transcription occurs in oocytes/early embryos?? Are the authors suggesting that NRDE-3 targets germline genes for silencing specifically at the times that zygotic transcription comes back on, or already in maturing oocytes? Is the transcription of most CSR-1 targets silenced in early embryos??

We appreciate the suggestions to check the function of NRDE-3 in oocytes. We tested this possibility and found it to be correct. NRDE-3 functions in oocytes for transcriptional repression by inhibiting RNA Pol II elongation. We added these data to Figure 8. We also attempted to do RT-qPCR, smFISH, and antiH3K9me3 Cut&Tag-seq on early embryos to further test the hypothesis that NRDE-3 acts with CSR-class 22G-RNAs in early embryos, but we either failed to obtain enough signal or failed to detect any significant difference (data not shown). Therefore, we think that the primary role for NRDE-3 bound to CSR-class 22G-RNAs may be for global transcriptional repression of oocytes prior to fertilization.

- Line 684-686: "In summary, this work investigating the role of SIMR granules in embryos, together with our previous study of SIMR foci in the germline (Chen and Phillips 2024), has identified a new mechanism for small RNA loading of nuclear Argonaute proteins in C. elegans". This statement appears overstated/incorrect since there is no evidence that SIMR-1 foci are required for sRNA loading of NRDE3. The authors should emphasize other models, as suggested above.

We have revised the text on line 869-871 to emphasize that SIMR granule regulate the localization of nuclear Argonaute proteins, rather than suggesting a direct role on controlling small RNA loading. We also edit the title, text, and legend for our model in Figure 9. 

Reviewer #3 (Recommendations for the authors):

Issues to be addressed:

- The authors show a switch in 22G RNA binding by NRDE-3 during embryogenesis. While the data is convincing, it would be great if it could be tested if the preferred NRDE-3 replacement model is indeed correct. This could be done relatively easily by giving NRDE-3 a Dendra tag, allowing one to colour-switch the maternal WAGO-3 pool before the zygotic pool comes up. Such data would significantly enhance the manuscript, as this would allow the authors to follow the fate of maternal NRDE-3 more precisely, perhaps identifying a period of sharp decline of maternal NRDE-3.

We think the NRDE-3 Dendra tag experiment suggested by the reviewer is a clever approach and we will consider generating this strain in the future. However, we feel that optimization of the color-switching tag between the maternal germline and the developing embryos is beyond the scope of this manuscript. To partially address the question about NRDE-3 fate during embryogenesis, we examined the single-cell sequencing data of C. elegans embryos from 1-cell to 16-cell stage (Tintori et al, 2016, Dev Cell; Visualization tool from John I Murray lab), as shown in Author response image 3 Panel A below, NRDE-3 transcript level increases as embryo develops, indicating that zygotic NRDE-3 is being actively expressed starting very early in development. We hypothesize that maternal NRDE-3 will either be diluted as the embryo develops or actively degraded during early embryogenesis. 

Author response image 3.

- Figure 3A: * should mark PGCs, but this seems incorrect. At the 8-cell stage there still is only one PGC (P4), not two, and at 100 cells there are only two, not three germ cells. Also, the identification of PGCs with a maker (PGL for instance) would be much more convincing.

We apologize for the confusion in Figure 3A. We changed the figure legend to clarify that the * indicate nuclear NRDE-3 localization in somatic cells for 8- and 100-cell stage embryos rather than the germ cells.  

- Overall, the authors should address colocalization more robustly. In the current manuscript, just one image is provided, and often rather zoomed-out. How robust are the claims on colocalization, or lack thereof? With the current data, this cannot be assessed. Pearson correlation, combined with line-scans through a multitude of granules in different embryos will be required to make strong claims on colocalization. This applies to all figures (main and supplement) where claims on different granules are derived from.

We thank reviewer 3 for this important suggestion. To better address the colocalization, we included insets of individual granules in Figure 2D and Figure 4. We also performed colocalization analysis by calculating the Pearson’s R value between different groups of proteins in Figure 4B, to highlight that SIMR-1 colocalizes with ENRI-2, NRDE-3(HK-AA), RDE-3, and RRF-1, while CSR-1 colocalizes with EGO-1.

For the proteins that lack colocalization in Figure 4—figure supplement 1, we also added insets of individual granules. Additionally, we included a new set of panels showing SIMR-1 localization compared to tubulin::GFP (Figure 4—figure supplement 1I) in response to a recent preprint (Jin et al, 2024, BioRxiv), which finds NRDE-3 (expressed under a mex-5 promoter) associating with pericentrosomal foci and the spindle in early embryos. We do not see SIMR-1 (or NRDE-3, data not shown) at centrosomes or spindles in wild-type conditions but made a similar observation for SIMR-1 in a mut-16 mutant (Figure 4E). All of the localization patterns were examined on at least 5 individual 100-cell staged embryos with same localization pattern.

- Figure 7: Its title is: Function of cytoplasmic granules. This is a much stronger statement than provided in the nicely balanced discussion. The role of the granules remains unclear, and they may well be just a reflection of activity, not a driver. While this is nicely discussed in the text, figure 7 misses this nuance. For instance, the title suggests function, and also the legend uses phrases like 'recruited to granule X'. If granules are the results of activity, 'recruitment' is really not the right way to express the findings. The nuance that is so nicely worded in the discussion should come out fully in this figure and its legend as well.

We have changed the title of Figure 7 (now Figure 9) to “Model for temporally- and developmentallyregulated NRDE-3 function” to deemphasize the role of the granules and to highlight the different functions of NRDE-3. Similarly, we have rephrased the text in the figure and legend and add a some details about our new results.

Minor:

Typo: line 663 Acaris

We corrected the typo.

eLife Assessment

This useful study presents findings on the efficacy and mechanisms of linalool protection against Saprolegnia parasitica oomycetes in the grass carp model. The evidence presented is solid since the methods, data and analyses broadly support the claims with only minor weaknesses. This work will be of great interest to scientists within the fields of aquaculture, ichthyology, microbiology, and drug discovery.

Reviewer #1 (Public review):

Summary:

The works seeks to investigate the efficacy of linalool as a natural alternative for combating Saprolegnia parasitica infections, which would provide great benefit to aquaculture. This paper shows the effect of linalool in vitro using a variety of techniques including changes in S. parasitica membrane integrity following linalool exposure and alterations in cell metabolism and ribosome function. Additionally, this work goes on to show that prophylactic and concurrent treatment of linalool at the time of S. parasitica infection can improve survival and tissue damage in vivo in their grass carp infection model. The conclusions of the paper are partially supported by the data with the corrections done by the authors improving clarity such that I believe there is merit in the work.

Reviewer #2 (Public review):

Summary:

In this study, the authors aimed to delineate the antimicrobial activity of linalool and tried to investigate the mode of action on linalool against S. parasitica infection. One of the main focus of this work was to identify the in vitro and in vivo mechanisms associated with the protective role of linalool against S. parasitica infection.

Strengths:

(1) Authors have used a variety of techniques to prove their hypothesis.<br /> (2) Adequate number of replicates were used in their studies.<br /> (3) Their findings showed a protective role of linalool against oomycetes and makes it an attractive future antibiotic in the aquaculture industry.

Weaknesses: The revised version of the manuscript is more thoroughly written with clearer explanations, however there are a few weaknesses in this manuscript.

(1) Although the introduction section was rewritten with rationale, it's still lengthy and not very much to the point.<br /> (2) The claim of linalool regulating the gut microbiota is based on the correlation analysis only. It's not super convincing and requires experimental validation to strengthen the claim.

Overall, the conclusions drawn by the authors are justified by the data. Importantly, this paper has discovered the novelty of the compound linalool as a potent antimicrobial agent and might open up future possibilities to use this compound in the aquaculture industry.

Author response:

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

Reviewer #1 (Public review):

(1) Adding microscopy of the untreated group to compare Figure 2A with would further strengthen the findings here.

First of all, we would like to thank Reviewer #1 for their comments and efforts on our manuscript. We have carefully revised it. We used a time-lapse method to capture images at 0 minutes, before any drugs were added. We will change '0 min' to 'untreated,' which will further strengthen the findings.

(2) Quantification of immune infiltration and histological scoring of kidney, liver, and spleen in the various treatment groups would increase the impact of Figure 4.

Thank you very much to Reviewer #1 for their comments and efforts on our manuscript. We have revised it carefully. We conducted quantitative analysis of immune infiltration in the kidney, liver, and spleen across different treatment groups. However, due to the extremely low number of abnormal cells in the negative control, treatment, and prophylactic groups, neither the instrument nor manual methods could reliably gate the cells. Consequently, quantification of immune infiltration and histological scoring were not performed.

(3) The data in Figure 6 I is not sufficiently convincing as being significant.

Thanks so much for Reviewer #1 comments and efforts for our manuscript. We have revised it carefully. Previous researches have shown that antibiotics and other drugs can cause alterations in gut microbiota. Therefore, we plan to study the effects of antibiotics on gut microbiota. To conduct this research, we need to isolate these microbes from the gut. Although this process is challenging, we still aim to explore the gut microbiota. If possible, we will continue to delve into interesting aspects of how antibiotics affect gut microbiota in future studies.

(4) Comparisons of the global transcriptomic analysis of the untreated group to the PC, LP, and LT groups would strengthen the author's claims about the immunological and transcriptomic changes caused by linalool and provide a true baseline.

Thanks so much for Reviewer #1 comments and efforts for our manuscript. We have revised it carefully. Due to the initial research design and data analysis strategy, we have focused on comparisons among the PC, LP, and LT groups to more directly explore the differences under various treatment conditions. Specifically, while the transcriptomic data from the untreated group could provide a basic reference, it has shown limited relevance to the core hypotheses of our study. Our research has aimed to investigate the immunological and transcriptomic changes among the treatment groups rather than comparing treated and untreated states. We believe that the current experimental design and data analysis have effectively revealed the mechanisms of linalool and that the additional comparisons among the treatment groups have further supported our conclusions. We hope the reviewer understands the rationale behind our experimental design. If there are additional suggestions, we are more than willing to further optimize the content of our manuscript.

Reviewer #2 (Public review):

(1) The authors have taken for granted that the readers already know the experiments/assays used in the manuscript. There was not enough explanation for the figures as well as figure legends.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. We will provide more detailed explanations of the experiments and assays used in the manuscript, as well as enhance the descriptions in the figure legends, to ensure that readers have a clear understanding of the figures and their context.

(2) The authors missed adding the serial numbers to the references.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. We will add serial numbers to the references to ensure proper citation and improve the clarity of our manuscript.

(3) The introduction section does not provide adequate rationale for their work, rather it is focused more on the assays done.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. We will add a section to the introduction that provides a rationale for our work, specifically focusing on the impact of plant extract on immunoregulation.

(4) Full forms are missing in many places (both in the text and figure legends), also the resolution of the figures is not good. In some figures, the font size is too small.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We will ensure that all abbreviations are expanded where necessary, both in the text and figure legends. Additionally, we will improve the resolution of the figures and increase the font size where needed to enhance clarity.

(5) There is much mislabeling of the figure panels in the main text. A detailed explanation of why and how they did the experiments and how the results were interpreted is missing.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. We will improve the labeling of the figure panels, provide detailed explanations of the experimental methods, including their rationale and interpretation, and clarify the connections between the methods.

(6) There is not enough experimental data to support their hypothesis on the mechanism of action of linalool. Most of the data comes from pathway analysis, and experimental validation is missing.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Actually, in our manuscript the transcriptomic data are not alone, and we carried out many experiments to substantiate the changes inferred from the transcriptomic data as SEM, TEM, CLSM, molecular docking, RT-qPCR, histopathological examinations. The detailed information is listed as below.

As shown in Figure 2, we combined the transcriptomic data related to membrane and organelle with SEM, TEM, and CLSM images. After deep analysis of these data and observation together, we illustrated that cell membrane may be a potential target for linalool.

As shown in Figure 3, we carried out molecular docking to explore the specific binding protein of linalool with ribosome which were screen out as potential target of linalool by transcriptomic data.

As shown in Figure 5, transcriptomic data illustrated that linalool enhanced the host complement and coagulation system. To substantiate these changes, we carried out RT-qPCR to detect those important immune-related gene expressions, and found that RT-qPCR analysis results were consistent with the expression trend of transcriptome analysis genes.

As shown in Figure 4 and 5, transcriptomics data revealed that linalool promoted wound healing tissue repair, and phagocytosis (Figure. 5E). To ensure these, we carried out histopathological examinations, and found that linalool alleviated tissue damage caused by S. parasitica infection on the dorsal surface of grass carp and enhancing the healing capacity (Figure. 4G).

Overall, we will conduct additional experiments to verify the mechanism of action of linalool in the future.

Reviewer #1 (Recommendations for the authors):

(1) Figure 1 Panel G is not referenced in the legend, this should be fixed

Thanks so much for Reviewer #1 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 1. The order of Panel F and G in Figure 1 is wrong. We have modified the order of Figure 1.

(2) Statistical comparisons between groups in Figure 4 Panels C-F is lacking and should be added.

Thanks so much for Reviewer #1 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 4 C-F. We have added statistical comparisons between groups in Figure 4 Panels C-F.

(3) Capitalize Kidney label in Figure 4G.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 4G. We have capitalized the K of kidney.

Reviewer #2 (Recommendations for the authors):

(1) The authors missed adding the serial numbers to the references. I could not go through the references to cross-check if they cited the right ones because it's extremely difficult to figure out which one corresponds to which reference number.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the references. We have added the serial numbers to the references.

(2) In the last paragraph of the introduction section, most of the techniques in the paper were summarized which does not go with the flow of the paper. The introduction should not be focused on the different techniques used the focus should be more on the rationale of the work. It would be nice if the last paragraph could be rewritten.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 85-94. We have added a section to the introduction that provides a rationale for our work, specifically focusing on the impact of plant extract on immunoregulation.

(3) The resolution of the figures is not good.

Thank you for your suggestion. We have revised it carefully. Please check all the figures. We have increased the resolution and size of all the figures.

(4) Mostly, the figure legends sound like results, with not enough explanation. Full forms are missing in many places which would make the readers go back to the text/other figures each time.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it throughout the manuscript and all the figure legends. We have added full names and abbreviations to both the manuscript and all the figure legends so that we don't make the readers go back to the text/other figures each time.

(5) Figure 1:

Figure 1A: there is not enough explanation for this panel. It's not clear from the text which other EOs than Linalool are referred to here. Which EOs were extracted from daidai flowers?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in the Figure 1A. Figure 1A is divided into “Essential oils (EOs)” and “The main compounds of EOs” to make it easier to distinguish.

Figure 1B: do the three different wells of each set represent three replicates? If so, are they biological/technical replicates? Also, I'm not sure how the MFC was determined from this figure (line 116) because clearly this panel only corresponds to the determination of MICs, not MFCs.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 126-130. The three different wells of each set represent three biological replicates. After adding 5 μL of resazurin dye, when the color of the wells turned to pink, the linalool concentration in the first non-pink well corresponded to the MIC. The culture liquid in the well where no mycelium growth was seen was marked onto the plate and incubated at 25°C for 7 days. The well with the lowest linalool concentration and no mycelium growth was identified as MFC.

Figure 1C: the figure legend says that the effect of linalool on mycelium growth inhibition was done over a 6hr timepoint but according to the figure the timepoint was 60hr. I am also confused about the concentrations of linalool used. Although a range of concentration from 0 to 0.4% is mentioned, I only see the time vs diameter curves for 7 concentrations.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 983 and Figure 1C. We have changed 6 h to 60 h in the figure legends. The reason why only the time vs diameter curves for 7 concentrations in Figure1C is that the growth inhibition of 0.4%, 0.2% and 0.1% linalool on mycelial growth is the same. As a result, the time vs diameter curves coincide. We have shown the time and diameter curves of 0.4%, 0.2% and 0.1% concentration with three dotted lines of different colors and sizes in Figure 1C.

Figure 1D: mislabeled as 1G in the figure panel.

Figures 1E and 1G: Figure 1E is missing and I do not see any figure legend for Figure 1G.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 1. The order of Panel F and G in Figure 1 is wrong. We changed the order of Figure 1 ABCDEF, no Figure G.

Overall, Figure 1 is very confusing and needs rewriting. Also, there is a need to add more explanation of the figure panels in the results section.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 1. We have corrected all the problems in Figure1. And we have added more explanation of the figure panels in the results section, and increased the correlation between methods, in order to show how to carry out the experiment logically and interpret the results, please check them in Line 126-130, 144-147, 174-179, 213-217, 343-345, 677-682.

(6) Figure 2:

The authors could justify the reason for doing the experiments before moving into the results they got.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the methods and results in the manuscript, please check them in Line 126-130, 144-147, 174-179, 213-217, 343-345, 677-682. We have added more explanation of the figure panels in the results section, and increased the correlation between methods, in order to show how to carry out the experiment logically and interpret the results.

What concentration of linalool was used?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 992-996. The mycelium treated with 6×MIC (0.3%) linalool was observed by Confocal laser scanning microscopy (CLSM), and the mycelium treated with 1×MIC 0.05% linalool was observed by Scanning Electron Microscope (SEM) and transmission electron microscopy (TEM).

The full form of DEGs has been mentioned later, but it should be mentioned in the figure legend of Figure 2 as this is the first time the term was used. Also, what is the full form of DEPs?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 168, 175, 182, 631, 998, 1001. The word DEPs in Figure 2I was incorrect, and we have changed DEPs to DEGs.

Is there a particular reason for looking into the cellular component rather than molecular function and biological processes in the GO analysis? (what I see is that Figure 2H indicates the prevalence of catalytic activity, binding, cellular, and metabolic processes as well). Also, there is not enough explanation of the observation from Figure 2I (both in the results section and figure legend).

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 174-179, 998-1002 (Figure 2I). The reason we looked at cellular components rather than molecular functions and biological processes in GO analysis is because we focused more on the effects of cell membranes and cell walls. These results are closely related to and echo the results of our scanning electron microscopy (SEM) and transmission electron microscopy (TEM), and also support the results of electron microscopy. Enough explanations have added to the results and figure legend section to explain the observations from Figure 2I.

(7) Figure 3:

Figures 3A and 3B: The adjusted p value is already indicated in the figures, so there is no need to add statistical significance (Asterix) to each bar. The resolution for these panels is not good and the font is too small.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 3A and 3B. We have removed statistical significance (Asterix) from Figure3A and 3B. If we are lucky, we will upload the clearest figures when the manuscript is published.

Figure 3C: the figure legend is missing (wrongly added as KEGG analysis, which should be network analysis). The numbering for the figure legends is wrong. What are the node sizes (5, 22, 40, 58) mentioned in the figure represent? Also, I wonder why ribosome biogenesis in eukaryotes has been indicated as the most enriched pathway despite its less connection to the other nodes.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 3C. Figure 3C is KEGG analysis generated by software, not network analysis. For the convenience of readers, we have made a new Figure of KEGG analysis.

Figure 3D: KEGG enrichment and GO analysis: global/local search? Which database was used as a reference?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the 633-635. Functional enrichment analysis was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. KEGG pathway analysis was conducted using Goatools.

Figure 3E: why were the RNA pol structures compared? The authors did not mention anything about this panel in their results.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the line 207. We found that many DEGs related to ribosome biogenesis (Figure 3D) and RNA polymerase (Figure 3E) are down expressed. Because RNA polymerase is closely related to ribosome biogenesis, the downregulation of RNA polymerase directly affects the synthesis of ribosome-related RNAs, including rRNA, mRNA, and tRNA, thereby inhibiting ribosome production. This relationship is particularly significant in cell growth, division, and the response to external environmental changes.

Figures 3F and 3G: please mention which model is illustrated (ribbon/sphere model).

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the line 1010-1015. The tertiary structure of NOP1 was displayed using a cartoon representation. Molecular docking of linalool with NOP1 was performed by enlarging the regions binding to the NOP1 activation pocket to showcase the detailed amino acid structures, which were presented using a surface model, while the small molecule was displayed with a ball-and-stick representation.

Figure 3H: this panel needs more explanation. Why were some of the ABC transporters upregulated while some were downregulated?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. It is a common phenomenon that microorganisms adjust the expression of genes related to substance transport in response to different environmental stimuli to optimize their survival strategies. The expression of ATP-binding cassette (ABC) transporters can be upregulated or downregulated due to various factors, such as environmental stimuli, metabolic demands, energy consumption, species specificity, and signaling molecules. This explains why some ABC transporters are upregulated while others are downregulated.

(8) Figure 4:

There was no statistical significance shown in the figures (D-F) which makes me wonder how they worked out that there was any significant increase/decrease, as mentioned in the text. What are the p values? What is the number of replicates? What concentration of linalool was used?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully.  Please check the Figure 4D-F. In this study, 4 groups were established: (1) Positive control (PC) group (10 fish infected with S. parasitica). (2) Linalool therapeutic (LT) group (10 fish infected with S. parasitica, soaked in 0.00039% linalool in a 20L tank for 7 days). (3) Linalool prophylactic (LP) group (10 uninfected fish soaked in 0.00039% linalool in a 20L tank for 2 days, followed by the addition of 1×10⁶ spores/mL secondary zoospores). (4) Negative control (NC) group (10 uninfected fish without linalool treatment). Each group had 3 replicate tanks. In each group, 8 fish were utilized for immunological assays, and on day 7, blood samples were collected from the tail veins using heparinized syringes and left to coagulate overnight at 4°C. Kits from Nanjing Jiancheng Institute (Nanjing, China) were used to measure lysozyme (LZY) activity, superoxide dismutase (SOD) activity, and alkaline phosphatase (AKP) activity.

(9) Figure 5:

Again, the resolution and font size are off. Please mention the full forms of the terms used in the figure legend. The interpretation of the in vivo protective mechanism of linalool is completely based on GO enrichment and KEGG pathway analysis (also some transcriptional analysis). The only wet lab validation done was by checking the mRNA level of some cytokines but that does not necessarily validate what the authors claim.

Thank you for your suggestion. We have revised it carefully. Please check all the figures and figure legend. We have increased the resolution and size of all the figures and used the full forms of the terms in figure legend. If we are lucky, we will upload the clearest figures when the manuscript is published. Currently, in the field of aquaculture research, mRNA quantification at the genetic level faces numerous challenges compared to model organisms like mice and zebra fish, primarily due to the lack of available antibodies. For instance, antibodies related to grass carp have not yet been commercialized, making protein-level studies and validations significantly more difficult. This lack of antibodies limits the progress of protein verification. However, we hope to design more experiments and validation tests in the future to gradually overcome these technical bottlenecks and provide stronger support for research in the future.

(10) Figure 6:

There is not enough explanation on why and how the experiments were done. It seems like the authors already presumed that the readers know the experiments. The interpretation of the PCA plot is not clear. Why are the quadrant sizes different? How was the heat map plotted? Also, the claim of linalool regulating the gut microbiota is only dependent on the correlation analysis and there is no wet lab validation for this. The data represented in this figure is not enough to prove their hypothesis and needs further investigation.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check the Figure 6. We will improve the labeling of the figure panels, provide detailed explanations of the experimental methods, including their rationale and interpretation, and clarify the connections between the methods.

The goal of PCoA is to preserve the distance relationships between samples as much as possible through the principal coordinates, thereby revealing the differences or patterns in microbial composition among different groups. For example, in our study, PCoA analysis demonstrated that the microbial compositions of the positive control (PC), linalool prophylactic (LP), and linalool therapeutic (LT) groups showed significant differences in the reduced dimensional space, possibly indicating that these treatments had a notable impact on the microbial community.

In our study, the heatmap was generated using the Majorbio Cloud Platform. This platform visualized the preprocessed microbial community data, providing an intuitive representation of the differences in microbial composition and relative abundance among samples. The platform automatically performed steps such as data normalization, color mapping, and clustering analysis, offering convenience for data analysis and interpretation.

Previous researches have shown that antibiotics and other drugs can cause alterations in gut microbiota. Therefore, we plan to study the effects of antibiotics on gut microbiota. To conduct this research, we need to isolate these microbes from the gut. Although this process is challenging, we still aim to explore the gut microbiota. If possible, we will continue to delve into interesting aspects of how antibiotics affect gut microbiota in future studies.

(11) Figure 7:

This figure does not clarify how they did the interpretation. The in vivo study does not phenocopy their in vivo studies.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. we have carefully reviewed and confirmed the current experimental design and data analysis. Although we have not made any changes to Figure 7, we have further clarified the interpretation of the results in the revised manuscript, especially concerning the discrepancies between the in vivo and in vitro studies. We have added more experimental background information to help better understand the possible reasons for these differences. We hope the reviewer will understand our explanation and we look forward to your further feedback.

(12) Minor comments:

Line 61: what's meant by "et al"?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 61. We have removed "et al".

Line 87-88: please add a citation referring to the earlier studies.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 109.

Line 151-152: the term "related to" has been used a couple of times. Mentioning it once in the beginning and avoiding repeating the same word might be better.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 168-171.We have rewritten this paragraph to avoid repeating the word “related to”.

How did they reconstitute the EO compounds?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. The EO compounds we used in our experiments were partially extracted from essential oils in the laboratory and partially purchased from ThermoFisher (USA).

Line 544: needs explanation of how there was a 2-fold dilution in the concentrations shown in the figure compared to the concentrations mentioned here.

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. We set the concentration of MIC assay for mycelium to be 0.8%, 0.4%, 0.2%, 0.1%, 0.05%, 0.025%, 0.0125%, and 0.00625%, and the concentration of MIC assay for spores to be 0.4%, 0.2%, 0.1%. 0.05%, 0.025%, 0.0125%, 0.00625%. Figure 1B shows the MIC determination of linalool on spores, while the MIC determination of mycelium is not shown.

Line 546: remove "were".

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 573. We have removed "were".

Line 555: what concentration of malachite green and tween 20 was used?

Thanks so much for Reviewer #2 comments and efforts for our manuscript. We have revised it carefully. Please check it in Line 579-580. 2.5mg /mL malachite green and 1% Tween 20 were used.

eLife Assessment

This useful study uses a model of Streptococcus suis (a pig pathogen) infection in mice using an intranasal route, the natural route of infection ignored in most of the literature. The study aims to understand how capsular polysaccharides (CPS) contribute to neuropathology and virulence. The findings suggest that the olfactory route may lead to meningitis before bacteremia occurs and that CPS down-regulation may play a role in this process. However, the study remains incomplete as presented.

Reviewer #1 (Public review):

Summary:

The manuscript by Wang et al. investigates the relationship between Streptococcus Suis (S. Suis) growth phases and levels of virulence factor, capsular polysaccharide (CPS), in the bacterial cell wall. They use an understudied mouse intranasal infection model to connect growth phase related CPS abundance to the pathogenicity of the bacteria in the nose, blood, and other organs. Adoptive transfer of serum against either CPS or V5 (five other virulence factors) reinforces their discovery of CPS levels on S. Suis in different organs and stages of infection. Vaccination against bacterial infections can be difficult, and understanding how the serotype of a bacterial pathogen changes between infection sights and systemic disease is critical. Further, understanding host-pathogen interactions at early time points in the upper respiratory tract may have broad implications for vaccine development. While some of the results are interesting and compelling, others are not supported by the data and require further experimental work.

Strengths:

The model of intranasal infection is compelling to expand upon work previously done in vitro and with systemic routes of infection. The histology and fluorescent imaging of the olfactory epithelium and olfactory bulb complement work in Figure 2 about the attachment of S. suis to epithelial cells and the bacterial burden over time in different organs of Figure 3. Histology was performed at 1 hour and 9 days after intranasal infection with stationary phase S. suis and drives home that this pathogen can invade the olfactory nerve and may potentially cause bacterial meningitis seen in some infected swine.

The adoptive transfer of either anti-CPS or anti-V5 to mice before infection at both longer (12 hr), and shorter (0.5 hr) time points is useful to demonstrate that the changes in cell wall composition between the NALT/CSF and blood compartments result in different efficacy in clearing bacteria from those locations. This is fundamental for the development of vaccines for the swine industry and begs those developing other bacterial vaccines to consider what virulence factors are the most useful as neutralizing antibody targets at the sight of bacterial invasion.

Demonstrating that the amount of CPS within the cell wall of S. Suis is related to the growth phase of the bacteria is an important consideration for vaccine development. While others had previously shown that CPS levels were higher in the blood than in the CNS, and that CPS decreases the invasion of epithelial cells, the close look at the olfactory epithelium at an early time point ties together in vitro findings. The control of a CPS-negative strain was critical to understanding their findings. The location and the microbial community that bacterial pathogens live within may change the growth phase and therefore also the cell wall components.

Weaknesses:

The authors present compelling data that is relevant to the development of anti-bacterial vaccinations and show a relationship between CPS levels and pathogenicity. However, the use of a laboratory murine model requiring acetic acid pre-treatment and a high i.n. dose. Therefore, the findings presented may not represent what occurs in swine. Furthermore, several conclusions are not supported by the data and require substantial new experimental support. Thus, major concerns remain that impact the validity of the findings.

Major concerns for the manuscript:

The intranasal infections were done with S. Suis in the stationary phase which has been shown to have less CPS on the cell wall. While this mimics the literature that shows S. Suis to have less CPS in the CNS, the difference in the pathogenesis of a log phase vs. stationary phage intranasal infection would be interesting. Especially because the bacteria is a part of the natural microbial community of swine tonsils, it is curious if the change in growth phase and therefore CPS levels may be a causative reason for pathogenic invasion in some pigs. To take this line of thought a step further, the authors should consider taking the bacteria from NALT/CSF and blood and compare the lag times bacteria from different organs take to enter a log growth phase to show whether the difference in CPS is because S. Suis in each location is in a different growth phase. If log phase bacteria were intranasally delivered, would it adapt a stationary phase life strategy? How long would that take? Lastly, the authors should be cautious about claims about S. suis downregulating CPS in the NALT for increased invasion and upregulating CPS to survive phagocytosis in blood. While it is true that the data shows that there are different levels of CPS in these locations, the regulation and mechanism of the recorded and observed cell wall difference is not investigated past the correlation to the growth phase. While mechanistic work is outside the scope of the current work, readers should keep in mind that these results may be explained multiple ways. In addition, the mouse model is used rather than the usual host of a pig. The NALTs of conventional pigs and SPF mice certainly have unique microbial communities and this may affect the pathogenesis of S. suis in the mouse, therefore influencing the results. Because the authors show a higher infection rate in the mouse with acetic acid, they may want to consider investigating what the mouse NALT microenvironment is naturally doing to exclude more bacterial invasion in future studies. Is it simply a host mismatch or is there something about the microbiome or steady-state immune system in the nose of mice that is different from pigs?

Author response:

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

Public Reviews:

Reviewer 1:

(1) Some conclusions are not completely supported by the present data, and at times the manuscript is disjoint and hard to follow. While the work has some interesting observations, additional experiments and controls are warranted to support the claims of the manuscript.

Thank you for the comments. We revised some of the claims and conclusions to be more objective and result-supportive.

(2) While the authors present compelling data that is relevant to the development of anti-bacterial vaccinations, the data does not completely match their assertions and there are places where some further investigation would further the impact of their interesting study.

We do not fully agree with the reviewer's comments. We have demonstrated that changes in CPS levels during infection are associated with pathogenesis, which will guide future studies on the underlying mechanisms. A significant amount of effort is required for studying mechanisms, which is beyond the scope of this research. We concur with the reviewer that assertions should be made cautiously until further studies are conducted. We have revised these assertions to align with the data and to avoid extrapolating the results (pages 7, lines 126, 133-136; page 11, lines 216-218; page 13, line 264; and page 18, lines 378-383).

(3) The difference in the pathogenesis of a log phase vs. stationary phage intranasal infection would be interesting. Especially because the bacteria is a part of the natural microbial community of swine tonsils, it is curious if the change in growth phase and therefore CPS levels may be a causative reason for pathogenic invasion in some pigs.

S. suis is a part of the natural microbial community of swine tonsils but not mouse NALT. It is interesting to know if CPS levels are low in pig tonsils since CPS is hydrophilic and not conducive to bacterial adhesion. In the study, mice were i.n. infected with a high dose of the bacteria, which could increase opportunities for dissemination (acidic acid may not be a contributor since with or without it is similar). S. suis getting into other body compartments from pig tonsils might be triggered by other conditions, such as viral coinfection, nasal cavity inflammation, cold weather, and decreased immunity.

Experiments with pig blood and phagocytes have shown that genes involved in the synthesis of CPS are upregulated in pig blood. In contrast, these genes are downregulated [1]. In addition, the absence of CPS correlated with increased hydrophobicity and phagocytosis, proposing that S. suis undergoes CPS phase variation and could play a role in the different steps of S. suis infection [2]. We showed direct evidence of encapsulation modulation associated with S. suis pathogenesis in mice. A pig infection model is required to confirm these findings.

(4) The authors should consider taking the bacteria from NALT/CSF and blood and compare the lag times bacteria from different organs take to enter a log growth phase to show whether the difference in CPS is because S. suis in each location is in a different growth phase. If log phase bacteria were intranasally delivered, would it adapt a stationary phase life strategy? How long would that take? 

What causes CPS regulation in vivo is not known. CPS changes in different culture stages, indicating that stress, such as nutrition levels, is one of the signals triggering CPS regulation. The microenvironment in the body compartments is far more complex than in vitro, in which host cells, immune factors and others may affect CPS regulation, individually or collectively. The reviewer’ question is important but the suggested experiment is impracticable since bacterial numbers taken from organs are few, and culturing the bacteria in vitro would obliterate the in vivo status.  

(5) Authors should be cautious about claims about S. suis downregulating CPS in the NALT for increased invasion and upregulating CPS to survive phagocytosis in blood. While it is true that the data shows that there are different levels of CPS in these locations, the regulation and mechanism of the recorded and observed cell wall difference are not investigated past the correlation to the growth phase.

We lower the tone and change the claim as “suggest a correlation between lower CPS in the NALT and a greater capacity for cellular association, whereas elevated CPS levels in the blood are linked to improved resistance against bactericidal activity. However, the mechanisms behind these associations remain unknown.” (page 7, lines 133-136).

(6) The mouse model used in this manuscript is useful but cannot reproduce the nasal environment of the natural pig host. It is not clear if the NALTs of pigs and mice have similar microbial communities and how this may affect the pathogenesis of S. Suis in the mouse. Because the authors show a higher infection rate in the mouse with acetic acid, they may want to consider investigating what the mouse NALT microenvironment is naturally doing to exclude more bacterial invasion. Is it simply a host mismatch or is there something about the microbiome or steady-state immune system in the nose of mice that is different from pigs?

It is a very interesting comment. The mice are SPF level. The microenvironment in SPF mouse NALT should be significantly different from conventional pig tonsils. Although NALT in mice resembles pig tonsils in function, many factors may contribute to the sensitivity to S. suis colonization in the pig nasal cavity, such as the microbiome and local steady-state immune system. More complex microbiota in tonsils could be one of the factors. Analyzing what makes S. suis inclined towards colonization in pig tonsils by SPF and conventional pigs are an ideal experiment to answer the question. 

(7) Have some concerns regarding the images shown for neuroinvasion because I think the authors mistake several compartments of the mouse nasal cavity as well as the olfactory bulb. These issues are critical because neuroinvasion is one of the major conclusions of this work.

Thank you for your comments. The olfactory epithelium (OE) is located directly underneath the olfactory bulb in the olfactory mucosa area and lines approximately half of the nasal cavities of the nasal cavity. The remaining surface of the nasal cavity is lined by respiratory epithelium, which lacks neurons. The olfactory receptor neuron in OE is stained green in the images by β-tubulin III, a neuron-specific marker. The respiratory epithelium is colorless due to the absence of nerve cells. Similarly, the green color stained by β-tubulin III identifies the olfactory bulb. The accuracy of the anatomic compartments of the mouse nasal cavity has been checked and confirmed by referring to related literature [3, 4].

References

(1) Wu Z, Wu C, Shao J, Zhu Z, Wang W, Zhang W, Tang M, Pei N, Fan H, Li J, Yao H, Gu H, Xu X, Lu C. The Streptococcus suis transcriptional landscape reveals adaptation mechanisms in pig blood and cerebrospinal fluid. RNA. 2014 Jun;20(6):882-98.

(2) Charland N, Harel J, Kobisch M, Lacasse S, Gottschalk M. Streptococcus suis serotype 2 mutants deficient in capsular expression. Microbiology (Reading). 1998 Feb;144 ( Pt 2):325-332.

(3) Pägelow D, Chhatbar C, Beineke A, Liu X, Nerlich A, van Vorst K, Rohde M, Kalinke U, Förster R, Halle S, Valentin-Weigand P, Hornef MW, Fulde M. The olfactory epithelium as a port of entry in neonatal neurolisteriosis. Nat Commun. 2018;9(1):4269.

(4) Sjölinder H, Jonsson AB. Olfactory nerve--a novel invasion route of Neisseria meningitidis to reach the meninges. PLoS One. 2010 Nov 18;5(11):e14034.

Reviewer 2:

(1) However, there are serious concerns about data collection and interpretation that require further data to provide an accurate conclusion. Some of these concerns are highlighted below:

Both reviewers were concerned about some of the interpretations of the results. We modified the interpretations in related lines throughout the manuscript (Please see the related responses to Reviewer 1).

(2) In figure 2, the authors conclude that high levels of CPS confer resistance to phagocytic killing in blood exposed S. suis. However, it seems equally likely that this is resistance against complement mediated killing. It would be important to compare S. suis killing in animals depleted of complement components (C3 and C5-9).

We thank the reviewer for the comment. The experiment should be Bactericidal Assay instead of anti-phagocytosis killing. CPS is a main inhibitor of C3b deposition [1]. It interferes with complement-mediated and receptor-mediated phagocytosis; and direct killing. Data in Figure 2C is expressed as “% of bacterial survival in whole blood” for clarity (page 8, Fig. 2C and page 23, lines 489-490).

(3) Intranasal administration non-CPS antisera provides a nice contrast to intravenous administration, especially in light of the recently identified "blood-olfactory barrier". Can the authors provide any insight into how long and where this antibody would be located after intranasal administration? Would this be antibody mediated cellular resistance, or something akin to simple antibody "neutralization"

Anti-V5 may not stay long locally following intranasal administration. Efficient reduction of S. suis colonization in NALT supports that anti-V5 could recognize and neutralize the bacteria in NALT quickly, thereby reducing further dissemination in the body. Antibody-mediated phagocytosis may not play a major role because neutrophils are mainly present in the blood but not in the tissues.  

(4) The micrographs in Figure 7 depict anatomy from the respiratory mucosa. While there is no histochemical identification of neurons, the tissues labeled OE are almost certainly not olfactory and in fact respiratory. However, more troubling is that in figures 7A,a,b,e, and f, the lateral nasal organ has been labeled as the olfactory bulb. This undermines the conclusion of CNS invasion, and also draws into question other experiments in which the brain and CSF are measured.

We understand the significance of your concerns and appreciate your careful review of Figure 7. The olfactory epithelium (OE) is situated directly beneath the olfactory bulb in the olfactory mucosa area and covers about half of the nasal cavity. This positioning allows information transduction between the olfactory and the olfactory epithelium. The remaining surface of the nasal cavity is lined with respiratory epithelium, which does not contain neurons and primarily serves as a protective barrier. In contrast, the olfactory epithelium consists of basal cells, sustentacular cells, and olfactory receptor neurons. The olfactory receptor neurons are specifically stained green in the images using β-tubulin III, a marker that is unique to neurons. The respiratory epithelium appears colorless due to the lack of nerve cells. Similarly, the green staining with β-tubulin III also highlights the olfactory bulb. The anatomical structures indicated in the images are consistent with those described in the literature [2, 3], confirming that the anatomy of the nasal cavity has been accurately identified.

(5) Micrographs of brain tissue in 7B are taken from distal parts of the brain, whereas if olfactory neuroinvasion were occurring, the bacteria would be expected to arrive in the olfactory bulb. It's also difficult to understand how an inflammatory process would be developed to this point in the brain -even if we were looking at the appropriate region of the brain -within an hour of inoculation (is there a control for acetic acid induced brain inflammation?). Some explanations about the speed of the immune responses recorded are warranted.

Thank you for highlighting this issue. Cerebrospinal fluid (CSF) flows into the subarachnoid space surrounding the spinal cord and the brain. There are direct connections from this subarachnoid space to lymphatic vessels that wrap around the olfactory nerves as they cross the cribriform plate towards the nasal submucosa. This connection allows for the drainage of CSF into the nasal submucosal lymphatics in mice [4, 5]. Bacteria may utilize this CSF outflow channel in the opposite direction, which explains the development of brain inflammation in the distal areas of brain tissue adjacent to the subarachnoid space. We have included additional relevant information in the revised manuscript (page 16, lines 323-325).

(6) The detected presence of S. suis in the CSF 0.5hr following intranasal inoculation is difficult to understand from an anatomical perspective. This is especially true when the amount of S. suis is nearly the same as that found within the NALT. Even motile pathogens would need far longer than 0.5hr to get into the brain, so it's exceedingly difficult to understand how this could occur so extensively in under an hour. The authors are quantifying CSF as anything that comes out of the brain after mincing. Firstly, this should more accurately be referred to as "brain", not CSF. Secondly, is it possible that the lateral nasal organ -which is mistakenly identified as olfactory bulb in figure 7- is being included in the CNS processing? This would explain the equivalent amounts of S. suis in NALT and "CSF".

The high dose of inoculation used in the experiment may explain the rapid presence of S. suis in the CSF. Mice exhibit low sensitivity to S. suis infection, and the range for the effective intranasal infectious dose is quite narrow. Higher doses lead to the quick death of the mice, while lower doses do not initiate an infection at all. The dose used in this study is empirical and is intended to facilitate the observation of the progression of S. suis infection in mice.

The NALT tissue and CSF samples are collected separately. After obtaining the NALT tissue, the nasal portion was carefully separated from the rest of the head along the line of the eyeballs. The brain tissue was then extracted from the remaining part of the head to collect the CSF, and it was lacerated to expose the subarachnoid space without being minced. This procedure aims to preserve the integrity of the brain tissue as much as possible. Further details about the CSF collection process can be found in the Materials and Methods section (page 24, lines 508-512).

(7) To support their conclusions about neuroinvasion along the olfactory route and /CSF titer the authors should provide more compelling images to support this conclusion: sections stained for neurons and S. suis, images of the actual olfactory bulb (neurons, glomerular structure etc).

Thank you. We respectfully disagree with the reviewer. We stained neurons using a neuron-specific marker to identify the anatomical structures of the olfactory bulb and olfactory epithelium (in green). We used an S. suis-specific antibody to highlight the bacteria present in these areas (in orange and red). The images, along with the bacteria found in the cerebrospinal fluid (CSF) and the brain inflammation observed early in the infection, strongly support our conclusion regarding brain invasion through the olfactory pathway. Please see the response to question 4 for further clarification.

References

(1) Seitz M, Beineke A, Singpiel A, Willenborg J, Dutow P, Goethe R, Valentin-Weigand P, Klos A, Baums CG. Role of capsule and suilysin in mucosal infection of complement-deficient mice with Streptococcus suis. Infect Immun. 2014 Jun;82(6):2460-71.

(2) Sjölinder H, Jonsson AB. Olfactory nerve--a novel invasion route of Neisseria meningitidis to reach the meninges. PLoS One. 2010 Nov 18;5(11):e14034.

(3) Pägelow D, Chhatbar C, Beineke A, Liu X, Nerlich A, van Vorst K, Rohde M, Kalinke U, Förster R, Halle S, Valentin-Weigand P, Hornef MW, Fulde M. The olfactory epithelium as a port of entry in neonatal neurolisteriosis. Nat Commun. 2018;9(1):4269.

(4) Yoon JH, Jin H, Kim HJ, Hong SP, Yang MJ, Ahn JH, Kim YC, Seo J, Lee Y, McDonald DM, Davis MJ, Koh GY. Nasopharyngeal lymphatic plexus is a hub for cerebrospinal fluid drainage. Nature. 2024 Jan;625(7996):768-777.

(5) Spera I, Cousin N, Ries M, Kedracka A, Castillo A, Aleandri S, Vladymyrov M, Mapunda JA, Engelhardt B, Luciani P, Detmar M, Proulx ST. Open pathways for cerebrospinal fluid outflow at the cribriform plate along the olfactory nerves. EBioMedicine. 2023 May;91:104558.

Response to Recommendations for the authors:

Reviewer 1:

Minor concerns for the manuscript:

(1) In the introduction, please consider giving a little more background about the bacteria itself and how it causes pathogenesis.

We appreciate your suggestion. We have included additional background on the virulent factors and the pathogenesis of the bacteria in the introduction to enhance understanding of the results (page 4, lines 63-69).

(2) Figure 2C would be more correct to say percent survival as the CFUs before and after are what are being compared and not if the bacteria is being phagocytosed or not. Flow cytometry of the leukocytes and a fluorescent S. Suis would show phagocytosis. Unless that experiment is performed, the authors cannot claim that there is a resistance to phagocytosis.

Thank you for your feedback. We agree with the reviewer that the experiment should be Bactericidal Assay rather than anti-phagocytosis killing. CPS interferes with complement-mediated phagocytosis and direct killing, and receptor-mediated phagocytosis. To enhance clarity, the data in Fig. 2C has been presented as “% of bacterial survival in whole blood” (page 8).  

(3) There are two different legends present for Figure 1. Please resolve.

We apologize for the oversight. The redundant figure legend has been removed (page 6).

(4) There are places such as in lines 194-195, that there are assertions and interpretations about the data that are not directly drawn from the data. These hypotheses are valuable, but please move them to the discussion.

Thank you for your suggestion. The hypothesis has been moved to the Discussion section (page 19, lines 402 - 405).

(5) In Figure 4B, higher resolution images would strengthen the ability of non-microbiologists to see the differences in CPS levels in the cell wall.

We achieved the highest resolution possible for clearer distinctions in CPS levels. To enhance the visualization of the different CPS levels in the images, we revised the description of the CPS changes in Figure 4B within the results section (page 11, lines 208-213).

(6) In Figure 5 there is no D. Further, the schematics throughout would be easier to parse with the text if the challenge occurred at time 0. Consider revising them for clarity.

Thank you for highlighting the error. We have removed "i.v + i.n (Fig. 5)" from Figure 5A and made adjustments to the schematic illustrations in Figures 5 and 6 as recommended by the reviewer (page 14).

(7) What is the control for the serum? The findings for figures 5 and 6 would be much stronger if a non- S. Suis isotype control serum was also infused.

We used a naive serum as a control to avoid interference from a non-S. suis isotype control that targets other surface molecules of S. suis serotypes.

(8) Figure 6 legend does not include the anti-CPS treatment.

Thank you. We have added anti-CPS serum in the legend (page 15, line 249).

(9) Figure 7 legend does not include the time point for panel 7A.

Thank you. The time point is shown on Fig.7A (page 17).

(10) Figure 7 should show OB micrographs or entire brain including the OB.

The neuron-specific marker, β-tubulin III, identifies the neuro cells in the olfactory bulb (OB) as shown in Fig. 7A. Unfortunately, we were unable to provide an image of the entire brain that includes the OB due to limitations in our section preparation. We apologize for the mislabeled structure in Fig. 7A, which may have caused confusion. We have corrected the labeling for consistency (see page 15, lines 257-260). Additionally, we included a drawing of the sagittal plane of the rodent's nose, depicting the compartments of the OB, olfactory epithelium (OE), nasal cavity (NC), and brain. This illustration, presented in Fig. 7B on page 17, aims to clarify the structural and functional connections between the nasopharynx and the CNS.

(11) Some conclusions may be better drawn if figures were to be consolidated. As noted above, the data at times feels disjointed and the importance is more difficult for readers to follow because data are presented further apart. Particularly figures 5 and 6 which are similar with different time points and controls of antisera administrative routes; placing these figures together would be an example of increasing continuity throughout the paper.

Thank you for the valuable suggestion. Figures 5 and 6, along with their related descriptions in the results section, have been combined for better cohesiveness (pages 14-15).

Reviewer #2:

To support their conclusions about neuroinvasion along the olfactory route and /CSF titer the authors should provide more compelling images to support this conclusion: sections stained for neurons and S. suis, images of the actual olfactory bulb (neurons, glomerular structure etc).

Please refer to our responses to Reviewer 1's Question 7, Reviewer 2's Questions 4 and 7 in the public reviews, and Reviewer 1's Question 10 in the authors' recommendations.

eLife Assessment

This valuable study reports the link between a disruption in testicular mineral (phosphate) homeostasis, FGF23 expression, and Sertoli cell dysfunction. The data supporting the conclusion are solid. This work will be of interest to biomedical researchers working on testis biology and male infertility. The assessment is based on the editors' critical evaluation of the authors' responses.

Reviewer #1 (Public review):

The authors have strengthened their conclusions by providing additional information about the specificity of their antibodies, but at the same time the authors have revealed concerning information about the source of their antibodies.

It appears that many of the antibodies used in this study have been discontinued because the supplier company was involved in a scandal of animal cruelty and all their goats and rabbits Ab products were sacrificed. The authors acknowledge that this is unfortunate but they also claim that the issue is out of their hands.

The authors' statement is false; the authors ought to not use these antibodies, just as the providing company chose to discontinue them, as<br /> those antibodies are tied to animal cruelty. The issue that the authors feel OK with using them is of concern. In short, please remove any results from unethical antibodies.

Removal of such results also best serves science. That is, any of their results using the discontinued antibodies means that the authors' results are non-reproducible and we should be striving to publish good, reproducible science.

For the antibodies that do not have unethical origins the authors claim that their antibodies have been appropriately validated, by "testing in positive control tissue and/or Western blot or in situ hybridization". This is good but needs to be expanded upon. It is a strong selling point that the Abs are validated and I want to see additional information in their Supplementary Table 2 stating for each Ab specifically:

(1) What +ve control tissue was used in the validation of each Ab and which species that +ve control came from. Likewise, if competition assays to confirm validity was used, please also specify.

(2) Which assay was the Ab validated for (WB, IHC, ELISA, all etc)

(3) For Antibodies that were validated for, or using WBs please let the reader know if there were additional bands showing.

(4) Include references to the literature that supports these validations. That is, please make it easy for the reader to appreciate the hard work that went into the validation of the Antibodies.

Finally, for the Abs, when the authors write that "All antibodies used have been validated by testing in positive control tissue and/or Western blot or in situ hybridization" I fail to understand what in situ hybridisation means in this context. I am under the impression that in situ hybridisation is some nucleic acid -hybridising-to-organ or tissue slice. Not polypeptide binding.

Author response:

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

Reviewer #1 (Public review):

The authors have strengthened their conclusions by providing additional information about the specificity of their antibodies, but at the same time the authors have revealed concerning information about the source of their antibodies.

It appears that many of the antibodies used in this study have been discontinued because the supplier company was involved in a scandal of animal cruelty and all their goats and rabbits Ab products were sacrificed. The authors acknowledge that this is unfortunate but they also claim that the issue is out of their hands.

The authors' statement is false; the authors ought to not use these antibodies, just as the providing company chose to discontinue them, as those antibodies are tied to animal cruelty. The issue that the authors feel OK with using them is of concern. In short, please remove any results from unethical antibodies.

Removal of such results also best serves science. That is, any of their results using the discontinued antibodies means that the authors' results are non-reproducible and we should be striving to publish good, reproducible science.

For the antibodies that do not have unethical origins the authors claim that their antibodies have been appropriately validated, by "testing in positive control tissue and/or Western blot or in situ hybridization". This is good but needs to be expanded upon. It is a strong selling point that the Abs are validated and I want to see additional information in their Supplementary Table 2 stating for each Ab specifically:

(1) What +ve control tissue was used in the validation of each Ab and which species that +ve control came from. Likewise, if competition assays to confirm validity was used, please also specify.

(2) Which assay was the Ab validated for (WB, IHC, ELISA, all etc)

(3) For Antibodies that were validated for, or using WBs please let the reader know if there were additional bands showing.

(4) Include references to the literature that supports these validations. That is, please make it easy for the reader to appreciate the hard work that went into the validation of the Antibodies.

Finally, for the Abs, when the authors write that "All antibodies used have been validated by testing in positive control tissue and/or Western blot or in situ hybridization" I fail to understand what in situ hybridisation means in this context. I am under the impression that in situ hybridisation is some nucleic acid -hybridising-to-organ or tissue slice. Not polypeptide binding.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Remove results that have been obtained by unethically-sourced antibody reagents.

Strengthen the readers' confidence about the appropriateness & validity of your antibodies.

First, we want to stress that reviewer 1 has raised his critique related to the used of antibodies from Santa Cruz biotechnology not only through the journal. The head of our department and two others were contacted by reviewer 1 directly without going through the journal or informing/approaching the corresponding or first author. It is our opinion that this debate and critique should be handled through the journal and editorial office and not with people without actual involvement in the project.

It is correct that we have purchased antibodies from Santa Cruz Biotechnologies both mouse, rabbit and goat antibodies as stated in the correspondence with the reviewer.

As stated in our previous rebuttal – the goat antibodies from Santa Cruz were discontinued due to inadequate treatment of goats after settling with the authorities in 2016.

https://www.nature.com/articles/nature.2016.19411

https://www.science.org/content/blog-post/trouble-santa-cruz-biotechnology

We have used 11 mouse, rabbit or goat antibodies from Santa Cruz biotechnologies in the manuscript as listed in supplementary table 2 of the manuscript and all of them have been carefully validated in other control tissues supported by ISH and/or WB and many of them already used in several publications by our group (https://pubmed.ncbi.nlm.nih.gov/34612843/, https://pubmed.ncbi.nlm.nih.gov/33893301/, https://pubmed.ncbi.nlm.nih.gov/32931047/, https://pubmed.ncbi.nlm.nih.gov/32729975/, https://pubmed.ncbi.nlm.nih.gov/30965119/, https://pubmed.ncbi.nlm.nih.gov/29029242/, https://pubmed.ncbi.nlm.nih.gov/23850520/, https://pubmed.ncbi.nlm.nih.gov/23097629/, https://pubmed.ncbi.nlm.nih.gov/22404291/, https://pubmed.ncbi.nlm.nih.gov/20362668/, https://pubmed.ncbi.nlm.nih.gov/20172873/,  and other research groups. All antibodies used in this manuscript were purchased before the whole world was aware of mistreatment of goats that was evident several years later.

We do not support animal cruelty in anyway but the purchase of antibodies from Santa Cruz biotechnologies were conducted long before mistreatment was reported. Moreover, antibodies from Santa Cruz biotechnologies are being used in thousands of publications annually. The company has been punished for their misconduct, and subsequently granted permission to produce antibodies from the relevant authorities again.

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Despite the study being a collation of important results likely to have an overall positive effect on the field, methodological weaknesses and suboptimal use of statistics make it difficult to give confidence to the study's message.

Strengths:

Relevant human and mouse models approached with in vivo and in vitro techniques.

Weaknesses:

The methodology, statistics, reagents, analyses, and manuscripts' language all lack rigour.

(1) The authors used statistics to generate P-values and Rsquare values to evaluate the strength of their findings.

However, it is unclear how stats were used and/or whether stats were used correctly. For instance, the authors write: "Gaussian distribution of all numerical variables was evaluated by QQ plots". But why? For statistical tests that fall under the umbrella of General Linear Models (line ANOVA, t-tests, and correlations (Pearson's)), there are several assumptions that ought to be checked, including typically:

(a) Gaussian distribution of residuals.

(b) Homoskedasticity of the residuals.

(c) Independence of Y, but that's assumed to be valid due to experimental design.

So what is the point of evaluating the Gaussian distribution of the data themselves? It is not necessary. In this reviewer's opinion, it is irrelevant, not a good use of statistics, and we ought to be leading by example here.

Additionally, it is not clear whether the homoscedasticity of the residuals was checked. Many of the data appear to have particularly heteroskedastic residuals. In many respects, homoscedasticity matters more than the normal distribution of the residuals. In Graphpad analyses if ANOVA is used but equal variances are assumed (when variances among groups are unequal then standard deviations assigned in each group will be wrong and thus incorrect p values are being calculated.

Based on the incomplete and/or wrong statistical analyses it is difficult to evaluate the study in greater depth.

We agree with the reviewer that we should lead by example and improve clarity on the use of the different statistical tests and their application. In response to the reviewer’s suggestion, we have extended the statistical section, focusing on the analyses used. Additionally, we have specified the statistical test used in the figure legends for each figure. Additionally, we did check for Gaussian distribution and homoskedasticity of residuals before conducting a general linear model test, and this has now been specified in the revised manuscript. In case the assumptions were not met, we have specified which non-parametric test we used. If the assumptions were not met, we specified which non-parametric test was used.

While on the subject of stats, it is worth mentioning this misuse of statistics in Figure 3D, where the authors added the Slc34a1 transcript levels from controls in the correlation analyses, thereby driving the intercept down. Without the Control data there does not appear to be a correlation between the Slc34a1 levels and tumor size.

We agree with the reviewer that a correlation analysis is inappropriate here and have removed this part of the figure.

There is more. The authors make statements (e.g. in the figure levels as: "Correlations indicated by R2.". What does that mean? In a simple correlation, the P value is used to evaluate the strength of the slope being different from zero. The authors also give R2 values for the correlations but they do not provide R2 values for the other stats (like ANOVAs). Why not?

We agree with the reviewer and have replaced the R2 values with the Pearson correlation coefficient in combination with the P value.

(2) The authors used antibodies for immunos and WBs. I checked those antibodies online and it was concerning:

(a) Many are discontinued.

Many of the antibodies we have used were from the major antibody provider Santa Cruz Biotechnology (SCBT). SCBT was involved in a scandal of animal cruelty and all their goats and rabbits were sacrificed, which explains why several antibodies were discontinued, while the mice antibodies were allowed to continue. This is unfortunate but out of our hands.

(b) Many are not validated.

We agree with the reviewer that antibody validation is essential. All antibodies used in this manuscript have been validated. The minimal validation has been to evaluate cellular expression in positive control tissue for instance bone, kidney, or mamma. Moreover, many of the antibodies have been used and validated in previous publications (doi: 10.1593/neo.121164, doi:10.1096/fj.202000061RR, doi: 10.1093/cvr/cvv187) including knockout models. Moreover, many antibodies but not all have been validated by western blot or in situ hybridization. We have included the following in the Materials and Methods section: “All antibodies used have been validated by testing in positive control tissue and/or Western blot or in situ hybridization”.

(c) Many performed poorly in the Immunos, e.g. FGF23, FGFR1, and Kotho are not really convincing. PO5F1 (gene: OCT4) is the one that looks convincing as it is expressed at the correct cell types.

We fail to understand the criticism raised by the reviewer regarding the specificity of these specific antibodies. We believe the FGF23 and Klotho antibodies are performing exceptionally well, and FGFR1 is abundantly expressed in many cell types in the testis. As illustrated in Figure 2E, the expression of Klotho, FGF23, and FGFR1 is very clear, specific, and convincing. FGF23 is not expressed in normal testis – which is in accordance with no RNA present there either. However, it is abundantly expressed in GCNIS where RNA is present. On the other hand, Klotho is abundantly expressed in germ cells from normal testis but not expressed in GCNIS.

(d) Others like NPT2A (product of gene SLC34A1) are equally unconvincing. Shouldn't the immuno show them to be in the plasma membrane?

If there is some brown staining, this does not mean the antibodies are working. If your antibodies are not validated then you ought to omit the immunos from the manuscript.

We acknowledge your concerns regarding the NPT2A, NPT2B, and NPT2C staining. While the NPT2A antibody is performing well, we understand your reservations about the other antibodies. It's worth noting that NPT2A is not expressed in normal testis (no RNA either) but is expressed in GCNIS where the RNA is also present. Although it is typically present in the plasma membrane, cytoplasmic expression can be acceptable as membrane availability is crucial for regulating NPT2A function, particularly in the kidney where FGF23 controls membrane availability. We are currently involved in a comprehensive study exploring these phosphate transporters in the organs lining the male reproductive tract. In functional animal models, we have observed very specific staining with this NPT2A antibody following exposed to high phosphate or FGF23. Additionally, we are conducting Western Blot analyses with this antibody, which reinforces our belief that the antibody has a specific binding.

Reviewer #2 (Public Review):

Summary:

This study set out to examine microlithiasis associated with an increased risk of testicular germ cell tumors (TGCT). This reviewer considers this to be an excellent study. It raises questions regarding exactly how aberrant Sertoli cell function could induce osteogenic-like differentiation of germ cells but then all research should raise more questions than it answers.

Strengths:

Data showing the link between a disruption in testicular mineral (phosphate)homeostasis, FGF23 expression, and Sertoli cell dysfunction, are compelling.

Weaknesses:

Not sure I see any weaknesses here, as this study advances this area of inquiry and ends with a hypothesis for future testing.

We thank the reviewer for the acknowledgment and highlighting that this is an important message that addresses several ways to develop testicular microlithiasis, which indicates that it is not only due to malignant disease but also frequent in benign conditions.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

I applaud the authors' approach to nomenclature for rodent and human genes and proteins (italicised for genes, all caps for humans, capitalised only for rodents, etc), but the authors frequently got it wrong when referring to genes or proteins. A couple of examples include:

(1) SLC34A1 (italics) refers to gene (correct use by the authors) but then again the authors use e.g. SLC34A1 (not italics) to refer to the protein product of SLC34A1(italics) gene. In fact, the protein product of the SLC34A1 (italics) gene is called NPT2A (non-italics).

(2) OCT4 (italics) refers to gene (correct use by the authors) but then again the authors use e.g. OCT4 (not italics) to refer to the protein product of OCT4 (italics)gene. In fact, the protein product of the OCT4 gene (italics) gene is called PO5F1(non-italics).

The problem with their incorrect and inconsistent nomenclature is widespread in the manuscript making further evaluation difficult.

Please consult a reliable protein-based database like Uniprot to derive the correct protein names for the genes. You got NANOG correct though.

We thank the reviewer for addressing this important point. We have corrected the nomenclature throughout the manuscript as suggested.

(3) The authors use the word "may" too many times. Also often in conjunction with words like "indicates", and "suggests". Examples of phrases that reflect that the authors lack confidence in their own results, conclusions, and understanding of the literature are:

"...which could indicate that the bone-specific RUNX2 isoform may also be expressed... "

"...which indicates that the mature bone may have been..."

Are we shielding ourselves from being wrong in the future because "may" also means "may not"? It is far more engaging to read statements that have a bit more tooth to them, and some assertion too. How about turning the above statements around, to :

"...which shows that the bone-specific RUNX2 isoform is also expressed... "

"...which reveals that the mature bone were..."

...then revisit ambiguous language ("may", "might" "possibly", "could", "indicate" etc.) throughout the manuscript?

It's OK to make a statement and be found wrong in the future. Being wrong is integral to Science.

Thank you for addressing this. We agree with the reviewer that it is fair to be more direct and have revised many of these vague phrases throughout the manuscript.

(4) The authors use the word "transporter" which in itself is confusing. For instance, is SLC34A1 an importer or an exporter of phosphate? Or both? Do SLC34As move phosphate in or out of the cells or cellular compartments? "Transporter" sounds too vague a word.

We understand that it might be easier for the reader with the term "importer". However, we should use the specific nomenclature or "wording" that applies to these transporters. The exact terminology is a co-transporter or sodium-dependent phosphate cotransporter as reported here (doi: 10.1152/physrev.00008.2019). Thus, we will use the terms “co-transporter” and “transporter” throughout the revised manuscript.

eLife Assessment

This study investigates a dietary intervention that employs a smartphone app to promote meal regularity, findings that have theoretical or practical implications for a subfield and may be clinically useful. The intervention to entice participants to adhere to specific meal times represents a restrictive diet (even though it does not ask to limit caloric intake) similar to a time-restricted feeding diet, while the control subjects are not experiencing or adhering to dietary restrictions. The authors report significant weight loss but did not rigorously assess caloric intake which remains a weakness of this study as food diaries are notoriously unreliable. While the concept is very interesting, the study is considered incomplete, and the rigor of the results should be strengthened in follow-up studies to add more stringent methods to assess caloric intake. Additionally, the study hypothesizes that the intervention resets the circadian clock. However, the study needs an objective method for assessing circadian rhythms, such as actigraphy, in addition to a subjective questionnaire.

Reviewer #3 (Public review):

In this study, the authors tested a dietary intervention focused on improving meal regularity. Participants first utilized a smartphone application to track their meal frequencies, and then they were asked to restrict their meal intake to times when they most often eat to enhance meal regularity for six weeks. This, supposedly, resulted in some weight loss, supposedly independent of changes in caloric intake.

The concept is appealing, and it is interesting to use a smartphone app in participants' typical everyday environment to regularize food intake. It asks from participants to stick to meal intake times that are supported in many cultures, and it asks them not to eat outside of what are likely unhealthy habits such as grazing a refrigerator late at night. In essence, this is a restrictive diet, not restricting caloric intake but the timing of food intake, and it has many parallel to time restricted feeding. It is important to note that there are many restrictive diets, and a common problem with restrictive diets is that while they allow one to lose a couple of pounds for a couple of months just as with this diet, the long-term success is very poor because they depend on restriction. This issue is still not discussed.

Further, why the participants lose weight, whether this is indeed due to a reduction in food intake as implied, or if the weight loss occurred without a reduction in caloric intake as first stated by the authors and now suggested remains to be determined as the method of food diary as a method to assess caloric intake lacks rigor as has been well established and has been shown again and again to be misleading even though many readers without that knowledge draw conclusions from such studies and they should best have been omitted.

The authors hypothesize that the intervention improves metabolism by improving circadian rhythmicity. That's plausible, but the study provides only a subjective questionnaire and lacks more objective measures such as actigraphy.

While the authors now state now that this as a pilot study, the study falls short of providing mechanistic insights into what underlies the weight loss and the many correlations provided do not make up for this weakness.

Overall, while this pilot study introduces an interesting approach to meal regularity, its limitations highlight the need for more rigorous studies to validate these findings.

(1) Unreliable method of caloric intake

The trial's reliance on self-reported caloric intake is problematic, as participants tend to underreport intake. As pointed out earlier by me and now cited in the revised manuscript, the NEJM paper (DOI: 10.1056/NEJM199212313272701) reported that some participants underreported caloric intake by approximately 50%, rendering such data unreliable and hence misleading. The question is, why include such unreliable data that is more misleading than informative at all? These data should have been omitted. More rigorous methods for assessing food intake should have been utilized. I understand this requires more effort, such as providing participants with meals, or using better methods that photograph and weigh the meals, etc., but it is certainly feasible. It has been done many times in other studies. Further, the control group was not asked to restrict their diet in any way, and hence, asking for a restriction in timing in the treatment group may be sufficient to reduce caloric intake and induce weight loss.<br /> Merely acknowledging the unreliability of self-reported caloric intake is insufficient, as it still leaves the reader with the impression that this weight loss is independent of caloric intake when, in reality, we actually have no idea if food intake contributes to it. A more robust approach to assessing food intake is imperative. Even if a decrease in caloric intake is observed through rigorous measurement, as I am convinced a more rigorous study would unveil testing this paradigm, this intervention may merely represent another restrictive diet among countless others that show that one may lose weight by going on a diet. Seemingly, any restrictive diet works for a few months. The trouble is they do not work long-term because they depend on restriction. I agree with the authors that their intervention seems common sense and has little downside, but one also needs to be realistic about the prospects of this intervention.

(2) Lack of objective data regarding circadian rhythm

The assessment of circadian rhythm using the MCTQ, a self-reported measure of chronotype, is subjective. More objective methods like actigraphy would have strengthened the study.

Actigraphy is considered better than a sleep questionnaire for assessing circadian rhythms because it provides objective data on activity patterns over time, offering a more accurate picture of sleep-wake cycles compared to subjective self-reported information from a questionnaire.

The authors' responses to my prior review are misleading.

I understand that this is a pilot study. Is it appropriate to point out weaknesses and flaws in the conclusion drawn from a pilot study? Absolutely, that is the reviewer's job.

I also understand that food intake can affect circadian rhythm, which was part of the rationale behind the study. Is it appropriate to criticize the study for not examining the effect of the intervention on circadian rhythm using objective measures provided by actigraphy? Yes, it is, as this would have provided mechanistic insights that are more rigorous. I understand that this was not the declared goal, but it should have been examined in a pilot study. To jump to the conclusion that based on prior studies, the intervention will improve circadian rhythms as the authors do is not rigorous and hence a weakness.

A less rigorous method, such as a food questionnaire, to assess caloric intake can result in inadequately supported and potentially misleading conclusions. By including it, the reader may conclude that there was no change in caloric intake when indeed we do not know. I disagree with the authors that this is a minor issue. The associations and correlations the authors provide do not solve the issue. Hence, to make it very clear, it remains to be studied if this intervention reduces weight by reducing caloric intake or other mechanisms. Including this data reduces the study's rigor as it suggests that there is no difference in food intake.

I did not suggest to only use an actimeter (which is a device); I suggested actigraphy. Actigraphy is widely recognized in the field for its utility in circadian rhythm research and provides objective data, while the questionnaire used is subjective. The authors do quote papers comparing their survey to actigraphy by correlation analysis, but the fundamental difference of the two approaches remains. Does an objective measure increase rigor compared to a subjective assessment? Yes, it does.

Similarly, I did not state "that any form of imposed diet appears to lead to weight loss over several months." I said that many forms of restrictive diets do induce weight loss of a similar magnitude to this diet.

The authors should have discussed the fundamental confounder of the study in that the treatment group is asked to restrict food intake to specific times while the control group is not asked to restrict in any way and the potential contribution of this to the weight loss observed.

Author response:

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

We would like to remind the editors and reviewers that the present project is a pilot study that does not claim to produce definitive results. Pilot studies are exploratory preliminary studies to test the validity of hypotheses, the feasibility of a study as well as the research methods and the study design. From our point of view, our hypotheses and the feasibility of the pilot study have been confirmed to such an extent that the implementation of a larger study is justified. At the same time, it became clear during the pilot that the methods and design need to be adapted in some areas in order to increase the reliability of the results - a finding that pilot studies are usually conducted to obtain. We discussed these limitations in detail in order to explain the planned changes in the follow-up study. What the reviewers and editors interpret as incompleteness is therefore due to the nature of a pilot study.  We consider it necessary that appropriate standards are taken into account in the evaluation of the present work.

In addition, we would like to make a counterstatement as to what our main claims, which should be used to assess the strength of evidence, are - and what they are not:

In the introduction, we describe the background that led to the formation of our hypotheses: Previous animal and human studies show that food, along with light, serves as the main Zeitgeber for circadian clocks. It has also been shown that chrononutrition can lead to weight loss and improved well-being. Based on this, we hypothesized that individualized meal timing can enhance these positive effects. This hypothesis has been validated on the basis of the available results. Contrary to what the editors and reviewers stated, the assumption that the observed beneficial effects are indeed related to an alteration or resetting of endogenous circadian rhythms was not intended to be investigated in this study and is not one of our main claims. This has already been sufficiently demonstrated and, in our view, need not and should not be repeated in every study on chrononutrition. Accordingly, this assumption was not formulated as a working hypothesis or main claim. It is described in the paper as a potential mechanism, the assumption of which is justified on the basis of previous studies. The lack of a corresponding examination and the erroneous insinuation that corresponding results were nevertheless listed by us in the paper as a main claim should therefore not be used as a criterion for downgrading the assessment of the strength of evidence.

The main criticism of our study is the collection of data using self-reported food and food quantities. This form of data collection is indeed prone to error, as there is little control over the accuracy of the reported data. However, we believe that this problem is limited in scope.

(1) Contrary to what the editors and reviewers claim, at no point do we write that we are convinced that food intake has not changed. On the contrary, in Figure 2 we explicitly show that there was a change in what some participants reported to us regarding their food intake. We make it clear throughout the text that we could not find any correlation between weight change and the changes in the reports of food quantities/meals. These statements are correct and only what are actual and formulated main claims should be included in the evaluation of the study.

(2) As previously stated, we conducted analyses that suggest that an unreported reduction in food intake is unlikely to be the cause of weight loss. For the most part, participants did not change their reporting behavior during the exploration and intervention phases. That is, participants who underreported food intake reported similar amounts in both phases of the study, but lost weight only in the intervention phase. To explain their weight loss with imprecise reporting, it would have to be assumed that these participants began to eat less in the intervention phase and at the same time report more in order to achieve similar calorie counts and food composition in the evaluation. We consider such behavior to be very unlikely, especially since it would apply to numerous participants.

(3) The editors and reviewers reduce the results to the absence of a correlation between weight loss and reported food quantity and composition. In their assessment of the significance of the findings, however, they ignore the fact that we did find a significant correlation in our analyses, namely between weight loss and an increase in the regularity of food intake. There is no correlation between an increase in regularity and a reduction in reported calories (R² = 0.01472). This is credible in our view, as it is unlikely that the more regularly participants ate, the more pronounced the error in their reports was (while in reality they ate less than before).

(4) We also had the requirement for the study design that the participants could carry out the intervention in their normal everyday life and environment in order to test and ensure implementation in real life. We consider it unrealistic to be able to monitor food intake continuously and without interruption over a period of several weeks under these conditions. We therefore see no alternative to self-reporting. As the reviewers and editors did not suggest any alternative methods of data collection that would fulfil the requirements of our study, we assume that, despite criticism and reservations, they generally agree with our assessment and take this into account in their evaluation.

It is still criticized that some confounding factors are present. The reviewer makes no reference to the fact that we either eliminated these in the last version submitted (age range), identified them as unproblematic (unmatched cohorts, menstrual cycle, shift work) or even deliberately used them in order to be able to test our hypothesis more validly (inclusion of individuals with normal weight, overweight, and obesity).

Besides, the use of actimeters to determine circadian rhythms as proposed by the editors and reviewers is not valid for this study and the requirement to use them to determine a circadian reset in the eLife assessment is misleading and inappropriate. This instrument only measures physical activity, but not the physiological parameters that are relevant for an investigation in this field of research.

For the assessment of chronotype alone, the MCTQ questionnaire is a valid instrument that has been validated several times against actimetry (e.g., DOIs: 10.1080/07420528.2022.2025821, 10.1080/07420528.2023.2202246, 10.1016/j.ijpsycho.2016.07.433, 10.1155/2018/5646848). The reviewer's statement that the MCTQ questionnaire is unreliable for determining chronotype is unsupported and incorrect.

Equally unproven is the statement that any form of imposed diet appears to lead to weight loss over a period of several months.

Nevertheless, in order to prevent further misunderstandings, we have revised our text in a number of places and clarified that our statements are not irrefutable assertions, but potential interpretations of the results obtained in the pilot study, which are to be analyzed in more detail with regard to the planned more comprehensive study.

eLife Assessment

This study provides a comprehensive exploration of the role of IL-1β signaling during development of lung injury induced by a combination of underlying inflammation and mechanical ventilation. The data are convincing, and while the translatability of the findings related to therapeutic hypothermia may be somewhat complicated, they have the potential to be very valuable to the field.

Reviewer #1 (Public review):

Summary:

The authors found that IL-1b signaling is pivotal for hypoxemia development and can modulate NETs formation in LPS+HVV ALI model.

Strengths:

They used IL1R1 ko mice and proved that IL1R1 is involved in ALI model proving that IL1b signalling leads towards ARDS. In addition, hypothermia reduces this effect, suggesting a therapeutic option.

Weaknesses:

(1) IL1R1 binds IL1a and IL1b. What would be the role of IL1a in this scenario?

(2) The authors depleted neutrophils using anti-Ly6G. What about MDSCs? Do these latter cells be involved in ARDS and VILI?

(3) The authors found that TH inhibited IL-1β release from macrophages led to less NETs formation and albumin leakage in the alveolar space in their lung injury model. A graphical abstract could be included suggesting a cellular mechanism.

(4) If Macrophages are responsible for IL1b release that via IL1R1 induces NETosis, what happens if you deplete macrophages? what is the role of epithelial cells?

Reviewer #2 (Public review):

Summary:

The manuscript by Nosaka et al is a comprehensive study exploring the involvement of IL1beta signaling in a 2-hit model of lung injury + ventilation, with a focus on modulation by hypothermia.

Strengths:

The authors demonstrate quite convincingly that interleukin 1 beta plays a role in the development of ventilator-induced lung injury in this model, and that this role includes the regulation of neutrophil extracellular trap formation. The authors use a variety of in vivo animal-based and in vitro cell culture work, and interventions including global gene knockout, cell-targeted knockout and pharmacological inhibition, which greatly strengthen the ability to make clear biological interpretations.

Weaknesses:

A primary point for open discussion is the translatability of the findings to patients. The main model used, one of intratracheal LPS plus mechanical ventilation is well accepted for research exploring the pathogenesis and potential treatments for acute respiratory distress syndrome (ARDS). However, the interpretation may still be open to question - in the model here, animals were exposed to LPS to induce inflammation for only 2 hours, and seemingly displayed no signs of sickness, before the start of ventilation. This would not be typical for the majority of ARDS patients, and whether hypothermia could be effective once substantial injury is already present remains an open question. The interaction between LPS/infection and temperature is also complicated - in humans, LPS (or infection) induces a febrile, hyperthermic response, whereas in mice LPS induces hypothermia (eg. Ganeshan K, Chawla A. Nat Rev Endocrinol. 2017;13:458-465). Given this difference in physiological response, it is therefore unclear whether hypothermia in mice and hypothermia in humans are easily comparable. Finally, the use of only young, male animals such as in the current study has been typical but may be criticised as limiting translatability to people.

Therefore while the conclusions of the paper are well supported by the data, and the biological pathways have been impressively explored, questions still remain regarding the ultimate interpretations.

Author response:

Public Reviews: 

Reviewer #1 (Public review): 

Summary: 

The authors found that IL-1b signaling is pivotal for hypoxemia development and can modulate NETs formation in LPS+HVV ALI model.  

Strengths: 

They used IL1R1 ko mice and proved that IL1R1 is involved in ALI model proving that IL1b signalling leads towards ARDS. In addition, hypothermia reduces this effect, suggesting a therapeutic option.  

We thank the Reviewer for recognizing the strengths of our study and their positive feedback.

Weaknesses: 

(1) IL1R1 binds IL1a and IL1b. What would be the role of IL1a in this scenario? 

Thank you for asking this question. We have addressed this in our previous paper (Nosaka et al. Front Immunol 2020;11; 207) where we used  anti-IL-1a and IL-1a KO mice (Nosaka et al. Front Immunol 2020;11; 207) in our model and found that neither anti-IL-1a treated mice nor IL-1a KO mice were protected. Thus, IL-1b plays a role in inducing hypoxemia during LPS+HVV but not IL-1a. We will now add this point in our revised manuscript discussion.

(2) The authors depleted neutrophils using anti-Ly6G. What about MDSCs? Do these latter cells be involved in ARDS and VILI?  

Anti-Ly6G neutrophils depletion may potentially affect G-MDSCs as well (Blood Adv 2022 Jul 29;7(1):73–86), however, we have not looked directly at G-MDSCs.  If these cells were depleted we would have expected to see an increase in inflammation, which we did not.   

Instead, anti-Ly6G treated mice were protected. Thus, we can not comment on any presumed role of G-MDSCs in LPS+HVV induced severe ALI model that we used.  

(3) The authors found that TH inhibited IL-1β release from macrophages led to less NETs formation and albumin leakage in the alveolar space in their lung injury model. A graphical abstract could be included suggesting a cellular mechanism.  

Thanks for summarizing our findings and the suggestion. Unfortunately, eLIFE does not publish a graphical abstract. We tried to mention this mechanism in the discussion.

(4) If Macrophages are responsible for IL1b release that via IL1R1 induces NETosis, what happens if you deplete macrophages? what is the role of epithelial cells?  

Previous studies have found that macrophage depletion is protective in several models of ALI (Eyal. Intensive Care Med. 2007;33:1212–1218., Lindauer.  J Immunol. 2009;183:1419–1426.), and other researchers have found that airway epithelial cells did not contribute to IL-1β secretion (Tang. PLoS ONE. 2012;7:e37689.). We have previously reported that epithelial cells produce IL-18 without LPS priming signal during LPS+HVV (Nosaka et al. Front Immunol 2020;11; 207). Thus, IL-18 is not sufficient to induce Hypoxemia as Saline+HVV treated mice do not develop hypoxemia (Nosaka et al. Front Immunol 2020;11; 207). We will now add this point to the revised discussion of the manuscript.

Reviewer #2 (Public review): 

Summary: 

The manuscript by Nosaka et al is a comprehensive study exploring the involvement of IL1beta signaling in a 2-hit model of lung injury + ventilation, with a focus on modulation by hypothermia. 

Strengths: 

The authors demonstrate quite convincingly that interleukin 1 beta plays a role in the development of ventilator-induced lung injury in this model, and that this role includes the regulation of neutrophil extracellular trap formation. The authors use a variety of in vivo animal-based and in vitro cell culture work, and interventions including global gene knockout, cell-targeted knockout and pharmacological inhibition, which greatly strengthen the ability to make clear biological interpretations. 

We thank the Reviewer for their positive feedback 

Weaknesses: 

A primary point for open discussion is the translatability of the findings to patients. The main model used, one of intratracheal LPS plus mechanical ventilation is well accepted for research exploring the pathogenesis and potential treatments for acute respiratory distress syndrome (ARDS). However, the interpretation may still be open to question - in the model here, animals were exposed to LPS to induce inflammation for only 2 hours, and seemingly displayed no signs of sickness, before the start of ventilation. This would not be typical for the majority of ARDS patients, and whether hypothermia could be effective once substantial injury is already present remains an open question. The interaction between LPS/infection and temperature is also complicated - in humans, LPS (or infection) induces a febrile, hyperthermic response, whereas in mice LPS induces hypothermia (eg. Ganeshan K, Chawla A. Nat Rev Endocrinol. 2017;13:458-465). Given this difference in physiological response, it is therefore unclear whether hypothermia in mice and hypothermia in humans are easily comparable. Finally, the use of only young, male animals such as in the current study has been typical but may be criticised as limiting translatability to people. 

Therefore while the conclusions of the paper are well supported by the data, and the biological pathways have been impressively explored, questions still remain regarding the ultimate interpretations.  

We agree with the reviewer that at two hours post LPS, there is only minimal pulmonary inflammation at that time (Dagvadorj et al Immunity 42, 640–653). This is a limitation to the experimental model we used in our study. Additionally, as the reviewer pointed out that LPS induces hyperthermia in human, but it is also well-established that physiological hypothermia occurs in humans with severe infections and sepsis (Baisse. Am J Emerg Med. 2023 Sep: 71: 134-138., Werner.  Am J Emerg Med. 2025 Feb;88:64-78.). Therefore, the difference between human and mouse responses to sepsis or infections may be more nuanced.  Furthermore, it is important to distinguish between physiological hypothermia (just <36°C) and therapeutic hypothermia (typically 32-34°C). We will add to the discussion whether hypothermia serves as a protective response, and the transition from normothermia to hyperthermia could have detrimental effects. We only used young male mice in our study as the Reviewer points out; we will also add this point to the revised discussion as a limitation of our study.

eLife Assessment

This study highlights ITCH as a regulator of SARS-CoV-2 replication by promoting K63-linked ubiquitination of M and E proteins. While the findings are potentially useful, the approaches are overly reliant on ectopic expression models and lack direct mechanistic evidence that ubiquitination of M and E has functional relevance. Accordingly, the strength of evidence is incomplete, as further experiments are needed to validate the findings and address potential confounding factors.

Reviewer #1 (Public review):

Summary:

The authors investigated the role of an E3 ubiquitin ligase ITCH in regulating the viral life cycle of SARS-CoV-2. The authors showed that ITCH mediates ubiquitination of the membrane (M) and envelope (E) proteins of SARS-CoV-2. Ubiquitination of E and M results in enhanced interactions between the structural proteins and redistribution of the structural proteins into autophagosomes. The authors claim that the enhanced interactions between structural proteins and trafficking of the structural proteins into autophagosomes contribute to SARS-CoV-2 replication and egress, prompting ITCH as a potential antiviral target. ITCH also alters the cellular distribution of host proteases important for spike cleavage which protect and stabilize spike with cleavage. The authors also demonstrated that SARS-CoV-2 replication is augmented by ITCH in which virus replication is significantly impaired in cells lacking ITCH expression.

Strengths:

The authors provided high-quality data with appropriate experimental controls to justify their claims and conclusions. The mechanistic analyses are excellent and presented in a logical manner. The investigation of the role of ubiquitination in coronavirus assembly and egress is novel as most previous studies focused on its role in mediating innate immune responses.

Weaknesses:

Although the authors showed that ITCH ubiquitinates E and M proteins, the claim that such ubiquitination promotes virion assembly and egress is circumstantial. The enhanced interaction between the structural proteins and targeting of ubiquitinated structural proteins into autophagosomes does not necessarily result in increased virion production and release as suggested by the authors. There is a disconnect between the ubiquitination of structural proteins and the role of ITCH in augmenting virus replication as shown in Fig. 6A and B. In addition, the authors showed that the catalytic activity of ITCH is important for the localization and maturation of host proteases. However, the mechanism behind is unknown. Also, it is unclear how protection of spike from cleavage conferred by ITCH explains its role in promoting replication as a lack of spike cleavage would inevitably compromise entry. The major weakness of the manuscript is the lack of experimental data that explains the molecular role of ITCH in relation to its phenotype observed during SARS-CoV-2 infection.

Reviewer #2 (Public review):

Summary:<br /> In this manuscript Qiwang Xiang et al. investigated the role of the E3 ubiquitin ligase ITCH in the life cycle of SARS-CoV-2. They claim the following:<br /> i) ITCH promotes virion assembly by interacting with E and M proteins and enhancing their K63-linked ubiquitination<br /> ii) ITCH-mediated ubiquitination promotes autophagosome-dependent secretion of viral particles.<br /> iii) ITCH stabilizes the viral spike protein by impairing its processing by furin and catepsin L proteases.<br /> The manuscript provides an interesting exploration of ITCH's role in the SARS-CoV-2 life cycle but requires additional work to strengthen key claims and address potential confounding factors.

Strengths:

The experiments are sufficiently clear in documenting that ITCH activity is critical for efficient SARS-CoV-2 replication and for M and E proteins K63-linked ubiquitination

Weaknesses:

• The manuscript does not convincingly demonstrate how ITCH-mediated ubiquitination of E and M impacts virus assembly and release. Identifying the specific lysine residues in M and E targeted by ITCH, and generating mutant VLPs or recombinant viruses, would strengthen the conclusions.<br /> • Most of the conclusions rely on ITCH overexpression data, which may have off-target effects on Golgi integrity and vesicular trafficking. For instance, figure 4F provides evidence of altered Golgi morphology and TGN46 fragmentation raising concerns that ITCH overexpression could indirectly mislocalize furin, affecting S1/S2 cleavage of the spike protein. In addition, inhibition of furin activity may also lead to off-target effects, given its role in processing numerous host proteins.<br /> • Similarly, ITCH overexpression is likely to indirectly affect cathepsin-L maturation. In addition, the manuscript does not clarify how impaired cathepsin L activity would influence virus assembly or release.<br /> • A major concern is also the lack of quantification and statistical analysis of immunofluorescence images throughout the manuscript, which undermines the reliability of these observations.

Reviewer #3 (Public review):

Summary:

Xiang et al. investigated the role of ubiquitin E3 ligase ITCH in SARS-CoV-2 replication. First, they described the role of ITCH on the structural proteins. Here, the ubiquitination of E and M (but not S) leads to an enhanced interaction and presumably virion assembly. In addition, E and M ubiquitination seems to be necessary for p62-guided sequestration into autophagosomes for secretion. Furthermore, ITCH regulates S proteolytic cleavage by changing furin localization and inhibiting CTSL protease maturation. In addition, SARS-CoV-2 infection upregulates ITCH phosphorylation, whereas knockout of ITCH reduces SARS-CoV-2 replication.

Strengths:

The proposed study is of interest to the virology community because it aims to elucidate the role of ubiquitination by ITCH in SARS-CoV-2 proteins. Understanding these mechanisms will address broadly applicable questions about coronavirus biology and enhance our knowledge of ubiquitination's diverse functions in cell biology.

Weakness:

The involvement of ubiquitin ligases in SARS-CoV-2 replication is not entirely new (see E3 Ubiquitin Ligase RNF5; Yuan et al., 2022; Li et al., 2023). While the data generally support the conclusions, additional work is needed to confirm the role of ITCH in SARS-CoV-2 replication in a biologically relevant context. The vast majority of data is based on transient overexpression experiments of ITCH, which ultimately leads to massive ubiquitination of several viral and host cell factors, including potentially low-affinity substrates not typically recognized under physiological conditions. In addition to that, nearly all experiments were done in cells co-overexpressing ITCH and the viral structural proteins (or cellular proteases) in HEK293T cells. Therefore, a proteomic analysis of protein ubiquitination in a) SARS-CoV-2-infected cells (ideally several cell types) and b) SARS-CoV-2-infected v2T-ITCH-KO cells would verify the ITCH-related ubiquitination of e.g., E and M and would strengthen the whole manuscript. In addition, the few key experiments using SARS-CoV-2 infected cells were performed in VeroE6 cells, which are neither human nor lung-derived. Only in one experiment were lung-derived Calu3 cells included.<br /> Moreover, the manuscript names ITCH as a central regulator of SARS-CoV-2 replication. If ITCH is beneficial for E and M interaction and thereby aids virion assembly, showing its effect on VLP production would be desirable. Clarifications regarding data acquisition and data analysis could strengthen the manuscript and its conclusions.

eLife Assessment

NCX1 is an important cardiac Ca2+/Na+ exchanger whose activity is tightly regulated. This manuscript describes the structural basis of activation by the lipid PIP2 and inhibition by binding of a small molecule to NCX1. These results provide key insights into NCX1 regulation and cellular Ca2+ signaling, but the evidence presented is still incomplete.

Reviewer #1 (Public review):

This study uses structural and functional approaches to investigate the regulation of the Na/Ca exchanger NCX1 by an activator, PIP2, and an inhibitor, SEA0400.

State-of-the-art methods are employed, and the data are of high quality and presented very clearly. The manuscript combines two rather different studies (one on PIP2; and one on SEA0400) neither of which is explored in the depth one might have hoped to form robust conclusions and significantly extend knowledge in the field.

The novel aspect of this work is the study of PIP2. Unfortunately, technical limitations precluded structural data on binding of the native PIP2, so an unnatural short-chained analog, di-C8 PIP2, was used instead. This raises the question of whether these two molecules, which have similar but very distinctly different profiles of activation, actually share the same binding pocket and mode of action. In an effort to address this, the authors mutate key residues predicted to be important in forming the binding site for the phosphorylated head group of PIP2. However, none of these mutations prevent PIP2 activation. The only ones that have a significant effect also influence the Na-dependent inactivation process independently of PIP2, thus casting doubt on their role in PIP2 binding, and thus identification of the PIP2 binding site. A more extensive mutagenic study, based on the di-C8 PIP2 binding site, would have given more depth to this work and might have been more revealing mechanistically.

The SEA0400 aspect of the work does not integrate particularly well with the rest of the manuscript. This study confirms the previously reported structure and binding site for SEA0400 but provides no further information. While interesting speculation is presented regarding the connection between SEA0400 inhibition and Na-dependent inactivation, further experiments to test this idea are not included here.

Reviewer #2 (Public review):

The study by Xue et al. reports the structural basis for the regulation of the human cardiac sodium-calcium exchanger, NCX1, by the endogenous activator PIP2 and the small molecule inhibitor SEA400. This well-written study contextualizes the new data within the existing literature on NCX1 and the broader NCX family. This work builds upon the authors' previous study (Xue et al., 2023), which presented the cryo-EM structures of human cardiac NCX1 in both inactivated and activated states. The 2023 study highlighted key structural differences between the active and inactive states and proposed a mechanism where the activity of NCX1 is regulated by the interactions between the ion-transporting transmembrane domain and the cytosolic regulatory domain. Specifically, in the inward-facing state and at low cytosolic calcium levels, the transmembrane (TM) and cytosolic domains form a stable interaction that results in the inactivation of the exchanger. In contrast, calcium binding to the cytosolic domain at high cytosolic calcium levels disrupts the interaction with the TM domain, leading to active ion exchange.

In the current study, the authors present two mechanisms explaining how both PIP2 stimulates NCX1 activity by destabilizing the protein's inactive state (i.e., by disrupting the interaction between the TM domain and the cytosolic domain) and how SEA400 stabilizes this interaction, thereby acting as a specific inhibitor of the system.

The first part of the results section addresses the effect of PIP2 and PIP2 diC8 on NCX1 activity. This is pertinent as the authors use the diC8 version of this lipid (which has a shorter acyl chain) in their subsequent cryo-EM structure due to the instability of native PIP2. I am not an electrophysiology expert; however, my main comment would be to ask whether there is sufficient data here to characterise fully the differences between PIP2 and PIP2 diC8 on NCX1 function. It appears from the text that this study is the first to report these differences, so perhaps this data needs to be more robust. The spread of the data points in Figure 1B is possibly a little unconvincing given that only six measurements were taken. Why is there one outlier in Figure 1A? Were these results taken using the same batch of oocytes? Are these technical or biological replicates? Is the convention to use statistical significance for these types of experiments?

I am also somewhat skeptical about the modelling of the PIP2 diC8 molecule. The authors state, "The density of the IP3 head group from the bound PIP2 diC8 is well-defined in the EM map. The acyl chains, however, are flexible and could not be resolved in the structure (Fig. S2)."

However, the density appears rather ambiguous to me, and the ligand does not fit well within the density. Specifically, there is a large extension in the volume near the phosphate at the 5' position, with no corresponding volume near the 4' phosphate. Additionally, there is no bifurcation of the volume near the lipid tails. I attempted to model cholesterol hemisuccinate (PDB: Y01) into this density, and it fits reasonably well - at least as well as PIP2 diC8. I am also concerned that if this site is specific for PIP2, then why are there no specific interactions with the lipid phosphates? How can the authors explain the difference between PIP2 and PIP2 diC8 if the acyl chains don't make any direct interactions with the TM domain? In short, the structures do not explain the functional differences presented in Figure 1.

The side chain densities for Arg167 and Arg220 are also quite weak. While there is some density for the side chain of Lys164, it is also very weak. I would expect that if this site were truly specific for PIP2, it should exhibit greater structural rigidity - otherwise, how is this specific?

Given this observation, have the authors considered using other PIP2 variants to determine if the specificity lies with PI4,5P2 as opposed to PI3,5P2 or PI3,4P2? A lack of specificity may explain the observed poor density.

I also noticed many lipid-like densities in the maps for this complex. Is it possible that the authors overlooked something? For instance, there is a cholesterol-like density near Val51, as well as something intriguing near Trp763, where I could model PIP2 diC8 (though this leads to a clash with Trp763). I wonder if the authors are working with mixed populations in their dataset. The accompanying description of the structural changes is well-written (assuming it is accurate).

I would recommend that the authors update the figures associated with this section, as they are currently somewhat difficult to interpret without prior knowledge of NCX architecture. My suggestions include:

- Including the density for the PIP2 diC8 in Figure 2A.

- Adding membrane boundaries (cytosolic vs. extracellular) in Figure 2B.

- Labeling the cytosolic domains in Figure 2B.

- Adding hydrogen bond distances in Figure 2A.

- Detailing the domain movements in Figure 2B (what is the significance of the grey vs. blue structures?).

The section on the mechanism of SEA400-induced inactivation is strong. The maps are of better quality than those for the PIP2 diC8 complex, and the ligand fits well. However, I noticed a density peak below F02 on SEA400 that lies within the hydrogen bonding distance of Asp825. Is this a water molecule? If so, is this significant?

Furthermore, there are many unmodeled regions that are likely cholesterol hemisuccinate or detergent molecules, which may warrant further investigation.

The authors introduce SEA400 as a selective inhibitor of NCX1; however, there is little to no comparison between the binding sites of the different NCX proteins. This section could be expanded. Perhaps Fig. 4C could include sequence conservation data.

Additionally, is the fenestration in the membrane physiological, or is it merely a hole forced open by the binding of SEA400? I was unclear as to whether the authors were suggesting a physiological role for this feature, similar to those observed in sodium channels.

Reviewer #3 (Public review):

NCXs are key Ca2+ transporters located on the plasma membrane, essential for maintaining cellular Ca2+ homeostasis and signaling. The activities of NCX are tightly regulated in response to cellular conditions, ensuring precise control of intracellular Ca2+ levels, with profound physiological implications. Building upon their recent breakthrough in determining the structure of human NCX1, the authors obtained cryo-EM structures of NCX1 in complex with its modulators, including the cellular activator PIP2 and the small molecule inhibitor SEA0400. Structural analyses revealed mechanistically informative conformational changes induced by PIP2 and elucidated the molecular basis of inhibition by SEA0400. These findings underscore the critical role of the interface between the transmembrane and cytosolic domains in NCX regulation and small molecule modulation. Overall, the results provide key insights into NCX regulation, with important implications for cellular Ca2+ homeostasis.

eLife Assessment

This valuable paper reports machine learning-based image analysis pipelines for the automated segmentation of micronuclei and the detection and sorting of micronuclei-containing cells. These are powerful new tools for researchers who study micronuclei and their physiologic consequences. The analysis of the new tools and their benchmarking is rigorous and convincing; applications and remaining limitations are well explained in the paper.

Reviewer #1 (Public review):

DiPeso et al. develop two tools to i) classify micronucleated (MN) cells, which they call VCS MN, and ii) segment micronuclei and nuclei with MNFinder. They then use these tools to identify transcriptional changes in MN cells.

The strengths of this study are:

- Developing highly specialized tools to speed up the analysis of specific cellular phenomena such as MN formation and rupture is likely valuable to the community and neglected by developers of more generalist methods.

- A lot of work and ideas have gone into this manuscript. It is clearly a valuable contribution.

- Combining automated analysis, single-cell labeling, and cell sorting is an exciting approach to enrich for phenotypes of interest, which the authors demonstrate here.

The authors addressed my original concerns related to the first version of this manuscript.

Reviewer #2 (Public review):

Summary:

Micronuclei are aberrant nuclear structures frequently seen following the missegregation of chromosomes. The authors present two image analysis methods, one robust and another rapid, to identify micronuclei (MN) bearing cells. To analyse their software efficacy, the authors study images of cells treated with MPS1 inhibitor to induce chromosome missegregation. Next, the authors use RNA-seq to assess the outcomes of their MN-identifying methods: they do not observe a transcriptomic signature specific to MN but find changes that correlate with aneuploidy status. Overall, this work offers new tools to identify MN-presenting cells, and it sets the stage with clear benchmarks for further software development.

Strengths:

Currently, there are no robust MN classifiers with a clear quantification of their efficiency across cell lines (mIoU score). The software presented here tries to address this gap. GitHub material (images, ground truth labels, tools, protocols, etc.) provided is a great asset to computational biologists. The method has been tested in more than one cell line. This method can help integrate cell biology and 'omics' data, making it suitable for multimodal studies.

Weaknesses:

Although the classifier outperforms available tools for MN segmentation by providing mIoU, it's not yet at a point where it can be reliably applied to functional genomics assays where we expect a range of phenotypic penetrance in most cell lines (e.g., misshapen, multinucleated, and lagging DNA in addition to micronucleated cells). The discussion considers the nature and proportion of MN in RPE1 cells, and how the classifier is well-suited for RPE1 that predominantly display MN structures. Whether the classifier can rigorously assign MN-presenting cells amidst drastic nuclear aberrancies following a spindle checkpoint loss needs to be tested in the future.

Reviewer #3 (Public review):

Summary:

The authors develop automated methods to visually identify micronuclei (MN) and MN-containing cells. The authors then use these methods to isolate MN-containing RPE-1 cells post-photoactivation and analyze transcriptional changes in cells with and without micronuclei. The authors find that RPE-1 cells with MN have similar transcriptomic changes as aneuploid cells and that MN rupture does not lead to vast changes in the transcriptome.

Strengths:

The authors develop a method that allows for automating measurements and analysis of micronuclei. This has been something that the field has been missing for a long time. Using such a method has the potential to greatly enhance the field's ability to analyze micronuclei and understand the downstream consequences. The authors also develop a method to identify cells with micronuclei in real-time, mark them using photoconversion, and then isolate them via cell sorting, which could change the way we isolate and study MN-containing cells, and the scale at which we do it. The authors use this method to look at the transcriptome. This method is very powerful as it can allow for the separation of a heterogenous population and subsequent analysis with a much higher sample number than previously possible.

Weaknesses:

The major weakness of this paper is the transcriptomic analysis of MN. There is in general large variance between replicates in experiments looking at cells with ruptured versus intact micronuclei. This limits our ability to assess if lack of changes are due to truly not having changes between these populations or experimental limitations. More transcriptomic analysis will be necessary to fully understand the downstream consequences of MN rupture.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

DiPeso et al. develop two tools to (i) classify micronucleated (MN) cells, which they call VCS MN, and (ii) segment micronuclei and nuclei with MMFinder. They then use these tools to identify transcriptional changes in MN cells.

The strengths of this study are:

(1) Developing highly specialized tools to speed up the analysis of specific cellular phenomena such as MN formation and rupture is likely valuable to the community and neglected by developers of more generalist methods.

(2) A lot of work and ideas have gone into this manuscript. It is clearly a valuable contribution.

(3) Combining automated analysis, single-cell labeling, and cell sorting is an exciting approach to enrich phenotypes of interest, which the authors demonstrate here.

Weaknesses:

(1) Images and ground truth labels are not shared for others to develop potentially better analysis methods.

We regret this omission and thank the reviewer for pointing it out. Both the images and ground truth labels for VCS MN and MNFinder are now available on the lab’s github page and described in the README.txt files. VCS MN: https://github.com/hatch-lab/fast-mn. MNFinder: https://github.com/hatch-lab/mnfinder.

(2) Evaluations of the methods are often not fully explained in the text.

The text has been extensively updated to include a full description of the methods and choices made to develop the VCS MN and MNFinder image segmentation modules.

(3) To my mind, the various metrics used to evaluate VCS MN reveal it not to be terribly reliable. Recall and PPV hover in the 70-80% range except for the PPV for MN+. It is what it is - but do the authors think one has to spend time manually correcting the output or do they suggest one uses it as is?

VCS MN attempts to balance precision and recall with speed to reduce the fraction of MN changing state from intact to ruptured during a single cell cycle during a live-cell isolation experiment. In addition, we chose to prioritize inclusion of small MN adjacent to the nucleus in our positive calls. This meant that there were more false positives (lower PPV) than obtained by other methods but allowed us to include this highly biologically relevant class of MN in our MN+ population. Thus, for a comprehensive understanding of the consequences of MN formation and rupture, we recommend using the finder as is. However, for other visual cell sorting applications where a small number of highly pure MN positive and negative cells is preferred, such as clonal outgrowth or metastasis assays, we would recommend using the slower, but more precise, MNFinder to get a higher precision at a cost of temporal resolution. In addition, MNFinder, with its higher flexibility and object coverage, is recommended for all fixed cell analyses.

Reviewer #2 (Public review):

Summary:

Micronuclei are aberrant nuclear structures frequently seen following the missegregation of chromosomes. The authors present two image analysis methods, one robust and another rapid, to identify micronuclei (MN) bearing cells. The authors induce chromosome missegregation using an MPS1 inhibitor to check their software outcomes. In missegregation-induced cells, the authors do not distinguish cells that have MN from those that have MN with additional segregation defects. The authors use RNAseq to assess the outcomes of their MN-identifying methods: they do not observe a transcriptomic signature specific to MN but find changes that correlate with aneuploidy status. Overall, this work offers new tools to identify MN-presenting cells, and it sets the stage with clear benchmarks for further software development.

Strengths:

Currently, there are no robust MN classifiers with a clear quantification of their efficiency across cell lines (mIoU score). The software presented here tries to address this gap. GitHub material (tools, protocols, etc) provided is a great asset to naive and experienced computational biologists. The method has been tested in more than one cell line. This method can help integrate cell biology and 'omics' studies.

Weaknesses:

Although the classifier outperforms available tools for MN segmentation by providing mIOU, it's not yet at a point where it can be reliably applied to functional genomics assays where we expect a range of phenotypic penetrance.

We agree that the MNFinder module has limitations with regards to the degree of nuclear atypia and cell density that can be tolerated. Based on the recall and PPV values and their consistency across the majority conditions analyzed, we believe that MNFinder can provide reliable results for MN frequency, integrity, shape, and label characteristics in a functional genomics assay in many commonly used adherent cell lines. We also added a discussion of caveats for these analyses, including the facts that highly lobulated nuclei will have higher false positive rates and that high cell confluency may require additional markers to ensure highly accurate assignment of MN to nuclei.

Spindle checkpoint loss (e.g., MPS1 inhibition) is expected to cause a variety of nuclear atypia: misshapen, multinucleated, and micronucleated cells. It may be difficult to obtain a pure MN population following MPS1 inhibitor treatment, as many cells are likely to present MN among multinucleated or misshapen nuclear compartments. Given this situation, the transcriptomic impact of MN is unlikely to be retrieved using this experimental design, but this does not negate the significance of the work. The discussion will have to consider the nature, origin, and proportion of MN/rupture-only states - for example, lagging chromatids and unaligned chromosomes can result in different states of micronuclei and also distinct cell fates.

We appreciate the reviewer’s comments and now quantify the frequency of other nuclear atypias and MN chromosome content in RPE1 cells after 24 h Mps1 inhibition (Fig. S1). In summary, we find only small increases in nuclear atypia, including multinucleate cells, misshapen nuclei, and chromatin bridges, compared to the large increase in MN formation. This contrasts with what is observed when mitosis is delayed using nocodazole or CENPE inhibitors where nuclear atypia is much more frequent. Importantly, after Mps1 inhibition, RPE1 cells with MN were only slightly more likely to have a misshapen nucleus compared to cells without MN (Fig. S1C).

Interestingly, this analysis showed that the VCS MN pipeline, which uses the Deep Retina segmenter to identify nuclei, has a strong bias against lobulated nuclei and frequently fails to find them (Fig. S2B). Therefore, the cell populations analyzed by RNAseq were largely depleted of highly misshapen nuclei and differences in nuclear atypia frequency between MN+ and MN- cells in the starting population were lost (Fig. S9A, compare to Fig. S1C). This strongly suggests that the transcript changes we observed reflect differences in MN frequency and aneuploidy rather than differences in nuclei morphology.

We agree with the reviewer that MN rupture frequency and formation, and downstream effects on cell proliferation and DNA damage, are sensitive to the source of the missegregated chromatin. In the revised manuscript we make clear that we chose Mps1 inhibition because it is strongly biased towards whole chromosome MN (Fig. S1E), limiting signal from DNA damage products, including chromosome fragments and chromatin bridges. This provides a base line to disambiguate the consequences of micronucleation and DNA damage in more complex chromosome missegregation processes, such as DNA replication disruption and irradiation. 

Reviewer #3 (Public review):

Summary:

The authors develop a method to visually analyze micronuclei using automated methods. The authors then use these methods to isolate MN post-photoactivation and analyze transcriptional changes in cells with and without micronuclei of RPE-1 cells. The authors observe in RPE-1 cells that MN-containing cells show similar transcriptomic changes as aneuploidy, and that MN rupture does not lead to vast changes in the transcriptome.

Strengths:

The authors develop a method that allows for automating measurements and analysis of micronuclei. This has been something that the field has been missing for a long time. Using such a method has the potential to advance micronuclei biology. The authors also develop a method to identify cells with micronuclei in real time and mark them using photoconversion and then isolate them via FACS. The authors use this method to study the transcriptome. This method is very powerful as it allows for the sorting of a heterogenous population and subsequent analysis with a much higher sample number than could be previously done.

Weaknesses:

The major weakness of this paper is that the results from the RNA-seq analysis are difficult to interpret as very few changes are found to begin with between cells with MN and cells without. The authors have to use a 1.5-fold cut-off to detect any changes in general. This is most likely due to the sequencing read depth used by the authors. Moreover, there are large variances between replicates in experiments looking at cells with ruptured versus intact micronuclei. This limits our ability to assess if the lack of changes is due to truly not having changes between these populations or experimental limitations. Moreover, the authors use RPE-1 cells which lack cGAS, which may contribute to the lack of changes observed. Thus, it is possible that these results are not consistent with what would occur in primary tissues or just in general in cells with a proficient cGAS/STING pathway.

We agree with the reviewer’s assessment of the limitations of our RNA-Seq analysis. After additional analysis, we propose an alternative explanation for the lower expression changes we observe in the MN+ and Mps1 inhibitor RNA-Seq experiments. In summary, we find that VCS MN has a strong bias against highly lobulated nuclei that depletes this class of cells from both the bulk analysis and the micronucleated cell populations (Fig. S9A). Based on this result, we propose that our analysis reduces the contribution of nuclear atypia to these transcriptional changes and that nuclear morphology changes are likely a signaling trigger associated with aneuploidy.

We believe that this finding strengthens our overall conclusion that MN formation and rupture do not cause transcriptional changes, as suppressing the signaling associated with nuclei atypia should increase sensitivity to changes from the MN. However, we cannot completely rule out that MN formation or rupture cause a broad low-level change in transcription that is obscured by other signals in the dataset.

As to cGAS signaling, several follow up papers and even the initial studies from the Greenburg lab show that MN rupture does not activate cGAS and does not cause cGAS/STING-dependent signaling in the first cell cycle (see citations and discussion in text). Therefore, we expect the absence of cGAS in RPE1 cells will have no effect in the first cell cycle, but could alter the transcriptional profile after mitosis. Although analysis of RPE1  cGAS+ cells or primary cells in these experiments will be required to definitively address this point, we believe that our interpretation of our RNAseq results is sufficiently backed up by the literature to warrant our conclusion that MN formation and rupture do not induce a transcriptional response in the first cell cycle.

Reviewer #1 (Recommendations for the authors):

I do not recommend additional experimental or computational work. Instead, I just recommend adapting the claims of the manuscript to what has been done. I am just asking for further clarification and minor rewriting.

(1) The manuscript is written like a molecular biology paper with sparse explanations of the authors' reasoning, especially in the development of their algorithms. I was often lost as to why they did things in one way or another.

The revised manuscript has thorough explanations and additional data and graphics defining how and why the VCS MN and MNFinder modules were developed. We hope that this clears up many of the questions the reviewer had and appreciate their guidance on making it more readable for scientists from different backgrounds.

(2) Evaluations of their method are often not fully explained, for example:

"On average, 75% of nuclei per field were correctly segmented and cropped."

"MN segments were then assigned to 'parent' nuclei by proximity, which correctly associated 97% of MN."

Were there ground truth images and labels created? How many? For example, I don't know how the authors could even establish a ground-truth for associating MNs to nuclei if MNs happened to be almost equidistant between two nuclei in their images.

I suggest a separate subsection early in the Results section where the underlying imaging data + labels are presented.

We added new sections to the text and figures at the beginning of the VCS MN and MNFinder subsections (Fig. S2 and Fig. S5) with specific information about how ground truth images and labels were generated for both modules and how these were broken up for training, validation, and testing.

We also added information and images to explain how ground truth MN/nucleus associations were derived. In summary, we took advantage of the fact that 2xDendra-NLS is present at low levels in the cytoplasm to identify cell boundaries. This combined with a subconfluent cell population allowed us to unambiguously group MN and nuclei for 98% of MN, we estimate. These identifications were used to generate ground truth labels and analyze how well proximity defines MN/nuclei groups (Fig.s S1 and S2).

(3) Overall, I find the sections long and more subtitles would help me better navigate the manuscript.

Where possible, we have added subtitles.

(4) Everything following "To train the model, H2B channel images were passed to a Deep Retina neural net ..." is fully automated, it seems to me. Thus, there seems to be no human intervention to correct the output before it is used to train the neural network. Therefore, I do not understand why a neural network was trained at all if the pipeline for creating ground truth labels worked fully automatically. At least, the explanations are insufficient.

We apologize for the initial lack of clarity in the text and included additional details in the revision. We used the Deep Retina segmenter to crop the raw images to areas around individual nuclei to accelerate ground truth labeling of MN. A trained user went through each nucleus crop and manually labeled pixels belonging to MN to generate the ground truth dataset for training, validation, and imaging in VCS MN (Fig. S2A).

(5) To my mind, the various metrics used to evaluate VCS MN reveal it not to be terribly reliable. Recall and PPV hover in the 70-80% range except for the PPV for MN+. It is what it is - but do the authors think one has to spend time manually correcting the output or do they suggest one uses it as is? I understand that for bulk transcriptomics, enrichment may be sufficient but for many other questions, where the wrong cell type could contaminate the population, it is not.

Remarks in the Results section on what the various accuracies mean for different applications would be good (so one does not need to wait for the Discussion section).

One of the strengths of the visual cell sorting system is that any image analysis pipeline can be used with it. We used VCS MN for the transcriptomics experiment, but for other applications a user could run visual cell sorting in conjunction with MNFinder for increased purity while maintaining a reasonable recall or use a pre-existing MN segmentation program that gives 100% purity but captures only a specific subgroup of micronucleated cells (e.g. PIQUE). 

To maintain readability, especially with the expansion of the results sections, we kept the discussion of how we envision using visual cell sorting for other MN-based applications in the discussion section.

(6) I am confused about what "cell" is referring to in much of the manuscript. Is it the nucleus + MNs only? Is it the whole cell, which one would ordinarily think it is? If so, are there additional widefield images, where one can discern cell boundaries? I found the section "MNFinder accurately ..." very hard to read and digest for this reason and other ambiguous wording. I suggest the authors take a fresh look at their manuscript and see whether the text can be improved for clarity. I did not find it an easy read overall, especially the computational part.

After re-examining how “cell” was used, we updated the text to limit its use to the MNFinder arm tasked with identifying MN-nucleus associations where the convex hull defined by these objects is used to determine the “cell” boundary. In all other cases we have replaced cell with “nucleus” because, as the reviewer points out, that is what is being analyzed and converted. We hope this is clearer.

(7) Post-FACS PPVs are not that great (Figure 3c). It depends on the question one wants to answer whether ~70% PPV is good enough. Again, would be good to comment on.

We added discussion of this result to the revision. In summary, a likely reason for the reduced PPV is that, although we maintain the cells in buffer with a Cdk1 inhibitor, we know that some proportion of the cells go through mitosis post-sorting. Since MN are frequently reincorporated into the nucleus after mitosis (Hatch et al, 2013; Zhang et al., 2015), we expect this to reduce the MN+ population. Thus, we expect that the PPV in the RNAseq population is higher than what we can measure by analyzing post-sorted cells that have been plated and analyzed later.

(8) I am thoroughly confused as to why the authors claim that their system works in the "absence of genetic perturbations" and why they emphasize the fact that their cells are non-transformed: They still needed a fluorescent label and they induce MNs with a chemical Mps1 inhibitor. (The latter is not a genetic manipulation, of course, but they still need to enrich MNs somehow. That is, their method has not been tested on a cell population in which MNs occur naturally, presumably at a very low rate, unless I missed something.) A more careful description of the benefits of their method would be good.

We apologize for the confusion on these points and hope this is clarified in the revision. We were comparing our system, which can be made using transient transfection, if desired, to current tools that disambiguate aneuploidy and MN formation by deleting parts of chromosomes or engineering double strand breaks with CRISPR to generate single chromosome-specific missegregation events. Most of these systems require transformed cancer cells to obtain high levels of recombination. In contrast, visual cell sorting can isolate micronucleated cells from any cell line that can exogenously express a protein, including primary cells and non-transformed cells like RPE1s.

Other minor points:

(1) The authors should not refer to "H2B channels" but to "H2B-emiRFP703 channels". It may seem obvious to the authors but for someone reading the manuscript for the very first time, it was not. I was not sure whether there were additional imaging modalities used for H2B/nucleus/chromatin detection before I went back and read that only fluorescence images of H2B-emiRFP703 were used. To put it another way, the authors are detecting fluorescence, not histones -- unless I misunderstood something.

To address this point, we altered the text to read “H2B-emiRFP703” when discussing images of this construct. For MNFinder some images were of cells expressing H2B-GFP, which has also been clarified.

(2) If the level of zoom on my screen is such that I can comfortably read the text, I cannot see much in the figure panels. The features that I should be able to see are the size of a title. The image panels should be magnified.

In the revision, the images are appended to the end at full resolution to overcome this difficulty. Thank you for your forbearance.

Reviewer #2 (Recommendations for the authors):

The methods are adequately explained. The Results text narrating experiments and data analysis is clear. Interpretation of a few results could be clarified and strengthened as explained below.

(1) RNAseq experiments are a good proof of principle. To strengthen their interpretation in Figures 4 and 6, I would recommend the authors cite published work on checkpoint/MPS1 loss-induced chromosome missegregation (PMID: 18545697, PMID: 33837239, PMC9559752) and consider in their discussion the 'origin' and 'proportion' of micronucleated cells and irregularly shaped nuclei expected in RPE1 lines. This will help interpret Figure 6 findings on aneuploidy signature accurately. Not being able to see an MN-specific signature could be due to the way the biological specimen is presented with a mixture of cells with 'MN only' or 'rupture' or 'MN along with misshapen nuclei'. These features may all link to aneuploidy rather than 'MN' specifically.

We appreciate the reviewer’s suggestion and added a new analysis of nuclear atypia after Mps1 inhibition in RPE1 cells to Fig. S1. Overall, we found that Mps1 inhibition significantly, but modestly, increased the proportion of misshapen nuclei and chromatin bridges. Multinucleate cells were so rare that instead of giving them their own category we included them in “misshapen nuclei.” These results are consistent with images of Msp1i treated RPE1 cells from He et al. 2019 and Santaguida et al. 2017 and distinct from the stronger changes in nuclear morphology observed after delaying mitosis by nocodazole or CENPE inhibition.

We also found that the Deep Retina segmenter used to identify nuclei in VCS MN had a significant bias against highly lobulated nuclei (Fig. S2B) that led to misshapen nuclei being largely excluded from the RNAseq analyses. As a result we found no enrichment of misshapen nuclei, chromatin bridges, or dead/mitotic nuclear morphologies in MN+ compared to MN- nuclei in our RNASeq experiments (Fig. S9A).

(2) As the authors clarify in the response letter, one round of ML is unlikely to result in fully robust software; additional rounds of ML with other markers will make the work robust. It will be useful to indicate other ML image analysis tools that have improved through such reiterations. They could use reviews on challenges and opportunities using ML approaches to support their statement. Also in the introduction, I would recommend labelling as 'rapid' instead of 'rapid and precise' method.

We updated the text to reference review articles that discuss the benefit of additional training for increasing ML accuracy and changed the text to “rapid.”

(3) The lack of live-cell studies does not allow the authors to distinguish the origin of MN (lagging chromatids or unaligned chromosomes). As explained in 1, considering these aspects in discussion would strengthen their interpretation. Live-cell studies can help reduce the dependencies on proximity maps (Figure S2).

The revised text includes new references and data (Fig. S1E) demonstrating that Mps1 inhibition strongly biases towards whole chromosome missegregation and that MN are most likely to contain a single centromere positive chromosome rather than chromatin fragments or multiple chromosomes.

(4) Mean Intersection over Union (mIOU) is a good measure to compare outcomes against ground truth. However, the mIOU is relatively low (Figure 2D) for HeLa-based functional genomics applications. It will help to discuss mIOU for other classifiers (non-MN classifiers) so that they can be used as a benchmark (this is important since the authors state in their response that they are the first to benchmark an MN classifier). There are publications for mitochondria, cell cortex, spindle, nuclei, etc. where IOU has been discussed.

We added references to classifiers for other small cellular structures. We also evaluated major sources of error in MNFinder found that false negatives are enriched in very small MN (3 to 9 pixels, or about 0.4 µm² – 3 µm², Fig. S6B). A similar result was obtained for VCS MN (Fig. S3B). Because small changes in the number of pixels identified in small objects can have outsized effects on mIoU scores, we suspect that this is exerting downward pressure on the mIoU value. Based on the PPV and recall values we identified, we believe that MNFinder is robust enough to use for functional genomics and screening applications with reasonable sample sizes.

(5) Figure 5 figure legend title is an overinterpretation. MN and rupture-initiated transcriptional changes could not be isolated with this technique where several other missegregation phenotypes are buried (see point 1 above).

We decided to keep the figure title legend based on our analysis of known missegregation phenotypes in Fig. S1 and S9 showing that there is no difference in major classes of nuclear atypia between MN+ and MN- populations in this analysis. Although we cannot rule out that other correlated changes exist, we believe that the title represents the most parsimonious interpretation.

Minor comments

(1) The sentence in the introduction needs clarification and reference. "However, these interventions cause diverse "off-target" nuclear and cellular changes, including chromatin bridges, aneuploidy, and DNA damage." Off-target may not be the correct description since inhibiting MPS1 is expected to cause a variety of problems based on its role as a master kinase in multiple steps of the chromosome segregation process. Consider one of the references in point 1 for a detailed live-cell view of MPS1 inhibitor outcomes.

We have changed “off-target” to “additional” for clarity.

(2) In Figure 3 or S3, did the authors notice any association between the cell cycle phase and MN or rupture presence? Is this possible to consider based on FACS outcomes or nuclear shapes?

Previous work by our lab and others have shown that MN rupture frequency increases during the cell cycle (Hatch et al., 2013; Joo et al., 2023). Whether this is stochastic or regulated by the cell cycle may depend on what chromosome is in the MN (Mammel et al., 2021) and likely the cell line. Unfortunately, the H2B-emiRFP703 fluorescence in our population is too variable to identify cell cycle stage from FACS or nuclear fluorescence analysis.

(3) Figure 5 - Please explain "MA plot".

An MA plot, or log fold-change (M) versus average (A) gene expression, is a way to visualize differently expressed genes between two conditions in an RNASeq experiment and is used as an alternative to volcano plots. We chose them for our paper because most of the expression changes we observed were small and of similar significance and the MA plot spreads out the data compared to a volcano plot and allowed a better visualization of trends across the population.

(4) Page 7: "our results strongly suggest that protein expression changes in MN+ and rupture+ cells are driven mainly by increased aneuploidy rather than cellular sensing of MN formation and rupture.". This is an overstatement considering the mIOU limits of the software tool and the non-exclusive nature of MN in their samples.

We agree that we cannot rule out that an unknown masking effect is inhibiting our ability to observe small broad changes in transcription after MN formation or rupture. However, we believe we have minimized the most likely sources of masking effects, including nuclear atypia and large scale aneuploidy differences, and thus our interpretation is the most likely one.

Reviewer #3 (Recommendations for the authors):

Overall, the authors need to explain their methods better, define some technical terms used, and more thoroughly explain the parameters and rationale used when implementing these two protocols for identifying micronuclei; primarily as this is geared toward a more general audience that does not necessarily work with machine learning algorithms.

(1) A clearer description in the methods as to how accuracy was calculated. Were micronuclei counted by hand or another method to assess accuracy?

We significantly expanded the section on how the machine learning models were trained and tested, including how sensitivity and specificity metrics were calculated, in both the results and the methods sections. The code used to compare ground truth labels to computed masks is also now included in the MNFinder module available on the lab github page. 

(2) Define positive predictive value.

The text now says “the positive predictive value (PPV, the proportion of true positives, i.e. specificity) and recall (the proportion of MN found by the classifier, i.e. sensitivity)…”.

(3) Why is it a problem to use the VCS MN at higher magnifications where undersegmentation occurs? What do the authors mean by diminished performance (what metrics are they using for this?).

We have included a representative image and calculated mIoU and recall for 40x magnification images analyzed by MNFinder after rescaling in Fig. 2A. In summary, VCS MN only correctly labeled a few pixels in the MN, which was sufficient to call the adjacent nucleus “MN+” but not sufficient for other applications, such as quantifying MN area. In addition, VCS MN did much worse at identifying all the MN in 40x images with a recall, or sensitivity, metric of 0.36. We are not sure why. Developing MNFinder provided a module that was well suited to quantify MN characteristics in fixed cell images, an important use case in MN biology.

(4) The authors should compare MN that are analyzed and not analyzed using these methods and define parameters. Is there a size limitation? Closeness to the main nucleus?

We added two new figures defining what contributes to module error for both VCS MN (Fig. S3) and MNFinder (Fig. S6). For VCS MN, false negatives are enriched in very large or very small MN and tend to be dimmer and farther from the nucleus than true positives. False positives are largely misclassification of small dim objects in the image as MN. For MNFinder, the most missed class of MN are very small ones (3-9 px in area) and the majority of false positives are misclassifications of elongated nuclear blebs as MN.

(5) Are there parameters in how confluent an image must be to correctly define that the micronucleus belongs to the correct cell? The authors discussed that this was calculated based on predicted distance. However, many factors might affect proper calling on MN. And the authors should test this by staining for a cytosolic marker and calculating accuracy.

We updated the text with more information about how the cytoplasm was defined using leaky 2x-Dendra2-NLS signal to analyze the accuracy of MN/nucleus associations (Fig. S2G-H). In addition, we quantified cell confluency and distance to the first and second nearest neighbor for each MN in our training and testing image datasets. We found that, as anticipated, cells were imaged at subconfluent concentrations with most fields having a confluency around 30% cell coverage (Fig. S2E) and that the average difference in distance between the closest nucleus to an MN and the next closest nucleus was 3.3 fold (Fig. S2F). We edited the discussion section to state that the ability of MN/nuclear proximity to predict associations at high cell confluencies would have to be experimentally validated.

(6) The authors measure the ratio of Dendra2(Red) v. Dendra2 (Green) in Figure 3B to demonstrate that photoconversion is stable. This measurement, to me, is confusing, as in the end, the authors need to show that they have a robust conversion signal and are able to isolate these data. The authors should directly demonstrate that the Red signal remains by analyzing the percent of the Red signal compared to time point 0 for individual cells.

We found a bulk analysis to be more powerful than trying to reidentify individual cells due to how much RPE1 cells move during the 4 and 8 hours between image acquisitions. In addition, we sort on the ratio between red and green fluorescence per cell, rather than the absolute fluorescence, to compensate for variation in 2xDendra-NLS protein expression between cells. Therefore, demonstrating that distinct ratios remained present throughout the time course is the most relevant to the downstream analysis.

To address the reviewer’s concern, we replotted the data in Fig. 3B to highlight changes over time in the raw levels of red and green Dendra fluorescence (Fig. S7D). As expected, we see an overall decrease in red fluorescence intensity, and complementary increase in green fluorescence intensity, over 8 hours, likely due to protein turnover. We also observe an increase in the number of nuclei lacking red fluorescence. This is expected since the well was only partially converted and we expect significant numbers of unconverted cells to move into the field between the first image and the 8 hour image.

(7) The authors isolate and subsequently use RNA-sequencing to identify changes between Mps1i and DMSO-treated cells. One concern is that even with the less stringent cut-off of 1.5 fold there is a very small change between DMSO and MPS1i treated cells, with only 63 genes changing, none of which were affected above a 2-fold change. The authors should carefully address this, including why their dataset sees changes in many more pathways than in the He et al. and Santaguida et al. studies. Is this due to just having a decreased cut-off?

The reviewer correctly points out that we observed an overall reduction in the strength of gene expression changes between our dataset of DMSO versus Mps1i treated RPE1 cells compared to similar studies. We suggest a couple reasons for this. One is that the log₂ fold changes observed in the other studies are not huge and vary between 2.5 and -3.8 for He et al., 3.3 and -2.3 for Santaguida et al., and -0.8 and 1.6 for our study. This variability is within a reasonable range for different experimental conditions and library prep protocols. A second is that our protocol minimizes a potential source of transcriptional change – nuclear lobulation – that is present in the other datasets.

For the pathway analysis we did not use a fold-change cut-off for any data set, instead opting to include all the genes found to be significantly different between control and Mps1i treated cells for all three studies. Our read-depth was higher than that of the two published experiments, which could contribute to an increased DEG number. However, we hypothesize that our identification of a broader number of altered pathways most likely arises from increased sensitivity due to the loss of covering signal from transcriptional changes associated with increased nuclear atypia. Additional visual cell sorting experiments sorting on misshapen nuclei instead of MN would allow us to determine the accuracy of this hypothesis.

(8) Moreover, clustering (in Figure 5E) of the replicates is a bit worrisome as the variances are large and therefore it is unclear if, with such large variance and low screening depth, one can really make such a strong conclusion that there are no changes. The authors should prove that their conclusion that rupture does not lead to large transcriptional changes, is not due to the limitations of their experimental design.

We agree with the reviewers that additional rounds of RNAseq would improve the accuracy of our transcriptomic analysis and could uncover additional DEGs. However, we believe the overall conclusion to be correct based on the results of our attempt to validate changes in gene expression by immunofluorescence. We analyzed two of the most highly upregulated genes in the ruptured MN dataset, ATF3 and EGR1. Although we saw a statistically significant increase in ATF3 intensity between cells without MN and those with ruptured MN, the fold change was so small compared to our positive control (100x less) that we believe it is it is more consistent with a small increase in the probability of aneuploidy rather than a specific signature of MN rupture.

(9) The authors also need to address the fact that they are using RPE-1 cells more clearly and that the lack of effect in transcriptional changes may be simply due to the loss of cGAS-STING pathway (Mackenzie et al., 2017; Harding et al., 2017; etc.).

As we discuss above in the public comments section, the literature is clear that MN do not activate cGAS in the first cell cycle after their formation, even upon rupture. Therefore, we do not expect any changes in our results when applied to cGAS-competent cells. However, this expectation needs to be experimentally validated, which we plan to address in upcoming work.

eLife Assessment

This valuable study introduces a new method for detecting RNA modification. Since it does not rely on chemical modification of RNA, which often results in RNA degradation and therefore loss of RNA molecules, it complements other approaches for detecting RNA modification, and it might be of particular interest for sites where modifications are found in only a minority of interrogated molecules. The information provided is incomplete, however, to allow for comparison with other methods, since there is uncertainty regarding false positive and false negative rates.

Reviewer #2 (Public review):

The fledgling field of epitranscriptomics has encountered various technical roadblocks with implications as to the validity of early epitranscriptomics mapping data. As a prime example, the low specificity of (supposedly) modification-specific antibodies for the enrichment of modified RNAs, has been ignored for quite some time and is only now recognized for its dismal reproducibility (between different labs), which necessitates the development of alternative methods for modification detection. Furthermore, early attempts to map individual epitranscriptomes using sequencing-based techniques are largely characterized by the deliberate avoidance of orthogonal approaches aimed at confirming the existence of RNA modifications that have been originally identified.

Improved methodology, the inclusion of various controls, and better mapping algorithms as well as the application of robust statistics for the identification of false-positive RNA modification calls have allowed revisiting original (seminal) publications whose early mapping data allowed making hyperbolic claims about the number, localization and importance of RNA modifications, especially in mRNA. Besides the existence of m6A in mRNA, the detectable incidence of RNA modifications in mRNAs has drastically dropped.

As for m5C, the subject of the manuscript submitted by Zhou et al., its identification in mRNA goes back to Squires et al., 2012 reporting on >10.000 sites in mRNA of a human cancer cell line, followed by intermittent findings reporting on pretty much every number between 0 to > 100.000 m5C sites in different human cell-derived mRNA transcriptomes. The reason for such discrepancy is most likely of a technical nature. Importantly, all studies reporting on actual transcript numbers that were m5C-modified relied on RNA bisulfite sequencing, an NGS-based method, that can discriminate between methylated and non-methylated Cs after chemical deamination of C but not m5C. RNA bisulfite sequencing has a notoriously high background due to deamination artifacts, which occur largely due to incomplete denaturation of double-stranded regions (denaturing-resistant) of RNA molecules. Furthermore, m5C sites in mRNAs have now been mapped to regions that have not only sequence identity but also structural features of tRNAs. Various studies revealed that the highly conserved m5C RNA methyltransferases NSUN2 and NSUN6 do not only accept tRNAs but also other RNAs (including mRNAs) as methylation substrates, which in combination account for most of the RNA bisulfite-mapped m5C sites in human mRNA transcriptomes. Is m5C in mRNA only a result of the Star activity of tRNA or rRNA modification enzymes, or is their low stoichiometry biologically relevant?

In light of the short-comings of existing tools to robustly determine m5C in transcriptomes, other methods, like DRAM-seq, aiming to map m5C independently of ex situ RNA treatment with chemicals, are needed to arrive at a more solid "ground state", from which it will be possible to state and test various hypotheses as to the biological function of m5C, especially in lowly abundant RNAs such as mRNA.

Importantly, the identification of >10.000 sites containing m5C increases through DRAM-Seq, increases the number of potential m5C marks in human cancer cells from a couple of 100 (after rigorous post-hoc analysis of RNA bisulfite sequencing data) by orders of magnitude. This begs the question, whether or not the application of these editing tools results in editing artefacts overstating the number of actual m5C sites in the human cancer transcriptome.

Author response:

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

Reviewer #2:

(1) The use of two m⁵C reader proteins is likely a reason for the high number of edits introduced by the DRAM-Seq method. Both ALYREF and YBX1 are ubiquitous proteins with multiple roles in RNA metabolism including splicing and mRNA export. It is reasonable to assume that both ALYREF and YBX1 bind to many mRNAs that do not contain m⁵C.

To substantiate the author's claim that ALYREF or YBX1 binds m⁵C-modified RNAs to an extent that would allow distinguishing its binding to non-modified RNAs from binding to m⁵C-modified RNAs, it would be recommended to provide data on the affinity of these, supposedly proven, m⁵C readers to non-modified versus m⁵C-modified RNAs. To do so, this reviewer suggests performing experiments as described in Slama et al., 2020 (doi: 10.1016/j.ymeth.2018.10.020). However, using dot blots like in so many published studies to show modification of a specific antibody or protein binding, is insufficient as an argument because no antibody, nor protein, encounters nanograms to micrograms of a specific RNA identity in a cell. This issue remains a major caveat in all studies using so-called RNA modification reader proteins as bait for detecting RNA modifications in epitranscriptomics research. It becomes a pertinent problem if used as a platform for base editing similar to the work presented in this manuscript.

The authors have tried to address the point made by this reviewer. However, rather than performing an experiment with recombinant ALYREF-fusions and m⁵C-modified to unmodified RNA oligos for testing the enrichment factor of ALYREF in vitro, the authors resorted to citing two manuscripts. One manuscript is cited by everybody when it comes to ALYREF as m⁵C reader, however none of the experiments have been repeated by another laboratory. The other manuscript is reporting on YBX1 binding to m⁵C-containing RNA and mentions PAR-CLiP experiments with ALYREF, the details of which are nowhere to be found in doi: 10.1038/s41556-019-0361-y.<br /> Furthermore, the authors have added RNA pull-down assays that should substitute for the requested experiments. Interestingly, Figure S1E shows that ALYREF binds equally well to unmodified and m⁵C-modified RNA oligos, which contradicts doi:10.1038/cr.2017.55, and supports the conclusion that wild-type ALYREF is not specific m⁵C binder. The necessity of including always an overexpression of ALYREF-mut in parallel DRAM experiments, makes the developed method better controlled but not easy to handle (expression differences of the plasmid-driven proteins etc.)

Thank you for pointing this out. First, we would like to correct our previous response: the binding ability of ALYREF to m⁵C-modified RNA was initially reported in doi: 10.1038/cr.2017.55, (and not in doi: 10.1038/s41556-019-0361-y), where it was observed through PAR-CLIP analysis that the K171 mutation weakens its binding affinity to m⁵C -modified RNA.

Our previous experimental approach was not optimal: the protein concentration in the INPUT group was too high, leading to overexposure in the experimental group. Additionally, we did not conduct a quantitative analysis of the results at that time. In response to your suggestion, we performed RNA pull-down experiments with YBX1 and ALYREF, rather than with the pan-DRAM protein, to better validate and reproduce the previously reported findings. Our quantitative analysis revealed that both ALYREF and YBX1 exhibit a stronger affinity for m⁵C -modified RNAs. Furthermore, mutating the key amino acids involved in m⁵C recognition significantly reduced the binding affinity of both readers. These results align with previous studies (doi: 10.1038/cr.2017.55 and doi: 10.1038/s41556-019-0361-y), confirming that ALYREF and YBX1 are specific readers of m⁵C -modified RNAs. However, our detection system has certain limitations. Despite mutating the critical amino acids, both readers retained a weak binding affinity for m⁵C, suggesting that while the mutation helps reduce false positives, it is still challenging to precisely map the distribution of m⁵C modifications. To address this, we plan to further investigate the protein structure and function to obtain a more accurate m⁵C sequencing of the transcriptome in future studies. Accordingly, we have updated our results and conclusions in lines 294-299 and discuss these limitations in lines 109-114.

In addition, while the m⁵C assay can be performed using only the DRAM system alone, comparing it with the DRAM^mutC control enhances the accuracy of m⁵C region detection. To minimize the variations in transfection efficiency across experimental groups, it is recommended to use the same batch of transfections. This approach not only ensures more consistent results but also improve the standardization of the DRAM assay, as discussed in the section added on line 308-312.

(2) Using sodium arsenite treatment of cells as a means to change the m⁵C status of transcripts through the downregulation of the two major m⁵C writer proteins NSUN2 and NSUN6 is problematic and the conclusions from these experiments are not warranted. Sodium arsenite is a chemical that poisons every protein containing thiol groups. Not only do NSUN proteins contain cysteines but also the base editor fusion proteins. Arsenite will inactivate these proteins, hence the editing frequency will drop, as observed in the experiments shown in Figure 5, which the authors explain with fewer m⁵C sites to be detected by the fusion proteins.

The authors have not addressed the point made by this reviewer. Instead the authors state that they have not addressed that possibility. They claim that they have revised the results section, but this reviewer can only see the point raised in the conclusions. An experiment would have been to purify base editors via the HA tag and then perform some kind of binding/editing assay in vitro before and after arsenite treatment of cells.

We appreciate the reviewer’s insightful comment. We fully agree with the concern raised. In the original manuscript, our intention was to use sodium arsenite treatment to downregulate NSUN mediated m⁵C levels and subsequently decrease DRAM editing efficiency, with the aim of monitoring m⁵C dynamics through the DRAM system. However, as the reviewer pointed out, sodium arsenite may inactivate both NSUN proteins and the base editor fusion proteins, and any such inactivation would likely result in a reduced DRAM editing. This confounds the interpretation of our experimental data.

As demonstrated in Appendix A, western blot analysis confirmed that sodium arsenite indeed decreased the expression of fusion proteins. In addition, we attempted in vitro fusion protein purification using multiple fusion tags (HIS, GST, HA, MBP) for DRAM fusion protein expression, but unfortunately, we were unable to obtain purified proteins. However, using the Promega TNT T7 Rapid Coupled In Vitro Transcription/Translation Kit, we successfully purified the DRAM protein (Appendix B). Despite this success, subsequent in vitro deamination experiments did not yield the expected mutation results (Appendix C), indicating that further optimization is required. This issue is further discussed in line 314-315.

Taken together, the above evidence supports that the experiment of sodium arsenite treatment was confusing and we determined to remove the corresponding results from the main text of the revised manuscript.

Author response image 1.

(3) The authors should move high-confidence editing site data contained in Supplementary Tables 2 and 3 into one of the main Figures to substantiate what is discussed in Figure 4A. However, the data needs to be visualized in another way then excel format. Furthermore, Supplementary Table 2 does not contain a description of the columns, while Supplementary Table 3 contains a single row with letters and numbers.

The authors have not addressed the point made by this reviewer. Figure 3F shows the screening process for DRAM-seq assays and principles for screening high-confidence genes rather than the data contained in Supplementary Tables 2 and 3 of the former version of this manuscript.

Thank you for your valuable suggestion. We have visualized the data from Supplementary Tables 2 and 3 in Figure 4A as a circlize diagram (described in lines 213-216), illustrating the distribution of mutation sites detected by the DRAM system across each chromosome. Additionally, to improve the presentation and clarity of the data, we have revised Supplementary Tables 2 and 3 by adding column descriptions, merging the DRAM-ABE and DRAM-CBE sites, and including overlapping m⁵C genes from previous datasets.

eLife Assessment

This important study shows how genetic variation is associated with fecundity following a period of reproductive diapause in female Drosophila. The work identifies the olfactory system as central to successful diapause with associated changes in longevity and fecundity. While the methods used are convincing, a limitation of the study, as of any other laboratory-based investigation is the challenge of demonstrating how well measures for fitness related to diapause and its recovery correlates with realities encountered during development in the wild.

Reviewer #1 (Public review):

Summary:

The paper begins with phenotyping the DGRP for post-diapause fecundity, which is used to map genes and variants associated with fecundity. There are overlaps with genes mapped in other studies and also functional enrichment of pathways including most surprisingly neuronal pathways. This somewhat explains the strong overlap with traits such as olfactory behaviors and circadian rhythm. The authors then go on to test genes by knocking them down effectively at 10 degrees. Two genes, Dip-gamma and sbb are identified as significantly associated with post-diapause fecundity, which they also find the effects to be specific to neurons. They further show that the neurons in the antenna but not arista are required for the effects of Dip-gamma and sbb. They show that removing antenna has a diapause specific lifespan extending effect, which is quite interesting. Finally, ionotropic receptor neurons are shown to be required for the diapause associated effects.

Strengths:

Overall I find the experiments rigorously done and interpretations sound. I have no further suggestions except an ANOVA to estimate heritability of the post-diapause fecundity trait, which is routinely done in the DGRP and offers a global parameter regarding how reliable phenotyping is.

Weaknesses:

A minor point is I cannot find how many DGRP lines are used.

Author response:

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

Reviewer #1 (Public Review): 

Summary: 

The paper begins with phenotyping the DGRP for post-diapause fecundity, which is used to map genes and variants associated with fecundity. There are overlaps with genes mapped in other studies and also functional enrichment of pathways including most surprisingly neuronal pathways. This somewhat explains the strong overlap with traits such as olfactory behaviors and circadian rhythm. The authors then go on to test genes by knocking them down effectively at 10 degrees. Two genes, Dip-gamma and sbb, are identified as significantly associated with post-diapause fecundity, and they also find the effects to be specific to neurons. They further show that the neurons in the antenna but not the arista are required for the effects of Dip-gamma and sbb. They show that removing the antenna has a diapause-specific lifespan-extending effect, which is quite interesting. Finally, ionotropic receptor neurons are shown to be required for the diapause-associated effects. 

Strengths and Weaknesses: 

Overall I find the experiments rigorously done and interpretations sound. I have no further suggestions except an ANOVA to estimate the heritability of the post-diapause fecundity trait, which is routinely done in the DGRP and offers a global parameter regarding how reliable phenotyping is. 

We added to the Methods: “We performed a one-way ANOVA to get the mean squares for between-group and withingroup variances and calculated broad-sense heritability using the formula: H² = MS_G - MS_E / MS_G + (k-1) MS_E where MS_G - Mean square between groups and MS_G - Mean square within groups and k - Number of individuals per group. Using this formula, the broad-sense heritability for normalized post-diapause fecundity was found to be 0.51.” 

We added to the Results: “The broad-sense heritability for normalized post-diapause fecundity was found to be 0.51 (see Methods).”

A minor point is I cannot find how many DGRP lines are used. 

Response: We screened 193 lines and have added that to the Results. 

Reviewer #2 (Public Review):

Summary

In this study, Easwaran and Montell investigated the molecular, cellular, and genetic basis of adult reproductive diapause in Drosophila using the Drosophila Genetic Reference Panel (DGRP). Their GWAS revealed genes associated with variation in post-diapause fecundity across the DGRP and performed RNAi screens on these candidate genes. They also analyzed the functional implications of these genes, highlighting the role of genes involved in neural and germline development. In addition, in conjunction with other GWAS results, they noted the importance of the olfactory system within the nervous system, which was supported by genetic experiments. Overall, their solid research uncovered new aspects of adult diapause regulation and provided a useful reference for future studies in this field.

Strengths:

The authors used whole-genome sequenced DGRP to identify genes and regulatory mechanisms involved in adult diapause. The first Drosophila GWAS of diapause successfully uncovered many QTL underlying post-diapause fecundity variations across DGRP lines. Gene network analysis and comparative GWAS led them to reveal a key role for the olfactory system in diapause lifespan extension and post-diapause fecundity.

Comments on revised version:

While the authors have addressed many of the minor concerns raised by the reviewers, they have not fully resolved some of the key criticisms. Notably, two reviewers highlighted significant concerns regarding the phenotype and assay of post-diapause fecundity, which are critical to the study. The authors acknowledged that this assay could be confounded by the 'cold temperature endurance phenotype,' potentially altering the interpretation of their results.

However, they responded by stating that it is not obvious how to separate these effects experimentally. This leaves the analysis in this research ambiguous, as also noted by Reviewer #3.

We should have clarified earlier that we actually chose to measure post-diapause fecundity in order to minimize any impact of ‘cold temperature endurance.” In fact, we chose post-diapause fecundity as the appropriate measure of successful diapause for both technical and conceptual reasons. Conceptually, the benefit of diapause is to perpetuate the species. It seems obvious to us that post-diapause fecundity is more relevant to species propagation than other measures of diapause such as how many egg chambers contain yolk or how many eggs are laid. Technically, we chose 5-week diapause and recovery based on pilot studies that showed that nearly all DGRP lines showed excellent survival at 5 weeks in diapause conditions. Therefore, our experimental design minimized as much as possible any effect of cold temperature endurance - in the sense of the ability to survive at 10°C - on our phenotype. 

We apologize for not clarifying that point earlier and have added this text to the Results: “We chose 5 weeks based on pilot studies that showed that nearly all DGRP lines showed excellent survival at 5 weeks in diapause conditions while exhibiting sufficient variation in post-diapause fecundity to carry out GWAS. Beyond 5 weeks, fecundity was low and there was insufficient variation to conduct a GWAS.”

Additionally, I raised concerns about the validity of prioritizing genes with multiple associated variants. Although the authors agreed with this point, they did not revise the manuscript accordingly. The statement that 'Genes with multiple SNPs are good candidates for influencing diapause traits' is not a valid argument within the context of population and quantitative genetics.

We apologize for neglecting to revise the manuscript accordingly. We have revised Supplemental Table: S4 and ranked the genes by p-value.

Author Response:

Reviewer #1 (Public Review):

[...] Strengths: This study utilized multiple in vitro approaches, such as proteomics, siRNA, and overexpression, to demonstrate that PCBP2 is an intrinsic factor of BMSC aging.

Weaknesses:

This study did not perform in vivo experiments.

Response: We will continue to conduct animal experiments in subsequent studies.

Reviewer #2 (Public Review):

[...] Weaknesses: It is unclear if PCBP2 can also function as an intrinsic factor for BMSC cells in female individuals. More work may be needed to further dissect the mechanism of how PCBP2 impacts FGF2 expression. Could PCBP2 impact the FGF2 expression independent of ROS?

Response: Thank you very much for your valuable comments, which is also the focus of our follow-up work. We will sort out the data and publish the relevant research results as soon as possible.

Additional context that would help readers interpret or understand the significance of the work: In the current work, the authors studied the aging process of BMSC cells, which are related to osteoporosis. Aging processes also impact many other cell types and their function, such as in muscle, skin, and the brain.

Response: Thank you very much for your valuable comments, we will continue to improve the writing logic of the article to make the article more understandable.

eLife Assessment

This useful manuscript reports mechanisms behind the increase in fecundity in response to sub-lethal doses of pesticides in the crop pest, the brown plant hopper. The authors hypothesize that the pesticide works by inducing the JH titer, which through the JH signaling pathway induces egg development. Evidence for this is, however, incomplete.

Reviewer #1 (Public review):

Summary:

Gao et al. has demonstrated that the the pesticide emamectin benzoate (EB) treatment of brown plathopper (BPH) leads to increased egg laying in the insect, which is a common agricultural pest. The authors hypothesize that EB upregulates JH titer resulting in increased fecundity.

Strengths:

The finding that a class of pesticide increases fecundity of brown planthopper is interesting.

Weaknesses:

(1) EB is an allosteric modulator of GluCl. That means it EB physically interacts with GluCl initiating a structural change in the cannel protein. Yet the authors here central hypothesis is about how EB can upregulate the mRNA of GluCl. I do not know whether there is any evidence that an allosteric modulator can function as a transcriptional activator for the same receptor protein. The basic premise of the paper sounds counterintuitive. This is a structural problem and should be addressed by the authors by giving sufficient evidence about such demonstrated mechanisms before.<br /> (2) I am surprised to see a 4th instar larval application or treatment with EB results in upregulation of JH in the adult stages. Complicating the results further is the observation that a 4th instar EB application results in an immediate decrease in JH titer. There is a high possibility that this late JH titer increase is an indirect effect.<br /> (3) The writing quality of the paper needs improvement. Particularly with respect to describing processes, and abbreviations. In several instances authors have not adequately described the processes they have introduced, thus confusing the readers.<br /> (4) In the section 'EB promotes ovarian development' the authors have shown that EB treatment results in increased detention of eggs which contradicts their own results which show that EB promotes egg laying. Again, this is a serious contradiction that nullifies their hypothesis.<br /> (5) Furthermore, the results suggest that oogenesis is not affected by EB application. The authors should devote a section to discussing how they are observing increased egg numbers in EB-treated insects while not impacting Oogenesis.<br /> (6) Met is the receptor of JH and to my understanding, remains mostly constant in terms of its mRNA or protein levels throughout various developmental periods in many different insects. Therefore, the presence of JH becomes the major driving factor for physiological events and not the presence of the receptor Met. Here the authors have demonstrated an increase in Met mRNA as a result of EB treatment. Their central hypothesis is that EB increases JH titer to result in enhanced fecundity. JH action will not result in the activation of Met. Although not contradictory to the hypothesis, the increase in mRNA content of Met is contrary to the findings of the JH field thus far.<br /> (7) As pointed out before, it is hard to rationalize how a 4th instar exposure to EB can result in upregulation of key genes involved in JH synthesis at the adult stage. The authors must consider providing a plausible explanation and discussion in this regard.<br /> (8) I have strong reservations against such an irrational hypothesis that Met (the receptor for JH) and JH-Met target gene Kr-h1 regulates JH titer (Line 311, Fig 3 supplemental 2D). This would be the first report of such an event on the JH field and therefore must be analysed to depth before it may go to publication. I strongly suggest the authors remove such claims from the manuscript without substantiating it.<br /> (9) Kr-h1 is JH/Met target gene. The authors demonstrate that silencing of Kr-h1 results in inhibition of FAMeT, which is a gene involved in JH synthesis. The feedback loop in JH synthesis is unreported. Authors must go ahead with a mechanistic detail of Kr-h1 mediated JH upregulation before this can be concluded. Mere qPCR experiments are not sufficient to substantiate a claim that is completely contrary to the current understanding of JH signalling pathway.<br /> (10) Authors have performed knockdowns of JHAMT, Met and Kr-h1 to demonstrate the effect of these factors on fecundity n BPH. Additionally, they have performed rescue experiments with EB application on these knockdown insects (Figure 3K-M). This I believe is a very flawed experiment. The authors demonstrate EB works through JHAMT in upregulating JH titer. In the absence of JHAMT, EB application is not expected to rescue the phenotype. But authors have reported a complete rescue here. In the absence of Met, the receptor of JH, either EB or JH is not expected to rescue the phenotype. But a complete rescue has been reported. These two experimental results contradict their own hypothesis.<br /> (11) A significant section of the paper deals with how EB upregulates JH titer. JH is a hormone synthesized in the Corpora Allata. Yet the authors have chosen to use the whole body for all of their experiments. Changes in the whole body for mRNA of those enzymes involved in JH synthesis does may not reflect on the situation in Corpora Allata. Although working with corpora Allata is challenging, discarding the abdomen and thorax region and working with the head and neck region of the insect is easily doable. Results from such sampling is always more convincing when it comes to JH synthesis studies.<br /> (12) The phenomenon reported was specific for BPH and not found in other insects. This limits the implications of the study.<br /> (13) Overall, the molecular experiments are very poorly designed and can at best be termed superficial. There are several contradictions within the paper and no discussion or explanation has been provided for that.

Comments on revisions:

(1) The onus of making the revisions understandable to the reviewers lies with the authors. In its current form, how the authors have approached the review is hard to follow, in my opinion. Although the authors have taken a lot of effort in answering the questions posed by reviewers, parallel changes in the manuscript are not clearly mentioned. In many cases, the authors have acknowledged the criticism in response to the reviewer, but have not changed their narrative, particularly in the results section.<br /> (2) In the response to reviewers, the authors have mentioned line numbers in the main text where changes were made. But very frequently, those lines do not refer to the changes or mention just a subsection of changes done. The problem is throughout the document making it very difficult to follow the revision and contributing to the point mentioned above.<br /> (3) The authors need to infer the performed experiments rationally without over interpretation. Currently, many of the claims that the authors are making are unsubstantiated. As a result of the first review process, the authors have acknowledged the discrepancies, but they have failed to alter their interpretations accordingly.<br /> (4) I would like to point to the fact that there are significant experimental modifications added to the manuscript. The decision from the first cycle of review was given on 8th Nov 2024. The authors re-submitted the manuscript on 20th Nov 2024. It just beats my understanding, how so many experiments can be done in such a short time. The rush in resubmission is evident in the writing quality as well. Which I think is now poorer than the original version.<br /> (5) The writing quality is still extremely poor.

Author response:

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

eLife assessment

This useful manuscript reports mechanisms behind the increase in fecundity in response to sub-lethal doses of pesticides in the crop pest, the brown plant hopper. The authors hypothesize that the pesticide works by inducing the JH titer, which through the JH signaling pathway induces egg development. Evidence for this is, however, inadequate.

We greatly appreciate your valuable comments and constructive suggestions for our work. All in all, the manuscript has been carefully edited and improved following your suggestions. We also provide more evidence to support our statements by conducting new experiments. First, we found that also EB treatment of adult females can stimulate egg-laying. Second, EB treatment in female adults increases the number of mature eggs in the ovary and ovarioles. Third, EB treatment in females enhances the expression of the kr-h1 gene in the whole body of BPH. Finally, EB treatment in female adults increases the JHIII titer, but has no impact on the 20E titer.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Gao et al. have demonstrated that the pesticide emamectin benzoate (EB) treatment of brown planthopper (BPH) leads to increased egg-laying in the insect, which is a common agricultural pest. The authors hypothesize that EB upregulates JH titer resulting in increased fecundity.

Strengths:

The finding that a class of pesticide increases the fecundity of brown planthopper is interesting.

We greatly appreciate your positive comments on our work.

Weaknesses:

(1) EB is an allosteric modulator of GluCl. That means EB physically interacts with GluCl initiating a structural change in the cannel protein. Yet the authors' central hypothesis here is about how EB can upregulate the mRNA of GluCl. I do not know whether there is any evidence that an allosteric modulator can function as a transcriptional activator for the same receptor protein. The basic premise of the paper sounds counterintuitive. This is a structural problem and should be addressed by the authors by giving sufficient evidence about such demonstrated mechanisms before.

Thank you for your question. As the reviewer points out, EB physically interacts with its target protein GluCl and thus affects its downstream signaling pathway. In the manuscript, we reported that EB-treated brown planthoppers display increased expression of GluCl in the adult stage (Fig. 5A). Actually, there are many studies showing that insects treated with insecticides can increase the expression of target genes. For example, the relative expression level of the ryanodine receptor gene of the rice stem borer, Chilo suppressalis was increased 10-fold after treatment with chlorantraniliprole, an insecticide which targets the ryanodine receptor (Peng et al., 2017). Besides this, in Drosophila, starvation (and low insulin) elevates the transcription level of the sNPF and tachykinin receptors (Ko et al., 2015; Root et al., 2011). In brown planthoppers, reduction in mRNA and protein expression of a nicotinic acetylcholine receptor α8 subunit is associated with resistance to imidacloprid (Zhang et al., 2015). RNA interference knockdown of α8 gene decreased the sensitivity of N. lugens to imidacloprid (Zhang et al., 2015). Hence, expression of receptor genes can be regulated by diverse factors including insecticide treatment. In our case, we found that EB can upregulate its target gene GluCl. However, we did not claim that EB functions as transcriptional activator for GluCl, and we still do not know why EB treatment changes the expression of GluCl in the brown planthopper. Considering our experiments are lasting several days, it might be an indirect (or secondary) effect caused by other factors, which change the expression of GluCl gene upon EB action of the channel. One reason is maybe that the allosteric interaction with GluCl by EB makes it dysfunctional and the cellular response is to upregulate the channel/receptor to compensate. We have inserted text on lines 738 - 757 to explain these possibilities.

(2) I am surprised to see a 4th instar larval application or treatment with EB results in the upregulation of JH in the adult stages. Complicating the results further is the observation that a 4th instar EB application results in an immediate decrease in JH titer. There is a high possibility that this late JH titer increase is an indirect effect.

Thank you for your question. Treatment with low doses or sublethal doses of insecticides might have a strong and complex impact on insects (Gandara et al., 2024; Gong et al., 2022; Li et al., 2023; Martelli et al., 2022). We kept the 4th instar of brown planthoppers feeding on EB for four days. They will develop to 5th instar after four days treatment, which is the final nymphal stage of BPH. Since the brown planthopper is a hemimetabolous insect, we cannot rule out the possibility that an indirect effect of treatment with EB results in the upregulation of JH in the adult stages. In this new revised manuscript, we investigated the impact of EB treatment in the adult stage. We found that female adults treated with EB also laid more eggs than controls (Figure 1-figure supplement 1A). The following experiments were performed in adults to address how EB treated stimulates egg-laying in adult brown planthopper.

(1) We found that EB treatment in adults increases the number of mature eggs in ovary (new Figure 2-figure supplement 1). We add this results in lines 234 – 238 and 281-285.

(2) We measured the JH titer after the female adults had been treated with EB. We found that EB can also increase the JH titer but has no impact on the 20E titer in the female adult (Figure 3-S3A and B). We add this results in lines 351 – 356 and 281-285.

(3) EB treatment in adults increases the gene expression of JHAMT and Kr-h1 (Figure 3-S3C and D). We add this results in lines 378 – 379, lines 387-390 and lines 457-462.

(3) The writing quality of the paper needs improvement. Particularly with respect to describing processes and abbreviations. In several instances the authors have not adequately described the processes they have introduced, thus confusing readers.

Thank you for your suggestion. We have thoroughly revised the paper to improve clarity.

(4) In the section 'EB promotes ovarian development' the authors have shown that EB treatment results in increased detention of eggs which contradicts their own results which show that EB promotes egg laying. Again, this is a serious contradiction that nullifies their hypothesis.

Thank you for pointing this out. We revised the figure 2B to show number of mature eggs in the ovary. The number of mature eggs in ovaries of females that fed on EB was higher than in control females. We also show that BPH fed with EB laid more eggs than controls. Thus, our results suggest that EB promotes ovary maturation (and egg production) and also increases egg laying (Figure 1 and Table S1). Thus, we found that EB treatment can increase both the production of eggs and increase egg laying. We add this results in lines 234 – 238.

(5) Furthermore, the results suggest that oogenesis is not affected by EB application. The authors should devote a section to discussing how they are observing increased egg numbers in EB-treated insects while not impacting Oogenesis.

Thank you for your suggestions, and apologies for the lack of clarity in our initial explanation. First, we found that EB treatment led to an increase in the number of eggs laid by female brown planthoppers (Figure 1). Through dissection experiments, we observed that EB-treated females had more mature eggs in their ovaries (Figure 2A and B), indicating that the increased egg-laying was due to a larger production of mature eggs in the ovaries after EB treatment. This is now explained on lines 229-238.

Additionally, since there is no systematic description of oogenesis in the brown planthopper, we were the first to observe the oogenesis process in this species using immunohistochemistry and laser confocal microscopy. Based on the developmental characteristics, we defined the different stages of oogenesis (Figure 2C, Figure 2-figure supplement 2). We did not observe any significant effect of EB treatment on the various stages of oogenesis, indicating that EB treatment does not impair normal egg development (Figure 2D). Instead, the increase in vitellogenin accelerates the production of mature eggs. This is now explained on lines 243-262.

During the maturation process, eggs require uptake of vitellogenin, and an increase in vitellogenin (Vg) content can accelerate egg maturation, producing more mature eggs. Our molecular data suggest that EB treatment leads to an upregulation of vg expression. Based on these findings, we conclude that the increase in egg-laying caused by EB treatment is due to the upregulation of vg (Figure 3I), which raises vitellogenin content, promoting the uptake of vitellogenin by maturing eggs and resulting in the production of more mature eggs. We have revised the text on lines 389-395 to clarify this point.

(6) Met is the receptor of JH and to my understanding, remains mostly constant in terms of its mRNA or protein levels throughout various developmental periods in many different insects. Therefore, the presence of JH becomes the major driving factor for physiological events and not the presence of the receptor Met. Here the authors have demonstrated an increase in Met mRNA as a result of EB treatment. Their central hypothesis is that EB increases JH titer to result in enhanced fecundity. JH action will not result in the activation of Met. Although not contradictory to the hypothesis, the increase in mRNA content of Met is contrary to the findings of the JH field thus far.

Thank you for your comment. Our results showed that EB treatment can mildly increase (about 2-fold) expression of the Met gene in brown planthoppers (Figure 3G). And our data indicated that Met and FAMeT expression levels were not influenced so much by EB compared with kr-h1 and vg (Figure 3H and I). We agree that JH action will not result in the increase of Met. However, we cannot rule out the possibility of other factors (indirect effects), induced by EB treatment that increase the mRNA expression level of Met. One recent paper reported that downregulation of transcription factor CncC will increase met expression in beetles (see Figure 6A in this reference) (Jiang et al., 2023). Many studies have reported that insecticide treatment will activate the CncC gene signaling pathway, which regulates detoxification gene expression (Amezian et al., 2023; Fu et al., 2024; Hu et al., 2021). Hence, it is possible that EB might influence the CncC gene pathway which then induces met expression. This EB effect on met upregulation may be similar to the upregulation of GluCl and some other secondary effects. We have discussed this on lines 725-738.

(7) As pointed out before, it is hard to rationalize how a 4th instar exposure to EB can result in the upregulation of key genes involved in JH synthesis at the adult stage. The authors must consider providing a plausible explanation and discussion in this regard.

Thank you for your comments. It must be mentioned that although we exposed the BPH to EB at 4th instar, we make the insect feed on the EB-treated rice plants for four days. After that, the insect will develop into 5^th instar, the final nymphal stage of brown planthopper. Since brown planthoppers do not have a pupal stage, this might cause the EB presented to the insects last a longer time even in the adult stage. Besides this, we found that EB treatment will increase the weight of adult females (Figure 1-figure supplement 3E and F), which indicates that EB might increase food intake in BPHs that might produce more insulin peptide. Insulin might increase the JH synthesis at the adult stage. In our revised study we also investigate EB impairment in adult BPHs. We found that, similar to the nymphal stage, EB treatment in adult BPHs also increases the egg laying. Furthermore, the JH titer was increased after treatment of BPH with EB in adults. Besides this, GluCl and kr-h1 genes were also up-regulated after EB treatment in the adult stage. We have discussed this on lines 739-746.

(8) I have strong reservations against such an irrational hypothesis that Met (the receptor for JH) and JH-Met target gene Kr-h1 regulate JH titer (Line 311, Fig 3 supplemental 2D). This would be the first report of such an event on the JH field and therefore must be analysed in depth. I strongly suggest the authors remove such claims from the manuscript without substantiating it.

Thank you for your suggestions and comments. We have changed our claims in this revised MS. We found that EB treatment can enhance Kr-h1 expression. We have no evidence to support that JH can induce met expression. We have rewritten the manuscript to avoid confusion (see text on lines 725-735).

(9) Kr-h1 is JH/Met target gene. The authors demonstrate that silencing of Kr-h1 results in inhibition of FAMeT, which is a gene involved in JH synthesis. A feedback loop in JH synthesis is unreported. It is the view of this reviewer that the authors must go ahead with a mechanistic detail of Kr-h1 mediated JH upregulation before this can be concluded. Mere qPCR experiments are not sufficient to substantiate a claim that is completely contrary to the current understanding of the JH signalling pathway.

Thank you for your suggestions and comments. We agree that only qPCR experiments are not enough to provide this kind of claim. More evidences need to be provided to support this. We have revised the MS to avoid confusion (see text on lines 725-735).

(10) The authors have performed knockdowns of JHAMT, Met, and Kr-h1 to demonstrate the effect of these factors on fecundity in BPH. Additionally, they have performed rescue experiments with EB application on these knockdown insects (Figure 3K-M). This, I believe, is a very flawed experiment. The authors demonstrate EB works through JHAMT in upregulating JH titer. In the absence of JHAMT, EB application is not expected to rescue the phenotype. But the authors have reported a complete rescue here. In the absence of Met, the receptor of JH, either EB or JH is not expected to rescue the phenotype. But a complete rescue has been reported. These two experimental results contradict their own hypothesis.

Thank you for your comments. We thought that this rescue is possible since knockdown of the genes is incomplete when using dsRNA injection (and residual gene expression allows for EB action). It is not a total knockout and actually, these genes still have a low level of expression in the dsRNA-injected insects. Since EB can upregulate the expression of JHAMT, Met, and Kr-h1, it is reasonable that EB treatment can rescue the down-regulation effects of these three genes and make fecundity completely rescued. We have clarified this on lines 411-413).

(11) A significant section of the paper deals with how EB upregulates JH titer. JH is a hormone synthesized in the Corpora Allata. Yet the authors have chosen to use the whole body for all of their experiment. Changes in the whole body for mRNA of those enzymes involved in JH synthesis may not reflect the situation in Corpora Allata. Although working with Corpora Allata is challenging, discarding the abdomen and thorax region and working with the head and neck region of the insect is easily doable. Results from such sampling are always more convincing when it comes to JH synthesis studies.

Thank you for your suggestions. Because the head is very difficult to separate from the thorax region in brown planthoppers as you can see in Author response image 1. We are now trying to answer how EB regulates JH synthesis using Drosophila as a model.

Author response image 1.

The brown planthopper

(12) The phenomenon reported was specific to BPH and not found in other insects. This limits the implications of the study.

Thank you for your comments. The brown planthopper is a serious insect pest on rice in Asia. Our findings can guide the use of this insecticide in the field. Besides this, our findings indicated that EB, which targets GluCl can impair the JH titer. Our findings added new implications for how a neuronal system influences the JH signaling pathway. We will further investigate how EB influences JH in the future and will use Drosophila as a model to study the molecular mechanisms.

(13) Overall, the molecular experiments are very poorly designed and can at best be termed superficial. There are several contradictions within the paper and no discussion or explanation has been provided for that.

Thank you for your comments. We have revised the paper according to your suggestions and added further explanation of our results in the discussion parts and hope the conclusions are better supported in the new version. We have discussed this on lines 725-746 and 778-799.

Reviewer #2 (Public Review):

The brown plant hopper (BPH) is a notorious crop pest and pesticides are the most widespread means of controlling its population. This manuscript shows that in response to sublethal doses of the pesticide (EB), BPH females show enhanced fecundity. This is in keeping with field reports of population resurgence post-pesticide treatment. The authors work out the mechanism behind this increase in fecundity. They show that in response to EB exposure, the expression of its target receptor, GluCl, increases. This, they show, results in an increase in the expression of genes that regulate the synthesis of juvenile hormone (JH) and JH itself, which, in turn, results in enhanced egg-production and egg-laying. Interestingly, these effects of EB exposure are species-specific, as the authors report that other species of plant hoppers either don't show enhanced fecundity or show reduced fecundity. As the authors point out, it is unclear how an increase in GluCl levels could result in increased JH regulatory genes.

We greatly appreciate your valuable comments and constructive suggestion to our work. We will try to figure out how EB interacts with its molecular target GluCl and then increases JH regulatory genes in the future work using Drosophila as models.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Overall, the molecular experiments are very poorly designed and can at best be termed superficial. There are several contradictions within the paper and no discussion or explanation has been provided for that.

The authors should consider a thorough revision.

Thank you for your comments. We have thoroughly revised the paper according to your suggestions and added further experiments and explanations of our results in the discussion parts.

Reviewer #2 (Recommendations For The Authors):

It would help the reader to have more schematics along with the figures. The final figure is helpful, but knowing the JH pathway, and where it acts would help with the interpretations as one reads the manuscript and the figures. The pathways represented in 4N or 5J are helpful but could be improved upon for better presentation.

It would be nice to have some discussion on how the authors think EB exposure results in an increase in GluCl expression, and how that in turn affects the expression of so many genes.

Thank you for your comments. We have thoroughly revised the paper according to your suggestions and added further experiments and explanations of how we think EB exposure results in an increase in JH titer and other genes in the discussion parts. We have added the test on lines 753-761.

References

Amezian, D., Fricaux, T., de Sousa, G., Maiwald, F., Huditz, H.-I., Nauen, R., Le Goff, G., 2023. Investigating the role of the ROS/CncC signaling pathway in the response to xenobiotics in Spodoptera frugiperda using Sf9 cells. Pesticide Biochemistry and Physiology 195, 105563.

Fu, B., Liang, J., Hu, J., Du, T., Tan, Q., He, C., Wei, X., Gong, P., Yang, J., Liu, S., Huang, M., Gui, L., Liu, K., Zhou, X., Nauen, R., Bass, C., Yang, X., Zhang, Y., 2024. GPCR–MAPK signaling pathways underpin fitness trade-offs in whitefly. Proceedings of the National Academy of Sciences 121, e2402407121.

Gandara, L., Jacoby, R., Laurent, F., Spatuzzi, M., Vlachopoulos, N., Borst, N.O., Ekmen, G., Potel, C.M., Garrido-Rodriguez, M., Böhmert, A.L., Misunou, N., Bartmanski, B.J., Li, X.C., Kutra, D., Hériché, J.-K., Tischer, C., Zimmermann-Kogadeeva, M., Ingham, V.A., Savitski, M.M., Masson, J.-B., Zimmermann, M., Crocker, J., 2024. Pervasive sublethal effects of agrochemicals on insects at environmentally relevant concentrations. Science 386, 446-453.

Gong, Y., Cheng, S., Desneux, N., Gao, X., Xiu, X., Wang, F., Hou, M., 2022. Transgenerational hormesis effects of nitenpyram on fitness and insecticide tolerance/resistance of Nilaparvata lugens. Journal of Pest Science.

Hu, B., Huang, H., Hu, S., Ren, M., Wei, Q., Tian, X., Esmail Abdalla Elzaki, M., Bass, C., Su, J., Reddy Palli, S., 2021. Changes in both trans- and cis-regulatory elements mediate insecticide resistance in a lepidopteron pest, Spodoptera exigua. PLOS Genetics 17, e1009403.

Jiang, H., Meng, X., Zhang, N., Ge, H., Wei, J., Qian, K., Zheng, Y., Park, Y., Reddy Palli, S., Wang, J., 2023. The pleiotropic AMPK–CncC signaling pathway regulates the trade-off between detoxification and reproduction. Proceedings of the National Academy of Sciences 120, e2214038120.

Ko, K.I., Root, C.M., Lindsay, S.A., Zaninovich, O.A., Shepherd, A.K., Wasserman, S.A., Kim, S.M., Wang, J.W., 2015. Starvation promotes concerted modulation of appetitive olfactory behavior via parallel neuromodulatory circuits. eLife 4, e08298.

Li, Z., Wang, Y., Qin, Q., Chen, L., Dang, X., Ma, Z., Zhou, Z., 2023. Imidacloprid disrupts larval molting regulation and nutrient energy metabolism, causing developmental delay in honey bee Apis mellifera. eLife

Martelli, F., Hernandes, N.H., Zuo, Z., Wang, J., Wong, C.-O., Karagas, N.E., Roessner, U., Rupasinghe, T., Robin, C., Venkatachalam, K., Perry, T., Batterham, P., Bellen, H.J., 2022. Low doses of the organic insecticide spinosad trigger lysosomal defects, elevated ROS, lipid dysregulation, and neurodegeneration in flies. eLife 11, e73812.

Peng, Y.C., Sheng, C.W., Casida, J.E., Zhao, C.Q., Han, Z.J., 2017. Ryanodine receptor genes of the rice stem borer, Chilo suppressalis: Molecular cloning, alternative splicing and expression profiling. Pestic. Biochem. Physiol. 135, 69-77.

Root, Cory M., Ko, Kang I., Jafari, A., Wang, Jing W., 2011. Presynaptic facilitation by neuropeptide signaling mediates odor-driven food search. Cell 145, 133-144.

Zhang, Y., Wang, X., Yang, B., Hu, Y., Huang, L., Bass, C., Liu, Z., 2015. Reduction in mRNA and protein expression of a nicotinic acetylcholine receptor α8 subunit is associated with resistance to imidacloprid in the brown planthopper, Nilaparvata lugens. Journal of Neurochemistry 135, 686-694.

eLife Assessment

This valuable study confirms the association between the human leukocyte antigen (HLA)-II region and tuberculosis (TB) susceptibility in genetically admixed South African populations, specifically identifying a near-genome-wide significant association in the HLA-DPB1 gene, which originates from KhoeSan ancestry. The evidence supporting the association between the HLA-II region and TB susceptibility is solid, and the work will be of interest to those studying the genetic basis of tuberculosis susceptibility/infection resistance.

Reviewer #1 (Public review):

Summary:

The authors aimed to confirm the association between the human leukocyte antigen (HLA)-II region and tuberculosis (TB) susceptibility within admixed African populations. Building upon previous findings from the International Tuberculosis Host Genetics Consortium (ITHGC), this study sought to address the limitations of small sample size and the inclusion of admixed samples by employing the Local Ancestry Allelic Adjusted (LAAA) model, as well as identify TB susceptibility loci in an admixed South African cohort.

Strengths:

The major strengths of this study include the use of multiple TB case-control datasets from diverse South African populations and ADMIXTURE for global ancestry inference.

Weaknesses:

The major weakness of this study include insufficient significant novel discoveries and reliance on cross-validation. The use of existing models did not add value to this study.

Appraisal:<br /> The authors achieved their aims. However, the results still needed to be further validated in the future.

Impact:<br /> The innovative use of the LAAA model and the comprehensive dataset in this study may make contributions to the field of genetic epidemiology.

Reviewer #2 (Public review):

Summary:

This manuscript is about using different analytical approaches to allow ancestry adjustments to GWAS analyses amongst admixed populations. This work is a follow-on from the recently published ITHGC multi-population GWAS (https://doi.org/10.7554/eLife.84394), with the focus on the admixed South African populations. Ancestry adjustment models detected a peak of SNPs in the class II HLA DPB1, distinct from the class II HLA DQA1 loci signficant in the ITHGC analysis.

Strengths:

Excellent demonstration of GWAS analytical pipelines in highly admixed populations. Particularly the utility of ancestry adjustment to improve study power to detect novel associations. Further confirmation of the importance of the HLA class II locus in genetic susceptibility to TB.

Weaknesses:

Limited novelty compared to the group's previous existing publications and the body of work linking HLA class II alleles with TB susceptibility in South Africa or other African populations. This work includes only ~100 new cases and controls from what has already been published. High resolution HLA typing has detected significant signals in both the DQA1 and DPB1 regions identified by the larger ITHGC and in this GWAS analysis respectively (Chihab L et al. HLA. 2023 Feb; 101(2): 124-137).<br /> Despite the availability of strong methods for imputing HLA from GWAS data (Karnes J et Plos One 2017), the authors did not confirm with HLA typing the importance of their SNP peak in the class II region. This would have supported the importance of this ancestry adjustment versus prior ITHGC analysis.<br /> The populations consider active TB and healthy controls (from high-burden presumed exposed communities) and do not provide QFT or other data to identify latent TB infection.

Important methodological points for clarification and for readers to be aware of when reading this paper:

(1) One of the reasons cited for the lack of African ancestry-specific associations or suggestive peaks in the ITHGC study was the small African sample size. The current association test includes a larger African cohort and yields a near-genome-wide significant threshold in the HLA-DPB1 gene originating from the KhoeSan ancestry. Investigation is needed as to whether the increase in power is due to increased African samples and not necessarily the use of the LAAA model as stated on lines 295 and 296?

Authors response - The Manhattan plot in Figure 3 includes the results for all four models: the traditional GWAS model (GAO), the admixture mapping model (LAO), the ancestry plus allelic (APA) model and the LAAA model. In this figure, it is evident that only the LAAA model identified the association peak on chromosome 6, which lends support the argument that the increase in power is due to the use of the LAAA model and not solely due to the increase in sample size.<br /> Reviewer comment - This data supports the authors conclusions that increase power is related to the LAAA model application rather than simply increase sample size.

(2) In line 256, the number of SNPs included in the LAAA analysis was 784,557 autosomal markers; the number of SNPs after quality control of the imputed dataset was 7,510,051 SNPs (line 142). It is not clear how or why ~90% of the SNPs were removed. This needs clarification.

Authors response:<br /> In our manuscript (line 194), we mention that "...variants with minor allele frequency (MAF) < 1% were removed to improve the stability of the association tests." A large proportion of imputed variants fell below this MAF threshold and were subsequently excluded from this analysis.

Reviewers additional comment: The authors should specify the number of SNPs in the dataset before imputation and indicate what proportion of the 784,557 remaining SNPs were imputed. Providing this information might help the reader better understand the rationale behind the imputation process.

(3) The authors have used the significance threshold estimated by the STEAM p-value < 2.5x10-6 in the LAAA analysis. Grinde et al. (2019 implemented their significance threshold estimation approach tailored to admixture mapping (local ancestry (LA) model), where there is a reduction in testing burden. The authors should justify why this threshold would apply to the LAAA model (a joint genotype and ancestry approach).

Authors response: We describe in the methods (line 189 onwards) that the LAAA model is an extension of the APA model. Since the APA model itself simultaneously performs the null global ancestry only model and the local ancestry model (utilised in admixture mapping), we thus considered the use of a threshold tailored to admixture mapping appropriate for the LAAA model.

Reviewers additional comment: While the LAAA model is an extension of the APA model, the authors describe the LAAA test as 'models the combination of the minor allele and the ancestry of the minor allele at a specific locus, along with the effect of this interaction,' thus a joint allele and ancestry effects model. Grinde et al. (2019) proposed the significance threshold estimation approach, STEAM, specifically for the LA approach, which tests for ancestry effects alone and benefits from the reduced testing burden. However, it remains unclear why the authors found it appropriate to apply STEAM to the LAAA model, a joint test for both allele and ancestry effects, which does not benefit from the same reduction in testing burden.

(4) Batch effect screening and correction (line 174) is a quality control check. This section is discussed after global and local ancestry inferences in the methods. Was this QC step conducted after the inferencing? If so, the authors should justify how the removed SNPs due to the batch effect did not affect the global and local ancestry inferences or should order the methods section correctly to avoid confusion.

Authors response: The batch effect correction method utilised a pseudo-case-control comparison which included global ancestry proportions. Thus, batch effect correction was conducted after ancestry inference. We excluded 36 627 SNPs that were believed to have been affected by the batch effect. We have amended line 186 to include the exact number of SNPs excluded due to batch effect.<br /> The ancestry inference by RFMix utilised the entire merged dataset of 7 510 051 SNPs. Thus, the SNPs removed due to the batch effect make up a very small proportion of the SNPs used to conduct global and local ancestry inferences (less than 0.5%). As a result, we do not believe that the removed SNPs would have significantly affected the global and local ancestry inferences. However, we did conduct global ancestry inference with RFMix on each separate dataset as a sanity check. In the Author response tables 1 and 2, we show the average global ancestry proportions inferred for each separate dataset, the average global ancestry proportions across all datasets and the average global ancestry proportions inferred using the merged dataset. The SAC and Xhosa cohorts are shown in two separate tables due to the different number of contributing ancestral populations to each cohort. The differences between the combined average global ancestry proportions across the separate cohorts does not differ significantly to the global ancestry proportions inferred using the merged dataset.

This is an excellent response and should remain accessible to readers to clarify this issue.

Author response:

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

Reviewer #1 (Public review): 

Summary: 

The authors aimed to confirm the association between the human leukocyte antigen (HLA)-II region and tuberculosis (TB) susceptibility within admixed African populations. Building upon previous findings from the International Tuberculosis Host Genetics Consortium (ITHGC), this study sought to address the limitations of small sample size and the inclusion of admixed samples by employing the Local Ancestry Allelic Adjusted (LAAA) model, as well as identify TB susceptibility loci in an admixed South African cohort. 

Strengths: 

The major strengths of this study include the use of six TB case-control datasets collected over 30 years from diverse South African populations and ADMIXTURE for global ancestry inference. The former represents comprehensive dataset used in this study and the later ensures accurate determination of ancestral contributions. In addition, the identified association in the HLA-DPB1 gene shows near-genomewide significance, enhancing the credibility of the findings. 

Weaknesses: 

The major weakness of this study includes insufficient significant discoveries and reliance on crossvalidation. This study only identified one variant significantly associated with TB status, located in an intergenic region with an unclear link to TB susceptibility. Despite identifying multiple lead SNPs, no other variants reached the genome-wide significance threshold, limiting the overall impact of the findings. The absence of an independent validation cohort, with the study relying solely on crossvalidation, is also a major limitation. This approach restricts the ability to independently confirm the findings and evaluate their robustness across different population samples. 

Appraisal: 

The authors successfully achieved their aims of confirming the association between the HLA-II region and TB susceptibility in admixed African populations. However, the limited number of significant discoveries, reliance on cross-validation, and insufficient discussion of model performance and SNP significance weaken the overall strength of the findings. Despite these limitations, the results support the conclusion that considering local ancestry is crucial in genetic studies of admixed populations. 

Impact:  

The innovative use of the LAAA model and the comprehensive dataset in this study make substantial contributions to the field of genetic epidemiology. 

Reviewer #2 (Public review): 

Summary: 

This manuscript is about using different analytical approaches to allow ancestry adjustments to GWAS analyses amongst admixed populations. This work is a follow-on from the recently published ITHGC multi-population GWAS (https://doi.org/10.7554/eLife.84394), with a focus on the admixed South African populations. Ancestry adjustment models detected a peak of SNPs in the class II HLA DPB1, distinct from the class II HLA DQA1 loci significant in the ITHGC analysis. 

Strengths: 

Excellent demonstration of GWAS analytical pipelines in highly admixed populations. Further confirmation of the importance of the HLA class II locus in genetic susceptibility to TB. 

Weaknesses: 

Limited novelty compared to the group's previous existing publications and the body of work linking HLA class II alleles with TB susceptibility in South Africa or other African populations. This work includes only ~100 new cases and controls from what has already been published. High-resolution HLA typing has detected significant signals in both the DQA1 and DPB1 regions identified by the larger ITHGC and in this GWAS analysis respectively (Chihab L et al. HLA. 2023 Feb; 101(2): 124-137). Despite the availability of strong methods for imputing HLA from GWAS data (Karnes J et Plos One 2017), the authors did not confirm with HLA typing the importance of their SNP peak in the class II region. This would have supported the importance of this ancestry adjustment versus prior ITHGC analysis. 

The populations consider active TB and healthy controls (from high-burden presumed exposed communities) and do not provide QFT or other data to identify latent TB infection. 

Important methodological points for clarification and for readers to be aware of when reading this paper: 

(1) One of the reasons cited for the lack of African ancestry-specific associations or suggestive peaks in the ITHGC study was the small African sample size. The current association test includes a larger African cohort and yields a near-genome-wide significant threshold in the HLA-DPB1 gene originating from the KhoeSan ancestry. The investigation is needed as to whether the increase in power is due to increased African samples and not necessarily the use of the LAAA model as stated on lines 295 and 296? 

Thank you for your comment. The Manhattan plot in Figure 3 includes the results for all four models: the traditional GWAS model (GAO), the admixture mapping model (LAO), the ancestry plus allelic (APA) model and the LAAA model. In this figure, it is evident that only the LAAA model identified the association peak on chromosome 6, which lends support the argument that the increase in power is due to the use of the LAAA model and not solely due to the increase in sample size. 

(2) In line 256, the number of SNPs included in the LAAA analysis was 784,557 autosomal markers; the number of SNPs after quality control of the imputed dataset was 7,510,051 SNPs (line 142). It is not clear how or why ~90% of the SNPs were removed. This needs clarification. 

Thank you for your recommendation. In our manuscript (line 194), we mention that “…variants with minor allele frequency (MAF) < 1% were removed to improve the stability of the association tests.” A large proportion of imputed variants fell below this MAF threshold, and were subsequently excluded from this analysis. Below, we show the number of imputed variants across MAF bins for one of our datasets [RSA(A)] to substantiate this claim:  

Author response image 1.

(3) The authors have used the significance threshold estimated by the STEAM p-value < 2.5x10^-6 in the LAAA analysis. Grinde et al. (2019 implemented their significance threshold estimation approach tailored to admixture mapping (local ancestry (LA) model), where there is a reduction in testing burden. The authors should justify why this threshold would apply to the LAAA model (a joint genotype and ancestry approach). 

Thank you for your recommendation. We describe in the methods (line 189 onwards) that the LAAA model is an extension of the APA model. Since the APA model itself simultaneously performs the null global ancestry only model and the local ancestry model (utilised in admixture mapping), we thus considered the use of a threshold tailored to admixture mapping appropriate for the LAAA model.  

(4) Batch effect screening and correction (line 174) is a quality control check. This section is discussed after global and local ancestry inferences in the methods. Was this QC step conducted after the inferencing? If so, the authors should justify how the removed SNPs due to the batch effect did not affect the global and local ancestry inferences or should order the methods section correctly to avoid confusion. 

Thank you for your comments. The batch effect correction method utilised a pseudo-case-control comparison which included global ancestry proportions. Thus, batch effect correction was conducted after ancestry inference. We excluded 36 627 SNPs that were believed to have been affected by the batch effect. We have amended line 186 to include the exact number of SNPs excluded due to batch effect. 

The ancestry inference by RFMix utilised the entire merged dataset of 7 510 051 SNPs. Thus, the SNPs removed due to the batch effect make up a very small proportion of the SNPs used to conduct global and local ancestry inferences (less than 0.5%). As a result, we do not believe that the removed SNPs would have significantly affected the global and local ancestry inferences. However, we did conduct global ancestry inference with RFMix on each separate dataset as a sanity check. In the tables below, we show the average global ancestry proportions inferred for each separate dataset, the average global ancestry proportions across all datasets and the average global ancestry proportions inferred using the merged dataset. The SAC and Xhosa cohorts are shown in two separate tables due to the different number of contributing ancestral populations to each cohort. The differences between the combined average global ancestry proportions across the separate cohorts does not differ significantly to the global ancestry proportions inferred using the merged dataset. 

Author response table 1.

Comparison of global ancestry proportions across the separate SAC datasets and the merged cohort.

Author response table 2.

Comparison of global ancestry proportions in the Xhosa dataset and the merged cohort. 

Reviewer #1 (Recommendations for the authors): 

Suggestions for Improved or Additional Experiments, Data, or Analyses:   

(1) It might be beneficial to consider splitting the data into separate discovery and validation cohorts rather than relying solely on cross-validation. This approach could provide a stronger basis for independently confirming the findings. 

Thank you for your suggestion. However, we are hesitant to divide our already modest dataset (n=1544) into separate discovery and validation cohorts, as this would reduce the statistical power to detect significant associations.

(2) Clearly stating the process of cross-validation in the methods section and reporting relevant validation statistics, such as accuracy, sensitivity, specificity, and area under the curve (AUC), would provide a more comprehensive assessment of the model's performance.  

Thank you for your recommendation. We would like to highlight this article, “GWAS in the southern African context” (1), which evaluated the performance of the LAAA model compared to other models in three- and five-way admixed populations. Given the thorough evaluation of the model’s performance in that study, we did not find it necessary to reassess its performance in this manuscript.   

(3) Analysing racial cohorts separately to see if you can replicate previous results and find significant markers in combined non-African populations that are not evident in African-only samples might be useful. 

Thank you for your suggestion. We would like to respectfully note that race is a social construct, and its use as a proxy for genetic ancestry can be problematic (2). In our study, we rather rely on genetic ancestry inferred using ancestry inference software to provide a more accurate representation of our cohort's genetic diversity. Additionally, our cohort consists mostly of a highly admixed population group, with some individuals exhibiting ancestral contributions from up to five different global populations. Therefore, it is not possible to categorize our samples into distinct “Africanonly” or “non-African” groups.

(4) It might be worthwhile to consider using polygenic risk scores (PRS) to combine multiple genetic influences. This approach could help in identifying cumulative genetic effects that are not apparent when examining individual SNPs.  

Thank you for your recommendation. While constructing a polygenic risk score (PRS) is beyond the scope of the current study, but an ongoing interest in our group, we recognize its potential value and will consider incorporating this approach in future research endeavours or a separate publication. A recent publication by Majara et al showed that that PRS accuracy is low for all traits and varies across ancestrally and ethnically diverse South African groups (3).

Recommendations for Improving the Writing and Presentation: 

Including a more thorough discussion of the methodological limitations, such as the challenges of studying admixed populations and the potential limitations of the LAAA model, would provide a more balanced perspective. 

Thank you for your suggestion. To provide a more balanced perspective, we included the limitations of our study in the discussion, from line 429 to like 451.

Minor Corrections to the Text and Figures: 

Including all relevant statistics would improve clarity. For example, providing confidence intervals for the odds ratios and discussing any observed trends or outliers would be beneficial. 

Thank you for your recommendation. We have added 95% confidence intervals to all odds ratios reported in Table 3. However, beyond the association peak identified in the HL-II region associated with the phenotype, we do not observe any other trends or outliers in or LAAA analysis.  

Reviewer #2 (Recommendations for the authors): 

Points for improvement: 

(1) Related to the different datasets and inclusions in previous publications, it would also be good to better understand the different numbers of cases and controls included across the previous and current analyses, or discussion thereof. For instance, the RSA(M) dataset includes 555/440 cases/controls for this analysis and only 410/405 cases/controls in the ITHGC analysis. Other discrepancies are noted across the other published datasets compared to those included in this analysis, and these always need to be detailed in a supplement or similar to better understand if this could have introduced bias or was in fact correct based on the additional ancestry-related restriction applied.  

Thank you for your comments. Table 1 of our manuscript lists number of individuals in the RSA(M) dataset, including related individuals. As described in line 131, related individuals were subsequently excluded during quality control: “Individual datasets were screened for relatedness using KING software (Manichaikul et al., 2010) and individuals up to second degree relatedness were removed.” The ITHGC only reported the number of unrelated individuals included their analyses, which would account for the discrepancies in the reported number of cases and controls.  

(2) The imbalance between cases and controls in this analysis is quite striking, and it is unusual to have the imbalance favour cases over controls. This contrasts with the ITHGC, where there are substantially more controls. There is no comment on how this could potentially impact this analysis. 

Thank you for your comment. We have included a note on our case-control imbalance in the discussion:

“While many studies discuss methods for addressing case-control imbalances with more controls than cases (which can inflate type 1 error rates (Zhou et al. 2018; Dai et al. 2021; Öztornaci et al. 2023), few address the implications of a large case-to-control ratio like ours (952 cases to 592 controls). To assess the impact of this imbalance, we used the Michigan genetic association study (GAS) power calculator (Skol et al. 2006). Under an additive disease model with an estimated prevalence of 0.15, a disease allele frequency of 0.3, a genotype relative risk of 1.5, and a default significance level of 7 × 10^-6, we achieved an expected power of approximately 75%. With a balanced sample size of 950 cases and 950 controls, power would exceed 90%, but it would drop significantly with a smaller balanced cohort of 590 cases and 590 controls. Given these results, we proceeded with our analysis to maximize statistical power despite the case-control imbalance.” 

Author response image 2.

Minor comments 

(1) Referencing around key points of TB epidemiology and disease states seems out of date, given recent epidemiology reviews and seminal nature or lancet review articles. Please update.  

Thank you for your suggestion. We have included the following recent publications in the introductory paragraph: 

Zaidi, S. M. A., Coussens, A. K., Seddon, J. A., Kredo, T., Warner, D., Houben, R. M. G. J., & Esmail, H. (2023). Beyond latent and active tuberculosis: a scoping review of conceptual frameworks. EClinicalMedicine, 66, 102332. https://doi.org/10.1016/j.eclinm.2023.102332

Menzies, N. A., Swartwood, N., Testa, C., Malyuta, Y., Hill, A. N., Marks, S. M., Cohen, T., & Salomon, J. A. (2021). Time Since Infection and Risks of Future Disease for Individuals with Mycobacterium tuberculosis Infection in the United States. Epidemiology, 32(1), 70–78. https://doi.org/10.1097/EDE.0000000000001271  

Cudahy, P. G. T., Wilson, D., & Cohen, T. (2020). Risk factors for recurrent tuberculosis after successful treatment in a high burden setting: a cohort study. BMC Infectious Diseases, 20(1), 789. https://doi.org/10.1186/s12879-020-05515-4  

Escombe, A. R., Ticona, E., Chávez-Pérez, V., Espinoza, M., & Moore, D. A. J. (2019). Improving natural ventilation in hospital waiting and consulting rooms to reduce nosocomial tuberculosis transmission risk in a low resource setting. BMC Infectious Diseases, 19(1), 88. https://doi.org/10.1186/s12879-019-3717-9  

Laghari, M., Sulaiman, S. A. S., Khan, A. H., Talpur, B. A., Bhatti, Z., & Memon, N. (2019). Contact screening and risk factors for TB among the household contact of children with active TB: a way to find source case and new TB cases. BMC Public Health, 19(1), 1274. https://doi.org/10.1186/s12889-0197597-0  

Matose, M., Poluta, M., & Douglas, T. S. (2019). Natural ventilation as a means of airborne tuberculosis infection control in minibus taxis. South African Journal of Science, 115(9/10). https://doi.org/10.17159/sajs.2019/5737

Smith, M. H., Myrick, J. W., Oyageshio, O., Uren, C., Saayman, J., Boolay, S., van der Westhuizen, L., Werely, C., Möller, M., Henn, B. M., & Reynolds, A. W. (2023). Epidemiological correlates of overweight and obesity in the Northern Cape Province, South Africa. PeerJ, 11, e14723. https://doi.org/10.7717/peerj.14723  

(2) Lines 46 to 48 appear to have two contradictory statements next to each other. The first says there are numerous GWAS investigating TB susceptibility; the second says there are sparse. Please clarify.

Thank you for bringing this to our attention. We have amended the lines as follows: 

“Numerous genome-wide association studies (GWASs) investigating TB susceptibility have been conducted across different population groups. However, findings from these studies often do not replicate across population groups (Möller & Kinnear, 2020; Möller et al., 2018; Uren et al., 2017).”

(3) Add ref in line 69 for two SAC populations.

Thank you for your recommendation. We have included the citation for the ITHGC meta-analysis paper here: 

“The authors described possible reasons for the lack of associations, including the smaller sample size compared to the other ancestry-specific meta-analyses, increased genetic diversity within African individuals and population stratification produced by two admixed cohorts from the South African Coloured (SAC) population (Schurz et al. 2024).”

(4) Write out abbreviations the first time they appear (Line 121).

Thank you for your recommendation. We have corrected the sentence as follows: 

“Monomorphic sites were removed. Individuals were screened for deviations in Hardy-Weinberg Equilibrium (HWE) for each SNP and sites deviating from the HWE threshold of 10-5 were removed.”

(5) It would be good in the supplement to see if there is a SNP peak in chromosome 20 with a hit that reached significance in the Bantu-speaking African ancestry.

Thank you for your recommendation. We have included a regional plot for the lead variant identified on chromosome 20 originating from Bantu-speaking African ancestry in the supplementary material (Supplementary Figure 3).

(6) It would be good to mention the p-values of rs28383206 from the ITHGC paper in this cohort for KhoeSan and Bantu-speaking African ancestries. 

Thank you for your suggestion. We have included the following paragraph from line 352:

“The lead variant identified in the ITHGC meta-analysis, rs28383206, was not present in our genotype or imputed datasets. The ITHGC imputed genotypes using the 1000 Genomes (1000G) reference panel (4). Variant rs28383206 has an alternate allele frequency of 11.26% in the African population subgroup within the 1000G dataset (https://www.ncbi.nlm.nih.gov/snp/rs28383206). However, rs28383206 is absent from our in-house whole-genome sequencing (WGS) datasets, which include Bantu-speaking African and KhoeSan individuals. This absence suggests that rs28383206 might not have been imputed in our datasets using the AGR reference panel, potentially due to its low alternate allele frequency in southern African populations. Our merged dataset contained two variants located within 800 base pairs of r_s28383206: rs482205_ (6:32576009) and rs482162 (6:32576019). However, these variants were not significantly associated with TB status in our cohort (Supplementary Table 1).” Supplementary Table 1 can be found in the supplementary material:

(7) It would improve the readability of the ancestry proportions listed on lines 236 and 237 if these population groups were linked with the corresponding specific population used in Figure 1, as has been done in Table 2.

Thank you for your suggestion. We have amended Figure 1 to include the corresponding population labels mentioned in Table 2.  

(8) In line 209, it is not clear why the number of alleles of a specific ancestry at a locus is referred to as a covariate in admixture mapping when the corresponding marginal effect is the parameter of interest. 

Thank you for bringing this to our attention. We have amended the description as follows: 

“(2) Local ancestry (LA) model:

This model is used in admixture mapping to identify ancestry-specific variants associated with a specific phenotype. The LA model evaluates the number of alleles of a specific ancestry at a locus and includes the corresponding marginal effect as a covariate in association analyses.”

(9) Table 3 would benefit from a column on whether the SNP was genotyped or imputed. 

Thank you for your suggestion. We have included a column indicating whether the SNP was genotyped or imputed, as well as an additional column with the INFO score for imputed genotypes. 

(10) The authors should remove the print and download icons in Figure 1 on lines 240 and 241.

Thank you for your suggestion. We have amended the figure as requested.  

(11) In the quality control, the authors use a more relaxed threshold for missingness in individuals (90%) and genotypes (5%) and have strayed away from the conventional 97%-98%. An explanation of the choice of these thresholds will be helpful to the reader.

Thank you for your suggestion. We aimed to use similar genotype and individual missingness thresholds outline by the ITHGC meta-analysis (which utilised a threshold of 10% for both genotype and individual missingness) and the previous LAAA analysis paper performed by Swart et al. in 2021. We have amended line 116 for more clarity: 

“Individuals with genotype call rates less than 90% and SNPs with more than 5% missingness were removed as described previously (5).”

References  

(1) Swart Y, van Eeden G, Uren C, van der Spuy G, Tromp G, Moller M. GWAS in the southern African context. Cold Spring Harbor Laboratory. 2022;

(2) Byeon YJJ, Islamaj R, Yeganova L, Wilbur WJ, Lu Z, Brody LC, et al. Evolving use of ancestry, ethnicity, and race in genetics research-A survey spanning seven decades. Am J Hum Genet. 2021 Dec 2;108(12):2215–23.

(3) Majara L, Kalungi A, Koen N, Tsuo K, Wang Y, Gupta R, et al. Low and differential polygenic score generalizability among African populations due largely to genetic diversity. HGG Adv. 2023 Apr 13;4(2):100184.

(4) Schurz H, Naranbhai V, Yates TA, Gilchrist JJ, Parks T, Dodd PJ, et al. Multi-ancestry metaanalysis of host genetic susceptibility to tuberculosis identifies shared genetic architecture. eLife. 2024 Jan 15;13.

(5) Swart Y, Uren C, van Helden PD, Hoal EG, Möller M. Local ancestry adjusted allelic association analysis robustly captures tuberculosis susceptibility loci. Front Genet. 2021 Oct 15;12:716558.

eLife Assessment

Studying several allergens in different mouse strains, the authors assessed the role of IgM in airway inflammatory responses and show that IgM deficient mice have reduced airway hyperresponsiveness. Although the findings are useful and interesting and among others show the expression of a protein that regulates actin in smooth cells, the study remains incomplete as the data and analyses only partly support their primary claim.

Reviewer #1 (Public review):

Summary:

The authors of this study sought to define a role for IgM in responses to house dust mites in the lung.

Strengths:

Unexpected observation about IgM biology.<br /> Combination of experiments to elucidate function.

Weaknesses:

Would love more connection to human disease

Reviewer #2 (Public review):

Summary:

The manuscript by Hadebe and colleagues describes a striking reduction in airway hyperresponsiveness in Igm-deficient mice in response to HDM, OVA and papain across the B6 and BALB-c backgrounds. The authors suggest that the deficit is not due to improper type 2 immune responses, nor an aberrant B cell response, despite a lack of class switching in these mice. Through RNA-Seq approaches, the authors identify few differences between the lungs of WT and Igm-deficient mice, but see that two genes involved in actin regulation are greatly reduced in IgM-deficient mice. The authors target these genes by CRISPR-Cas9 in in vitro assays of smooth muscle cells to show that these may regulate cell contraction. While the study is conceptually interesting, there are a number of limitations, which stop us from drawing meaningful conclusions.

Strengths:

Fig. 1. The authors clearly show that IgMKO mice have striking reduced AHR in the HDM model, despite the presence of a good cellular B cell response.

Weaknesses:

Due to several technical and experimental limitations, it is unclear what leads to the reduction in airway hyperresponsiveness in IGM-KO mice. The limitations as outlined previously remain.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary: The authors of this study sought to define a role for IgM in responses to house dust mites in the lung.

Strengths:

Unexpected observation about IgM biology

Combination of experiments to elucidate function

Weaknesses:

Would love more connection to human disease

We thank the reviewer for these comments. At the time of this publication, we have not made a concrete link with human disease. While there is some anecdotal evidence of diseases such as Autoimmune glomerulonephritis, Hashimoto’s thyroiditis, Bronchial polyp, SLE, Celiac disease and other diseases in people with low IgM. Allergic disorders are also common in people with IgM deficiency, other studies have reported as high as 33-47%. The mechanisms for the high incidence of allergic diseases are unclear as generally, these patients have normal IgG and IgE levels. IgM deficiency may represent a heterogeneous spectrum of genetic defects, which might explain the heterogeneous nature of disease presentations. 

Reviewer #2 (Public Review):

Summary:

The manuscript by Hadebe and colleagues describes a striking reduction in airway hyperresponsiveness in Igm-deficient mice in response to HDM, OVA and papain across the B6 and BALB-c backgrounds. The authors suggest that the deficit is not due to improper type 2 immune responses, nor an aberrant B cell response, despite a lack of class switching in these mice. Through RNA-Seq approaches, the authors identify few differences between the lungs of WT and Igm-deficient mice, but see that two genes involved in actin regulation are greatly reduced in IgM-deficient mice. The authors target these genes by CRISPR-Cas9 in in vitro assays of smooth muscle cells to show that these may regulate cell contraction. While the study is conceptually interesting, there are a number of limitations, which stop us from drawing meaningful conclusions.

Strengths:

Fig. 1. The authors clearly show that IgMKO mice have striking reduced AHR in the HDM model, despite the presence of a good cellular B cell response.

Weaknesses:

Fig. 2. The authors characterize the cd4 t cell response to HDM in IGMKO mice.<br /> They have restimulated medLN cells with antiCD3 for 5 days to look for IL-4 and IL-13, and find no discernible difference between WT and KO mice. The absence of PBS-treated WT and KO mice in this analysis means it is unclear if HDM-challenged mice are showing IL-4 or IL-13 levels above that seen at baseline in this assay.

We thank the Reviewer for this comment. We would like to mention that a very minimal level of IL-4 and IL-13 in PBS mice was detected. We have indicated with a dotted line on the Figure to show levels in unstimulated or naïve cytokines. Please see Author response image 1 below from anti-CD3 stimulated cytokine ELISA data. The levels of these cytokines are very low and are not changed between WT and IgM^-/- mice, this is also true for PMA/ionomycin-stimulated cells.

Author response image 1.

The choice of 5 days is strange, given that the response the authors want to see is in already primed cells. A 1-2 day assay would have been better.

We agree with the reviewer that a shorter stimulation period would work. Over the years we have settled for 5-day re-stimulation for both anti-CD3 and HDM. We have tried other time points, but we consistently get better secretion of cytokines after 5 days.

It is concerning that the authors state that HDM restimulation did not induce cytokine production from medLN cells, since countless studies have shown that restimulation of medLN would induce IL-13, IL-5 and IL-10 production from medLN. This indicates that the sensitization and challenge model used by the authors is not working as it should.

We thank the reviewer for this observation. In our recent paper showing how antigen load affects B cell function, we used very low levels of HDM to sensitise and challenge mice (1 ug and 3 ug respectively). See below article, Hadebe et al., 2021 JACI. This is because Labs that have used these low HDM levels also suggested that antigen load impacts B cell function, especially in their role in germinal centres. We believe the reason we see low or undetectable levels of cytokines is because of this low antigen load sensitisation and challenge. In other manuscripts we have published or about to publish, we have shown that normal HDM sensitisation load (1 ug or 100 ug) and challenge (10 ug) do induce cytokine release upon restimulation with HDM. See the below article by Khumalo et al, 2020 JCI Insight (Figure 4A).

Sabelo Hadebe, Jermaine Khumalo, Sandisiwe Mangali, Nontobeko Mthembu, Hlumani Ndlovu, Amkele Ngomti, Martyna Scibiorek, Frank Kirstein, Frank Brombacher. Deletion of IL-4Ra signalling on B cells limits hyperresponsiveness depending on antigen load. doi.org/10.1016/j.jaci.2020.12.635).

Jermaine Khumalo, Frank Kirstein, Sabelo Hadebe, Frank Brombacher. IL-4Rα signalling in regulatory T cells is required for dampening allergic airway inflammation through inhibition of IL-33 by type 2 innate lymphoid cells. JCI Insight. 2020 Oct 15;5(20):e136206. doi: 10.1172/jci.insight.136206

The IL-13 staining shown in panel c is also not definitive. One should be able to optimize their assays to achieve a better level of staining, to my mind.

We agree with the reviewer that much higher IL-13-producing CD4 T cells should be observed. We don’t think this is a technical glitch or non-optimal set-up as we see much higher levels of IL-13-producing CD4 T cells when using higher doses of HDM to sensitise and challenge, say between 7 -20% in WT mice (see Author response image 2, lung stimulated with PMA/ionomycin+Monensin, please note this is for illustration purposes only and it not linked to the current manuscript, its merely to demonstrate a point from other experiments we have conducted in the lab).

Author response image 2.

In d-f, the authors perform a serum transfer, but they only do this once. The half life of IgM is quite short. The authors should perform multiple naïve serum transfers to see if this is enough to induce FULL AHR.

We thank the reviewer for this comment. We apologise if this was not clear enough on the Figure legend and method, we did transfer serum 3x, a day before sensitisation, on the day of sensitisation and a day before the challenge to circumvent the short life of IgM. In our subsequent experiments, we have now used busulfan to deplete all bone marrow in IgM-deficient mice and replace it with WT bone marrow and this method restores AHR (Figure 3).

This now appears in line 165 to 169 and reads

“Adoptive transfer of naïve serum

Naïve wild-type mice were euthanised and blood was collected via cardiac puncture before being spun down (5500rpm, 10min, RT) to collect serum. Serum (200mL) was injected intraperitoneally into IgM-deficient mice. Serum was injected intraperitoneally at day -1, 0, and a day before the challenge with HDM (day 10).”

The presence of negative values of total IgE in panel F would indicate some errors in calculation of serum IgE concentrations.

We thank the reviewer for this observation. For better clarity, we have now indicated these values as undetected in Figure , as they were below our detection limit.

Overall, it is hard to be convinced that IgM-deficiency does not lead to a reduction in Th2 inflammation, since the assays appear suboptimal.

We disagree with the reviewer in this instance, because we have shown in 3 different models and in 2 different strains and 2 doses of HDM (high and low) that no matter what you do, Th2 remains intact. Our reason for choosing low dose HDM was based on our previous work and that of others, which showed that depending on antigen load, B cells can either be redundant or have functional roles. Since our interest was to tease out the role of B cells and specifically IgM, it was important that we look at a scenario where B cells are known to have a function (low antigen load). We did find similar findings at high dose of HDM load, but effects on AHR were not as strong, but Th2 was not changed, in fact in some instances Th2 was higher in IgM-deficient mice.

Fig. 3. Gene expression differences between WT and KO mice in PBS and HDM challenged settings are shown. PCA analysis does not show clear differences between all four groups, but genes are certainly up and downregulated, in particular when comparing PBS to HDM challenged mice. In both PBS and HDM challenged settings, three genes stand out as being upregulated in WT v KO mice. these are Baiap2l1, erdr1 and Chil1.

Noted

Fig. 4. The authors attempt to quantify BAIAP2L1 in mouse lungs. It is difficult to know if the antibody used really detects the correct protein. A BAIAP2L1-KO is not used as a control for staining, and I am not sure if competitive assays for BAIAP2L1 can be set up. The flow data is not convincing. The immunohistochemistry shows BAIAP2L1 (in red) in many, many cells, essentially throughout the section. There is also no discernible difference between WT and KO mice, which one might have expected based on the RNA-Seq data. So, from my perspective, it is hard to say if/where this protein is located, and whether there truly exists a difference in expression between wt and ko mice.

We thank the reviewer for this comment. We are certain that the antibody does detect BAIAP2L1, we have used it in 3 assays, which we admit may show varying specificities since it’s a Polyclonal antibody. However, in our western blot, the antibody detects 1 band at 56.7kDa and no other bands, apart from what we think are isoforms. We agree that BAIAP2L1 is expressed by many cell types, including CD45+ cells and alpha smooth muscle negative cells and we show this in our supplementary Figure 9. Where we think there is a difference in expression between WT and IgM-deficient mice is in alpha-smooth muscle-positive cells. We have tested antibodies from different companies, and we find similar findings. We do not have access to BAIAP2L1 KO mice and to test specificity, we have also used single stain controls with or without secondary antibody and isotype control which show no binding in western blot and Immunofluorescence assays and Fluorescence minus one antibody in Flow cytometry, so that way we are convinced that the signal we are seeing is specific to BAIAP2L1.

Fig. 5 and 6. The authors use a single cell contractility assay to measure whether BAIAP2L1 and ERDR1 impact on bronchial smooth muscle cell contractility. I am not familiar with the assay, but it looks like an interesting way of analysing contractility at the single cell level.

The authors state that targeting these two genes with Cas9gRNA reduces smooth muscle cell contractility, and the data presented for contractility supports this observation. However, the efficiency of Cas9-mediated deletion is very unclear. The authors present a PCR in supp fig 9c as evidence of gene deletion, but it is entirely unclear with what efficiency the gene has been deleted. One should use sequencing to confirm deletion. Moreover, if the antibody was truly working, one should be able to use the antibody used in Fig 4 to detect BAIAP2L1 levels in these cells. The authors do not appear to have tried this.

We thank the reviewer for these observations. We are in a process to optimise this using new polyclonal BAIAP2L1 antibodies from other companies, since the one we have tried doesn’t seem to work well on human cells via western blot. So hopefully in our new version, we will be able to demonstrate this by immunofluorescence or western blot.

Other impressions:

The paper is lacking a link between the deficiency of IgM and the effects on smooth muscle cell contraction.

The levels of IL-13 and TNF in lavage of WT and IGMKO mice could be analysed.

We have measured Th2 cytokine IL-13 in BAL fluid and found no differences between IgM-deficient mice and WT mice challenged with HDM (Author response image 1). We could not detected TNF-alpha in the BAL fluid, it was below detection limit.

Author response image 3.

IL-13 levels are not changed in IgM-deficient mice in the lung. Bronchoalveolar lavage fluid in WT or IgM-deficient mice sensitised and challenged with HDM. TNF-a levels were below the detection limit.

Moreover, what is the impact of IgM itself on smooth muscle cells? In the Fig. 7 schematic, are the authors proposing a direct role for IgM on smooth muscle cells? Does IgM in cell culture media induce contraction of SMC? This could be tested and would be interesting, to my mind.

We thank the Reviewer for these comments. We are still trying to test this, unfortunately, we have experienced delays in getting reagents such as human IgM to South Africa. We hope that we will be able to add this in our subsequent versions of the article. We agree it is an interesting experiment to do even if not for this manuscript but for our general understanding of this interaction at least in an in vitro system.

Reviewer #3 (Public Review):

Summary:

This paper by Sabelo et al. describes a new pathway by which lack of IgM in the mouse lowers bronchial hyperresponsiveness (BHR) in response to metacholine in several mouse models of allergic airway inflammation in Balb/c mice and C57/Bl6 mice. Strikingly, loss of IgM does not lead to less eosinophilic airway inflammation, Th2 cytokine production or mucus metaplasia, but to a selective loss of BHR. This occurs irrespective of the dose of allergen used. This was important to address since several prior models of HDM allergy have shown that the contribution of B cells to airway inflammation and BHR is dose dependent.

After a description of the phenotype, the authors try to elucidate the mechanisms. There is no loss of B cells in these mice. However, there is a lack of class switching to IgE and IgG1, with a concomitant increase in IgD. Restoring immunoglobulins with transfer of naïve serum in IgM deficient mice leads to restoration of allergen-specific IgE and IgG1 responses, which is not really explained in the paper how this might work. There is also no restoration of IgM responses, and concomitantly, the phenotype of reduced BHR still holds when serum is given, leading authors to conclude that the mechanism is IgE and IgG1 independent. Wild type B cell transfer also does not restore IgM responses, due to lack of engraftment of the B cells. Next authors do whole lung RNA sequencing and pinpoint reduced BAIAP2L1 mRNA as the culprit of the phenotype of IgM^-/- mice. However, this cannot be validated fully on protein levels and immunohistology since differences between WT and IgM KO are not statistically significant, and B cell and IgM restoration are impossible. The histology and flow cytometry seems to suggest that expression is mainly found in alpha smooth muscle positive cells, which could still be smooth muscle cells or myofibroblasts. Next therefore, the authors move to CRISPR knock down of BAIAP2L1 in a human smooth muscle cell line, and show that loss leads to less contraction of these cells in vitro in a microscopic FLECS assay, in which smooth muscle cells bind to elastomeric contractible surfaces.

Strengths:

(1) There is a strong reduction in BHR in IgM-deficient mice, without alterations in B cell number, disconnected from effects on eosinophilia or Th2 cytokine production

(2) BAIAP2L1 has never been linked to asthma in mice or humans

Weaknesses:

(1) While the observations of reduced BHR in IgM deficient mice are strong, there is insufficient mechanistic underpinning on how loss of IgM could lead to reduced expression of BAIAP2L1. Since it is impossible to restore IgM levels by either serum or B cell transfer and since protein levels of BAIAP2L1 are not significantly reduced, there is a lack of a causal relationship that this is the explanation for the lack of BHR in IgM-deficient mice. The reader is unclear if there is a fundamental (maybe developmental) difference in non-hematopoietic cells in these IgM-deficient mice (which might have accumulated another genetic mutation over the years). In this regard, it would be important to know if littermates were newly generated, or historically bred along with the KO line.

We thank the reviewer for asking this question and getting us to think of this in a different way. This prompted us to use a different method to try and restore IgM function and since our animal facility no longer allows irradiation, we opted for busulfan. We present this data as new data in Figure 3. We had to go back and breed this strain and then generated bone marrow chimeras. What we have shown now with chimeras is that if we can deplete bone marrow from IgM-deficient mice and replace it with congenic WT bone marrow when we allow these mice to rest for 2 months before challenge with HDM (new Supplementary Figure 6 a-c) We also show that AHR (resistance and elastance) is partially restored in this way (Figure 3 a and b) as mice that receive congenic WT bone marrow after chemical irradiation can mount AHR and those that receive IgM-deficient bone marrow, can’t mount AHR upon challenge with HDM. If the mice had accumulated an unknown genetic mutation in non-hematopoietic cells, the transfer of WT bone marrow would not make a difference. So, we don’t believe the colony could have gained a mutation that we are unaware of. We have also shipped these mice to other groups and in their hands, this strains still only behaves as an IgM only knockout mice. See their publication below.

Mark Noviski, James L Mueller, Anne Satterthwaite, Lee Ann Garrett-Sinha, Frank Brombacher, Julie Zikherman 2018. IgM and IgD B cell receptors differentially respond to endogenous antigens and control B cell fate. eLife 2018;7:e35074. DOI: https://doi.org/10.7554/eLife.35074 we have also added methods for bone marrow chimaeras and added results sections and new Figures related to this methods.

Methods (line 171-182).

“Busulfan Bone marrow chimeras

WT (CD45.2) and IgM^-/- (CD45.2) congenic mice were treated with 25 mg/kg busulfan (Sigma-Aldrich, Aston Manor, South Africa) per day for 3 consecutive days (75 mg/kg in total) dissolved in 10% DMSO and Phosphate buffered saline (0.2mL, intraperitoneally) to ablate bone marrow cells. Twenty-four hours after last administration of busulfan, mice were injected intravenously with fresh bone marrow (10x10⁶ cells, 100mL) isolated from hind leg femurs of either WT (CD45.1) or IgM^-/- mice(33). Animals were then allowed to complement their haematopoietic cells for 8 weeks. In some experiments the level of bone marrow ablation was assessed 4 days post-busulfan treatment in mice that did not receive donor cells. At the end of experiment level of complemented cells were also assessed in WT and IgM^-/- mice that received WT (CD45.1) bone marrow.”

Results (line 491-521)

“Replacement of IgM-deficient mice with functional hematopoietic cells in busulfan mice chimeric mice restores airway hyperresponsiveness.

We then generated bone marrow chimeras by chemical radiation using busulfan(33). We treated mice three times with busulfan for 3 consecutive days and after 24 hrs transferred naïve bone marrow from congenic CD45.1 WT mice or CD45.2 IgM^-/- mice (Fig. 3a and Supplementary Fig. 5a). We showed that recipient mice that did not receive donor bone marrow after 4 days post-treatment have significantly reduced lineage markers (CD45+Sca-1+) or lineage negative (Lin-) cells in the bone marrow when compared to untreated or vehicle (10% DMSO) treated mice (Supplementary Figure 5b-c). We allowed mice to reconstitute bone marrow for 8 weeks before sensitisation and challenge with low dose HDM (Figure 3a). We showed that WT (CD45.2) recipient mice that received WT (CD45.1) donor bone marrow had higher airway resistance and elastance and this was comparable to IgM^-/- (CD45.2) recipient mice that received donor WT (CD45.1) bone marrow (Figure 3b). As expected, IgM^-/- (CD45.2) recipient mice that received donor IgM^-/- (CD45.2) bone marrow had significantly lower AHR compared to WT (CD45.2) or IgM^-/- (CD45.2) recipient mice that received WT (CD45.1) bone marrow (Figure 3b). We confirmed that the differences observed were not due to differences in bone marrow reconstitution as we saw similar frequencies of CD45.1 cells within the lymphocyte populations in the lungs and other tissues (Supplementary Fig. 5d). We observed no significant changes in the lung neutrophils, eosinophils, inflammatory macrophages, CD4 T cells or B cells in WT or IgM^-/- (CD45.2) recipient mice that received donor WT (CD45.1/CD45.2) or IgM^-/- (CD45.2) bone marrow when sensitised and challenged with low dose HDM (Fig. 3c)

Restoring IgM function through adoptive reconstitution with congenic CD45.1 bone marrow in non-chemically irradiated recipient mice or sorted B cells into IgM^-/- mice (Supplementary Fig.  6a) did not replenish IgM B cells to levels observed in WT mice and as a result did not restore AHR, total IgE and IgM in these mice (Supplementary Fig.  6b-c).”

The 2 new figures are

Figure 3 which moved the rest of the Figures down and Supplementary Figure 5, which also moved the rest of the supplementary figures down.

Discussion appears in line 757-766 of the untracked version of the article.

To resolve other endogenous factors that could have potentially influenced reduced AHR in IgM-deficient mice, we resorted to busulfan chemical irradiation to deplete bone marrow cells in IgM-deficient mice and replace bone marrow with WT bone marrow. While it is well accepted that busulfan chemical irradiation partially depletes bone marrow cells, in our case it was not possible to pursue other irradiation methods due to changes in ethical regulations and that fact that mice are slow to recover after gamma rays irradiation. Busulfan chemical irradiation allowed us to show that we could mostly restore AHR in IgM-deficient recipient mice that received donor WT bone marrow when challenged with low dose HDM.

(2) There is no mention of the potential role of complement in activation of AHR, which might be altered in IgM-deficient mice 

We thank the reviewer for this comment. We have not directly looked at complement in this instance, however, from our previous work on C3-/- mice, there have been comparable AHR to WT mice under the HDM challenge.

(3) What is the contribution of elevated IgD in the phenotype of the IgM-deficient mice. It has been described by this group that IgD levels are clearly elevated

We thank the reviewer for this question. We believe that IgD is essentially what drives partial class switching to IgG, we certainly have shown that in the case of VSV virus and Trypanosoma congolense and Trypanosoma brucei brucei that elevated IgD drive delayed but effective IgG in the absence of IgM (Lutz et al, 2001, Nature). This is also confirmed by Noviski studies where they show that both IgM and IgD do share some endogenous antigens, so its likely that external antigens can activate IgD in a similar manner to prompt class switching.

(4) How can transfer of naïve serum in class switching deficient IgM KO mice lead to restoration of allergen specific IgE and IgG1?

We thank the Reviewer for these comments, we believe that naïve sera transferred to IgM deficient mice is able to bind to the surface of B cells via IgM receptors (FcμR / Fcα/μR), which are still present on B cells and this is sufficient to facilitate class switching. Our IgM^-/- mouse lacks both membrane-bound and secreted IgM, and transferred serum contains at least secreted IgM which can bind to surfaces via its Fc portion. We measured HDM-specific IgE and we found very low levels, but these were not different between WT and IgM^-/- adoptively transferred with WT serum. We also detected HDM-specific IgG1 in IgM^-/- transferred with WT sera to the same level as WT, confirming a possible class switching, of course, we can’t rule out that transferred sera also contains some IgG1. We also can’t rule out that elevated IgD levels can partially be responsible for class switched IgG1 as discussed above.

In the discussion line 804-812, we also added the following

“We speculate that IgM can directly activate smooth muscle cells by binding a number of its surface receptors including FcμR, Fcα/μR and pIgR(52-54). IgM binds to FcμR strictly, but shares Fcα/μR and pIgR with IgA(5,52,54). Both Fcα/μR and pIgR can be expressed by non-structural cells at mucosal sites(54,55). We would not rule out that the mechanisms of muscle contraction might be through one of these IgM receptors, especially the ones expressed on smooth muscle cells(54,55). Certainly, our future studies will be directed towards characterizing the mechanism by which IgM potentially activates the smooth muscle.”

We have discussed this section under Discussion section, line 731 to 757. In addition, since we have now performed bone marrow chimaeras we have further added the following in our discussion in line 757-766.

To resolve other endogenous factors that could have potentially influenced reduced AHR in IgM-deficient mice, we resorted to busulfan chemical irradiation to deplete bone marrow cells in IgM-deficient mice and replace bone marrow with WT bone marrow. While it is well accepted that busulfan chemical irradiation partially depletes bone marrow cells, in our case it was not possible to pursue other irradiation methods due to changes in ethical regulations and that fact that mice are slow to recover after gamma rays irradiation. Busulfan chemical irradiation allowed us to show that we could mostly restore AHR in IgM-deficient recipient mice that received donor WT bone marrow when challenged with low dose HDM.

We removed the following lines, after performing bone marrow chimaeras since this changed some aspects.

Our efforts to adoptively transfer wild-type bone marrow or sorted B cells into IgM-deficient mice were also largely unsuccessful partly due to poor engraftment of wild-type B cells into secondary lymphoid tissues. Natural secreted IgM is mainly produced by B1 cells in the peritoneal cavity, and it is likely that any transfer of B cells via bone marrow transfer would not be sufficient to restore soluble levels of IgM(3,10).

(5) Alpha smooth muscle antigen is also expressed by myofibroblasts. This is insufficiently worked out. The histology mentions "expression in cells in close contact with smooth muscle". This needs more detail since it is a very vague term. Is it in smooth muscle or in myofibroblasts.

Response: We appreciate that alpha-smooth muscle actin-positive cells are a small fraction in the lung and even within CD45 negative cells, but their contribution to airway hyperresponsiveness is major. We also concede that by immunofluorescence BAIAP2L1 seems to be expressed by cells adjacent to alpha-smooth muscle actin (Fig. 5b), however, we know that cells close to smooth muscle (such as extracellular matrix and myofibroblasts) contribute to its hypertrophy in allergic asthma.

James AL, Elliot JG, Jones RL, Carroll ML, Mauad T, Bai TR, et al. Airway Smooth Muscle Hypertrophy and Hyperplasia in Asthma. Am J Respir Crit Care Med [Internet]. 2012;185:1058–64. Available from: https://doi.org/10.1164/rccm.201110-1849OC

(6) Have polymorphisms in BAIAP2L1 ever been linked to human asthma?

No, we have looked in asthma GWAS studies, at least summary statics and we have not seen any SNPs can could be associated with human asthma.

(7) IgM deficient patients are at increased risk for asthma. This paper suggests the opposite. So the translational potential is unclear

We thank the reviewer for these comments. At the time of this publication, we have not made a concrete link with human disease. While there is some anecdotal evidence of diseases such as Autoimmune glomerulonephritis, Hashimoto’s thyroiditis, Bronchial polyp, SLE, Celiac disease and other diseases in people with low IgM. Allergic disorders are also common in people with IgM deficiency as the reviewer correctly points out, other studies have reported as high as 33-47%. The mechanisms for the high incidence of allergic diseases are unclear as generally, these patients have normal or higher IgG and IgE levels. IgM deficiency may represent a heterogeneous spectrum of genetic defects, which might explain the heterogeneous nature of disease presentations.

eLife Assessment

This important study describes how a single effector of the Type Six Secretion System (T6SS) has two distinct functions, which may contribute to bacterial survival and the development of novel antibacterials. The authors utilized various methods in biochemistry, microbiology, and microscopy to produce convincing data supporting their claims about the protein's function; however, they could clarify the implications for non-experts to enhance the accessibility of this work. This manuscript is of interest to those studying T6SS, particularly those interested in effectors and bacterial enzymes.

Reviewer #1 (Public review):

Summary:

The manuscript performs a comprehensive biochemical, structural, and bioinformatic analysis of TseP, a type 6 secretion system effector from Aeromonas dhakensis that includes identification of a domain required for secretion and residues conferring target organism specificity. Through targeted mutations, they have expanded the target range of a T6SS effector to include a gram-positive species, which are not typically susceptible to T6SS attack. Although this is not the first dual domain effector to be described, this is the first time anyone has been able to modify a T6SS effector to have an expanded target species range.

Strengths:

The thorough dissection of TseP activity and modulation of target specificity represent a novel contribution to the field of antibacterial research.

Weaknesses:

Although the mechanistic activity of TseP is fully dissected here, there are some unaddressed questions regarding the importance/evolution of the dual activity domain organization. For example, does the modified Gram-positive targeting TseP effector still kill Gram-negative bacteria in bacterial mixtures? And if so, what is the evolutionary benefit of having a TseP that cannot target Gram-positives? And can something be inferred about the biology of Aeromonas from this?

Comments on revisions:

The comments and critiques from the initial submission have been addressed. However, some of them have only been addressed in the author's rebuttal. Some of the discussion particularly regarding the validity of using E. coli PG, the ability for TseP_C4+ to still kill E. coli, and the advantages of having dual domain function effectors probably should be present in the actual manuscript.

Reviewer #2 (Public review):

Summary:

Wang et al. investigate the role of TseP, a Type VI secretion system (T6SS) effector molecule, revealing its dual enzymatic activities as both an amidase and a lysozyme. This discovery significantly enhances the understanding of T6SS effectors, which are known for their roles in interbacterial competition and survival in polymicrobial environments. TseP's dual function is proposed to play a crucial role in bacterial survival strategies, particularly in hostile environments where competition between bacterial species is prevalent.

Strengths:

(1) The dual enzymatic function of TseP is a significant contribution, expanding the understanding of T6SS effectors.<br /> (2) The study provides important insights into bacterial survival strategies, particularly in interbacterial competition.<br /> (3) The findings have implications for antimicrobial research and understanding bacterial interactions in complex environments.

Weaknesses:

(1) The manuscript assumes familiarity with previous work, making it difficult to follow. Mutants and strains need clearer definition and references.<br /> (2) Figures lack proper controls, quantification, and clarity in some areas, notably in Figures 1A and 1C.<br /> (3) The Materials and Methods section is poorly organized, hindering reproducibility. Biophysical validation of Zn²⁺ interaction and structural integrity of proteins need to be addressed.<br /> (4) Discrepancies in protein degradation patterns and activities across different figures raise concerns about data reliability.

Comments on revisions:

The authors have addressed most of the comments, significantly improving the manuscript. They provided clear details of mutant constructs and strains, including additional references and a revised strain. Individual data points and statistical analyses were added to key figures, ensuring transparency and reproducibility. Supplemental data, such as protein purification details and loading controls, were included to address concerns about experimental reliability. However, the authors did not perform new experiments, such as isothermal titration calorimetry (ITC) to demonstrate the interaction between Zn²⁺ and TsePN or stop-flow spectroscopy to examine enzymatic kinetics, which could have further strengthened the manuscript. I trust these aspects will be addressed in future studies.

The revised Materials and Methods section was significantly improved, providing detailed protocols for bioinformatics analyses, microscopic imaging, and enzymatic assays.

These revisions provide a clearer and more robust presentation of TseP's dual enzymatic functions and their implications in bacterial competition. The manuscript now represents a significant contribution to understanding T6SS effectors, and I recommend it for publication in its current form.

Reviewer #3 (Public review):

Summary:

Type VI secretion systems (T6SS) are employed by bacteria to inject competitor cells with numerous effector proteins. These effectors can kill injected cells via an array of enzymatic activities. A common class of T6SS effector are peptidoglycan (PG) lysing enzymes. In this manuscript, the authors characterize a PG-lysing effector-TseP-from the pathogen Aeromonas dhakensis. While the C-terminal domain of TseP was known to have lysozyme activity, the N-terminal domain was uncharacterized. Here, the authors functionally characterize TsePN as a zinc-dependent amidase. This discovery is somewhat novel because it is rare for PG-lysing effectors to have amidase and lysozyme activity. In the second half of the manuscript, the authors utilize a crystal structure of the lysozyme TsePC domain to inform the engineering of this domain to lyse gram-positive peptidoglycan.

Strengths:

The two halves of the manuscript considered together provide a nice characterization of a unique T6SS effector and reveal potentially general principles for lysozyme engineering.

Weaknesses:

The advantage of fusing amidase and lysozyme domains in a single effector is not discussed but would appear to be a pertinent question.

Comments on revisions:

The authors have adequately addressed my previous comments. The authors did not conduct any additional experiments to address the comments made by other reviewers. However, in most cases it seems that paring down the strength of claims made in the text or adding data to the supplement is sufficient to address these concerns.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The manuscript performs a comprehensive biochemical, structural, and bioinformatic analysis of TseP, a type 6 secretion system effector from Aeromonas dhakensis that includes the identification of a domain required for secretion and residues conferring target organism specificity. Through targeted mutations, they have expanded the target range of a T6SS effector to include a gram-positive species, which is not typically susceptible to T6SS attack.

Strengths:

All of the experiments presented in the study are well-motivated and the conclusions are generally sound.

Thank you.

Weaknesses:

There are some issues with the clarity of figures. For example, the microscopy figures could have been more clearly presented as cell counts/quantification rather than representative images. Similarly, loading controls for the secreted proteins for the westerns probably should be shown.

Also, some of the minor/secondary conclusions reached regarding the "independence" of the N and C term domains of the TseP are a bit overreaching.

We thank the reviewer for pointing out the issues and have carefully revised the manuscript accordingly. We acknowledge the reviewer’s concern regarding the independence of the N- and C-terminal domains, and have toned down the relevant claims.

Reviewer #2 (Public review):

Summary:

Wang et al. investigate the role of TseP, a Type VI secretion system (T6SS) effector molecule, revealing its dual enzymatic activities as both an amidase and a lysozyme. This discovery significantly enhances the understanding of T6SS effectors, which are known for their roles in interbacterial competition and survival in polymicrobial environments. TseP's dual function is proposed to play a crucial role in bacterial survival strategies, particularly in hostile environments where competition between bacterial species is prevalent.

Strengths:

(1) The dual enzymatic function of TseP is a significant contribution, expanding the understanding of T6SS effectors.

(2) The study provides important insights into bacterial survival strategies, particularly in interbacterial competition.

(3) The findings have implications for antimicrobial research and understanding bacterial interactions in complex environments.

Thank you.

Weaknesses:

(1) The manuscript assumes familiarity with previous work, making it difficult to follow. Mutants and strains need clearer definitions and references.

Thank you for raising the issue. We have revised the manuscript accordingly to improve the clarity by including more detailed descriptions of the mutants and strains, along with references to prior work where relevant, to improve clarity.

(2) Figures lack proper controls, quantification, and clarity in some areas, notably in Figures 1A and 1C.

We have now added the controls as requested by reviewers.

(3) The Materials and Methods section is poorly organized, hindering reproducibility. Biophysical validation of Zn²⁺ interaction and structural integrity of proteins need to be addressed.

We have now included more details in the Materials and Methods section. While we recognize the importance of biophysical validation of the Zn²⁺ interaction, this analysis lies beyond the primary scope of the current study. We plan to investigate the role of Zn²⁺ interaction and the EF-hand domain in greater depth as part of our follow-up studies. Thank you for this suggestion.

(4) Discrepancies in protein degradation patterns and activities across different figures raise concerns about data reliability.

We acknowledge the concern about discrepancies in protein degradation patterns. TseP exhibits inherent instability, which might explain the observed variations. We have added an explanation in the detailed response letter and the manuscript.

Reviewer #3 (Public review):

Summary:

Type VI secretion systems (T6SS) are employed by bacteria to inject competitor cells with numerous effector proteins. These effectors can kill injected cells via an array of enzymatic activities. A common class of T6SS effector are peptidoglycan (PG) lysing enzymes. In this manuscript, the authors characterize a PG-lysing effector-TseP-from the pathogen Aeromonas dhakensis. While the C-terminal domain of TseP was known to have lysozyme activity, the N-terminal domain was uncharacterized. Here, the authors functionally characterize TsePN as a zinc-dependent amidase. This discovery is somewhat novel because it is rare for PG-lysing effectors to have amidase and lysozyme activity.

In the second half of the manuscript, the authors utilize a crystal structure of the lysozyme TsePC domain to inform the engineering of this domain to lyse gram-positive peptidoglycan.

Strengths:

The two halves of the manuscript considered together provide a nice characterization of a unique T6SS effector and reveal potentially general principles for lysozyme engineering.

Thank you.

Weaknesses:

The advantage of fusing amidase and lysozyme domains in a single effector is not discussed but would appear to be a pertinent question. Labeling of the figures could be improved to help readers understand the data.

Thank you for the suggestions. We have revised the manuscript and figures to improve clarity.

The advantage of having dual-domain functions relative to having just one of the two functions is likely for increasing competitive fitness. Although such dual functional cell-wall targeting effectors have not been characterized prior to this study, there are some examples that dual functions are encoded by the same secretion module, for example the VgrG1-TseL pair in Vibrio cholerae. The C-terminal of VgrG1 not only catalyzes actin crosslinking but also recognizes and delivers the downstream encoded lipase effector TseL through direct interaction. In this context, the VgrG1-TseL pair also represent a dual-functional module. Therefore, it is likely that fusing effector domains and coupling effector functions are parallel strategies for the evolution of T6SS effectors.

eLife Assessment

This manuscript reports an important new statistical method for calculating the significance of correlations between two time-series, which provides more accuracy than other methods when the data has few replicates. The proposed method solves a real-life problem that is frequently encountered and is broadly applicable to many realistic datasets in many experimental contexts. The technique is supported with compelling mathematical derivations as well as analysis of both computer-generated and previously published experimental data.

Reviewer #1 (Public review):

Summary:

The manuscript puts forward a statistical method to more accurately report the significance of correlations within data. The motivation for this study is two-fold. First, the publication of biological studies demands the report of p-values, and it is widely accepted that p-values below the arbitrary threshold of 0.05 give the authors of such studies justification to draw conclusions about their data. Second, many biological studies are limited by the number of replicate samples that are feasible, with replicates of less than 5 typical. The authors report a statistical tool that uses a permute-match approach to calculate p-values. Notably, the proposed method reduces p-values from around 0.2 to 0.04 as compared to a standard permutation test with a small sample size. The approach is clearly explained, including detailed mathematical explanations and derivations. The advantage of the approach is also demonstrated through analysis of computer-generated synthetic data with specified correlation and analysis of previously published data related to fish schooling. The authors make a clear case that this method is an improvement over the more standard approach currently used, and also demonstrate the impact of this methodology on the ability to obtain p-values that are the standard for biological research. Overall, this paper is very strong. While the subject matter seems somewhat specialized, I would make the case that this will be an important study that has broad general interest to readers. The findings are very general and applicable to many research contexts. Experimentalists also want to report accurate p-values in their work and better understand how these values are calculated. Although I believe the previous statement is true, I am not sure that many research groups doing biological work are reading specialized statistics journals regularly. Therefore a useful and broadly applicable statistical tool is well placed in this journal.<br /> Strengths:

The proposed method is broadly applicable to many realistic datasets in many experimental contexts.

The power of this method was demonstrated with both real experimental data and "synthetic" data. The advantages of the tool are clearly reported. The zebrafish data is a great example dataset.

The method solves a real-life problem that is frequently encountered by many experimental groups in the biological sciences.

The writing of the paper is surprisingly clear, given the technical nature of the subject matter. I would not at all consider myself a statistician or mathematician, but I found the text easy to follow. The authors did an impressive job guiding the reader through material that would often be difficult to grasp. The introduction was also well-written and clearly motivated the goals of the study.

Weaknesses:

A few changes could be made if the manuscript is revised. I would consider all of these points minor, but the paper could be improved if these points were addressed.

(1) The caption of Figure 2 doesn't seem to mention panel D. Figure A-2 also does not mention C in the caption.

(2) Figure 2D is a little hard to follow. First, the definition of "Power" is not clear, and I couldn't find the precise definition in the text. Second, the legend for the different lines in 2D is only given in Figure A-2. Perhaps a portion of the caption for Figure 2 is missing?

(3) The concept of circular variance for the fish data was heard to understand/visualize. The equation on line 326 did not help much. If there is a very simple picture that could be added near line 326 that helps to explain Ct and theta, that could be a big help for some readers who do not work on related systems. The analysis performed is understandable, the reader just has to accept that circular variance captions the degree of alignment of the fish.

(4) For the data discussed in Figure 3, I wasn't 100% sure how the time windows were selected. In the caption, it says "time series to different lengths starting from the first frame". So the 20 s time window was from t=0 to t= 20 s. Would a different result be obtained if a different 20 s window was chosen (from t = 4 min to t = 4 min 20 s just to give a specific example). I suppose by chance one of the time windows would give a p-value less than the target 0.05, that wouldn't be surprising. Maybe a random time window should be selected (although I am not indicating what was reported was incorrect)? A little more discussion on this aspect of the study may be helpful.