Author Response

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

eLife assessment

This is a valuable study of Eph-Ephrin signaling mechanisms generating pathological changes in amyotropic lateral sclerosis. There are exciting findings bearing on the role of glial cells in this pathology. The study emerges with solid evidence for a novel astrocyte-mediated mechanism for disease propagation. It may help identify potential therapeutic targets.

Response to Editor’s decision letter: Drs. Huang and Zaidi: Thank you for considering this re-revision of our manuscript for potential publication in eLife. We have addressed the remaining comments of reviewer #2. We have included detailed response-to-reviewer comments below to address each of these remaining specific points from reviewer #2, and we have highlighted all the changes in the manuscript text (using a red font color) made in response to these comments. Based on the reviewers’ critiques, we feel our re-working of the manuscript has made for a greatly improved study.

Reviewer #1 (Recommendations For The Authors):

Reviewer comment: All questions/concerns have been addressed.

Response: We thank Reviewer #1 for the previous helpful comments that we used to improve our manuscript. As Reviewer #1 has no new comments, we have provided no additional responses to address this reviewer’s input. Instead, we only focus (in this new “Response to Reviewer Comments” document) on the remaining points from Reviewer #2 below.

Reviewer #2 (Recommendations For The Authors):

Overall, the authors have addressed most concerns raised in the prior review. A couple of very minor points remain, which would improve the clarity of the report.

Reviewer comment 1: The abstract has not been edited and still emphasizes that astrocyte-mediated upregulation in ephrinB2 signaling underlies pathogenicity in mutant SOD1-associated ALS. There is certainly sufficient evidence to suggest a large role for astrocytes, however, without a thorough investigation of other key cell types in the spinal cord, this cannot be concluded specifically. Especially given that a non-specific promoter (U6) was employed in the viral constructs.

Response: We apoplogize for this mistake. In response to the reviewer’s previous comment in the first round of review, we made changes throughout the manuscript to address this issue; however, we failed to do this in the Abstract. In this re-revised manucript, we now also make the necessary changes to the Abstract.

Reviewer comment 2: It is interesting to note that a non-specific promoter, U6, exhibited such large specificity to astrocytes in the cord as compared to neurons (Fig 2M). This is worth discussing briefly in the discussion and how this result compares to those in the literature.

Response: We have now added a brief discussion of this issue to the Discussion section, including describing our previous studies that used the Gfa2 promotor to achieve astrocyte-specific transduction when employing viral vectors in the rodent spinal cord.

Reviewer comment 3: I appreciate the authors including a supplemental figure on the expression of ephrinA4 receptors in the cervical ventral horn. Unfortunately, the quality of this image is very poor in conveying the receptor expression. The detailed discussion point on the expression of EphB receptors in the cervical ventral horn should be sufficient for readers to take into consideration.

Response: We have now removed this supplemental figure and keep only the text from the rerevised manuscript.

Reviewer comment 4: A few instances of motor neuron diameter being attributed to a 200μm2 size remain (e.g. pg 14).

Response: We have corrected this issue throughout the re-revised manuscript. The correct information is: somal diameter greater than 20 μm.

Reviewer comment 5: It is still a little unclear in the result text as to when assessment of lentiviral transduction was conducted following intraspinal injections.

Response: We have now added this detail about the time point of assessing transduction to both the Results section and the Materials/Methods section.

Reviewer comment 6: Some figures are missing markers of significance (e.g. Fig 2M).

Response: Below are our comments about significance markers for each graph in all figures.

Figure 1:

Panel E: We have now added asterisks for any statistically-significant comparisons. In addition, we provide the details of this statistical analysis in the text of the re-revised manuscript.

Figure 2:

Panel M: We have now added asterisks for statistical comparisons, as well as details in the text.

Panel N: The asterisk was already shown in the previous version of the figure.

Figure 3:

Panels B and G: The asterisks were already shown in the previous version of the figure.

Figure 4:

All panels: There are no significant differences; therefore, no asterisks are needed.

Figure 5:

Panel F and G: The asterisks were already shown in the previous version of the figure.

Panel H: The difference is not statistically-signficant.

Figure 6: No graphs are shown in this figure.

Reviewer comment 7: Since a wild type mouse control has not been included in the quantification of diaphragm NMJ innervation with and without ephrin knock-down, it would be useful to include a description or discussion on the phenotype of NMJ denervation exhibited in the SOD1G93A mouse model of ALS.

Response: We have now added description of diaphragm NMJ denervation that occurs in SOD1G93A mice, in particular at the age/time point of our NMJ analysis.

Author Response

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

eLife Assessment

This valuable manuscript investigates the roles of DKK3 in AD synapse integrity. Although previous work has identified the involvement of Wnt and DKK1 in synaptic physiology, this study provides compelling evidence that suppression of DKK3 rescues the changes in excitatory synapse numbers, as well as memory deficits in an established AD model mice. The authors provide both gain and loss of function data that support the main conclusion and advance our understanding of the mechanisms by which Wnt pathway mediates early synaptic dysfunction in AD models.

Public Reviews:

Reviewer #1 (Public Review):

In this study, Nuria Martin-Flores, Marina Podpolny and colleagues investigate the role of Dickkopf-3 (DKK3), a Wnt antagonist in synaptic dysfunction in Alzheimer's disease. Loss of synapses is a feature of Alzheimer's and other forms of dementia such as frontotemporal dementia and linked amyotrophic lateral sclerosis (FTD). The authors utilise a broad range of experimental approaches. They show that DKK3 levels are increased in Alzheimer's disease and that this occurs early in disease. This is an important finding since early disease changes are believed to be the most important. They also show increases in DKK3 in transgenic mouse models of Alzheimer's disease and that DKK3 knockdown restores synapse number and memory in one such model. Finally, they link these DKK3 increases to loss of excitatory synapses via the blockade of the Wnt pathway and subsequent activation of GSK3B; GSK3B is strongly linked to both Alzheimer's disease and FTD. The quality of the data is good and the conclusions well supported by these data. There are no major weaknesses. The findings support studies that target the Wnt pathway as a potential therapeutic for Alzheimer's disease.

Reviewer #2 (Public Review):

This manuscript by Martin-Flores et al., has examined the role of DKK3 in Alzheimer's disease, focusing on the regulation of synaptic numbers. By using human AD brain databases and tissue samples, the authors showed that DKK3 protein and mRNA levels are increased in the brains of AD patients. DKK3 is expressed in the excitatory neurons in WT mouse brains and accumulates at atrophic neurites around amyloid plaques in AD mouse brains. Interestingly, secretion of DKK3 appears to be regulated by NMDAR antagonist as well as chemical LTD. Through gain and loss of function studies, the authors showed that DKK3 regulates the number of excitatory as well as inhibitory synapses with distinct downstream pathways. Finally, the authors investigated the contribution of DKK3 to synaptic changes in AD and found that DKK3 loss of function rescues both the excitatory and inhibitory synaptic defects, resulting in the improvement of memory function in J20 mice.

Overall, the data is clearly presented and deals with novel roles of DKK3 in controlling excitatory and inhibitory synapses. The finding that shRNA expression of DKK3 in AD model mice rescues synaptic phenotypes and memory impairment is potentially interesting and may provide a new strategy for AD treatment.

We would like to thank the Editors and the Reviewers for their very insightful suggestions. We are delighted to receive very positive reviews of our manuscript. In response to the comments made by the reviewers, we have carried out an extensive revision of our manuscript. In the revised manuscript, we have addressed all the comments made by the reviewers.

Recommendations for the authors:

Reviewer #1:

My only comment regards the role of GSK3B activation in synaptic dysfunction and its targets. GSK3B is a Tau kinase but is also involved in IP3 receptor delivery of Ca2+ to mitochondria. This delivery is major regulator of mitochondrial ATP production and synaptic function is heavily dependent on ATP. Both Alzheimer's disease and FTD insults have been linked to GSK3B activation -see for e.g. Szabo EMBO R 2023, Gomez-Suaga Aging Cell 2022. It might be valuable to readers for the authors to speculate briefly on potential GSK3B synaptic targets in the Discussion.

We appreciate the reviewer for this suggestion. In the Discussion, we now included how GSK3β may contribute to synaptic dysfunction and loss in the context of increased DKK3 levels and in Alzheimer’s disease.

Reviewer #2:

1. In Fig 1B, the authors showed that soluble DKK3 levels were increased in Braak 1-3 patients, while no changes were observed in Braak 4-5. If the secretion of DKK3 is dependent on NMDAR activity, does this data imply that Braak 4-5 patients have reduced NMDAR activity in general, resulting in the reduced DKK3 release even with the increased mRNA levels? It would be interesting to test this hypothesis in a mouse AD model.

In Figure 1B, we analyzed the levels of soluble and insoluble DKK3 in the hippocampus of AD patients at different disease stages based on their Braak stages. As the reviewer indicated, soluble levels of DKK3 were increased in patients with Braak I-III but not at later stages. Importantly, DKK3 levels were also elevated in Braak IV-VI patients, but only in the insoluble fraction (Figure 1C), suggesting that DKK3 could accumulate within Aβ aggregates. Based on these findings, we cannot conclude that DKK3 release is reduced at later stages of the disease in patients.

To explore the underlying mechanisms regulating DKK3 levels, we used cultured hippocampal neurons and AD mouse brain slices. In mouse models, we have demonstrated that extracellular DKK3 levels (secreted DKK3 fraction) depends on NMDAR activation early in the disease progression (Figure 2E, F). Moreover, we also provide new data showing that antagonizing NMDAR partially blocks the increase of DKK3 extracellular levels induced by oligomeric Aβ (see response to question 4 of this reviewer and Figure S2G, H). It is well established that oligomeric Aβ promotes hyperexcitability through, in part, the aberrant activation of NMDAR (Li S et al., 2011, PMID: 21543591; Mucke L and Selkoe DJ et al., 2012, PMID: 22762015). In line with this, NMDAR blockers prevent Aβ-induced synapse loss and improve cognition in AD models (Hu NW et al., 2009, PMID: 19918059; Ye C et al., 2004, PMID: 15288443). In addition, an NMDAR antagonist is currently approved as a drug treatment for AD patients (Cumming J 2021, PMID: 33441154). Together, our findings in dissociated neurons, AD mouse brain and human samples indicate that soluble Aβ oligomers promote the release of DKK3 through NMDAR activation and suggest that this mechanism might also be occurring in the brain of AD patients.

1. Recent work (Yuan et al., 2022, Nature) has shown that dystrophic neurites/axonal spheroids found around Aβ deposits are filled with neuronal endolysosomes. Are DKK3 in ThioS positive amyloid plaques located in endolysosomes of these axonal spheroids? If so, does this data mean that DKK3 in Fig 2B-D represents the entrapped DKK3 protein population that fails to be secreted from dystrophic neurites?

The reviewer points an interesting question. Our results show that secretion of DKK3 is increased in two AD models before substantial plaque load. Later in the disease, DKK3 accumulates in dystrophic neurites (visualized as axonal spheroids) surrounding amyloid plaques. To address if DKK3 protein is located in vesicles of the endolysosomal pathway within axonal spheroids, we performed co-localization analyses of DKK3 and the endolysosomal marker LAMP1. We found that DKK3 colocalized with LAMP1 (Figure 2D) indicating the presence of DKK3 in axonal spheroids. These results indeed suggest that DKK3 is present in abnormally enlarged vesicles in dystrophic neurites around Aβ plaques. This could affect the axonal transport of DKK3. Given that proteins present in dystrophic neurites have been correlated with defects in bidirectional transport in the axon (Stokin GB et al., 2005, PMID: 15731448; Sadleir KR et al., 2016, PMID: 26993139), both DKK3 turnover and secretion could be affected.

1. Why does only LTD induce DKK3 release? Why not general activation of neuronal activity? It would be important to test the relationship between DKK3 secretion and neuronal activity with optogenetics and chemogenetics.

We tested whether neuronal activity triggered increased extracellular DKK3 levels by subjecting neurons to chemical long-term potentiation (cLTP) or long-term depression (cLTD). However, only cLTD increased extracellular DKK3, which we then confirmed in brain slices (Figure S3). This finding is not unexpected as it is well described that different patterns of activity can lead to different molecular outcomes. For example, high-frequency stimulation (HFS; an activity pattern that resembles LTP) and low-frequency stimulation (LFS; a different activity pattern resembling LTD) leads to opposing effects on surface levels of the Wnt receptor Frizzled-5 (Fz5) (Sahores M et al., 2010, PMID: 20530549). Furthermore, cLTP increases Fz5 s-acylation, an important post-translational modification that regulates the surface levels of Fz5, whereas cLTD decreases it (Teo S et al., 2023, PMID: 37557176). Another example is the BDNF receptor TrkB. Surface TrkB is increased by tetanic stimulation, which also induces LTP as HFS or cLTP, but not by LFS (Du J et al., 2000, PMID: 10995446). Our findings suggest that DKK3 might contribute to synaptic changes underlying cLTD. Future experiments using chemogenetics or optogenetics might elucidate the role of DKK3 in activity-induced synaptic changes.

1. Are Abeta oligomer treatment-dependent increases in DKK3 protein levels in the cellular lysate and the extracellular fraction also suppressed by APV?

Our results in AD mice indicate that increased DKK3 release is dependent on NMDAR activation. To investigate if amyloid-β oligomers (Aβo) increase DKK3 levels in the cell lysate and extracellular fractions through NMDAR, we blocked these receptors in hippocampal neurons using AP-V (Figure S2G, H). In these experiments, we use a lower concentration of Aβo (200nM of Aβ1-42) to avoid any potential cytotoxic effect. In line with our previous results using a higher concentration of Aβo, we observed that Aβo markedly increased DKK3 levels both in the cell lysate and in the extracellular fraction compared to the reverse Aβ42-1 control peptide. Kruskal-Wallis with Dunn’s test showed a trend to a reduced levels of DKK3 in the extracellular fraction when we compared neurons treated with Aβo and APV with those neurons treated with Aβ and vehicle (p = 0.0726). However, this reduced levels of DKK3 in the extracellular fraction reached statistical significance using a t-test (p = 0.0384). No differences were observed between the reverse control peptide and Aβo and APV conditions. These results suggest that blockade of the NMDAR partially occludes the ability of Aβo to increase DKK3 levels in the extracellular fraction.

1. Why does DKK3 shRNA only downregulate inhibitory synapses but not excitatory synapses in the WT brain slice? Does this mean that in the WT brain, other DKK proteins (without changes in their expression as shown in Fig S6) are sufficiently expressed and compensate for the roles of DKK3 in excitatory synapse integrity?

The reviewer points out an interesting result. In J20 mice, DKK3 knockdown affects both excitatory and inhibitory synapse density (Figure 6B, C). In Figure 3B, D, we show that in vivo downregulation of DKK3 leads to an increased number of inhibitory synapses without affecting excitatory ones in the brain of WT animals. These results indicate that in a healthy brain (WT), DKK3 is required for the maintenance of inhibitory synapses but not for excitatory synapses under our experimental conditions. Furthermore, DKK3 partially shares the mechanism of action with DKK1 as both DKK proteins promote excitatory synapse loss through the Wnt/GSK3β pathway (Figure 4A-C) (Marzo A et al., 2016, PMID: 27593374). Therefore, it is possible that endogenous DKK1 levels in the hippocampus could compensate for the reduced expression of DKK3 resulting in the lack of changes in excitatory synapse number when DKK3 is knockdown in WT animals.

1. Manipulating DKK3 in WT brains only affects Gephyrin but not VGAT, but in J20, both Gephyrin and VGAT seem to be affected by DKK3 shRNA (Fig 6). The authors need to provide the pre vs post synapse number in Fig 6 and discuss the potential differences.

We have now included the quantification of excitatory and inhibitory pre- and postsynaptic puncta for 4-months old (Figure S6B, C) and 9-months old (Figure S6D, E) WT and J20 mice. At 4-months old, the density of Homer1 puncta for excitatory synapses and both vGAT and Gephyrin for inhibitory synapses was increased and decreased respectively by knocking down DKK3 in the J20 mice. At 9-months, strong trends were observed in all the synaptic markers when downregulating DKK3, but significance was only reached for Homer1 puncta.

1. Where are the Wnt receptors expressed? Are they exclusively expressed in neurons? Can the authors exclude the potential involvement of glial cells in this process?

In neurons, Wnt receptors can be expressed in the synaptic terminals. For example, Wnt receptor Frizzled-5 is located at the presynaptic terminal and the dendritic shaft but not at spines (Sahores M et al., 2010, PMID: 20530549; McLeod F et al., 2018, PMID: 29694885), whereas Frizzled-7 is located at the dendritic shaft and spines (McLeod F et al., 2018, PMID: 29694885). In addition, the Wnt co-receptor LRP6 is present at both pre- and postsynaptic sites in excitatory synapses (Jones ME et al., 2023, PMID: 36638182). Kremen1, another receptor for Dkk proteins, is also highly expressed in the brain and our unpublished superresolution results show that this receptor is present in both pre- and postsynaptic sites of 53% of excitatory and 30% of inhibitory synapses. However, these receptors are not exclusively expressed in neurons and many of them are also highly expressed in astrocytes (Zhang Y et al., 2016, PMID: 25186741). Based on the literature and our findings, we cannot rule out the possibility that DKK3 may signal to other cell types such as astrocytes, which could also contribute to changes in synapse density. However, recombinant DKK3 induces structural and functional changes in excitatory and inhibitory synapses within 3-4h (Figure 3), suggesting that DKK3 acts on neurons leading to synaptic changes.

1. Does the shRNA treatment of DKK3 affect the size and number of amyloid plaques in the AD mice?

We thank the reviewer for raising this very important question. We have now evaluated the impact of DKK3 knockdown in Aβ pathology in the J20 mice. We did not observe differences in the Aβ coverage nor the averaged number and size of Aβ plaques when DKK3 was silenced in the CA3 (Figure S6F). Therefore, the changes we observe in excitatory and inhibitory synapse density around plaques after knocking down DKK3 are unlikely to be due to changes in Aβ plaques.

Author Response

eLife assessment

This study presents a valuable finding on the distinct subpopulation of adipocytes during brown-to-white conversion in perirenal adipose tissue (PRAT) at different ages. The evidence supporting the claims of the authors is convincing, although specific lineage tracing of this subpopulation of cells and mechanistic studies would expand the work. The work will be of interest to scientists working on adipose and kidney biology.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors performed single nucleus RNA-seq for perirenal adipose tissue (PRAT) at different ages. They concluded a distinct subpopulation of adipocytes arises through brown-to-white conversion and can convert to a thermogenic phenotype upon cold exposure.

Strengths:

PRAT adipose tissue has been reported as an adipose tissue that undergoes browning. This study confirms that brown-to-white and white-to-beige conversions also exist in PRAT, as previously reported in the subcutaneous adipose tissue.

We did not observe any white-to-beige conversion in PRAT under regular condition. The adipocyte population that arises from brown-to-white conversion (mPRAT-ad2) can respond to cold and restore their UCP1 expression. However, brown adipocytes that arise from the mPRAT-ad2 subpopulation after cold exposure have a distinct transcriptome to that of cold-induced beige adipocyte in iWAT (Figure S6K) and are more related to iBAT brown adipocytes (Figure 6E).

Weaknesses:

1. There is overall a disconnection between single nucleus RNA-seq data and the lineage chasing data. No specific markers of this population have been validated by staining.

We are not sure what “this population” refers to. We suspect it is the Ucp1-&Cidea+ mPRAT-ad2 adipocyte subpopulation. If so, we did not identify specific markers for these adipocytes as shown in Figure 1H and statement in the Discussion. mPRAT-ad2 is negative for Ucp1 and Cyp2e1, which are markers for mPRAT-ad1 and mPRAT-ad3&4, respectively. Therefore, we plan to stain the mPRAT with Ucp1, Cyp2e1 and Perilipin (a pan adipocyte marker) antibodies. Cells that are Perilipin+&Ucp1-&Cyp2e1- will represent the mPRAT-ad2 subpopulation.

1. It would be nice to provide more evidence to support the conclusion shown in lines 243 to 245 "These results indicated that new BAs induced by cold exposure were mainly derived from UCP1- adipocytes rather than de novo ASPC differentiation in puPRAT". Pdgfra-negative progenitor cells may also contribute to these new beige adipocytes.

Our sequencing data and many previous studies (Angueira et al., 2021; Burl et al., 2022; Dong et al., 2022) have shown that Pdgfra is a marker for all ASPCs. We will also check adipocyte labelling pattern of mPRAT in the PdgfraCre;Ai14 mice. If all adipocytes are Tomato+, it suggests that adipocytes in mPRAT are all derived from Pdgfra-expressing cells. Also, the cold-induced adipocytes in mPRAT resemble more to the brown adipocytes of iBAT than the beige adipocytes of iWAT (Figure 6E and S6K).

Angueira, A.R., Sakers, A.P., Holman, C.D., Cheng, L., Arbocco, M.N., Shamsi, F., Lynes, M.D., Shrestha, R., Okada, C., Batmanov, K., et al. (2021). Defining the lineage of thermogenic perivascular adipose tissue. Nat Metab 3, 469-484. 10.1038/s42255-021-00380-0.

Burl, R.B., Rondini, E.A., Wei, H., Pique-Regi, R., and Granneman, J.G. (2022). Deconstructing cold-induced brown adipocyte neogenesis in mice. Elife 11. 10.7554/eLife.80167.

Dong, H., Sun, W., Shen, Y., Balaz, M., Balazova, L., Ding, L., Loffler, M., Hamilton, B., Kloting, N., Bluher, M., et al. (2022). Identification of a regulatory pathway inhibiting adipogenesis via RSPO2. Nat Metab 4, 90-105. 10.1038/s42255-021-00509-1.

1. The UCP1Cre-ERT2; Ai14 system should be validated by showing Tomato and UCP1 co-staining right after the Tamoxifen treatment.

We will inject Ucp1CreERT2;Ai14 mice at 1- and 6-month-old of age with tamoxifen and collect one day after the last injection to check the overlap between the Tomato signal and UCP1 immunofluorescent staining.

Reviewer #2 (Public Review):

Summary:

In the present manuscript, Zhang et al utilize single-nuclei RNA-Seq to investigate the heterogeneity of perirenal adipose tissue. The perirenal depot is interesting because it contains both brown and white adipocytes, a subset of which undergo functional "whitening" during early development. While adipocyte thermogenic transdifferentiation has been previously reported, there remain many unanswered questions regarding this phenomenon and the mechanisms by which it is regulated.

Strengths:

The combination of UCP1-lineage tracing with the single nuclei analysis allowed the authors to identify four populations of adipocytes with differing thermogenic potential, including a "whitened" adipocyte (mPRAT-ad2) that retains the capacity to rapidly revert to a brown phenotype upon cold exposure. They also identify two populations of white adipocytes that do not undergo browning with acute cold exposure.

Anatomically distinct adipose depots display interesting functional differences, and this work contributes to our understanding of one of the few brown depots present in humans.

Weaknesses:

The most interesting aspect of this work is the identification of a highly plastic mature adipocyte population with the capacity to switch between a white and brown phenotype. The authors attempt to identify the transcriptional signature of this ad2 subpopulation, however, the limited sequencing depth of single nuclei somewhat lessens the impact of these findings. Furthermore, the lack of any form of mechanistic investigation into the regulation of mPRAT whitening limits the utility of this manuscript. However, the combination of well-executed lineage tracing with comprehensive cross-depot single-nuclei presented in this manuscript could still serve as a useful reference for the field.

The sequencing depth of our data is comparable, if not better than previously published snRNA-seq studies on adipose tissue (Burl et al., 2022; Sarvari et al., 2021; Sun et al., 2020). Therefore, the depth of our data has reached the limit of the 3’ sequencing methods. Unfortunately, due to size limitation of the adipocytes, it is also not feasible to sort them for Smart-seq.

Burl, R.B., Rondini, E.A., Wei, H., Pique-Regi, R., and Granneman, J.G. (2022). Deconstructing cold-induced brown adipocyte neogenesis in mice. Elife 11. 10.7554/eLife.80167.

Sarvari, A.K., Van Hauwaert, E.L., Markussen, L.K., Gammelmark, E., Marcher, A.B., Ebbesen, M.F., Nielsen, R., Brewer, J.R., Madsen, J.G.S., and Mandrup, S. (2021). Plasticity of Epididymal Adipose Tissue in Response to Diet-Induced Obesity at Single-Nucleus Resolution. Cell Metab 33, 437-453 e435. 10.1016/j.cmet.2020.12.004.

Sun, W., Dong, H., Balaz, M., Slyper, M., Drokhlyansky, E., Colleluori, G., Giordano, A., Kovanicova, Z., Stefanicka, P., Balazova, L., et al. (2020). snRNA-seq reveals a subpopulation of adipocytes that regulates thermogenesis. Nature 587, 98-102. 10.1038/s41586-020-2856-x.

Author Response

Public Reviews:

Roget et al. build on their previous work developing a simple theoretical model to examine whether ageing can be under natural selection, challenging the mainstream view that ageing is merely a byproduct of other biological and evolutionary processes. The authors propose an agent-based model to evaluate the adaptive dynamics of a haploid asexual population with two independent traits: fertility timespan and mortality onset. Through computational simulations, their model demonstrates that ageing can give populations an evolutionary advantage. Notably, this observation arises from the model without invoking any explicit energy tradeoffs, commonly used to explain this relationship.

The model’s results are based on both numerical simulations and formal mathematical analysis.

Additionally, the theoretical model developed here indicates that mortality onset is generally selected to start before the loss of fertility, irrespective of the initial values in the population. The selected relationship between the fertility timespan and mortality onset depends on the strength of fertility and mortality effects, with larger effects resulting in the loss of fertility and mortality onset being closer together. By allowing for a trans-generational effect on ageing in the model, the authors show that this can be advantageous as well, lowering the risk of collapse in the population despite an apparent fitness disadvantage in individuals. Upon closer examination, the authors reveal that this unexpected outcome is a consequence of the trans-generational effect on ageing increasing the evolvability of the population (i.e., allowing a more effective exploration of the parameter landscape), reaching the optimum state faster.

The simplicity of the proposed theoretical model represents both the major strength and weakness of this work. On one hand, with an original and rigorous methodology, the logic of their conclusions can be easily grasped and generalised, yielding surprising results. Using just a handful of parameters and relying on direct competition simulations, the model qualitatively recapitulates the negative correlation between lifespan and fertility without requiring energy tradeoffs. This alone makes this work an important milestone for the rapidly growing field of adaptive dynamics, opening many new avenues of research, both theoretically and empirically.

We thank the reviewers and editor for highlighting the importance of the work presented here.

On the other hand, the simplicity of the model also makes its relationship with living organisms difficult to gauge, leaving open questions about how much the model represents the reality of actual evolution in a natural context.

We presented both in results and discussion how the mathematical trade-offs between fertility and survival time give rise to (xb, xd) configuration representative of existing aging modes.

In particular, a more explicit discussion of how the specifics of the model can impact the results and their interpretation is needed. For example, the lack of mechanistic details on the trans-generational effect on ageing makes the results difficult to interpret.

We discussed the role of the transgenerational Lansing effect played to its function, there is no need for a particular mechanism beyond that function of transgenerational negative effect. We reinforce this in the discussion by adding the following sentence “Regarding the nature of the transgenerational effect, our model is agnostic and the mere transmission of any negative effect would be sufficient to exert the function. “

Even if analytical results are obtained, most of the observations appear derived from simulations as they are currently presented. Also, the choice of parameters for the simulations shown in the paper and how they relate to our biological knowledge are not fully addressed by the authors.

The long time limit of the system with and without the Lansing effect is based on analytical results later confirmed using numerical simulations. The choice of parameters is explained in the introduction as being the minimum ones for defining a living organism. As for the parameters’ values, our numerical analysis gives a solution for any ib, id, xb and xd on R+, making the choice of initial value a mere random decision.

Finally, the conclusions of evolvability are insufficiently supported, as the authors do not show if the wider genotypic variability in populations with the ageing trans-generational effect is, in fact, selected.

We do not show nor claim that evolvability per se is selected for but that the apparent advantage given by this transgenerational effect seems to be mediated by an increased genotypic/phenotypic variability conferred to the lineage that we interpreted as evolvability.

Author Response

Reviewer #1 (Public Review):

De Seze et al. investigated the role of guanine exchange factors (GEFs) in controlling cell protrusion and retraction. In order to causally link protein activities to the switch between the opposing cell phenotypes, they employed optogenetic versions of GEFs which can be recruited to the plasma membrane upon light exposure and activate their downstream effectors. Particularly the RhoGEF PRG could elicit both protruding and retracting phenotypes. Interestingly, the phenotype depended on the basal expression level of the optoPRG. By assessing the activity of RhoA and Cdc42, the downstream effectors of PRG, the mechanism of this switch was elucidated: at low PRG levels, RhoA is predominantly activated and leads to cell retraction, whereas at high PRG levels, both RhoA and Cdc42 are activated but PRG also sequesters the active RhoA, therefore Cdc42 dominates and triggers cell protrusion. Finally, they create a minimal model that captures the key dynamics of this protein interaction network and the switch in cell behavior.

We thank reviewer #1 for this assessment of our work.

The conclusions of this study are strongly supported by data. Perhaps the manuscript could include some further discussion to for example address the low number of cells (3 out of 90) that can be switched between protrusion and retraction by varying the frequency of the light pulses to activate opto-PRG.

The low number of cells being able to switch can be explained by two different reasons:

1) first, we were looking for clear inversions of the phenotype, where we could see clear ruffles in the case of the protrusion, and clear retractions in the other case. Thus, we discarded cells that would show in-between phenotypes, because we had no quantitative parameter to compare how protrusive or retractile they were. This reduced the number of switching cells

2) second, we had a limitation due to the dynamic of the optogenetic dimer used here. Indeed, the control of the frequency was limited by the dynamic of unbinding of the optogenetic dimer. This dynamic of recruitment (~20s) is comparable to the dynamics of the deactivation of RhoA and Cdc42. Thus, the differences in frequency are smoothed and we could not vary enough the frequency to increase the number of switches. Thanks to the model, we can predict that decreasing the unbinding rate of the optogenetic tool should allow us to increase the number of switching cells.

We will add further discussion of this aspect to the manuscript.

Also, the authors could further describe their "Cell finder" software solution that allows the identification of positive cells at low cell density, as this approach will be of interest for a wide range of applications.

There is a detailed explanation of the ‘Cell finder’ in the method sections. It is also available on github at https://github.com/jdeseze/cellfinder and currently in development to be more user-friendly and properly commented.

Reviewer #2 (Public Review):

Summary:

This manuscript builds from the interesting observation that local recruitment of the DHPH domain of the RhoGEF PRG can induce local retraction, protrusion, or neither. The authors convincingly show that these differential responses are tied to the level of expression of the PRG transgene. This response depends on the Rho-binding activity of the recruited PH domain and is associated with and requires (co?)-activation of Cdc42. This begs the question of why this switch in response occurs. They use a computational model to predict that the timing of protein recruitment can dictate the output of the response in cells expressing intermediate levels and found that, "While the majority of cells showed mixed phenotypes irrespectively of the activation pattern, in few cells (3 out of 90) we were able to alternate the phenotype between retraction and protrusion several times at different places of the cell by changing the frequency while keeping the same total integrated intensity (Figure 6F and Supp Movie)."

Strengths:

The experiments are well-performed and nicely documented. However, the molecular mechanism underlying the shift in response is not clear (or at least clearly described). In addition, it is not clear that a prediction that is observed in ~3% of cells should be interpreted as confirming a model, though the fit to the data in 6B is impressive.

Overall, the main general biological significance of this work is that RhoGEF can have "off target effects". This finding is significant in that an orthologous GEF is widely used in optogenetic experiments in drosophila. It's possible that these findings may likewise involve phenotypes that reflect the (co-)activation of other Rho family GTPases.

We thank reviewer #2 for having assessed our work. Indeed, the main finding of this work is the change in the GEF function upon its change in concentration, which could be explained with a simple model supported by quantitative data. We think that the mechanism of the switch is quite clear, supported by the data showing the double effect of the PH domain and the activation of Cdc42. The few cells that are able to switch phenotype have to be seen as an honest data confirming that 1) concentration is indeed the main determinant of the protein’s function, and the switch is hard to obtain (which is also predicted by the model) 2) the two underlying networks are being activated at different timescales, which leaves some space for differential activation in the same cell. We are here limited by the dynamic of the optogenetic tool, as explained in the response to reviewer #1, and the intrinsic cell-to-cell variability.

Regarding the interpretation of our results as RhoGEF “off target effects”, we think that it might be too reductive. As said in the discussion, we proposed that the dual role of the RhoGEF could have physiological implications on the induction of front protrusions and rear retractions. While we do not demonstrate it here, it opens the door for further investigation.

Weaknesses:

The manuscript makes a number of untested assumptions and the underlying mechanism for this phenotypic shift is not clearly defined.

We may not have been clear in our manuscript, but we think that the underlying mechanism for this phenotypic shift is clearly explained and backed up by the data and the literature. It relies on 1) the ability of PRG to activate both RhoA and Cdc42 and 2) the ability of the PH domain to directly bind to active RhoA (which is, as shown in the manuscript, necessary but not sufficient for protrusions to happen). The model succeeds in reproducing the data of RhoA with only one free parameter and two independently fitted ones. The fact that activation of RhoA and Cdc42 lead to retraction and protrusion respectively is known since a long time. Thus, we think that the switch is clearly and quantitatively explained.

This manuscript is missing a direct phenotypic comparison of control cells to complement that of cells expressing RhoGEF2-DHPH at "low levels" (the cells that would respond to optogenetic stimulation by retracting); and cells expressing RhoGEF2-DHPH at "high levels" (the cells that would respond to optogenetic stimulation by protruding). In other words, the authors should examine cell area, the distribution of actin and myosin, etc in all three groups of cells (akin to the time zero data from figures 3 and 5, with a negative control). For example, does the basal expression meaningfully affect the PRG low-expressing cells before activation e.g. ectopic stress fibers? This need not be an optogenetic experiment, the authors could express RhoGEF2DHPH without SspB (as in Fig 4G).

We thank reviewer #2 for this suggestion. PRG-DHPH is known to affect the phenotype of the cell as shown in Valon et al., 2017. Thus, we really focused on the change implied by the change in optoPRG expression, to understand the phenotype difference. However, we agree that this could be an interesting data to add and will do the experiments for the revised version of the manuscript.

Relatedly, the authors seem to assume ("recruitment of the same DH-PH domain of PRG at the membrane, in the same cell line, which means in the same biochemical environment." supplement) that the only difference between the high and low expressors are the level of expression. Given the chronic overexpression and the fact that the capacity for this phenotypic shift is not recruitment-dependent, this is not necessarily a safe assumption. The expression of this GEF could well induce e.g. gene expression changes.

We agree with reviewer #2 that there could be changes in gene expression. In the next point of this supplementary note, we had specified it, by saying « that overexpression has an influence on cell state, defined as protein basal activity or concentration before activation. » We are sorry if it was not clear and will change this sentence for the new version.

One of the interests of the model is that it does not require any change in absolute concentrations, beside the GEF. The model is thought to be minimal and fits well and explains the data with very few parameters. We don’t show that there is no change in concentration but we show that it is not required to invoke it.

We will add in the revised version of the manuscript a paragraph discussing this question.

The third paragraph of the introduction, which begins with the sentence, "Yet, a large body of works on the regulation of GTPases has revealed a much more complex picture with numerous crosstalks and feedbacks allowing the fine spatiotemporal patterning of GTPase activities" is potentially confusing to readers. This paragraph suggests that an individual GTPase may have different functions whereas the evidence in this manuscript demonstrates, instead, that a particular GEF can have multiple activities because it can differentially activate two different GTPases depending on expression levels. It does not show that a particular GTPase has two distinct activities. The notion that a particular GEF can impact multiple GTPases is not particularly novel, though it is novel (to my knowledge) that the different activities depend on expression levels.

We thank the reviewer for this remark and didn’t intended to confuse the readers. Indeed, we think that this manuscript confirms the canonical view on the GTPases (as most optogenetic experiments did in the past years). We show here that it is more complicated at the level of the GEF. We agree that this is not particularly novel. However, to our knowledge, there is no example of such clear phenotypic control, explained solely by the change in concentration.

We think that the last paragraph of the introduction is quite clear in the fact that it is the GEF itself that switches its function, and not the Rho-GTPases, but we will reconsider the phrasing of this paragraph for the revised version.

Concerning the overall model summarizing the authors' observations, they "hypothesized that the activity of RhoA was in competition with the activity of Cdc42"; "At low concentration of the GEF, both RhoA and Cdc42 are activated by optogenetic recruitment of optoPRG, but RhoA takes over. At high GEF concentration, recruitment of optoPRG lead to both activation of Cdc42 and inhibition of already present activated RhoA, which pushes the balance towards Cdc42."

These descriptions are not precise. What is the nature of the competition between RhoA and Cdc42? Is this competition for activation by the GEFs? Is it a competition between the phenotypic output resulting from the effectors of the GEFs? Is it competition from the optogenetic probe and Rho effectors and the Rho biosensors? In all likelihood, all of these effects are involved, but the authors should more precisely explain the underlying nature of this phenotypic switch. Some of these points are clarified in the supplement, but should also be explicit in the main text.

We are going to precise these descriptions for the revised version of the manuscript. The competition between RhoA and Cdc42 was thought as a competition between retraction due to the protein network triggered by RhoA (through ROCK-Myosin and mDia-bundled actin) and the protrusion triggered by Cdc42 (through PAK-Rac-ARP2/3-branched Actin). We will make it explicit in the main text.

Author Response

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

eLife assessment

The findings of this study are valuable as they provide new insights into the role of acetylcholine in modulating sensory processing in the auditory cortex. This paper reports a systematic measurement of cell activity in the auditory cortex before and after applying ACh during an oddball and cascade sequence of auditory stimuli in anesthetized rats. The results presented are solid given the rigorous experimental design and statistical analysis. The conclusions are provocative and will interest researchers in auditory neuroscience and neuromodulation, as well as clinicians and individuals with auditory processing disorders. However, the findings support multiple interpretations, beyond that offered by the authors.

Our reply: First and foremost, we would like to thank the editors and reviewers for their constructive criticisms, as well as their thoughtful and thorough evaluations of our manuscript. We greatly appreciate their assessment about the novelty and general significance in our study and have revised the manuscript according to their recommendations. In the following we include detailed responses and revisions based on the reviewer’s recommendations.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study examined the impact of exogenous microapplication of acetylcholine (Ach) on metrics of novelty detection in the anesthetized rat auditory cortex. The authors found that the majority of units showed some degree of modulation of novelty detection, with roughly similar numbers showing enhanced novelty detection, suppressed novelty detection, or no change. Enhanced novelty responses were driven by increases in repetition suppression. Suppressed novelty responses were driven by deviance suppression. There were no compelling differences seen between auditory cortical subfields or layers, though there was heterogeneity in the Ach effects within subfields. Overall, these findings are important because they suggest that fluctuations in cortical Ach, which are known to occur during changes in arousal or attentional states, will likely influence the capacity of individual auditory cortical neurons to respond to novel stimuli.

Strengths:

The work addresses an important problem in auditory neuroscience. The main strengths of the study are that the work was systematically done with appropriate controls (cascaded stimuli) and utilizes a classical approach that ensures that drug application is isolated to the micro-environment of the recorded neuron. In addition, the authors do not isolate their study to only the primary auditory cortex, but examine the impact of Ach across all known auditory cortical subfields.

Our reply: Thank you very much for these supportive comments and the appreciation of our work.

Weaknesses:

1. As acknowledged by the authors, this study explicitly examines a phenomenon of high relevance to active listening but is done in anesthetized animals, limiting its applicability to the waking state.

Our reply: We agree; and indeed, this weakness was already recognized in the original manuscript but is now emphasized in the discussion.

1. The authors do not make any attempt to determine, by spike shape/duration, if their units are excitatory or inhibitory, which may explain some of the variance of the data.

Our reply: This is a very interesting question, and in fact, we have previously estimated whether neurons are excitatory or inhibitory based on the spike shape (Pérez-Gonzalez et al., 2021). Originally, we sought to implement a similar analysis here and tried to estimate if the recorded units were excitatory or inhibitory based on the spike shapes. But when we tried to perform this analysis, we found that in many cases the recordings had captured occasional spikes from other neurons. This caveat had introduced alterations in the average spike shape, and thus precluded an accurate categorization. Therefore, we decided to discard this analysis for the sake of correctness. This weakness is further commented on in the discussion.

1. The application of exogenous Ach, potentially in supra-physiological amounts, makes this study hard to extrapolate to a behaving animal. A more compelling design would be to block Ach, particularly at particular receptor types, to determine the effect of endogenous Ach.

Our reply: We agree again with the reviewer; this weakness was already acknowledged, but this is now further highlighted in discussion where we comment that future studies should analyze the effect of muscarinic- and nicotinic- receptors and blockade them to potentially observe more physiologically-comparable effects. Moreover, this issue is also related to a comment raised by reviewer#2 on a possible ‘dose-response relationship’ issue.

Reviewer #2 (Public Review):

Summary:

In this study, the authors investigate the effect of ACh on neuronal responses in the auditory cortex of anesthetized rats during an auditory oddball task. The paradigm consisted of two pure tones (selected from the frequency responses at each recording site) presented in a pseudo-random sequence. One tone was presented frequently (the "standard" tone) and the other infrequently (the "deviant" tone). The authors found that ACh enhances the detection of unexpected stimuli in the auditory environment by increasing or decreasing the neuronal responses to deviant and standard tones.

Strengths:

The study includes the use of appropriate and validated methodology in line with the current state-of-the-art, rigorous statistical analysis, and the demonstration of the effects of acetylcholine on auditory processing.

Our reply: Thank you very much for these supportive comments and the appreciation of our work.

Weaknesses:

The study was conducted in anesthetized rats, and further research is needed to determine the behavioral relevance of these findings.

Our reply: We agree; and indeed, this weakness was already recognized but is now emphasized in discussion.

Reviewer #1 (Recommendations For The Authors):

As outlined above, breaking out the units into those that are putative excitatory or inhibitory cells would be helpful, if possible. Other critiques are minor:

1. "Acetylcholine", "ACh" and "Ach" are used throughout the manuscript. Please define the chosen abbreviation at first use, and be consistent.

2. Line 116, remove comma after "ACh".

3. Line 123, I would add "in the rat at the end of the first sentence since the species was not mentioned up to this point.

4. Fig 2 - it would be useful in the Figure (not just in the text) to label red as being the deviant tone and blue as being the standard.

5. In many Figures (e.g., Fig 5), the term "effect" is found in the legend rather than "ACh". It would seem more intuitive to label these as "ACh".

6. The AUC and MI interpretations are not clear. Both are metrics that quantify similarity but the authors state that when these values decrease the neurons are less able to discriminate between them (i.e., they are more similar). Some clarifying text would be useful.

7. L276 - should "SI increase" be "SI decrease"?

8. L285 - would replace "solely" with "primarily".

9. Fig 7 - the authors may consider indicating with a label what the difference is between A and C compared to B and D.

10. L634 - why were only females used?

11. L646 - "bran" should be "brain".

12. L649 - "homoeothermic" should be "homeothermic".

13. L661 - "allowed to generate" should be "allowed the generation of".

14. L670 - no need for both "about" and "approximately".

15. L681 - please state what the search stimuli were.

16. L688 - should be "closed-field".

17. L754 - add a hyphen to "time-consuming".

Our reply: Thanks so much for the detailed proofreading of the manuscript and suggestions. All them have been clarified or implemented and corrected in the text.

Reviewer #2 (Recommendations For The Authors):

The authors could investigate the effects of different doses of ACh on auditory processing to determine if there is a dose-response relationship.

Our reply: We agree that this is an interesting question also relate to a matter raised by Reviewer#1 that could be linked to the issue of ‘exogenous Ach’.

The study only investigated the effects of ACh on neuronal responses during an auditory oddball task. It would be interesting to investigate the effects of ACh on other aspects of auditory processing, such as sound localization or the discrimination of tones.

Our reply: We agree that, while these aspects of auditory processing are very fascinating, they were outside the scope of the study, and not directly related to predictive coding and precision, so each one of these characteristics would be a full, future project in itself.

The authors could provide more context on the significance of their findings for individuals with auditory processing disorders.

Our reply: Thanks for the suggestion. It remains unclear how abnormal brainstem and cortical processing associated with auditory processing disorders arises (Moore, 2006, 2012). While we are not aware of any known direct connection between auditory processing disorders and acetylcholine, individuals with auditory processing disorders do have difficulties with auditory selective attention, so perhaps one could speculate that ACh, by modulating SSA/prediction error, could have some impact on encoding salient events, and if disrupted could lead to problems with selective attention. Moore (2012) speculated that auditory processing disorders may arise from unbalanced processing in bottom-up and top-down contributions.

Since ACh has been implicated in some neurogenerative diseases and neurodevelopmental disorders, we have also added in the Discussion dialogue about a possible relationship between the modulatory effect of ACh on predictive coding (which involves bottom-up and top-down contributions) and auditory processing disorders. We also cite the recent work by Felix and colleagues (2019) which is the only study we have found on the effects of ACh on auditory processing disorders where they analyzed altered temporal processing at the level of the brainstem in α7-subunit of the nicotinic acetylcholine receptor (α7-nAChR)-deficient mice. After studying α7-nAChR knockout mice of both sexes and wild-type colony controls, they concluded that the malfunction of the CHRNA7 gene that encodes the α7-nAChR may contribute to degraded spike timing in the midbrain, which may underlie the observed timing delay in the ABR signals. These authors propose that their findings are consistent with a role for the α7-nAChR in types of neurodevelopmental and auditory processing disorders. There is also evidence on cholinergic system disfunction being related to the pathophysiology of Alzheimer’s disease (Pérez-González et al., 2022). For instance, disfunction of the synapses of cholinergic neurons in the hippocampus and nucleus basalis of Meynert, as well as decreased choline acetyltransferase activity, is associated to memory disorders in Alzheimer’s disease (Hampel et al., 2018). Also, A Alzheimer’s disease D patients show reduced amounts of the vesicular ACh transporter in some brain areas (Aghourian et al., 2017). Finally, cholinesterase inhibitors seem to have some favorable effect in the treatment of Alzheimer’s disease patients (Sharma, 2019).

Aghourian M, Legault-Denis C, Soucy J-P, Rosa-Neto P, Gauthier S, Kostikov A, et al. 2017. Quantification of brain cholinergic denervation in Alzheimer’s disease using PET imaging with [18F]-FEOBV. Mol. Psychiatry 22:1531–1538. doi: 10.1038/mp.2017.183

Felix RA 2nd, Chavez VA, Novicio DM, Morley BJ, Portfors CV. 2019. Nicotinic acetylcholine receptor subunit α7-knockout mice exhibit degraded auditory temporal processing. J Neurophysiol. 122(2):451-465. doi: 10.1152/jn.00170.2019.

Hampel H, Mesulam M-M, Cuello AC, Khachaturian AS, Vergallo A, Farlow MR, et al. 2018. Revisiting the Cholinergic Hypothesis in Alzheimer’s Disease: emerging Evidence from Translational and Clinical Research. J. Prev. Alzheimers Dis. 6:1–14. doi:10.14283/jpad.2018.43

Moore DR. 2006. Auditory processing disorder (APD)-potential contribution of mouse research. Brain Res. 1091:200–206.

Moore DR. 2012. Listening difficulties in children: bottom-up and top-down contributions. J Commun Disord. ;45:411–418.

Pérez-González D, Parras GG, Morado-Díaz CJ, Aedo-Sánchez C, Carbajal GV, Malmierca MS. 2021. Deviance detection in physiologically identified cell types in the rat auditory cortex. Hear Res. 2021 Jan;399:107997. doi: 10.1016/j.heares.2020.107997.

Pérez-González D, Schreiner TG, Llano DA and Malmierca MS. 2022. Alzheimer’s Disease, Hearing Loss, and Deviance Detection. Front. Neurosci. 16:879480. doi: 10.3389/fnins.2022.879480

Sharma K. 2019. Cholinesterase inhibitors as Alzheimer’s therapeutics. Mol. Med. Rep. 20:1479–1487. doi:10.3892/mmr.2019.1 0374

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript by Heyndrickx et al describes protein crystal formation and function that bears similarity to Charcot-Leyden crystals made of galectin 10, found in humans under similar conditions. Therefore, the authors set out to investigate CLP crystal formation and their immunological effects in the lung. The authors reveal the crystal structure of both Ym1 and Ym2 and show that Ym1 crystals trigger innate immunity, activated dendritic cells in the lymph node, enhancing antigen uptake and migration to the lung, ultimately leading to induction of type 2 immunity.

Strengths:

We know a lot about expression levels of CLPs in various settings in the mouse but still know very little about the functions of these proteins, especially in light of their ability to form crystal structures. As such data presented in this paper is a major advance to the field.

Resolving the crystal structure of Ym2 and the comparison between native and recombinant CLP crystals is a strength of this manuscript that will be a very powerful tool for further evaluation and understanding of receptor, binding partner studies including the ability to aid mutant protein generation.

The ability to recombinantly generate CLP crystals and study their function in vivo and ex vivo has provided a robust dataset whereby CLPs can activate innate immune responses, aid activation and trafficking of antigen presenting cells from the lymph node to the lung and further enhances type 2 immunity. By demonstrating these effects the authors directly address the aims for the study. A key point of this study is the generation of a model in which crystal formation/function an important feature of human eosinophilic diseases, can be studied utilising mouse models. Excitingly, using crystal structures combined with understanding the biochemistry of these proteins will provide a potential avenue whereby inhibitors could be used to dissolve or prevent crystal formation in vivo.

The data presented flows logically and formulates a well constructed overall picture of exactly what CLP crystals could be doing in an inflammatory setting in vivo. This leaves open a clear and exciting future avenue (currently beyond the scope of this work) for determining whether targeting crystal formation in vivo could limit pathology.

Weaknesses:

Although resolving the crystal structure of Ym2 in particular is a strength of the authors work, the weaknesses are that further work or even discussion of Ym2 versus Ym1 has not been directly demonstrated. The authors suggest Ym2 crystals will likely function the same as Ym1, but there is insufficient discussion (or data) beyond sequence similarity as to why this is the case. If Ym1 and Ym2 crystals function the same way, from an evolutionary point, why do mice express two very similar proteins that are expressed under similar conditions that can both crystalise and as the authors suggest act in a similar way. Some discussion around these points would add further value.

We agree with reviewer. We have further elaborated the discussion section including these points, stating clearly that more research needs to be done using Ym2 crystals before we can draw parallels in vivo.

Additionally, the crystal structure for Ym1 has been previously resolved (Tsai et al 2004, PMID 15522777) and it is unclear whether the data from the authors represents an advance in the 3D structure from what is previously known.

The crystal structure of Ym1 has indeed been previously solved, and we refer to that paper. In addition, we also provide the crystal structure of in vitro grown Ym1, ashowing biosimilarity. This, for the field of crystallography is a major finding, since it validates the concept that crystal structures generated in vitro can reflect in vivo grown structures. Moreover, the in vivo crystallization of Ym2 was unknown prior to this work, and is now clear as revealed by the ex vivo X-ray crystallography. The strength of our story is that we can now compare Ym1 and Ym2 crystals structures in detail.

Whilst also generating a model to understand Charcot-Leyden crystals (CLCs), the authors fail to discuss whether crystal shape may be an important feature of crystal function. CLCs are typically needle like, and previous publications have shown using histology and TEM that Ym1 crystals are also needle like. However, the crystals presented in this paper show only formation of plate like structures. It is unclear whether these differences represent different methodologies (ie histology is 2D slides), or differences in CLP crystals that are intracellular versus extracellular. These findings highlight a key question over whether crystal shape could be important for function and has not been addressed by the authors.

In contrast to the bipyramidal, needle-like CLC crystals formed by human galectin-10 protein (hexagonal space group P6522), the in vivo grown Ym1 and Ym2 crystals we were able to isolate for X-ray diffraction experiments had a plate-like morphology with identical crystallographic parameters as recombinant Ym1/Ym2 crystals (space group P21). We note that depending on the viewing orientation of the thin plate-like Ym1 crystals, they may appear needle-like in histology and TEM images. In addition, we can fully not exclude that both Ym1 or Ym2 may crystallize in vivo in different space groups (which could result in different crystal morphologies for Ym1/Ym2) but we have no data to support this. It is finally also a possibility that plate like structures can break up in vivo along a long axis as a result of mechanical forces, and end up as rod-or needle like shapes.

Ym1/Ym2 crystals are often observed in conditions where strong eosinophilic inflammation is present. However, soluble Ym1 delivery in naïve mice shows crystal formation in the absence of a strong immune response. There is no clear discussion as to the conditions in which crystal formation occurs in vivo and how results presented in the paper in terms of priming or exacerbating an immune response align with what is known about situations where Ym1 and Ym2 crystals have been observed.

Although Ym1 and Ym2 crystals are often observed in mice at sites of eosinophilic inflammation, they are not made by eosinophils, but mainly by macrophages and epithelial cells, respectively. In vitro, protein crystallization typically starts from supersaturated solutions that support crystal nucleation. Several factors such as temperature and pH can affect the solubility of Ym1 and Ym2 in vivo and thus affect the nucleation and crystallization process. For Ym1 and Ym2 we noticed in vitro that a small drop in pH facilitates the crystallization process. Although the physiological pH is 7.4, during inflammation, there is a drop in pH. This drop in pH is the result of the infiltration and activation of inflammatory cells in the tissue, which leads to an increased energy and oxygen demand, accelerated glucose consumption via glycolysis and thus increased lactic acid secretion. In addition, we cannot exclude that in vivo, the nucleation process for Ym1/Ym2 is facilitated by interaction with ligands in the extracellular space (e.g. polysaccharide ligands or other – yet to be identified – specific ligands to Ym1/Ym2).

Reviewer #2 (Public Review):

Summary:

This interesting study addresses the ability of Ym1 protein crystals to promote pulmonary type 2 inflammation in vivo, in mice.

Strengths:

The data are extremely high quality, clearly presented, significantly extending previous work from this group on the type 2 immunogenicity of protein crystals.

Weaknesses:

There are no major weaknesses in this study. It would be interesting to see if Ym2 crystals behave similarly to Ym1 crystals in vivo. Some additional text in the Introduction and Discussion would enrich those sections.

We agree that this would be interesting to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Suggestions for improved experiments and to strengthen findings:

I think additional data on the ability of Ym2 crystals to induce an immune response would be advantageous. I'm not by any means suggesting the authors repeat all the experiments with Ym2 crystals, but even just the ability to show that Ym2 could promote type 2 immunity in the acute OVA model, would help to strengthen the argument that these crystals in general function in a similar way. Alternatively, a discussion on whether these protein crystals may function in different scenarios/tissues or conditions could help in light of additional data

We agree that this is an interesting point to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Measuring IL-33 in lung tissue is difficult to interpret as cells will express intracellular IL-33 that is not active and may explain why the results in Fig 2D are not overly convincing. It could just be that Ym1 crystals are changing the number of cells expressing IL-33 (e.g macrophages, or type 2 pneumocytes) Did the authors also measure active IL-33 release in the BAL fluid which may give a better indication of Ym1's ability to activate DAMPs?

We also measured active IL-33 release in the BAL fluid, but due to the limited sample availability we could only measure this in one of the two repeat experiments, resulting in non-significant results for the BAL fluid. However, certainly for the 6h timepoint we saw a similar trend in the BAL fluid as in the lung tissue, meaning higher levels of IL-33 in the Ym1 crystal group compared to the PBS and soluble Ym1 group.

Crystals in Fig 2F staining with Ym1 appear to be brighter in the soluble Ym1 group. Is this related to increased packing of Ym1 in the crystals formed in vivo as opposed to those formed in vitro? Aside from reduced amount of crystals that form when you give soluble Ym1, could the type of crystal also be influencing the ability of soluble Ym1 crystals to generate an immune response?

Our X-ray diffraction experiments show that the packing of Ym1 is identical for in vivo and in vitro grown crystals. Possibly the apparent difference in brightness is caused by stochastic staining by the antibody. In this regard we note that the crystals formed from soluble Ym1 after 24h also can appear as less bright in a similar fashion as recombinant Ym1 crystals.

Overall, the data and writing of the manuscript is presented to a very high standard

A few minor points:

• Fig 2F - a little unsure what the number in the left top corner of the images represented.

These numbers represent the picture numbers generated by the software, but as they don’t have any added value for the story, we removed these numbers from the images.

• Not clear why two different expression vectors were used - one for Ym1 and one for Ym2?

Because we observed that recombinant Ym2 is more poorly secreted in the mammalian cell culture supernatant as compared to recombinant Ym1, we produced Ym2 with an N-terminal hexahistidine-tag followed by a Tobacco Etch Virus (TEV)-protease cleavage site to facilitate its purification.

Reviewer #2 (Recommendations For The Authors):

The authors briefly outline in their Introduction potential Sources of Ym1/2 in vivo, highlighting monocytes, M2 macrophages, alveolar macrophages, neutrophils and epithelial cells. Do DCs also make detectable/meaningful amounts of Ym1/2 in vivo, particularly in type 2 settings?

In the introduction we only highlighted the main cellular sources of Ym1 and Ym2, but there is literature available stating/showing that Ym1/2 is not only expressed by macrophages, neutrophils, monocytes and epithelial cells, but can also be induced in DCs and mast cells. We added the word ‘mainly’ to this sentence in the introduction, to make clear that macrophages, neutrophils and monocytes are not the only sources of Ym1.

Given the nicely demonstrated similarity of recombinant Ym1 and Ym2 crystals, I think it is important for the authors to include at least initial data on the outcome of recombinant Ym2 crystal admin to mice, in comparison to their Ym1 data.

We agree that this is an interesting point to investigate, however, we choose to not include recombinant Ym2 crystal data in this report. However, we have further elaborated the discussion section including this point.

Given the generation of crystals following in vivo administration of soluble Ym1, albeit at a lower level than when crystals were administered, it would be interesting to see if increased concentrations of soluble material show a dose dependent increase in lung inflammation readouts.

We agree that this would be an interesting point to investigate. Alongside this we could also titrate down the crystal dose, to see if there is a dose dependent decrease in lung inflammation readouts. However, at this time, we choose to not investigate this further.

I couldn't easily follow the authors' Discussion about potential ability of anti Ym-1/2 Abs to dissolve Ym1/2 crystals (similar to what they have demonstrated for Abs vs Gal10 crystals). Have they addressed this possibility experimentally? If so, addition of such data to the manuscript would be extremely interesting, given the obvious potential Ym1/2 crystal dissolving Abs for investigation of the role of these in a range of different murine models of type 2 inflammation.

We agree that the phrasing of this part of the discussion can be unclear/confusing. We rephrased this part to make it clearer. However, we did not address the possibility of Ym1/2 crystal dissolving antibodies experimentally.

In the Results section, the authors briefly comment on the pro-type 2 nature of Ym1 crystals in relation to their previous work with uric acid and Gal10 crystals, proposing that the pulmonary type 2 response may be a 'generic response to crystals of different chemical composition'. The Discussion would be enriched by deeper exploration of this comment.

We have further elaborated the discussion section including this point.

Author Response

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

We thank the reviewers for their thorough reading of the manuscript and insightful comments. We have responded to both the “public review” and the “recommendations” and feel that the manuscript is now significantly strengthened.

Public Review comments

Reviewer #1:

Weaknesses:

1. The abstract does not discuss the reduction of E-gel consumption that occurs after multiple days of exposure to the THC formulation, but rather implies that a new model for chronic oral self-administration has been developed. Given that only two days of consumption was assessed, it is not clear if the model will be useful to determine THC effects beyond the acute measures presented here. The abstract should clarify that there was evidence of reduced consumption/aversive effects with repeated exposures.

Thank you for your observation. We have added language to address this in the manuscript and the abstract. The model developed in the manuscript is an acute exposure model, with the intention of further chronic exposure adaptations to be developed separately (page 2, line 29).

1. In the results section, the authors sometimes describe effects in terms of the concentration of gel as opposed to the dose consumed in mg/kg, which can make interpretation difficult. For example, the text describing Figure 1i states that significant effects on body temperature were achieved at 4 mg CTR-gel and 5 mg THC-gel, but were essentially equivalent doses consumed? It would be helpful to describe what average dose of THC produced effects given that consumption varied within each group of mice assigned to a particular concentration.

We thank the reviewer for this comment and have edited our text to clarify our results. For example, this point is further emphasized by the correlation of the data in Figure1l-n showing the relationship between individual consumption and behavioral readouts (page 11, line 225-226).

1. The description of the PK data in Figure 3 did not specify if sex differences were examined. Prior studies have found that males and females can exhibit stark differences in brain and plasma levels of THC and metabolites, even when behavioral effects are similar. However, this does depend on species, route, timing of tissue collection. It would be helpful to describe the PK profile of males and females separately.

We did compare sex dependent effects and found no significant effects after THC E-gel consumption. We’ve added additional language to address this point in the discussion (Supplementary tables T1 and T2).

1. In Figure 5, it is unclear how the predicted i.p. THC dose could be 30 mg/kg when 30 mg/kg was not tested by the i.p. route according to the figure, and if it had been it would have likely been almost zero acoustic startle, not the increased startle that was observed in the 2 hr gel group. It seems more likely that it would be equivalent to 3 mg/kg i.p. Could there be an error in the modeling, or was it based on the model used for the triad effects? This should be clarified.

We apologize for the confusion created by that data, and it has now been updated for clarity. The original ~30mg/kg was not a predicted dose consumed, but rather an expected dose consumed based on individual male v. female consumption data in Supplemental Figure S1b. For clarity on the figure, we’ve instead placed dashed lines that draw attention only to the predicted startle response expected from our THC-E-gel model. We have also updated the text which hopefully makes this clearer.

Reviewer #2:

Weaknesses:

Certainly, more THC mediated behavioral outcomes could have been tested, but the work presents a proof-of-concept study to investigate acute THC treatment.

It would have been interesting if this application form is also possible for chronic treatment regimen

We agree that a chronic treatment regimen and additional behavioral outcomes is the next, most exciting step for expanding this oral THC-E-gel consumption model, and something we are actively pursuing.

Reviewer #3:

Weaknesses:

The main weaknesses of the manuscript revolve around clarification of the Methods section. All of these weaknesses are described in the "Recommendations to authors" section. Revising the manuscript would account for many of these weaknesses.

Thank you for carefully reading through our methodology. We have made edits according to everything brought up in the recommendation section of reviewer comments.

Recommendations for Authors

Reviewer #1:

Minor edits to the text:

Abstract: "intraperitoneal contingent" should be "intraperitoneal noncontingent".

Line 221, this sentence needs editing for clarity.

Lines 249-250, incomplete sentence.

Line 284, the word "activity" is missing from "locomotor between mice".

Lines 299-301, incomplete sentence.

Thank you for finding these mistakes. All these recommendations have been incorporated into the final publication.

Reviewer #2:

1. The typical THC tetrad includes catalepsy. Why was this behavioral outcome not monitored?

We felt that locomotion, analgesia, and body temperature were robust behavioral readouts for monitoring cannabimimetic responses and that acoustic startle served as an additional, novel means of understanding THC-E-gel effects.

1. Please specify the exact substrain of C57BL/6 (i.e., J or N or some other)

C57BL/6J mice were used for the publication. This clarification has been made in the methods section.

1. Figure S3 is not mentioned in the result part, but only in the discussion.

Figure S3 is now referenced in the main body of the Results section.

1. It might be interesting to follow up the issue that the individual THC consumption is considerable, as depicted in Fig. 1e (at high dose). This will presumably also lead to different behavioral responses. Or is there individual metabolism, also difference male vs. female?

Thank you for the suggestion. We agree that the distribution of THC doses consumed (calculation based on weight) would be worth further investigating and have now included language about this (page 20, line 436). Please note that we did not find a sex difference (Supplemental Figure S1b), but it would be exciting to discover some biologically relevant cause such as individual absorption or metabolism

Reviewer #3:

Major

1. Methods: Were the observers of experiments blinded to animal treatment? Why or why not?

Multiple investigators performed the behavioral measurements and were not blinded to mouse treatments, but the dose consumed by each mouse remained blind. Thus, because animals consumed THC gelatin of their own volition while having ad libitum access, we performed the correlational analysis presented in Figure 1 l-n.

1. Methods: The authors could consider relating their study design to the ARRIVE guidelines and providing a statement as to whether their study adheres to these guidelines. Related to this, were mice provided with any environmental enrichment during the study?

We followed the ARRIVE guidelines with exception to investigator blinding (described above). Please note that mice were not provided with additional environmental enrichment during the study, a point that we specified in our methods (page 5, line 91).

1. Methods / Results: In the Methods it is stated that the triad of cannabimimetic behaviors was measured 1 h post-injection or immediately after gelatin exposure. Why were these timepoints chosen? Perhaps this wording should be revised because measurements of cannabimimetic effects were taken several times after drug exposure. Peak i.p. drug may occur earlier than 1 h whereas peak oral drug effect is likely to occur over a longer time period (i.e., not immediately after) due to delays of absorption and first pass metabolism. Is it possible that the authors have underestimated oral drug effects by selecting these timepoints? Please discuss.

We observed a reduction in locomotion activity starting 1 h following the beginning of exposure to the gelatin (Figure 2), suggesting initial cannabimimetic changes. Based on this observable response we chose to measure all cannabimimetic behaviors immediately following gelatin exposure. The exposure timeline for i.p. injection (1 h post-injection) was selected based on a standard published protocol (Metna-Laurent et al, 2017).

a. Pharmacodynamics: Related to this and because the aim of this paper is to establish a rodent oral dose model, could the authors discuss the need for better characterization of the time course of drug effects? For example, how might anti-nociception or locomotor activity vary following THC E-gel consumption? This is somewhat addressed in the locomotion time course in Figure 2G but could be elaborated on or discussed in more detail.

We agree that future studies should include additional time points measuring behavioral changes. This important point is now emphasized in the discussion (page 21, line 455).

b. Pharmacokinetics: Related to this point above, have the authors considered collecting blood or tissue samples from their i.p.-injected animals to assess drug pharmacokinetics as they relate to drug effect and as compared to oral THC consumption? I am not suggesting the authors conduct a completely new study for this manuscript; however, this could be raised as a future study and/or as a weakness of the current study.

We did not measure blood and tissue concentrations after i.p. administration due to the number of studies reporting these values by our co-author, Dr. Daniele Piomelli, that established these pharmacokinetic measures. Thus, we chose to reference these studies. Please note that repeating such measurements would be labor intensive, unnecessary use federal NIH resources and animals, while being very redundant to the existing literature.

c. Minor, but related to these points: In the results, page 14 line 299: the first sentence of this paragraph is confusing as written. The Reviewer recognizes that the authors are relating the pharmacokinetic work to previously published findings, but still thinks that measuring and comparing THC levels from their cohort of i.p.-injected animals would have benefitted the present study.

Thank you, this edit has been made in the manuscript.

1. Methods, Histology: The methods as described do not contain sufficient detail regarding THC and THC metabolite quantification. In addition, it is not clear from this section what Histology was performed and how (no histology results appear in the manuscript). Please add more detail to this section of the Methods.

We apologize for this typo and have corrected it in the methods section of the manuscript.

1. Methods / Results: The statistics section requires additional detail regarding the rationale for tests being performed on different datasets. In addition, a description of the curve fitting used for data in figures 1H-J, 4B-D, and S4 would be helpful to the reader.

Thank you, we have updated and provided more information regarding the curve fitting that was used in the methods and results section for the respective figure panels (page 9, line 183-184).

Minor

1. Throughout: The use of the phrase "high" dose is somewhat arbitrary and not defined relative to other doses of the THC formulation throughout the manuscript. The Reviewer suggests simply stating that THC was used, specifying the dose, or justifying in the Abstract and/or Introduction the classification of "high" based on relevant literature.

Thank you for the observation. We have removed this ambiguity by specifically mentioning the dose that was consumed (e.g., abstract page 2, line 20).

1. Abstract: define "CB1" in the abstract. Although this is a common abbreviation within the field, its use should be defined.

We have added this definition in the abstract for clarification.

1. Figure 2: why are the consumption panels B, C, and D given separate labels but the locomotor data are all labeled together as panel G?

Thank you for the observation, we have adjusted the labeling, so it is equal for both sets of panels.

Author Response

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

Thank you very much for forwarding these two important reviews on our paper. Please find hereby our point-by-point responses addressing the ideas, arguments and points of concern raised by the reviewers. We provide explanation of how these points have been incorporated in the paper.

We feel the review process has been a useful exercise and that the paper has greatly benefited in terms of clarity and accessibility. It is our hope that our findings may ignite renewed interest on unexplored and “unexpected” aspects of great ape vocal communication, inspire novel research, and invite bold new advances on the long-standing puzzle of language origins and evolution. In several relevant sections, we have also sought to explicitly address the point of doubt raised in eLife’s editorial assessment, published alongside the reviewed preprint of our paper. The editorial assessment stated that “…However the evidence provided to support the major claims of the paper is currently incomplete. Specifically, it is not yet clear how the rhythmic structuring found in these long calls is more similar to human language recursion per se rather than isochrony as a broader, more common phenomenon.” To directly clarify this point, we provide now various examples of how recursion is distinct from repetition, using everyday objects for an intuitive understanding (e.g., lines 43-51). We have also expanded the discussion to better contextualise and clarify the implications of our findings on language evolution theory. We hope this will help addressing the implicit request for clarification in the previous editorial assessment.

Thank you very much for your kind and dedicated attention in the processing of our study.

Public Reviews:

Reviewer #1 (Public Review):

This study investigates the structuring of long calls in orangutans. The authors demonstrate long calls are structured around full pulses, repeated following a regular tempo (isochronic rhythm). These full pulses are themselves structured around different sub-pulses, themselves repeated following an isochronic rhythm. The authors argue this patterning is evidence for self-embedded, recursive structuring in orangutang long calls.

The analyses conducted are robust and compelling and they support the rhythmicity the authors argue is present in the long calls. Furthermore, the authors went above and beyond and confirmed acoustically the sub-categories identified were accurate.

We thank the reviewer for this important support regarding our methods and findings.

However, I believe the manuscript would benefit from a formal analysis of the specific recursive patterning occurring in the long call. Indeed, as of now, it is difficult for the reader to identify what the authors argue to be recursion and distinguish it from simple repetitions of motifs, which is essential.

We agree with the reviewer that the distinction between repetition and recursion is very important for the adequate interpretation of our findings. Following the reviewer’s point (and the Editorial Assessement), we have now rephrased several passages in the initial paragraph of the paper for added clarity, where recursion is introduced and explained. We now also provide various new examples of recursion in everyday life and popular culture to better illustrate in an easy and accessible way the fundamental nature of recursion. We then use two of these common examples (computer folders and Russian dolls) to specifically distinguish repetition from recursion.

Although the authors already discuss briefly why linear patterning is unlikely, the reader would benefit from expanding on this discussion section and clarifying the argument here (a lay terminology might help).

Corrected accordingly.

I believe an illustration here might help. In the same logic, I believe a tree similar to the trees used in linguistics to illustrate hierarchical structuring would help the reader understand the recursive patterning in place here. This would also help get the "big picture", as Fig 1A is depicting a frustratingly small portion of the long call.

We completely understand the reviewer’s concern here. As proposed by the reviewer, and in addition to changes in the Introduction (see above) and Discussion (see below), we have now added a new figure in the Discussion to help the reader get the “big picture” of our findings.

We have also made revisions throughout the Introduction and Discussion to simplify the text, clarify our exposition and facilitate the reader better and intuitively understand the nature and relevance of our results.

Notwithstanding these comments, this paper would provide crucial evidence for recursion in the vocal production of a non-human ape species. The implication it would have would represent a key shift in the field of language evolution. The study is very elegant and well-constructed. The paper is extremely well written, and the point of view adopted is original, well-argued and compelling.

We are humbled by the reviewer’s words, and we thank the reviewer for attributing these qualities to our paper. This feedback reassures us of the disruptive potential that these and similar future findings may have on our understanding of language evolution.

Reviewer #2 (Public Review):

I am not qualified to judge the narrow claim that certain units of the long calls are isochronous at various levels of the pulse hierarchy. I will assume that the modelling was done properly. I can however say that the broad claims that (i) this constitutes evidence for recursion in non-human primates, (ii) this sheds light on the evolution of recursion and/or language in humans are, when not made trivially true by a semantic shift, unsupported by the narrow claims. In addition, this paper contains errors in the interpretation of previous literature.

We report the first confirmed case of “vocal sequences within vocal sequences” in a wild nonhuman primate, namely a great ape. The currently prevailing models of language evolution often rest on the (purely theorical) premise that such structures do not exist in any animal bar humans. We find the discovery of such structures in a wild great ape exciting, remarkable, and promising. We regret that the reviewer does not share this sentiment with us. We feel that the statement that these findings are trivial and narrow is unfounded.

In order to clarify and better communicate the significance of our findings, we now explain in more detail in the Introduction and Discussion how the discovery of nested isochrony in wild orangutans promises to stimulate new series of studies in nature and captivity. Our findings dovetail nicely with previous captive studies that have shown that animals can learn how to recognise recursive patterns and invite new research efforts for the investigation of recursive abilities in the wild and in the absence of human priming and in nonhuman primates.

The main difficulty when making claims about recursion is to understand precisely what is meant by "recursion" (arguably a broader problem with the literature that the authors engage with). The authors offer some characterization of the concept which is vague enough that it can include anything from "celestial and planetary movement to the splitting of tree branches and river deltas, and the morphology of bacteria colonies". With this appropriately broad understanding, the authors are able to show "recursion" in orangutans' long calls. But they are, in fact, able to find it everywhere.

The reviewer is correct in highlighting that recursion is ubiquitous in nature and this is something that we explicitly state in the paper. This only makes it the more surprising that, when it comes to vocal combinatorics, recursion has only been described in human language and music, but in no other animals. If studies providing such evidence are known to reviewer, we kindly request their corresponding references.

In the new revised version, we have paid attention to this aspect raised by the reviewer, and we have sought to disambiguate that our observations pertain to temporal recursion. This clarification will hopefully allow a better understanding of our results.

The sound of a plucked guitar string, which is a sum of self-similar periodic patterns, count as recursive under their definition as well.

The example pointed out here by reviewer is factually correct; sound harmonics represent a recursive pattern of a fundamental frequency. (In fact, we explain this phenomenon in the Discussion.) The reviewer’s comment seems to offer an analogy to oscillatory phenomena in the physiology of the vocal folds, and so, it is misplaced with regards to our present study, which focused vocal sequences. Admittedly, this misinterpretation may have been implicitly caused by our wording and we apologise for this. We now refer to “vocal combinatorics” instead of “vocal production” throughout the paper to avoid the reader considering that our findings pertain to the physiology of the vocal folds.

One can only pick one's definition of recursion, within the context of the question of interest: evolution of language in humans. One must try to name a property which is somewhat specific to human language, and not a ubiquitous feature of the universe we live in, like self-similarity. Only after having carved out a sufficiently distinctive feature of human language, can we start the work of trying to find it in a related species and tracing its evolutionary history. When linguists speak of recursion, they speak of in principle unbounded nested structure (as in e.g., "the doctor's mother's mother's mother's mother ..."). The author seems to acknowledge this in the first line of the introduction: "the capacity to iterate a signal within a self-similar signal" (emphasis added). In formal language theory, which provides a formal and precise definition of one notion of recursivity appropriate for human language, unbounded iteration makes a critical difference: bounded "nested structures" are regular (can be parsed and generated using finite-state machines), unbounded ones are (often) context-free (require more sophisticated automaton). The hierarchy of pulses and sub-pulses only has a fixed amount of layers, moreover the same in all productions; it does not "iterate".

The reviewer explains here how recursion, in its fully fledged form in modern language(s), is defined by linguistics. We fully agree and do not contest such descriptions and definitions in any way. These descriptions and definitions aim to describe how recursion operates today, not how it evolved. Nor do these descriptions and definitions generate data-driven, testable predictions about precursors or proto-states of recursion as used by modern language-able humans. This is scientifically problematic and heuristically unsatisfying regarding the open question of language evolution.

Following human-specific definitions for recursion, as proposed by the reviewer, cannot per se be used to undertake a comparative approach to evolution because they leave nothing to compare recursion with in other (wild) species. Using human-specific definitions unavoidably leads to black-and-white notions that language is always absolutely present in humans and always absolutely absent in other animals, regardless of their degree of relatedness to humans. It is unpreventable that these descriptions flout foundational principles of evolution, such as descent with modification and shared ancestry.

This conceptual problem is not new. Less than a century ago, it was believed that humans were the only tool-user (thousands of examples are known today in nonhuman animals, including fish and invertebrates), and later, that humans were the only cultural animal (today it is known that migrating caribou and fruit flies can establish traditions based on social learning). We must follow in the footsteps of those who have helped redefine human nature in the past. As famously stated by Louis Leakey when presented with evidence for chimpanzee tool-use collected by Jane Goodall, “Now we must redefine tool, redefine man, or accept chimpanzees as human”. Therefore, as a matter of course, we must redefine recursion, embracing empirically (other than purely theoretically) definitions that allow recursion to take on forms and functions different from that of modern language-able humans.

Another point is that the authors don't show that the constraints that govern the shape of orangutans long calls are due to cognitive processes.

The reviewer is indeed correct. This does not, however, refute our findings. We do not directly show that cognitive processes govern recursion in orangutan long calls. Instead, we show that the observed patterns cannot be explained by simple bodily or motoric processes, excluding therefore low-level explanations. With more than 50 years of accumulated field experience in primatology, this was the only possible way that our team found to go about conducting research and analyses on natural behaviour, in the wild, with a critically endangered primate. We would be very interested in learning from the reviewer what ethical and non-invasive methods, specific locations in the wild, and type of behavioural or socio-ecological data could be otherwise viably used to demonstrate what the reviewer requests. If other scientists believe that the patterns observed in wild orangutan long calls – three independent, but simultaneously-occurring recursive motifs – can be generated based on low-level physiological mechanisms alone, the burden of proof resides with them.

Any oscillating system will, by definition, exhibit isochrony.

We disagree with this statement. The example provided above by the reviewer him/her-self disproves the statement: a guitar string when struck is an oscillating system but it is not isochronic nor is it combinatorial. Isochrony cannot be established with single events, only with event sequences (in practice, ideally >3).

For instance, human trills produce isochronouns or near isochronous pulses. No cognitive process is needed to explain this; this is merely the physics of the articulators. Do we know that the rhythm of the pulses and sub-pulses in orangutans is dictated by cognition as opposed to the physics of the articulators?

The reviewer seems to misinterpret our results here. Our focus is on vocal combinatorics, not vocal fold oscillation (see previous response). We have now reworded all instances where the text could be unclear.

Even granting the authors' unjustified conclusion that wild orangutans have "recursive" structures and that these are the result of cognition, the conclusions drawn by the authors are too often fantastic leaps of induction. Here is a cherry-picked list of some of the far-fetched conclusions: - "our findings indicate that ancient vocal patterns organized across nested structural strata were likely present in ancestral hominids". Does finding "vocal patterns organized across nested structural strata" in wild orangutans suggest that the same were present in ancestral hominids?

Following the reviewer’s comment, we have now rephrased and toned down this passage, stating that such structures “may have been present” in ancestral hominids. We are grateful to the reviewer for this comment.

• "given that isochrony universally governs music and that recursion is a feature of music, findings (sic.) suggest a possible evolutionary link between great ape loud calls and vocal music". Isochrony is also a feature of the noise produced by cicadas. Does this suggest an evolutionary link between vocal music and the noise of cicadas?

We apologise, but it is unclear what the reviewer is exactly suggesting or proposing here. It seems as though it is believed that cicadas are as phylogenetically related to humans as great apes are. Our last common ancestor with great apes diverged about 10mya, but with cicadas 600mya. The last common ancestor with great apes was a great ape (or hominid). The human-cicada last common ancestor would have looked like a worm (it is probable it would already have a nervous bulge at the head, or “brain”). In order to avoid similar misinterpretations, we have now clarified in several instances that our study and interpretation of results are based on shared ancestry within the Hominid family.

It seems that the reviewer may be also misinterpreting our findings. We do not simply report isochrony in a wild great ape (multiple references for isochronous calls in primate are provided in the Discussion). We report isochrony within isochrony in three non-exclusive rhythmic arrangements. In case the reviewer knows of a study on cicadas, or any non-human species, showing recursive sound combinatorics of this nature, we kindly request the citation. We can only hope that such new cases may be gradually unveiled in wild animals to help propel our general understanding of possible ways of how insipient recursive vocal combinatorics in ancient hominids could have given rise to recursion as used today by language-able modern humans.

Finally, some passages also reveal quite glaring misunderstandings of the cited literature. For instance:

• "Therefore, the search for recursion can be made in the absence of meaning-base operations, such as Merge, and more generally, semantics and syntax". It is precisely Chomsky's (disputable) opinion that the main operation that govern syntax, Merge, has nothing to do with semantics. The latter is dealt within a putative conceptual-intentional performance system (in Chomsky's terminology), which is governed by different operations.

Following the reviewer’s comment, we have now removed “meaning-base operations, such as Merge, and more generally” from the target sentence in order to avoid confusion. Thank you.

• "Namely, experimental stimuli have consisted of artificial recursive signal sequences organized along a single temporal scale (though not structurally linear), similarly with how Merge and syntax operate". The minimalist view advocated by Chomsky assumes that mapping a hierarchichal structure to a linear order (a process called linearizarion) is part of the articulatory-perceptual system. This system is likewise not governed by Merge and is not part of "syntax" as conceived by the Chomskyan minimalists.

Following the reviewer’s comment, we have not omitted the target sentence for added clarity.

Reviewer #1 (Recommendations For The Authors):

L55-67: I feel there is a step missing in the logic of the argumentation here. The studies cited by the authors here are mostly about syntactic-like structuring but not recursion. Hence when the authors mention in the next sentence that these studies investigate the perception of recursive signalling, it seems incorrect. I agree with the logic, but the references do not seem appropriate. I would further suggest that if there are no other references, that would make the introduction of the study here even easier: there is very little work investigating this capacity in non-human animals, let alone on a production perspective, therefore, the study conducted here is paramount and fills this important gap in the literature.

We are grateful to Reviewer #1 for these comments, and we are honoured to hear that our findings are filling a literature gap. We have now carefully revised the manuscript, hopefully, streamlining our line of reasoning and improving the paper’s overall readability. We agree that there is very little work investigating the spontaneous “production” of recursion in nonhuman animals. We decided to better detail the logic of our paper by clarifying the difference between recursion and repetition and clarifying that the motifs that we identify in wild orangutan represent a case of "temporal recursion".

L59: Johan J should be removed (same in discussion).

Removed, thanks.

L60: For example is repeated twice, here and L55.

We have rephrased this part of the manuscript, thanks.

L72-73: If we consider the Watson et al., 2020 study an example of recursive perception (which I do not think is true), this was conducted using a passive design - i.e. with no active training.

We have rephrased this part of the manuscript, thanks.

L240-241: Again, non-adjacent dependency processing does not equal recursion.

We agree that non-adjacent dependency processing does not equal recursion. We have now clarified this section accordingly.

L269: one of the most.

Corrected, thanks.

L296: add space after settings.

Corrected, thanks.

Reviewer #2 (Recommendations For The Authors):

In addition to the public portion of the review, I advise the authors' to substantially alter their style of writing. The language used is not accurate and the intended meaning is often not clear. This makes it hard for any reader to follow the authors' reasoning fully. Below I list only a few of the egregious examples but the examples abound:

• "this hints at a neuro-cognitive or neuro-computational transformation in the human brain" what meaning do the author assign to "neuro-cognitive" and "neuro-computational" ? what difference do they place between the two (so that they would be disjoined.) ? What "transformation" are we talking about ? From what to what ?

• " However, recursive signal structures can also unfold in other manners, such as across nested temporal scales and in the absence of semantics (Fitch, 2017a), as in music." what is meant here by nested temporal scales ?

• "The simultaneous occurrence of non-exclusive recursive patterns excludes the likelihood that orangutans concatenate long calls and their subunits in linear structure without any recursive processes": isn't there a more straightforward way to say "excludes the likelihood"? What is meant by "non-exclusive recursive patterns"?

It seems that Reviewer #2 does not share our writing style. Nonetheless, we have tried to meet the reviewer halfway, clarifying throughout the new revised version our definitions, our line of argument, our motivations, our results, the context of our findings in what is known about recursion in animals, and the implication of our discovery for language evolution theory.

Author Response

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

We agree with the reviewer that the statistics are buried in a dense excel file without a read-me page. We will address this by making a summary excel page for p-values during the production process.

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

eLife assessment

This important study uses genomically-engineered glypican alleles to demonstrate convincingly that Dally (not Dally-like protein [Dlp]) is the key contributor to formation of the Dpp/BMP morphogen gradient in the wing disc of Drosophila. The authors provide solid genetic evidence that, surprisingly, the core domain of Dally appears to suffice to trap Dpp at the cell surface. They conclude with a model according to which Dally modulates the range of Dpp signaling by interfering with Dpp's internalization by the Dpp receptor Thickveins.

Public Reviews:

Reviewer #1 (Public Review):

How morphogens spread within tissues remains an important question in developmental biology. Here the authors revisit the role of glypicans in the formation of the Dpp gradient in wing imaginal discs of Drosophila. They first use sophisticated genome engineering to demonstrate that the two glypicans of Drosophila are not equivalent despite being redundant for viability. They show that Dally is the relevant glypican for Dpp gradient formation. They then provide genetic evidence that, surprisingly, the core domain of Dally suffices to trap Dpp at the cell surface (suggesting a minor role for GAGs). They conclude with a model that Dally modulates the range of Dpp signaling by interfering with Dpp's degradation by Tkv. These are important conclusions, but more independent (biochemical/cell biological) evidence is needed.

As indicated above, the genetic evidence for the predominant role of Dally in Dpp protein/signalling gradient formation is strong. In passing, the authors could discuss why overexpressed Dlp has a negative effect on signaling, especially in the anterior compartment. The authors then move on to determine the role of GAG (=HS) chains of Dally. They find that in an overexpression assay, Dally lacking GAGs traps Dpp at the cell surface and, counterintuitively, suppresses signaling (fig 4 C, F). Both findings are unexpected and therefore require further validation and clarification, as outlined in a and b below.

a. In loss of function experiments (dallyDeltaHS replacing endogenous dally), Dpp protein is markedly reduced (fig 4R), as much as in the KO (panel Q), suggesting that GAG chains do contribute to trapping Dpp at the cell surface. This is all the more significant that, according to the overexpression essays, DallyDeltaHS seems more stable than WT Dally (by the way, this difference should also be assessed in the knock-ins, which is possible since they are YFP-tagged). The authors acknowledge that HS chains of Dally are critical for Dpp distribution (and signaling) under physiological conditions. If this is true, one can wonder why overexpressed dally core 'binds' Dpp and whether this is a physiologically relevant activity.

According to the overexpression assay, DallyDeltaHS seems more stable than WT Dally (Fig. 4B’, E’, 5A’, B’). As the reviewer suggested, we addressed the difference using the two knock-in alleles and found that DallyDeltaHS is more stable than WT Dally (Fig.4 L, M inset), further emphasizing the insufficient role of core protein of Dally for extracellular Dpp distribution.

In summary, we showed that, although Dally interacts with Dpp mainly through its core protein from the overexpression assay (Fig. 4E, I), HS chains are essential for extracellular Dpp distribution (Fig. 4R). Thus, the core protein of Dally alone is not sufficient for extracellular Dpp distribution under physiological conditions. These results raise a question about whether the interaction of core protein of Dally with Dpp is physiologically relevant. Since the increase of HS upon dally expression but not upon dlp expression resulted in the accumulation of extracellular Dpp (Fig. 2) and this accumulation was mainly through the core protein of Dally (Fig. 4E, I), we speculate that the interaction of the core protein of Dally with Dpp gives ligand specificity to Dally under physiological conditions.

To understand the importance of the interaction of core protein of Dally with Dpp under physiological conditions, it is important to identify a region responsible for the interaction. Our preliminary results overexpressing a dally mutant lacking the majority of core protein (but keeping the HS modified region intact) showed that HS chains modification was also lost. Although this is consistent with our results that enzymes adding HS chains also interact with the core protein of Dally (Fig. 4D), the dally mutant allele lacking the core protein would hamper us from distinguishing the role of core protein of Dally from HS chains.

Nevertheless, we can infer the importance of the interaction of core protein of Dally with Dpp using dally[3xHA-dlp, attP] allele, where dlp is expressed in dally expressing cells. Since Dally-like is modified by HS chains but does not interact with Dpp (Fig. 2, 4), dally[3xHA-dlp, attP] allele mimics a dally allele where HS chains are properly added but interaction of core protein with Dpp is lost. As we showed in Fig.3O, S, the allele could not rescue dallyKO phenotypes, consistent with the idea that interaction of core protein of Dally with Dpp is essential for Dpp distribution and signaling and HS chain alone is not sufficient for Dpp distribution.

b. Although the authors' inference that dallycore (at least if overexpressed) can bind Dpp. This assertion needs independent validation by a biochemical assay, ideally with surface plasmon resonance or similar so that an affinity can be estimated. I understand that this will require a method that is outside the authors' core expertise but there is no reason why they could not approach a collaborator for such a common technique. In vitro binding data is, in my view, essential.

We agree with the reviewer that a biochemical assay such as SPR helps us characterize the interaction of core protein of Dally and Dpp (if the interaction is direct), although the biochemical assay also would not demonstrate the interaction under the physiological conditions.

However, SPR has never been applied in the case of Dpp, probably because purifying functional refolded Dpp dimer from bacteria has previously been found to be stable only in low pH and be precipitated in normal pH buffer (Groppe J, et al., 1998)(Matsuda et al., 2021). As the reviewer suggests, collaborating with experts is an important step in the future.

Nevertheless, SPR was applied for the interaction between BMP4 and Dally (Kirkpatrick et al., 2006), probably because BMP4 is more stable in the normal buffer. Although the binding affinity was not calculated, SPR showed that BMP4 directly binds to Dally and this interaction was only partially inhibited by molar excess of exogenous HS, suggesting that BMP4 can interact with core protein of Dally as well as its HS chains. In addition, the same study applied Co-IP experiments using lysis of S2 cells and showed that Dpp and core protein of Dally are co-immunoprecipitated, although it does not demonstrate if the interaction is direct.

In a subsequent set of experiments, the authors assess the activity of a form of Dpp that is expected not to bind GAGs (DppDeltaN). Overexpression assays show that this protein is trapped by DallyWT but not dallyDeltaHS. This is a good first step validation of the deltaN mutation, although, as before, an invitro binding assay would be preferable.

Our overexpression assays actually showed that DppDeltaN is trapped by DallyWT and by dallyDeltaHS at similar levels (Fig. 5C), indicating that interaction of DppDeltaN and HS chains of Dally is largely lost but DppDeltaN can still interact with core protein of Dally.

We thank the reviewer for the suggesting the in vitro experiment. Although we decided not to develop biophysical experiments such as SPR for Dpp in this study due to the reasons discussed above, we would like to point out that our result is consistent with a previous Co-IP experiment using S2 cells showing that DppDeltaN loses interaction with heparin (Akiyama2008).

However, in contrast to our results, the same study also proposed by Co-IP experiments using S2 cells that DppDeltaN loses interaction with Dally (Akiyama2008). Although it is hard to conclude since western blotting was too saturated without loading controls and normalization (Fig. 1C in Akiyama 2008), and negative in vitro experiments do not necessarily demonstrate the lack of interaction in vivo. One explanation why the interaction was missed in the previous study is that some factors required for the interaction of DppDeltaN with core protein of Dally are missing in S2 cells. In this case, in vivo interaction assay we used in this study has an advantage to robustly detect the interaction.

Nevertheless, the authors show that DppDeltaN is surprisingly active in a knock-in strain. At face value (assuming that DeltaN fully abrogates binding to GAGs), this suggests that interaction of Dpp with the GAG chains of Dally is not required for signaling activity. This leads to authors to suggest (as shown in their final model) that GAG chains could be involved in mediating the interactions of Dally with Tkv (and not with Dpp. This is an interesting idea, which would need to be reconciled with the observation that the distribution of Dpp is affected in dallyDeltaHS knock-ins (item a above). It would also be strengthened by biochemical data (although more technically challenging than the experiments suggested above). In an attempt to determine the role of Dally (GAGs in particular) in the signaling gradient, the paper next addresses its relation to Tkv. They first show that reducing Tkv leads to Dpp accumulation at the cell surface, a clear indication that Tkv normally contributes to the degradation of Dpp. From this they suggest that Tkv could be required for Dpp internalisation although this is not shown directly. The authors then show that a Dpp gradient still forms upon double knockdown (Dally and Tkv). This intriguing observation shows that Dally is not strictly required for the spread of Dpp, an important conclusion that is compatible with early work by Lander suggesting that Dpp spreads by free diffusion. These result show that Dally is required for gradient formation only when Tkv is present. They suggest therefore that Dally prevents Tkv-mediated internalisation of Dpp. Although this is a reasonable inference, internalisation assays (e.g. with anti-Ollas or anti-HA Ab) would strengthen the authors' conclusions especially because they contradict a recent paper from the Gonzalez-Gaitan lab.

Thanks for suggesting the internalization assay. As we discussed in the discussion, our results suggest that extracellular Dpp distribution is severely reduced in dally mutants due to Tkv mediated internalization of Dpp (Fig. 6). Thus, extracellular Dpp available for labelling with nanobody is severely reduced in dally mutants, which can explain the reduced internalization of Dpp in dally mutants in the internalization assay. Therefore, we think that the nanobody internalization assay would not distinguish the two contradicting possibilities.

The paper ends with a model suggesting that HS chains have a dual function of suppressing Tkv internalisation and stimulating signaling. This constitutes a novel view of a glypican's mode of action and possibly an important contribution of this paper. As indicated above, further experiments could considerably strengthen the conclusion. Speculation on how the authors imagine that GAG chains have these activities would also be warranted.

Thank you very much!

Reviewer #2 (Public Review):

The authors are trying to distinguish between four models of the role of glypicans (HSPGs) on the Dpp/BMP gradient in the Drosophila wing, schematized in Fig. 1: (1) "Restricted diffusion" (HSPGs transport Dpp via repetitive interaction of HS chains with Dpp); (2) "Hindered diffusion" (HSPGs hinder Dpp spreading via reversible interaction of HS chains with Dpp); (3) "Stabilization" (HSPGs stabilize Dpp on the cell surface via reversible interaction of HS chains with Dpp that antagonizes Tkv-mediated Dpp internalization); and (4) "Recycling" (HSPGs internalize and recycle Dpp).

To distinguish between these models, the authors generate new alleles for the glypicans Dally and Dally-like protein (Dlp) and for Dpp: a Dally knock-out allele, a Dally YFP-tagged allele, a Dally knock-out allele with 3HA-Dlp, a Dlp knock-out allele, a Dlp allele containing 3-HA tags, and a Dpp lacking the HS-interacting domain. Additionally, they use an OLLAS-tag Dpp (OLLAS being an epitope tag against which extremely high affinity antibodies exist). They examine OLLAS-Dpp or HA-Dpp distribution, phospho-Mad staining, adult wing size.

They find that over-expressed Dally - but not Dlp - expands Dpp distribution in the larval wing disc. They find that the Dally[KO] allele behaves like a Dally strong hypomorph Dally[MH32]. The Dally[KO] - but not the Dlp[KO] - caused reduced pMad in both anterior and posterior domains and reduced adult wing size (particularly in the Anterior-Posterior axis). These defects can be substantially corrected by supplying an endogenously tagged YFP-tagged Dally. By contrast, they were not rescued when a 3xHA Dlp was inserted in the Dally locus. These results support their conclusion that Dpp interacts with Dally but not Dlp.

They next wanted to determine the relative contributions of the Dally core or the HS chains to the Dpp distribution. To test this, they over-expressed UAS-Dally or UAS-Dally[deltaHS] (lacking the HS chains) in the dorsal wing. Dally[deltaHS] over-expression increased the distribution of OLLAS-Dpp but caused a reduction in pMad. Then they write that after they normalize for expression levels, they find that Dally[deltaHS] only mildly reduces pMad and this result indicates a major contribution of the Dally core protein to Dpp stability.

Thanks for the comments. We actually showed that compared with Dally overexpression, Dally[deltaHS] overexpression only mildly reduces extracellular Dpp accumulation (Fig. 4I). This indicates a major contribution of the Dally core protein to interaction with Dpp, although the interaction is not sufficient to sustain extracellular Dpp distribution and signaling gradient.

The "normalization" is a key part of this model and is not mentioned how the normalization was done. When they do the critical experiment, making the Dally[deltaHS] allele, they find that loss of the HS chains is nearly as severe as total loss of Dally (i.e., Dally[KO]). Additionally, experimental approaches are needed here to prove the role of the Dally core.

Since the expression level of Dally[deltaHS] is higher than Dally when overexpressed, we normalized extracellular Dpp distribution (a-Ollas staining) against GFP fluorescent signal (Dally or Dally[deltaHS]). To do this, we first extracted both signal along the A-P axis from the same ROI in the previous version. The ratio was calculated by dividing the intensity of a-Ollas staining with the intensity of GFP fluorescent signal at a given position x. The average profile from each normalized profile was generated and plotted using the script described in the method (wingdisc_comparison.py) as other pMad or extracellular staining profiles.

Although this analysis provides normalized extracellular Dpp accumulation at different positions along the A-P axis, we are more interested in the total amount of Dpp or DppDeltaN accumulation upon Dally or dallyDeltaHS expression. Therefore, in the revised ms, we decided to normalize total amount of extracellular Dpp against the level of Dally or Dally[deltaHS] by dividing total signal intensity of extracellular Dpp staining (ExOllas staining) by total GFP fluorescent signal (Dally or Dally[deltaHS]) around the Dpp producing cells in each wing disc. Statistical analysis showed that accumulation of extracellular Dpp is only slightly reduced without HS chains (Fig.4I), indicating that Dally interacts with Dpp mainly through its core protein.

We agree with the reviewer that additional experimental approaches are needed to address the role of the core protein of Dally. As we discussed in the response to the reviewer1, to understand the importance of the interaction of core protein of Dally with Dpp, it is important to identify a region responsible for the interaction. Our preliminary results overexpressing a dally mutant lacking the majority of core protein (but keeping the HS modified region intact) showed that HS chains modification was also lost. Although this is consistent with our results that enzymes adding HS chains also interact with the core protein of Dally (Fig. 4D), the dally mutant allele lacking the core protein would hamper us from distinguishing the role of the core protein of Dally from HS chains.

Nevertheless, we can infer the importance of the interaction of core protein of Dally with Dpp using dally[3xHA-dlp, attP] allele, where dlp is expressed in dally expressing cells. Since Dally-like is modified by HS chains but does not interact with Dpp (Fig. 2, 4), dally[3xHA-dlp, attP] allele mimics a dally allele where HS chains are properly added but interaction of core protein with Dpp is lost. As we showed in Fig.3O, S, the allele could not rescue dallyKO phenotypes, consistent with the idea that interaction of core protein of Dally with Dpp is essential for Dpp distribution and signaling.

Prior work has shown that a stretch of 7 amino acids in the Dpp N-terminal domain is required to interact with heparin but not with Dpp receptors (Akiyama, 2008). The authors generated an HA-tagged Dpp allele lacking these residues (HA-dpp[deltaN]). It is an embryonic lethal allele, but they can get some animals to survive to larval stages if they also supply a transgene called “JAX” containing dpp regulatory sequences. In the JAX; HA-dpp[deltaN] mutant background, they find that the distribution and signaling of this Dpp molecule is largely normal. While over-expressed Dally can increase the distribution of HA-dpp[deltaN], over-expression of Dally[deltaHS] cannot. These latter results support the model that the HS chains in Dally are required for Dpp function but not because of a direct interaction with Dpp.

Our overexpression assays actually showed that both Dally and Dally[deltaHS] can accumulate Dpp upon overexpression and the accumulation of Dpp is comparable after normalization (Fig. 5C), consistent with the idea that interaction of DppdeltaN and HS chains are largely lost. As the reviewer pointed out, these results support the model that the HS chains in Dally are required for Dpp function but not because of a direct interaction with Dpp.

In the last part of the results, they attempt to determine if the Dpp receptor Thickveins (Tkv) is required for Dally-HS chains interaction. The 2008 (Akiyama) model posits that Tkv activates pMad downstream of Dpp and also internalizes and degrades Dpp. A 2022 (Romanova-Michaelides) model proposes that Dally (not Tkv) internalizes Dpp.

To distinguish between these models, the authors deplete Tkv from the dorsal compartment of the wing disc and found that extracellular Dpp increased and expanded in that domain. These results support the model that Tkv is required to internalize Dpp.

They then tested the model that Dally antagonizes Tkv-mediated Dpp internalization by determining whether the defective extracellular Dpp distribution in Dally[KO] mutants could be rescued by depleting Tkv. Extracellular Dpp did increase in the D vs V compartment, potentially providing some support for their model. However, there are no statistics performed, which is needed for full confidence in the results. The lack of statistics is particularly problematic (1) when they state that extracellular Dpp does not rise in ap>tkv RNAi vs ap>tkv RNAi, dally[KO] wing discs (Fig. 6E) or (2) when they state that extracellular Dpp gradient expanded in the dorsal compartment when tkv was dorsally depleted in dally[deltaHS] mutants (Fig. 6I). These last two experiments are important for their model but the differences are assessed only visually. In fact, extracellular Dpp in ap>tkv RNAi, dally[KO] (Fig. 6B) appears to be lower than extracellular Dpp in ap>tkv RNAi (Fig. 6A) and the histogram of Dpp in ap>tkv RNAi, dally[KO] is actually a bit lower than Dpp in ap>tkv RNAi, But the author claim that there is no difference between the two. Their conclusion would be strengthened by statistical analyses of the two lines.

We provided statistics for all the quantifications for pMad and extracellular Dpp distribution as supplementary data. In the previous version, we argued that extracellular Dpp level in ap>tkvRNAi, dallyKO (Fig.6B) does not increase compared with that in ap>tkvRNAi (Fig.6A). Statistical analysis (t-test) showed that the extracellular Dpp level in Fig. 6B is similar to or lower than that in Fig. 6A (Fig. 6E), confirming our conclusion. Statistical analysis (t-test) also confirmed that extracellular Dpp distribution expanded when tkv was knocked down in dallyHS mutants (Fig. 6I).

Strengths:

1. New genomically-engineered alleles

A considerable strength of the study is the generation and characterization of new Dally, Dlp and Dpp alleles. These reagents will be of great use to the field.

Thanks. We hope that these resources are indeed useful to the field.

1. Surveying multiple phenotypes

The authors survey numerous parameters (Dpp distribution, Dpp signaling (pMad) and adult wing phenotypes) which provides many points of analysis.

Thanks!

Weaknesses:

1. Confusing discussion regarding the Dally core vs HS in Dpp stability. They don't provide any measurements or information on how they "normalize" for the level of Dally vs Dally[deltaHS]? This is important part of their model that currently is not supported by any measurements.

We explained how we normalized in the above section and updated the method section in the revised ms.

1. Lacking quantifications and statistical analyses:

a. Why are statistical significance for histograms (pMad and Dpp distribution) not supplied? These histograms provide the key results supporting the authors' conclusions but no statistical tests/results are presented. This is a pervasive shortcoming in the current study.

Thanks. We provided t-test analyses together with the raw data as supplementary data.

b. dpp[deltaN] with JAX transgene - it would strengthen the study to supply quantitative data on the percent survival/lethal stage of dpp[deltaN] mutants with or without the JAK transgene

In this study, we are interested in the role of dpp[deltaN] during the wing disc development. Therefore, we decided not to perform the detailed analysis on the percent survival/lethal stage of dpp[deltaN] mutants with or without the JAX transgene in the current study. Nevertheless, the fact that dpp[deltaN] allele is maintained with a balanced stock and JAX;dpp[deltaN] allele can be maintained as homozygous stock indicates that the lethality of dpp[deltaN] allele comes from the early stages. Indeed, our preliminary results showed that pMad signal is severely lost in the dpp[deltaN] embryo without JAX (data not shown), indicating that the allele is lethal at early embryonic stages.

c. The graphs on wing size etc should start at zero.

Thanks. We corrected this in the current ms.

d. The sizes of histograms and graphs in each figure should be increased so that the reader can properly assess them. Currently, they are very small.

Thanks. We changed the sizes in the current ms.

The authors' model is that Dally (not Dlp) is required for Dpp distribution and signaling but that this is not due to a direct interaction with Dpp. Rather, they posit that Dally-HS antagonize Tkv-mediated Dpp internalization. Currently the results of the experiments could be considered consistent with their model, but as noted above, the lack of statistical analyses of some parameters is a weakness.

Thanks. We now performed and provided the statistical analyses in the revised ms.

One problematic part of their result for me is the role of the Dally core protein (Fig. 7B). There is a mis-match between the over-expression results and Dally allele lacking HS (but containing the core). Finally, their results support the idea that one or more as-yet unidentified proteins interact with Dally-HS chains to control Dpp distribution and signaling in the wing disc.

Our results simply suggest that Dpp can interact with Dally mainly through core protein but this interaction is not sufficient to sustain extracellular Dpp gradient formation under physiological conditions (dallyDeltaHS) (Fig. 4Q). We find that the mis-match is not problematic if the role of Dally is not simply mediated through interaction with Dpp. We speculate that interaction of Dpp and core protein of Dally is transient and not sufficient to sustain the Dpp gradient without HS chains of Dally stabilizing extracellular Dpp distribution by blocking Tkv-mediated Dpp internalization.

There is much debate and controversy in the Dpp morphogen field. The generation of new, high quality alleles in this study will be useful to Drosophila community, and the results of this study support the concept that Tkv but not Dally regulate Dpp internalization. Thus the work could be impactful and fuel new debates among morphogen researchers.

Thanks.

The manuscript is currently written in a manner that really is only accessible to researchers who work on the Dpp gradient. It would be very helpful for the authors to re-write the manuscript and carefully explain in each section of the results (1) the exact question that will be asked, (2) the prior work on the topic, (3) the precise experiment that will be done, and (4) the predicted results. This would make the study more accessible to developmental biologists outside of the morphogen gradient and Drosophila communities.

Thanks. We modified texts and changed the order of Fig.5. We hope that the changes make this study more accessible to developmental biologists outside of the field.

Author Response

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

We thank the reviewers for their feedback. Our response and a summary of the changes made to the manuscript are shown below. In addition to the changes made in response to the reviewer’s comments, we made the following changes to improve the manuscript:

• We updated figures 8 and 9 using data with improved preprocessing and source reconstruction. We now also include graphical network plots. This helps in the cross method (figure 8 vs 9) and cross dataset (figure 9 vs 10) comparison.

• We added funding acknowledgments and a credit author statement.

Reviewer #1 (Public Review):

Summary:

These types of analyses use many underlying assumptions about the data, which are not easy to verify. Hence, one way to test how the algorithm is performing in a task is to study its performance on synthetic data in which the properties of the variable of interest can be apriori fixed. For example, for burst detection, synthetic data can be generated by injected bursts of known durations, and checking if the algorithm is able to pick it up. Burst detection is difficult in the spectral domain since direct spectral estimators have high variance (see Subhash Chandran et al., 2018, J Neurophysiol). Therefore, detected burst lengths are typically much lower than injected burst lengths (see Figure 3). This problem can be solved by doing burst estimation in the time domain itself, for example, using Matching Pursuit (MP). I think the approach presented in this paper would also work since this model is also trained on data in the time domain. Indeed, the synthetic data can be made more "challenging" by injecting multiple oscillatory bursts that are overlapping in time, for which a greedy approach like MP may fail. It would be very interesting to test whether this method can "keep up" as the data is made more challenging. While showing results from brain signals directly (e.g., Figure 7) is nice, it will be even more impactful if it is backed up with results obtained from synthetic data with known properties.

We completely agree with the reviewer that testing the methods using synthetic data is an important part of validating such an approach. Each of the original papers that apply these methods to a particular application do this. The focus of this manuscript is to present a toolbox for applying these methods rather than to introduce/validate the methods themselves. For a detailed validation of the methods, the reader should see the citations. For example, the following paper introduces the HMM as a method for oscillatory burst detection:

• A.J. Quinn, et al. “Unpacking transient event dynamics in electrophysiological power spectra”. Brain topography 32.6 (2019): 1020-1034. See figures 2 and 3 for an evaluation of the HMM’s performance in detecting single-channel bursts using synthetic data.

We have added text to paragraph 2 in section 2.5 to clarify this burst detection method has been validated using simulated data and added references.

I was wondering about what kind of "synthetic data" could be used for the results shown in Figure 8-12 but could not come up with a good answer. Perhaps data in which different sensory systems are activated (visual versus auditory) or sensory versus movement epochs are compared to see if the activation maps change as expected. We see similarities between states across multiple runs (reproducibility analysis) and across tasks (e.g. Figure 8 vs 9) and even methods (Figure 8 vs 10), which is great. However, we should also expect the emergence of new modes specific to sensory activation (say auditory cortex for an auditory task). This will allow us to independently check the performance of this method.

The following papers study the performance of the HMM and DyNeMo in detecting networks using synthetic data:

• D. Vidaurre, et al. “Spectrally resolved fast transient brain states in electrophysiological data”. Neuroimage 126 (2016): 81-95. See figure 3 in this paper for an evaluation of the HMM’s performance in detecting oscillatory networks using simulation data.

• C. Gohil, et al. “Mixtures of large-scale dynamic functional brain network modes”. Neuroimage 263 (2022): 119595. See figures 4 and 5 for an evaluation of DyNeMo performance in detecting overlapping networks and long-range temporal structure in the data.

We have added text to paragraph 2 in section 2.5 to clarify these methods have been well tested on simulated data and added references.

The authors should explain the reproducibility results (variational free energy and best run analysis) in the Results section itself, to better orient the reader on what to look for.

Considering the second reviewer’s comments, we moved the reproducibility results to the supplementary information (SI). This means the reproducibility results are no longer part of the main figures/text. However, we have added some text to help the reader understand what aspects indicate the results are reproducible in section 2 of the SI.

Page 15: the comparison across subjects is interesting, but it is not clear why sensory-motor areas show a difference and the mean lifetime of the visual network decreases. Can you please explain this better? The promised discussion in section 3.5 can be expanded as well.

It is well known that the frequency and amplitude of neuronal oscillations changes with age. E.g. see the following review: Ishii, Ryouhei, et al. "Healthy and pathological brain aging: from the perspective of oscillations, functional connectivity, and signal complexity." Neuropsychobiology 75.4 (2018): 151-161. We observe older people have more beta activity and less alpha activity. These changes are seen in time-averaged calculations, i.e. the amplitude of oscillations are calculated using the entire time series for each subject.

The dynamic analysis presented in the paper provides further insight into how changes in the time-averaged quantities can occur through changes in the dynamics of frequency-specific networks. The sensorimotor network, which is a network with high beta activity, has a higher fractional occupancy. This indicates the change we observe in time-average beta power may be due to a longer amount of time spent in the sensorimotor network. The visual network, which is a network with high alpha activity, shows reduced lifetimes, which can explain the reduced time-averaged alpha activity seen with ageing.

We hope the improved text in the last paragraph of section 3.5 clarifies this. It should also be taken into account that the focus of this manuscript is the tools rather than an in-depth analysis of ageing. We use the age effect as an example of the potential analysis this toolbox enables.

Reviewer #2 (Public Review):

Summary:

The authors have developed a comprehensive set of tools to describe dynamics within a single time-series or across multiple time-series. The motivation is to better understand interacting networks within the human brain. The time-series used here are from direct estimates of the brain's electrical activity; however, the tools have been used with other metrics of brain function and would be applicable to many other fields.

Strengths:

The methods described are principled, and based on generative probabilistic models.

This makes them compact descriptors of the complex time-frequency data.

Few initial assumptions are necessary in order to reveal this compact description.

The methods are well described and demonstrated within multiple peer-reviewed articles.

This toolbox will be a great asset to the brain imaging community.

Weaknesses:

The only question I had was how to objectively/quantitatively compare different network models. This is possibly easily addressed by the authors.

We thank the reviewer for his/her comments. We address the weaknesses in our response in the “Recommendations For The Authors” section.

Reviewer #1 (Recommendations For The Authors):

Figure 2 legend: Please add the acronym for LCMV also.

We have now done this.

Section 2.5.1 page 8: the pipeline is shown in Figure 4, not 3.

This has been fixed.

Reviewer #2 (Recommendations For The Authors):

This is a great paper outlining a resource that can be applied to many different fields. I have relatively minor comments apart from one.

How does one quantitatively compare network descriptors (from DyNeMo and TDE-HMM for example)? At the moment the word 'cleaner' (P17) is used, but is there any non-subjective way? (eg Free energy/ cross validation etc). At the moment it is useful that one method gives a larger effect size (in a comparison between groups).. but could the authors say something about the use of these methods as more/less faithful descriptors of the data? Or in other words, do all methods generate datasets (from the latent space) that can be quantitatively compared with the original data?

In principle, the variational free energy could be used to compare models. However, because we use an approximate variational free energy (an exact measure is not attainable) for DyNeMo and an exact free energy for the HMM, it is possible that any differences we see in the variational free energy between the HMM and Dynemo are caused by the errors in its approximation. This makes it unreliable for comparing across models. That said, we can still use the variational free energy to compare within models. Indeed, we use the variational free energy for quantitative model comparisons when we select the best run to analyse from a set of 10.

One viable approach for comparing models is to assess their performance on downstream tasks. In this manuscript, examples of downstream tasks are the evoked network response and the young vs old group difference. We argue a better performance in the downstream task indicates a more useful model within that context. This performance is a quantitative measure. Note, there is no straightforward answer to which is the best model. It is likely different models will be useful for different downstream tasks.

In terms of which model provides a more faithful description of the data. The more flexible generative model for DyNeMo means it will generate more realistic data. However, this doesn’t necessarily mean it’s the best model (for a particular downstream task). Both the HMM and DyNeMo provide complementary descriptions that can be useful.

We have clarified the above in paragraph 5 of section 4.

Other comments:

• Footnote 6 - training on concatenated group data seems to be important. It could be more useful in the main manuscript where the limitations of this could be discussed.

By concatenating the data across subjects, we learn a group-level model. By doing this, we pool information across all subjects to estimate the networks. This can lead to more robust estimates. We have moved this footnote to the main text in paragraph 1 of section 2.5 and added further information.

• In the TDE burst detection section- please expand on why/how a specific number of states was chosen.

As with the HMM dynamic network analysis, the number of states must be pre-specified. For burst detection, we are often interested in an on/off type segmentation, which can be achieved with a 2 state HMM. However, if there are multiple burst types, these will all be combined into a single ‘on’ state. Therefore, we might want to increase the number of states to model multiple burst types. 3 was chosen as a trade-off to stay close to the on/off description but allow the model to learn more than 1 burst type. We have added text discussing this in paragraph 4 of section 4.

• Normally the value of free energy is just a function of the data - and only relative magnitude is important. I think figures (eg 7c) would be clearer if the offset could be removed.

We agree only the relative magnitude is important. We added text clarifying this in section 2 of the SI. We think it would still be worthwhile to include the offset so that future users can be sure they have correctly trained a model and calculated the free energy.

• Related to the above- there are large differences in model evidence shown between sets. Yet all sets are the same data, and all parameter estimates are more or less the same. Could the authors account for this please (i.e. is there some other parameter that differentiates the best model in one set from the other sets, or is the free energy estimate a bit variable).

We would like to clarify only the model parameters for the best run are shown in the group-level analysis. This is the run with the lowest variational free energy, which is highlighted in red. We have now clarified this in the caption of each figure. The difference in free energy for the best runs (across sets) is relatively small compared to the variation across runs within a set. If we were to plot the model parameters for each of the 10 runs in a set, we would see more variability. We have now clarified this in section 2 of the SI.

Also note, the group analysis usually involves taking an average. Small differences in the variational free energy could reflect small differences in subject-specific model parameters, which are then averaged out, giving virtually identical group effects.

• And related once again, if the data are always the same, I wonder if the free-energy plots and identical parameter estimates could be removed to free up space in figures?

The reproducibility results have now been moved to the supplementary information (SI).

• When citing p-values please specify how they are corrected (and over what please eg over states, nodes, etc?). This would be useful didactically as I imagine most users will follow the format of the presentation in this paper.

We now include in the caption further details of how the permutation significance testing was done.

• Not sure of the value of tiny power maps in 9C. Would consider making it larger or removing it?

The scale of these power maps is identical to part (A.I). We have moved the reproducibility analysis to the SI, enlarged the figure and added colour bars. We hope the values are now legible.

• Figure 3. I think the embedding in the caption doesn't match the figure (+-5 vs +-7 lags). Would be useful to add in the units of covariance (cii).

The number of embeddings in the caption has been fixed. Regarding the units for the covariances, as this is simulated data there aren’t really any units. Note, there is already a colour bar to indicate the values of each element.

• Minimize variational free energy - it may be confusing for some readers that other groups maximize the negative free energy. Maybe a footnote?

We thank the reviewer for their suggestion. We have added a footnote (1).

• Final question- and related to the Magnetoencephalography (MEG) data presented. These data are projected into source space using a beamformer algorithm (with its own implicit assumptions and vulnerabilities). Would be interested in the authors' opinion on what is standing between this work and a complete generative model of the MEG data - i.e. starting with cortical electrical current sources with interactions modeled and a dynamic environmental noise model (i.e. packing all assumptions into one model)?

In principle, there is nothing preventing us from including the forward model in the generative model and training on sensor level MEG data. This would be a generative model starting from the dipoles inside the brain to the MEG sensors. This is under active research. If the reviewer is referring to a biophysical model for brain activity, the main barrier for this is the inference of model parameters. However, note that the new inference framework presented in the DyNeMo paper (Gohil, et al. 2022) actually makes this more feasible. Given the scope of this manuscript is to present a toolbox for studying dynamics with existing methods, we leave this topic as future work.

Author Response

We are delighted that the reviewers found our work to have merit and we are thankful for their careful reviews and suggestions for experiments and changes to the text to further improve this study.

Author Response

We would like to express our thorough gratitude to the editors and reviewers, for the helpful comments and valuable suggestions, which provided us an opportunity to further address our research. Prior to submitting our final revision, here we provide our preliminary responses for the comments. Please find our detailed responses to the reviewers’ recommendations below.

Reviewer #1 (Public Review):

Summary:

This study examines the spatial and temporal patterns of occurrence and the interspecific associations within a terrestrial mammalian community along human disturbance gradients. They conclude that human activity leads to a higher incidence of positive associations.

Strengths:

The theoretical framework of the study is brilliantly introduced. Solid data and sound methodology. This study is based on an extensive series of camera trap data. Good review of the literature on this topic.

Weaknesses:

The authors use the terms associations and interactions interchangeably.

Response: This is not the case. In fact, we state specifically that "... interspecific associations should not be directly interpreted as a signal of biotic interactions between pairs of species…" However, co-occurrence can be an important predictor of likely interactions, such as competition and predation. We stand by our original text.

It is not clear what the authors mean by "associations". A brief clarification would be helpful.

Response: Our specific definition of what is meant here by spatial association can be found in the Methods section. To clarify, the calculation of the index of associations is based on the covariance for the two species of the residuals (epsilon) after consideration of all species-specific response to known environmental covariates. These covariances are modelled to allow them to vary with the level of human disturbance, measured as human presence and human modification. After normalization, the final index of association is a correlation value that varies between -1 (complete disassociation) and +1 (complete positive association).

Also, the authors do not delve into the different types of association found in the study. A more ecological perspective explaining why certain species tend to exhibit negative associations and why others show the opposite pattern (and thus, can be used as indicator species) is missing.

Response: Suggesting the ecological underpinnings of the associations observed here would mainly be speculation at this point, but the associations demonstrated in this analysis do suggest promising areas for the more detailed research suggested.

Also, the authors do not distinguish between significant (true) non-random associations and random associations. In my opinion, associations are those in which two species co-occur more or less than expected by chance. This is not well addressed in the present version of the manuscript.

Response: Results were considered to be non-random if correlation coefficients (for spatial association) or overlap (for temporal association) fell outside of 95% Confidence Intervals. This is now stated clearly in the Methods section. In Supplementary Figures S2 and S3, p<0.01 levels are also presented.

The obtained results support the conclusions of the study.

Anthropogenic pressures can shape species associations by increasing spatial and temporal co-occurrence, but above a certain threshold, the positive influence of human activity in terms of species associations could be reverted. This study can stimulate further work in this direction.

Reviewer #2 (Public Review):

Summary:

This study analyses camera trapping information on the occurrence of forest mammals along a gradient of human modification of the environment. The key hypotheses are that human disturbance squeezes wildlife into a smaller area or their activity into only part of the day, leading to increased co-occurrence under modification. The method used is joint species distribution modelling (JSDM).

Strengths:

The data source seems to be very nice, although since very little information is presented, this is hard to be sure of. Also, the JSDM approach is, in principle, a nice way of simultaneously analysing the data.

Weaknesses:

The manuscript suffers from a mismatch of hypotheses and methods at two different levels.

1. At the lower level, we first need to understand what the individual species do and "like" (their environmental niche). That information is not presented, and the methods suggest that the representation of each species in the JSDM is likely to be extremely poor.

Response: The response of each species to the environmental covariates provides a window into their environmental niche, encapsulated in the beta coefficients for each environmental covariate. This information is presented in Figure 2.

1. The hypothesis clearly asks for an analysis of the statistical interaction between human disturbance and co-occurrence. Yet, the model is not set up this way, and the authors thus do a lot of indirect exploration, rather than direct hypothesis testing.

Response: Our JSDM model is set up specifically to examine the effect of human disturbance on co-occurrence, after controlling for shared responses to environmental variables. It directly tests the first hypothesis, since, if increase in indices of human disturbance had not tended to increase the measured spatial correlations between species as detected by the model, we would have rejected our stated hypothesis that human modification of habitats results in increased positive spatial associations between species.

Even when the focus is not the individual species, but rather their association, we need to formulate what the expectation is. The hypotheses point towards presenting the spatial and the temporal niche, and how it changes, species for species, under human disturbance. To this, one can then add the layer of interspecific associations.

Response: Examining each species one by one and how each one responds to human disturbance would miss the effects of any meaningful interactions between species. The analysis presented provides a means to highlight associations that would have been overlooked. Future research could go on to analyze the strongest associations in the community and the strongest effects of human disturbance so as to uncover the underlying interactions that give rise to them and the mechanisms of human impact. We believe that this will prove to be a much more productive approach than trying to tackle this problem species by species and pair by pair.

The change in activity and space use can be analysed much simpler, by looking at the activity times and spatial distribution directly. It remains unclear what the contribution of the JSDM is, unless it is able to represent this activity and spatial information, and put it in a testable interaction with human disturbance.

The topic is actually rather complicated. If biotic interactions change along the disturbance gradient, then observed data are already the outcome of such changed interactions. We thus cannot use the data to infer them! But we can show, for each species, that the habitat preferences change along the disturbance gradient - or not, as the case may be.

Then, in the next step, one would have to formulate specific hypotheses about which species are likely to change their associations more, and which less (based e.g. on predator-prey or competitive interactions). The data and analyses presented do not answer any of these issues.

Response: We suggest that the so-called “simpler” approach described above is anything but simple, and this is precisely what the Joint Species Distribution Model improves upon. As pointed out in the Introduction, simply examining spatial overlap is not enough to detect a signal of meaningful biotic interaction, since overlap could be the result of similar responses to environmental variables. With the JSDM approach, this would not be considered a positive association and would then not imply the possible existence of meaningful interaction.

Another more substantial point is that, according to my understanding of the methods, the per-species models are very inappropriate: the predictors are only linear, and there are no statistical interactions (L374). There is no conceivable species in the world whose niche would be described by such an oversimplified model.

Response: While interaction terms can be included in the JSDM, this would considerably increase the complexity of the models. In previous work, we have found no strong evidence for the importance of interaction terms and they do not improve the performance of the models.

We have no idea of even the most basic characteristics of the per-species models: prevalences, coefficient estimates, D2 of the model, and analysis of the temporal and spatial autocorrelation of the residuals, although they form the basis for the association analysis!

Response: The coefficient estimates for response to environmental variables used in the JSDM are provided in Figure 2.

Why are times of day and day of the year not included as predictors IN INTERACTION with niche predictors and human disturbance, since they represent the temporal dimension on which niches are hypothesised to change?

Also, all correlations among species should be shown for the raw data and for the model residuals: how much does that actually change and can thus be explained by the niche models?

The discussion has little to add to the results. The complexity of the challenge (understanding a community-level response after accounting for species-level responses) is not met, and instead substantial room is given to general statements of how important this line of research is. I failed to see any advance in ecological understanding at the community level.

Response: We agree that the community-level response to human disturbance is a complex topic, and we believe it is also a very important one. This research and its support of the spatial compression hypothesis, while not providing definitive answers to detailed mechanisms, opens up new lines of inquiry that makes it an important advance. For example, the strong effects of human disturbance on certain associations that were detected here could now be examined with the kind of detailed species by species and pair by pair analysis that this reviewer appears to demand.

Author Response

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

Reviewer #1 (Public Review):

The manuscript has helped address a long-standing mystery in splicing regulation: whether splicing occurs co- or post-transcriptionally. Specifically, the authors (1) uniquely combined smFISH, expansion microscopy, and live cell imaging; (2) revealed the ordering and spatial distribution of splicing steps; and (3) discovered that nascent, not-yet-spliced transcripts move more slowly around the transcription site and undergo splicing as they move through the clouds. Based on the experimental results, the authors suggest that the observation of co-transcriptional splicing in previous literature could be due to the limitation of imaging resolution, meaning that the observed co-transcriptional splicing might actually be post-transcriptional splicing occurring in proximity to the transcription site. Overall, the work presented here clearly provides a comprehensive picture of splicing regulation.

Major points:

1. Linearity of expansion microscopy. For Figure 2B, it would be helpful to display the same sample before and after expansion, just like Supplementary Figure 3, but with a transcription site and "cloud". In the current version, the transcription site looks quite different in the not-expanded (more green dots on the left) and expanded image (more green dots on the top).

We thank the reviewer for this comment on linearity of expansion. Based on our prior manuscript (Chen et al 2015 Nature Methods. PMID: 27376770), we expect expansion microscopy to yield isotropic expansion. Indeed, as shown in Supplemental Figure 3, we confirmed that expansion of nuclei (3B, top) and transcripts (3B, bottom) is isotropic. Additionally, before splicing inhibition, we demonstrated the linearity of expansion for a transcription site (3B, left), shown at standard resolution with intron stain. The images shown in Figure 2B are meant solely to illustrate the change in resolution upon expansion, and are not meant to imply spatial matching between the expanded and unexpanded image. We apologize for the confusion and have clarified this in the figure legend for Figure 2.

We also point the reader towards Supplemental Figure 4, in which we validate the use of expansion microscopy in these findings. We show that transcription sites in expanded samples were the same size as those imaged using stochastic optical reconstruction microscopy (STORM), demonstrating that expansion did not significantly alter the morphology of the site.

1. FISH dot colocalization. What is the colocalization rate of FISH dots in general under experimental conditions? In addition, in Figures 2C and 2G, why do some 3'exon dots not have co-localized 5'exon dots?

We thank the reviewer for asking for these important clarifications. Under standard (non-expanded) conditions, our colocalization of 3’ and 5’ spots varies by gene, but more than 75% of intron spots colocalize with exon spots for the vast majority of transcripts we evaluated. The percentage of colocalization for each gene and intron can be found in column 4 of Table 1.

Regarding the second point—these individual images may not reflect the actual quantitative number of spot counts at the site, as these transcription sites have a sizable Z dimension that is difficult to capture in one image, and certain dyes are more easily visually distinguished in contrasted images than others. These factors may cause some 3’ spots to appear without a corresponding co-localized 5’ spot in these images. We refer the reviewer to Supplemental Figure 4C for quantitative spot counting of an expanded transcription site, for which there are a similar number of 3’ end and 5’ end spots within the entire Z-stacked image. Importantly, these transcription site clouds contain longer, unspliced transcripts, potentially leading to further separation between the 5’ and 3’ ends of a single transcript when compared to a cytoplasmic, spliced transcript (quantified in Figure 2I).

1. It would be helpful if the authors uploaded a few examples of live cell imaging movies.

Certainly! Please refer to the new Supplementary Movies 1-3 for representative examples of live cell imaging data.

1. It is recommended to double-check the text for errors.

We apologize for errors in the original manuscript, and have made the appropriate corrections.

Reviewer #2 (Public Review):

Allison Coté et al. investigated the ordering and spatial distribution of nascent transcripts in several cells using smFISH, expansion microscopy, and live-cell imaging. They find that pre-mRNA splicing occurs post-transcriptionally at the clouds around the transcription start site, termed the transcription site proximal zone. They show that pre-mRNA may undergo continuous splicing when they pass through the zone after transcription. These data suggest a unifying model for explaining previously reported co-transcriptional splicing events and provide a direction for further study of the nature of the slow-moving zone around the transcription start site.

This paper is well-written. The findings are very important, and the data supports the conclusions well. However, some aspects of the image and description need to be clarified and revised.

The authors describe Figure 4E and 4F results in the main text as that "we performed RNA FISH simultaneously with immunofluorescence for SC35, a component of speckles, and saw that this compartmentalized pre-mRNA did indeed appear near nuclear speckles both before (Supplementary Figure 6C) and after (Figure 4E) splicing inhibition." However, no SC35 staining is shown in the Figure 4E. A similar situation happened in describing Figure 4F.

We thank the reviewer for noting this error. We mistakenly called in text for Figure 4E, when we meant to refer to Figure 4G, which shows combined RNA FISH and SC35 immunofluorescence show compartmentalization within nuclear speckles. Figures 4E and 4F do not show SC35 immunofluorescence. We have altered the text and figure captions accordingly. Recommendations for the authors: please note that you control which revisions to undertake from the public reviews and recommendations for the authors Reviewer #1 (Recommendations For The Authors):

Minor points:

1. For Figures, it would be better to mark co-transcriptional and proximal post-transcriptional splicing in a clearer way. Like in Figure 1A, the simulated RNA FISH signals are almost identical across two conditions, which is a bit confusing. Overlapping and close proximity shall be better illustrated in related figures.

We thank the reviewer for these suggestions. We have iterated these figures through multiple revisions and have found that these diagrams tend to resonate the most, so we have elected to keep them as is, but we do appreciate the suggestion.

1. May include some details of expansion microscopy in the last paragraph of the Introduction. For example, why introduce expansion microscopy? To what level it can help overcome the diffraction limit?

We thank the reviewer for this comment, and have added additional text to this paragraph to further set up the use of expansion microscopy.

1. Double-check the formatting. Some sub-titles are in Bold, some in Italic.

We apologize for any formatting errors, and have made the appropriate corrections.

1. Please double-check the writing. I find many incompatible parts across the manuscript. For example, as described in the Figure 1D caption, there aren't "first" and "second" graphs in the figure. Moreover, some writings require additional refinement. For instance, in the Introduction part, the paragraph discussing RNA imaging, various techniques (such as FISH and live imaging), and concerns (such as microscopy resolution, chromatin fraction, and limitations related to reporter genes) are intertwined without clear indexing or logical structuring. Similar cases in other paragraphs too. Last but not least, I can even find repetitive sentences across the manuscript. For instance, I believe that the authors forgot to delete "By distinguishing the separate fluorescent signals from probes bound to exons and introns, we could visualize splicing intermediates (represented by colocalized intron and exon spots) relative to the site of transcription (represented by bright colocalized intron and exon spots) and fully spliced products (represented by exon spots alone)." in the first paragraph of the Results part, as the exact same sentence re-occurs right after. I've only listed a few examples here. Please refine the manuscript.

We apologize for any errors in the original manuscript, and have made the appropriate corrections.

Reviewer #2 (Recommendations For The Authors):

1. The sentence "By distinguishing the separate fluorescent signals from probes bound to exons and introns, we could visualize splicing intermediates (represented by colocalized intron and exon spots) relative to the site of transcription (represented by bright colocalized intron and exon spots) and fully spliced products (represented by exon spots alone)." is accidentally repeated twice, one of them should be deleted.

We apologize for this duplication, and have made the appropriate correction.

Author Response

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

We thank the two reviewers very much for their careful review and valuable comments. Upon these comments, the following revisions have been made. First, we have performed a new analysis on human accelerated regions (HARs) recently reported by the Zoonomia Project. Second, we have presented more data on experimentally detected and computationally predicted DBSs of MALAT1, NEAT1, and MEG3. Third, we have added details on the RNA-seq data processing and subsequent differential expression testing to the Materials and Methods section. Fourth, we have clarified some details on the human ancestor sequence and the use of parameters and thresholds. Six new citations are added. In addition, we have also carefully polished the main text. We hope these revisions, together with the Responses-to-Reviewers, would help the reader better get the information from the paper.

eLife assessment

In this valuable manuscript, the authors attempt to examine the role of long non-coding RNAs (lncRNAs) in human evolution, through a set of population genetics and functional genomics analyses that leverage existing datasets and tools. Although the methods are at times inadequate - for example, suitable methods and/or relevant controls are lacking at many points, and selection is inferred sometimes too quickly - the results nonetheless point towards a possible contribution of long non-coding RNAs to the evolution of human biology and they suggest clear directions for future, more rigorous study.

Public Reviews:

Reviewer #1 (Public Review):

Summary

While DNA sequence divergence, differential expression, and differential methylation analysis have been conducted between humans and the great apes to study changes that "make us human", the role of lncRNAs and their impact on the human genome and biology has not been fully explored. In this study, the authors computationally predict HSlncRNAs as well as their DNA Binding sites using a method they have developed previously and then examine these predicted regions with different types of enrichment analyses. Broadly, the analysis is straightforward and after identifying these regions/HSlncRNAs the authors examined their effects using different external datasets.

Strengths/weaknesses

By and large, the analysis performed is dependent on their ability to identify HSlncRNAs and their DBS. I think that they have done a good job of showing the performance metrics of their methods in previous publications. Thereafter, they perform a series of enrichment-type analyses that have been used in the field for quite a while now to look at tissue-specific enrichment, or region-specific enrichment, or functional enrichment, and I think these have been carried out well. The authors achieved the aims of their work. I think one of the biggest contributions that this paper brings to the field is their annotation of these HSlncRNAs. Thus a major revisionary effort could be spent on applying their method to the latest genomes that have been released so that the community could get a clean annotation of newly identified HSlncRNAs (see comment 2).

Comments

1. Though some of their results about certain HSlncRNAs having DBSs in all genes is rather surprising/suspicious, I think that broadly their process to identify and validate DBSs is robust, they have multiple lines of checks to identify such regions, including functional validation. These predictions are bound to have some level of false positive/negative rate and it might be nice to restate those here and on what experiment/validation data these were conducted. However, the rest of their analysis comprises different types of enrichment analysis which shouldn't be affected by outlier HSlncRNAs if indeed their FPR/FNR are low.

2. There are now several new genomes available as part of the Zoonomia consortium and 240 Primate consortium papers released. These papers have re-examined some annotations such as Human Accelerated Regions (HARs) and found with a larger dataset as well as better reference genomes, that a large fraction of HARs were actually incorrectly annotated - that is that they were also seen in other lineages outside of just the great apes. If these papers have not already examined HSlncRNAs, the authors should try and re-run the computational predictions with this updated set and then identify HSlncRNAs there. This might help to clarify their signal and remove lncRNAs that might be present in other primates but are somehow missing in the great apes. This might also help to mitigate some results that they see in section 3 of their paper in comparing DBS distances between archaics and humans.

Responses:

(1) Thanks for the good suggestion. We have checked the Zoonomia reported genomes and found that new primate genomes are monkeys and lemurs but not apes (Zoonomia Consortium. Nature 2023. https://doi.org/10.1038/s41586-020-2876-6), and the phylogenetic relationships between monkeys and humans are much more remote than those between apes and humans. In addition, the Zoonomia project did target identifying new lncRNA genes.

(2) We have examined the Zoonomia-reported HARs (Keough et al. Science 2023. DOI: 10.1126/science.abm1696). Of the 312 HARs reported by Keough et al, 8 overlap 26 DBSs of 14 HS lncRNAs; moreover, DBSs greatly outnumber HARs, suggesting that HAR and DBS are different sequences with different functions.

(3) In the revised manuscript, a new paragraph (the second one) has been added to the section “HS lncRNAs regulate diverse genes and transcripts” to describe the HAR analysis result.

1. The differences between the archaic hominins in their DBS distances to modern humans are a bit concerning. At some level, we expect these to be roughly similar when examining African modern humans and perhaps the Denisovan being larger when examining Europeans and Asians, but they seem to have distances that aren't expected given the demography. In addition, from their text for section 3, they begin by stating that they are computing two types of distances but then I lost track of which distance they were discussing in paragraph 3 of section 3. Explicitly stating which of the two distances in the text would be helpful for the reader.

Responses:

(1) Upon the archaic human genomes, the genomic distances from the three modern humans are shorter to Denisovan than to Altai Neanderthal; however, upon the related studies we cite, the phylogenetic relationship between the three modern humans is more remote to Denisovan than to Altai Neanderthal. Thus, the finding that 2514 and 1256 DBSs have distances >0.034 in Denisovans and Altai Neanderthals is not unreasonable. The numbers of DBSs, of course, depend on the cutoff of 0.034, which is somewhat subjective but not unreasonable.

(2) The second paragraph is added to the Discussion, discussing parameters and cutoffs.

(3) Regarding the two types of distance, the distances computed in the first way were not further analyzed because, as we note, “This anomaly may be caused by that the human ancestor was built using six primates without archaic humans”.

1. Isn't the correct control to examine whether eQTLs are more enriched in HSlncRNA DBSs a set of transcription factor binding sites? I don't think using just promoter regions is a reasonable control here. This does not take away from the broader point however that eQTLs are found in DBSs and I think they can perform this alternate test.

Responses:

Indeed, TFBSs are more comparable to DBSs than promoters. However, many more methods have been developed to predict TFBSs than to predict DBSs, making us concerned about TFBS prediction's reliability. Since most QTLs in DBSs are mQTLs (Supplementary Table 13), but many QTLs in TFBSs are eQTLs (Flynn et al. PLoS Genetics 2021. DOI: 10.1371/journal.pgen.1009719), it is pretty safe to conclude that DBSs are enriched in mQTLs.

1. In the Discussion, they highlight the evolution of sugar intake, which I'm not sure is appropriate. This comes not from GO enrichment but rather from a few genes that are found at the tail of their distribution. While these signals may be real, the evolution of traits is often highly polygenic and they don't see this signal in their functional enrichment. I suggest removing that line. Moreover, HSlncRNAs are ones that are unique across a much longer time frame than the transition to agriculture which is when sugar intake rose greatly. Thus, it's unlikely to see enrichment for something that arose in the past 6000-7000 years would in the annotation that is designed to detect human-chimp or human-neanderthal level divergence.

Responses:

(1) The Discussion on human adaptation to high sugar intake is based on both enriched GO terms (Supplementary Table 4, 7) and a set of genes in modern humans with the most SNP-rich DBSs (Table 2). These glucose-related GO terms are not at the tail of the list because, of the 614 enriched GO terms (enriched in genes with strongest DBSs), glucose metabolism-related ones are ranked 208, 212, 246, 264, 504, 522, 591, and of the 409 enriched GO terms (enriched in the top 1256 genes in Altai Neanderthals), glucose metabolism-related ones are ranked 152 and 217.

(2) Indeed, there are other top-ranked enriched GO terms; some (e.g., neuron projection development (GO:0031175) and cell projection morphogenesis (GO:0048858)) have known impact on human evolution, but the impact of others (e.g., cell junction organization (GO:0034330)) remain unclear. We specifically report human adaptation to high sugar intake because the DBSs in related genes show differences in modern humans (Table 2).

Reviewer #2 (Public Review):

Lin et al attempt to examine the role of lncRNAs in human evolution in this manuscript. They apply a suite of population genetics and functional genomics analyses that leverage existing data sets and public tools, some of which were previously built by the authors, who clearly have experience with lncRNA binding prediction. However, I worry that there is a lack of suitable methods and/or relevant controls at many points and that the interpretation is too quick to infer selection. While I don't doubt that lnc RNAs contribute to the evolution of modern humans, and certainly agree that this is a question worth asking, I think this paper would benefit from a more rigorous approach to tackling it.

At this point, my suggestions are mostly focused on tightening and strengthening the methods; it is hard for me to predict the consequence of these changes on the results or their interpretation, but as a general rule I also encourage the authors to not over-interpret their conclusions in terms of what phenotype was selected for when as they do at certain points (eg glucose metabolism).

Responses:

(1) Now, we use more cautious wording to describe the results.

(2) A paragraph (the second one) is added to Discussion to explain parameters and cutoffs.

(3) We make the caution at the end of the third paragraph that “We note that these are findings instead of conclusions, and they indicate, suggest, or support something revealing the primary question of what genomic differences critically determine the phenotypic differences between humans and apes and between modern and archaic humans”.

I note some specific points that I think would benefit from more rigorous approaches, and suggest possible ways forward for these.

1. Much of this work is focused on comparing DNA binding domains in human-unique long-noncoding RNAs and DNA binding sites across the promoters of genes in the human genome, and I think the authors can afford to be a bit more methodical/selective in their processing and filtering steps here. The article begins by searching for orthologues of human lncRNAs to arrive at a set of 66 human-specific lncRNAs, which are then characterised further through the rest of the manuscript. Line 99 describes a binding affinity metric used to separate strong DBS from weak DBS; the methods (line 432) describe this as being the product of the DBS or lncRNA length times the average Identity of the underlying TTSs. This multiplication, in fact, undoes the standardising value of averaging and introduces a clear relationship between the length of a region being tested and its overall score, which in turn is likely to bias all downstream inference, since a long lncRNA with poor average affinity can end up with a higher score than a short one with higher average affinity, and it's not quite clear to me what the biological interpretation of that should be. Why was this metric defined in this way?

Responses:

(1) Binding affinity and length of all DBSs of HS lncRNAs are given in Supplementary Table 2 and 3. Since a triplex (say, 100 bp in length) may have 50% or 70% of nucleotides bound, it is necessary to differentiate binding affinity and length, and the two measures can differentiate DBSs of the same length but with different binding affinity and DBSs with the same binding affinity but different length.

(2) Differentiating DBSs into strong and weak ones is somewhat subjective, accurately differentiating them demands experimental data that are currently unavailable, and it is advisable to separately analyze strong and weak DBSs because they may likely influence different aspects of human evolution.

1. There is also a strong assumption that identified sites will always be bound (line 100), which I disagree is well-supported by additional evidence (lines 109-125). The authors show that predicted NEAT1 and MALAT1 DBS overlap experimentally validated sites for NEAT1, MALAT1, and MEG3, but this is not done systematically, or genome-wide, so it's hard to know if the examples shown are representative, or a best-case scenario.

Responses:

(1) We do not assume/think that identified sites will always be bound. Instead, lncRNA/DBS binding is highly context-dependent (including tissue-specific).

(2) An extra supplementary table (Supplementary Table 15) is added to show what predicted DBSs overlap experimentally detected DBSs for NEAT1, MALAT1, and MEG3. By the way, it is more accurate to say “experimentally detected” than “experimentally validated”, because experimental data have true/false positives and true/false negatives, and different sequencing protocols (for detecting lncRNA/DNA binding) may generate somewhat different results.

It's also not quite clear how overlapping promoters or TSS are treated - are these collapsed into a single instance when calculating genome-wide significance? If, eg, a gene has five isoforms, and these differ in the 3' UTR but their promoter region contains a DBS, is this counted five times, or one? Since the interaction between the lncRNA and the DBS happens at the DNA level, it seems like not correcting for this uneven distribution of transcripts is likely to skew results, especially when testing against genome-wide distributions, eg in the results presented in sections 5 and 6. I do not think that comparing genes and transcripts putatively bound by the 40 HS lncRNAs to a random draw of 10,000 lncRNA/gene pairs drawn from the remaining ~13500 lncRNAs that are not HS is a fair comparison. Rather, it would be better to do many draws of 40 non-HS lncRNAs and determine an empirical null distribution that way, if possible actively controlling for the overall number of transcripts (also see the following point).

Responses:

(1) We analyzed each and every GENCODE-annotated transcript (Supplementary Table 2). For example, if a gene has N TSS and N transcripts, DBSs are predicted in N promoter regions. When analyzing gene expression in tissues, each and every transcript is analyzed.

(2) Ideally, it would be better to do many draws, but statistically, a huge number is needed due to the number of total genes in the human genome.

(3) We feel that doing many draws of 40 non-HS lncRNAs and determining an empirical null distribution is not as straightforward as comparing HS lncRNA-target transcript pairs (45% show significant expression correlation) with random lncRNA-random transcript pairs (2.3% show significant expression correlation).

1. Thresholds for statistical testing are not consistent, or always well justified. For instance, in line 142 GO testing is performed on the top 2000 genes (according to different rankings), but there's no description of the background regions used as controls anywhere, or of why 2000 genes were chosen as a good number to test? Why not 1000, or 500? Are the results overall robust to these (and other) thresholds? Then line 190 the threshold for downstream testing is now the top 20% of genes, etc. I am not opposed to different thresholds in principle, but they should be justified.

Responses:

(1) The over-representation analysis using g:Profiler was applied to the top and bottom 2000 genes with the whole genome as the background. The number “2000” was chosen somewhat subjectively. If more or fewer genes were chosen, more or fewer enriched GO terms would be identified, but GO terms with adjusted P-values <0.05 would be quite stable.

(2) A paragraph (the second one) is added to the Discussion to explain parameters and cutoffs.

Likewise, comparing Tajima's D values near promoters to genome-wide values is unfair, because promoters are known to be under strong evolutionary constraints relative to background regions; as such it is not surprising that the results of this comparison are significant. A fairer comparison would attempt to better match controls (eg to promoters without HS lncRNA DBS, which I realise may be nearly impossible), or generate empirical p-values via permutation or simulation.

Responses:

We examined Tajima’s D in DBSs (Supplementary Figure 9) and in HS lncRNA genes (Supplementary Figure 18). We compared the Tajima’s D values with the genome-wide background in both cases.

1. There are huge differences in the comparisons between the Vindija and Altai Neanderthal genomes that to me suggest some sort of technical bias or the such is at play here. e.g. line 190 reports 1256 genes to have a high distance between the Altai Neanderthal and modern humans, but only 134 Vindija genes reach the same cutoff of 0.034. The temporal separation between the two specimens does not seem sufficient to explain this difference, nor the difference between the Altai Denisovan and Neanderthal results (2514 genes for Denisovan), which makes me wonder if it is a technical artefact relating to the quality of the genome builds? It would be worth checking.

Responses:

(1) The cutoff of 0.034 was chosen upon that DBSs in the top 20% (4248) genes in chimpanzees have distances larger than this cutoff, and accordingly, 4248, 1256, 2514, and 134 genes have DBSs distances >0.034 in chimpanzees, Altai Neanderthals, Denisovans, and Vindija Neanderthals. These numbers of genes qualitatively agree with the phylogenetic distances from chimpanzees, archaic humans to modern humans. If a percentage larger or smaller than 20% (e.g., 10% or 30%) is chosen, and so is a cutoff X, the numbers of genes with DBSs distance >X would not be 4248, 1256, 2514, and 134, but could still qualitatively agree with the phylogenetic distances from chimpanzees, archaic humans to modern humans.

(2) The second paragraph in the Discussion now explains the parameters and cutoffs.

1. Inferring evolution: There are some points of the manuscript where the authors are quick to infer positive selection. I would caution that GTEx contains a lot of different brain tissues, thus finding a brain eQTL is a lot easier than finding a liver eQTL, just because there are more opportunities for it. Likewise, claims in the text and in Tables 1 and 2 about the evolutionary pressures underlying specific genes should be more carefully stated. The same is true when the authors observe high Fst between groups (line 515), which is only one possible cause of high Fst - population differentiation and drift are just as capable of giving rise to it, especially at small sample sizes.

Responses:

(1) We analyzed brain tissues separately instead of taking the whole brain as a tissue, see Supplementary Table 12 and Figure 3.

(2) We make the caution at the end of the third paragraph that “We note that these are findings instead of conclusions, and they indicate, suggest, or support something revealing the primary question of what genomic differences critically determine the phenotypic differences between humans and apes and between modern and archaic humans”.

Reviewer #1 (Recommendations For The Authors):

Some figures are impossible to see/read so I wasn't able to evaluate them - Fig, 1B, 1E, 1F are small and blurry.

Responses:

High-quality figures are provided.

Typo in line 178: in these archaic humans, the distances of HS lncRNAs are smaller than the distances of DBSs.

Responses:

This is not a typo. We use “distance per base” to measure whether HS lncRNAs or their DBSs have evolved more from archaic humans to modern humans. See also Supplementary Note 4 and 5.

Reviewer #2 (Recommendations For The Authors):

1. There's some inconsistency in the genome builds and the database versions used, eg, sometimes panTro4 is used and sometimes panTro5 (line 456). Likewise, the version of GENCODE used is very old (18), the current version is 43. The current version contains 19928 lncRNAs, which is a big difference relative to what is being tested!

Responses:

(1) panTro4 was used to search orthologues of human lncRNAs; this time-consuming work started several years ago when the version of GENCODE was V18 (see Lin et al., 2019).

(2) Regarding “the version of GENCODE used is very old (V18)”, we have later examined the 4396 human lncRNAs reported in GENCODE V36 and found that the set of 66 HS lncRNAs remains the same.

(3) The counterparts of HS lncRNAs’ DBSs in chimpanzees were predicted recently using panTro5.

1. Table 1: What does 'mostly' mean in this context? I understand that it refers to sequence differences between humans and the other genomes, but what is the actual threshold, and how is it defined?

Responses:

The title of Table 1 is “Genes with strongest DBSs and mostly changed sequence distances from modern humans to archaic humans and chimpanzees”. Instead of using two cutoffs, choosing genes with the two features seems easy and sensible.

1. Line 117: The methods do not include information on the RNA-seq data processing and subsequent DE testing.

Responses:

The details are added to the section “Experimentally validating DBS predictiom” (The reads were aligned to the human GRCh38 genome using Hiasat2 (Kim et al., 2019), and the resulting sam files were converted to bam files using Samtools (Li et al., 2009). Stringtie was used to quantify gene expression level (Pertea et al., 2015). Fold change of gene expression was computed using the edgeR package (Robinson et al., 2010), and significant up- and down-regulation of target genes after DBD knockout was determined upon |log2(fold change)| > 1 with FDR < 0.1).

1. Line 180: I looked at the EPO alignment and it's not clear to me what 'human ancestor' means, but it may well explain the issues the authors have with calculating distances (I agree those numbers are weird). Is it the reconstructed ancestral state of humans at around 300-200,000 years ago (coalescence of most human uniparental lineages), or the inferred sequence of the human-chimpanzee most recent common ancestor? If it's the former, it's not surprising it skews results towards shorter distances for modern humans, since the tree distance from that point to archaic hominins is significantly larger than to modern humans.

Responses:

The “human ancestor” is constructed by the EBI team upon the genomes of six primates in the Ensembl website. We find that the reconstructed ancestral state of humans may be unlikely around 300,000-200,000 years, and may be much earlier. We also find that many DNA sequences of the “human ancestor” are low-confidence calls (i.e., the ancestral states are supported by only one primate’s sequence).

1. Line 221: SNP-rich DBS: Is this claim controlled for the length of the DBS?

Responses:

No. Long DBSs tend to have more SNPs. When comparing the same DBS in modern humans, archaic humans, and chimpanzees, both the length and SNP number reflect evolution, so it is not necessary to control for the length.

1. Given that GTEx is primarily built off short-read data and it is impossible to link binding of a lncRNA to a DBS with its impact with a specific transcript

Responses:

As written in the section “Examining the tissue-specific impact of HS lncRNA-regulated gene expression”, we calculated the pairwise Spearman's correlation coefficient between the expression of an HS lncRNA (the representative transcript, median TPM value > 0.1) and the expression of each of its target transcripts (median TPM value > 0.1) using the scipy.stats.spearmanr program in the scipy package. The expression of an HS lncRNA gene and a target transcript was considered to be significantly correlated if the |Spearman's rho| > 0.3, with Benjamini-Hochberg FDR < 0.05.

1. Line 429: should TTO be TFO?

Responses:

Here TTO should be TFO; the typo is corrected.

1. Methods, section 7: Some of the text in this section should perhaps be moved to the results section?

Responses:

Each of the two paragraphs in Methods’ section 7 is quite large, and some contents in Supplementary Notes are also very relevant. Thus, moving them to the Results section could make the Results too lengthy and specific.

1. Line 587: GTEx is built from samples of primarily European ancestry and has poor representation of African ancestry and negligible representation of Asian ancestry (see the GTEx v8 paper supplement). This means that it is basically impossible to find a non-European population-specific eQTL in GTEx, which in turn impacts these results.

Responses:

(1) Indeed, this is a serious issue of data analysis, and this issue cannot be solved until more Africans are sequenced.

(2) Anyway, one can still find considerable African-specific eQTLs in GTEx, such as rs28540058 (with frequency of 0, 0, 0.13 in CEU, CHB, YRI) and rs58772997 (with frequency of 0, 0, 0.12 in CEU, CHB, YRI (see Supplementary Table12 and Supplementary Figure 22).

Author Response

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

eLife assessment

The finding that Fusicoccin (FC-A) promotes locomotor recovery after spinal cord injury is useful, and the idea of harnessing small molecules that may affect protein-protein interactions to promote axon regeneration is interesting and worthy of study. However, the main methods, data, and analyses are inadequate to support the primary claim of the manuscript that a 14-3-3-Spastin complex is necessary for the observed FC-A effects.

Response: We appreciate the eLife editorial and review team for consideration and evaluation of our manuscript. In light of the feedback from the editors and reviewers, we recognize that certain aspects of the title and key conclusions require further refinement. We have shown that 14-3-3, through its interaction with phosphorylated spastin, inhibits the degradation of spastin. Also, we have demonstrated that 14-3-3 can enhance spastin's microtubule-severing ability in cell lines. Furthermore, our work has illustrated the significant roles of 14-3-3 and spastin in the repair process of spinal cord injury. However, there is currently insufficient direct evidence to confirm the cooperation between 14-3-3 and spastin during axon regeneration and the recovery of spinal cord injury. Moreover, we have not provided conclusive evidence of their simultaneous action in injured axons, mediating changes in microtubule dynamics. Consequently, we have re-evaluated the manuscript's title and primary conclusions, and have made relevant modifications. For more detailed information, please refer to the reviewer's comments.

Public Reviews:

Reviewer #1 (Public Review):

The present work establishes 14-3-3 proteins as binding partners of spastin and suggests that this binding is positively regulated by phosphorylation of spastin. The authors show evidence that 14-3-3 - spastin binding prevents spastin ubiquitination and final proteasomal degradation, thus increasing the availability of spastin. The authors measured microtubule severing activity in cell lines and axon regeneration and outgrowth as a prompt to spastin activity. By using drugs and peptides that separately inhibit 14-3-3 binding or spastin activity, they show that both proteins are necessary for axon regeneration in cell culture and in vivo models in rats.

The following is an account of the major strengths and weaknesses of the methods and results.

Major strengths

-The authors performed pulldown assays on spinal cord lysates using GST-spastin, then analyzed pulldowns via mass spectrometry and found 3 peptides common to various forms of 14-3-3 proteins. In co-expression experiments in cell lines, recombinant spastin co-precipitated with all 6 forms of 14-3-3 tested. The authors could also co-immunoprecipitate spastin-14-3-3 complexes from spinal cord samples and from primary neuronal cultures.

-By protein truncation experiments they found that the Microtubule Binding Domain of spastin contained the binding capability to 14-3-3. This domain contained a putative phosphorylation site, and substitutions that cannot be phosphorylated cannot bind to 14-3-3.

-Overexpression of GFP-spastin shows a turn-over of about 12 hours when protein synthesis is inhibited by cycloheximide. When 14-3-3 is co-overexpressed, GFP-spastin does not show a decrease by 12 hours. When S233A is expressed, a turn-over of 9 hours is observed, suggesting that phosphorylation increases the stability of the protein. In support of that notion, the phospho-mimetic S233D makes it more stable, lasting as much as the over-expression of 14-3-3.

-By combining FCA with Spastazoline, authors claim that FCA increased regeneration is due to increased spastin activity in various models of neurite outgrowth and regeneration in cell culture and in vivo, the authors show impressive results on the positive effect of FCA in regeneration, and that this is abolished when spastin is inhibited.

Major weaknesses

1. The present manuscript suggests that 14-3-3 and spastin work in the same pathway to promote regeneration. Although the manuscript contains valuable evidence in support for a role of 14-3-3 and spasting in regeneration, the conclusive evidence is difficult to generate, and is missing in the present manuscript. For example, there are simpler explanations for the combined effect of FC-A and spastazoline. The FC-A mechanism of action can be very broad, since it will increase the binding of all 14-3-3 proteins with presumably all their substrates, hence the pathways affected can rise to the hundreds. The fact that spastazoline abolishes FC-A effect, may not be because of their direct interaction, but because spastin is a necessary component of the execution of the regeneration machinery further downstream, in line with the fact that spastazoline alone prevented outgrowth and regeneration, and in agreement with previous work showing that normal spastin activity is necessary for regeneration.

With this in mind, I consider the title and most major conclusions of the manuscript related to these two proteins acting together for the observed effects are overstated.

Response: We appreciate and acknowledge the reviewers' considerations. Our results demonstrated that the spastin inhibitor, spastazolin, almost completely inhibited axon regeneration and the spinal cord injury repair process. This, in turn, leads to the disappearance of any promoting effect on spinal cord injury repair when spastin function is compromised. While we have provided evidence that the expression levels of spastin are moderately increased at the injury site in mice after treatment with FC-A following spinal cord injury, the conclusion that FC-A promotes spinal cord injury repair through the direct interaction between 14-3-3 and spastin still lacks direct evidence. Therefore, we have made appropriate modifications to the manuscript's title and main conclusions.

1. Authors show that S233D increases MT severing activity, and explain that it is related to increased binding to 14-3-3. An alternative explanation is that phosphorylation at S233 by itself could increase MT severing activity. The authors could test if purified spastin S233D alone could have more potent enzymatic activity.

Response: We appreciate the considerations of the reviewer. We believe that supplementing in vitro experiments to assess whether S233D affects spastin's microtubule severing function can more intuitively demonstrate whether phosphorylation of spastin at S233 affects its microtubule severing function; however, spastin forms hexamers through its AAA domain to exert ATPase activity and cut microtubules. Current research has reported that mutation sites leading to changes in microtubule severing function are mainly located within spastin's AAA domain (affecting spastin's ATPase activity, amino acids 342-599), such as E356A, G370R, N386K, K388R, E442Q, K427R and R562Q. Furthermore, studies have shown that mutating 11 phosphorylation sites in spastin's MIT and MTBD regions to alanine does not affect spastin's microtubule severing function, including human S268 (Rat Ser233) (Phosphorylation mutation impairs the promoting effect of spastin on neurite outgrowth without affecting its microtubule severing ability. Eur J Histochem. doi: 10.4081/ejh.2023.3594). Additionally, we also provided supplementary experiments in cell lines which showed that both spastin S233A and S233D could effectively sever microtubules (Fig.S2).

1. The interpretation of the authors cannot explain how Spastin can engage in MT severing while bound to 14-3-3 using its Microtubule Binding Domain.

Response: We appreciate the considerations of the expert reviewer. The IP experiments with truncated fragments suggest that the binding region of 14-3-3 with spastin is located within the region (215-336 amino acids) in spastin. Furthermore, experiments involving site-directed mutagenesis confirm that the actual binding site of 14-3-3 with spastin is the S233 site, rather than its MTBD region (270-328). Therefore, we have made corrections in the manuscript. We also indicate that 14-3-3 enhances spastin's protein levels by binding to the S233 site, which may be due to 14-3-3 masking the ubiquitination sites near spastin S233 (K206 or K254). Our further experiments also demonstrate that 14-3-3 inhibits the ubiquitination degradation pathway of phosphorylated spastin.

1. Also, the term "microtubule dynamics", which is present in the title and in other major conclusions, is overstated. Although authors show, in cell lines, changes in microtubule content, it is far from evidence for changes in "MT dynamics" in the settings of interest (i.e. injured axons).

Response: We appreciate and acknowledge the rigorous feedback. While our manuscript demonstrated the regulatory role of 14-3-3 and spastin in microtubule dynamics in cell lines, we lack direct evidence of these changes in microtubule dynamics within injured axons. Therefore, we have made appropriate modifications to the title, main conclusions, and related statements in our manuscript.

1. In the same lines, the manuscript lacks evidence for the changes of MT content and/dynamics as a function of the proposed 14-3-3 - Spastin pathway.

Response: We appreciate and concur with the opinions of the expert reviewer. The observed changes in microtubule dynamics in spinal cord injury were related to the overall alterations in microtubule dynamics within the spinal cord injury site. We still lack direct evidence that 14-3-3, in conjunction with spastin, alters the microtubule dynamics within axons during the process of regeneration. Therefore, we have made modifications to the manuscript.

Reviewer #2 (Public Review):

Summary:

The idea of harnessing small molecules that may affect protein-protein interactions to promote axon regeneration is interesting and worthy of study. In this manuscript Liu et al. explore a 14-3-3-Spastin complex and its role in axon regeneration.

Strengths:

Some of the effects of FC-A on locomotor recovery after spinal cord contusion look interesting

Weaknesses:

The manuscript falls short of establishing that a 14-3-3-Spastin complex is important for any FC-A-dependent effects and there are several issues with data quality that make it difficult to interpret the results. Importantly, the effects of the spastin inhibitor has a major impact on neurite outgrowth suggesting that cells simply cannot grow in the presence of the inhibitor and raising serious questions about any selectivity for FC-A - dependent growth. Aspects of the histology following spinal cord injury were not convincing.

Response: We appreciate the rigorous review by the expert reviewers. In response to the feedback from reviewer 1, we lack direct evidence to demonstrate that the reparative effect of FC-A on spinal cord injury is mediated by the combined action of 14-3-3 and spastin. We have accordingly made the necessary changes to our manuscript. Additionally, due to upload limitations, the resolution of our tissue slices related to spinal cord injury in the manuscript is relatively low. To address this, we have supplemented relevant images which was enlarged in the supplementary materials (Fig. S7-9), Also, the original confocal files and images were uploaded.

Furthermore, our manuscript does not suggest that the reparative effect of FC-A in spinal cord injury selectively impacts the interaction between 14-3-3 and spastin. Therefore, we have modified our claims (title and conclusions) to ensure a more precise statement. Despite the fact that our axonal markers do not fully align, our evidence still strongly supports the role of FC-A in promoting nerve regeneration after spinal cord injury. Additionally, we will further optimize our immunohistochemistry methods.

Reviewer #3 (Public Review):

Summary:

The current manuscript shows that 14-3-3 are binding partners of spastin, preventing its degradation. It is additionally shown, using complementary methods, that both 14-3-3 and spastin are necessary for axon regeneration in vitro and in vivo. While interesting in vitro and vivo data is provided, some of the claims of the authors are not convincingly supported.

Major strengths:

Very interesting effect of FC-A in functional recovery after spinal cord injury.

Major Weaknesses:

Some of the in vitro data, including colocalizations, and analysis of microtubule severing fall short to support the claims of the authors.

The in vivo selectivity of FC-A towards spastin is not adequately supported by the data presented. There are aspects of the spinal cord injury site histology that are unclear.

Response: Reviewer 3's comments align with those of Reviewers 1 and 2.

Reviewer #1 (Recommendations For The Authors):

-The new blots presented in Fig. 3N lacks corresponding labels as for antibodies used for IP and IB and molecular weight markers.

Response: We appreciate the reviewer's feedback. We have made the corresponding modifications in the figure.

Reviewer #2 (Recommendations For The Authors):

The authors have addressed many of the specific concerns shared with the authors in the first round of review but several issues remain with the manuscript.

1. Fig. 1D - the interpretation that spastin co-localizes with 14-3-3 proteins in hippocampal neurons is still tenuous since 14-3-3 uniformly labels the cell.

Response: We appreciate the reviewer's consideration. Upon re-examining the source files, we found that the predominant reason for 14-3-3 showing a ubiquitous cellular distribution was excessive brightness and insufficient contrast. After appropriate adjustments, we discovered that 14-3-3 exhibits characteristic distribution in axons, including aggregation at growth cone and specific locations in the axon shaft. We have made the relevant changes in the revised version.

1. Line 336. The meaning of the following statement is unclear "To further identify which isoform of 14-3-3 interacts with spastin, we generated six 14-3-3 isoforms in rats (β、γ、ε、ζ、η、θ ), then purified GST fusion 14-3-3 proteins (Figure 1G).

Response: Sorry for any confusing statement. We obtained gene fragments of six 14-3-3 isoforms from rat brain cDNA and inserted these fragments into the pEGX-5X-3 vector. Subsequently, GST 14-3-3 fusion proteins were expressed and purified in vitro. We have made the corresponding revisions in the revised version.

1. Line 341. The authors still fall short of showing that spastin and 14-3-3 interact directly thus it may be more accurate to say that they form a complex.

Response: Thank you for the reviewer's advice. We have made the corresponding corrections in the manuscript.

1. Line 388. Please clarify 2th and the meaning of "moderately" - "S233D) was moderately expressed in primary hippocampal neurons at 2th DIV." While it is specified that the transfection dosage and duration were meticulously controlled - it is unclear what the criteria was for establishing the appropriate moderate dosage.

Response: Sorry about the mistake, it should be "2nd" instead of "2th". In order to establish a model for overexpressing spastin to promote neuronal neurite growth, we transfected 0.2 µg of plasmid into 1 well (1×104 cells/cm2, 24-well plate), with a transfection duration controlled at 24 hours.

1. Line 395 - It is unclear if S233D is toxic as there seem to be no measurements of cell survival.

Response: We have supplemented relevant experiments (See comment 6) based on comment 6 and found that Spastin S233D can promote neuronal neurite growth. The corresponding descriptions have been revised.

1. The pro-growth effects of S233A still does not seem to fit the narrative and the results would have been more convincing if dosage was better controlled to establish any differences between WT and S233A Spastin.

Response: We appreciate the constructive comments from the reviewer. In order to better illustrate the role of spastin S233 in neuronal growth, we have made appropriate adjustments to our experimental conditions based on previous experiments. Cells were transfected with plasmids expressing non-fused GFP and spastin and the relevant S233 mutants at a transfection dose of 0.2 µg into 1 well (1×104 cells/cm2, 24-well plate), duration was controlled at 12 hours. Due to the low expression state of the overexpressed protein, GFP (ab290 antibody for IF) was then stained to trace neuronal morphology. The experimental results demonstrate that spastin promotes neuronal neurite growth, and the dephosphorylation mutant of spastin (spastin S233A) significantly attenuates its neurite-promoting effect compared to wild-type spastin. Conversely, the phosphorylation mutant spastin S233D further enhances the promotion of neuronal neurite growth. We have also made corrections to the relevant statements in the manuscript.

1. The reason for examining protection in response to glutamate is not well rationalized based on known spastin functions. The interpretation of this experiment is unclear with respect to effects on protection vs repair.

Response: Thank you for the reviewer's consideration. We suppose that spastin may be involved in both protective and repair processes. Existing studies suggest that spastin can control store-operated calcium entry (SOCE) by altering endoplasmic reticulum morphology (doi: 10.1093/brain/awac122, doi: 10.3389/fphys.2019.01544), which may indicate its role in regulating calcium overload. Additionally, due to the critical role of spastin in axon growth, it is also essential for neuronal repair after injury. Therefore, we have not strictly distinguished between these two concepts here.

1. It is unclear if Spastazoline simply blocks any type of growth and it is thus difficult to conclude that FC-A functions through a 14-3-3-spastin effect based on the current data.

Response: We have re-evaluated and modified the title and main conclusions of the manuscript based on the reviewer's comments and the existing evidence, as responded to in reviewer 1's comments.

1. The access of FC-A to the CNS with the current protocol has not been clearly established and the effects of FC_A on spastin expression seem to mirror the profile of the control condition.

Response: We agree with the reviewer's comments. The expression trend of spastin after FC-A treatment is consistent with that of the control group, with a slight increase in its expression level compared to the control group.

1. The NF and 5-HT staining is not convincing labelling fibres.

Response: We appreciate the reviewer's comments. We believe that the reason for the incomplete axon staining is closely related to the thickness of the tissue sections. In our future research, we will further optimize our axon labeling methods.

Reviewer #3 (Recommendations For The Authors):

Figure 1D: Both spastin and 14-3-3 label the entire neuron which is rather unusual. Conditions of immunfluorescence should be improved. As it is, this image should not be used to claim colocalization.

Response: We appreciate the reviewer's consideration. In response to comment 1 from the expert reviewer 2, we have re-examined the source files and identified that the primary reason for the overall cell-wide distribution of 14-3-3 and spastin is due to excessive brightness and a lack of sufficient contrast. After making appropriate adjustments, we found that 14-3-3 and spastin exhibit characteristic localization within the axon (concentrated in a particular region of the axon shaft and the growth cone). We have made corresponding revisions in the revised version of the manuscript.

Figure S2: The experimental setup and data provided is not adequate to infer microtubule severing.

Response: We appreciate the reviewer's guidance. We have improved the relevant experiments and used a 100X objective lens to observe the microtubule structures more clearly.

Figure 2 I-K: The functional effect of spastin S233A and S233D on neurite outgrowth does not correlate with a function of 14-3-3 and thus does not support the central hypothesis of the manuscript. Minor: The images selected as representative show differences in neurite length and branching that are not portrayed in the graphs.

Response: Thank you for the reviewer’s comment. Similar to the response to the reviewer 2's comment 6, in order to better illustrate the role of spastin S233 in neurite outgrowth, we made corresponding adjustments to our experimental conditions. Cells were transfected with plasmids expressing non-fused GFP and spastin and the relevant S233 mutants at a transfection dose of 0.2 µg into 1 well (1×104 cells/cm2, 24-well plate), duration was controlled at 12 hours. Due to the low expression state of the overexpressed protein, GFP (ab290 antibody for IF) was then stained to trace neuronal morphology. The experimental results demonstrate that spastin promotes neuronal neurite growth, and the dephosphorylation mutant of spastin (spastin S233A) significantly attenuates its neurite-promoting effect compared to wild-type spastin. Conversely, the phosphorylation mutant spastin S233D further enhances the promotion of neuronal neurite growth. We have also made corrections to the relevant statements in the manuscript.

Figure 5 J and L: The quality, resolution and size of the images is insufficient to support the claims of the authors. As it is, one cannot interpret the data. It is very hard to envisage, even considering the explanation provided by the authors, that spinal cords where spastazoline was used correspond to contusion as a complete discontinuity between the rostral and caudal spinal cord tissue is present.

Response: Due to limitations in file uploads, we encountered issues with the resolution of the tissue slices related to spinal cord injury. To address this, we have adjusted the size and resolution of the corresponding images in the supplementary materials (Fig.S7-S9 ) and included the original confocal files and images.

Additionally, it's important to note that the tissue slices we presented do not represent all layers of the spinal cord, and not all layers exhibit discontinuity. Our slices are taken longitudinally at the dorsal site of the lesion area. The dorsal slices represent areas closer to the injury site, while deeper slices correspond to areas distant from the injury site. Therefore, we selected areas closer to the injury site to reflect the repair process following injury.

Figure 7B: Similar comment to spianl cord images provided in Figure 5. NF and MBP are not supposed to colocalize as they label different cell types...

Response: We appreciate the comments from the expert reviewer, and we agree with their suggestions. We will further optimize our axon labeling methods. The excessive brightness and lack of contrast primarily led to the non-specific labeling of other cell types with the MBP antibody. In fact, our primary goal was to highlight the injured areas by enhancing the fluorescence intensity of the images, which inadvertently resulted in neglecting the exclusion of non-specific staining. Therefore, we have made appropriate adjustments to the images to better visualize the distribution of myelin sheaths.

Author Response

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

Reviewer #1 (Public Review):

The importance of the role of sexual behavior, specifically ejaculation rates, is worth emphasizing for the formation of pair bonds in prairie voles. It suggests that the role of sexual behavior in contributing to the strength of pair bonds should be explored more. It is also important to add that males and females in the study were screened for sexual receptivity. It would therefore be important to identify characteristics of animals that did not mate under the laboratory conditions used that may add depth and complexity to what was identified in the current study. The identification of brain regions for pair bond maintenance centered around the amygdala was also intriguing.

Thank you for pointing some interpretations of our findings that can be emphasized in the Discussion. We added the following sentences to the Discussion:

“Our findings, along with this previous work, support the hypothesis that sexual behavior plays a key role in driving pair-bond strength. However, the current study focused on animals that were screened for sexual receptivity, which may have limited variation in sexual behavior across pairs. An intriguing direction for future research will be to test how this variation contributes to bond strength.”

We also emphasized amygdala in relation to pair bond maintenance. We added the following sentence to the Discussion:

“These brain regions, and especially amygdala, will be important candidates for future research on neural regulation of pair-bond maintenance.”

The issue of the lack of a strong presence of the reward circuitry (nucleus accumbens) in the final models is also worth more discussion. Perhaps it has been overly emphasized in the past, but there are strong results from other studies pointing to the importance of reward circuitry.

Thank you for this suggestion. There is a section in the Results that analyses accumbens in more detail than other brain regions. Accumbens did not survive our corrections for multiple statistical tests, however it was significant at early timepoint without these corrections. This Results paragraph states the following:

“Although the nucleus accumbens did not survive multiple test corrections in our ROI analysis (q=0.17), it was significant in univariate analysis (p=0.03), particularly when focused on the 2.5 and 6h timepoints (two sample t-test: t=2.53, p=0.01, Video 2). Furthermore, voxel-level comparisons revealed significant sites within the ventral striatum and the posterior nucleus accumbens (Figure 2A, Figure Supplement 1b-c, Video 2).”

We added Supplementary File 4, which contains model comparison results for accumbens and all other ROIs. We also added more detail on nucleus accumbens to the Discussion:

“Pairing drove increased c-Fos expression in the ventral pallidum, a major node in reward circuity, as well as in the paraventricular nucleus and the medial preoptic area, modulators of reward. This is consistent with a large body of work implicating neuropeptide actions on reward circuits in the formation of bonds (Walum & Young, 2018; Young & Wang, 2004). Conspicuously missing from our list, however, is significant pairing induced c-Fos induction in the nucleus accumbens. One possibility is that an absence of significant accumbens IEG induction reflects the limitations of using c-Fos and other immediate early genes as indicators of neural activity. It is known that some neuronal populations can be active without expressing c-Fos (Sheng & Greenberg, 1990). Indeed, although a variety of studies implicate the accumbens in bond formation (Amadei et al., 2017; Aragona et al., 2006; Scribner et al., 2020), previous work finds only weak c-Fos induction in the prairie vole accumbens during bonding (Curtis & Wang, 2003). Another possibility is that there was heterogeneous activation in the accumbens that was not captured by the precision of our atlas. Consistent with this interpretation, found that the accumbens was significant in univariate tests, as well as in voxel-level analyses. Overall, our results do not conflict with pharmacological, electrophysiological, and calcium-imaging data on the role of the nucleus accumbens in prairie vole bonding (Amadei et al., 2017; Aragona et al., 2006; Scribner et al., 2020). Instead, the absence of significant effects at the level of the entire nucleus accumbens together with the presence of anatomically restricted voxel-level significance suggests substantial anatomical heterogeneity in the contributions of the nucleus accumbens to bond formation.”

Please discuss the consequences of creating the behavioral data for pair bond formation by subtracting same-sex pairs interactions from the opposite-sex interactions. What sources of information are removed by using this approach?

One limitation of our study’s approach is that we are unable to fully separate information related to social novelty from mating experience. Thank you for pointing out that we should touch on this sort of caveat in the paper. We added several sentences to the Discussion:

“It seems likely that sensory and motor areas were important for social processes related to both pair-bonding and reunion with same-sex cagemates, such as investigation and recognition. Our study design, however, highlights differences between treatments, and in order to detect such effects, it might be necessary to compare mating and bonding pairs to animals left in complete isolation.”

We reiterate the point in a new paragraph we added to the Discussion to explicitly provide caveats regarding our data:

“Before offering a synthesis of our findings, it would be useful to acknowledge a few caveats. First, as noted above, IEG induction does not capture all relevant neural activity. Second, the design of our experiment, which controlled for social interaction, likely excluded many circuits important to both pair bonding and sibling social interactions. Third, c-Fos activity within a given brain region may nevertheless rely on distinct cell types, and so the absence of sex differences in c-Fos immunoreactivity does not definitively rule out the sexually dimorphic circuits hypothesized in the “dual function hypothesis” (de Vries, 2004). Lastly, the current study focused on animals that were screened for sexual receptivity, which may have limited the variation in sexual behavior across opposite-sex pairs.”

Time 0 is when the barrier is removed after a two-hour exposure. Please speculate on what is going on during the two-hour exposure. Time zero is potentially more than the time of mating. Is it possible that aggression is being decreased during this timepoint that represents mating? Could it also be a measure of the outcome of an initial compatibility assessment by the male and female?

Thank you for this interesting observation. While the opaque divider prevented physical social interactions, it is possible that animals picked up on auditory or olfactory cues. We did not detect group differences in movement patterns and vocalization rates from the 0 h timepoint group (Figure 2). These findings suggest that potential partner detection and assessment occurred in a similar way for both experiment groups. It is unlikely that this period represents a decrease in aggression, since unbonded prairie voles are not known to be aggressive towards conspecifics. However, the idea that animals may potentially use olfactory or auditory cues to assess each other is an interesting idea, and one that we cannot rule out. We added a brief statement to the Methods “Experiment Design” section about the possibility that the two hours prior to divider removal (0 h timepoint) could represent more than an acclimation period:

“It is important to note that the opaque divider in the acclimation period prevented physical interactions, but it is possible that animal pairs may have detected each other through olfactory or auditory cues.”

We also mention this in the revised Discussion in the context of the PFC cluster, which not only differed between mating and non-mating groups, but also showed differences between isolated (0h) and socially interacting animals (sibs and mates, 2.5h-22h):

“A fourth cluster (“PFC,” green) is composed of prelimbic, infralimbic and olfactory cortex; activity in the vole prefrontal cortex is known to be modulated by hypothalamic oxytocin, and to shape bonding through projections to the nucleus accumbens (Amadei et al., 2017; Burkett et al., 2016; Horie et al., 2020). The pattern of activity in this cluster, however, indicates that it was due in part to differences between the isolated animals (0h) and other time points (Figure 4—figure supplement 1 and Figure 4—figure supplement 2). Because animals in the isolated condition were in a compartment adjacent to either an opposite sexed individual or a familiar former cagemate, we cannot rule out that olfactory or auditory cues may have made animals aware of the presence of a potential social partner. Indeed, we interpret this dimension as capturing appetitive aspects of behaviors associated with investigation of the animal isolated from the subject by the barrier.”

Reviewer #2 (Public Review):

An important caveat to this study not mentioned by the authors is that c-fos provides a snapshot of neural activity and that important populations of neurons could be active and not express c-fos. Thus observed correlations are likely to be robust, but the absence of differences (in say accumbens) may just reflect the limits of c-fos estimation of neural activity. Similarly, highly coordinated neural activity between males and females might still be driven by different mechanisms if different cell types were activated within a specific region.

We now discuss limitations of c-Fos in the Discussion paragraph that focuses on accumbens:

“The absence of significant accumbens IEG induction may reflect the limitations of using c-Fos and other immediate early genes as indicators of neural activity. It is known that some neuronal populations can be active without expressing c-Fos (Sheng & Greenberg, 1990). Indeed, although a variety of studies implicate the accumbens in bond formation (Amadei et al., 2017; Aragona et al., 2006; Scribner et al., 2020), previous work finds only weak c-Fos induction in the prairie vole accumbens during bonding (Curtis & Wang, 2003).”

We also include the following sentence in a new Discussion paragraph that focuses on caveats to our findings:

“Before offering a synthesis of our findings, it would be useful to acknowledge or reiterate a few caveats. First, as noted above, IEG induction does not capture all relevant neural activity (Sheng & Greenberg 1990). Second, the design of our experiment, which controlled for social interaction, likely excluded many circuits important to both pair bonding and sibling social interactions. Third, c-Fos activity within a given brain region may nevertheless rely on distinct cell types, and so the absence of sex differences in c-Fos immunoreactivity does not definitively rule out the sexually dimorphic circuits hypothesized in the “dual function hypothesis” (de Vries, 2004). Lastly, the current study focused on animals that were screened for sexual receptivity, which may have limited the variation in sexual behavior across opposite-sex pairs.”

Recommendations for the authors:

It appears as if df is missing from some statistical reporting.

Thank you for pointing this out. We went through the manuscript and added in sample sizes to statistical reporting.

Reviewer #1 (Recommendations for the authors):

It is surprising that the cortex was not more extensively identified as being involved in pair bonding, but perhaps this is because the emphasis for choosing brain areas in the cortical region is biased towards olfactory regions. Please discuss. It may also be worth noting that brain regions associated with perception may be important in all of these processes, but selected out because of the design.

Thank you for this observation. We agree that some cortical regions may not have been identified due to the study design. For example, social processes related to both pair bonding and cagemate recognition likely rely on overlapping circuits. It is also important to note here that our analysis approach identified the “most” significant regions. This means that several candidate regions did not survive the statistical threshold used to select regions. We now discuss the cortex in more detail in the Discussion, where we also identify the regions that approached significance but did not survive multiple test corrections:

“Although the PFC and other olfactory cortical areas formed a cluster, we did not find widespread c-Fos induction throughout the cortex in response to pairing. It seems likely that sensory and motor areas were important for social processes related to both pair-bonding and reunion with same-sex cagemates, such as investigation and recognition. Our study design, however, highlights differences between treatments, and in order to detect such effects, it might be necessary to compare mating and bonding pairs to animals left in complete isolation. Moreover, several cortical regions that did not survive corrections for multiple tests may have been identified in a less stringent analysis. Several subregions within the isocortex, hippocampal formation, and cortical subplate had statistical models that approached significance (i.e., p-values < 0.1) prior to multiple test corrections. These subregions were found within primary somatosensory area, primary auditory area, dorsal and ventral auditory areas, primary visual area, anteromedial visual area, agranular insular area, temporal association areas, ectorhinal area, postsubiculum, and basomedial amygdala. Frontal cortex subregions were within the agranular insular area and orbital area, as well as additional subregions in prelimbic and infralimbic areas of the PFC.”

Same-sex siblings were isolated for 4-5 days and then repaired. This is a creative way of dealing with this, but was any aggression displayed in the same-sex pairs? Are there bonds or preferences among same-sex individuals? Could the isolation have set the stage for neural changes associated with migrating from the natal group? 4-5 days of isolation is not trivial.

Thank you for these questions. We did not witness aggression between same-sex pairs. We had recorded ‘aggression’ events (lunges and chases) during the 1 h behavioral observation epochs and found that these rates were nearly zero for all sibling timepoint groups (events/h per focal animal in mean ± sd: 2.5 h group = 0.58 ± 1.53, 6 h group = 0.17 ± 0.48, 22 h group = 0.25 ± 0.44).

The question about peer relationships is a good one. Previous literature does suggest that prairie voles can develop preferences for familiar same-sex individuals (e.g., Beery et al. 2018 Front. Behav. Neuro., Lee et al. 2019 Front. Behav. Neuro). Thus, we want to reiterate here that our study design tests for differences between these baseline levels of affiliation with pair bonding in a reproductive context.

It is possible that the period of isolation prior to experiments may have set the stage for neural changes associated with migration from the natal group. Testing this possibility is outside the scope of the current study. We want to point out here that animals were separated from their natal groups several weeks prior to the experiment. Animals were weaned at 21 days and put into same-sex cages, and then experiments occurred between 8-12 weeks of age. All experiment groups went through the same weaning and co-housing conditions.

Pg 26, Line 655: "better" is listed twice in the sentence and only one is needed

Thank you for catching this typo. This is fixed.

Reviewer #2 (Recommendations for the authors):

Why was it necessary to bring voles into estrus when they are induced ovulators? The authors need to state how voles were brought into estrus.

Thank you for this suggestion. We explained estrus induction in the Methods, but this explanation could be missed because it was within the “Behavioral procedures” section. We put the paragraph about estrus induction into a new section called “Estrus induction and animal selection”. We also elaborated on the final sentences of this paragraph to provide a clearer rationale:

“We used this mating assay to restrict study subjects to voles that showed lordosis (females) or mounting behavior (males). By selecting voles who showed sexual behavior, we could control the estrus state and timing of mating across the 0, 2.5, 6 and 22 h study groups. This selection process also ensured that animals assigned to the same-sex sibling pair and opposite-sex mating pair groups had similar sexual motivation and experience.”

I assume in the final manuscript the authors will release the availability of the atlas? Making the atlas public seems to be in the spirit of the eLife publishing model.

The prairie vole reference brain, atlas, and atlas annotation labels, are now included on the Figshare repository site. We updated the Data and code availability section to clarify this.

Reviewer #3 (Recommendations for the authors):

Please clarify in the Methods if same-sex sibling females were also estrogen primed. If not, could the estrogen exposure cause Fos differences?

Thank you for this suggestion. All females were estrogen primed. We refined the Methods section “Estrus induction and animal selection” to make this part of the study design clearer. We edited one of the sentences to say “During this isolation period, all females were induced into estrus[...]” We also added a couple sentences at the end of this paragraph:

“By selecting voles who showed sexual behavior, we could control the estrus state and timing of mating across the 0, 2.5, 6 and 22 h study groups. This selection process also ensured that animals assigned to the same-sex sibling pair and opposite-sex mating pair groups had similar sexual motivation and experience.”

Author Response:

We thank the reviewers for their careful comments. We sincerely agree with the comments from both reviewers, and noticed the word “cell transplantation”, throughout the manuscript including the title, was confusing. We will revise the manuscript to clarify the aim of the study, and to express the conclusion more straightforwardly.

Response to reviewers:

We interpret the data of the present study as the color of each RPE cell is a temporal condition which does not necessarily represent the quality (e.g. for cell transplantation) of the cells. We consider this may be applicable not only in vitro but also in vivo, although we do not know whether RPE shows heterogeneous level of pigmentation in vivo.

As our concern for iPSC-RPE is always about their quality for cell transplantation, maybe we haven’t fairly evaluated the scientific significance obtained from the present study.

Another thing we noticed was, although we used the term “cell transplantation” to explain what we meant by “quality” of the cells, we agree this was confusing. The aim of the study was not to show how the pigmentation level of transplant-RPE affects the result of cell transplantation, but to show the heterogeneous gene expression of iPSC-derived RPE cells, and the less correlation of the heterogeneity with the pigmentation level. We went through the manuscript, including the title, to more straightforwardly lead this conclusion: the degree of pigmentation had some but weak correlation with the expression levels of functional genes, and the reason for the weakness of the correlation may be because the color is a temporal condition (as we interpreted from the data) that is different from more stable characteristics of the cells.

We agree that “cell transplantation” in the title (and other parts) was misleading. So, we will change the title, and removed the phrase that led as if the aim of the study was to show something about cell transplantation or in vivo results.

Also, to face scientifically significant results obtained from the present study appropriately, we will discuss more about the correlation of the pigmentation level with some functional genes, and brought this as one of the conclusions of the manuscript.

Author Response:

Reviewer #1 (Public Review):

[...] Weaknesses:

1. I feel the authors need to justify why flow-crushing helps localization specificity. There is an entire family of recent papers that aim to achieve higher localization specificity by doing the exact opposite. Namely, MT or ABC fRMRI aims to increase the localization specificity by highlighting the intravascular BOLD by means of suppressing non-flowing tissue. To name a few:

Priovoulos, N., de Oliveira, I.A.F., Poser, B.A., Norris, D.G., van der Zwaag, W., 2023. Combining arterial blood contrast with BOLD increases fMRI intracortical contrast. Human Brain Mapping hbm.26227. https://doi.org/10.1002/hbm.26227.

Pfaffenrot, V., Koopmans, P.J., 2022. Magnetization Transfer weighted laminar fMRI with multi-echo FLASH. NeuroImage 119725. https://doi.org/10.1016/j.neuroimage.2022.119725

Schulz, J., Fazal, Z., Metere, R., Marques, J.P., Norris, D.G., 2020. Arterial blood contrast ( ABC ) enabled by magnetization transfer ( MT ): a novel MRI technique for enhancing the measurement of brain activation changes. bioRxiv. https://doi.org/10.1101/2020.05.20.106666

Based on this literature, it seems that the proposed method will make the vein problem worse, not better. The authors could make it clearer how they reason that making GE-BOLD signals more extra-vascular weighted should help to reduce large vein effects.

The empirical evidence for the claim that flow crushing helps with the localization specificity should be made clearer. The response magnitude with and without flow crushing looks pretty much identical to me (see Fig, 6d). It's unclear to me what to look for in Fig. 5. I cannot discern any layer patterns in these maps. It's too noisy. The two maps of TE=43ms look like identical copies from each other. Maybe an editorial error?

The authors discuss bipolar crushing with respect to SE-BOLD where it has been previously applied. For SE-BOLD at UHF, a substantial portion of the vein signal comes from the intravascular compartment. So I agree that for SE-BOLD, it makes sense to crush the intravascular signal. For GE-BOLD however, this reasoning does not hold. For GE-BOLD (even at 3T), most of the vein signal comes from extravascular dephasing around large unspecific veins, and the bipolar crushing is not expected to help with this.

The authors would like to clarify that the velocity-nulling gradient is NOT designed to suppress all the contributions from intravascular blood. Instead, we tried to find a balance so that the VN gradient maximally suppressed the macrovascular signal in unspecific veins but minimally attenuated the microvascular signal in specific capillary bed. We acknowledge the reviewer's concern regarding the potential extravascular contributions from large, non-specific vessels. This aspect will be thoroughly evaluated and addressed in the revised manuscript. Additionally, we will make clarifications in other parts that may have cause the reviewer’s misunderstandings.

1. The bipolar crushing is limited to one single direction of flow. This introduces a lot of artificial variance across the cortical folding pattern. This is not mentioned in the manuscript. There is an entire family of papers that perform layer-fmri with black-blood imaging that solves this with a 3D contrast preparation (VAPER) that is applied across a longer time period, thus killing the blood signal while it flows across all directions of the vascular tree. Here, the signal cruising is happening with a 2D readout as a "snap-shot" crushing. This does not allow the blood to flow in multiple directions. VAPER also accounts for BOLD contaminations of larger draining veins by means of a tag-control sampling. The proposed approach here does not account for this contamination.

Chai, Y., Li, L., Huber, L., Poser, B.A., Bandettini, P.A., 2020. Integrated VASO and perfusion contrast: A new tool for laminar functional MRI. NeuroImage 207, 116358. https://doi.org/10.1016/j.neuroimage.2019.116358

Chai, Y., Liu, T.T., Marrett, S., Li, L., Khojandi, A., Handwerker, D.A., Alink, A., Muckli, L., Bandettini, P.A., 2021. Topographical and laminar distribution of audiovisual processing within human planum temporale. Progress in Neurobiology 102121. https://doi.org/10.1016/j.pneurobio.2021.102121

If I would recommend anyone to perform layer-fMRI with blood crushing, it seems that VAPER is the superior approach. The authors could make it clearer why users might want to use the unidirectional crushing instead.

We acknowledge that the degree of velocity nulling varies across the cortical folding pattern. We intend to discuss potential solutions to address this variance, and these may be implemented in the revised manuscript as appropriate. Furthermore, we will provide a comprehensive discussion on the advantages and disadvantages of both CBV-based and BOLD-based approaches.

1. The comparison with VASO is misleading. The authors claim that previous VASO approaches were limited by TRs of 8.2s. The authors might be advised to check the latest literature of the last years. Koiso et al. performed whole brain layer-fMRI VASO at 0.8mm at 3.9 seconds (with reliable activation), 2.7 seconds (with unconvincing activation pattern, though), and 2.3 (without activation). Also, whole brain layer-fMRI BOLD at 0.5mm and 0.7mm has been previously performed by the Juelich group at TRs of 3.5s (their TR definition is 'fishy' though).

Koiso, K., Müller, A.K., Akamatsu, K., Dresbach, S., Gulban, O.F., Goebel, R., Miyawaki, Y., Poser, B.A., Huber, L., 2023. Acquisition and processing methods of whole-brain layer-fMRI VASO and BOLD: The Kenshu dataset. Aperture Neuro 34. https://doi.org/10.1101/2022.08.19.504502

Yun, S.D., Pais‐Roldán, P., Palomero‐Gallagher, N., Shah, N.J., 2022. Mapping of whole‐cerebrum resting‐state networks using ultra‐high resolution acquisition protocols. Human Brain Mapping. https://doi.org/10.1002/hbm.25855

Pais-Roldan, P., Yun, S.D., Palomero-Gallagher, N., Shah, N.J., 2023. Cortical depth-dependent human fMRI of resting-state networks using EPIK. Front. Neurosci. 17, 1151544. https://doi.org/10.3389/fnins.2023.1151544

The authors are correct that VASO is not advised as a turn-key method for lower brain areas, incl. Hippocampus and subcortex. However, the authors use this word of caution that is intended for inexperienced "users" as a statement that this cannot be performed. This statement is taken out of context. This statement is not from the academic literature. It's advice for the 40+ user base that wants to perform layer-fMRI as a plug-and-play routine tool in neuroscience usage. In fact, sub-millimeter VASO is routinely being performed by MRI-physicists across all brain areas (including deep brain structures, hippocampus etc). E.g. see Koiso et al. and an overview lecture from a layer-fMRI workshop that I had recently attended: https://youtu.be/kzh-nWXd54s?si=hoIJjLLIxFUJ4g20&t=2401

Thus, the authors could embed this phrasing into the context of their own method that they are proposing in the manuscript. E.g. the authors could state whether they think that their sequence has the potential to be disseminated across sites, considering that it requires slow offline reconstruction in Matlab? Do the authors think that the results shown in Fig. 6c are suggesting turn-key acquisition of a routine mapping tool? In my humble opinion, it looks like random noise, with most of the activation outside the ROI (in white matter).

Those literatures will be included and discussed in the revised manuscript. Furthermore, we are considering the exclusion of the LGN results presented in Figure 6, as they may divert attention from the primary focus of the study.

We are enthusiastic about sharing our imaging sequence, provided its usefulness is conclusively established. However, it's important to note that without an online reconstruction capability, such as the ICE, the practical utility of the sequence may be limited. Unfortunately, we currently don’t have the manpower to implement the online reconstruction. Nevertheless, we are more than willing to share the offline reconstruction codes upon request.

1. The repeatability of the results is questionable. The authors perform experiments about the robustness of the method (line 620). The corresponding results are not suggesting any robustness to me. In fact, the layer profiles in Fig. 4c vs. Fig 4d are completely opposite. The location of peaks turns into locations of dips and vice versa. The methods are not described in enough detail to reproduce these results. The authors mention that their image reconstruction is done "using in-house MATLAB code" (line 634). They do not post a link to github, nor do they say if they share this code.

It is not trivial to get good phase data for fMRI. The authors do not mention how they perform the respective coil-combination. No data are shared for reproduction of the analysis.

There may have been a misunderstanding regarding the presentation in Figure 4, which illustrates the impact of TEs and the VN gradient. To enhance clarity and avoid further confusion, we will redesign this figure for improved comprehension.

The authors are open to sharing the MATLAB codes associated with our study. However, we were limited by manpower for refining and enhancing the readability of these codes for broader use.

Regarding the coil combination, we utilized an adaptive coil combination approach as described in the paper by Walsh DO, Gmitro AF, and Marcellin MW, titled 'Adaptive reconstruction of phased array MR imagery' (Magnetic Resonance in Medicine 2000; 43:682-690). The MATLAB code for this method was implemented by Dr. Diego Hernando. We will include a link for downloading this code in the revised manuscript for the convenience of interested readers.

1. The application of NODRIC is not validated. Previous applications of NORDIC at 3T layer-fMRI have resulted in mixed success. When not adjusted for the right SNR regime it can result in artifactual reductions of beta scores, depending on the SNR across layers. The authors could validate their application of NORDIC and confirm that the average layer-profiles are unaffected by the application of NORDIC. Also, the NORDIC version should be explicitly mentioned in the manuscript.

Akbari, A., Gati, J.S., Zeman, P., Liem, B., Menon, R.S., 2023. Layer Dependence of Monocular and Binocular Responses in Human Ocular Dominance Columns at 7T using VASO and BOLD (preprint). Neuroscience. https://doi.org/10.1101/2023.04.06.535924

Knudsen, L., Guo, F., Huang, J., Blicher, J.U., Lund, T.E., Zhou, Y., Zhang, P., Yang, Y., 2023. The laminar pattern of proprioceptive activation in human primary motor cortex. bioRxiv. https://doi.org/10.1101/2023.10.29.564658

During our internal testing, we observed that the NORDIC denoising process did not alter the activation patterns. These findings will be incorporated into the revised manuscript. The details of NORDIC will be provided as well.

Reviewer #2 (Public Review):

[...] The well-known double peak feature in M1 during finger tapping was used as a test-bed to evaluate the spatial specificity. They were indeed able to demonstrate two distinct peaks in group-level laminar profiles extracted from M1 during finger tapping, which was largely free from superficial bias. This is rather intriguing as, even at 7T, clear peaks are usually only seen with spatially specific non-BOLD sequences. This is in line with their simple simulations, which nicely illustrated that, in theory, intravascular macrovascular signals should be suppressible with only minimal suppression of microvasculature when small b-values of the VN gradients are employed. However, the authors do not state how ROIs were defined making the validity of this finding unclear; were they defined from independent criteria or were they selected based on the region mostly expressing the double peak, which would clearly be circular? In any case, results are based on a very small sub-region of M1 in a single slice - it would be useful to see the generalizability of superficial-bias-free BOLD responses across a larger portion of M1.

Given the individual variations in the location of the M1 region, we opted for manual selection of the ROI. In the revised manuscript, we plan to explore and implement an independent criterion for ROI selection to enhance the objectivity and reproducibility of our methodology.

As repeatedly mentioned by the authors, a laminar fMRI setup must demonstrate adequate functional sensitivity to detect (in this case) BOLD responses. The sensitivity evaluation is unfortunately quite weak. It is mainly based on the argument that significant activation was found in a challenging sub-cortical region (LGN). However, it was a single participant, the activation map was not very convincing, and the demonstration of significant activation after considerable voxel-averaging is inadequate evidence to claim sufficient BOLD sensitivity. How well sensitivity is retained in the presence of VN gradients, high acceleration factors, etc., is therefore unclear. The ability of the setup to obtain meaningful functional connectivity results is reassuring, yet, more elaborate comparison with e.g., the conventional BOLD setup (no VN gradients) is warranted, for example by comparison of tSNR, quantification and comparison of CNR, illustration of unmasked-full-slice activation maps to compare noise-levels, comparison of the across-trial variance in each subject, etc. Furthermore, as NORDIC appears to be a cornerstone to enable submillimeter resolution in this setup at 3T, it is critical to evaluate its impact on the data through comparison with non-denoised data, which is currently lacking.

We appreciate the reviewer’s comments. Those issues will be addressed carefully.

Reviewer #3 (Public Review):

[...] Weaknesses: - Although the VASO acquisition is discussed in the introduction section, the VN-sequence seems closer to diffusion-weighted functional MRI. The authors should make it more clear to the reader what the differences are, and how results are expected to differ. Generally, it is not so clear why the introduction is so focused on the VASO acquisition (which, curiously, lacks a reference to Lu et al 2013). There are many more alternatives to BOLD-weighted imaging for fMRI. CBF-weighted ASL and GRASE have been around for a while, ABC and double-SE have been proposed more recently.

The principal distinction between DW-fMRI and our methodology lies in the level of the b-value employed. DW-fMRI typically measures cellular swelling by utilizing a b-value greater than 1000 s/mm^2 (e.g. 1800). Conversely, our Velocity Nulling functional MRI (VN-fMRI) approach continues to assess hemodynamic responses, utilizing a smaller b-value specifically for the suppression of signals from draining veins. In addition, other layer-fMRI methods will be discussed.

• The comparison in Figure 2 for different b-values shows % signal changes. However, as the baseline signal changes dramatically with added diffusion weighting, this is rather uninformative. A plot of t-values against cortical depth would be much more insightful.
• Surprisingly, the %-signal change for a b-value of 0 is not significantly different from 0 in the gray matter. This raises some doubts about the task or ROI definition. A finger-tapping task should reliably engage the primary motor cortex, even at 3T, and even in a single participant.
• The BOLD weighted images in Figure 3 show a very clear double-peak pattern. This contradicts the results in Figure 2 and is unexpected given the existing literature on BOLD responses as a function of cortical depth.

In our study, the TE in Figure 2 is shorter than that in Figure 3 (33 ms versus 43 ms). It has been reported in the literature that BOLD fMRI with a shorter TE tends to include a greater intravascular contribution. Acknowledging this, we plan to repeat the experiments with a controlled TE to ensure consistency in our results.

• Given that data from Figures 2, 3, and 4 are derived from a single participant each, order and attention affects might have dramatically affected the observed patterns. Especially for Figure 4, neither BOLD nor VN profiles are really different from 0, and without statistical values or inter-subject averaging, these cannot be used to draw conclusions from.

The order of the experiments were randomized to ensure unbiased results.

It is important to note that the error bars presented in Figures 2, 3, and 4 do not represent the standard deviation of the residual fitting error. Instead, they illustrate the variation across voxels within a specific layer. This approach may lead to the error bars being influenced by the selection of the Region of Interest (ROI). In light of this, we intend to refine our statistical methodologies in the revised manuscript to address this issue.

• In Figure 5, a phase regression is added to the data presented in Figure 4. However, for a phase regression to work, there has to be a (macrovascular) response to start with. As none of the responses in Figure 4 are significant for the single participant dataset, phase regression should probably not have been undertaken. In this case, the functional 'responses' appear to increase with phase regression, which is contra-intuitive and deserves an explanation.
• Consistency of responses is indeed expected to increase by a removal of the more variable vascular component. However, the microvascular component is always expected to be smaller than the combination of microvascular + macrovascular responses. Note that the use of %signal changes may obscure this effect somewhat because of the modified baseline. Another expected feature of BOLD profiles containing both micro- and microvasculature is the draining towards the cortical surface. In the profiles shown in Figure 7, this is completely absent. In the group data, no significant responses to the task are shown anywhere in the cortical ribbon.
• Although I'd like to applaud the authors for their ambition with the connectivity analysis, I feel that acquisitions that are so SNR starved as to fail to show a significant response to a motor task should not be used for brain wide directed connectivity analysis.

We agree that exploring brain-wide directed functional connectivity may be overly ambitious at this stage, particularly before the VN-fMRI technique has been comprehensively evaluated and validated. In the revised manuscript, we will focus more on examining the characteristics of the layer-dependent BOLD signal rather than delving into layer-dependent functional connectivity.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors developed computational models that capture the electrical and Ca2+ signaling behavior in mesenteric arterial cells from male and female mice. A baseline model was first formulated with eleven transmembrane currents and three calcium compartments. Sex-specific differences in the L-type calcium channel and two voltage-gated potassium channels were then tuned based on experimental measurements. To incorporate the stochastic ion channel openings seen in smooth muscle cells under physiological conditions, noise was added to the membrane potential and the sarcoplasmic Ca2+ concentration equations. Finally, the models were assembled into 1D vessel representations and used to investigate the tissue-level electrical response to an L-type calcium channel blocker.

Strengths:

A major strength of the paper is that the modeling studies were performed on three different scales: individual ionic currents, whole-cell, and 1D tissue. This comprehensive computational framework can help provide mechanistic insight into arterial myocyte function that might be difficult to achieve through traditional experimental methods.

The authors aimed to develop sex-specific computational models of mesenteric arterial myocytes and demonstrate their use in drug-testing applications. Throughout the paper, model behavior was both validated by experimental recordings and supported by previously published data. The main findings from the models suggested that sex-specific differences in membrane potential and Ca2+ handling are attributable to variability in the gating of a small number of voltage-gated potassium channels and L-type calcium channels. This variability contributes to a higher Ca2+ channel blocker sensitivity in female arterial vessels. Overall, the study successfully met the aims of the paper.

Thank you for your insightful review and for recognizing the strengths of our study. We appreciate your encouraging comment regarding our multi-scale approach. Indeed, we believe that by systematically connecting these scales—individual ionic currents, whole-cell, and 1D tissue—we can integrate and reconcile experimental and clinical data. We anticipate that this approach will not only provide mechanistic insights into arterial myocyte function that may not be easy to glean from traditional experimental methods but will also facilitate the translation of this information into the development of therapeutic interventions.

Weaknesses:

A main weakness of the paper, as addressed by the authors, is the simplicity of the 1D vessel model; it does not take into account various signaling pathways or interactions with other cell types which could impact smooth muscle electrophysiology.

Thank you for highlighting areas for improvement in our study. The strength of computational modeling lies in its iterative nature, allowing us to introduce and examine variables in a systematic manner. While our current model is simplified and does not contain all details, the modular nature of the build will allow continuous expansion to add the important elements described by the reviewer. We are enthusiastic about progressively enriching the model in subsequent studies, introducing signaling pathways in a step-by-step manner, and ensuring their validation with rigorous experimental data.

Another potential shortcoming is the use of mouse data for optimizing the model, as there could be discrepancies in signaling behavior that limit the translatability to human myocyte predictions.

We appreciate this important comment. Our model was parametrized using data from mouse mesenteric artery smooth muscle cells as initial proof of concept. Mouse arteries are a good representation of human arteries, as they have similar intravascular pressure-myogenic tone relationships, resting membrane potentials, and express similar ionic channels (e.g., CaV1.2, BK channels, RyRs, etc) (PMID: 28119464, PMID: 29070899, PMID: 23232643). In response to the reviewer, we have modified the discussion section of the manuscript to specifically note the mouse is not identical to the human but does share some common important features that make mice a good approximate model.

Reviewer #2 (Public Review):

In this study, Hernandez-Hernandez et al developed a gender-dependent mathematical model of arterial myocytes based on a previous model and new experimental data. The ionic currents of the model and its sex difference were formulated based on patch-clamp experimental data, and the model properties were compared with single-cell and tissue scale experimental results. This is a study that is of importance for the modeling field as well as for experimental physiology.

Thank you for the comment. In fact, we developed a model that incorporates sex-dependent differences that allowed for male and female models. It’s an important distinction as sex is a biological variable and gender is a self-ascribed characteristic.

Reviewer #3 (Public Review):

Summary:

This hybrid experimental/computational study by Hernandez-Hernandez sheds new light on sex-specific differences between male and female arterial myocytes from resistance arteries. The authors conduct careful experiments in isolated myocytes from male and female mice to obtain the data needed to parameterize sex-specific models of two important ionic currents (i.e., those mediated by CaV1.2 and KV2.1). Available experimental data suggest that KV1.5 channel currents from male and female myocytes are similar, but simulations conducted in the novel Hernandez-Hernandez sex-specific models provide a more nuanced view. This gives rise to the first of the authors' three key scientific claims: (1) In males, KV1.5 is the dominant current regulating membrane potential; whereas, in females, KV2.1 plays a primary role in voltage regulation. They further show that this (2) the latter distinction drives drive sex-specific differences in intracellular Ca2+ and cellular excitability. Finally, working with one-dimensional models comprising several copies of the male/female myocyte models linked by resistive junctions, they use simulations to (3) predict that the sensitivity of arterial smooth muscle to Ca2+ channel-blocking drugs commonly used to treat hypertension is heightened in female compared to male cells.

Strengths:

The Methodology is described in exquisite detail in straightforward language that will be easy to understand for most if not all peer groups working in computational physiology. The authors have deployed standard protocols (e.g., parameter fitting as described by Kernik et al., sensitivity analysis as described by Sobie et al.) and appropriate brief explanations of these techniques are provided. The manoeuvre used to represent stochastic effects on voltage dynamics is particularly clever and something I have not personally encountered before. Collectively, these strengthen the credibility of the model and greatly enrich the manuscript.

We appreciate your comment highlighting the robustness of our methodology. Your acknowledgment of our approach to represent stochastic effects on voltage dynamics is especially encouraging. Indeed, noise is a fundamental component of physiological systems, including in vascular myocytes

Broadly speaking, the Results section describes findings that robustly support the three key scientific claims outlined in my summary. While there is certainly room for further discussion of some nuanced points as outlined below, it is evident these experiments were carefully designed and carried out with care and intentionality. In the present version of the manuscript, there are a few figures in which experimental data is shown side-by-side with outputs from the corresponding models. These are an excellent illustration of the power of the authors' novel sex-specific computational simulation platform. I think these figures will benefit from some modest additional quantitative analysis to substantiate the similarities between experimental and computational data, but there is already clear evidence of a good match.

We sincerely appreciate your constructive feedback on the Results section. We have included additional quantitative analysis to substantiate the similarities between experimental and computational data. We agree with the reviewer that the suggestion on the potential value of a more quantitative assessment. As such we have updated the figure to include an in-depth analysis that provides greater insights and solidifies the power of our simulation predictions when compared to experimental results. A detailed analysis of the male and female data as well as the male and female simulations are summarized in the text as follows:

Baseline membrane potential is -40 mV in male myocytes compared to -30 mV. The frequency of hyperpolarization transients (THs) is 1 Hz in male and 2.5 Hz in female cells for the specific baseline membrane potential shown in Figure 5 A-B. In the range of membrane potentials from -50 mV to -30 mV the frequency increases from 1-2.8Hz which is identical to the experimental frequency range.

Areas for Improvement:

The authors used experimental data from a prior publication to calibrate their model of the BKCa current. As indicated in the manuscript, these data are for channel activity measured in a heterologous expression system (Xenopus oocytes). A similar principle applies to other major ion channels/pumps/etc. Is it possible there might be relevant sex-specific differences in these players as well? In the context of the present work, this feels like an important potential caveat to highlight, in case male/female differences in the activity of BKCa or other currents might influence model-predicted differences (e.g., the relative importance of KV1.5 and KV2.1). This should be discussed, and, if possible, related to the elegant sensitivity analysis presented in Fig. 5C (which shows, for example, that the models are relatively insensitive to variation in GBK).

We fully agree with the reviewer - an important caveat to highlight is the unknown sex-specific differences in all the other players regulating membrane potential and calcium signaling. While our initial assessments indicated that the contribution of BKCa channels to the total voltage-gated K+ current (IKvTOT) was small within the physiological range of -50 mV to -30 mV, further analysis of spontaneous transient outward currents revealed sex-specific variations. We have investigations underway to explore if BKCa channel expression and organization may be also sex-dependent.

The authors state that their model can be expanded to 2D/3D applications, "transitioning seamlessly from single-cell to tissue-level simulations". I would like to see more discussion of this. For example, given the modest complexity of the cell-scale model, how considerable would the computational burden be to implement a large network model of a subset of the human female or male arterial system? Are there sex-specific differences in vessel and/or network macro-structure that would need to be considered? How would this influence feasibility? Rather than a 1D cable as implemented here, I imagine a multi-scale implementation would involve the representation of myocytes wrapped around vessels. How would the behavior of such a system differ from the authors' presented work using a 1D representation of 100 myocytes coupled end-to-end? Could these differences partially explain why the traces in Fig. 8D are smoother than those in Fig. 8C? From my standpoint, discussing these points would enrich the paper.

We appreciate the reviewer’s thoughtful and forward-looking ideas! Indeed, we are very interested to extend the model to incorporate a number of these important items.

Our choice for the 1D cable model was driven by its anatomical relevance to the structure of third and fourth-order mesenteric arteries. These arteries possess a singular layer of vascular myocytes encircling the lumen in a cylindrical arrangement. When we conceptualize this structure as unrolled or viewed laterally, it aligns with a flat, rectangular form, closely paralleling our 1D cable implementation. One option is to expand this into a 2D representation by connecting multiple 1D cables together. Another option would be to connect the 1D cable end-to-end to create a ring to represent a cross section. While these approaches would appear to be different geometries, in either case, the dynamics will remain consistent because the cells comprising the tissue are the same. There is no propagating impulse (for example – although even then in a 2D homogenous tissue, a planar wave is identical in 1D), and the only effect will be an increase in electrotonic load (sink) from neighboring cells, which can readily be approximated in 1D by increasing coupling or modification of the boundary conditions.

We totally agree that future investigation should include exploration into the potential sex-specific differences in vessel and/or network macro-structure, as these factors may critically impact predictions and indeed the difference in traces observed between Fig. 8D and Fig. 8C may well involve “insulating” effects of vessel layers and interaction between various cell types and other structural factors. In particular, the contribution of endothelial cells in modulating membrane potential in vascular myocytes might be one such influential factor. In future studies, we are also keen to investigate blood flow regulation where a 3D configuration might become necessary.

The nifedipine data presented in Fig. 9 are quite compelling, and a nice demonstration of the potential power of the new models. How does this relate to what is known about the clinical male/female responses to nifedipine? Are there sex differences in drug efficacy?

Thank you for your comment regarding Fig. 9.

It is well known that sex-specific differences in pharmacokinetics and pharmacodynamics influence antihypertensive drug responses [PMID: 8651122., PMID: 22089536]. Previous studies, notably by Kloner et al., have illustrated this point quantitatively, highlighting a more pronounced diastolic BP response in women (91.4%) compared to men (83%) when treated with dihydropyridine-type channel blockers, such as amlodipine/nifedipine. Importantly, this distinction persisted even after adjusting for confounding factors such as baseline BP, age, weight, and dosage per kilogram [PMID: 8651122]. An interesting observation from Kajiwara et al. emphasizes that vasodilation-related adverse symptoms occur significantly more frequently in younger women (<50 years) compared to their male counterparts, suggesting a heightened sensitivity to dihydropyridine-type calcium channel blockers [PMID: 24728902].

While our findings resonate with clinical observations, a word of caution is in order. Our data suggest that, in the mouse model, nifedipine elicits distinct sex-specific effects. Importantly, future research should test the direct translatability and implications of these observations in human subjects.

Reviewer #1 (Recommendations For The Authors):

1. Cellular simulations with noise: It might be useful to also include in this section how noise was introduced specifically into the [Ca]SR equations.

We agree. The manuscript now includes an expanded explanation of how noise was incorporated into the model. This includes the addition of Equation 6 into section 2.4 "Cellular simulations with noise" to describe how noise was specifically integrated into the [Ca]SR equations. Please see LINE 355.

1. For equation 14, the description might be confusing. RCG and Ri are not explicitly included.

Thank you – this has been corrected.

1. In the paragraph starting with, "Having explored the regulation of graded membrane potential..." , the references to Figure 7C-D do not seem to match the content of the text. Namely, the figures show female versus male responses to nifedipine, which is not introduced until the next paragraph. Additionally, the graphs in 7C-D do not have the panels titled and the y-axes labeled.

We apologize for the error. We have modified the text and figures to address these issues.

1. Perhaps give more detail on how the effects of nifedipine were mathematically simulated at the ionic current level.

Good suggestion. Briefly, previous studies [PMID: 1329564] have shown that at the therapeutic dose of nifedipine (i.e., about 0.1 μM) L-type Cav1.2 channel currents are reduced by about 70%. Accordingly, we decreased ICaL in our mathematical simulations by the same extent. It is known that dihydropyridine-type channel blockers exhibit a voltage-dependent behavior, predominantly binding to the inactivated state. In smooth muscle cells, these blockers initiate inhibition quickly within a voltage range of -60 to -40 mV. This range aligns with the membrane potential baseline of vascular muscle cells (PMID: 8388295), ensuring the blockers are effective without the need of inducing significant depolarization. Therefore, the voltage dependency of dihydropyridine-type channel blockers can be neglected.

1. For the simulations with 400 uncoupled myocytes, the methods stated that the "gap junctional resistance [was set] to zero". Did the authors mean to use "conductivity" or am I misunderstanding?

Thank you for bringing up this issue with the term "gap junctional resistance." We now state that the "gap junctional conductivity" was set to zero to indicate no electrical communication/coupling.

1. Address whether there are differences-such as in cell geometry, degree of sex-based ionic current changes, and frequency of spontaneous hyperpolarization-between mice and human smooth muscle myocytes that could limit the predictive capability of the model.

Excellent point. Our model was parametrized using data from mouse mesenteric artery smooth muscle cells as initial proof of concept. In general terms, mouse arteries are a good animal model for human arteries, as they have similar intravascular pressure-myogenic tone relationships, resting membrane potentials, and express similar ionic channel (e.g., CaV1.2, BK channels, RyRs, etc) (PMID: 28119464, PMID: 29070899). Unfortunately, these studies have largely been done in male arteries and myocytes. Thus, while we recognize that the physiological distinctions between mice and humans could introduce variances in the model's outcomes. Our model offers valuable insights into the sex-specific mechanisms of KV2.1 and CaV1.2 channels in controlling membrane potential and Ca2+ dynamics in mice. It has been shown that sex-specific differences in pharmacokinetics and pharmacodynamics influence antihypertensive drug responses [[PMID: 8651122., PMID: 22089536]. Previous studies, notably by Kloner et al., have illustrated this point quantitatively, highlighting a more pronounced diastolic BP response in women (91.4%) compared to men (83%) when treated with dihydropyridine-type channel blockers, such as amlodipine/nifedipine. Importantly, this distinction persisted even after adjusting for confounding factors such as baseline BP, age, weight, and dosage per kilogram [PMID: 8651122]. An interesting observation from Kajiwara et al. emphasizes that vasodilation-related adverse symptoms occur significantly more frequently in younger women (<50 years) compared to their male counterparts, suggesting a heightened sensitivity to dihydropyridine-type calcium channel blockers [PMID: 24728902].

While our findings resonate with clinical observations, a word of caution is in order. Our data suggest that, in the mouse model, nifedipine elicits distinct sex-specific effects. Importantly, future research should test the direct translatability and implications of these observations in human subjects.

1. "A virtual drug-screening system that can model drug-channel interactions" (pg 32) sounds very novel.

Thank you for highlighting this. We recognize the typo in our manuscript and have made the necessary corrections to ensure clarity and accuracy.

Reviewer #2 (Recommendations For The Authors):

The manuscript is well written. I only have some minor comments:

1. In the patch clamp experiments, there is no information on the recovery of the ionic currents. Is recovery important or not in arterial myocytes? This question is related to the results shown in Figs 5-7. In Fig.5, is the oscillation caused by noise alone or a spontaneous oscillation (such as the oscillation in Fis.6-7) modulated by noise? In general, recovery is an important parameter for the frequency of spontaneous oscillations. It seems to me that the spontaneous oscillations in Fig.8 are mainly noise-driven since they disappear after the cells are coupled through gap junctions.

One important aspect of the oscillatory behavior of the smooth muscle cells is the very long timescales, with fluctuations occurring on the order of seconds. But the majority of ion channels are operating and recovering on the order of milliseconds, so a reasonable approximation is that most ion channels in the cell are operating at steady state at low voltages.

Oscillations in Fig.5: Both the intrinsic oscillations and the noise play key roles in shaping in the oscillations.

The intrinsic deterministic dynamics of the model cells are oscillatory (as seen in Figures 6-7), but the noise can trigger sparks early or delay them, which leads to substantial fluctuations in the inter-spark intervals. Therefore, the spontaneous oscillations are technically modulated by the noise rather than driven by the noise. Nevertheless, in both cases, recovery dynamics play an essential role in shaping the oscillations and determining their frequency

Note however that, when an excitable system is around the bifurcation for oscillations and noise is included, the "firing" statistics in the oscillatory state and the non-oscillatory state are indistinguishable for moderate to high levels of noise.

Noise Exclusion in Figures 6-7: To offer a clear and undistracted interpretation of the results, noise was intentionally omitted from Figures 6-7. This was done to ensure that the primary phenomena under investigation were not obscured. While we recognize the significance of incorporating all elements, including noise, in simulating biological systems, in this case we prioritized a clear point to be made in this context.

Oscillations in Fig.8: Your observation regarding Fig.8 is insightful. Here, uncoupled cells indeed display a spontaneous oscillatory behavior. As documented in previous research, this behavior is not an artifact resulting from cell isolation from the vessel but represents an intrinsic characteristic vital for maintaining electrical signals. The noise in the cells leads to substantial fluctuations in the inter-spike intervals. Because the noise in each cell is uncorrelated, it acts to desynchronize the activity of the cells. Therefore, instead of synchronizing the activity of the cells, the gap junction coupling quenches the large-scale oscillations (the spikes), creating lower amplitude irregular oscillations.

1. The calcium level is much higher in women than in men as shown in Figs.7 and 9. Do women have higher arterial pressure than men?

We thank the reviewer for the observation regarding the calcium levels in Figs.7 and 9. All data presented comes from both male and female C57BL/6J animal models, forming the foundation of our experimental framework.

From earlier studies by the Santana lab (PMID: 32015129), distinct sex-specific differences were found between male and female vascular mesenteric vessels. When the endothelium was removed from small arteriole segments and these segments were subsequently pressurized within a range of 20–120 mmHg, the female arterioles exhibited a pronounced myogenic response in comparison to the male ones. This brings to the forefront the marked sex-based differences, especially in the context of vascular smooth muscle activity.

Yet, when examining the behavior of whole, intact vessels, a different picture emerges. Despite clear sex-specific differences in conditions with the endothelium removed, these distinctions become less pronounced in whole, intact vessels. In essence, both male and female mice exhibit analogous arterial pressure patterns. This suggests possible compensatory mechanisms related to the caliber and structure of the small vessels.

To address the core issue: Despite our data showing higher calcium levels in female samples, it doesn't necessarily imply females consistently exhibit higher arterial pressure across all physiological scenarios.

1. In Fig.9, where is the intravascular pressure (a variable or a parameter) in the mathematical model?

In our model, the intravascular pressure effects are implicitly introduced by modulating the conductance of the non-selective cation currents (INSCC). Specifically, the increase in INSCC is our way of simulating the effects of pressure-induced membrane depolarization. This approach allows us to capture the physiological response to intravascular pressure changes without explicitly introducing it as a separate parameter in the model. We have modified the manuscript to ensure that this rationale is clarified.

1. In Eq.14, the given units of Rmyo (Ohmcm) and Rg (Ohmcmcm) are different, but Eq.14 implies they should have the same unit.

We sincerely appreciate the reviewer's meticulous observation regarding the units discrepancy in Eq.14. We have revised the manuscript to correct the error.

Reviewer #3 (Recommendations For The Authors):

Suggestions for improved or additional experiments, data, or analyses:

Fig. 5 A-B: This is a beautiful qualitative comparison between experimental and simulation data! I think it would be even more impactful if the authors carried out some quantitative analysis of the similarity between male/female experimental/simulation data. For example, the "resting" Vm levels (approx. -30 mV and -40 mV for females and males, respectively) and the peak levels of Vm hyperpolarization could be compared, as well as the frequency of transient hyperpolarization events. It seems like the female model is much more prone to intervals of relative quiescence (i.e., absence of transient hyperpolarization events - e.g., from ~5-6.5 s). Is this consistent with the duration of such ranges in the experimental data (e.g., from 0 to 2.5 s in Fig. 5A).

Thank you for your positive remarks concerning the qualitative comparison in Fig. 5 A-B. We are indeed enthusiastic about the parallels we've identified between experimental and simulation outcomes. We agree with the reviewer that the suggestion on the potential value of a more quantitative assessment. As such we have updated the figure to include an in-depth analysis that provides greater insights and solidifies the power of our simulation predictions when compared to experimental results. A detailed analysis of the male and female data as well as the male and female simulations are summarized in the text as follows:

Baseline membrane potential is -40 mV in male myocytes compared to -30 mV. The frequency of hyperpolarization transients (THs) is 1 Hz in male and 2.5 Hz in female cells for the specific baseline membrane potential shown in Figure 5 A-B. In the range of membrane potentials from -50 mV to -30 mV the frequency increases from 1-2.8Hz which is identical to the experimental frequency range.

• Fig. 7 C-D: Likewise, it would be helpful to quantitatively characterize male/female differences in the model's response to simulated Ca channel blockade (e.g., rate of transient hyperpolarization events, relative levels of ICa and [Ca]i).

Thank you for the constructive feedback on Fig. 7 C-D. We appreciate the emphasis on a quantitative approach to solidify our understanding and have modified the results as follows:

Next, we simulated the effects of calcium channel blocker nifedipine on ICa at a steady membrane potential of -40 mV in male and female simulations. Briefly, previous studies70 have shown that at the therapeutic dose of nifedipine (i.e., about 0.1 μM) L-type Cav1.2 channel currents are reduced by about 70%. Accordingly, we decreased ICa in our mathematical simulations by the same extent. In Figure 7C-D, we show the predicted male (gray) and female (pink) time course of membrane voltage at -40 mV (top panel), ICa (middle panel), and [Ca2+]i (lower panel). First, we observed that in both male and females 0.1 μM nifedipine modifies the frequency of oscillation in the membrane potential, by causing a reduction in oscillation frequency. Second, both male and female simulations (middle panels) show that 0.1 μM nifedipine caused a reduction of ICa to levels that are very similar in male and female myocytes following treatment. Consequently, the reduction of ICa causes both male and female simulations to reach a very similar baseline [Ca2+]i of about 85 nM (lower panels). As a result, simulations provide evidence supporting the idea that CaV1.2 channels are the predominant regulators of intracellular [Ca2+] entry in the physiological range from -40 mV to -20 mV. Importantly, these predictions also suggest that clinically relevant concentrations of nifedipine cause larger overall reductions in Ca2+ influx in female than in male arterial myocytes.

Recommendations for improving the writing and presentation:

When I accessed the GitHub repository linked in section 2.7 (Aug 17, 13:30 PT) it only contained a LICENSE file and none of the described codes and model equations appeared to be publicly available. I would like to access and examine these files. Based on the Clancy lab's excellent track record for making their work publicly available, I have no doubt that the published files will be complete, thoroughly documented, and ready for implementation in studies to reproduce or extend the work described in this manuscript.

https://github.com/ClancyLabUCD/sex-specific-responses-to-calcium-channel-blockers-in-mesenteric-vascular-smooth-muscle

We sincerely apologize for the omission regarding the GitHub repository. It was never our intention to omit the crucial files that should accompany our manuscript. We deeply regret any inconvenience this may have caused in your review process.

We deeply value transparency and the importance of making our work accessible to fellow researchers and the wider community. As you rightly pointed out, the Clancy lab has always been committed to ensuring that our work is available publicly, and this instance is no exception. Please find all codes and documentation here:

Minor corrections to the text and figures:

The introduction is somewhat lengthy, and some of the material contained therein might be more suitable to be merged into the Discussion instead (e.g., paragraphs on negative feedback regulation and the recent study by O'Dwyer et al.).

Thank you – we have updated the introduction but left some foundational work descriptions intact.

• Page 6, section 1.1: There is a missing word (mice?) in the first sentence.

• Page 11, under Eqn. 7: Luo is misspelled as Lou. (Also twice on Page 20.)

Thank you – these have been corrected.

Figs. 2-3: As a colorblind person, it was somewhat challenging for me to differentiate between the red and black lines. Choosing a higher-contrast colour pairing would be beneficial. For some reason, this is not so much of an issue for other figures that use the red/black scheme later in the manuscript (e.g., Figs. 5, 7-8).

We truly appreciate your feedback on the color contrast used in our figures. Accessibility and clarity are crucial to us, and we regret any difficulty you encountered due to the color choices. Based on your valuable feedback, we have included different color pairings in our visual representations to ensure they are comprehensible to all readers, including those who are colorblind.

Fig. 2-3: I am also confused about the use of symbols to indicate significant differences in these plots. In Fig. 2, ** is defined in the legend but not used in the figure. In both figures, the symbols are placed above/below specific sets of points, but it is unclear whether large differences for other x-axis values are statistically significant (e.g., -20 mV in Fig. 3B, +40 mV in Fig. 2C, etc.) This should be clarified.

Thank you – we now have included all the significant differences in the data discussed in the manuscript.

Page 22: The authors state that they "introduced noise into the [Ca]SR..." but the specifics of this approach are not described. As with other aspects of the Methods section, it would be suitable to provide a brief description of the technique used in ref. 40, perhaps added to section 2.4.

Thank you – it has been corrected.

Fig.7 C-D: Axis labels and units are missing. Even though the labels and units will be inferred by most readers, it would be helpful to include them here (at least in C).

Thank you for pointing out the inconsistency between the textual references and Figure 7C-D. We have added the corrected figure.

Page 32: "...the first step toward the development of a virtual drug-screaming system..." I think the authors mean drug-screening. As a side note, this is immediately in the running for the best typo I've ever seen as a peer reviewer.

<good laugh> Thank you for pointing out this error, and we sincerely appreciate your sense of humor about it. You are indeed correct; the intended word is "drug-screening." We have corrected this typo in the manuscript. We're grateful for your thorough review and the light-hearted way you brought this to our attention.

Author Response

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

We would like to thank the reviewers for their strong interest in our studies and their excellent suggestions for improvement.

Reviewer #1:

Weaknesses:

Comment 1. The authors identified NPR-15 and ASJ neurons that are involved in both molecular and behavioral responses to pathogen attack. This finding, by itself, is significant. However, how the NPR-15/ASJ circuit regulates the interplay between the two defense strategies was not explored. Therefore, emphasizing the interplay in the title and the abstract is misleading.

Response to comment 1. We have removed the word “interplay.”

Comment 2. Although the discovery of a single GPCR regulating both immunity and avoidance behavior is significant and novel, NPR-15 is not the first GPCR identified with these functions. Previously, the same lab reported that the GPCR OCTR-1 also regulates immunity and avoidance behavior through ASH and ASI neurons respectively (PMID: 29117551). This point was not mentioned in the current manuscript.

Response to Comment 2. We’d like to clarify that it remains unclear whether OCTR-1 itself controls both immunity and behavior (PMID: 29117551). The reference study showed that OCTR-1-expressing neurons ASH and ASI control immunity and behavior, respectively. We modified the manuscript to make this point clearer: “While OCTR-1-expressing neurons ASI play a role in avoidance (34), the specific role of OCTR-1 in ASH and ASI neurons remains unclear. “

Comment 3. The authors discovered that NPR-15 regulates avoidance behavior via the TRPM gene, GON-2. Only two factors (GON-2 and GTL-2) were examined in this study, and GON-2 happens to function through the intestine.

Response to comment 3. We studied GON-2 and GTL-2 because a recent screen of intestinal TRPM genes showed that they are the only two involved in the control of pathogen avoidance. We modified the manuscript to make this rationale clearer: “Because transient receptor potential melastatin (TRPM) ion channels, GON-2 and GTL-2, are required for pathogen avoidance (32), we studied whether they may be part of the NPR-15 pathway that controls pathogen avoidance”

Comment 3b. It is possible that NPR-15 may broadly regulate multiple effectors in multiple tissues. Confining the regulation to the amphid sensory neuron-intestinal axis, as stated in the title and elsewhere in the manuscript, is not accurate.

Response to comment 3b. We agree that NPR-15 may broadly regulate multiple effectors in different tissues. Indeed, we have shown that the transcriptional activity of ELT-2, HLH-30, DAF-16, and PMK-1 is higher in npr-15 than in WT animals. We found that expression of NPR-15 only in ASJ cells rescues both the survival and behavioral phenotypes of npr-15 animals (Figs. 4F and 5C).

Comment 4. The C. elegans nervous system is simple, and hermaphrodites only have 302 neurons. Individual neurons possessing multiple regulatory functions is expected. Whether this is conserved in mammals and other vertebrates is unknown, because in higher animals, neurons and neuronal circuits could be more specialized.

Response to Comment 4. We agreed. We have removed the statements discussing conservation in that manner.

Comment 5. A key question, that is, why would NPR-15 suppress immunity (which is bad for defense) but enhance avoidance behavior (which is good for defense), is not addressed or explained. This could be due to temporal regulation, for example, upon pathogen exposure, NPR-15 could regulate behavior to avoid the pathogen, but after infection, NPR-15 could suppress excessive immune responses or quench the responses for the resolution of infection.

Response to comment 5. We found that NPR-15 controls the expression of immune genes in the absence of an infection. Without further experiments, we think it would be too speculative to discuss the possibility of a temporal regulation. However, we modified the manuscript to address the control of both molecular and behavioral immunity by NPR-15. The revised discussion reads: “Our findings shed light on the role of NPR-15 in the control of the immune response. NPR-15 seems to suppress specific immune genes while activating pathogen avoidance behavior to minimize potential tissue damage and the metabolic energy cost associated with activating the molecular immune response against pathogen infections. Overall, the control of immune activation is essential for maintaining homeostasis and preventing excessive tissue damage caused by an overly aggressive and energy-costly response against pathogens (60-63).”

Comment 6. Discussion appears timid in scope and contains some repetitive statements. Point 5 can be addressed in the Discussion.

Response to comment 6. We have removed repetitive concepts and modified the discussion as mentioned in the response to point 5.

Comment 7. Overall, the authors presented an impactful study that identified specific molecules and neuronal cells that regulate both molecular and behavioral immune responses to pathogen attack. Most conclusions are supported by solid evidence. However, some statements are overreaching, for example, regulation of the interplay between molecular and behavioral immune responses was emphasized but not explored. Nonetheless, this study reported a significant and novel discovery and has laid a foundation for investigating such an interplay in the future.

Response to comment 7: We removed the statements that may have appeared to be overreaching and addressed the weakness raised by the reviewer. The revised discussion reads “Our findings shed light on the role of NPR-15 in the control of the immune response. NPR-15 seems to suppress specific immune genes while activating pathogen avoidance behavior to minimize potential tissue damage and the metabolic energy cost associated with activating the molecular immune response against pathogen infections. Overall, the control of immune activation is essential for maintaining homeostasis and preventing excessive tissue damage caused by an overly aggressive and energy-costly response against pathogens (60-63).”

Recommendations for the authors:

Recommendations 1. The title, abstract and some statements in the main text need to be re-written to reflect the fact that regulation of the interplay between molecular and behavioral immune responses was not explored in this study.

Response to recommendations 1. We modified the title and abstract accordingly.

Recommendations 2. It should be mentioned in the manuscript that OCTR-1 is the first GPCR that was identified to regulate both immunity and avoidance behavior.

Response to recommendation 2. We addressed this issue as discussed in the response to comment 2.

Recommendations 3. Repetitive statements should be removed from Discussion.

Response to recommendations 3. The statements were removed.

Recommendations 4. It is surprising to see that pmk-1 RNAi did not affect the survival of npr-15(tm12539) animals against S. aureus because PMK-1 has a general role in defense against S. aureus infection.

Response to recommendations 4. We agree. However, the RNAi studies were validated using mutants (Fig. S3B).

Recommendations 4b. Also, the rationale for using skn-1 RNAi as a control was not given. These need to be explained adequately in the manuscript.

Response to recommendations 4b. There’s no need to include skn-1 RNAi and we removed the data.

Recommendations 5. The conclusion that the lack of avoidance behavior by NPR-15 loss-of-function is independent of immunity and neuropeptide genes was drawn entirely based on experiments with RNAi of individual genes. Functional redundancy among genes could render RNAi of individual genes ineffective, thus masking the dependence of avoidance behavior on these genes. More experiments are needed to support this conclusion, or the wording of the conclusion need to be changed.

Response to recommendations 5. We modified the conclusion to address this issue: “Given the possibility of functional redundancy among these genes, we cannot rule out the possibility that different combinations may play a role in controlling avoidance behavior.”

Recommendations 6. What is representation factor in Fig. 2B and 2C?

Response to recommendations 5. Figure 2B shows significantly enriched terms with a Q value < 0.1, sorted by P values. Figure 2 C shows the representation factor that is calculated using a tool, http://nemates.org/MA/progs/overlap_stats.html. The calculation is based on the number of genes in set 1, the number of genes in set 2, and the Overlap between set 1 and set 2, as well as the number of genes in the genome.

We corrected the Figure legends and included the corresponding information in Material and Methods.

Recommendations 7. The legend of Fig. 6 was wrong and should be changed to 'GPCR/NPR-15 suppressed immune response and enhanced avoidance behavior via sensory neurons'.

Response to recommendations 7. Thank you for pointing this out. We changed the legend.

Reviewer #2:

Comments 1. There is some variance in lawn occupancy of wt strains between the different trials in WT animals (e.g. in Fig. 1: 25 for wt vs 60% for npr mutant; S1c 5% for wt and 60% for npr mutant).

Response to comment 1. We appreciate the observation. We did notice some variation in both the WT and npr-15(tm12539) animals during our study. Notably, the variation appeared to be more in the WT compared to the npr-15(tm12539) animals. However, it's important to note that these variations did not significantly affect the outcome of our findings. We calculated the means, standard deviation, and standard error across different experimental trials that are presented in the manuscript (Table S2) (new Table). It's worth noting that these variations did not significantly impact the observed differences in lawn occupancy between the wild-type (WT) and npr-15 mutant strains.

We addressed this issue in the revised manuscript: “Interestingly, we noticed that the variation in lawn occupancy is greater in WT than in npr-15(tm12539) animals across experiments (Table S2), which suggests that the strong lack of avoidance of npr-15(tm12539) somehow counteracts the experimental variation”

Comment 2. Does this reflect rates of migration or re-occupancy in WT?

Response to comment 2. We did not observe any re-occupancy in either the WT or npr-15 animals at 24-hour time points (which we mostly use in this study) or beyond. To address the comment, we performed a new experiment and found that the re-occupancy of npr-15 mutants is comparable to that of WT animals at 4 hours post-exposure (Figure S1B).

Comment 3. Does pathogen avoidance persist and/or the rate of avoidance differ in npr mutant worms?

Response to comment 3. As illustrated in new Figure S1B, the avoidance behavior in response to pathogens remained consistent even when we extended our observations up to 48 hours (Figure S1B).

Comment 4. if animals were exposed then re-exposed, could the authors to determine whether a learned avoidance was similarly affected by this mutation by assessing rate changes?

Response to comment 4. We conducted the proposed experiment and observed that the WT animals learned to avoid the pathogen but not npr-15(tm12539) mutants (Figure S1C). The revised manuscript reads: “We also found that npr-15(tm12539) exhibited reduced learned avoidance compared to WT animals (Figure S1C).”

Comment 5: Is there any difference in gene expression of animals that have migrated off the lawn to those remaining on the lawn (e.g. in partial lawn experiments?).

Response to comment 5. This is an interesting question that has not been addressed in the field yet. While we think the study is exciting, we believe that it is outside the scope of our work. All the gene expression studies performed here are in non-avoiding conditions.

Comment 6. No concerns but the P values in the legends are a pain to read. Why not put them in figures as in above figures.

Response to comment 6. We included the P values as suggested.

Recommendations for the authors:

Recommendation 1. Fig. 1/S1. Comments: There is some variance in lawn occupancy of wt strains between the different trials in WT animals (e.g. in Fig. 1: 25 for wt vs 60% for npr mutant; S1c 5% for wt and 60% for npr mutant).

Response to recommendation 1. We addressed this issue as discussed in the response to comment 1.

Recommendation 2. Fig. 1/S1. Comments. Does this reflect rates of migration or re-occupancy in WT?

Response to recommendation 2. We have responded to this issue in comment 2.

Recommendations 3. Fig. 1/S1. Comments. Does pathogen avoidance persist and/or the rate of avoidance differ in npr mutant worms.

Response to recommendation 3. We have responded to this issue in comment 3.

Recommendation 4. Fig. 1/S1. Comments B. and if animals were exposed then re- exposed, could the authors to determine whether a learned avoidance was similarly affected by this mutation by assessing rate changes?

Response to recommendation 4: We have responded to this issue in comment 4 above.

Recommendation 5. Fig. 2/S2. Comment: Is there any difference in gene expression of animals that have migrated off the lawn to those remaining on the lawn (e.g. in partial lawn expts?).

Response to recommendation 5. We have responded to this issue in comment 5 above.

Recommendation 6. Fig. 3/S3. Comment. No concerns but the P values in the legends are a pain to read. Why not put them in figures as in above figures.

Response to recommendation 6. We included the P values.

Recommendation 7. Fig. 5. Comments: The authors suggest that the ASJ/NPR15 effect to limit avoidance acts via inhibition of GON-2 in the intestine. The observation that GON-2 inhibition effects on pathogen avoidance occur independently of neurons could suggest that it is a redundant way of accomplishing the same thing, which then makes one wonder if or what the connection is exists between the neuron and the gut. The effect of ASJ via NPR on pathogen avoidance is not neuropeptide dependent, which they show. So how the neuronal-gut communication works. Specific Transmitters... perhaps.

Response to Recommendation 7 Fig. 5. Thanks for this observation. To address the recommendation, we modified the discussion: “Our research additionally indicates that the regulation of NPR-15-mediated avoidance is not influenced by intestinal immune and neuropeptide genes. Given the potential for functional redundancy and our focus on genes upregulated in the absence of NPR-15, we cannot entirely rule out the possibility that unexamined immune effectors or neuropeptides, not transcriptionally controlled by NPR-15, might be involved. Different intestinal signals may also participate in the NPR-15 pathway that controls pathogen avoidance.”

Recommendation 8. Comment. Since ASJ neurons control entry into dauer, perhaps isn't surprising that DAF-16 showed up as an NPR-15. induced factor (and dauer worms are resistant to a lot of stressors); that said dauer hormones might be involved as well. Is there any evidence that DAF-16 down-regulates GON-2 expression (see Murphy, Kenyon et al. 2005), and along these lines would GON-2 RNAi work in a DAF-16 mutant? I think addressing these issues are the subject of future studies.

Response to recommendation 8. We checked the data in the study by Murphy, Kenyon et al., and found that the gon-2 gene was not downregulated.

Recommendation 9. Minor: Regarding the description to Fig. 5. "Consistently with our previous findings, we found that only " The adverb form of consistent should not be used here.

Response to recommendation 9. Thank you for pointing this out. The description of Figure 5 was corrected.

Author Response

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

Reviewer #1:

A weakness of the paper is that the power of the model is illustrated for only one specific set of parameters, added in a stepwise manner and the comparison to one specific empirical TGM, assumed to be prototypical; And that this comparison remains descriptive. (That is could a different selection of parameters lead to similar results and is there TGM data which matches these settings less well.)

The fact that the comparisons in the paper are descriptive is a central point of criticism from both reviewers. As mentioned in my preliminary response, I intentionally did not optimise the model to a specific TGM or show an explicit metric of fitness. As I now explicitly mention in the new experimental section of the paper:

“The previous analyses were descriptive in the sense that they did not quantify how much the generated TGMs resembled a specific empirical TGM. This was deliberate, because empirical TGMs vary across subjects and experiments, and I aimed at characterising them as generally as possible by looking at some characteristic features in broad terms. For example, while TGMs typically have a strong diagonal and horizontal/vertical bars of high accuracy, questions such as when these effects emerge and for how long are highly dependent on the experimental paradigm. For the same reason, I did not optimise the model hyperparameters, limiting myself to observing the behaviour of the model across some characteristic configurations”

And, in the Discussion:

“The demonstrations here are not meant to be tailored to a specific data set, and are, for the most part, intentionally qualitative. TGMs do vary across experiments and subjects; and the hyperparameters of the model can be explicitly optimised to specific scientific questions, data sets, and even individuals. In order to explore the space of configurations effectively, an automatic optimisation of the hyperparameter space using, for instance, Bayesian optimisation (Lorenz, et al., 2017) could be advantageous. This may lead to the identification of very specific (spatial, spectral and temporal) features in the data that may be neurobiologically interpreted.”

Nonetheless, it is possible to fit the model to a specific TGMs by using a explicit metric of fitness. For illustration, this is what I did in the new experimental section Fitting and empirical TGM, where I used correlation with an empirical TGM to optimise two temporal parameters: the rise slope and the fall slope. As can be seen in the Figure 8, the correlation with the empirical TGM was as high as 0.7, even though I did not fit the other parameters of the model. As mentioned in the paragraph above, more sophisticated techniques such as Bayesian optimisation might be necessary for a more exhaustive exploration, but this would be beyond the scope of the current paper.

I would also like to point out that fitting the parameters in a step-wise manner was a necessity for interpretation. I suggest to think of the way we use F-tests in regression analyses as a comparison: if we want to know how important a feature is, we compare the model with and without this feature and see how much we loss.

It further remained unclear to me, which implications may be drawn from the generative model, following from the capacities to mimic this specific TGM (i) for more complex cases, such as the comparison between experimental conditions, and (ii) about the complex nature of neural processes involved.

Following on the previous points, the object of this paper (besides presenting the model and the associated toolbox) was not to mimic a specific TGM, but to characterise the main features that we generally see across studies in the field. To clarify this, I have added Figure 2 (previously a Supplemental Information figure), and added the following to the Results section:

“Figure 2 shows a TGM for an example subject, where some archetypal characteristics are highlighted. In the experiments below, specifically, I focus on the strong narrow diagonal at the beginning of the trial, the broadening of accuracy later in the trial, and the vertical/horizontal bars of higher-than-chance accuracy. Importantly, this specific example in Figure 2 is only meant as a reference, and therefore I did not optimise the model hyperparameters to this TGM (except in the last subsection), or showed any quantitative metric of similarity.”

I mention the possibility of using the model to explore more complex cases in the Introduction, although doing so here would be out of scope:

“Other experimental paradigms, including motor tasks and decision making, can be investigated with genephys”

Towards this end, I would appreciate (i) a more profound explanation of the conclusions that can be drawn from this specific showcase, including potential limitations, as well as wider considerations of how scientists may empower the generative model to (ii) understand their experimental data better and (iii) which added value the model may have in understanding the nature of underlying brain mechanism (rather than a mere technical characterization of sensor data).

To better illustrate how to use genephys to explore a specific data set, I have added a section (Fitting an empirical TGM) where I show how to fit specific hyperparameters to an empirical TGM in a simple manner.

In the Introduction, I briefly mentioned:

“This (not exhaustive) list of effects was considered given previous literature (Shah, et al., 2004; Mazaheri & Jensen, 2006; Makeig, et al., 2002; Vidaurre, et al., 2021), and each effect may be underpinned by distinct neural mechanisms. For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question”

In the Discussion, I have further commented:

“Genephys has different available types of effect, including phase resets, additive damped oscillations, amplitude modulations, and non-oscillatory responses. All of these elements, which may relate to distinct neurobiological mechanisms, are configurable and can be combined to generate a plethora of TGMs that, in turn, can be contrasted to specific empirical TGMs. This way, we can gain insight on what mechanisms might be at play in a given task.

The demonstrations here are not meant to be tailored to a specific data set, and are, for the most part, intentionally qualitative. TGMs do vary across experiments and subjects; and the hyperparameters of the model can be explicitly optimised to specific scientific questions, data sets, and even individuals. In order to explore the space of configurations effectively, an automatic optimisation of the hyperparameter space using, for instance, Bayesian optimisation (Lorenz, et al., 2017) could be advantageous. This may lead to the identification of very specific (spatial, spectral and temporal) features in the data that may be neurobiologically interpreted. “

On p. 15 "Having a diversity of frequencies but not of latencies produces another regular pattern consisting of alternating, parallel bands of higher/lower than baseline accuracy. This, shown in the bottom left panel, is not what we see in real data either. Having a diversity of latencies but not of frequencies gets us closer to a realistic pattern, as we see in the top right panel." The terms frequency and latency seem to be confused.

The Reviewer is right. I have corrected this now. Thank you.

Reviewer #2:

The results of comparisons between simulations and real data are not always clear for an inexperienced reader. For example, the comparisons are qualitative rather than quantitative, making it hard to draw firm conclusions. Relatedly, it is unclear whether the chosen parameterizations are the only/best ones to generate the observed patterns or whether others are possible. In the case of the latter, it is unclear what we can actually conclude about underlying signal generators. It would have been different if the model was directly fitted to empirical data, maybe of different cognitive conditions. Finally, the neurobiological interpretation of different signal properties is not discussed. Therefore, taken together, in its currently presented form, it is unclear how this method could be used exactly to further our understanding of the brain.

This critique coincides with that of Reviewer 1. In the current version, I made more clear the fact that I am not fitting a specific empirical TGM and why, and that, instead, I am referring to general features that appear broadly throughout the literature. See more detailed changes below.

Regarding whether the chosen parameterizations are the only/best ones to generate the observed patterns, the Discussion reflects this limitation:

“Also importantly, I have shown that standard decoding analysis can differentiate between these explanations only to some extent. For example, the effects induced by phase-resetting and the use of additive oscillatory components are not enormously different in terms of the resulting TGMs. In future work, alternatives to standard decoding analysis and TGMs might be used to disentangle these sources of variation (Vidaurre, et al., 2019). ”

And

“Importantly, the list of effects that I have explored here is not exhaustive …”

Of course, since the list of signal features I have explored is not exhaustive, it cannot be claimed without a doubt that these features are the ones generating the properties we observe in real TGMs. The model, however, is a step forward in that direction, as it provides us with a tool to at least rule out some causes.

Firstly, it was not entirely clear to me from the introduction what gap exactly the model is supposed to fill: is it about variance in neural responses in general, about which signal properties are responsible for decoding, or about capturing stability of signals? It seems like it does all of these, but this needs to be made clearer in the introduction. It would be helpful to emphasize exactly what insights the model can provide that are unable to be obtained with the current methods.

I have now made this explicit in in the Introduction, as suggested:

“To gain insight into what aspects of the signal underpin decoding accuracy, and therefore the most stable aspects of stimulus processing, I introduce a generative model”

To help illustrating what insights the model can provide, I have added the following sentence as an example:

“For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question.”

Furthermore, I was unclear on why these specific properties were chosen (lines 71 to 78). Is there evidence from neuroscience to suggest that these signal properties are especially important for neural processing? Or, if the logic has more to do with signal processing, why are these specific properties the most important to include?

To clarify this the text now reads:

“In the model, when a channel responds, it can do it in different ways: (i) by phase-resetting the ongoing oscillation to a given target phase and then entraining to a given frequency, (ii) by an additive oscillatory response independent of the ongoing oscillation, (iii) by modulating the amplitude of the stimulus-relevant oscillations, or (iv) by an additive non-oscillatory (slower) response. This (not exhaustive) list of effects was considered given previous literature (Shah, et al., 2004; Mazaheri & Jensen, 2006; Makeig, et al., 2002; Vidaurre, et al., 2021), and each effect may be underpinned by distinct neural mechanisms”

The general narrative and focus of the paper could also be improved. It might help to start off with an outline of what the goal is at the start of the paper and then explicitly discuss how each of the steps works toward that goal. For example, I got the idea that the goal was to capture specific properties of an empirical TGM. If this was the case, the empirical TGM could be placed in the main body of the text as a reference picture for all simulated TGMs. For each simulation step, it could be emphasized more clearly exactly which features of the TGM is captured and what that means for interpreting these features in real data.

Thank you. To clarify the purpose of the paper better, I have brought Figure 2 to the front (before a Supplementary Figure), and in the first part of Results I have now added:

“Figure 2 shows a TGM for an example subject, where some archetypal characteristics are highlighted. In the experiments below, specifically, I focus on the strong narrow diagonal at the beginning of the trial, the broadening of accuracy later in the trial, and the vertical/horizontal bars of higher-than-chance accuracy. Importantly, this specific example in Figure 2 is only meant as a reference, and therefore I did not optimise the model hyperparameters to this TGM (except in the last subsection), or showed any quantitative metric of similarity. ”

I have enunciated the goals more clearly in the Introduction:

“To gain insight into what aspects of the signal underpin decoding accuracy, and therefore the most stable aspects of stimulus processing, …”

Relatedly, it would be good to connect the various signal properties to possible neurobiological mechanisms. I appreciate that the author tries to remain neutral on this in the introduction, but I think it would greatly increase the implications of the analysis if it is made clearer how it could eventually help us understand neural processes.

The Reviewer is right in pointing out that I preferred to remain neutral on this. While I have still kept that tone of neutrality throughout the paper, I have now included the following sentence as an example of a neurobiological question that could be investigated with the model:

“For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question.”

And, more generally,

“Genephys has different available types of effect, including phase resets, additive damped oscillations, amplitude modulations, and non-oscillatory responses. All of these elements, which may relate to distinct neurobiological mechanisms, are configurable and can be combined to generate a plethora of TGMs that, in turn, can be contrasted to specific empirical TGMs. This way, we can gain insight on what mechanisms might be at play in a given task. ”

Line 57: this sentence is very long, making it hard to follow, could you break up into smaller parts?

Thank you. The sentence is fragmented now.

Please replace angular frequencies with frequencies in Hertz for clarity.

Here I have preferred to stick to angular frequencies because it is more general than if I talk about Hertz, because that would entail having a specific sampling frequency. I think doing so would create confusion precisely of the sorts that I am trying to clarify in this revision: that is, that these results are not specific of one TGM but reflect general features that we see broadly in the literature.

There are quite some types throughout the paper, please recheck

Thank you. I have revised and have made my best to clear them out.

Author Response

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

Reviewer #1:

A weakness of the paper is that the power of the model is illustrated for only one specific set of parameters, added in a stepwise manner and the comparison to one specific empirical TGM, assumed to be prototypical; And that this comparison remains descriptive. (That is could a different selection of parameters lead to similar results and is there TGM data which matches these settings less well.)

The fact that the comparisons in the paper are descriptive is a central point of criticism from both reviewers. As mentioned in my preliminary response, I intentionally did not optimise the model to a specific TGM or show an explicit metric of fitness. As I now explicitly mention in the new experimental section of the paper:

“The previous analyses were descriptive in the sense that they did not quantify how much the generated TGMs resembled a specific empirical TGM. This was deliberate, because empirical TGMs vary across subjects and experiments, and I aimed at characterising them as generally as possible by looking at some characteristic features in broad terms. For example, while TGMs typically have a strong diagonal and horizontal/vertical bars of high accuracy, questions such as when these effects emerge and for how long are highly dependent on the experimental paradigm. For the same reason, I did not optimise the model hyperparameters, limiting myself to observing the behaviour of the model across some characteristic configurations”

And, in the Discussion:

“The demonstrations here are not meant to be tailored to a specific data set, and are, for the most part, intentionally qualitative. TGMs do vary across experiments and subjects; and the hyperparameters of the model can be explicitly optimised to specific scientific questions, data sets, and even individuals. In order to explore the space of configurations effectively, an automatic optimisation of the hyperparameter space using, for instance, Bayesian optimisation (Lorenz, et al., 2017) could be advantageous. This may lead to the identification of very specific (spatial, spectral and temporal) features in the data that may be neurobiologically interpreted.”

Nonetheless, it is possible to fit the model to a specific TGMs by using a explicit metric of fitness. For illustration, this is what I did in the new experimental section Fitting and empirical TGM, where I used correlation with an empirical TGM to optimise two temporal parameters: the rise slope and the fall slope. As can be seen in the Figure 8, the correlation with the empirical TGM was as high as 0.7, even though I did not fit the other parameters of the model. As mentioned in the paragraph above, more sophisticated techniques such as Bayesian optimisation might be necessary for a more exhaustive exploration, but this would be beyond the scope of the current paper.

I would also like to point out that fitting the parameters in a step-wise manner was a necessity for interpretation. I suggest to think of the way we use F-tests in regression analyses as a comparison: if we want to know how important a feature is, we compare the model with and without this feature and see how much we loss.

It further remained unclear to me, which implications may be drawn from the generative model, following from the capacities to mimic this specific TGM (i) for more complex cases, such as the comparison between experimental conditions, and (ii) about the complex nature of neural processes involved.

Following on the previous points, the object of this paper (besides presenting the model and the associated toolbox) was not to mimic a specific TGM, but to characterise the main features that we generally see across studies in the field. To clarify this, I have added Figure 2 (previously a Supplemental Information figure), and added the following to the Results section:

“Figure 2 shows a TGM for an example subject, where some archetypal characteristics are highlighted. In the experiments below, specifically, I focus on the strong narrow diagonal at the beginning of the trial, the broadening of accuracy later in the trial, and the vertical/horizontal bars of higher-than-chance accuracy. Importantly, this specific example in Figure 2 is only meant as a reference, and therefore I did not optimise the model hyperparameters to this TGM (except in the last subsection), or showed any quantitative metric of similarity.”

I mention the possibility of using the model to explore more complex cases in the Introduction, although doing so here would be out of scope:

“Other experimental paradigms, including motor tasks and decision making, can be investigated with genephys”

Towards this end, I would appreciate (i) a more profound explanation of the conclusions that can be drawn from this specific showcase, including potential limitations, as well as wider considerations of how scientists may empower the generative model to (ii) understand their experimental data better and (iii) which added value the model may have in understanding the nature of underlying brain mechanism (rather than a mere technical characterization of sensor data).

To better illustrate how to use genephys to explore a specific data set, I have added a section (Fitting an empirical TGM) where I show how to fit specific hyperparameters to an empirical TGM in a simple manner.

In the Introduction, I briefly mentioned:

“This (not exhaustive) list of effects was considered given previous literature (Shah, et al., 2004; Mazaheri & Jensen, 2006; Makeig, et al., 2002; Vidaurre, et al., 2021), and each effect may be underpinned by distinct neural mechanisms. For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question”

In the Discussion, I have further commented:

“Genephys has different available types of effect, including phase resets, additive damped oscillations, amplitude modulations, and non-oscillatory responses. All of these elements, which may relate to distinct neurobiological mechanisms, are configurable and can be combined to generate a plethora of TGMs that, in turn, can be contrasted to specific empirical TGMs. This way, we can gain insight on what mechanisms might be at play in a given task.

The demonstrations here are not meant to be tailored to a specific data set, and are, for the most part, intentionally qualitative. TGMs do vary across experiments and subjects; and the hyperparameters of the model can be explicitly optimised to specific scientific questions, data sets, and even individuals. In order to explore the space of configurations effectively, an automatic optimisation of the hyperparameter space using, for instance, Bayesian optimisation (Lorenz, et al., 2017) could be advantageous. This may lead to the identification of very specific (spatial, spectral and temporal) features in the data that may be neurobiologically interpreted. “

On p. 15 "Having a diversity of frequencies but not of latencies produces another regular pattern consisting of alternating, parallel bands of higher/lower than baseline accuracy. This, shown in the bottom left panel, is not what we see in real data either. Having a diversity of latencies but not of frequencies gets us closer to a realistic pattern, as we see in the top right panel." The terms frequency and latency seem to be confused.

The Reviewer is right. I have corrected this now. Thank you.

Reviewer #2:

The results of comparisons between simulations and real data are not always clear for an inexperienced reader. For example, the comparisons are qualitative rather than quantitative, making it hard to draw firm conclusions. Relatedly, it is unclear whether the chosen parameterizations are the only/best ones to generate the observed patterns or whether others are possible. In the case of the latter, it is unclear what we can actually conclude about underlying signal generators. It would have been different if the model was directly fitted to empirical data, maybe of different cognitive conditions. Finally, the neurobiological interpretation of different signal properties is not discussed. Therefore, taken together, in its currently presented form, it is unclear how this method could be used exactly to further our understanding of the brain.

This critique coincides with that of Reviewer 1. In the current version, I made more clear the fact that I am not fitting a specific empirical TGM and why, and that, instead, I am referring to general features that appear broadly throughout the literature. See more detailed changes below.

Regarding whether the chosen parameterizations are the only/best ones to generate the observed patterns, the Discussion reflects this limitation:

“Also importantly, I have shown that standard decoding analysis can differentiate between these explanations only to some extent. For example, the effects induced by phase-resetting and the use of additive oscillatory components are not enormously different in terms of the resulting TGMs. In future work, alternatives to standard decoding analysis and TGMs might be used to disentangle these sources of variation (Vidaurre, et al., 2019). ”

And

“Importantly, the list of effects that I have explored here is not exhaustive …”

Of course, since the list of signal features I have explored is not exhaustive, it cannot be claimed without a doubt that these features are the ones generating the properties we observe in real TGMs. The model, however, is a step forward in that direction, as it provides us with a tool to at least rule out some causes.

Firstly, it was not entirely clear to me from the introduction what gap exactly the model is supposed to fill: is it about variance in neural responses in general, about which signal properties are responsible for decoding, or about capturing stability of signals? It seems like it does all of these, but this needs to be made clearer in the introduction. It would be helpful to emphasize exactly what insights the model can provide that are unable to be obtained with the current methods.

I have now made this explicit in in the Introduction, as suggested:

“To gain insight into what aspects of the signal underpin decoding accuracy, and therefore the most stable aspects of stimulus processing, I introduce a generative model”

To help illustrating what insights the model can provide, I have added the following sentence as an example:

“For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question.”

Furthermore, I was unclear on why these specific properties were chosen (lines 71 to 78). Is there evidence from neuroscience to suggest that these signal properties are especially important for neural processing? Or, if the logic has more to do with signal processing, why are these specific properties the most important to include?

To clarify this the text now reads:

“In the model, when a channel responds, it can do it in different ways: (i) by phase-resetting the ongoing oscillation to a given target phase and then entraining to a given frequency, (ii) by an additive oscillatory response independent of the ongoing oscillation, (iii) by modulating the amplitude of the stimulus-relevant oscillations, or (iv) by an additive non-oscillatory (slower) response. This (not exhaustive) list of effects was considered given previous literature (Shah, et al., 2004; Mazaheri & Jensen, 2006; Makeig, et al., 2002; Vidaurre, et al., 2021), and each effect may be underpinned by distinct neural mechanisms”

The general narrative and focus of the paper could also be improved. It might help to start off with an outline of what the goal is at the start of the paper and then explicitly discuss how each of the steps works toward that goal. For example, I got the idea that the goal was to capture specific properties of an empirical TGM. If this was the case, the empirical TGM could be placed in the main body of the text as a reference picture for all simulated TGMs. For each simulation step, it could be emphasized more clearly exactly which features of the TGM is captured and what that means for interpreting these features in real data.

Thank you. To clarify the purpose of the paper better, I have brought Figure 2 to the front (before a Supplementary Figure), and in the first part of Results I have now added:

“Figure 2 shows a TGM for an example subject, where some archetypal characteristics are highlighted. In the experiments below, specifically, I focus on the strong narrow diagonal at the beginning of the trial, the broadening of accuracy later in the trial, and the vertical/horizontal bars of higher-than-chance accuracy. Importantly, this specific example in Figure 2 is only meant as a reference, and therefore I did not optimise the model hyperparameters to this TGM (except in the last subsection), or showed any quantitative metric of similarity. ”

I have enunciated the goals more clearly in the Introduction:

“To gain insight into what aspects of the signal underpin decoding accuracy, and therefore the most stable aspects of stimulus processing, …”

Relatedly, it would be good to connect the various signal properties to possible neurobiological mechanisms. I appreciate that the author tries to remain neutral on this in the introduction, but I think it would greatly increase the implications of the analysis if it is made clearer how it could eventually help us understand neural processes.

The Reviewer is right in pointing out that I preferred to remain neutral on this. While I have still kept that tone of neutrality throughout the paper, I have now included the following sentence as an example of a neurobiological question that could be investigated with the model:

“For example, it is not completely clear the extent to which stimulus processing is sustained by oscillations, and disentangling these effects can help resolving this question.”

And, more generally,

“Genephys has different available types of effect, including phase resets, additive damped oscillations, amplitude modulations, and non-oscillatory responses. All of these elements, which may relate to distinct neurobiological mechanisms, are configurable and can be combined to generate a plethora of TGMs that, in turn, can be contrasted to specific empirical TGMs. This way, we can gain insight on what mechanisms might be at play in a given task. ”

Line 57: this sentence is very long, making it hard to follow, could you break up into smaller parts?

Thank you. The sentence is fragmented now.

Please replace angular frequencies with frequencies in Hertz for clarity.

Here I have preferred to stick to angular frequencies because it is more general than if I talk about Hertz, because that would entail having a specific sampling frequency. I think doing so would create confusion precisely of the sorts that I am trying to clarify in this revision: that is, that these results are not specific of one TGM but reflect general features that we see broadly in the literature.

There are quite some types throughout the paper, please recheck

Thank you. I have revised and have made my best to clear them out.

Author Response

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

We are grateful for the comments and suggestions from the reviewers and have followed the recommendation in producing our revised manuscript. We have modified the text and performed additional statistical analysis as detailed below, which we believe has improved the overall manuscript.

Reviewer #1 (Public Review):

Establishing direct links between the neuronal connectivity information of connectomics datasets with circuit physiology and behavior and exciting current research area in neurobiology. Until recently, studies of aggression in Drosophila had been conducted largely in males, and many of the neurons involved in this behavior are male-specific clusters. Since the currently available fly brain connectomes come from female brains, their applicability for the study of the circuitry underlying aggressive behavior is very limited.

The authors have previously used the Janelia hemibrain connectome paired with behavior analysis to show that activating either the aIPg or pC1d cell types can induce short-term aggression in females, while activation of other PC1 clusters (a-c and e) does not. Here they expand on those findings, showing that optogenetic stimulation of aIPg neurons was sufficient to promote an aggressive internal state lasting at least 10 minutes following a 30-second activation. In addition, the authors show that while stimulation of PC1d alone is not sufficient to induce this persistent aggressive state, simultaneous activation of PC1d + PC1e is, suggesting a synergistic effect. Connectomics analysis performed in the authors' previous study had shown that PC1d and aIPg are interconnected. However, silencing pC1d neuronal activity did not reduce aIPg-evoked persistent aggression, indicating that the aggressive state did not depend on pC1d-aIPg recurrent connectivity.

The conclusions are well supported by the data, and the results presented in this manuscript represent an important contribution to our understanding of the neuronal circuitry underlying female aggression.

Reviewer #1 (Recommendations For The Authors):

1. Previously, the authors have shown that the activation of PC1e alone does not induce female aggression. In this study, they investigate the role of aIPg, PC1d, or PC1d+e on aggression persistence, but they do not explore the effect of activation of PC1e alone. It is possible that PC1e activation may not produce an immediate short-term effect but could lead to a gradual increase in aggression over time, potentially explaining at least in part the observed effect upon PC1d+e activation. Incorporating an examination of the long-term impact of PC1e activation on aggression could provide valuable information.

We did perform mixed pair experiments with the pC1e-SS1 line from the Schretter et al. (2020) paper and did not find any significant changes in aggression over time in this setup as well. We have now added a reference to these experiments in the revised submission in lines 135 to 136.

1. Some important controls are missing: flies with the genetic combinations employed in the activation experiments shown in Figure 2 but in the absence of activation and under the exact same conditions and for a similar observation period.

For Figure 2, we used an empty split-Gal4 driver as a genetic control for our activation paradigms. As these flies contain the same number of copies of mini-white while not labeling the targeted cell types, we believe that they provide an appropriate control for these experiments. The control information is specified in all figure legends as well.

1. The quantification shown in Fig 3- Supplementary Figure 1 shows no effect during stimulation (13 s + 15s), but based on the plots of Figure 3, there may be an effect of silencing PC1d on aIPg-induced aggression during the initial 13 second period. Those two time periods (13 s vs 15 s) could be quantified separately to determine if this is the case.

We examined the two stimulation periods separately and did not find any significant differences in either period (13s period, p = 0.2978; 15s period, p = 0.6650). We have now added this into the figure legend for Figure 3 and Figure 3 supplement 1.

1. Expression of Kir2.1 in pC1d neurons while aIPg neurons were activated did not suppress aggression after aIPg stimulation, suggesting that connections from pC1d neurons are not necessary for the persistent aggressive state promoted by aIPg. Since previously the authors have shown that TNT-mediated inhibition of aIPg reduces aggression, the reciprocal experiment would be informative: determining if stimulation of PC1d+e no longer produces persistent aggression when aIPg neurons are silenced.

In this manuscript, we were primarily testing if the connections from aIPg to pC1d were necessary for the persistent aggressive state induced by aIPg activation. Therefore, we believe the suggested experiment is beyond the scope of the current manuscript.

1. How many times was each experiment repeated? This is important information and should be in the methods section for each type of experiment or in each figure legend.

We have now added this information in the appropriate figure legends.

1. Determining the effect on persistent aggression of silencing sNPF (for example via RNAi or Crispr-Cas9 mediated mutagenesis) in aIPG neurons would be an important addition to the manuscript. If peptidergic signaling is underlying the persistence phenotype of aIPg neurons, that would explain why the recurrent connectivity found between those cells and the PC1 cluster does not play a role.

We agree with the reviewer that this would be a logical next step in extending this work.

Reviewer #2 (Public Review):

The mechanisms that mediate female aggression remain poorly understood. Chiu, Schretter, and colleagues, employed circuit dissection techniques to tease apart the specific roles of particular doublesex and fruitless expressing neurons in the fly Drosophila in generating a persistent aggressive state. They find that activating the fruitless positive alPg neurons, generated an aggressive state that persisted for >10min after the stimulation ended. Similarly, activating the doublesex positive pC1de neurons also generated a persistent state. Activating pC1d or pC1e individually did not induce a persistent state. Interestingly, while neural activation of alPGs and pC1d+e neurons induced persistent behavioural states it did not induce persistent activity in the neurons being activated.

The conclusions of this paper are well supported by the data, there were only a few points where clarification might help:

1. Figure 3 is a little confusing. This is a circuit behavioural epistasis experiment where the authors activate alPg with CsChrimson while inhibiting pC1d with Kir2.1. In Fig. 2 flies were separated for 10 min following stimulation which allowed for identification of a persistent state. However, in Fig 3 it appears as if flies were allowed to freely interact during and immediately post-stimulation. It is unclear why flies were not separated as in Fig. 2, which makes it difficult to compare the two results. Some discussion of this point would help. Also, from the rasters it appears as if inhibition of pC1d reduced aggression induced by alPg during the stimulation period. Is this true?

We thank the reviewer for pointing out the need for clarification and we have modified the legend in Figure 3 to address the points raised. The flies were allowed to freely interact during the experiments shown in Figure 3 and we have added this information to the figure legend. To obtain a high level of aggressive behavior that would make it easier to observe a suppression of aggression, the epistasis experiments were performed with freely moving same-genotype pairs. The level of aggression triggered by the generation 1 LexA line labeling aIPg was lower than that observed when using with the aIPg-SS GAL4 line. The experiment was performed as in Schretter et al. (2020) where we found that aIPg activation induced persistent fighting in same genotype pairs. We have added a brief explanation in lines 152 to 155.

Inhibition of pC1d does not significantly reduce the overall aggression induced by aIPg stimulation in the 13s + 15s period. We also examined the differences within the two stimulation periods and did not find any significant differences (13s period, p = 0.2978; 15s period, p = 0.6650). We have now added this information to the figure legends for Figure 3 and Figure 3 supplement 1.

1. pC1e neurons also have recurrent connectivity with alPg neurons. It might help to also discuss the potential role of this arm of the microcircuit.

We thank the review for this suggestion. The number of synapses that aIPg sends back to pC1e is a very low proportion of its total output (0.177%). However, based on the experiments that we have performed, we cannot rule out that this microcircuit might contribute to maintaining persistence. We have added this point into the discussion in lines 210 to 211.

Reviewer #2 (Recommendations For The Authors):

1. Line 129-130: A citation for group-housed flies showing lower aggression would be helpful.

We have now added in the reference to Chiu et al. (2021), as they showed this effect for females, in line 130.

1. Figure 2 - figure supplement 1: In the legend, change "when pC1d neurons were stimulation" to "when pC1d neurons were stimulated".

We thank the reviewer for finding this error and have now corrected this.

Reviewer #3 (Public Review):

Two studies published in 2020 independently identified the alPg, pC1d, and pC1e neurons to be involved in initiating and maintaining a state of aggression in female Drosophila. Both studies combined behavioural analyses, optogenitic manipulation of neurons, and connectomics. One of these studies proposed that the extensive interconnections seen between the alPg and pC1d+e neurons might represent a recurrent motif known to support persistent behvioural states in other systems. In this manuscript, the authors test this idea and report that their data do not support it. Specifically, they report that alPg or pC1d+e (but not pC1d alone) can initiate a persistent state of aggression. But they find that the persistent aggressive state is maintained even when the pC1d neurons are inactivated. Finally, they show that neither of these neurons themselves sustains neuronal activity upon stimulation, nor do either of them induce a persistent activity in the other. Together, their data suggest that the recurrent connection between alPg and pC1d is not what supports the persistent state. The data underlying these claims are convincing. A possibility to explore before ruling out recurrent motifs (at this circuit level) in maintaining aggression is that the connections between alPg and pC1e can compensate for the loss of pC1e. Overall, the study is important and will be of interest to those who study the circuit basis of persistent behavioural states, but also to neuroscientists in general.

Reviewer #3 (Recommendations For The Authors):

I enjoyed reading this manuscript for its clarity in writing and data presentation.

I would like the authors to comment on the possibility that pC1e can compensate for the loss of pC1d. It is possible that if they silence both pC1d+e in the context of alPg activation, the persistent aggression is lost?

We agree with the reviewer that this is an intriguing hypothesis. In order to examine if pC1e does compensate for pC1d, we would need to also activate pC1e while inhibiting pC1d. However, such an experiment is not currently possible as we do not have a LexA line that specifically labels either pC1d or pC1e alone.

For the pC1d+e silencing experiments, we were primarily testing to see if the most prominent recurrent connection, which is between pC1d and aIPg, was responsible for the behavioral persistence. We agree with the reviewer that this would be a logical follow up experiment to be performed in the future.

Have the authors looked for activity in the pC1e neuron upon simulation of alPg? (Deutsch et al 2020 observed many regions in the brain that maintained sustained activity upon pC1d+e stimulation.)

We have not examined this activity. We agree that this would be a good follow up experiment; however, we believe it is beyond the scope of the current work.

Would the more appropriate experiment in Figure 4c be the co-stimulation of pC1d+e while imaging from alPg?

For these experiments, we were testing to see if the most prominent recurrent connection, which is between pC1d and aIPg, was responsible for the behavioral persistence. We agree with the reviewer that this would be a good follow up experiment

Author Response

Reviewer #1 (Public Review):

Summary:

This study uses whole genome sequencing to characterise the population structure and genetic diversity of a collection of 58 isolates of E. coli associated with neonatal meningitis (NMEC) from seven countries, including 52 isolates that the authors sequenced themselves and a further 6 publicly available genome sequences. Additionally, the study used sequencing to investigate three case studies of apparent relapse. The data show that in all three cases, the relapse was caused by the same NMEC strain as the initial infection. In two cases they also found evidence for gut persistence of the NMEC strain, which may act as a reservoir for persistence and reinfection in neonates. This finding is of clinical importance as it suggests that decolonisation of the gut could be helpful in preventing relapse of meningitis in NMEC patients.

Strengths:

The study presents complete genome sequences for n=18 diverse isolates, which will serve as useful references for future studies of NMEC. The genomic analyses are high quality, the population genomic analyses are comprehensive and the case study investigations are convincing.

We agree

Weaknesses:

The NMEC collection described in the study includes isolates from just seven countries. The majority (n=51/58, 88%) are from high-income countries in Europe, Australia, or North America; the rest are from Cambodia (n=7, 12%). Therefore it is not clear how well the results reflect the global diversity of NMEC, nor the populations of NMEC affecting the most populous regions.

The virulence factors section highlights several potentially interesting genes that are present at apparently high frequency in the NMEC genomes; however, without knowing their frequency in the broader E. coli population it is hard to know the significance of this.

We acknowledged the limitations of our NMEC collection in the Discussion. We agree the prevalence of virulence factors in our collection is interesting. The limited size of our collection prevented further evaluation of the prevalence of these virulence factors in a broader E. coli population.

Reviewer #2 (Public Review):

Summary:

In this work, the authors present a robust genomic dataset profiling 58 isolates of neonatal meningitis-causing E. coli (NMEC), the largest such cohort to be profiled to date. The authors provide genomic information on virulence and antibiotic resistance genomic markers, as well as serotype and capsule information. They go on to probe three cases in which infants presented with recurrent febrile infection and meningitis and provide evidence indicating that the original isolate is likely causing the second infection and that an asymptomatic reservoir exists in the gut. Accompanying these results, the authors demonstrate that gut dysbiosis coincides with the meningitis.

Strengths:

The genomics work is meticulously done, utilizing long-read sequencing.

The cohort of isolates is the largest to be sampled to date.

The findings are significant, illuminating the presence of a gut reservoir in infants with repeating infection.

We agree

Weaknesses:

Although the cohort of isolates is large, there is no global representation, entirely omitting Africa and the Americas. This is acknowledged by the group in the discussion, however, it would make the study much more compelling if there was global representation.

We agree. In the Discussion we state this is likely a reflection of the difficulty in acquiring isolates causing neonatal meningitis, in particular from countries with limited microbiology and pathology resources.

Reviewer #3 (Public Review):

Summary:

In this manuscript, Schembri et al performed a molecular analysis by WGS of 52 E. coli strains identified as "causing neonatal meningitis" from several countries and isolated from 1974 to 2020. Sequence types, virulence genes content as well as antibiotic-resistant genes are depicted. In the second part, they also described three cases of relapse and analysed their respective strains as well as the microbiome of three neonates during their relapse. For one patient the same E. coli strain was found in blood and stool (this patient had no meningitis). For two patients microbiome analysis revealed a severe dysbiosis.

Major comments:

Although the authors announce in their title that they study E. coli that cause neonatal meningitis and in methods stipulate that they had a collection of 52 NMEC, we found in Supplementary Table 1, 29 strains (therefore most of the strains) isolated from blood and not CSF. This is a major limitation since only strains isolated from CSF can be designated with certainty as NMEC even if a pleiocytose is observed in the CSF. A very troubling data is the description of patient two with a relapse infection. As stated in the text line 225, CSF microscopy was normal and culture was negative for this patient! Therefore it is clear that patient without meningitis has been included in this study.

We have reviewed the clinical data for our 52 NMEC isolates, noting that for some of the older Finish isolates we relied on previous publications. This data is shown in Table S1. To address the Reviewer’s comment, we have added the following text to the methods section (new text underlined).

‘The collection comprised 42 isolates from confirmed meningitis cases (29 cultured from CSF and 13 cultured from blood) and 10 isolates from clinically diagnosed meningitis cases (all cultured from blood).’

Patient 2 was initially diagnosed with meningitis based on a positive blood culture in the presence of CSF pleocytosis (>300 WBCs, >95% polymorphs). We understand there may be some confusion with reference to a relapsed infection, which we now more accurately describe as recrudescent invasive infection in the revised manuscript.

Another major limitation (not stated in the discussion) is the absence of clinical information on neonates especially the weeks of gestation. It is well known that the risk of infection is dramatically increased in preterm neonates due to their immature immunity. Therefore E. coli causing infection in preterm neonates are not comparable to those causing infection in term neonates notably in their virulence gene content. Indeed, it is mentioned that at least eight strains did not possess a capsule, we can speculate that neonates were preterm, but this information is lacking. The ages of neonates are also lacking. The possible source of infection is not mentioned, notably urinary tract infection. This may have also an impact on the content of VF.

We agree. In the Discussion we now note the following (new text underlined):

‘… we did not have clinical data on the weeks of gestation for all patients, and thus could not compare virulence factors from NMEC isolated from preterm versus term infants.’

Submission to Medrxiv, a requirement for review of our manuscript at eLife, necessitated the removal of some patient identifying information, including precise age and detailed medical history.

Sequence analysis reveals the predominance of ST95 and ST1193 in this collection. The high incidence of ST95 is not surprising and well previously described, therefore, the concluding sentence line 132 indicating that ST95 E. coli should exhibit specific virulence features associated with their capacity to cause NM does not add anything. On the contrary, the high incidence of ST1193 is of interest and should have been discussed more in detail. Which specific virulence factors do they harbor? Any hypothesis explaining their emergence in neonates?

We compared the virulence factors of ST95 and ST1193 and summarized this information in Figure 4. We also discussed how the K1 polysialic acid capsule in ST95 and ST1193 could contribute to the emergence of these STs in NM. Specifically, we stated the following: ‘We speculate this is due to the prevailing K1 polysialic acid capsule serotype found in ST95 and the newly emerged ST1193 clone [22, 37] in combination with other virulence factors [15, 28, 29] (Figure 4) and the immature immune system of preterm infants.’

In the paragraph depicted the VF it is only stated that ST95 contained significantly more VF than the ST1193 strains. And so what? By the way "significantly" is not documented: n=?, p=?

We compared the prevalence of known virulence factors between ST95 and ST1193, and showed that ST95 strains in our collection contained significantly more virulence factors than the ST1193 strains. The P-value and the statistical test used were included in Supplementary Figure 3. To address the reviewers concern, we have now also added this to the main manuscript text as follows (new text underlined):

‘Direct comparison of virulence factors between ST95 and ST1193, the two most dominant NMEC STs, revealed that the ST95 isolates (n = 20) contained significantly more virulence factors than the ST1193 isolates (n=9), p-value < 0.001, Mann-Whitney two-tailed unpaired test (Supplementary Table 1, Supplementary Figure 3).’

The complete sequence of 18 strains is not clear. Results of Supplementary Table 2 are presented in the text and are not discussed.

NMEC isolates that were completely sequenced in this study are indicated in bold and marked with an asterisk in Figure 1. This information is indicated in the figure legend and was provided in the original submission. All information regarding genomic island composition and location, virulence genes and plasmid and prophage diversity is included in Supplementary Table 2. This information is highly descriptive and thus we elected not to include it as text in the main manuscript.

46 years is a very long time for such a small number of strains, making it difficult to put forward epidemiological or evolutionary theories. In the analysis of antibiotic resistance, there are no ESBLs. However, Ding's article (reference 34) and other authors showed that ESBLs are emerging in E. coli neonatal infection. These strains are a major threat that should be studied, unfortunately, the authors haven't had the opportunity to characterize such strains in their manuscript.

We agree 46 years is a long time-span. The study by Ding et al examined 56 isolates comprised of 25 different STs isolated in China from 2009-2015, with ST1193 (n=12) and ST95 (n=10) the most common. Our study examined 58 isolates comprised of 22 different STs isolated in seven different geographic regions from 1974-2020, with ST1193 (n=9) and ST95 (n=20) the most common. Thus, despite differences in the geographic regions from which isolates in the two studies were sourced, there are similarities in the most common STs identified. The fact that we observed less antibiotic resistance, including a lack of ESBL genes, in ST1193 is likely due to the different regions from which the isolates were sourced. We acknowledged and discussed the potential of ST1193 harbouring multidrug resistance including ESBLs in our manuscript as follows:

‘Concerningly, the ST1193 strains examined here carry genes encoding several aminoglycoside-modifying enzymes, generating a resistance profile that may lead to the clinical failure of empiric regimens such as ampicillin and gentamicin, a therapeutic combination used in many settings to treat NM and early-onset sepsis [35, 36]. This, in combination with reports of co-resistance to third-generation cephalosporins for some ST1193 strains [22, 34], would limit the choice of antibiotic treatment.’

Second part of the manuscript:

The three patients who relapsed had a late neonatal infection (> 3 days) with respective ages of 6 days, 7 weeks, and 3 weeks. We do not know whether they are former preterm newborns (no term specified) or whether they have received antibiotics in the meantime.

As noted above, patient ages were not disclosed to comply with submission to Medrxiv, a requirement for review of our manuscript at eLife.

Patient 1: Although this patient had a pleiocytose in CSF, the culture was negative which is surprising and no explanation is provided. Therefore, the diagnosis of meningitis is not certain. Pleiocytose without meningitis has been previously described in neonates with severe sepsis. Line 215: no immunological abnormalities were identified (no details are given).

We respectfully disagree with the reviewer. The diagnosis of meningitis is made unequivocally by the presence of a clearly abnormal CSF microscopy (2430 WBCs) and an invasive E. coli from blood culture. This does not seem controversial to the authors. We had believed it unnecessary to include this corroborative evidence, but have added the following to support our assertion:

‘The child was diagnosed with meningitis based on a cerebrospinal fluid (CSF) pleocytosis (>2000 white blood cells; WBCs, low glucose, elevated protein), positive CSF E. coli PCR and a positive blood culture for E. coli (MS21522).’

On the contrary, the authors are surprised by the statement that CSF pleocytosis occurs in neonatal sepsis ‘without meningitis’ and do not know of any definitions of neonatal meningitis that are not tied to the presence of a CSF pleocytosis. Furthermore, the later isolation of E. coli from the CSF during the relapsed infection re-enforces the initial diagnosis.

Patient 2: This patient had a recurrence of bacteremia without meningitis (line 225: CSF microscopy was normal and culture negative!). This case should be deleted.

In a similar vein to the previous comment, we respectfully assert that this patient has clear evidence of meningitis (330 WBCs in the CSF, taken 24h after initiation of antibiotic treatment). In this case, molecular testing was not performed as, under the principle of diagnostic stewardship, it was not considered necessary by the clinical microbiologists and treating clinicians following the culture of E. coli in the bloodstream. We agree that this is not a case of recurrent meningitis, but our intention was to highlight the recrudescence of an invasive infection (urinary sepsis requiring admission to hospital and intravenous antibiotics) which we hypothesise has arisen from the intestinal reservoir. We did not state that all patients suffered from relapsed meningitis.

Despite this, to address this reviewers concern, we have changed all reference to ‘relapsed infection’ to now read ‘recrudescent invasive infection’ in the revised manuscript.

Patient 3: This patient had two relapses which is exceptional and may suggest the existence of a congenital malformation or a neurological complication such as abscess or empyema therefore, "imaging studies" should be detailed.

This patient underwent extensive imaging investigation to rule out a hidden source. This included repeated MRI imaging of head and spine, CT imaging of head and chest, USS imaging of abdomen and pelvis and nuclear medicine imaging to detect a subtle meningeal defect and CSF leak. All tests were normal, and no abscess or empyema found.

We have modified the text to include this information:

Text in original submission: ‘Imaging studies and immunological work-up were normal.’

New text in revised manuscript (underlined): ‘Extensive imaging studies including repeated MRI imaging of the head and spine, CT imaging of the head and chest, ultrasound imaging of abdomen and pelvis, and nuclear medicine imaging did not show a congenital malformation or abscess. Immunological work-up did not show a known primary immunodeficiency. At two years of age, speech delay is reported but no other developmental abnormality.’

The authors suggest a link between intestinal dysbiosis and relapse in three patients. However, the fecal microbiomes of patients without relapse were not analysed, so no comparison is possible. Moreover, dysbiosis after several weeks of antibiotic treatment in a patient hospitalized for a long time is not unexpected. Therefore, it's impossible to make any assumption or draw any conclusion. This part of the manuscript is purely descriptive. Finally, the authors should be more prudent when they state in line 289 "we also provide direct evidence to implicate the gut as a reservoir [...] antibiotic treatment". Indeed the gut colonization of the mothers with the same strain may be also a reservoir (as stated in the discussion line 336). Finally, the authors do not discuss the potential role of ceftriaxone vs cefotaxime in the dysbiosis observed. Ceftriaxone may have a major impact on the microbiota due to its digestive elimination.

We addressed the limitations of our study in the Discussion, including that we did not have access to urine or stool samples from the mother of the infants that suffered recrudescence, and thus cannot rule out mother-to-child transmission as a mechanism of reinfection. We have now added that we did not have clinical data on the weeks of gestation for all patients, and thus could not compare virulence factors from NMEC isolated from preterm versus term infants. The limitations of our study are summarised as follows in the Discussion (new text underlined):

‘This study had several limitations. First, our NMEC strain collection was restricted to seven geographic regions, a reflection of the difficulty in acquiring strains causing this disease. Second, we did not have access to a complete set of stool samples spanning pre- and post-treatment in the patients that suffered NM and recrudescent invasive infection. This impacted our capacity to monitor E. coli persistence and evaluate the effect of antibiotic treatment on changes in the microbiome over time. Third, we did not have access to urine or stool samples from the mother of the infants that suffered recrudescence, and thus cannot rule out mother-to-child transmission as a mechanism of reinfection. Finally, we did not have clinical data on the weeks of gestation for all patients, and thus could not compare virulence factors from NMEC isolated from preterm versus term infants.’

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Shibl et al., studied the possible role of dicarboxylate metabolite azelaic acid (Aze) in modulating the response of different bacteria, it was used as a carbon source by Phycobacter and possibly toxic for Alteromonas. The experiments were well conducted using transcriptomics, transcriptional factor coexpression networks, uptake experiments, and chemical methods to unravel the uptake, catabolism, and toxicity of Aze on these two bacteria. They identified a putative Aze TRAP transporter in bacteria and showed that Aze is assimilated through fatty acid degradation in Phycobacter. Meanwhile, in Alteromonas it is suggested that Aze inhibits the ribosome and/or protein synthesis, and that efflux pumps shuttles Aze outside the cytoplasm. Further on, they demonstrate that seawater amended with Aze selects for microbes that can catabolize Aze.

Major strengths:

The manuscript is well written and very clear. Through the combination of gene expression, transcriptional factor co-expression networks, uptake experiments, and chemical methods Shibl et al., showed that Aze has a different response in two bacteria.

Major weakness:

There is no confirmation of the Aze TRAP transporters through mutagenesis.

Impact on the field:

Metabolites exert a significant influence on microbial communities in the ocean, playing a crucial role in their composition, dynamics, and biogeochemical cycles. This research highlights the intriguing capacity of a single metabolite to induce contrasting responses in distinct bacterial species, underscoring its role in shaping microbial interactions and ecosystem functions.

We thank the reviewer for their comments on the paper and we appreciate their suggestion to confirm the activity of Aze TRAP transporters through mutagenesis. We agree that this would be a valuable addition to the study, and we mention in the text that “Despite numerous attempts, our efforts to knock-out azeTSL in Phycobacter failed.”

The success rate of mutagenesis experiments is often low and time-consuming. There are a few reasons why our knock-out experiments with Phycobacter have not been successful. Despite using several modified protocols for electroporation, no Phycobacter colonies grew on the antibiotic plate. We then tried the homologous recombination approach for conjugation but were not successful in selecting for Phycobacter cells, even when grown in high salinity conditions that favor Phycobacter and disfavor the carrier, E. coli . While we would love to include a mutagen to confirm the function of this cluster, the task seems to be unattainable at the moment .

Reviewer #2 (Public Review):

This study explores the breadth of effects of one important metabolite, azelaic acid, on marine microbes, and reveals in-depth its pathway of uptake and catabolism in one model bacterial strain. This compound is known to be widely produced by phytoplankton and plants, and to have complex effects on associated microbiomes.

This work uses transcriptomics to assay the response of two strains that show contrasting responses to the metabolite: one catabolizes the compound and assimilates the carbon, while the other shows growth inhibition and stress response. A highly induced TRAP transporter, adjacent to a previously identified regulator, is inferred to be the specific uptake system for azelaic acid. However the transport function was not directly tested via genetic or biochemical methods. Nevertheless, this is a significant finding that will be useful for exploring the distribution of azelaic acid uptake capability across metagenomes and other bacteria.

The authors use pulse-chase style metabolomics experiments to beautifully demonstrate the fate of azelaic acid through catabolic pathways. They also measure an assimilation rate per cell, though it remains unclear how this measured rate relates to natural systems. The metabolomics approach is an elegant way to show carbon flux through cells, and could serve as a model for future studies.

The study seeks to extend the results from two model strains to complex communities, using seawater mesocosm experiments and soil/Arabidopsis experiments. The seawater experiments show a community shift in mesocosms with added azelaic acid. However, the mechanisms for the shift were not determined; further work is necessary to demonstrate which community members are directly assimilating the compound vs. benefitting indirectly or experiencing inhibition. In my opinion the soil and Arabidopsis experiments are quite preliminary. I appreciate the authors' desire to broaden the scope beyond marine systems, but I believe any conclusions regarding different modes of action in aquatic vs terrestrial microbial communities are speculative at this stage.

This work is a nice illustration of how we can begin to tease apart the effects of chemical currencies on marine ecosystems. A key strength of this work is the combination of transcriptomics and metabolomics methods, along with assaying the impacts of the metabolite on both model strains of bacteria and whole communities. Given the sheer number of compounds that probably play critical roles in community interactions, a key challenge for the field will be navigating the tradeoffs between breadth and depth in future studies of metabolite impacts. This study offers a good compromise and will be a useful model for future studies.

We thank the reviewer for their thoughtful comments on the manuscript. We appreciate their feedback on the breadth of effects of Aze on marine microbes, and their insights into the strengths and limitations of our study.

We agree that the specific mechanisms underlying community-level shifts in seawater mesocosm experiments with added Aze are not yet fully understood and we believe such work is beyond the scope of this paper and warrants an in-depth study of its own. This can perhaps be conducted at a larger scale by using a combination of meta-omics and targeted enrichment to identify the community members directly assimilating Aze, as well as those that are benefitting indirectly or experiencing inhibition.

We also agree that the soil and Arabidopsis experiments are exploratory. However, we believe that these experiments are a valuable first step in highlighting the potential for Aze to have different modes of action in aquatic versus terrestrial microbial communities. Our interest in contrasting bacterial molecular responses in terrestrial plant rhizospheres and marine algal phycospheres stems from the fact that both environments share similar molecules and related bacteria, yet exhibit significantly different evolutionary histories and fluid dynamic profiles (Seymour et al 2017, Nature Microbiol ). Although more is known about Aze in Arabidopsis than phytoplankton, there are still gaps in this knowledge. For example, recent work has shown that Aze and derivatives can be secreted into soil (Korenblum et al 2020, PNAS ), but whether Aze directly influences microbial communities in soil as we have shown in seawater has not been explored. Thus, we feel our preliminary experiments in soil are important to provide such a distinction with seawater. Additional studies in these systems to further investigate the importance of Aze, which were beyond the scope of this current work, would be quite beneficial.

Reviewer #1 (Recommendations For The Authors):

General comments:

A complete supplemental file of differentially expressed genes should be provided in the supplemental. Please add tables with the entire DESeq output for Aze additions in the genomes of Phycobacter (0.5 and 8 h) and Alteromonas (0.5 h). While it makes sense to focus the paper on Aze related genes, the full dataset should be made available in a more curated form than just the raw reads in the SRA.

We thank the reviewer for this suggestion. We have included three more sheets in Supplementary Table 1 file where readers can find the entire DESeq outputs of Phycobacter (0.5 and 8 h) and Alteromonas (0.5 h) experiments.

Specific comments:

• L82 indicates the TRAP transporter for Aze. Looking at the table for gene expression of Phycobacter there are 26 significantly enriched transport genes at 0.5 h other than the putative Aze TRAP transporter. Even though the TRAP transporter is likely transporting Aze, it would be good to let the readers know that other transporters showed transcript enrichment.

Thank you for this helpful comment. We modified the sentence accordingly to read as follows: “Among 26 enriched transporter genes in our dataset, a C 4 - dicarboxylate tripartite ATP-independent periplasmic (TRAP) transporter substrate-binding protein (INS80_RS11065) was the most and the third most upregulated gene in Phycobacter grown on Aze at 0.5 and 8 hours, respectively.”

• Figure 1: There are many genes enriched from -1 to 1. Is there a cut off, p-val (can you add it to the caption)? It would be good to have a dashed line or something that indicates the -1 and 1 log2 fold change in the figure.

We thank the reviewer for this suggestion. We added the following sentence to the legend of Fig. 1: “Genes were considered DE with a p -adjusted value of < 0.05 and a log2 fold-change of ≥ ±0.50.”

• Supplementary tables: Add a title on all the supplementary tables. It's hard to tell what each one of the tables means without looking at the text and content of each tables.

A short descriptive title is now added to all supplementary tables.

• Not sure if it matters, though Table S1 was not available in the attached files, though it is in the complete pdf.

Table S1 is now in the attached files and the DESeq output has been added to it as suggested in the general comment above.

Reviewer #2 (Recommendations For The Authors):

Here I offer some more specific suggestions and comments on the methods and presentation.

I recommend being careful throughout with the language regarding conclusions. For instance, the study does not directly demonstrate the activity of the TRAP transporter (as mentioned above), and does not directly demonstrate that the bacteria that increase in abundance in the mesocosm experiments are actually assimilating azelaic acid.

We thank the reviewer for this comment. We agree that further studies are required to get definitive answers regarding the direct activity of the transporter genes and direct assimilation of Aze by bacteria in the mesocosm. These complex experiments would require establishing a reproducible workflow for knocking out genes and further isotope labeling experiments to track Aze assimilation in a natural setting. To that end, we were keen on using language throughout the manuscript indicating that transporter activity is putative. We went through the manuscript again to make sure it was clear that the transporter activity is putative at this time and is not confirmed. For the mesocosms, we cannot rule out that the changes in community structure is not due to other factors besides Aze. We have added this sentence in the discussion of the mesocosm experiments to indicate that the observed changes in microbial community cannot be directly attributed to Aze activity and may be a byproduct of other mechanisms.

Additionally, I find the soil and plant experiments to be very preliminary, and would personally recommend removing them from the manuscript. This is of course the authors' choice, but I find they detract from an otherwise more solid story. I wonder whether 16 hours was sufficient to see community changes and whether adding azelaic acid directly into the plant is necessary or relevant. The study does not measure any plant immune responses so I caution against drawing conclusions about the mechanism. It seems the connection to plant immunity was already shown in the literature, in which case I'm not sure whether these experiments presented here really add anything new to the paper.

We thank the reviewer for these comments. Our 16-hour sampling time point (similar to the seawater experiment) represents an overnight incubation period that should allow sufficient change in the natural microbial composition yet avoids the long-term succession of microbes with high metabolic capacities that may outcompete the rest of the community at long incubation periods. Deciding on this length of incubation was also informed by the uptake rate of Aze and its influence on either bacteria assimilating it as a carbon source or being inhibited by it.

Since no significant changes were observed in the soil, it was necessary to test the hypothesis that the plant host might be indirectly influencing the rhizosphere microbial communities by infiltrating A. thaliana leaves with Aze. As the reviewer mentions, the association between Aze and plant immunity was previously shown; however, the overall influence on the microbial community has not been fully explored yet. The soil and plant experiments were meant to serve an exploratory purpose and we find them necessary to keep in the manuscript as a first step in comparing the mode of action of Aze within marine and terrestrial ecosystems. They are by no means the answer to what role Aze plays in soil systems, but rather they are the starting point. We hope that our results encourage some readers to investigate similar common metabolites to further elucidate the molecular underpinnings of microbial modulation in both environments.

Regarding the transcriptomics data, I am not clear on why the "expression ratio" -- i.e. the fraction of pathway genes that were differentially abundant -- was used. I would not expect all transcripts in a pathway to behave the same way in response to a perturbation, due to variation in half-life/stability, post-transcriptional and post-translational regulation, etc. I recommend removing the expression ratio (right panel) from Figure 1. The left panel shows the data more clearly and more directly.

We thank the reviewer for their insight and we agree that not all transcripts in a pathway behave the same way. However, we find the expression ratio panel visually informative to highlight the importance of a pathway in response to Aze, taking into consideration the total number of key genes involved in a pathway. For example, despite the larger number of DE genes associated with the Amino Acid Metabolism & Degradation pathway compared to the Fatty Acid Degradation pathway, the expression ratio for the former in each transcriptome is lower than its Fatty Acid Degradation counterpart, indicating that the response of key fatty acid degradation genes to Aze is more pronounced. We have qualified the reasons for including expression ratios in Figure 1 legend.

Overall I enjoyed reading the manuscript and applaud the authors on a nice contribution to this important field.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Peng et al develop a computational method to predict/rank transcription factors (TFs) according to their likelihood of being pioneer transcription factors--factors that are capable of binding nucleosomes--using ChIP-seq for 225 human transcription factors, MNase-seq and DNase-seq data from five cell lines. The authors developed relatively straightforward, easy to interpret computational methods that leverage the potential for MNase-seq to enable relatively precise identification of the nucleosome dyad. Using an established smoothing approach and local peak identification methods to estimate positions together with identification of ChIP-seq peaks and motifs within those peaks which they referred to as "ChIP-seq motifs", they were able to quantify "motif profiles" and their density in nucleosome regions (NRs) and nucleosome depleted regions (NDRs) relative to their estimated nucleosome dyad positions. Using these profiles, they arrived at an odd-ratio based motif enrichment score along with a Fisher's exact test to assess the odds and significance that a given transcription factor's ChIP-seq motifs are enriched in NRs compared to NDRs, hence, its potential to be a pioneer transcription factor. They showed that known pioneer transcription factors had among the highest enrichment scores, and they could identify a number of relatively novel pioneer TFs with high enrichment scores and relatively high expression in their corresponding cell line. They used multiple validation approaches including (1) calculating the ROC-AUC and Matthews correlation coefficient (MCC) and generating ROC and precision-recall curves associated with their enrichment score based on 32 known pioneer TFs among their 225 TFs which they used as positives and the remaining TFs (among the 225) as negatives; (2) use of the literature to note that known pioneer TFs that acted as key regulators of embryonic stem cell differentiation had a highest enrichment scores; (3) comparison of their enrichment scores to three classes of TFs defined by protein microarray and electromobility shift assays (1. strong binder to free and nucleosomal DNA, 2. weak binder to free and nucleosomal DNA, 3. strong binding to free but not nucleosomal DNA); and (4) correlation between their calculated TF motif nucleosome end/dyad binding ratio and relevant data from an NCAP-SELEX experiment. They also characterize the spatial distribution of TF motif binding relative to the dyad by (1) correlating TF motif density and nucleosome occupancy and (2) clustering TF motif binding profiles relative to their distance from the dyad and identifying 6 clusters.

The strengths of this paper are the use of MNase-seq data to define relatively precise dyad positions and ChIP-seq data together with motif analysis to arrive at relatively accurate TF binding profiles relative to dyad positions in NRs as well as in NDRs. This allowed them to use a relatively simple odds ratio based enrichment score which performs well in identifying known pioneer TFs. Moreover, their validation approaches either produced highly significant or reasonable, trending results.

The weaknesses of the paper are relatively minor, and the authors do a good job describing the limitations of the data and approach.

Reviewer #2 (Public Review):

In this study, the authors utilize a compendium of public genomic data to identify transcription factors (TF) that can identify their DNA binding motifs in the presence of nuclosome-wrapped chromatin and convert the chromatin to open chromatin. This class of TFs are termed Pioneer TFs (PTFs). A major strength of the study is the concept, whose premise is that motifs bound by PTFs (assessed by ChIP-seq for the respective TFs) should be present in both "closed" nucleosome wrapped DNA regions (measured by MNase-seq) as well as open regions (measured by DNAseI-seq) because the PTFs are able to open the chromatin. Use of multiple ENCODE cell lines, including the H1 stem cell line, enabled the authors to assess if binding at motifs changes from closed to open. Typical, non-PTF TFs are expected to only bind motifs in open chromatin regions (measured by DNaseI-seq) and not in regions closed in any cell type. This study contributes to the field a validation of PTFs that are already known to have pioneering activity and presents an interesting approach to quantify PTF activity.

For this reviewer, there were a few notable limitations. One was the uncertainty regarding whether expression of the respective TFs across cell types was taken into account. This would help inform if a TF would be able to open chromatin. Another limitation was the cell types used. While understandable that these cell types were used, because of their deep epigenetic phenotyping and public availability, they are mostly transformed and do not bear close similarity to lineages in a healthy organism. Next, the methods used to identify PTFs were not made available in an easy-to-use tool for other researchers who may seek to identify PTFs in their cell type(s) of interest. Lastly, some terms used were not define explicitly (e.g., meaning of dyads) and the language in the manuscript was often difficult to follow and contained improper English grammar.

Reviewer #3 (Public Review):

Peng et al. designed a computational framework for identifying pioneer factors using epigenomic data from five cell types. The identification of pioneer factors is important for our understanding of the epigenetic and transcriptional regulation of cells. A computational approach toward this goal can significantly reduce the burden of labor-intensive experimental validation.

The authors have addressed my previous comments.

The main issue identified in this re-review is based on the authors' additional experiments to investigate the reproducibility of the pioneer factors identified in the previously analysis that anchored on H1 ESCs.

The additional analysis that uses the other four cell types (HepG2, HeLa-S3, MCF-7, and K562) as anchors reveals the low reproducibility/concordance and high dependence on the selection of anchor cell type in the computational framework. In particular, now several stem cell related TFs (e.g. ESRRB, POU5F1) are ranked markedly higher when H1 ESC is not used as the anchor cell type as shown in Supplementary Figure 5.

Of note, the authors have now removed the shape labels that denote Yamanaka factors in Figure 2c (revised manuscript) that was presented in the main Figure 2a in the initial submission. The NFYs and ESRRB labels in Supplementary 4a are also removed and the boxplot comparing NFYs and ESRRB with other TF are also removed in this figure. Removing these results effectively hides the issues of the computational framework we identified in this revision. Please justify why this was done.

In summary, these new results reveal significant limitations of the proposed computational framework for identifying pioneer factors. The current identifications appear to be highly dependent on the choice of cell types.

Response: We thank all reviewers for their thoughtful and constructive comments and suggestions, which helped us to strengthen our paper. Following the suggestions, we have further addressed the reviewer’s comments and the detailed responses are itemized below.

Reviewer #1 (Recommendations For The Authors):

The following few minor mistakes/discrepancies/omissions should be addressed:

1. In Figure 3, the Nucleosome Occupancy curves and legend are orange and the Binding Motif Profiles are blue; however, the y-axis label for Nucleosome occupancy profile is blue, and the y-axis label of Binding motif profile is orange. The colors seem to be switched, or I'm missing something.

Response: We thank the reviewer for pointing it out. We have changed the colors to make it consistent.

1. The text at the bottom of p. 11 of the main manuscript describing Supplementary Fig. 5 states: "If we repeat our anaysis by redefining differentially open regions as those closed in differentiated cell lines and open in H1 embryonic cell line, then ESSRB and Yamanaka pioneer transcription factor POU5F1 (OCT4) showed significantly higher enrichment scores (Supplementary Figure 5)." However, Supplementary Fig. 5 legend states: "Enrichment analysis of different TFs using the differentially open from one cell line (shown in the title) and conserved open regions from other four cell lines.". These two descriptions of the differential chromatin criteria used in the analysis don't appear to match. The description in the text is the one that makes much more sense to me. The legend should be written a little more clearly and reflect the statement in the main text. One can see from the cut and paste the "analysis" is also misspelled.

Response: We have rewritten the legend of Supplementary Figure 5 to make it clear and consistent. The misspelling has also been corrected.

1. It might be helpful to add that a random classifier would yield a constant precision recall (PR) curve (as a function of Recall) with the Precision = P/(P+N) or the fraction of positives for all plotted PR curves which in the case of Fig. 2a is 32/225 = 0.142, for example.

Response: We thank the reviewer for the suggestions. We have added the fraction of positives for Figure 2.

1. On p. 17 line 513, the authors refer to "Supplementary 7, 9 and 13". I'm assuming it's "Supplementary Tables 7, 9 and 13".

Response: It has been corrected.

1. On p. 18 line 539, "essays" should be "assays".

Response: It has been corrected.

Reviewer #2 (Recommendations For The Authors):

We are satisfied with the revisions in this version of the manuscript.

Reviewer #3 (Public Review):

The main issue identified in this re-review is based on the authors' additional experiments to investigate the reproducibility of the pioneer factors identified in the previously analysis that anchored on H1 ESCs.

The additional analysis that uses the other four cell types (HepG2, HeLa-S3, MCF-7, and K562) as anchors reveals the low reproducibility/concordance and high dependence on the selection of anchor cell type in the computational framework. In particular, now several stem cell related TFs (e.g. ESRRB, POU5F1) are ranked markedly higher when H1 ESC is not used as the anchor cell type as shown in Supplementary Figure 5.

Of note, the authors have now removed the shape labels that denote Yamanaka factors in Figure 2c (revised manuscript) that was presented in the main Figure 2a in the initial submission. The NFYs and ESRRB labels in Supplementary 4a are also removed and the boxplot comparing NFYs and ESRRB with other TF are also removed in this figure. Removing these results effectively hides the issues of the computational framework we identified in this revision. Please justify why this was done.

In summary, these new results reveal significant limitations of the proposed computational framework for identifying pioneer factors. The current identifications appear to be highly dependent on the choice of cell types.

Response: We would like to clarify that our enrichment score used for TF classification, defined by Equation 3, is expected to be cell-type specific. The value of the enrichment score is modulated by a number of factors beyond the property of a TF to act as a PTF, such as the abundance of a given TF in a given cell line, cell type-specific nucleosome binding maps and interactions with other TFs. Thus, it is expected that the enrichment scores calculated for the same TF in different cell lines should be quantitatively different. Following the initial suggestion of Reviewer 3, we have diversified our analysis by using different cell lines as anchors. This analysis showed that most PTFs that we identified could be confirmed based on different cell lines, when comparing the relative enrichment scores within each cell line. On the other hand, it is not expected that the values of enrichment scores of a given TF should be similar across different cell lines.

Regarding a specific comment about ESRRB and POU5F1, these TFs are known pioneer factors with roles in reprogramming of somatic cells into induced pluripotent stem cells and suppressing cell differentiation. They have the ability to open closed chromatin regions in the differentiated cell lines. Therefore, if we redefine the differentially open regions as those closed in differentiated cell lines and open in H1 embryonic cell line, these pioneer factors are expected to have high enrichment scores. Indeed, our new results validated the roles of these PTFs in cell reprogramming. As mentioned above, their enrichment scores in different cell lines are not expected to be the same.

We also would like to clarify that no results were removed during the update of the figures, and all modifications of the manuscript following the suggestions of the reviewers were only made to improve the figures and make them clearer and the message more straightforward.

Author Response

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

We would like to thank the Editors and Reviewers for their additional comments and constructive feedback on our manuscript. We have made minor adjustments to the figures and texts based on their suggestions, including improved images in Figure 1 and correction of figure labels.

Reviewer #1 (Public Review):

In their previous paper (Lari et al, 2019; Azra Lari Arvind Arul Nambi Rajan Rima Sandhu Taylor Reiter Rachel Montpetit Barry P Young Chris JR Loewen Ben Montpetit (2019) A nuclear role for the DEAD-box protein Dbp5 in tRNA export eLife 8:e48410.) as well as in the current manuscript the authors states that Dbp5 is involved in the export of tRNA that is independent of and parallel to Los1. They state that Dbp5 binds to the tRNA independent of known tRNA export proteins. The obtained conclusion is both intriguing and innovative, since it suggests that there are other variables, beyond the ones previously identified as tRNA factors, that might interact with Dbp5 to facilitate the export process. In order to find out additional factors aiding this process the authors may employ total RNA-associated protein purification (TRAPP) experiments ( Shchepachevto et al., 2019; Shchepachev V, Bresson S, Spanos C, Petfalski E, Fischer L, Rappsilber J, Tollervey D. Defining the RNA interactome by total RNA-associated protein purification. Mol Syst Biol. 2019 Apr 8;15(4):e8689. doi: 10.15252/msb.20188689. PMID: 30962360; PMCID: PMC6452921) to identify extra factors involved in conjunction with Dbp5. The process elucidates hitherto uninvestigated tRNA export components that function in conjunction with Dbp5.

Author Response: We greatly appreciate this suggestion and agree with the reviewer that identification of the composition of the export competent Dbp5 containing tRNA complex is a critical next step for understanding the mechanism of Dbp5 mediated tRNA export, which will form the foundation of a future investigation in the laboratory and warrants its own study.

Reviewer #1 (Public Review):

Various reports suggest that eukaryotic translation elongation factor 1 eEF1A is involved tRNA export Bohnsack et al., 2002 (Bohnsack MT, Regener K, Schwappach B, Saffrich R, Paraskeva E, Hartmann E, Görlich D. Exp5 exports eEF1A via tRNA from nuclei and synergizes with other transport pathways to confine translation to the cytoplasm. EMBO J. 2002 Nov 15;21(22):620515. doi: 10.1093/emboj/cdf613. PMID: 12426392; PMCID: PMC137205), Grosshans etal., 2002; Grosshans H, Hurt E, Simos G. An aminoacylation-dependent nuclear tRNA export pathway in yeast. Genes Dev. 2000 Apr 1;14(7):830-40. PMID: 10766739; PMCID: PMC316491). The presence of mutations in eEF1A has been seen to hinder the nuclear export process of all transfer RNAs (tRNAs). eEF1A has been shown to interact with Los1 aiding in tRNA export. The authors can also explore the crosstalk between Dbp5 and eEF1A in this study. Additionally, suppressor screening analysis in dbp5R423A , los1∆dbp5R423A los1∆msn∆dbp5R423A could shed more light on this.

Author Response: Thank you for this suggestion and raising an important possible role for Dbp5 in eEF1A mediated tRNA export. Based on more recent investigation of eEF1A function in tRNA export (PMID: 25838545), it is likely that eEF1A functions in re-export of charged tRNAs specifically (likely in conjunction with Msn5). The current manuscript has largely focused on the role of Dbp5 in pre-tRNA export, but a more careful mechanistic characterization of Dbp5 and re-export will be conducted in follow-up studies given the physical interaction between Dbp5 and spliced tRNAs we previously reported. Similarly, suppressor screens with the Dbp5 and los1Δmsn5Δ mutants will likely be a useful tool in identifying additional tRNA export factors and we thank the reviewer for this suggestion.

Reviewer #1 (Public Review):

The addition of Gle1 is potentially novel but it's unclear why the authors didn't address the potential involvement of IP6.

Author Response: The text has been revised to highlight the importance of InsP6 in Gle1 mediated activation of Dbp5. This includes referencing InsP6 throughout the manuscript during discussions of Gle1 activation of Dbp5 and lines 401-404 discussing the potential role for the small molecule in regulating mRNA and tRNA export in different cellular contexts (e.g., stress and disease).

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Weaknesses:

Reviewer comment: Here, the activity of SWIFT molecules was assessed in single cell types with or without BKlotho expression. Ultimately, the ability of the SWIFT molecules to activate Wnt signaling in a cell type-specific manner should be tested in the context of many different cellular identities that express BKlotho to different extents. It would be good to demonstrate that Wnt activation by SWIFT correlates with BKlotho expression level in multiple cell types - such data would strengthen the claim of cell-type specificity.

Response: We agree with the reviewer’s comment, it would be interesting to correlate the signaling level to the expression levels of βKlotho. The tools to carry out such an experiment are not currently available, as this would require a culture system that allows efficient growth of different cell types, and the reagents to detect both the receptor protein levels of βKlotho (as well as FZD/LRP) and signaling levels. We did perform an additional experiment to further support this targeting approach using a 2-layered (transwell) cell culture system. In this culture system, one cell type is put into the top well and the other cell type is put into the bottom well. Molecules to be tested were added to the media which is shared and freely diffuse across the two cell types. In this 2-layer cell system, the results again demonstrate the ability of the SWIFT molecules to specifically induce signaling only in βKlotho expressing hepatoma Huh7 cells and not in non-targeting HEK293 cells. This new data is included as Fig. 3H in the revised manuscript.

Reviewer comment: The study does not address whether the targeted cells express FGFR1c/2c/3c and whether the FGF21 full-length moiety or the 39F7 IgG moiety of SWIFT molecules could unintentionally activate FGF signaling in these cells.

Response: We agree with the reviewer’s comment. The receptor βKlotho and its binders (FGF21 and 39F7) were used to test the BRAID/SWIFT concept, the effects on FGF signaling were not the focus of the current study. This comment has now been added to the revised manuscript in the discussion. Inclusion of αGFP controls in the study also suggests the observed reporter activity in the targeted scenario is unlikely caused indirectly by any unexpected FGF signaling.

Reviewer #2 (Public Review):

Weaknesses:

Reviewer comment: The study shows the SWIFT approach works in vitro using cell lines, primary human hepatocytes, and human intestinal organoids, but it lacks an in vivo animal model or clinical validation. The applicability of this approach to therapy is still unknown.

Response: The βKlotho binder, 39F7, is specific to the human receptor and does not cross react with mouse. Unfortunately, we are not able to test these SWIFTs in a mouse model.

Reviewer comment: The success of SWIFT depends on the presence and expression of the bridging receptor (βKlotho) on target cells. The approach may fail if the target receptor is not expressed or available.

Response: We agree with the reviewer, the SWIFT molecules should not induce signaling on cells where bridging receptor is not expressed, therefore, achieving target cell specificity. As pointed out by the reviewer, finding the right bridging receptor on the target cell is critical.

Reviewer #1 (Recommendations For The Authors):

Reviewer comment 1: One way to further validate the specificity of SWIFT molecules would be to apply them to a mix of different cell types and quantify BKlotho level and Wnt reporter activity at the single cell level, potentially through imaging, FACS, or transcriptomics.

Response: We agree with the reviewer’s comment, it would be interesting to correlate the signaling level to the expression levels of βKlotho. The tools to carry out such an experiment are not currently available, as this would require a culture system that allows efficient growth of different cell types, and the reagents to detect both the receptor protein levels of βKlotho (as well as FZD/LRP) and signaling levels. We did perform an additional experiment to further support this targeting approach using a 2-layered (transwell) cell culture system. In this culture system, one cell type is put into the top well and the other cell type is put into the bottom well. Molecules to be tested were added to the media which is shared and freely diffuse across the two cell types. In this 2-layer cell system, the results again demonstrate the ability of the SWIFT molecules to specifically induce signaling only in βKlotho expressing hepatoma Huh7 cells and not in non-targeting HEK293 cells. This new data is included as Fig. 3H in the revised manuscript.

Reviewer comment 2: The experiments presented demonstrate activation of one signaling pathway in cells specifically expressing a target receptor rather than demonstrating "the feasibility of combining different signaling pathways" as stated in the abstract.

Response: We thank the reviewer for pointing this out and have adjusted the sentence accordingly.

Reviewer comment 3: What are the biological consequences of activating Wnt signaling in cells expressing BKlotho and why is that of interest? Could these biological outcomes be used as an additional, perhaps more consequential, readout for SWIFT activity?

Response: βKlotho is expressed on several different cell types that include hepatocytes, WAT, BAT, and certain regions in CNS. Our studies here focused on the WNT signaling pathway, and βKlotho/FGF21/39F7 receptor ligand system was used to illustrate the BRAID/SWIFT cell targeting concept. Whether these molecules may additional modulate endocrine FGF signaling and metabolic homeostasis, and whether there is any interaction between βKlotho and Wnt signaling pathways could be the subject of future studies. This is now added to the revised manuscript.

Reviewer comment 4: The manuscript would benefit from a careful review to improve wording and address grammatical errors.

Response: We thank the reviewer for this suggestion, and we have now had another round of language editing by a professional service.

Reviewer #2 (Recommendations For The Authors):

Reviewer comment 1. The expression of KLB in Fig 3G and 4B seems way too low and may not represent the amount on the cell surface. Did the authors validate the expression on the cell surface?

Response: In both figures we have displayed the expression level normalized to housekeeping gene ACTB. Housekeeping genes such as ACTB can be among the most abundant transcripts in a cell. The observation that KLB mRNA detection is below ACTB mRNA levels is expected and we would argue not too low. The average real-time PCR cycle threshold (Ct) for KLB in Huh7 and primary hepatocytes was 18 and 24 respectively. To avoid any confusion, we have now displayed the expression data normalized to HEK293 and intestinal organoids as a fold difference in a new Figure 3G and 4B.

Comment 2. Fig 3G needs statistical significance.

Response: We thank the reviewer for highlighting this, we have now included the statistical analysis in an updated Figure 3G.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

In the manuscript there is not much comparison between the crystal and cryoEM structures provided, and on inspection they appear to be very similar. The crystal structures also reveal parts of the CC domains in Las1, which is not present in the cryoEM structures. It is interesting the CC domains in Sc and Cj are quite different as illustrated in Figure 4B. They also seem to be somewhat disconnected from the rest of the complex (more so for Cj), even though that's not apparent in Figures 2-4. Despite this, it would be very useful to show the cryoEM densities when describing the catalytic site and C-terminal domain interactions, for example, as this can be very useful to increase confidence in the model and proposed mechanisms.

We thank the reviewer for this suggestion. We have added a figure (Figure 5- Figure supplement 3) to show cryo-EM and crystal densities of key amino acids, when describing the catalytic site and C-terminal domain interactions. In analyzing the interaction between Las1 and Grc3, we have also provided additional comparisons of the crystal structure and the cryo-EM structure (Figure 5, Figure 5-figure supplement 1, 2 and 3, Figure 6, Figure 6-figure supplement 1).

The description of the complex as a butterfly is engaging, and from a certain angle it can be made to look as such; this was also described previously in (Pillon et al., 2019, NSMB) for the same complex from a different organism (Ct). However, it is a bit misleading, because the complex is actually C2 symmetric. Under this symmetry, the 'body' would consist of two 'heads' one pointing up, one down facing towards the back, and one wing would have its back toward the viewer, the other the front. The structures presented here in Sc and Cj seem quite similar to the previous structure of the same complex in Ct, though the latter was only solved with cryoEM, and was also lacking the structure of the CC domain in Las1.

We thank the reviewer for pointing out this issue. We have re-wrote these sentences and changed the butterfly description of Las1-Grc3 complex in the revised manuscript.

For the model suggested in Figure 8, perhaps in the 'weak activity' state, the LCT in Las1 could still be connected to Grc3, via the LCT, rather than disconnected as shown. This could facilitate faster assembly of the 'high activity' state. The complex is described as 'compact and stable', but from the structure and this image, it appears more dynamic, which would serve its purpose and the illustrated model better. The two copies of HEPN appear to have more connective area, meaning they are indeed more likely to remain assembled in the 'weak activity' state. On the other hand, HEPN in one protein appears to have less binding surface with PNK in Grc3, and even less so with the CTD (both PNK and CTD being from the other associated protein), meaning these bindings could release easily to form the 'weak activity' state.

There is also the potential to speculate that the GCT is bound to HEPN near the catalytic area in the 'weak activity' state. The reduced activity when the GCT residues are replaced by Alanine could then be explained by the complex not being able to assemble as quickly upon binding of the substrate, as it could if the GCT remained bound, rather than by a conformational change that it induces upon binding. The conformational change is also likely to be influenced by the combined binding of PNK and CTD in the assembled state, which also contact HEPN, rather than by GCT alone.

We thank the reviewer for this suggestion. We have revised our model in the new Figure 8 of our revised manuscript. We apologize for the un-clarity description of the 'weak activity' state in our model. In fact, we believe that Las1 is in a "weakly activity" state before binding to Grc3 and is in a "highly activity" state when it forms a complex with Grc3. We strongly agree that the Las1-Grc3 complex is more dynamic than compact and stable, so it is easy to change its active state. We have changed our description and revised our model in the revised manuscript.

When comparing the structure of the HEPN domain in the lone Las1 protein to the structure of Las1-HEPN in the Las1-Grc3 complex, it is mentioned that 'large conformational changes are observed'. These could be described a bit better. The conformational change is ~3-4Å C-alpha RMSD across all ~150 residues in the domain (~90 residues forming a stable core that only changes by ~1Å). There is also a shift in the associated HEPN domain in Las1B domain compared to the bound HEPN in the Las1-Grc3 complex, as shown in Figure 7D: ~1Å shift and ~12degrees rotation. This does point to the conformation of HEPN changing upon complex formation, as does the relative positions of the HEPN domains in Las1A and Las1B. The conformational change and relative shift could indeed by key for the catalysis of the substrate as mentioned.

We thank the reviewer for this great suggestion. We have replaced the sentence describing the conformational changes in our revised manuscript.

Overall, the structures presented should be very useful in further study of this system, even though the exact dynamics and how the substrate is bound are aspects that are perhaps not fully clear yet. The addition of the structures of the CC domain in two different organisms and the Las1 HEPN domain (not in complex with Grc3) as new structural information should allow for increasing our understanding of the overall complex and its mechanism.

We thank this reviewer for these encouraging comments, which helped us with greatly improving our manuscript.

Reviewer #2 (Public Review):

In this manuscript, Chen et al. determined the structural basis for pre-RNA processing by Las1-Grc3 endoribonuclease and polynucleotide kinase complexes from S. cerevisiae (Sc) and C. jadinii (Cj). Using a robust set of biochemical assays, the authors identify that the sc- and CjLas1-Grc3 complexes can cleave the ITS2 sequence in two specific locations, including a novel C2' location. The authors then determined X-ray crystallography and cryo-EM structures of the ScLas1-Grc3 and CjLas1-Grc3 complexes, providing structural insight that is complimentary to previously reported Las1-Grc3 structures from C. thermophilum (Pillon et al., 2019, NSMB). The authors further explore the importance of multiple Las1 and Grc3 domains and interaction interfaces for RNA binding, RNA cleavage activity, and Las1-Grc3 complex formation. Finally, evidence is presented that suggests Las1 undergoes a conformational change upon Grc3 binding that stabilizes the Las1 HEPN active site, providing a possible rationale for the stimulation of Las1 cleavage by Grc3.

Several of the conclusions in this manuscript are supported by the data provided, particularly the identification and validation of the second cleavage site in the ITS2. However, several aspects of the structural analysis and complimentary biochemical assays would need to be addressed to fully support the conclusions drawn by the authors.

We thank the reviewer for the positive comments.

• There is a lack of clarity regarding the number of replicates performed for the biochemical experiments throughout the manuscript. This information is critical for establishing the rigor of these biochemical experiments.

We apologize for not providing the detailed information on the number of replicates of biochemical experiments. All the biochemical experiments were repeated three times. We have provided this information in the figure legends.

• The authors conclude that Rat1-Rai1 can degrade the phosphorylated P1 and P2 products of ITS2 (lines 160-162, Figure 1H). However, the data in Fig. 1H shows complete degradation of 5'Phos-P2 and 5'Phos-P4 of ITS2, while the P1 and 5'Phos-P3 fragments remain in-tact. Additional clarification for this discrepancy should be provided.

We thank the reviewer for pointing out this issue. “phosphorylated P1 and P2 products” should be “phosphorylated P2 and P4 products”. We have corrected this clerical error. In addition, we have also provided an explanation for why phosphorylated P3 product show only partial degradation. We suspect that P3 product may be too short to completely degrade.

• The authors determined X-ray crystal structures of the ScLas1-Grc3 (PDB:7Y18) and CjLas1-Grc3 (PDB:7Y17) complexes, which represents the bulk of the manuscript. However, there are major concerns with the structural models for ScLas1-Grc3 (PDB:7Y18) and CjLas1-Grc3 (PDB:7Y17). These structures have extremely high clashscores (>100) as well as a significant number of RSRZ outliers, sidechain rotamer outliers, bond angle outliers, and bond length outliers. Moreover, both structures have extensive regions that have been modeled without corresponding electron density, and other regions where the model clearly does not fit the experimental density. These concerns make it difficult to determine whether the structural data fully support several of the conclusions in the manuscript. A more careful and thorough reevaluation of the models is important for providing confidence in these structural conclusions.

We thank the reviewer for pointing out this issue. We have used the cryo-EM datasets to further validate our conclusions of the manuscript. We analyzed the active site of Las1-Grc3 complex and the interactions between Las1 and Grc3 using the cyro-EM structures and presented new figures (Figure 5- Figure supplement 1, Figure 5- Figure supplement 2, Figure 5- Figure supplement 3, Figure 6- Figure supplement 1) in our revised manuscript. Both the refinement and validation statistical parameters of the cryo-EM datasets are within a reasonable range (Table 2), which will provide confidence for our structure conclusions. The X-ray crystal structures of ScLas1-Grc3 (PDB:7Y18) and CjLas1-Grc3 (PDB:7Y17) complexes has high calshscores and many outliers, which is mainly due to the great flexibility of Las1-Grc3 complex, especially the CC domain of Las1. We have improved our crystal structure models with better refinement and validation of statistical parameters. The clashscores of ScLas1-Grc3 complex and CjLas1-Grc3 complex are 25 and 45, respectively. There are no rotamer outliers and C-beta outliers to report for both ScLas1-Grc3 complex and CjLas1-Grc3 complex.

• The presentation of the cryo-EM datasets is underdeveloped in the results section drawing and the contribution of these structures towards supporting the main conclusions of the manuscript are unclear. An in-depth comparison of the structures generated from X-ray crystallography and cryo-EM would have greatly strengthened the structural conclusions made for the ScLas1-Grc3 and CjLas1-Grc3 complexes.

We thank the reviewer for this suggestion. We have performed structural comparisons between X-ray crystal structure and cyro-EM structure in analyzing the active site of Las1-Grc3 complex and the interactions between Las1 and Grc3 (Figure 5- Figure supplement 1, Figure 5- Figure supplement 2, Figure 6- Figure supplement 1). We have also added a figure (Figure 5- Figure supplement 3) to show cryo-EM and crystal densities of the Las1 active site as well as the key amino acids for Las1 and Grc3 interactions. These comparisons and densities have greatly strengthened our structural conclusions.

• The authors conclude that truncation of the CC-domain contributes to Las1 IRS2 binding and cleavage (lines 220-222, Fig. 4C). However, these assays show that internal deletion of the CC-domain alone has minimal effect on cleavage (Fig 4C, sample 3). The loss in ITS2 cleavage activity is only seen when truncating the LCT and LCT+CC-domain (Fig 4C, sample 2 and 4, respectively). Consistently, the authors later show that Las1 is unable to interact with Grc3 when the LCT domain is deleted (Fig. 6 and Fig. 6-figure supplement 2). These data indicate the LCT plays a critical role in Las1-Grc3 complex formation and subsequent Las1 cleavage activity. However, it is unclear how this data supports the stated conclusion that the CC-domain is important for LasI cleavage.

Our EMSA data shows that the CC domain contributes to the binding of ITS2 RNA (Figure 4D), suggesting that the CC domain may play a role of ITS2 RNA stabilization in the Las1 cutting reaction. The in vitro RNA cleavage assays (Figure 4C) indicate that the LCT is important for Las1 cleavage because it plays a critical role in the formation of the Las1-Grc3 complex. Compared with LCT, the CC domain, although not particularly important for Las1 cutting ITS2, still has some influence (Fig 4C, sample 1 and 3, sample 2 and 4,). Therefore, we conclude that the CC domain may mainly play a role in the stabilization of ITS2 RNA, thereby enhancing ITS2 RNA cleavage.

• The authors conclude that the HEPN domains undergo a conformational change upon Grc3 binding, which is important for stabilization of the Las1 active site and Grc3-mediated activation of Las1. This conclusion is based on structural comparison of the HEPN domains from the CjLas1-Grc3 complex (PDB:7Y17) and the structure of the isolated HEPN domain dimer (PDB:7Y16). However, it is also possible that the conformational changes observed in the HEPN domain are due to truncation of the Las1 CC and CGT domains. A rationale for excluding this possibility would have strengthened this section of the manuscript.

We thank the reviewer for pointing out this issue. We agree that the complete Las1 structure information is helpful in illuminating the conformational activation of the Las1 by Grc3. We screened about 1200 crystallization conditions with full-length Las1 proteins, but ultimately did not obtain any crystals, probably due to flexibility. The CC domain exhibits a certain degree of flexibility, which has not been observed in the structure obtained from electron microscopy. The LCT is involved in binding to the CTD domain of Grc3. The coordination of the active center of HEPN domains by LCT and CC domains is unlikely due to the limited nuclease activity observed in full-length Las1. The conformational changes of the active center are essential for HEPN nuclease activation. Our structure shows that the GCTs of Grc3 interact with the active residues of Las1 HEPN domains, which probably induce conformational changes in the active center of the HEPN domain to activate Las1. Of course, we cannot exclude the possibility that truncation of the Las1 CC and LCT domains will result in little conformational change in the HEPN domains. We have explained this possibility in our revised manuscript.

Reviewer #1 (Recommendations For The Authors):

1) It would be very useful to show the cryoEM densities when describing the catalytic site and C-terminal domain interactions.

The new Figure 5-figure supplement 2 have showed the Cyro-EM densities of the catalytic site of ScLas1 and the C-terminal domain of ScGrc3.

2) "ScLas1 cleaves the 33-nt ITS2 at C2 site to theoretically generate a 10-nt 5′-terminal product and a 23-nt 3′-terminal product (Figure 1A). Our merger data shows that the final 5′-terminal and 3′-terminal product bands are at nearly the same horizontal position on the gel (Figure 1B), indicating that they are similar in size." These two sentences seem to contradict, i.e. 10-nt and 23-nt are similar in size even though they are different lengths?

We apologize for the contradiction in these two sentences mentioned above. We have re-wrote these two sentences in the revised manuscript.

3) We observed four cleavage bands of approximately 23-nt (P2), 14-nt (P3), 10-nt (P1), and 9-nt (P4) in length (Figure 1C). "

Figure 1C. The bands show 23 nt, 22 nt, 21nt, 14 nt, 13nt, and 11nt, so this text does not seem to describe the figure.

We have re-wrote this sentence in the revised manuscript.

4) "We obtained similar cleavage results with a longer 81-nt ITS2 RNA substrate 6 (Figure 1D, E). " Figure 1D,E. The lengths in Figure 1E do not correspond to all bands in Figure 1E, e.g. the 13 nt band, though the others do, e.g. 14 nt, 30nt, 37nt, etc.

In order to better evaluate the size of the cut product, we used an RNA marker as a comparison. The RNA marker will have more bands than the cleavage products. To further confirm the cleavage site of C2′, we also mapped the cleavage sites of the 81-nt ITS2 using reverse transcription coupling sequencing methods (Figure 1F).

5) In Figure 3, domains are colored different but it's hard to know which are different proteins.

We have added a diagram in Figure 3 to show the Las1-Grc3 complex structure, and it is now clear how Las1 and Grc3 are assembled into a tetramer.

6) Line 267. "we screened a lot of crystallization conditions with full-length Las1 proteins" How many? Rough numbers ok, but 'a lot' is not very informative

We have provided the approximate numbers of crystallization conditions in our revised manuscript.

Reviewer #2 (Recommendations For The Authors):

1) The authors missed an excellent opportunity to compare and contrast the ScLas1-Grc3 and CjLas1-Grc3 complex structures presented here with that of the previously determined CtLas1-Grc3 structure (Pillon et al., 2019, NSMB). For example, His130 in the ScLas1-Grc3 complex active site adopts a similar conformation to His142 in the TcLas1-Grc3 complex active site (Pillon et al., 2019, NSMB). Interestingly, the analogous His134 active site residue in the CjLas1-Grc3 adopts an alternative (maybe inactive) conformation. This observation could provide a structural rationale for the activation of scLas1 and TcLas1 by Grc3, while also providing a rationale for the fairly weak activation of CjGrc3 by CjGrc3.

We thank the reviewer for this suggestion. We have performed structural comparisons between ScLas1-Grc3, CjLas1-Grc3 and CtLas1-Grc3 complexes, especially the Las1 nuclease active center. We added two figures (Figure7-figure supplement 3A and 3B) in the revised manuscript to contrast and highlight the conformational differences of active amino acids in active centers between ScLas1-Grc3, CtLas1-Grc3 and CjLas1-Grc3. These structural comparisons provide stronger evidence that further reinforces the conclusions of our manuscript.

2) Can the authors speculate as to whether the structural data can provide any insight into how the Las1-Grc3 may cleave both C2 and C2' positions in the ITS2 RNA? This commentary would further strengthen the discussion section of the manuscript.

We thank the reviewer for this suggestion. We have provided a speculation in the discussion section of the revised manuscript.

We think that the structural data may provide some insight into how Las1-Grc3 complex cleaves ITS2 RNA at both C2 and C2' positions. The Las1-Grc3 tetramer complex has one nuclease active center and two kinase active centers. The nuclease active center consists of two Las1 molecules in a symmetric manner, while the kinase active center consists of only one Grc3 molecule. The ITS2 RNA is predicted to form a stem-loop structure. The symmetrical nuclease active center recognizes the stem region of ITS2 RNA and makes it easy to perform C2 and C2' cleavages on both sides of the stem. C2 and C2' cleavage products are further phosphorylated by two Grc3 kinase active centers, respectively.

3) The method used for the plasmid generation, expression, and purification of the Las1 truncations and the Las1 and Grc3 point mutants should be provided in the methods section.

The method used for the plasmid generation, expression, and purification of the Las1 truncations and the Las1 and Grc3 point mutants have be provided in the methods section.

4) The exact amino acid cutoffs for the truncated forms of Las1 used for the biochemical assays in Fig. 4 should be provided.

We have provided the exact amino acid cutoffs for the truncated forms of Las1 in the figure legend of Figure 4C.

5) The models associated with the cryo-EM datasets should be deposited in the PDB.

The models associated with the Cryo-EM datasets have be deposited in the PDB with the following accession codes: 8J5Y (ScLas1-Grc3 complex), and 8J60 (CjLas1-Grc3 complex).

6) Lines 232-234: Arg129 should be changed to His134.

We have corrected it.

7) Figure 5B: the bottom half of the HEPN active site has been labeled incorrectly. The labels should be Arg129, His130, and His134 (from left to right).

We have corrected it.

8) Line 252: "multitudinous" should be changed to "multiple."

We have corrected it.

Author Response

We are grateful to the reviewers for their thorough and thoughtful critiques, including their agreement on the significant value of this dataset. We intend to respond to their comments in full with a revision in the near future. However, we would like to make an initial comment at this stage. A key concern raised by the reviewers was that the analyses described do not adequately support the claim that "movie-watching data can identify retinotopic regions" (quoted from R2, similar sentiment expressed by R1). To be clear, we agree with this assessment. Our primary aim was not to identify visual areas with movie-watching data. Rather, our focus was on how movies can reveal fine-grained organization in infant visual cortex, which would support their potential utility for understanding the development of dynamic visual processing. To demonstrate this potential, we tested and found that maps of visual activity generated from movies are significantly similar to those generated by a retinotopy task. Nevertheless, we did not intend to argue that movie-based maps are sufficiently accurate to replace task-based retinotopic maps when defining visual areas, nor did we test this possibility. We accept that this point was unclear in the original manuscript and will make edits to avoid this miscommunication. We also plan to incorporate the reviewers’ many other helpful recommendations, including addressing concerns about the clarity of the presentation and double dipping, as well as adding several new analyses we hope will provide greater confidence in the findings and interpretation.

Author Response

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

We would like the reviewers for their positive and useful comments. Below please find our answers to the issues raised.

Reviewer #1 (Public Review):

Overall, the experiments are well-designed and the results of the study are exciting. We have one major concern, as well as a few minor comments that are detailed in the following.

Major:

1) The authors suggest that "Visuomotor experience induces functional and structural plasticity of chandelier cells". One puzzling thing here, however, is that mice constantly experience visuomotor coupling throughout life which is not different from experience in the virtual tunnel. Why do the authors think that the coupled experience in the VR induces stronger experience-dependent changes than the coupled experience in the home cage? Could this be a time-dependent effect (e.g. arousal levels could systematically decrease with the number of head-fixed VR sessions)? The control experiment here would be to have a group of mice that experience similar visual flow without coupling between movement and visual flow feedback.

Either change would be experience-dependent of course, but having the "visuomotor experience dependent" in the title might be a bit strong given the lack of control for that. We would suggest changing the pitch of the manuscript to one of the conclusions the authors can make cleanly (e.g. Figure 4).

Although the plasticity is induced by the visuomotor experience in the tunnel, we agree that we do not know what aspect of the repeated exposure to the virtual tunnel caused the plasticity. We cannot rule out that it was the exposure to the visual stimuli alone that caused it. Therefore, we rephrased sentences that suggested that it was the coupling between visual stimuli and motor behavior that was responsible for the plasticity. We also changed the title to “Experience Shapes Chandelier Cell Function and Structure in the Visual Cortex”.

We do believe that training the mice in the virtual tunnel does significantly increase experience with coupled visuomotor activity, though. In their home cage, mice are mostly active in the dark and there is litle space to run.

Minor:

2). "ChCs shape the communication hierarchy of cortical networks providing visual and contextual information." We are not sure what this means.

We thank the reviewer for helping to raise clarity and we rephrased this sentence to: “…ChCs may establish a hierarchical relationship among cortical networks.”

3) "respond to locomotion and visuomotor mismatch, indicating arousal-related activity" This is not clear. We think we understand what the authors mean but would suggest rephrasing.

Agreed. We rephrased this sentence to: "...respond to events that are known to increase arousal levels, such as locomotion and visuomotor mismatch.”

4) 'based on morphological properties revealed that 87% (287/329) of labeled neurons were ChCs" Please specify the morphological properties used for the classification somewhere in the methods.

We added that the neurons were positioned at the border of L1 and L2 and had a dendrite reaching into layer 1.

5) We may have missed this - in the patch clamp experiment (Fig.1 H-K), please add information about how many mice/slices these experiments were performed in.

We have added the information to the legend of Fig. 1.

6) "These findings suggest that the rabies-labeled L1-4 neurons providing monosynaptic input to ChCs are predominantly inhibitory neurons". We are not sure this conclusion is warranted given the sparse set of neurons labelled and the low number of cells recorded in the paired patch experiment. We would suggest properly testing (e.g. stain for GABA on the rabies data) or rephrasing.

We weakened the statement to: “These findings suggest that the rabies-labeled L1-4 neurons providing monosynaptic input to ChCs may include many inhibitory neurons.”

7) Figure 2E. A direct comparison of dF/F across different cell types can be subject to a problematic interpretation. The transfer function from spikes to calcium can be different from cell type to cell type. Additionally, the two cell populations have been marked with different constructs (despite the fact that it's the same GECI) further reducing the reliability of dF/F comparisons. We would recommend using a different representation here that does not rely on a direct comparison of dF/F responses (e.g. like the "response strength" used in Figure 3B). Assuming calcium dynamics are different in ChCs and PyCs - this similarity in calcium response is likely a coincidence.

We have removed the quantification in this figure.

8) If ChCs are more strongly driven by locomotion and arousal, then it's a bit counterintuitive that at the beginning of the visual corridor when locomotion speed consistently increases, the activity of ChCs consistently decreases. This does not appear to be driven by suppression by visual stimuli as it is present also in the first and last 20cm of the tunnel where there are no visual stimuli. How do the authors explain this?

We do believe that this is suppression driven by visual stimuli. Although on average the strongest visual suppression happens between 20-80 cm, neurons that have their receptive fields toward the center of the visual field will already respond to the stimuli before the mouse reaches the 20 cm location of the tunnel. In addition, although the visual stimuli are the strongest sensory inputs, the background of the visual part of the tunnel has a black and white noise patern, which might already mildly suppress ChC activity. Both arguments are supported by the observation that the visual PyCs (V-PyCs, blue line) in Fig. 4D are already activated at the beginning of the tunnel and that the activity of V-PyCs matches well with the suppression of ChC activity.

9) The authors mention that "ChC responses underwent sensory-evoked plasticity during the repeated visual exposure, even though the visual stimuli were different from those encountered during training in the virtual tunnel". How would this work? And would this mean all visual responses are reduced? What is special about the visual experience in the virtual tunnel? It does not inherently differ from visual experience in the home cage, given that the test stimuli (full field gratings) are different from both.

As mentioned in our answer to point 1, the exposure to visual stimuli is strongly increased since, firstly, they are presented during the dark phase when the mice are most active and, secondly, they do not get these types of visual inputs in their home cage.

10) Just as a point to consider for future experiments: For the open-loop control experiments, the visual flow is constant (20cm/s) - ideally, this would be a replay of the running speed the mouse previously generated to match statistics.

We agree with this point and will implement replay of earlier sessions in future experiments.

11) We would recommend specifying the parameters used for neuropil correction in the methods section.

This is described on page 24, under “preprocessing”. We also refer to the analysis package (Spectral Segmentation - SpecSeg) in which the neuropil correction as used by us here is explained in more detail.

12) If we understand correctly, the F0 used for the dF/F calculation is different from that used for division. Why is this?

We apologize for this mistake, which was based on an older version of the software. We have now corrected this in the revised manuscript.

13) Authors compare neuronal responses using "baseline-corrected average". Please specify the parameters of the baseline correction (i.e. what is used as baseline here).

In the revised version we have now beter explained this in the methods, page 24, under “Passive Sessions”.

Reviewer #2 (Public Review):

Summary:

Seignete et al. investigated the potential roles of axo-axonic (chandelier) cells (ChCs) in a sensory system, namely visual processing. As introduced by the authors, the axo-axonic cell type has remained (and still is) somehow mysterious in its function. Seignete and colleagues leveraged the development of a transgenic mouse line selective for ChC, and applied a very wide range of techniques: transsynaptic rabies tracing, optogenetic input activation, in vitro electrophysiology, 2-photon recording in vivo, behavior and chemogenetic manipulations, to precisely determine the contribution of ChCs to the primary visual cortex network.

The main findings are 1) the identification of synaptic inputs to ChC, with a majority of local, deep layer principal neurons (PN), 2) the demonstration that ChC is strongly and synchronously activated by visual stimuli with low specificity in naive animals, 3) the recruitment of ChC by arousal/visuomotor mismatch, 4) the induction of functional and structural plasticity at the ChC-PN module, and, 5) the weak disinhibition of PNs induced by ChCs silencing. All these findings are strongly supported by experimental data and thoroughly compared to available evidence.

Strengths:

This article reports an impressive range of very demanding experiments, which were well executed and analyzed, and are presented in a very clear and balanced manner. Moreover, the manuscript is well- writen throughout, making it appealing to future readers. It has also been a pleasure to review this article.

In sum, this is an impressive study and an excellent manuscript, that presents no major flaws.

Notably, this study is one of the first studies to report on the activities and potential roles of axo-axonic cells in an active, integrated brain process, beyond locomotion as reported and published in V1. This type of research was much awaited in the fields of interneuron and vision research.

Weaknesses:

There are no fundamental weaknesses; the later mainly concern the presentation of the main results. The main weakness may be that the different sections appear somehow disconnected conceptually.

Additionally, some parts deserve a more in-depth clarification/simplification of concepts and analytic methods for scientists outside the subfield of V1 research. Indeed, this paper will be of key interest to researchers of various backgrounds.

Reviewer #3 (Public Review):

Summary:

The authors set out to characterize the anatomical connectivity profile and the functional responses of chandelier cells (ChCs) in the mouse primary visual cortex. Using retrograde rabies tracing, optogenetics, and in vitro electrophysiology, they found that the primary source of input to ChCs are local layer 5 pyramidal cells, as well as long-range thalamic and cortical connections. ChCs provided input to local layer 2/3 pyramidal neurons, but did not receive reciprocal connections.

With two-photon calcium imaging recordings during passive viewing of drifting gratings, the authors showed that ChCs exhibit weakly selective visual responses, high correlations within their own population, and strong responses during periods of arousal (assessed by locomotion and pupil size). These results were replicated and extended in experiments with natural images and prediction of receptive field structure using a convolutional neural network.

Furthermore, the authors employed a learned visuomotor task in a virtual corridor to show that ChCs exhibit strong responses to mismatches between visual flow and locomotion, locomotion-related activation (similar to what was shown above), and visually-evoked suppression. They also showed the existence of two clusters of pyramidal neurons with functionally different responses - a cluster with "classically visual" responses and a cluster with locomotion- and mismatch-driven responses (the later more correlated with ChCs). Comparing naive and trained mice, the authors found that visual responses of ChCs are suppressed following task learning, accompanied by a shortening of the axon initial segment (AIS) of pyramidal cells and an increase in the proportion of AIS contacted by ChCs. However, additional controls would be required to identify which component(s) of the experimental paradigm led to the functional and anatomical changes observed.

Finally, using a chemogenetic inactivation of ChCs, the authors propose weak connectivity to pyramidal cells (due to small effects in pyramidal cell activity). However, these results are not unequivocally supported, as the baseline activity of ChCs before inactivation is considerably lower, suggesting a potentially confounding homeostatic plasticity mechanism might already be operating.

Strengths:

The authors bring a comprehensive, state-of-the-art methodology to bear, including rabies tracing, in vivo two-photon calcium imaging, in vitro electrophysiology, optogenetics and chemogenetics, and deep neural networks. Their analyses and statistical tests are sound and for the most part, support their claims. Their results are in line with previous findings and extend them to the primary visual cortex.

Weaknesses:

• Some of the results (e.g. arousal-related responses) are not entirely surprising given that similar results exist in other cortical areas.

We agree that previous studies have shown arousal-related responses of ChC cells and our study confirms those findings. However, this is not the main message of the article and we present many findings that are novel.

• Control analyses regarding locomotion paterns before and atier learning the task (Figure 5), and additional control experiments to identify whether functional and anatomical changes following task learning were due to learning, repeated visual exposure, exposure to reward, or visuomotor experience would strengthen the claims made.

In figure 5 we excluded running trials, so locomotion paterns are unlikely to play a major role. We agree that testing what are the factors that contribute to the observed plasticity are important to investigate in future experiments.

• The strength of the results of the chemogenetics experiment is impacted by the lower baseline activity of ChCs that express the KORD receptor. At present, it is not possible to exclude the presence of homeostatic plasticity in the network before the inactivation takes place.

Although we do not know why there is a difference in the baseline df/f (e.g. expression levels), we consider it unlikely that expression of the KORD receptor itself without exposure to the ligand causes reduction of ChC activity. Moreover, we are not sure how homeostatic plasticity in the network would occur selectively in KORD-expressing ChCs. Finally, we do not find evidence for a relationship between lower ChC calcium signals and the effects of ChC silencing on PyC activity. We performed an additional analysis in which we correlated baseline ChC activity (before salvinorin B injection) with the effect of ChC silencing on PyC activity (post – pre) across mice, and found that this correlation was not significant (R = 0.41, p = 0.18).

Reviewer #1 (Recommendations For The Authors):

In the spirit of openness of the scientific discussion, all our feedback and recommendations to the authors are included in the public reviews.

Reviewer #2 (Recommendations For The Authors):

Most of my comments and suggestions concern the presentation of the data, to (hopefully) help and convey as clearly as possible the messages of this important article.

Main

The main weakness of the paper may be that the different sections appear somehow disconnected conceptually. This is particularly true for:

-structural plasticity: how can we link this finding with the rest of the study? Are there ways to correlate this finding with physiological recordings in individual animals, or to directly test whether particular functional types of PNs (visual, non-visual) undergo plasticity at their AIS?

This is a very interesting question that may be addressed in future experiments.

-the indirect finding suggesting that ChC weakly inhibits PNs using chemogenetic silencing of PNs. Do chemogenetic manipulations of ChCs affect PN responses in visual paradigm and/or modify the induction of structural plasticity at the ChC-AIS connection?

This is also a very interesting question for future work.

Additionally, some parts would deserve a more in-depth clarification/simplification of concepts and analytic methods (OSI, DSI, MEI...) for scientists outside the subfield of V1 research. Indeed, this paper will be of key interest to researchers of various backgrounds.

In the revised manuscript we briefly explain what an MEI is when first introduced, and introduce the abbreviations OSI and DSI at the correct location. We believe orientation and direction selectivity are well-known concepts for the audience reading this article.

Minor

These are discussed by order of appearance in the text.

Abstract

The alternative interpretation of error/mismatch negativity to explain ChC activation deserves to appear in the abstract. Arousal consistency in prediction should be in the introduction. "In mice running in a virtual tunnel, ChCs respond strongly to locomotion and halting visual flow, suggesting arousal-related activity."

This comment holds for the end of the introduction and the beginning of the discussion, as well.

"These findings suggest that ChCs provide an arousal-related signal to layer 2/3 pyramidal cells that may modulate their activity". This statement appears to be in contradiction with the weak effect mentioned just before. This comment holds for the end of the introduction.

The full sentence was: “These findings suggest that ChCs provide an arousal-related signal to layer 2/3 pyramidal cells that may modulate their activity and/or gate plasticity of L2/3 PyCs in V1.” Our results show that activity of layer 2/3 pyramidal cells is modulated (albeit weakly) and it is well possible that ChCs regulate plasticity at the AIS. Therefore, we do not believe that this statement contradicts the weak direct effect of ChCs on layer 2/3 pyramidal cell activity. Therefore , we think that this statement does not contradict the weak direct effect of ChCs on layer 2/3 pyramidal cell activity.

We changed the last sentence of the introduction to “Our findings suggest that ChCs predominantly respond to arousal related to locomotion or unexpected events/stimuli, and act to weakly modulate activity and/or gate plasticity of L2/3 PyCs in V1.”

Introduction First paragraph

Coming from a field outside of vision research, it is not obvious to me what has been learned from interneuron classes in the past. An example would be welcome in the introduction.

The literature on the role of different interneuron types in visual processing and plasticity is too large to pick one or two examples. For the sake of conciseness, we have therefore provided some important references and reviews for the interested readers (references 1 to 10).

Interneuron "subtypes" seem to refer to main classes (e.g. PV+): please rephrase accordingly (ChC being a type and PV+ ChC a subtype).

We changed interneuron “subtypes” to “types” and left L2/3 pyramidal cell “subtypes” unchanged.

Second paragraph

Beyond the reversal potential of GABA-ARs at the axon initial segment, GABA may inhibit action potential generation in various conditions (Lipkin et al. 2023, DOI: 10.1523/JNEUROSCI.0605-23.2023 : should be cited).

We added this citation.

Fourth paragraph

"ChCs alter the number of synapses at the AIS based on the activity of their postsynaptic targets": the concept of alteration is too vague to let the reader grasp the concept: could the authors rephrase?

We have rephrased the sentence to:

“…ChCs increase the number of synapses at the AIS if their postsynaptic targets are chemogenetically activated…”

Results 1) ChCs receive input from long-range sources and L5 PyCs in V1 It is not clear how morphological identification of ChC was performed. Did dendrites and/or axons of starter cells occasionally overlap as can be expected, complicating the cell-by-cell morphological classification?

"Most labeled neurons were located on the border between L1 and L2/3 and displayed typical ChC morphology": maybe clarify that this concerns neurons expressing eYFP-TVA?

We assessed the location (at the border of L1 and L2) and spatial distribution of the labeled cells and whether they had a dendrite extending upwards towards into L1. We have now indicated this in the results section and clarified that these neurons express eYFP-TVA.

-Likewise the following would benefit from clarification " This is further supported by the distributed localization of the labeled neurons": it would also help here to remind the reader of the labelling (presumably retrogradely-labeled mCherrry+ neurons).

We have now clarified in the text that these are mCherry+ neurons labeled by the rabies virus

2) Chandelier cells are modulated by arousal and show high correlations

-The authors indicate that the results "(suggest) that ChCs distribute a synchronized signal during high arousal." : it would be stronger to defend this claim by showing a higher ChC-ChC correlation during "arousal" vs. baseline (i.e. analyze high arousal epochs outside of movement). It may be difficult to perform this analysis due to low fluorescence changes outside running episodes, but this should be discussed accordingly. In this respect, the title of the section is more in line with the data presented.

We believe our statement is correct. The activity of ChCs is highly synchronized and their firing rates increase during arousal. We do not state that synchronization increases with arousal.

-A brief explanation of DSI and OSI meaning would be nice for the audience that will definitely extend beyond vision research given the importance of this study.

See above

3) ChCs are weakly selective to visual information

-I may very well miss the point, but the equivalence in response strength among cell classes (Fig3B) seems inconsistent with the wider distribution of high response strength in ChCs (Fig3C). Perhaps a graphical representation taking into account the distribution of single data points in Fig3B would help resolve this discrepancy.

This is because in panel C the response strengths are normalized. We now also state this in the legend to avoid confusion.

-"clearly oriented edge-like paterns with sharp ON and OFF regions": it would help if a representative example was highlighted in Figure 3F.

The majority of L2/3 pyramidal MEIs presented in this panel show this patern.

-It is interesting and surprising that properties of ChCs appear more distinct from those of L5 PNs than from those of L2-3 PNs (Fig 3G-J), given the fact that V1 ChCs were found by the authors to derive their inputs from V1 L5 PNs (please see comments of the discussion for this specific point).

How ChCs respond based on L5 input depends strongly on how the connections between L5 and ChCs are organized. Similarity between responses of L5 and ChC neurons is not required.

4) Locomotion and visuomotor mismatch drive chandelier cell activity in a virtual tunnel This is the least convincing part in terms of presentation.

-It is unclear where/when visuomotor mismatch has been induced in the tunnel: please clarify in the text and in Fig 4B.

We realized that the title of the paragraphs was indeed confusing. In fig. 4A-D and the first paragraph about the virtual tunnel, we do not discuss the visuomotor mismatch. This comes later, when we describe the results in Fig. 4E. The titles have been changed.

-No result on visuomotor mismatch is reported in the text of this section, while this is presented in the subsequent section: this needs to be corrected (merge this section with the next?).

We agree, apologies for the confusion. See above.

-It would be interesting to further analyze responses to CS and US. Regarding the US: is water rewarding in non-water-restricted mice? This should be mentioned.

We realized that we did not mention that the mice were water restricted during behavioral training and during the imaging sessions when mice performed the virtual tunnel task. We have now added this to the methods section. Sorry for the omission.

-Along this line: was water sometimes omited? This would provide a complementary way to test the prediction error theory for ChC activation with an alternative modality.

We never omited the water reward. It would be interesting to test this in a future experiment.

5) ChCs have similar response properties as non-visual PyCs

• It would help to explicitly mention that in Ai65 mice, only Cre and Flp+ cells express tdTomato (here Vipr2 and PV+).

We added the following sentence: “In these mice, tdTomato was only expressed in cells expressing both Vipr2 and PV.”

6) Visuomotor experience in the virtual tunnel induces plasticity of ChC-AIS connectivity

• In relation to the previous section, Jung et al. (doi.org/10.1038/s41593-023-01380-x) recently reported that motor learning reduced ChC-ChC synchrony in M2. Did the author observe a similar change in ChC- ChC synchrony with visual experience/habituation to the task? If available, these data should be reported to help build a clearer picture of ChC functions in the neocortex.

We tested this and also found reduced correlations between ChCs in trained mice vs naïve mice. We added this as text on p14 in the results section.

• The low number of ChC boutons' appositions per AIS may be misleading: "While the average number of ChC boutons per AIS remained constant (~2-3 ChC boutons/AIS)"). It would be helpful to make it clear that these are "virally" labelled boutons, as opposed to absolute numbers, if compared with the detailed quantification of Schneider-Mizell et al, 2021 (7.4 boutons per AIS in average; doi: 10.7554/eLife.73783.).

We added "virally labeled"

• It may be difficult to clearly isolate boutons in light microscopic images of ChC boutons. could the authors comment on this and explain how they solved this issue (in the methods section for instance)?

We elaborated on our definition of a bouton under confocal microscopy conditions. We also added that the analysis was performed under blinded conditions for the experimenter (i.e. the experimenter did not know whether the images came from trained or untrained mice).

• Is there any suggestion for heterogeneity/selectivity for a subset of PNs (the distribution does not seem to show this, though)? It would be interesting to discuss this and try to link this finding to the rest of the study a bit more directly. Future work could also investigate if genetically defined PN types undergo different pre-synaptic plasticity at their AISs (e.g. work cited by the authors by O'Toole et al, 2023 doi: 10.1016/j.neuron.2023.08.015 -this reference can be updated as well, since the work has been published in the meantime).

In our data, we did not find evidence for heterogeneity or selectivity of targeting, also not in the physiology using KORD (see below). We do agree that it is an interesting question and deserves atention in future experiments. We also updated the reference.

7) ChCs weakly inhibit PyC activity independent of locomotion speed

The authors state that "recent work in adult mice has reported hyperpolarizing and shunting effects in prelimbic cortex, S1 and hippocampus (18, 26, 27)": however, to my knowledge studies presented in refs 26 & 27 found reduced activity/firing of PNs upon optogenetic activation of ChCs in vivo, but did not perform intracellular recordings to assess GABA-A reversal potential at the AIS. I would like to kindly ask the authors to correct this sentence.

If the polarity of responses is discussed, they may rather refer to the corresponding literature including Rinetti Vargas et al (doi: 10.1016/j.celrep.2017.06.030), Lipkin et al (doi: 10.1523/JNEUROSCI.0605- 23.2023), and Khirug et al (doi: 10.1523/JNEUROSCI.0908-08.2008.).

We added the reference to Lipkin et al and changed the sentence so that it matches the references..

• In an atempt to link findings from several parts of the article, did the authors investigate whether chemogenetic effects were different in visual vs non-visual PNs? As ChCs are functionally related to visual PNs, one might indeed speculate that these cells are synaptically connected.

We did not find evidence for selectivity in the chemogenetic effect. We compared the chemogenetic effect to locomotion modulation (see text accompanying Fig 7.) – based on our observation that non- visual PyCs were more strongly modulated by locomotion (see Fig. 4) – but did not find any significant correlation.

• " We first looked at the average activity of neurons in both essions.": sessions

Thank you for noticing. We corrected this.

Discussion

Summary of findings

-It would be worthwhile to include in the summary the finding of mismatch-related activity, that appears to explain more convincingly ChC activation than arousal per se (with the data available).

We updated the summary of the discussion accordingly.

-Moreover, the last part of the article (weak inhibition of PNs by ChCs), despite being very important, is not mentioned.

We now mention this in the summary of the discussion (“Finally, ChCs only weakly inhibit PyCs.”)

Discussion of findings

-" Optogenetic activation of cortical feedback": it is not clear what the authors mean by cortical feedback. As RS was retrogradely labeled, this region may rather provide feedforward inhibition to V1 via ChCs.

Retrosplenial cortex is a higher order cortical area and only provides feedback to V1.

-"This means that each ChC receives input from many L5 PyCs, which could explain the low selectivity of ChC responses we observed to natural images compared to those of L2/3 and L5 PyCs". : perhaps state explicitly that the convergence of many PN inputs each carrying different RF/visual properties "averages out" in ChC (as you do a few lines below for MEI).

At this point, we do not know how the connections from L5 to ChCs are organized. Whether this converge results in “average out” is therefore not so certain. We have made an atempt to clarify the situation. (“This convergence of L5 PyC inputs, if not strongly organized, could explain the low selectivity of ChC responses we observed to natural images compared to those of L2/3 and L5 PyCs.”)

-"However, we did not identify neuromodulatory inputs to ChCs in our rabies tracing experiment. Possibly, these inputs act predominantly through extrasynaptic receptors and were therefore not labeled by the transsynaptic rabies approach.": here, the authors should cite the work by Lu et al (doi: 10.1038/nn.4624) which found basal forebrain (diagonal band of Broca) cholinergic inputs to ChC of the PFC in the Nkx2.1CreER mouse model. Moreover, the authors should discuss potential technical differences (?) responsible for this discrepancy. Beyond the extrasynaptic release of neuromodulators, rabies strains may display different tropism profiles for neuron classes.

We have now added a sentence discussing this and added the reference in the revised manuscript.

-The section dedicated to prediction error is particularly interesting and relevant. In my opinion, this interpretation should be further emphasized in the abstract and summary of findings paragraph in the discussion (as already indicated).

Yes, we agree and have added some emphasis.

-" These findings are thus in contrast with the general notion that ChCs exert powerful control over PyC output (28, 78), but consistent with computational simulations predicting a relatively small inhibitory effect of GABAergic innervation of the AIS, possibly involving shunting inhibition (79, 80)." These findings are also consistent with results from PFC and dCA1 studies showing, with electrophysiological recordings combined with optogenetic stimulation of ChCs, that a small proportion of putative PNs was inhibited upon ChC stimulation (doi: 10.1038/nn.4624 doi: 10.1016/j.neuron.2021.09.033).

Perhaps the effect of ChCs is limited in all these experiments by a suboptimal efficiency of ChC targeting. Moreover, inhibition might be restricted to a subset of PNs carrying a specific function. This could be discussed.

We added an explanation for the weak effects of silencing to the discussion and stated that our results are in line with findings in PFC and CA1. (“One explanation for the weak effects we observed is the high variability in the number of GABAergic boutons that PyCs receive at their AISs. Possibly, only a smaller fraction of PyCs with high numbers of AIS synapses are inhibited when ChCs are active. Indeed, we find that only a small fraction of PyCs increased their activity upon chemogenetic silencing of ChCs, in line with findings by others showing that manipulating ChC activity in vivo has relatively weak effects on small populations of PyCs (27, 28).”)

Although we cannot rule out that ChC targeting is suboptimal in our and other experiments, the expression of the KORD receptor as visualized by mCyRFP1 fluorescence appeared very strong. In addition, the common notion in the ChC field is that ChCs exert powerful control over PyC firing. Even suboptimal labeling should in that case show clear inhibitory effects. Similar experiments with PV+ interneurons would show very convincing inhibition, even if labeling is suboptimal. To keep the discussion concise, we prefer to leave this particular point out.

-" ChC activation could prevent homeostatic AIS shortening of L2/3 PyCs if their activity occurs during behaviorally relevant, arousal inducing events": this postulate seems to be very interesting but is not very clear and lacks some mechanistic speculation.

We considered elaborating more on this hypothesis. However – given that it is merely a speculation at this point – we do not wish to lengthen the discussion further on this point.

• A reference to previous studies demonstrating high levels of synchronous ChC activities is missing: the authors may cite Dudok et al., Schneider-Mizell et al., and Jung et al. (and discuss a change in synchrony with learning or habituation in the case of this study; see above).

We have now also referred to these papers in the context of high correlations between ChCs.

Methods

Beyond references to reagents (eg antibodies, viruses), lot numbers should be provided whenever this is possible. Indeed, there might be strong lot-to-lot variations in specificity and efficiency.

Reviewer #3 (Recommendations For The Authors):

Major:

• (Figure 5) Control analysis missing. Mice before and after training in VR will almost definitely exhibit different running paterns when viewing driftng gratings. Since ChCs are strongly modulated by locomotion, assess whether results depend on changes in running.

Although we did not compare locomotion paterns before and after training, we removed all trials in which the mice were running (see methods). Therefore, we can exclude that these results are caused by changes in running behavior.

• (Figure 5 & 6) What would happen with simple passive visual experience, not in a visuomotor task? What if there was no reward? What if there was an open-loop experiment with random reward? To which specific aspect of the experiment are the results atributable?

These are indeed very interesting questions that may be tested in future experiments.

(Figure 7 B, H) The pre-injection ChC activity in the KORD group is less than 50% of that in control mice! Discuss the effect of such a shift in baseline. Plasticity of PyCs even before ChC inactivation?

See answer to the above question in the public section of reviewer 3.

• (Figure 3 H) Contrast tuning results, as far as I understand, come only from the CNN. However, if I understood correctly, during the passive viewing of gratings there were already different contrasts. Why not show contrast tuning there? Do the results disagree?

We did indeed show stimuli at different contrasts during the passive viewing of gratings. Although the results from those recordings were not optimal for defining contrast sensitivity, they also showed that ChC responses were less modulated by contrast than PyCs.

Minor: - (Figure 3) Explain the potential impact of different indicators 8m vs 6f due to different baselines and dynamics.

We believe there is no impact of different indicators, because for the CNN analyses we estimated spikes using CASCADE. This toolbox is specifically designed to generalize across different calcium indicators. Although GCaMP8m was not included in their training set, the wide variety of indicators used provides a solid basis for generalizable spike estimation. Importantly, comparisons between L2/3 PyCs and ChCs also would not be affected by this concern.

• (Figure 4) NV-PyCs. Would you call all of these mismatch-responsive neurons? Discuss the difference in the percentage of neurons (more than 50% of total PyCs here, compared to significantly less - up to 40% in previous studies, as far as I'm aware)

Not all NV-PyCs appeared to be mismatch-responsive neurons.

• (Figure 6 D) No error bars?

This is a representation of the fraction of all contacted AISs, which has no error bars indeed.

• (Figure 6 E-F and H-I) These pairs of panels contain essentially the same information. The first panel of each pair seems redundant.

We prefer to keep both plots in place, as in this case the skewness of the histogram can be helpful, which is less clear in the boxplot (which in itself displays the quantiles beter).

• The equation for direction tuning still has ang_ori, instead of ang_dir which I'm assuming should be there.

Thank you for noticing, we corrected it.

• The response for drifting gratings is calculated from a different interval (0.2-1.2s) compared to natural images (0-0.5s). Why?

Because we used spike probability in the case of the natural images to shorten the signal, and the visual stimuli were presented for 0.5 s (instead of 1 s as with the gratings).

Very minor:

• It would be helpful for equations to have numbers.

Done

• Sparsity equation. Beter to have it as a general equation, with N instead of 40. Then below it can be explained that N is the number of images = 40.

Done

• "The similarity of these MEIs with those we found for ChCs is in line with the idea that ChCs are driven by input from a large number of L5 PyCs (but do not exclude alternative explanations)." - in parenthesis it should be does not exclude.

Corrected.

• "In contrast, the response strength of PyCs was only mildly and non-significantly reduced after training"

• statistically non-significant..

Corrected.

"We first looked at the average activity of neurons in both essions." - sessions

Corrected.

• (Figure 7 C) Explain what points and error bars represent

Done.

Author Response

Reviewer #2 (Public Review):

The study from Gumaste et al investigates whether mice can use changes of intermittency, a temporal odor feature, to locate an odor source. First, the study tries to demonstrate that mice can discriminate between low and high intermittency and that their performance is not affected by the odor used or the frequency of odor whiffs. Then, they show that there is a correlation between glomerular responses (OSNs and mitral cells) and intermittency. Finally, they conclude that sniffing frequency impacts the behavioral discrimination of intermittency as well as its neural representation. Overall, the authors seek to demonstrate that intermittency is an odor-plume property that can inform olfactory navigation.

The paper explored an interesting question, the use of intermittency of an odor plume as a behavioral cue, which is a new and intriguing hypothesis. However, it falls short in demonstrating that the animal is actually sensitive to intermittency but not other flow parameters, and is missing some important details.

Major concerns

1) One of the cornerstones of this paper consists in showing that mice are behaviorally able to distinguish among different intermittency values (high or low), across a variety of different stimuli and without confounds such as the number of whiffs or concentration. However, I could not find in the paper a convincing explanation of how these confounds were tested. It is clear that the authors repeat their measurements in different conditions (low or high concentration, and different whiff numbers) but it is not specified how: do the authors mix all stimuli in the same session, and so the animals simply generalize across all the stimuli and only consider intermittency for the behavioral choices? Or do authors repeat different sessions for different parameters? For example: do they perform two separate sessions with low concentration and high concentration? If this last one is the case, I would argue that this is not enough proof that animals generalize across concentrations, as the animals might simply use concentration as a cue and change the decision criteria at each session. Please clarify.

We appreciate the reviewer pointing out our oversight in including this information in the manuscript. Trials of the two gain values (which modulate the maximum concentration) are presented interleaved within a session. These trials are solely separated for post-session analysis to test the effect of gain on animal performance. To make this point clearer we have included the following text on line 952 of the manuscript:

“Additionally, trials of a gain of 0.5 and a gain of 1 are interwoven randomly during the session with each unique stimulus being presented at both a gain of 0.5 and 0.1. Thus, after the initial engagement trials, animals are presented with a total of 28 trials at a gain of 0.5 and 28 trials at a gain of 0.1.”

Additionally, to address one of the reviewer’s overarching points, that the manuscript “falls short in demonstrating that the animal is actually sensitive to intermittency but not other flow parameters,” we would like to highlight that through our olfactometer design (described in the Olfactometer Design subsection of the Methods section and illustrated in Figure 1C) the flow rate is held constant throughout the experiment. To further ensure that the animal is not using flowrate or other experimental conditions to perform the task, we tested all animals on a “no odor” condition in which the vial of odor is replaced with a vial of mineral oil. In this condition, their hit rate significantly lowered, as shown in Figure 2C and described in Lines 240- 245:

“Animals’ hit rate also significantly decreased when tested on the Go/No-Go task with the odor vial replaced with mineral oil (n=12 mice, two-sample t-test Naturalistic: odor hit rate = 0.87 ±0.01, no odor hit rate= 0.23 ±0.05, p<0.0001; two-sample t-test Binary Naturalistic: odor hit rate= 0.89±0.01, no odor hit rate= 0.18±0.07, p<0.0001; two-sample t-test Synthetic: odor hit rate= 0.86±0.007, no odor hit rate= 0.23±0.07, p<0.0001), confirming that mice are using odor to perform the task.”

2) It looks to me that the measure of intermittency strongly depends on the set. What is the logic of setting a specific threshold? Do the results hold when this threshold changes within a reasonable range? The same questions (maybe even more important) go for the measure of glomerular intermittence. Unfortunately, a sensitivity analysis for both measures is missing, which makes it hard to interpret the results.

We assume the reviewer suggests that we could have tested discrimination at various Intermittency thresholds. This is indeed wat we did, though not by varying the threshold parametrically (due to abovementioned time constraints), but rather qualitatively/categorically. We tested our mice on 3 stimulus "types" (Figure 1F): actual continuous plume concentration traces (naturalistic), thresholded traces (binarized by threshold 0.1) and square wave (odor agnostic periodic binary). Further, each was tested at 2 gain levels. Figure 2B demonstrates mice discriminate similarly across these 3 widely differing stimuli, while traces were spanning most of the range of possible intermittencies. Reducing the threshold by 1 or 2 orders would skew the range of trials toward many more CS+ trials. We hence conclude that the mice are robustly discriminating and that the paradigm chosen and its associated constraints provide a reasonable test of "intermittency space".

We agree nonetheless that future work should address your suggestion directly by implementing an alternate paradigm. For example, in such a paradigm, mice may be trained to discriminate high vs low intermittencies at varying absolute levels (e.g. 1 vs 0.9 and 0.1 vs 0), etc., however that was well outside the scope of what we aimed to test.

See Figure 1- Supplement 1A. We varied the threshold half a log unit around the 0.1 threshold used in the neuro-behavioral research. As expected, the higher the odor threshold, the more left-shifted the curve. You can see that the monotonic relationship is qualitatively the same across thresholds.

3) The logic of choosing the decision boundary for the discrimination task is not clear: low intermittency is considered to be below 0.15 and high intermittency is considered to be between 0.2 and 0.8. Do these values correspond to natural intermittency distribution? How were these values chosen?

Intermittency drops as function of distance from the source (downwind). It also has a close to normal (with kurtosis) distribution across wind, peaking at the center (see e.g. Crimaldi 2002, Connor 2018). So, animals may encounter any and all intermittencies (0-1). Given our Go/No-Go paradigm we had to set a CS-/CS+ boundary. Typically, to generate an adequate psychometric curve using this paradign, either the CS- or CS+ stimuli need to represent a wide range of values of which the animals are required to compare against a narrow range (or single value). Again, bounded by effective behavioral paradigm design, the number of CS+ and CS- trials need to be even in order to appropriately motivate animals to engage in the task. Thus, considering the entire range of intermittency values animals can encounter while navigating through a plume in conjunction with effective behavioral design, we arrived at our chosen values for low and high intermittency.

As you can see in Figure 1- Supplement 1A (and also reviewer #1, comment 2), I=0.15 is roughly at the knee where the monotonic decrease begins to asymptote. This is roughly true for all 3 concentration thresholds. Consequently, I=0.2-0.8 effectively samples the region where intermittency clearly relates to distance to the source, which is where we hypothesize animals.

4) Only 2 odors were used in the whole study and some results were in disagreement between the two odors. By looking at only two odors it is very difficult to make a general conclusion about intermittency encoding in the OB.

We agree 2 odors are limited, but we were constrained in terms of number of tests that we could run on our cohort of animals. Nonetheless intermittency of both odors is clearly discriminable. As explained to comment 3 by Reviewer 1:

“We indeed considered several odorants and associated properties. Given time constrains we were limited to 2 stimuli of which we had to vary many parameters (type, I, gain, sniffing) in assessing both discrimination and neural processing.”

“Additionally, these two odorants recruit glomeruli in different regions of the dorsal olfactory bulb, have different functional groups and elicit different spatiotemporal response properties in the olfactory bulb (Figure 6- figure supplement 1A, stated on line 507). Both odorants are fruit-associated odors with neutral preference indices (Saraiva et al., 2016, Fletcher, 2012). Thus, while we do not explore a panel of odorants, we do explore the generalizability of intermittency processing with two distinct odorants.”

We decided to test 2 monomolecular odorants (2-heptanone and methyl valerate) as these have been widely used in rodent olfactory bulb imaging, providing distinct and clear glomerular response patterns. They are both fruity smelling odors, implying a relationship to edible food (at least, for humans). Methyl valerate is a methyl ester of pentanoic acid with a fruity (apple) smell and 2-Heptanone is a ketone with a fruity (green banana) smell.

5) Assuming that all the above issues are resolved, one can conclude that intermittency can be perceived by an animal. The study puts a strong accent on the fact that this feature could be used for navigation. I understand that it is extremely hard to demonstrate that this feature is actually used for navigation, however, the analysis of relevance of this measure is missing. Even if it is used in navigation, most probably this would be in combination with other features, thus its relative importance needs to be discussed, or even better, established.

We fully appreciate the reviewers reasoning. Our approach indeed intended to establish a conditio sine qua non: if mice could not discriminate these stimuli they would likely not be able to use intermittency in general for navigation (at least for the odorants tested, for the intermittency ranges tested). We show however that they can, and hence they could use it. To demonstrate their use of intermittency alone or combined with other modalities or properties is well beyond the scope of this manuscript and we agree is a very interesting endeavor.

We discussed other temporal properties on line 58-71 and 657-664 and other general properties on lines 46-56. The relative roles were briefly addressed on lines 664-676 and we hesitate to speculate beyond this.

Author Response

Reviewer #1 (Public Review):

With MERGEseq, the authors sought to develop a scalable and accessible method for getting both projectome and transcriptome information at the single-cell level from multiple projection targets within a single animal. MERGEseq uses a retro rAAV2 to deliver a 15-nucleotide barcode driven by a CAG promoter with co-expression of eGFP to enrich barcoded cells using FACS. Injection of this rAAV2 in distinct regions (with each injection region distinguished by a unique barcode that is specific to the virus used) allows retrograde trafficking and expression of the barcodes in cells that project to the injected region. In this manuscript, rAAVs harboring 5 unique barcodes were stereotactically delivered to 5 targets of the mouse: dorsomedial striatum (DMS), mediodorsal thalamic nucleus (MD), basal amygdala (BLA), lateral hypothalamus (LH), and agranular insular cortex (AI). After a 6-week period to allow for viral transduction and expression, the ventromedial prefrontal cortex (vmPFC) was harvested for scRNAseq. vmPFC scRNAseq data were validated against previously published PFC datasets, demonstrating that MERGEseq does not disrupt transcript expression and identifies the same principal cell types as annotated in previous studies. Importantly, MERGEseq enabled the identification of cell types in the vmPFC that project to distinct areas, with separation occurring largely based on cell type and cortical layer. The application of stringent criteria for barcode index determination is rigorous and improves confidence that barcoded cells are correctly identified. The observation that all barcoded cells were excitatory is consistent with prior work, although it is not clear if viral tropism contributes to this in some way. In a parallel experiment, FAC-sorted cells (vmPFC cells expressing EGFP) were isolated as a comparison. Notably, EGFP+ cells were exclusively excitatory neurons, consistent with literature showing PFC projection neurons are excitatory. Next, barcode analysis was combined with transcriptional identification of neuronal subtypes to define general projection patterns and single-cell projection patterns, which were validated by the DMS and MD in situ using retrograde tracing in combination with RNA FISH. MERGEseq data were also used to identify transcriptional differences between neurons with dedicated and bifurcated projections. DMS+LH and DMS+MD projecting neurons had distinct transcriptional profiles, unlike cells with other targets. RNA FISH for marker gene Pou3f and retrograde tracing from DMS+LH projecting cells demonstrate enrichment of this gene in this projection population. Finally, machine-learning was used to predict projection targets based on transcriptional profiles. In this dataset, 50 highly variable genes (HVGs) were optimal for predicting projection patterns, though this might vary in different circuits. Overall, the results of this manuscript are well presented and include rigorous validation for select vmPFC targets with in situ techniques. The application of unique barcodes for retro-AAV delivery is an accessible tool that other labs can implement to study other brain circuits.

Ultimately, MERGEseq is a subtle conceptual advancement over VECTORseq (retro-AAV delivered transgenes rather than barcodes, in combination with scRNAseq) that offers higher confidence in the described projectome diversity in comparison. The use of a retrograde AAV inherently limits the number of projection areas that can be assessed, a weakness compared to anterograde approaches such as MAPseq/BARseq. However, BARseq demands more time and resources; further, the use of the highly toxic Sindbis virus limits the application of this technique. This manuscript builds upon previous work by utilizing machine learning to predict projection targets. BARseq2 could be used to rigorously validate predicted projectomes and gain single-cell information regarding target neurons. Overall, MERGEseq is an accessible technique that can be used across many animal models and serve as an important starting point to define circuits at the single-cell level.

We thank reviewer for the comprehensive review. We are grateful for reviewer’s recognition of the conceptual advancement of MERGE-seq and the rigorous criteria we applied for projection barcode determination. We have revised the Introduction to highlight advancements in our method. We also discussed the balance of transcriptomic comprehensiveness against spatial resolution in the revised Discussion. Reviewer’s comments have been invaluable in enhancing the clarity and depth of our manuscript.

Reviewer #2 (Public Review):

Investigating the relationship between transcriptomic profiles, their axonal projection and collateralization patterns will help define neuronal cell types in the mammalian central nervous system. The study by Xu et al. combined multiple retrograde viruses with barcodes and single-cell RNA-sequencing (MERGE-seq) to determine the projection and collateralization patterns of transcriptomically defined ventral medial prefrontal cortex (vmPFC) projection neurons. They found a complex relationship: the same transcriptomically defined cell types project to multiple target regions, and the same target region receives input from multiple transcriptomic types of vmPFC neurons. Further, collateralization patterns of vmPFC to the five target regions they investigated are highly non-random.

While many of the biological conclusions are not surprising given recent studies on the collateralization patterns of vmPFC neurons using single neuron tracing and other methods that integrate transcriptomics and projections, MERGE-seq provides validation, at the single cell level, collateralization patterns of individual vmPFC neurons, and thus offer new and valuable information over what has been published. The method can also be used to study collateralization patterns of other neuron types.

Some of the conclusions the authors draw depend on the efficiency of retrograde labeling, which was not determined. Without quantitative information on retrograde labeling efficiency, and unless such efficiency is close to 100%, these conclusions are likely misleading.

We thank reviewer for recognizing the contributions of our MERGE-seq technique in advancing the understanding of projection patterns of vmPFC neurons. We concur that while our conclusions align with previous findings, our single-cell level analysis provides additional depth to the existing knowledge of the field. We acknowledge the challenge to quantify retrograde labeling efficiency to draw quantitive conclusions based on our findings. Alternatively, we have used fMOST-based single-neuron tracing data and analysis to validate our projection patterns and ensure the robustness of our conclusions in the revised manuscript. We also more explicitly clarified the limitations of the quantitive conclusion drawn from MERGE-seq in the revised Discussion. The insights of reviewer are greatly appreciated and will inform the improvement of our research methodology.

Reviewer #3 (Public Review):

This manuscript describes a multiplexed approach for the identification of transcriptional features of neurons projecting to specific target areas at the single-cell level. This approach, called MERGE-seq, begins with multiplexed retrograde tracing by injecting distinctly barcoded rAAV-retro viruses into different target areas. The transcriptomes and barcoding of neurons in the source area are then characterized by single-cell RNA sequencing (scRNAseq) on the 10xGenomics platform. The projection targets of barcoded neurons in the source area can be inferred by matching the detected barcodes to the barcode sequences to of rAAV-retro viruses injected into the target areas.

The authors validated their approach by injecting five rAAV-retro GFP viruses, each encoding a different barcode, into five known targets of the ventromedial prefrontal cortex (vmPFC). The transcriptomes and barcoding of vmPFC neurons were then analyzed by scRNA-seq with or without enrichment of retrogradely labeled neurons based on GFP fluorescence. The authors confirmed the previously described heterogeneity of vmPFC neurons. In addition, they showed that most transcriptionally defined cell types project to multiple targets and that the five targets received projections from multiple transcriptomic types. The authors further characterized the transcriptomic features of barcoded vmPFC neurons with different projection patterns and defined Pou3f1 as a marker gene of neurons extending collateral branches to the dorsomedial striatum and lateral hypothalamus.

Overall, the results of the manuscript are convincing: the transcriptomic vmPFC cell types defined by scRNAseq in this study appear to correlate well with previous studies, the bifurcated projection patterns inferred by barcoding are validated using dual-color retro-AAV tracing, and marker genes for projection-specific cell subclasses are validated in retrogradely labeled vmPFC using RNA FISH for marker detection.

The concept of combining retrograde tracing and scRNAseq is not new. Previous studies have applied recombinase-expressing viruses capable of retrograde labeling, such as CAV, rabies virus, and AAV2-Retro, to retrogradely label and induce the expression of fluorescence markers in projection neurons, therefore facilitating enrichment and analysis of neurons projecting to a specific target. Multiplexed analysis can be achieved with the combination of different reporter viruses or viruses expressing different recombinases and appropriate reporter mouse lines. The advantages of MERGE-seq include that no transgenic lines are required and that it could be applied at even higher levels of multiplexity.

We thank reviewer for the insightful review of our manuscript and the recognition of the advantages of MERGE-seq. We appreciate reviewer acknowledged the robust validation of the method through dual-color retro-AAV tracing and RNA FISH, and the confirmation of previous findings on vmPFC neuronal heterogeneity and collateral projection patterns. We provided additional joint analysis with fMOST-based single-neuron projectome data (Gao et al., 2022, Nature Neuroscience) to further validate the projection patterns (>= 3 targets) that cannot be easily validated with dual-color retro-AAV tracing.

However, previously existing datasets that have already profiled this region with scRNAseq have not been utilized to their full extent. Therefore, for the proper context with prior literature, bioinformatic integration of these scRNAseq and prior scRNAseq data is needed.

Moreover, robust detection of barcodes in neurons labeled by barcoded AAV-retro viruses remains a challenge. The authors should clearly discuss the difficulties with barcode detection in this approach, as well as discuss potential solutions, which are important for others interested in its approach.

While this study is limited to the five known targets of vmPFC, the results suggest that MERGE-seq is a valuable tool that could be used in the future to characterize projection targets and transcriptomes of neurons in a multiplexed manner. As MERGE-seq uses AAVs to deliver barcodes, this method has the potential for application in model organisms for which transgenic lines are not available. Further improvements in experimental design and data analysis should be considered when applying MERGE-seq to poorly characterized source areas or with increased multiplexity of target areas.

In summary, this is a valuable approach, but the authors should clearly provide the context for their study within the existing literature, transparently discuss the limitations of MERGE-seq, as well as suggest improvements for the future.

We appreciate your positive assessment of MERGE-seq as a valuable approach with future potential. As recommended, we have performed integration analysis with existing vmPFC scRNA-seq studies, including Bhattacherjee et al., 2019, Lui et al., 2021, Yao at al., 2021, and specifically recently published MERFISH data of PFC (Bhattacherjee et al., 2023).

In the revised Discussion, we have transparently addressed the current limitations of MERGE-seq, including imperfect retrograde labeling efficiency, variable barcode recovery rates and cell loss during dissociation. We also addressed the challenges in detecting and recovering projection barcodes and suggested potential solutions such as using FAC-sorted EGFP-negative cells for control and applying single-molecule FISH techniques. We sincerely appreciate reviewer’s rigorous and insightful feedback, which has substantially strengthened our manuscript.

Author Response

Reviewer #1 (Public Review):

In this paper, the authors develop new models of sequential effects in a simple Bernoulli learning task. In particular, the authors show evidence for both a "precision-cost" model (precise posteriors are costly) and an "unpredictabilitycost" model (expectations of unpredictable outcomes are costly). Detailed analyses of experimental data partially support the model predictions.

Strengths:

• Well-written and clear.

• Addresses a long-standing empirical puzzle.

• Rigorous modeling.

Weaknesses:

• No model adequately explains all of the data.

• New empirical dataset is somewhat incremental.

• Aspects of the modeling appear weakly motivated (particularly the unpredictability model).

• Missing discussion of some relevant literature.

We thank Reviewer #1 for her/his positive comments on our work and her/his comments and suggestions.

Reviewer #2 (Public Review):

This paper argues for an explanation of sequential effects in prediction based on the computational cost of representing probability distributions. This argument is made by contrasting two cost-based models with several other models in accounting for first- and second-order dependencies in people's choices. The empirical and modeling work is well done, and the results are compelling.

We thank Reviewer #2 for her/his positive comments on our work.

The main weaknesses of the paper are as follows:

1) The main argument is against accounts of dependency based on sensitivity to statistics (ie. modeling the timeseries as having dependencies it doesn't have). However, such models are not included in the model comparison, which makes it difficult to compare these hypotheses.

Many models in the sequential-effects literature (Refs. [7-12] in the manuscript) are ‘leaky-integration’ models that interpret sequential effects as resulting from an attempt to learn the statistics of a sequence of stimuli, through exponentiallydecaying counts of the simple patterns in the sequence (e.g., single stimuli, repetitions, and alternations). In some studies, the ‘forgetting’ of remote observations that results from the exponential decay is justified by the fact that people live in environments that are usually changing: it is thus natural that they should expect that the statistics underlying the task’s stimuli undergo changes (although in most experiments, they do not), and if they expect changes, then they should discard old observations that are not anymore relevant. This theoretical justification raises the question as to why subjects do not seem to learn that the generative parameters in these tasks are in fact not changing — all the more as other studies suggest that subjects are able to learn the statistics of changes (and consistently they are able to adapt their inference) when the environment does undergo changes (Refs. [42,57]).

Our models are derived from a different approach: we derive behavior from the resolution of a problem of constrained optimization of the inference process. It is not a phenomenological model. When the constraint that weighs on the inference process is a cost on the precision of the posterior, as measured by its entropy, we find that the resulting posterior is one in which remote observations are ‘forgotten’, through an exponentially discount, i.e., we recover the predictions of the leaky-integration models, which past studies have empirically found to be reasonably good accounts of sequential effects. (Thus these models are already in our model comparison.) In our framework, the sequential effects do not stem from the subjects’ irrevocable belief that the statistics of the stimuli change from time to time, but rather from the difficulty that they have in representing precise belief; a rather different theoretical justification.

Furthermore, we show that a large fraction of subjects are not best-fitted by precision-cost models (i.e., they are not best-fitted by leaky integration), but instead they are best fitted by unpredictability-cost models. These models suggest a different explanation of sequential effects: that they result from the subjects favoring predictable environments, in their inference. In the revised version of the manuscript, we have made clearer that the derivation of the optimal posterior under a precision cost results in the exponential forgetting of remote observations, as in the leaky-integration models. We mention it in the abstract, in the Introduction (l. 76-78), in the Results when presenting the precision-cost models (l. 264-278), and in the Discussion (l.706-716).

2) The task is not incentivized in any way. Since incentives are known to affect probability-matching behaviors, this seems important. In particular, we might expect incentives would trade off against computational costs - people should increase the precision of their representations if it generates more reward.

We thank Reviewer #2 for her/his attention to our paper and for her/his comments. As for the point on the models, see answer above (point 1).

As for the point on incentivization: we agree that it would be very interesting to measure whether and to which extent the performance of subjects increases with the level of incentivization. Here, however, we wanted, first, to establish that subjects’ behavior could be understood as resulting from inference under a cost, and second, to examine the sensitivity of their predictions to the underlying generative probability — rather than to manipulating a tradeoff involving this cost (e.g. with financial reward). We note that we do find that subjects are sensitive to the generative probability, which implies that they exhibit some degree of motivation to put some effort in the task (which is the goal of incentivization), in spite of the lack of economic incentives. But it would indeed be interesting to know how the potential sensitivity to reward interacts with the sensitivity to the generative probability. Furthermore, as Reviewer #2 mentions, some studies show that incentives affect probability-matching behavior: it is then unclear whether the introduction of incentives in our task would change the inference of subjects (through a modification of the optimal trade-off that we model); or whether it would change their probability-matching behavior, as modeled by our generalized probability-matching response-selection strategy; or both. Note that we disentangled both aspects in our modeling and that our conclusions are about the inference, not the response-selection strategy. We deem the incentivization effects very much worth investigating; but they fall outside of the scope of our paper.

We now mention this point in the Discussion of the revised manuscript (l. 828-840).

3) The sample size is relatively small (20 participants). Even though a relatively large amount of data is collected from each participant, this does make it more difficult to evaluate the second-order dependencies in particular (Figure 6), where there are large error bars and the current analysis uses a threshold of p < .05 across a large number of tests hence creating a high false-discovery risk.

Indeed we agree with Reviewer #2 that as the number of tests increases, so does the probability that at least one null hypothesis is rejected at a given level, even if the null hypothesis is correct. But in the panels a, b and c of Figure 6, about half of the tests are rejected, which is very unlikely under the null hypothesis that there is no effect of the stimulus history on the prediction, all the more as the signs of the non-significant results are in most cases consistent with the direction of the significant results. (In panel e, which reports a finer analysis in which the number of subjects is essentially divided by 2, about a fourth of the tests are rejected, and here also the non-significant results are almost all in the same direction as the significant ones.)

However, we agree that there remains a risk of false discovery, thus we applied a Bonferroni-Holm-Šidák correction to the p-values in order to mitigate this risk. With these more conservative p-values, a lower number of tests are rejected, but in most cases in Fig. 6abc the effects remain significant. In particular, we are confident that there is a repulsive effect of the third-to-last stimulus in the case of Fig. 6c, while there is an attractive effect in the other cases.

In the revised manuscript, Figure 6 now reports whether the tests are rejected when the p-values are corrected with the Bonferroni-Holm-Šidák correction.

(We also applied this correction to the p-values of the tests in Fig. 2, which has more data: the corrected p-values are all below 1e-13, which we now indicate in the caption of this figure.)

4) In the key analyses in Figure 4, we see model predictions averaged across participants. This can be misleading, as the average of many models can produce behavior outside the class of functions the models themselves can generate. It would be helpful to see the distribution of raw model predictions (ideally compared against individual data from humans). Minimally, showing predictions from representative models in each class would provide insight into where specific models are getting things right and wrong, which is not apparent from the model comparison.

In the main text of the original manuscript, we showed the behavior of the pooled responses of the best-fitting models, and we agree with Reviewer #2 that it did not make clear to the reader that the apparent ability of the models to reproduce the subjects’ behavioral patterns was not a misleading byproduct of the averaging of different models. In the original version of the manuscript, we had put a figure showing the behavior of each individual model (each cost type with each Markov order) in the Methods section of the paper; but this could easily be overlooked, and indeed it would be beneficial for the reader to be shown the typical behaviors of the models, in the main text. We have reorganized the presentation of the models’ behaviors: the first panels in Fig. 4 (in the main text) are now dedicated to showing the individual sequential effects of the precision-cost and of the unpredictabilitycost models with Markov order 0 and 1. The Figure 4 is reproduced in the response to Reviewer #1, above, along with comments on the sequential effects produced by these models (and also on the impact of the generalized probability-matching response-selection strategy, in comparison with the traditional probability matching). We believe that this figure makes clearer how the individual models are able to reproduce the patterns in subjects’ predictions — in particular it shows that this ability of the models is not just an artifact of the averaging of many models, as was the legitimate concern of Reviewer #2. We have left the illustration of the firstorder sequential effects of the other models (with Markov order 2 and 3) in the Methods section (Fig. 7), so as not to overload Fig. 4, and because they do not bring new critical conceptual points.

As for the higher-order sequential effects, the updated Figure 5, also reproduced above in the responses to Reviewer #1, now includes the sequential effects obtained with the precision-cost model of a Bernoulli observer (m=0), in addition to the precision-cost model of a Markov observer (m=1) and to the unpredictabilitycost model of a Markov observer (m=3), in order to better illustrate the behaviors of the different models. The higher-order sequential effects of the other models can be found in Fig. 8 in Methods.

Reviewer #3 (Public Review):

This manuscript offers a novel account of history biases in perceptual decisions in terms of bounded rationality, more specifically in terms of finite resources strategy. Bridging two works of literature on the suboptimalities of human decision-making (cognitive biases and bounded rationality) is very valuable per se; the theoretical framework is well derived, building upon the authors' previous work; and the choice of experiment and analysis to test their hypothesis is adequate. However, I do have important concerns regarding the work that do not enable me to fully grasp the impact of the work. Most importantly, I am not sure whether the hypothesis whereby inference is biased towards avoiding high precision posterior is equivalent or not to the standard hypothesis that inference "leaks" across time due to the belief that the environment is not stationary. This and other important issues are detailed below. I also think that the clarity and architecture of the manuscript could be greatly improved.

We thank Reviewer #3 for her/his positive comments on our work and her/his comments and suggestions.

1) At this point it remains unclear what is the relationship between the finite resources hypothesis (the only bounded rationality hypothesis supported by the data) and more standard accounts of historical effects in terms of adaptation to a (believed to be) changing environment. The Discussion suggests that the two approaches are similar (if not identical) at the algorithmic level: in one case, the posterior belief is stretched (compared to the Bayesian observer for stationary environments) due to precision cost, in other because of possible changes in the environment. Are the two formalisms equivalent? Or could the two accounts provide dissociable predictions for a different task? In other words, if the finite resources hypothesis is not meant to be taken as brain circuits explicitly minimizing the cost (as stated by the authors), and if it produces the same type of behavior as more classical accounts: is the hypothesis testable experimentally?

We agree with Reviewer #3 that the relation between our approach and other approaches in the literature should be made clearer to the reader.

Since the 1990s, in the psychology and neuroscience literature, many models of perception and decision-making have featured an exponential decay of past observations, resulting in an emphasis, in decisions, of the more recent evidence (‘leaky integration’, Refs. [7-12, 76-86]). In the context of sequential effects, this mechanism has found a theoretical justification in the idea that people believe that statistics typically change, and thus that remote observations should indeed be discarded [8,12]. In inference tasks with binary signals, in which the optimal Bayesian posterior is in many cases a Beta distribution whose two parameters are the counts of the two signals, one way to conveniently incorporate a forgetting mechanism is to replace these counts with exponentially-filtered counts, in which more recent observations have more weight (e.g., Ref. [12]).

Our approach to sequential effects is not grounded in the history of leakyintegration models: we assume, first, that subjects attempt at learning the statistics of the signals presented to them (this is also the assumption in many studies [712]), and second, that their inference is subject to a cost, which prevents them from reaching the optimal, Bayesian posterior; but under the constraint of this cost, they choose the optimal posterior. We formalize this as a problem of constrained optimization.

The two formalisms are thus not equivalent. Beyond the fact that we clearly state the problem which we assume the brain is solving, we do not propose that the origin of sequential effects resides in an adaptation to putatively changing environments: instead, we assume that they originate in a cognitive cost internal to the decision-maker. If this cost is proportional to the entropy of the posterior, as in our precision cost, then the optimal approximate posterior is one in which remote observations are ‘forgotten’ through an exponential filter, as in the leakyintegration models. In other words, in the context of this task and with this kind of cost, the models are, as Reviewer #3 writes, identical at the algorithmic level. As for the unpredictability cost, it does not result in a solution that resembles leaky integration; about half the subjects, however, are best fitted by unpredictabilitycost models. We thus provide a different rationale for sequential effects — that the brain favors predictive environment, in its inference — and this alternative account is successful in capturing the behavior of a large fraction of the subjects.

In the revised manuscript, we now clarify that the precision cost results in leaky integration, in the abstract, in the Introduction (l. 76-78), in our presentation of the precision-cost models (Results section, l. 264-275), and in the Discussion (l. 706716). (We also refer Reviewer #3 to our response to the first comment of Reviewer #2, above.)

Finally, Reviewer #3 asks the interesting question as to whether the “two accounts provide dissociable predictions for a different task”. Given that the leakyintegration approach is justified by an adaptation to potential changes, and our approach relies on the hypothesis that precision in beliefs is costly, one way to disentangle the two would be to eliminate the sequential nature of the task and presenting instead observations simultaneously. This would eliminate the mere notion of change across time. In this case, the leaky account would predict that subjects’ inference becomes optimal (because the leak should disappear in the absence of change), while in the second approach the precision cost would still weigh on the inference, and result in approximate posteriors that are “wider” (less precise) than the optimal one. The resulting divergence in the predictions of these models is very interesting, but out of the scope of this study on sequential effects.

2) The current analysis of history effects may be confounded by effects of the motor responses (independently from the correct response), e.g. a tendency to repeat motor responses instead of (or on top of) tracking the distribution of stimuli.

We thank Reviewer #3 for pointing out the possibility that subjects may have a tendency to repeat motor responses that is not related to their inference.

We note that in Urai et al., 2017, as in many other sensory 2AFC tasks, successive trials are independent: the stimulus at a given trial is a random event independent of the stimulus at the preceding trial; the response at a given trial should in principle be independent of the stimulus at the preceding trial; and the response at the preceding trial conveys no information about the response that should be given at the current trial (although subjects might exhibit a serial dependency in their responses). By contrast, in our task an event is more likely than not to be followed by the same event (because observing this event suggests that its probability is greater than .5); and a prediction at a given trial should be correlated with the stimuli at the preceding trials, and with the predictions at the preceding trials. In a logit model (or any other GLM), this would mean that the predictors exhibit multicollinearity, i.e., they are strongly correlated. Multicollinearity does not reduce the predictive power of a model, but it makes the identification of parameters extremely unreliable: in other words, we wouldn’t be able to confidently attribute to each predictor (e.g., the past observations and the past responses) a reliable weight in the subjects’ decisions. Furthermore, our study shows that past stimuli can yield both attractive and repulsive effects, depending on the exact sequence of past observations. To capture this in a (generalized) linear model, we would have to introduce interaction terms for each possible past sequence, resulting in a very high number of parameters to be identified.

However, this does not preclude the possibility that subjects may have a motor propensity to repeat responses. In order to take this hypothesis into account, we examined the behavior and the ability to capture subjects’ data of models in which the response-selection strategy allows for the possibility of repeating, or alternating, the preceding response. Specifically, we consider models that are identical to those in our study, except for the response-selection strategy, which is an extension of the generalized probability-matching strategy, in which a parameter eta, greater than -1 and lower than 1, determines the probability that the model subject repeats its preceding response, or conversely alternates and chooses the other response. With probability 1-|η|, the model subject follows the generalized probability-matching response-selection strategy (parameterized by κ). With probability |η|, the model subject repeats the preceding response, if η > 0, or chooses the other response, if η < 0. We included the possibility of an alternation bias (negative η), but we find that no subject is best-fitted by a negative η, thus we focus on the repetition bias (positive η). We fit the models by maximizing their likelihoods, and we compared, using the Bayesian Information Criterion (BIC), the quality of their fit to that of the original models that do not include a repetition propensity.

Taking into account the repetition bias of subjects leaves the assignment of subjects into two families of inference cost mostly unchanged. We find that for 26% of subjects the introduction of the repetition propensity does not improve the fit (as measured by the BIC) and can therefore be discarded. For 47% of subjects, the fit is better with the repetition propensity (lower BIC), and the best-fitting inference model (i.e., the type of cost, precision or unpredictability, and the Markov order) is the same with or without repetition propensity. Thus for 73% (=26+47) of subjects, allowing for a repetition propensity does not change the inference model. We also find that the best-fitting parameters λ and κ, for these subjects, are very stable, when allowing or not for the repetition propensity. For 11% of subjects, the fit is better with the repetition propensity, and the cost type of the inference model is the same (as without the repetition propensity), but the Markov order changes. For the remaining 16%, both the cost type and the Markov order change.

Thus for a majority of subjects, the BIC is improved when a repetition propensity is included, suggesting that there is indeed a tendency to repeat responses, independent of the subjects’ inference process and generative stimulus probability. In Figure 7, in Methods, we show the behavior of the models without repetition propensity, and with repetition propensity, with a parameter η = 0.2 close to the average best-fitting value of eta across subjects. We show, in Methods, that (i) the unconditional probability of a prediction A, p(A), is the same with and without repetition propensity, and that (ii) the conditional probabilities p(A|A) and p(A|B) when η≠0 are weighted means of the unconditional probability p(A) and of the conditional probabilities when eta=0 (see p. 47-49 of the revised manuscript).

In summary, our results suggest that a majority of subjects do exhibit a propensity to repeat their responses. Most subjects, however, are best-fitted by the same inference model, with or without repetition propensity, and the parameters λ and κ are stable, across these two cases; this speaks to the robustness of our model fitting. We conclude that the models of inference under a cost capture essential aspects of the behavioral data, which does not exclude, and is not confounded by, the existence of a tendency, in subjects, to repeat motor responses.

In the revised manuscript, we present this analysis in Methods (p.47-49), and we refer to it in the main text (l. 353-356 and 400-406).

3) The authors assume that subjects should reach their asymptotic behavior after passively viewing the first 200 trials but this should be assessed in the data rather than hypothesized. Especially since the subjects are passively looking during the first part of the block, they may well pay very little attention to the statistics.

The assumptions that subjects reach their asymptotic behavior after being presented with 200 observations in the passive trials should indeed be tested. To that end, we compared the behavior of the subjects in the first 100 active trials with their behavior in the remaining 100 active trials. The results of this analysis are shown in Figure 9.

For most values of the stimulus generative probability, the unconditional proportions of predictions A, in the first and the second half (panel a, solid and dashed gray lines), are not significantly different (panel a, white dots), except for two values (p-value < 0.05; panel a, filled dots). Although in most cases the difference between the two is not significant, in the second half the proportions of prediction A seem slightly closer to the extremes (0 and 1), i.e., closer to the optimal proportions. As for the sequential effects, they appear very similar in the two halves of trials. We conclude that for the purpose of our analysis we can reasonably consider that the behavior of the subjects is stationary throughout the task.

4) The experiment methods are described quite poorly: when is the feedback provided? What is the horizontal bar at the bottom of the display? What happens in the analysis with timeout trials and what percentage of trials do they represent? Most importantly, what were the subjects told about the structure of the task? Are they told that probabilities change over blocks but are maintained constant within each block?

We thank Reviewer #3 for her/his close attention to the details of our experiment. Here are the answers to the reviewer’s questions:

• The feedback (i.e., a lightning strike on the left or the right rod, with the rod and the battery turning yellow if the strike is on the side predicted by the subject,) is immediate, i.e., it is provided right after the subject makes a prediction, with no delay. We now indicate this in the caption of Figure 1.

• The task is presented to the subjects as a game in which predicting the correct location of the lightning strike results in electric power being collected in the battery. The horizontal bar at the bottom of the display is a gauge that indicates the amount of power collected in the current block of trials. It has no operational value in the task. We now mention it in the Methods section (l. 872-874).

• The timeout trials were not included in the analysis. The timeout trials represented 1.27% of the trials, on average (across subjects); and for 95% of the subjects the timeout trials represented less than 2.5% of the trials. This information was added in Methods (l. 887-889).

• Each new block of trials was presented to the subject as the lightning strikes occurring in a different town. The 200 passive trials at the beginning of each block, in which subjects were asked to observe a sequence of 200 strikes, were presented as the ‘track record’ for that town, and the instructions indicated that it was ‘useful’ to know this track record. No information was given on the mechanism governing the locations of the strikes. In the main text of the revised manuscript, we now include these details when describing the task (p. 6).

Author Response

Reviewer #1 (Public Review):

Sun et al. investigated the circuit mechanism of a novel type of synaptic plasticity in the projection from the visual cortex to the auditory cortex (VC-AC), which is thought to play an important role in visuo-auditory associative learning. The key question behind this paper is what is the role of CCK positive projection from the entorhinal cortex in the plasticity of VC-AC projections? They discover that the strength of VC-AC projections does not change when pairing the stimulation of this pathway with the acoustic stimulation of the auditory cortex (AC) unless CCK is applied to the AC or CCK positive projection from the entorhinal cortex to auditory cortex (EC-AC) is optogenetically stimulated. In contrast, optogenetically stimulating VC-AC projections, which express a lower level of CCK than the EC-AC projection, do not induce such synaptic plasticity. Interestingly, the data also indicates that even if the EC-AC pathway is stimulated 500ms ahead of the pairing of stimulating VC-AC pathway and the AC, the VC-AC synaptic strength can still be potentiated, consistent with the long-lasting nature of CCK as a neuropeptide. By performing a fear conditioning assay, the authors demonstrate that the CCK signaling is indeed required for the association of visual and auditory cues.

The proposed mechanism is interesting because it not only helps explain the heterosynaptic plasticity of the visual-auditory projection but also will provide insight into how the entorhinal cortex as an association area contributes to the association of visual and auditory cues. Nevertheless, this study suffers from the lack of a few key experiments, which prevents drawing a conclusion on the contribution of CCK release from the EC-AC projection to the plasticity of the VC→AC projection.

We are grateful for the constructive comments provided by the reviewers and appreciate the significant effort they have dedicated to reviewing our manuscript. To enhance our study and strengthen our conclusions, we have made the following revisions in response to their feedback.

1) One main conclusion from figures 1-3 is that CCK released from the EC-AC projection is required for the plasticity of VC-AC projection in addition to pairing VALS with noise/electrical stimulation. But the data in those figures cannot exclude alternative explanations that CCK alone or the pairing CCK with either VALS or noise are sufficient to make the VC-AC synaptic connection more potent. It concerns the mechanism underlying the effect of CCK: CCK may function simply as a neuromodulator to regulate the excitatory synaptic transmission, but not to promote long term synaptic plasticity.

Thanks for the valuable comment and pointing out the weakness. In response to the comment, we have conducted additional control experiments to reinforce our conclusions. These include: For Figure 1G, we introduced three control groups: CCK alone (Figure1-figure supplement 1F-G), CCK + presynaptic activation of VC-to-AC inputs (Figure 1-figure supplement 1H-I), and CCK + postsynaptic firing induced by noise (Figure 1-figure supplement 1J-K). Our findings from these control experiments indicate that in all three scenarios, there was no potentiation of the VC-to-AC inputs. Further details can be found in Figure 1-figure supplement 1F-K.

For Figure 2E, we introduced three control groups: HFS laser EC-to-AC alone (Figure 2-figure supplement 1H-I), HFS laser EC-to-AC + presynaptic activation of VC-to-AC inputs (Figure 2-figure supplement 1L-M), and HFS laser + postsynaptic firing induced by noise (Figure 2-figure supplement 1P-Q). And we found that in all three scenarios, the VC-to-AC inputs were not significantly potentiated. Please see details in Figure 2-figure supplement 1.

Given that our in vivo results already demonstrated that neither HFS laser EC-to-AC alone, nor its combination with presynaptic or postsynaptic activation, potentiated the VC-to-AC inputs, we did not replicate these control groups in our ex vivo setup. These additional experiments enhance the robustness of our findings and address the initial concerns raised.

2) Similar issue exists in Fig. 2H and 3J. Without proper controls, it is impossible to tell whether all three conditions (HFLSEA, VALA, noise/electrical stimulation) are necessary for potentiated AC responses to acoustic/electrical stimulation.

Same as above, we have conducted additional control experiments to reinforce our conclusions. These include:

For Figure 2H, we also tested the noise response in the above three control groups: HFS laser EC to AC alone (Figure 2-figure supplement 1J-K), HFS laser EC-to-AC + presynaptic activation of VC-to-AC inputs (Figure 2-figure supplement 1N-O), and HFS laser + postsynaptic firing induced by noise (Figure 2-figure supplement 1R-S). And we found that fEPSPs evoked by noise stimuli were significantly potentiated after HFS laser EC-to-AC + Post (Figure 2-figure supplement 1R-S). However, there was no potentiation observed following HFS laser EC-to-AC alone (Figure 2-figure supplement 1J-K) and HFS laser EC-to-AC + Pre (Figure 2-figure supplement 1N-O).

These results suggest that both HFS laser targeting the EC-to-AC projection and noise-induced AC firing are required to potentiate the AC's response to acoustic stimuli. In contrast, activation of the VC-to-AC projection is not necessary. This finding aligns with our previous research (Li et al., 2014).

Given the similarity in experimental design, we opted not to replicate these specific control groups in our ex vivo setup.

These additional control experiments have been crucial in reinforcing the conclusions of our study.

3) Fig. 2E and 3G show that the stimulation of CCK-positive EC-AC projection is required for the plasticity of VC-AC projection. Considering most EC-AC projection neurons co-release glutamate and CCK, however, we cannot tell if CCK or glutamate or both matter to this type of plasticity. Even though the long delay in Fig 5B is consistent with the neuropeptide nature of CCK, direct experimental evidence is needed, since it is where the novelty of the paper is.

Thank you for your constructive feedback. In response to the suggestions, for Figure 2E, we have incorporated two additional experiments: one with a CCKB receptor (CCKBR) antagonist and another with ACSF infused into the AC prior to HFS laser EC-to-AC + Pre/Post Pairing (Figures 2N-P). Our findings demonstrate that the CCKBR antagonist effectively inhibited the potentiation of the VC-to-AC inputs following the HFS laser EC-to-AC + Pre/Post Pairing. Conversely, ACSF did not exhibit this inhibitory effect. For further information, please refer to Figures 2N-P. Given the similarity in experimental design, we opted not to replicate these groups in our ex vivo setup.

4) In Fig. 6, the authors examined the necessity of CCK for the generation of the visuo-auditory association. The experimental approach of injection CCK receptor blocker or CCK-4 is not specific to the EC-AC pathway. There is neither a link between VC-AC plasticity nor this behavioral result. Thus, the explanatory power of this experiment is limited in the context set up by the first 5 figures.

Thank you for highlighting this area for improvement. To enhance the explanatory power of our behavioral experiments, we conducted the following additional studies:

1) Assessing the Necessity of CCK+ EC-to-AC Projection in Establishing Visuo-Auditory Association:

We bilaterally injected AAV9-syn-DIO-hM4Di-eYFP or AAV9-syn-DIO-eYFP into the EC and implanted cannulae in the AC of Cck Ires-Cre mice. During the encoding phase, we inactivated the CCK+ EC-to-AC pathway via CNO infusion into the AC. Our results show that this inactivation prevents the behavioral establishment of an association between the visual stimulus (VS) and auditory stimulus (AS), without affecting the fear conditioning memory to the AS (Figure 6B, beige).

2) Determining the Role of VC-to-AC Projection in Establishing Visuo-Auditory Association: We bilaterally injected AAV9-syn-hM4Di-eYFP or AAV9-syn-eYFP into the visual cortex (VC) and also implanted cannulae in the AC of Cck Ires-Cre mice. Inactivating the VC-to-AC pathway during the encoding phase with CNO infusion in the AC, we observed that this inactivation hinders the establishment of a behavioral association between VS and AS, but does not interfere with the fear conditioning memory to the AS (Figure 6B, red).

3) Investigating the Importance of CCK+ EC-to-AC Projection in Recalling Recent Visuo-Auditory Association:

Again, AAV9-syn-DIO-hM4Di-eYFP or AAV9-syn-DIO-eYFP was injected bilaterally into the EC, and cannulae were implanted in the AC of Cck Ires-Cre mice. By inactivating the CCK+ EC-AC pathway during the retrieval phase with CNO infusion into the AC, we found that such inactivation disrupted the recall of the recent association between VS and AS behaviorally, yet did not affect the fear conditioning memory to the AS (Figure 6D, beige).

4) Assessing the Necessity of VC-to-AC Projection in Recalling Recent Association Memory: For this experiment, AAV9-syn-hM4Di-eYFP or AAV9-syn-DIO-eYFP was injected bilaterally into the VC, and cannulae were placed in the AC of Cck Ires-Cre mice. Inactivating the VC-AC pathway during the retrieval phase with CNO infusion in the AC led to the discovery that this inactivation disrupted the behavioral recall of the recent association between VS and AS but did not disrupt the fear conditioning memory to the AS (Figure 6D, red).

These additional experiments significantly contribute to our understanding of the roles played by the CCK+ EC-AC and VC-AC projections in both the establishment and recall of visuo-auditory associative memories.

5) In page 16, line 322-326, the authors concluded that to induce the plasticity of VC→AC projection, Delay 1 should be longer than 10 ms and Delay 2 should be longer than 0 ms. This conclusion was not fully supported by the data from Figure 5B-D, because there is no data point between -65 ms and 10 ms for Delay 1 (for example 0 ms), and no negative values for Delay 2.

We rewrote this paragraph and hope it is more accurate now.

“Taken together, our study indicates that significant potentiation of the VC-to-AC inputs can be observed (Figure 5D, black cube) across five pairing trials with a 10-second inter-trial interval, under certain tested conditions: (i) the frequency of repetitive laser stimulation of the CCK+ entorhinal cortex (EC) to AC projection was maintained at 10 Hz or higher (as we did not test frequencies between 1 to 10 Hz), (ii) Delay 1 was set within the tested range of 10 to 535 ms (noting the absence of data between -65 to 10 ms), and (iii) Delay 2 was within the range of 0 to 200 ms (acknowledging that negative values for Delay 2 were not explored).”

Reviewer #2 (Public Review):

The manuscript by Sun et al., investigates the synaptic plasticity underlying visuo-auditory association. Through a series of in vivo and ex vivo electrophysiology recordings, the authors show that high-frequency stimulation (HFLS) of the cholecystokinin (CCK) positive neurons in the entorhino-auditory projection paired with an auditory stimulus can evoke long-term potentiation (LTP) of the visuo-auditory projection. However, LTP of the visuo-auditory projection could not be elicited by HFLS of the visuo-auditory projection itself or by an unpaired stimulus. They further demonstrate that auditory stimulus pairing with CCK is required to elicit LTP of the visuo-auditory projection as well as visuo-auditory association in a fear conditioning behavioral experiment. As they found elevated expression of CCK in entorhinal neurons which project to the auditory cortex, they conclude that HFLS of the entorhino-auditory projection causes CCK release.

Strengths:

The authors use an elegant approach with Chrimson and Chronos to stimulate different auditory inputs in the same mouse in vivo and also in slice and demonstrate that potentiation of the visuo-auditory projection is dependent on HFLS of the entorhino-auditory projection paired with auditory stimulus. Furthermore, they test several parameters in a systematic fashion, generating a comprehensive analysis of the plasticity changes that regulate visuo-auditory association.

Weaknesses:

In their previous publications (Chen et al., 2019; Li et al., 2014; Zhang et al., 2020), it has been established that HFLS of the entorhino-auditory projection and CKK release are important for visuo-auditory association via electrophysiology and behavioral experiments. The Chrimson and Chronos approach was applied by Zhang et al., 2020, where they already found that the visuo-auditory projection was potentiated through HFLS of entorhino-neocortical fibers. This manuscript extends those findings by testing different parameters of pairing, which may not represent a major conceptual advance. Unlike the electrophysiological recordings, drug infusion is used in behavioral manipulations to show that HFLS of the entorhino-auditory projection is important for visuo-auditory association. While the use of drugs to inhibit CKK receptors is important, it does not directly demonstrate that CCK release from the entorhino-auditory is necessary.

We deeply appreciate the reviewer's constructive and insightful feedback. Building on our previous work (Zhang et al., 2020), which highlighted the potentiation of the VC-to-AC projection through high-frequency laser stimulation (HFS laser) of entorhino-neocortical fibers, our current study probes further into the intricacies of this process. We have thoroughly explored the specific conditions necessary for the potentiation of the VC-to-AC projection, assessing a wide range of parameters.

A significant advancement in our current research is the elucidation of why HFS of the VC-to-AC pathway alone fails to induce potentiation, whereas HFS of the EC-to-AC pathway, coupled with Pre/Post Pairing, is effective. This critical distinction is linked to the heightened expression of CCK in EC neurons projecting to the AC, in contrast to those from the VC. In this revised version of our study, we have also demonstrated that HFS laser stimulation of the EC-to-AC CCK+ projection induces the release of endogenous CCK in the AC using a combination of a CCK sensor and fiber photometry.

Behaviorally, our revised research emphasizes the vital role of the CCK+ EC-AC projection in both establishing and retrieving visuo-auditory memories, thereby highlighting its fundamental importance in memory processing. Moreover, our study confirms that the CCK+ EC-AC projection is not only crucial for memory formation and retrieval but also indicates that the VC-to-AC projection is the anatomical basis for establishing visuo-auditory associations and serves as the principal storage site for visuo-auditory associative memory. These findings represent significant strides in our understanding of synaptic plasticity and memory mechanisms.

For the behavioral part, to build the link that HFS laser of the EC-to-AC CCK+ projection is important for visuo-auditory association in the behavioral context, we conducted the following additional behavioral studies (for details please see the response to comment 4 of reviewer 1):

1) Assessing the Necessity of CCK+ EC-to-AC Projection in Establishing Visuo-Auditory Associative memories, by inactivating the pathway with inhibitory DREADD during the encoding phase.

2) Investigating the Importance of CCK+ EC-to-AC Projection in Recalling Visuo-Auditory Association, by inactivating the pathway with inhibitory DREADD during the retrieving phase.

Author Response

Reviewer #1 (Public Review):

This paper combines an array of techniques to study the role of cholecystokinin (CCK) in motor learning. Motor learning in a pellet reaching task is shown to depend on CCK, as both global and locally targeted CCK manipulations eliminate learning. This learning deficit is linked to reduced plasticity in the motor cortex, evidenced by both slice recordings and two-photon calcium imaging. Furthermore, CCK receptor agonists are shown to rescue motor cortex plasticity and learning in knockout mice. While the behavioral results are clear, the specific effects on learning are not directly tested, nor is the specificity pathway between rhinal CCK neurons and the motor cortex. In general, the results present interesting clues about the role of CCK in motor learning, though the specificity of the claims is not fully supported.

Since all CCK manipulations were performed throughout learning, rather than after learning, it is not clear whether it is learning that is affected or if there is a more general motor deficit. Related to this point, Figure 1D appears to show a general reduction in reach distance in CCK-/- mice. A general motor deficit may be expected to produce decreased success on training day 1, which does not appear to be the case in Figure 1C and Figure 2B, but may be present to some degree in Figure 5B. Or, since the task is so difficult on day 1, a general motor deficit may not be observable. It is therefore inconclusive whether the behavioral effect is learning-specific.

Thanks for your comments and suggestions.

We have tested the basic movement ability of CCK-/- and WT mice and we found that there were no significant difference between CCK-/- and WT in terms of stride length, stride time, step cycle ratio and grasp force (Figure S1C, S1D, S1E, S1F). Besides, we also have tested the performance of mice injected with CCKBR antagonist or injected with hM4Di together with clozapine after learned the task (Figure S2D, S8D). The performance of mice before and after antagonist injection or chemogenetic manipulation were comparable. These results suggested that all the CCK manipulations did not cause general defects to the movement ability of mice.

The paper implicates motor cortex-projecting CCK neurons in the rhinal cortex as being a key component in motor learning. However, the relative importance of this pathway in motor learning is not pinned down. The necessity of CCK in the motor cortex is tested by injecting CCK receptor antagonists into the contralateral motor cortex (Figure 2), though a control brain region is not tested (e.g. the ipsilateral motor cortex), so the specificity of the motor cortex is not demonstrated.

Thanks for your comments and suggestions.

In this study, we focus on the role played by CCK from the rhinal cortex to the motor cortex, and how CCK affects motor learning. The single pellet reaching task was selected to study the role of CCK from the rhinal cortex to the motor cortex in motor skill learning and the motor cortex is considered as the main area generates motor memory when training in this task (Komiyama et al., 2010; Peters et al., 2014; Richard et al., 2019). We emphasized that the importance of the motor cortex in motor learning, not meant that other brain areas where also receive CCK-positive neural projections from the rhinal cortex, for example hippocampus (spatial memory), are not important for the performance of this task. In fact, specifically inhibiting the projection from the rhinal cortex to the contrallateral motor cortex is not enough to suppress the motor learning ability of, but inhibiting projecting in both sides (contro- and ipsi-lateral) could suppress the learning ability of mice, suggesting that the whole motor cortex is critical for motor skill learning (Figure 6, S8). In this paper, we studied the relationship between the rhinal cortex and the motor cortex and the role played by CCK in this circuit. The specificity of the motor cortex is task-dependent, not the main purpose in this study.

The learning-related source of CCK in the motor cortex is also unclear, since even though it is demonstrated that CCK neurons in the rhinal cortex project to the motor cortex in Figure 4D, Figure 4C shows that there is also a high concentration of CCK neurons locally within the motor cortex. Likewise, the importance of the projection from the rhinal cortex to the motor cortex is not specifically tested, as rhinal CCK neurons targeted for inactivation in Figure 5 include all CCK cells rather than motor cortex-projecting cells specifically.

Thanks for your comments and suggestions.

The specificity of the CCK-projection from the rhinal cortex to the motor cortex for motor skill learning was studies using chemogenetic methods in the revised version of the manuscript. We first determined that over 98% of neurons in the rhinal cortex that projected to the motor cortex are CCK positive (Figure 6A, S6A, S6B). Next, we injected the retro-Cre virus in the motor cortex and the Cre-dependent hM4Di in the rhinal cortex in C57BL/6 mice to specifically inhibit the CCK neurons from the rhinal cortex to the motor cortex. Compared to two control groups, the learning ability of the experimental group was significant suppressed, suggesting that CCK projections from the rhinal cortex to the motor cortex are critical for motor skill learning (Figure 6). Detailed description was added in the part of "Result" in the manuscript.

CCK is suggested to play a role in producing reliable activity in the motor cortex through learning through two-photon imaging experiments. This is useful in demonstrating what looks like normal motor cortex activity in the presence of CCK receptor antagonist, indicating that the manipulations in Figure 2 are not merely shutting off the motor cortex. It is also notable that, as the paper points out, the activity appears less variable in the CCK manipulations (Figure 3G). However, this could be due to CCK manipulation mice having less-variable movements throughout training. The Hausdorff distance is used for quantification against this point in Figure 1E, though the use of the single largest distance between trajectories seems unlikely to give a robust measure of trajectory similarity, which is reinforced by the CCK-/- traces looking much less variable than WT traces in Figure 1D. The activity effects may therefore be expected from a general motor deficit if that deficit prevented the mice from normal exploratory movements and restricted the movement (and activity) to a consistently unsuccessful pattern.

Thanks for your comments and suggestions.

To totally suppress CCK receptors in the motor cortex, the antagonist is unavoidable to diffuse to the adjacent brain areas as the motor cortex is not regularly circular. But the area inhibited most should be the motor cortex. We applied the chemogenetics method to further determine the specificity of the motor cortex in the motor skill learning. Specific projection from the RC to the MC was inhibited bilaterally, which suppressed the motor learning ability.

For a wild-type mouse, neurons were activated when it try to get the food pellet. Neuronal pattern corresponding to each trial will be remembered, and the patterns corresponding to successful movements will tend to be repeated. Manipulations of CCK prevented neurons from remembering the pattern they tried and repeated the pattern they tried before no matter it is successful or not. This is corresponding to the neuron-activation pattern showed in figure 3D, 3E and 3G, the population activities (neuronal activities) are comparable, while the trial-to-trial population correlation is a little bit higher for the CCK-manipulation groups on Day 1. In terms of the behavior, manipulations of CCK decreased the possibility to explore the best path to get food pellets and just repeating a reach for the food pellet like it was the first time. Besides, many tests including the movement ability of CCK-/-, performance of antagonist injection group and chemogenetics manipulation group after learning indicated that CCK-manipulation did not affect the basic movement ability.

Hausdorff distance is the greatest of all the distances from a point in one set to the closest point in the other set. It is not just the largest distance between two trajectories, but comprehensively takes all points in each trajectory into consideration. Hausdorff distance is widely used to assess the variation of two trajectories. The similarity of the shapes of trajectories is not applied for analysis because it is not very effective to assess the performance of a mouse. The fixed location of the initial site and food site makes all trajectories are single lines in the same direction, thus, the shapes of the trajectories are very similar among different trials. Two trajectories with similar shape but far from each other (big Hausdorff distance) should be treated as big variation because, in terms of the final results, they are quite different (success vs. miss). Therefore, Hausdorff distance is more reliable to be applied for assessment of the performance of mice.

Finally, slice experiments are used to demonstrate the lack of LTP in the motor cortex following CCK knockout, which is rescued by CCK receptor agonists. This is a nice experiment with a clear result, though it is unclear why there are such striking short-term depression effects from high-frequency stimulation observed in Figure 6A that are not observed in Figure 1H. Also, relating to the specificity of the proposed rhinal-motor pathway, these experiments do not demonstrate the source of CCK in the motor cortex, which may for example originate locally.

Thanks for your comments.

1. Because CCK4 is a small molecule, which degrades very fast with half-time less than 1 min in the rat serum and 13 min in the human serum, we injected the drug into the electrode recording dishes, while the ACSF was stopped flowing, leading to a relatively low oxygen condition. As it showed in Figure 6A, it cost about 15 min for the brain slices to recover. Compared with CCK4 manipulation, the depression of vehicle group is stronger, which could be due to the effects of CCK4 induced LTP after HFS compensated the depression.

2. In the motor cortex, many CCK-positive neurons are γ-aminobutyric acid-ergic (GABAergic) neurons, in which the role played by CCK is not very clear (Whissell et al., 2015). However, evidence showed that GABA may inhibit the release of CCK in the neocortex (Yaksh et al., 1987). Many glutamatergic neurons in the neocortex also express CCK (Watakabe et al., 2012). In this study, the stimulation electrode was placed on the layer 1, where receives most CCK projections from the rhinal cortex, to release CCK from the rhinal cortex, but can not rule out the possibility that some CCK may release from the local CCK neurons (Figure 4B). We focused on the importance of CCK for neural plasticity in the motor cortex, but did not aim to figure out the role played by the cortical CCK-positive neurons, including inhibitory and excitatory neurons, in neuronal plasticity and motor skill learning by this experiment.

Therefore, the specificity of the projections from the rhinal cortex to the motor cortex was further studied by chemogenetic manipulation. Inhibiting the activity of the projections suppressed the learning ability compared with two types of control manipulations, indicating the CCK projections from RC to the MC is critical for motor skill learning.

Reviewer #2 (Public Review):

This study aims to test whether and if so, how cholecystokinin (CCK) from the mice rhinal cortex influences neural activity in the motor cortex and motor learning behavior. While CCK has been previously shown to be involved in neural plasticity in other brain regions/behavioral contexts, this work is the first to demonstrate its relationship with motor cortical plasticity in the context of motor learning. The anatomical projection from the rhinal cortex to the motor cortex is also a novel and important finding and opens up new opportunities for studying the interactions between the limbic and motor systems. I think the results are convincing to support the claim that CCK and in particular CCK-expressing neurons in the rhinal cortex are critical for learning certain dexterous movements such as single pellet reaching. However, more work needs to be done, or at least the following concerns should be addressed, to support the hypothesis that it is specifically the projection from the rhinal cortex to the motor cortex that controls motor learning ability in mice.

1)Because CCK is expressed in multiple brain regions, as the authors recognized, results from the CCK knock-out mice could be due to a global loss of neural plasticity. In comparison, the antagonist experiment is in my opinion the most convincing result to support the specific effect of CCK in the motor cortex. However, it is unclear to me whether the CCK knock-out mice exhibited an impaired ability to learn in general, i.e., not confined to motor skills. For instance, it would be very valuable to show whether these mice also had severe memory deficits; this would help the field to understand different or similar behavioral effects of CCK in the case of global vs. local loss of function. If the CCK knock-out mice only exhibited motor learning deficits, that would be surprising but also very interesting given previous studies on its effect in other brain areas.

Thanks for your comments. According to the studies in our lab, we found that CCK is critical for the neural plasticity in the auditory cortex, hippocampus and the amygdala and CCK-/- mice performed much worse than wildtype mice in associative, spatial and fear memory (Li et al.,2014; Chen et al., 2019; Su et al. 2019; Feng et al. 2021).

2) Related to my last point, I believe that normal neural plasticity should be essential to motor skill learning throughout development not just during the current task. Thus, it would be important to show whether these CCK knock-out mice present any motor deficits that could have resulted from a lack of CCK-mediated neural plasticity during development. If not, the authors should explain how this normal motor learning during development is consistent with their major hypothesis in this study (e.g., is CCK not critical for motor learning during early development).

Thanks for your comments and suggestions.

Development is mainly gene-guided which prepares the physical structure for learning, while learning is dependent on the neural plasticity and a period of experience (such as motor training in this research). Besides, development is deemed as "experience-expectant", using common environmental information, while learning is "experience-dependent", sensitive to the specific individual experiences (Greenough et al., 1987; Galván, 2010). Moreover, development costs longer time to form a specific ability of a species in general. The role of CCK plays in the development is not clear. Duchemin et al. (1987) studied the CCK gene expression level in the brain of rats pre- and postnatally. They found that the CCK mRNA was detectable on embryonic day 14 (E14) and gradually increased to the maximum level on postnatal day 14 (P14), indicating that CCK might participate in the development of rats. Paolo et al. (2007) mapped the expression of CCK in the mouse brain. Plentiful CCK expression was observed at E12.5 in the thalamus and spinal cord and by E17.5 CCK expression extended to the cortex, hippocampus and hypothalamus, suggesting that CCK might also regulate the development of mice. Paolo et al. (2004) found that CCK suppressed the migration of GnRH-1 through CCK-A receptor in the brain. Besides, postnatal early learning may participate in development. CCK-B receptor antagonist administration (postnatal 6 hours) suppressed the infant sheep get motor preference, indicating that CCK might be important for the development of mother preference of sheep. However, what the role CCK played in the development of motor system is not known.

In this study, the performance of both CCK-/- and WT mice is at the same level without significant difference on Day one, in terms of the percentage of "miss", "no-grasp", "drop" and "success". Besides, the movement abilities, including stride length, stride time, step cycle ratio and grasp force, were comparable for both CCK-/- and WT mice (Figure S1C, S1D, S1E, S1F), suggesting that knockout of cck gene did not affect the basic movement ability. This could be because the development of basic movement ability is not learning-guided, but is physical structure-determined. However, all these tests were on physical level, but how CCK affected the motor system on the molecular and cellular level is not known. Therefore, we further applied CCK-BR antagonist and chemogenetic method to study the role of CCK in the motor learning.

3)Lines 198-200 and Fig. 2C: The authors found that the vehicle group showed significantly increased "no grasp" behavior, and reasoned that the implantation of a cannula may have caused injuries to the motor cortex. In order to support their reasoning and make the control results more convincing, I think it would be helpful to show histology from both the antagonist and control groups and demonstrate motor cortical injury in some mice of the vehicle group but not the antagonist group. Otherwise, I'm a bit concerned that the methods used here could be a significant confounding factor contributing to motor deficits.

Thanks for your comments and suggestions.

The injury of the motor cortex can not be avoided, because the cannula was inserted below the surface of the cortex (Figure S2C). The significantly increased "no-grasp" rate is because the improvement of miss rate of the Vehicle group, which turned to "no-grasp" but failed to further improve to drop or success, while for the Antagonist group, there is no significant improving from "miss" to "no-grasp", leaving no change in the "no grasp".

4) The authors showed that chemogenetic inhibition of CCK neurons in the rhinal cortex impaired motor skill learning in the pellet-reaching task. However, we know that the rhinal cortex projects to multiple brain regions besides the motor cortex (e.g., other cortical areas and the hippocampus). Thus, the conclusion/claim that the observed behavioral deficits resulted from inhibited rhinal-motor cortical projections is not strongly supported without more targeted loss-of-function or rescue experiments.

It would also be very informative to the field to compare the specific behavioral deficits, if any, of inhibiting specific downstream targets of the rhinal CCK neurons. As a concrete example, the hippocampus may be involved in learning more sophisticated motor skills (as the authors pointed out in the Discussion) besides the motor cortex. It would be a critical result if the authors could either show or exclude the possibility that the motor learning deficits observed in CCK-/- mice were at least partially due to the inhibition of hippocampal plasticity. This echoes my earlier point (point 1) that it is unclear whether the effect of lacking CCK in knock-out mice is specific in the motor cortex or engages multiple brain regions.

Lastly, because Fig. 4 only showed histology in the rhinal and motor cortices, I am not sure whether the motor cortex solely receives CCK input from the rhinal cortex. A more comprehensive viral tracing result could be important to both supporting the circuit-specificity of the observed behavior in this study and providing a clearer picture of where the motor cortex receives CCK inputs.

Thanks for your comments.

The specificity of the CCK-projection from the rhinal cortex to the motor cortex for motor skill learning was studies using chemogenetic methods in the revised version of the paper. We first determined that over 98% of neurons in the rhinal cortex that projected to the motor cortex are CCK positive (Figure 6A, S6A, S6B). Next, we injected the retro-Cre virus in the motor cortex and the Cre-dependent hM4Di in the rhinal cortex in C57BL/6 mice to specifically inhibit the CCK neurons from the rhinal cortex to the motor cortex. Compared to two control groups, the learning ability of the experimental group was significantly suppressed, suggesting that CCK projections from the rhinal cortex to the motor cortex are critical for motor skill learning (Figure 6). Detailed description was added in the part of "Result" in the manuscript.

In this study, we focus on the role played by CCK from the rhinal cortex, and how CCK affects motor learning. The single pellet reaching task was selected to study the role of CCK from the rhinal cortex in motor skill learning and the motor cortex is considered as the main area generates motor memory when training in this task (Komiyama et al., 2010; Peters et al., 2014; Richard et al., 2019). We emphasized that the importance of the contrallateral motor cortex in motor learning, not meant that other brain areas where also receive CCK-positive neural projections from the rhina cortex, for example hippocampus (spatial memory), are not important for the performance of this task. In fact, specifically inhibiting the projection from the rhinal cortex to the contrallateral motor cortex is not enough to suppress the motor learning ability, but inhibiting projecting in both sides (contro- and ipsi-lateral) could suppress the learning ability of mice, suggesting that the whole motor cortex is critical for motor skill learning (Figure 6, S8). In our lab, we found that CCK projection from the entorhinal cortex to the hippocampus is critical for spatial memory formation (Su et al., 2019). Impaired hippocampus, to some extent, affected the performance in single pellet reaching task (Shwuhuey et al., 2007). Therefore, manipulation of CCK projections from the rhinal cortex to the hippocampus may also affect the performance in the single pellet reaching task. In this paper, we aim to study the relationship between the rhinal cortex and the motor cortex and the role played by CCK in this circuit. Other brain areas involved in the single pellet reaching task are not the core concern in this study.

The motor cortex also receive CCK projections from other cortices, such as the contrallateral motor cortex, the deep layer of visual cortex and auditory cortex, and thalamus (Figure S4).

5) I am glad to see the CCK4 rescue experiment to demonstrate the sufficiency of CCK in promoting motor learning. However, the rescue experiment lacked specificity: IP injection did not allow specific "gain of function" in the motor cortex but instead, the improved learning ability in CCK knock-out mice could be a result of a global effect of CCK4 across multiple brain regions. CCK4 injection specifically targeted at the motor cortex would be necessary to support the sufficiency of CCK-regulated neuroplasticity in the motor cortex to promote motor learning.

Thanks for your comments.

First, the specificity of the circuit were studied by injecting a Cre virus in the MC and a Cre-dependent hM4Di virus in the RC. After injection with clozapine, the motor learning ability were significantly suppressed compared with the saline control and the control virus combined with clozapine.

Besides, we emphasized that the importance of the motor cortex in motor learning, not meant that other brain areas where also receive CCK-positive neuronal projections from the rhinal cortex, for example hippocampus (spatial memory), are not important for the performance of this task. Specific infusion the drug into the motor cortex is hard to rescue the motor learning ability of CCK-/- mice because the motor cortex is very large, varying from AP: -1.3 to 2.46 mm and ML: ±0.5 to ±2.75 mm and other areas receiving CCK projections from the rhinal cortex also could be important for motor learning. Actually, we tried to inject CCK into the motor cortex through a drug cannula, but the result showed that it is hard to compensate the knock out of cck gene in the whole brain, and rescue the motor learning ability (Figure S11D, S11E). Moreover, cannula implantation causes inescapable injury to the motor cortex, because the cannula must be inserted into the brain, so that the drug could be infused into the brain. This injury may affect the performance in the task, as the motor cortex is very critical for motor learning. Therefore, it is not the best method to be applied for motor skill rescuing.

Furthermore, CCK4 molecules can be transported to the whole brain by i.p. injection, as CCK4 is capable to pass through brain blood barrier, which compensates the knockout of cck gene in the whole brain, leading to the rescuing of motor learning ability. Furthermore, i.p. injection is widely accepted for drug discovery because it is very convenient, simply manipulated and does not causes any direct injury on the brain. Thus, we applied i.p. injection not only for whole brain CCK compensation, but also for the further study of the application in drug discovery.

Reviewer #3 (Public Review):

The authors elucidated the roles of cholecystokinin (CCK)-expressing excitatory neurons, which project from the rhinal cortex to the motor cortex, in motor skill learning. The authors found CCK knock-out mice exhibited learning defects in the pellet reaching task while the baseline success rate of the knock-out mice was similar to that of the wild-type mice. Application of a CCK B receptor (CCKBR) antagonist into the motor cortex lowered the success rate in the motor task. The authors found the population activity which was observed in the in vivo calcium imaging during motor learning was elevated after motor learning, but this increase disappeared in CCK knock-out mice and animals with CCKBR antagonist administration. Anterograde and retrograde viral tracing revealed that CCK-expressing excitatory neurons in the rhinal cortex projected to the motor cortex. Chemogenetic inhibition of the CCK-expressing neurons in the rhinal cortex lowered the ability for motor learning. The application of a CCKBR agonist increased the motor learning ability of CCK knock-out animals as well as long-term potentiation (LTP) observed in the slice of the motor cortex.

However, the manuscript contains several shortcomings:

First, the "Discussion" has several statements that are only supported weakly by the results, for example, ll. 429-431, ll. 432-433, and ll. 447-448. In addition, most of the sentences in this section are not divided into subsections. The paragraphs should be composed in multiple subsections with appropriate subheadings, even though the initial section summarizing the results can lack a subheading.

Thanks for your suggestions. The statements were revised and the discussion was divided into subsections.

Second, it would be important that the authors showed which area(s) of the brain is affected by the CCKBR antagonist in the experiments described in ll. 166-206 and Fig. 2. The authors injected the drug into the motor cortex, but the chemical can spread to neighboring cortical areas (e.g. somatosensory cortex) or wider brain regions. If so, the blockade of the CCKBR in the brain areas other than the motor cortex could cause the defects of the motor task learning observed in these experiments. I think it is desirable that such a possibility should be excluded. Conversely, it is possible that the antagonist had an effect on a limited subarea of the motor cortex (e.g. only the primary motor cortex (M1)). In this case, the information about the field altered by the CCKBR blocker would be useful to interpret the results of the learning defects.

Thanks for your comments and suggestions.

The drug cannula was implanted in the motor cortex (coordinates: AP, 1.4 mm, ML, -/+1.6 mm, DV, 0.25 - 0.3 mm) contralateral to the dominant hand of the mice (Figure S2C). To totally inhibit CCKBR in the motor cortex, we injected over-dosage of antagonist into the motor cortex. Thus, we cannot totally exclude the possibility that some antagonist spread to the neighboring cortices. However, the fact is that the motor cortex is very large, varying from AP: -1.3 to 2.46 mm and ML: ±0.5 to ±2.75 mm. It is not easily to spread out of the motor cortex with high concentration.

Third, the authors need to show bilateral data about their anterograde and retrograde tracking of CCK-expressing neurons in the rhinal cortex. In ll. 290-292, they described as follows: "Both anterograde and retrograde tracking results indicated that CCK-expressing neurons in the rhinal cortex projecting to the motor cortex were asymmetric, showing a preference for the ipsilateral hemisphere." However, they provided only unilateral data for the anterograde (Fig. 4B) and the retrograde (Fig. 4D) experiments.

Thanks for your comments. Both anterograde and retrograde tracking data from bilateral hemisphere were added to the supplementary file (Figure S4).

Fourth, unilateral (contralateral to the dominant forelimb) experiments are needed in the chemogenetic inhibition of the CCK neurons. In ll. 301-338 and Fig. 5, the authors inhibited the CCK -expressing neurons in both hemispheres by injecting the virus into both sides. However, the CCKBR antagonist injection into the motor cortex contralateral to the dominant forelimb caused defects in motor learning ability, as described in ll. 166-206. The authors also observed that the population neuronal activity in the motor cortex contralateral to the dominant forelimb changed in accordance with the improvement of the motor skill in ll. 208-269. Therefore, it may be the case that inhibition of CCK neurons only in the side contralateral to the dominant forelimb - not bilaterally, as the authors did - could cause the lowered ability of motor learning. Such unilateral inhibition can be carried out by unilateral injection of the virus. In relation to the point above, in the chemogenetic inhibition experiments, it would be important to show which neurons in which cortical area is inhibited. This could be done by examining the distributions of the mCherry-labeled somata in the rhinal cortex using histochemistry.

Thanks for your comments and suggestions.

The specific of the CCK-projection from the rhinal cortex to the motor cortex for motor skill learning was studied using chemogenetic methods in the revised version of the paper. We first determined that over 98% of neurons in the rhinal cortex that projected to the motor cortex are CCK positive by retrograde virus injection and immunostaining (Figure 6A, S6A, S6B). Next, we injected the retro-Cre virus in the motor cortex and the Cre-dependent hM4Di in the rhinal cortex in C57BL/6 mice to specifically inhibit the CCK neurons from the rhinal cortex to the motor cortex. Compared to two control groups, the learning ability of the experimental group was significant suppressed, suggesting that CCK projections from the rhinal cortex to the motor cortex are critical for motor skill learning (Figure 6). Furthermore, we also injected the retro-Cre virus into the single site of the motor cortex controlateral to the dominant forelimb together with Cre-dependent hM4Di virus in the rhinal cortex. The result showed that after injection of clozapine, the motor learning ability was not significantly suppressed, suggesting that the bilateral motor cortex is important for motor skill learning. This is consistent with the previous findings that the increased GluA1 expression were observed bilaterally in the motor cortex after training in the single pellet reaching task. Detailed description was added in the part of "Result" in the manuscript.

Fifth, it would be valuable to further examine differences in task performance across sessions and groups. The paragraph in ll. 138-153 needs a comparison of the "miss" rates of CCK-/- animals between Day 1 vs. Day 6 (related to ll. 429- 431). This paragraph also needs comparisons of the "no-grasp" and "drop" rates of CCK-/- animals between Day 1 vs. Day 6 (related to ll. 432- 433). The paragraph in ll. 175-190 needs comparisons of success rates between Day 1 and Day 5/6 within the antagonist group (related to ll. 447-448).

Thanks for your comments. The comparisons were made in the revised manuscript.

Author Response

Reviewer #1 (Public Review):

Strengths:

The study addresses an intriguing research question that fills a gap in existing literature, and was carefully designed and well-executed, with a series of experiments and control experiments.

We thank the reviewer for the positive statement about the conception and execution of the study as well as the potential interest to the community within a broader field.

Weaknesses:

1) My main concern is the null effect of precision estimation pattern between cued and un-cued trials. It is well established that relative to the un-cued stimuli, the cued stimuli obtain more attentional resource and this study claimed serial attentional resource allocation during parallel feature value tracking. However, all Experiments 3a-c did not find any difference in precision estimates between these two types of trials.

We would like to annotate that the terminology „cued versus uncured trials“ in the usual sense of distinguishing between stimuli being attended versus unattended is admittedly somewhat misleading in the current work. In cued and uncured trials of the present experiments 3a-c the allocation of attention is equal. The difference is that the color stream that is attended first is defined (knowable) in the cued but not in the uncued trials. In all cases subjects had to track both color streams and report any of the probed streams as accurately as possible. In other words, the overall allocation of attention in cued and uncured trials is the same. Also, the „cue“ did not provide any information regarding the following probe (no indication of likelihood for a probe in that stream as in an attention experiment). It was entirely irrelevant and was therefore expected not to alter subjects overall performance – as confirmed by the mentioned null-result. The performed test shows, that the reported bias of ~2:1 does not depend on whether in one set of the trials one stream is cued or not. The sole purpose of the “cue” was to subconsciously redirect attention briefly towards that particular stream at the start of each trial in order to ‘phase-reset’ any process, switching/oscillating feature-based resources over time. Performance imbalance across streams is hereby not altered by this phase-reset but remains constant since precision ratio is estimated across a large number of trials and durations. To clarify this issue, we rephrased relevant descriptions in the methods section.

2) Results of Exp.1 in the main text were different from those in Figure.

Thank you for spotting that error. We have corrected the figure accordingly.

3) It would be helpful to add more details for the assignation of response 1 and response 2 to target 1 and target 2, respectively, in all experiments.

For Experiment 2 and 3 only one response per trial was required by the subjects. This design was chosen to avoid potentially ambiguous response-target assignments.

However in the first experiment, as the reviewer points out, subjects gave two color estimates (one for each of the tracked color streams) within each trial. Given that we intend to split subjects’ target-response differences (precisions) into two distributions (based on the idea that each stream is being maintained by an independent attentional resource), there are two possible ways of assigning responses:

(1) We split responses into a best and worst independent of which response was given first.

(2) Alternatively, we assign target-response pairs based on the order of response. The assumption would be, that the first response would be the one with the highest confidence and would be paired with the target closest. This pairing would occur independent of the second response, which is consequently paired with the remaining target. This leaves open the possibility of the second target-response difference being better than the first one due to resource fluctuations. In general, this strategy would be less ‘rigid’ in dividing the two precision-responses into ‘good’ and ‘bad’ responses and was consequently chosen.

To avoid problems arising from the ambiguity of target-response assignments, in all following experiments (2/3), subjects were required to give one response per trial only. We will go into further detail on this issue with reviewer 3 as well, including a numerical example. The logic behind the target-response assignments in experiment 1 has been described in more detail in the methods.

Reviewer #2 (Publlic Review):

The authors asked the question about whether and how changing feature values within the same feature dimensions are tracked. Using a series of behavioral studies combined with modeling approaches, the authors report interesting results regarding a robust, uneven distribution of attentional resources between two changing feature values (in a 2:1 ratio), alternating at 1 Hz. Although the results are clear, it is important to rule out the possible biases due to computational processes. The results advanced our understanding of how parallel tracking of multiple feature values within the same dimension is achieved.

We thank the reviewer for the summary, including the potential impact on the field and we look forward to clarify methodological imprecisions.

Reviewer #3 (Public Review):

The study is interesting and the results are informative in how well people can report colors of two superimposed dot clouds. It reveals that there are trade-offs between reporting two colors. However, I have a few basic but major concerns with the present study and its conclusions about people's abilities to continuously track color values and the rate at which attention may be allocated across the two streams which I am outlining below.

We thank the reviewer for the positive description of our findings and look forward to address any remaining issues.

1) The first concern regards the task that was used to measure continuous tracking of feature values, which in my view is ambiguous in whether it truly assesses active tracking of features or rather short-term memory of the last-seen colors. Specifically, participants were viewing two colored dot clouds that then turned gray, and were asked to report each of the colors they saw using continuous report. The test usually occurred after 6-8s (in Exp. 1 &2), so while not completely predictable, participants could easily perform the task without tracking both feature streams continuously and simply perform the color report based on the very last colors they saw. In other words, it does not seem necessary to know which color belonged to which stream, or what color it was before, to perform the task successfully. Thus, it is unclear to what extent this task is actually measuring active tracking, the same way tracking of spatial locations in multiple-object tracking tasks has been studied, which is the literature that the authors are trying to draw parallels to. In multiple-object tracking tasks, targets and nontarget objects look identical and so to keep track of which of the moving objects are targets, participants need to attend to them actively and selectively. (Similarly, the original feature-tracking study by Blaser et al., at least in their main experiment, people were asked to track an object superimposed on a second object which required continuous and selective tracking of that object).

The reviewer addresses a very fundamental point regarding ‘tracking’ in general: Does tracking rely on attentional processes or mere perception.

The reviewer posits that subjects may simply ‘report based on the very last color they saw’ without the need to track both features streams continuously. Our argument supported by a broad literature on change blindness, inattentional blindness and related phenomena (c.f. Rensink, 2000) is, that one cannot consciously report a changing feature-value without continuously attending to it, in particular when it moves around randomly in feature space. The report of a feature value at a random unpredictable time t by ‘identifying it’ includes its attentive processing immediately before t. Since the time of the probing identification is random, it must continue throughout the trial. We do also rule out any strategy in which subjects only start tracking after some time (the probe appears between 6-8sec after trial onset) since such a strategy would involve processes of temporal attention as well and increase difficulty.

Lastly, the reviewer refers to Blaser et al. as an example in which attentive tracking would be required, since ‘an object [is] superimposed on a second object’. We do absolutely agree. However, the same design principle applies in the current experiment: Two objects with separate values in feature space, that continuously change, are superimposed, that is, spatially inseparable. We do believe that the continuous movement of the feature values through color space separates this work from previous feature-tracking studies like Re et al., in which the presented features remained static. The latter work gives rise to alternate explanations in terms of working memory (mentioned in the next point of the reviewer). Once feature values keep changing and are relevant, a process of updating their internal representations in order to grant access is required (i.e. attention).

2) The main claim that tracking two colors relies on a shared and strictly limited resource is primarily based on the relation between the two responses people give, such that the first response about one color tends to be higher accuracy than for the second response of the other color across participants. In my view, this is a relatively weak version of looking at trade-offs in resources, and it would have been more compelling to show such trade-offs at a single-trial level, or assess them with well-established methods that have been developed to look at attentional bottlenecks such as attention-operating characteristics that allow quantifying the cost of adding an additional task in a precise and much more direct manner.

The reviewer suggests showing trade-offs at a single trial level within subject, which is in essence what we have done in experiment 1. Testing both streams simultaneously, however, has the drawback of introducing interference effects during the report (Reporting the first stream may degrade the precision of reporting the second stream) as well as the mentioned ambiguity between targets and responses. The second and third experiment circumvent this by probing only one color stream, as to analyze the data with a minimal set of assumptions. As the dependent measure of ‘precision’ fluctuates highly across trials, we have to estimate an overall tracking resource by creating a ‘precision’ distribution across many trials.

3) Finally, the data of the last experiment is taken as evidence that feature-based selection oscillates at 1Hz between the two streams. This is based on response errors changing across time points with respect to an exogenous cue that is thought to "reset" attentional allocation to one stream. Only one of three data sets (which uses relatively sparse temporal sampling) shows a significant interaction between cue and time, and given that there was no a priori prediction of when such interaction should occur, this result begs for a replication to ensure that this is not a false positive result. Furthermore, based on the analyses done in the paper, it may very well be the case that the presumed "switching rate" is entirely non-oscillatory based on a recent very important paper by Geoffrey Brookshire (2022, Nature Human Behavior) that demonstrates that frequency analysis are not just sensitive to periodic but also aperiodic temporal structures. The paper also has a series of suggested analyses that could be used here to further test the current conclusions.

The reviewer is absolutely correct in doubting the oscillatory nature of the results in Exp3. Importantly, in our discussion we do not claim that a regular periodicity of the attentional process maintains both color streams. In contrast, we stress the point of ‘one-feature at a time’, indicating a constraint that entails alternation between two representations. We do not presume any sort of regularity of this process but, instead, consider the switching being determined by the recurrent processing of tuning towards one of the two relevant values. Our interpretation is therefore largely in line with Brookshires criticism of previous attentional oscillation studies. In fact, we entirely share the doubtful interpretation of attentional oscillations that transfer mathematical modelling onto functional processes. In our study we use the tool of Fourier transformation in a mere methodological manner, in order to quantify alternations between our color streams but not to imply an underlying oscillatory process. We cannot draw conclusions about underlying attentional oscillations especially since we quantify the alternation/switch only across one full and one half period, in exp3a and exp3b respectively.

We make the distinction between oscillations as a methodological tool and functional cognitive process more clear in the paper.

Author Response

eLife Assessment:

The fluorescently tagged SYT-1 mouse line will be useful for the field. Importantly, the authors used a comprehensive set of immunohistochemical and physiological experiments to demonstrate that the fluorescence tagging did not alter the function of SYT-1. These are important control experiments that will make the strain useful for physiological experiments in the future. However, the advance of this manuscript is less clear.

We thank the editor for raising this point. In the revised manuscript, we performed additonal experiments including testing the expression level of Syt1-TDT and testing the co-labeling of Syt1-TDT with synaptic marker in situ. We also dicussed the advantage of our model compared with the existed ones in line 285 to 300 in the section of discusion. Briefly, we conclude the advance of our models as follows: First, the Syt1-TDT could label synapse in situ, especially in glomerular layer of olfactory bulb (compared with B6SJL-Tg(Thy1-Syt1/ECFP)1Sud/J (Han et al. 2005)). Second, we provided a potential usage of our model in the study of electrophysiological recording and imaging in vivo, as the electrophyiological properties of neurons from Syt1-TDT mice are normal (not be analyzed in B6.Cg-Tg(Thy1-YFP/Syp)10Jrs/J and B6;CBA-Tg(Thy1-spH)21Vnmu/J (Umemori et al. 2004; Li et al. 2005)), which might be result from the relative low expression of Syt1-TDT compared with the native Syt1. Third, the neurons from the transgenic mice can be used in ASF screening by skiping the procedure of immunostaining. It will save the cost of time, reagents and work.

Reviewer #1 (Public Review):

In this manuscript, Zhang and colleagues created a transgenic mouse strain that expresses SYT-1-tdt in all neurons. They showed that the labelled SYT-1 colocalizes with multiple synaptic markers and label synapses in different regions. More importantly, they showed that the transgenic expression does not alter synaptic function using ephys assays. This is a straightforward paper that generated a useful reagent that will be used broadly.

We are grateful for the reviewer’s positive comments.

Reviewer #2 (Public Review):

Yang et al. produced a transgenic mouse line (Syt1-TDT) that could be used for labeling both excitatory and inhibitory synaptic sites in cultured neurons and in vivo neurons. The strength of the current study is to provide a series of thorough analyses to claim the applicability of this mouse line in the relevant neuroscience research field(s). The weakness is the potential impact/usefulness of this mouse line. To strengthen the merit of this mouse line, the authors should present evidence showing its advantage over other similar genetic approaches.

We thank the reviewer for raising this point. To strengthen the merit of this mouse line, we tested the application of Syt1-TDT in labeling synapse in situ. We found that the Syt1-TDT is highly overlapped with synapsin in the brain slice, especially in hippocampus, cerebellum and olfactory bulb, which suggest a potential usage of our model in imaging synapse in vivo. We also compared our transgenic model with the existed ones in line 285 to 300 in the section of discussion in the revised manuscript:

“Several fluorescently tagged synaptic protein transgenic mice model, such as YFP tagged synaptophysin and pHluorin tagged synaptobrevin have been developed to label synapses [49, 50]. While these models can label synapse well, it lacks the functional analysis of neurotransmitter release in the overexpressed neurons as synaptophysin and synaptobrevin were reported to play a role in regulating neurotransmitter release. Considering the overexpression of synaptobrevin or synaptophysin were reported to promote neurite elongation or enhance neurotransmitter secretion, the synaptic organization and synaptic transmission might be changed in these models. Weiping Han et al. in their previous work [47] have generated transgenic mice expressing a Syt1-ECFP fusion protein. The Syt1-ECFP mice expressed the fluorescent protein ECFP in the cortex, midbrain, and cerebellum. However, the expression pattern in their model showed some difference with ours: In the olfactory bulb, the Syt1-TDT signals were highly enriched in glomerular layer in our model, which was not observed in the previously reported Syt1-ECFP transgenic mice [47]. It suggested a potential application of our model in labeling synapse in glomerular layer of olfactory bulb compared with Syt1-ECFP transgenic mice.”

Reviewer #3 (Public Review):

Yang and colleagues provide a thorough characterization of a transgenic mouse model expressing fluorescently tagged synaptotagmin. In particular, they present key controls validating this mouse model as a tool, including co-localization of the tagged synaptotagmin with other synaptic markers as well as normalcy of synaptic transmission mediated by synaptic terminals expressing the tagged synaptotagmin. Importantly, the authors present data on the potential use of neuronal cultures obtained from these mice in synaptic co-culture assays. In these assays, synaptic cell adhesion molecules expressed on non-neuronal cell lines such as HEK-293 cells or COS cells are used to test the sufficiency of these molecules to trigger synapse assembly. This mouse model will be a useful addition to existing models expressing fluorescently-tagged synaptic vesicle proteins such as synaptophysin, synaptotagmin as well as synaptobrevin.

We are grateful for the reviewer’s positive comments.

Author Response

Reviewer #1 (Public Review):

Bakoyiannis et al. investigated the distinct contribution of ventral hippocampal outputs to the nucleus accumbens and medial prefrontal cortex on memory in mice exposed to a high-fat diet (HFD) beginning in adolescence. The authors first characterize the hippocampal to accumbens or mPFC circuits using intersectional viral approaches. They then replicate their previous finding that adolescent HFD contributes to the overactivation of the ventral hippocampus during contextual learning via quantification of c-fos+ cells. In this manuscript, the authors further explore the distinct contribution of these two outputs from the ventral hippocampus using chemogenetics to specifically inhibit one circuit or the other. Interestingly, the authors find that inhibition of either circuit returns c-fos+ cell number to control levels, but the effects on memory are dissociable. They demonstrate that inhibition of output to the NAc rescues HFD-induced deficits on object recognition, while inhibition of mPFC outputs rescues HFD-induced deficits on object location recall. The authors further confirmed that chemogenetic manipulations resulted in alterations in c-fos+ cells that were specific to CA1, and not CA3 or DG. Behaviorally, they excluded any contribution of anxiety on recall, finding no effect on the elevated plus maze.

The strengths of this manuscript include robust behavioral findings that can be attributed to specific circuits. The conclusions of this paper are largely well supported by the data, although some of the methods could provide more detail and the statistical approaches used for analysis need improvement.

We thank the Reviewer for thoroughly summarizing the main results of the study and for providing the comments that we address below.

Reliance on only one measure of anxiety to exclude this as a confound on recall performance is a weakness of the manuscript. To be more convincing that anxiety is not a confound, more than one behavioral assay should be performed.

Reviewer #2 (Public Review):

Bakoyiannis et al. aim to analyze the impact of high-fat diet (HFD) intake during the preadolescent period on memory performances by optogenetically manipulating the circuits responsible for related memory performances. In previous work, they showed the possibility to rescue object-based memory impairments in HFD-exposed animals by silencing the ventral hippocampus (vHPC). Here they investigated further the projections to the nucleus accumbens (NAc) and medial prefrontal cortex (mPFC), 2 of the main monosynaptic targets of the vHPC.

They used a precise strategy to target and manipulate only vHPC cells that project to either NAc or mPFC. They found that preadolescent HFD can induce different types of memory deficits related to different vHPC pathways. In particular, they found that silencing vHPC-NAc, but not vHPC-mPFC, pathway restored HFD-induced object recognition memory deficit. On the other side, silencing vHPC to mPFC, but not vHPC-NAc, pathway rescued HFD-induced object location memory deficits. Moreover, these pathways do not control anxiety-like behaviours since their inactivation has no effect on anxiety levels.

We thank the Reviewer for summarizing the findings of the study and for their positive comments on our manuscript.

The conclusions of the manuscript are mostly supported by the results, but there are some points and controls that need to be addressed and clarified:

• While identifying the relevance of hippocampal cells projecting to NAc and mPFC, a missing control is to verify the activity of vHPC not projecting to these 2 regions in normal conditions or when the investigated pathways are manipulated. This control is essential to refine and bring novel results related to their previous discovery that vHPC overall is involved in the process.

• A downstream effect of their optogenetic manipulation on NAc and mPFC cellular populations should be shown if they want to claim that their chemogenetic inhibition decrease the activation of the pathway and not only of vHPC projecting neurons.

New c-Fos experiments were performed. Please see our response to points 4-5-6 in the “Essential Revision” section.

Reviewer #3 (Public Review):

"Obesogenic diet induces circuit-specific memory deficits in mice" by Bakoyiannis et al., investigates the role of specific ventral hippocampal circuits (specifically to nucleus accumbens and mPFC) in high-fat diet-induced memory deficits. The authors had previously shown that increases in activity in the ventral hippocampus accompany high-fat diet-induced memory deficits, and that inhibition of activity thereby normalizes those memory deficits. In this manuscript, the authors extend these findings to specific projections, showing that they normalize different types of memories by inhibiting the two different pathways.

The strengths of the paper include the pathway-specific manipulations that reveal a difference between the two types of memory. The results are a modest step forward for the field of feeding and learning and memory and would be of interest to that subgroup of neuroscientists. However, the paper also has a number of weaknesses which I detail below.

We thank the Reviewer for summarizing the finding of our study and for the positive feedback.

1) First, the authors show an effect of cfos from both pathways in Figure 2 on object learning. However, the inactivation studies show a pathway-specific effect on object recognition and object location, with no experiments to delineate how this divergence occurs. The authors do not specify whether they compared cfos in the control group between NAc and mPFC projections (presumably they did some controls with each injection), which might reveal differences.

We have added new groups and presented/analyzed the results for each pathway (either vHPC-NAc pathway or vHPC-mPFC pathway) separately for c-Fos (new Figure 2 and Figure 2-Figure Supplement 1) or behaviours (new Figure 3 and Figure 3-Figure Supplement 1). Please see our responses to points 2, 4-5-6 and 9 in the “Essential Revision” section.

2) Related to this, it is unclear how the pathways end up diverging for memory if they do not show any differences in cfos during training. Perhaps there are pathway-specific differences in cfos following the ORM and OLM tests? It is difficult to support the claim that there are pathway differences in memory following inactivation if we do not see any pathway-specific change in activity.

We thank the Reviewer for this comment. Please see our answer to point 7 in the “Essential Revision” section above.

3) Figure 2 and Figure 3 are also hard to interpret because of the usage of a 1-way ANOVA which is not the appropriate statistical test when there are two independent variables (HFD and DREADD manipulation). Indeed, noticing the statistical test also reveals that a critical control missing: HFD -, hM4di+CNO +. It is possible that inactivation simply brings down cfos levels regardless of diet. While this might benefit memory in the case of HFD, it is critical to know whether the manipulation is specific to the overactivation caused by HFD or just provides a general decrease in activity.

Based on this comment we added new HFD-hM4di+CNO+ groups and modified statistical analyses accordingly. Indeed, inactivation of each pathway (vHPC-NAc or vHPC-mPFC) decreases c-Fos in both HFD+ and HFD- (CD+) groups (new Figure 2) whereas it has opposite effect on behaviors, improving memory performance in HFD+ groups but impairing or having no effect in HFD- (CD+) groups (new Figure 3). We have corrected this in the manuscript (please see our responses to points 2 and 9 of “Essential Revision” section).

Author Response

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

eLife assessment

This paper reports the fundamental discovery of adrenergic modulation of spontaneous firing through the inhibition of the Na+ leak channel NALCN in cartwheel cells in the dorsal cochlear nucleus. This study provides unequivocal evidence that the activation of alpha-2 adrenergic or GABA-B receptors inhibit NALCN currents to reduce neuronal excitability. The evidence supporting the conclusions is compelling, the electrophysiological data is high quality and the experimental design is rigorous.

Public Reviews:

Reviewer #1 (Public Review):

This study uses electrophysiological techniques in vitro to address the role of the Na+ leak channel NALCN in various physiological functions in cartwheel interneurons of the dorsal cochlear nucleus. Comparing wild type and glycinergic neuron-specific knockout mice for NALCN, the authors show that these channels 1) are required for spontaneous firing, 2) are modulated by noradrenaline (NA, via alpha2 receptors) and GABA (through GABAB receptors), 3) how the modulation by NA enhances IPSCs in these neurons.

This work builds on previous results from the Trussell's lab in terms of the physiology of cartwheel cells, and from other labs in terms of the role of NALCN channels, that have been characterized in more and more brain areas somewhat recently; for this reason, this study could be of interest for researchers that work in other preparations as well. The general conclusions are strongly supported by results that are clearly and elegantly presented.

I have a few comments that, in my opinion, might help clarify some aspects of the manuscript.

1. It is mentioned throughout the manuscript, including the abstract, that the results suggest a closed apposition of NALCN channels and alpha2 and GABAB receptors. From what I understand, this conclusion comes from the fact that GABAB receptors activate GIRK channels through a membrane-delimited mechanism. Is it possible that these receptors converge on other effectors, for example adenylate cyclase (see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6374141/).

We have now tested the role of adenylyl cyclase modulation in the control of NALCN, by saturating the cells with a cAMP analogue 8-Br-cAMP and found no effect on the NA response. These data are included in the paper. While further experiments are necessary, these results argue in favor of a direct gating by G-proteins.

1. In Figure 2G, the neurons from NALCN KO mice appear to reach a significantly higher frequency than those from WT (figure 2E, 110 vs. 70 spikes/s). Was this higher frequency a feature of all experiments? The results mention a rundown of peak firing rate due to whole-cell dialysis, but, from what I understand, the control conditions should be similar for all experiments.

The peak firing rates in control solutions for WT and KO CWC are not statistically different.

1. Also in Figure 2, the firing patterns for neurons from WT and NALCN KO mice appear to be quite different, with spikes appearing to be generated during the hyperpolarization of the bursts in the second half of the current step for WT neurons but always during the depolarization in KO neurons. Was this always the case? If so, could NALCN channels be involved in this type of firing? Along these lines, it would be interesting to show an example of a firing pattern of neurons from WT mice in the presence of NA, which inhibits NALCN channels.

The specific pattern of spikes in CWC is quite variable from trial-to-trial or cell-to-cell, as it is dependent on multiple CaV and calcium dependent K channels subtypes, and is not dependent on the genotypes used here. The primary effects observed in the KO are in background firing and sensitivity to NA, both reflected alterations in rheobase. The firing pattern example requested was shown in the raster plot of fig 2B2.

1. It might be interesting to discuss how the hyperpolarization induced by the activation of GIRK channels and inhibition of NALCN channels could have different consequences due to their opposite effect on the input resistance.

We considered this as a point of discussion, but decided that making sense of it would depend on assumptions about the location of the channels (dendritic vs somatic, distance to AIS) that we do not have data for. For example, a dendritic increase in resistance through NALCN block, leading to a hyperpolarization of the soma, might have actions similar to a somatic hyperpolarizing conductance increase by GIRK, as far as the voltage at the AIS is concerned.

Reviewer #2 (Public Review):

This is a very interesting paper with several important findings related to the working mechanism of the cartwheel cells (CWC) in the dorsal cochlear nucleus (DCN). These cells generate spontaneous firing that is inhibited by the activation of α2-adrenergic receptors, which also enhances the synaptic strength in the cells, but the mechanisms underlying the spontaneous firing and the dual regulation by α2-adrenergic receptor activation have remained elusive. By recording these cells with the NALCN sodium-leak channel conditionally knocked, the authors discovered that both the spontaneous firing and the regulation by noradrenaline (NA) require NALCN. Mechanistically, the authors found that activation of the adrenergic receptor or GABAB receptor inhibits NALCN. Interestingly, these receptor activations also suppress the low [Ca2+] "activation" of NALCN currents, suggesting crosstalk between the pathways. The finding of such dominant contribution of the NALCN conductance to the regulation of firing by NA is somewhat surprising considering that NA is known to regulate K+ conductances in many other neurons.

The studies reveal the molecular mechanisms underlying well known regulations of the neuronal processes in the auditory pathway. The results will be important to the understanding of auditory information processing in particular, and, more generally, to the understanding of the regulation of inhibitory neurons and ion channels. The results are convincing and are clearly presented.

Reviewer #3 (Public Review):

The study by Ngodup and colleagues describes the contribution of sodium leak NALCN conductance on the effects of noradrenaline on cartwheel interneurons of the DCN. The manuscript is very well-written and the experiments are well-controlled. The scope of the study is of high biological relevance and recapitulates a primary finding of the Khaliq lab (Philippart et al., eLife, 2018) in ventral midbrain dopamine neurons, that Gi/o-coupled receptors inhibit NALCN current to reduce neuronal excitability. Together these studies provide unequivocable evidence for NALCN as a downstream target of these receptors. There are no major concerns. I have only minor suggestions:

Minor

1. As introduced in the introduction, NALCN is inhibited by extracellular calcium which has led to some discourse of the relevance of NALCN when recorded in 0.1 mM calcium. A strength of this study is the effect of NA on NALCN is recorded in physiological levels of calcium (1.2 mM). I suggest including the concentration of extracellular calcium in the aCSF in the Results section instead of relying on the reader to look to the Methods.

Done.

1. It would be interesting to include the basal membrane properties of the KO compared to wildtype, including membrane resistance and resting membrane potential. From the example recording in Figure 2, one might think that the KOs have lower membrane resistance, so it is interesting that the 2 mV hyperpolarization produced similar effects on rheobase. In addition, from the example in Figure 2G, it appears that NA has an effect on firing frequency with large current injection in the KO. Is this true in grouped data and if so, is there any speculation into how this occurs?

We have included in the text a comparison of the input resistance in WT and KO. These were not different. This should not be too surprising given the wide range of values between animals, and the necessity to compare populations. Measurements of resting potential are complicated by the fact that CWC are normally spontaneously active. As was discussed in the text, peak firing frequency declined with time during recording in both control and KO, necessitating normalization as shown in Fig 2E-H.

1. Please expand on the rationale for why GABAB and alpha2 must be physically close to NALCN. To my knowledge, the mechanism by which these receptors inhibit NALCN is not known. Must it be membrane-delimited?

Given the known membrane delimited modulation of GIRK by GABAB, and that alpha2 and GABAB receptors appear to share the same population of NALCN channels, and that alpha2 receptors do not appear to target GIRK channels, we felt the simplest explanation would be coupling through G-proteins, with spatial segregation of different receptor/channel pools providing the means for separating GIRK and NALCN effects. Given that the alpha2 receptor is a Gi/o GPCR, we have now included in the revision new experiments using 8-Br-cAMP, as discussed above. These showed no effect on the NA response, consistent with a direct effect membrane delimited of G-proteins. We acknowledge however that further experiments are warranted.

Reviewer #1 (Recommendations For The Authors):

1. I suggest labeling the voltage traces in Figure 2 with WT and KO for easier comprehension; in addition, I suggest adding the average data to the plots in Figure 2, as in Figure 2-supplementary Figure 1 panel F.

We have added the figure labels as requested. We chose not to add the average data as we noticed that averaging the full FI plots led to a smearing of the curves and a distortion in the apparent rheobase. Thus, we instead measured the rheobase for individual cells and report their average.

1. For readers that are not familiar with the field, more details should be given about the electrical stimulation to evoke IPSCs in cartwheel cells, and what they represent.

Done.

1. The methods should mention if and how the concentrations of divalents were adjusted in the experiments with 0.1 extracellular Ca2+

Done.

Reviewer #2 (Recommendations For The Authors):

I only have several minor comments.

1. The total lack of spontaneous firing in CWCs in the NALCN KO (Fig. 1) is interesting and provides an opportunity to probe the in vivo function of such spontaneous firing. Besides being a little smaller, do the mutant mice have any sign of abnormality in sound signal processing?

Figure 1 – Figure supplement 1 showed that there are no effects on auditory brainstem responses in the KO.

1. Figs. 3&4 (and several other figures with voltage-clamp recordings), a line indicating zero current level would be useful.

Done

1. page 7, "Outward current generated by suppression of NALCN": it might be better to state as "Outward response generated by suppression of NALCN", as the authors correctly pointed out that the NA-induced apparently outward current response is largely a result of an inhibition of NALCN-mediated inward Na+ current. One way to clarify this might be to record at the Nernst potential of K+ to isolate the contribution of Na+ currents (unclear if K+- or Cs+-based pipette was used in the experiment in Fig 3).

Text has been modified.

1. Figs. 5,6&7: do the dashed lines indicate initial current level or zero current level?

Initial current. See legends.

1. The labeling of some of the bar graphs can be made more clear. For example, in Fig. 2K, the right two columns should be labeled as WT as well. Fig. 3C & Fig. 4C, the left two columns should be labeled as WT and the right two as KO.

Added labels to Fig 2 as requested.

1. Figs. 5-7: The suppression of low extracellular [Ca2+]-induced NALCN-dependent current by NA and baclofen is very interesting. As the tonic inhibition of NALCN by extracellular Ca2+ is likely through a Ca2+-sensing GPCR (CaSR) and G-proteins (lowering [Ca2+] releases the inhibition and generates inward current) (Lu et al. 2010), the action of NA and baclofen may all converge onto the same G-protein dependent pathway of the Ca2+-sensing receptor. I'd include this in the discussion to provide a potential mechanistic explanation of the interesting observation.

This is indeed an interesting idea. We prefer not to discuss here, as 1) the source of Ca2+ sensitivity of the channel seems to be controversial (Chua et al 2020), and 2) the effect of Ca2+ reduction is enormously slower than the effect of the modulators (Fig 5-7), implying distinct mechanisms.

Reviewer #3 (Recommendations For The Authors):

Typos/general comments

1. Figure 2 would be easier to comprehend with WT and KO labels as in the other figures. Done

2. Page 11, size of the IPSCs in NA is missing the minus sign.

Corrected.

1. Is the y-axis correct on Figure 8B? This looks like it is doubling the size of the IPSC.

Thank you for catching this mistake. The formula used to calculate % change was in error. We have corrected all the data analysis in the figure, which fortunately did not change the conclusion. Regarding the axis, note that the measurement was % change, not ratio of drug vs control.

Author Response

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

We thank the reviewers and editors for their constructive comments on the manuscript. We have extensively revised the manuscript based on these concerns and comments. The followings are the specific answers.

Public Reviews:

Reviewer #1 (Public Review):

In the manuscript "Long‐read single‐cell sequencing reveals expressions of hypermutation clusters of isoforms in human liver cancer cells", S. Liu et al present a protocol combining 10x Genomics single-cell assay with Element LoopSeq synthetic long-read sequencing to study single nucleotide variants (SNVs) and gene fusions in Hepatocellular carcinoma (HCC) at single‐cell level. The authors were the first to combine LoopSeq synthetic long‐read sequencing technology and 10x Genomics barcoding for single cell sequencing. For each cell and each somatic mutation, they obtain fractions of mutated transcripts per gene and per each transcript isoform. The manuscript states that these values (as well as gene fusion information) provide better features for tumor-normal classification than gene expression levels. The authors identified many SNVs in genes of the human major histocompatibility complex (HLA) with up to 25 SNVs in the same molecule of HLA‐DQB1 transcript. The analysis shows that most mutations occur in HLA genes and suggests evolution pathways that led to these hypermutation clusters. Yet, very little is said about novel isoforms and alternative splicing in HCC cells, differences in isoform ratio between cells carrying different mutations, or diversity of alternative isoforms across cells. While the manuscript by Liu et al. presents a promising combination of technologies, it lacks significant insights, a comprehensive introduction, and has significant problems with data description and presentation.

Answer: Thanks for the precious suggestion. Our long-read single-cell sequencing has discovered an average of 442 novel isoform transcripts per benign liver cell and 450 novel isoform transcripts per HCC cell per SCANTI v1.2 analysis. These are stated in the revised manuscript. The alternative splicing was detected by differential isoform expression as demonstrated in supplemental figures 6 and 7 and supplemental tables 8-11. The examples of differences in isoform ratio between cells carrying different mutations are now shown by DOCK8 and STEAP4 (figure 5 in the revised manuscript). A new section was added in the results to discuss the mutation expression of these two genes. The diversity of isoforms of the selected genes is shown in Supplemental Figure 10.

This study showed how mutations in the same allele evolved in liver cancer. In particular, HLA hypermutations were found to develop from some specific sites of the molecules into large clusters of mutations in the same molecules. A new paragraph of introduction was added about the role of mutations in human cancer development. We also revised the figures to present the information better. All the HLA genes expressed only one known isoform, as shown in Figure 4 and Supplemental Figure 3, regardless of mutations.

Major comments:

1. The introduction section is scarce. It lacks description of important previous works focused on clustered mutations in cancers (for example, PMID35140399), on deriving the process of cancer development through somatic evolution (PMID32025013, from single cell data PMID32807900). Moreover, some key concepts e.g. mutational gene expression and mutational isoform expression are not defined. The introduction and the abstract contain slang expressions e.g. "protein mutation', a combination of terms I teach my students not to use.

Answer: We appreciate the reviewer for the idea of more solid background introduction and term definition. We added a new paragraph in the introduction section to introduce the role of mutations and hypermutations in human cancers. Some important work has been cited. We added a new section in the "Methods" to define "mutation gene expression share" and "mutation isoform expression share". "Protein mutation" has been replaced by "genetic mutation".

1. In the results section, to select the mutations of interest, the authors apply UMAP dimensionality reduction to the mutation isoforms expression and cluster samples in UMAP space, then select the mutations that are present only in one cluster, then apply UMAP to the selected mutations only and cluster the samples again. The motivation for such a procedure seems unclear, could it be replaced with a more straightforward feature selection?

Answer: Thanks for raising up this important question. The goal of the analysis is an unbiased classification of the cell populations in the samples. We found that by removal of mutated isoform expressions that were at similar levels of all cells, the UMAP clustering generated clear segregation of three population cells. When the unique mutated isoform expressions from each group were applied, it generated highly distinct 8 groups of cells, with each group having a distinct mutation isoform expression pattern. If we force known knowledge into the mix of the analysis, it may generate unwanted bias. Specifically, the first UMAP was performed in an unbiased way to cluster cells, while the second step is a supervised approach by selecting the unique mutations in each cluster to identify the classifiers. The second UMAP matches the Benign/HCC labeling well.

1. As I understand, the first "mutated isoform"-based UMAP clustering was built from expression levels of 205 "mutational isoforms". What was the purpose and outcome of the second "mutated isoform"based UMAP clustering (Figure 2E)? In the manuscript the authors just describe the clusters and do not draw any conclusions or use the results of the clustering anywhere further.

Answer: Thanks for pointing this out. Figure 2E was generated from unique mutation isoform expressions in groups A, B, and C from Figure 2D. The purpose of Figure 2E is to investigate whether these unique mutation isoforms can further classify the cell populations free of prior biological knowledge. We added a sentence in the revision to clarify the purpose of the clustering. The conclusion from this analysis, including Figure 2F and Figure 3 (which is an extension of Figure 2E), is that HLA mutation isoform expressions dominated the classifications of cell populations.

1. The authors just cluster the data three times based on expression levels of different sets of "mutational isoforms" and describe the clusters. What do we need to gather from these clustering attempts besides the set of 113 mutations used for further analysis? What was the point of the reclusterings? Did the authors observe improvement of the classification at each step?

Answer: Thanks for asking this important question. The improvement of re-clustering to classify cell populations is the obvious segregation of 8 different groups of cells without any manual classification through prior knowledge. The distances among groups were far apart in comparison to the first clustering (figure 2B). Detailed subclassifications were achieved on cell populations that otherwise could not be segregated based on the first clustering.

1. The alignment of short reads generated from hypermutated transcriptomes is non-trivial. The proposed approach could address the issue without the need for whole genome sequencing and offer insights about the cancer development through somatic evolution. Why didn't the authors use modern phylogenetic approaches in the "Evolution of mutations in HLA molecules" section or at least utilize the already performed clustering to infer cell lineages?

Answer: We appreciate for the great question. For a single molecule mutation evolution, single gene clustering may not produce a desirable and robust effect. A simple evolution snowball chart in Figure 4B may be easier to be understood.

1. I am not sure I understood the definition of "mutated gene expression levels" and "mutated isoform expression levels" in the "Mutational gene expression and fusion transcript enhanced transcriptome clustering of benign hepatocytes and HCC" section. The authors mention that gene lists included all the isoforms within the same range of standard deviation. If I understand it correctly, they are equal if there is only one expressed transcript isoform. In that case, this overlap is not surprising at all.

Answer: We thank the reviewer for the great question. The definition of mutation gene expression level, mutation isoform expression level, and fusion gene expression level are now defined in the "Methods" section. In all HLA mutation transcripts, there were multiple transcripts with or without mutations for a single dominant isoform.

1. "To investigate the roles of gene expression alterations that were not accompanied with isoform expression changes, UMAP analyses were performed based on the non‐overlapped genes." Venn diagrams (Sup Figure 8) show that there are much less "non-overlapped genes" than "genes that showed both gene and isoform level changes" for each SD threshold (for example, for SD>=0.8 59 vs 275). Could that be the reason why clustering based on the former group is worse i.e the cancer and normal cells are separated less clearly?

Answer: The number of (attributes) genes could be a contributing factor in the segregation of cell populations. However, the number of attributes is not the underlying reason for worse performance for gene only classifier because much smaller isoforms/genes (22) overlap in SD>=1 outperformed a large number of genes (59) with SD>=0.8. It suggested that 59 gene expression classifier is less efficient in segregating the cell populations. To address this concern, we took SD>=0.8 as an example for demonstration if we subsampled the 275 overlapped genes/isoforms to 59 (equal to 59 non-overlapped genes in terms of number), we can still get better separation than the 59 DEG only. We repeated this subsampling process for three times. Similar results were found. The new data were inserted into supplemental Figure 8

Reviewer #2 (Public Review):

In the present study, Liu et al present an analysis of benign and HCC liver samples which were subjected to a new technology (LOOP-Seq) and paired WES. By integrating these data, the authors find isoforms, fusions and mutations which uniquely cluster within HCC samples, such as in the HLA locus, which serve as candidate leads for further investigation. The main appeal of the study is in the potential of LOOPSeq as a method to present isoform-resolved data without actually performing long-read sequencing. While this presents an exciting new method, the current study lacks systematic comparisons with other technologies/data to test the robustness, reproducibility and utility of LOOPSeq. Further, this study could be further improved by giving more physiologic context and examples from the analyses, thus providing a new resource to the HCC community. A few suggestions based on these are below:

Answer: We appreciate the reviewer to raise up all the important questions and the great suggestions. The LOOPseq technology was compared with Oxford nanopore and PacBio long-read sequencing in our previous study. We have cited analysis in the introduction section of the paper. HLA mutation clusters in the single molecules are our finding with major physiological significance since these mutations may help liver cancer cells evade immune surveillance. We have extensively discussed the potential impact of these mutations on cancer development in the discussion. In addition, we added a new section of DOCK8 and STEAP4 mutation expressions in the results (page 11, new Figure 5) that are highly relevant to the pathogenesis of HCC.

1. A primary consideration is that this seems to be the first implementation of LOOP-Seq, where the technology, while intriguing, has not been evaluated systematically. It seems like a standard 10x workflow is performed, where exons are selectively pulled down and amplified. Subsequent ultra-deep sequencing is assumed to give isoform-resolution of the sc-seq data. To demonstrate the utility of the approach it would benefit the study to compare the isoform-resolved results with studies where long-read sequencing was actually performed (ex: https://journals.lww.com/hep/Fulltext/2019/09000/Long_Read_RNA_Sequencing_Identifies_Alternativ e.19.aspx, https://www.jhep-reports.eu/article/S2589-5559(22)00021-0/fulltext, https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1010342). Presumably, a fair amount of overlap should occur to justify the usage.

Answer: We have discussed the utility of the methodology in comparison with the previous studies by these three groups in the revision (results, page 12).

1. Related to this point, the sc-seq cell types and benign vs HCC genes should be compared with the wealth of data available for HCC sc-seq (https://www.nature.com/articles/s41467-022-322833, https://www.nature.com/articles/s41598-021-84693-w). These seem to be important to benchmark the technology in order to demonstrate that the probe-based selection and subsequent amplification does not bias cell type definition and clustering. In particular, https://www.nature.com/articles/s41586021-03974-6 seems quite relevant to compare mutational landscapes from the data.

Answer: This is a great point. The consistency probe-based analysis was demonstrated in our previous analyses and the analyses mentioned in the comments. We further discussed it in the results section of the paper (page 12).

1. From the initial UMAP clustering, it will be important to know what the identities are of the cells themselves. Presumably, there is quite a bit of immune cells and hepatocytes, but without giving identities, downstream mechanistic interpretation is difficult.

Answer: When mutation analyses were combined with cell marker analysis, i.e., immune marker positive but negative in HLA mutation, we found only one bona fide immune cell in the HCC sample. Thus, immune cells may not be significant in the current analysis.

1. In general, there are a fair amount of broad analyses, such as comparisons of hierarchical clustering of cell types, but very little physiologic interpretations of what these results mean. For example, among the cell clusters from Fig 6, knowing the pathways and cell annotations would help to contextualize these results. Without more biologically-meaningful aspects to highlight, most of the current appeal for the manuscript is dependent on the robustness of LOOP-seq and its implementation.

Answer: To address this comment, a new pathway analysis was performed on the cluster results of Figure 6. A new supplemental table was generated. The results are now discussed on page 13.

1. Many of the specific analyses are difficult and the methods are brief. Especially given that this technology is new and the dataset potentially useful, I would strongly recommend the authors set up a git repository, galaxy notebook or similar to maximize utility and reproducibility

Answer: The script file has been uploaded to GIT to facilitate the reproducibility of the analysis. We also added a new pipeline description script in the methods (pages 19-20).

1. The authors claim that clustering between benign and HCC samples was improved by including isoform & gene (Suppl fig 8). This seems like an important conclusion if true, especially to justify the use of longread implementation. Given that the combination of isoform + gene presents ~double the number of variables on which to cluster, it would be important to show that the improved separation on UMAP distance is actually due to the isoforms themselves and not just sampling more variables from either gene or isoform

Answer: The number of (attributes) genes could be a contributing factor in the segregation of cell populations. However, the number of attributes is not the underlying reason for worse performance for gene only classifier because much smaller isoforms/genes (22) overlap in SD>=1 outperformed a large number of genes (58) with SD>=0.8. It suggested that 58 gene expression classifier is less efficient in segregating the cell populations. To address this comment, we performed random subsampling to reduce the isoform/gene overlap iterates, similar results were obtained. A new supplemental figure was generated to reflect the new analyses.

1. SQANTI implementation to identify fusions relevant for the HCC/benign comparison. How do the fusions compare with those already identified for HCC? These analyses can be quite messy when performed on WES alone so it seems that having such deep RNA-seq would improve the capacity to see which fused genes are strongly expressed/suppressed. This doesn't seem as evident from current analysis. There are quite a bit of WES datasets which could be compared: https://www.nature.com/articles/ng.3252, https://www.nature.com/articles/s41467-01803276-y

Answer: Exome sequencing is not an ideal tool to identify fusion genes. Very few fusion genes have been discovered based on RNA sequencing so far. The fusion genes discovered in the study appeared mostly novel. No exome sequencing was involved in the identification of fusion genes.

1. Figure 4 is fairly unclear. The matrix graphs showing gene position mutations are tough to interpret and make out. Usually, gene track views with bars or lollipop graphs can make these results more readily interpretable. Also, how Figure 4 B infers causal directions from mutations is unclear.

Answer: We appreciate the reviewer for pointing this out. We have revised the diagram in Figure 4A to reflect the proper distance between the mutations in HLA-DQB1 NM_002123. Since these are the positions in the same alleles (protein), the gene track view or lollipop graph may not show that properly. The mutation clusters started from an isolated mutation, and mutation did not revert to wild type sequence after occurring. Based on these two principles, we showed several mutation accumulation pathways leading to hypermutation clusters.

Reviewer #3 (Public Review):

The Liu, et al. manuscript focuses on the interesting topic of evaluating in an almost genome-wide-scale, the number of transcriptional isoforms and fusion gene are present in single cells across the annotated protein coding genome. They also seek to determine the occurrences of single nucleotide variations/mutations (SNV) in the same isoform molecule emanating from the same gene expressed in normal and normal and hepatocellular carcinoma (HCC) cells. This study has been accomplished using modified LoopSeq long‐read technology (developed by several of the authors) and single cell isolation (10X) technologies. While this effort addresses a timely and important biological question, the reader encounters several issues in their report that are problematic.:

1. Much of the analysis of the evolution of mutations results and the biological effects of the fusion genes is conjecture and is not supported by empirical data. While their conclusions leave the reader with a sense that the results obtained from the LoopSeq has substantive biological implications. However, they are extended interpretations of the data. For example: The fusion protein likely functions as a decoy interference protein that negatively impacts the microtubule organization activity of EML4.(pg 9)... and other statements presented in a similar fashion.

Answer: We thank the reviewer for the helpful comment. The mutation results were experimentally validated by exome sequencing on the same samples. Furthermore, these mutations were filtered by requiring their presence in three different transcriptomes. The biological significance of these mutations is probably the subject of investigation in the next phase. Since a large number of HLA mutations did not occur overnight, the analysis of the accumulation pathways for these mutations was warranted, given the extensive evidence of such a process. The impact of mutations on HLA molecules appeared obvious and should be discussed. For ACTR2-EML4 fusion, we revised it as "The loss of microtubule binding domain may negatively impact the microtubule organization activity of EML4 domain of the fusion protein." We only discussed the obvious impact due to the loss of a large protein domain.

2, LoopSeq has the advantage of using short read sequencing analyses to characterize the exome capture results and thus benefits from low error rate compared to standard long-read sequencing techniques. However, there is no evidence obtained from standard long read sequencing that the isoforms observed with LoopSeq are obtained with parallel technologies such as long read technologies. It is not made clear how much discordance there is in comparing the LoopSeq results are with either PacBio or ONT long read technologies.

Answer: The comparative analyses among LOOPSeq, Oxford nanopore, and PacBio sequencing were performed in our previous study. We have cited the study in our introduction.

1. There is no proteome evidence (empirically derived or present in proteome databases) from the HCC and normal samples that confirms the presence or importance of the identified novel isoforms, nor is there support that indicate that changes in levels HLA genes translate to effects observed at the protein level. Since the stability and transport differences of isoforms from the same gene are often regulated at the post-transcriptional level, the biological importance of the isoform variations is unclear.

Answer: Given the transcriptome sequencing data, we can only focus on the isoform variation analysis but not directly link to the protein level variation because of the post-transcriptional level regulation. We discussed this in the revised manuscript (page 14).

4 It is unclear why certain thresholds were chosen for standard deviation (SD) <0.4 (page 5), SD >1.0 (pg 11).

Answer: The threshold is flexible and arbitrary. We showed different thresholds, and the same conclusion holds. We just choose the thresholds with better separation and a reasonable number of genes/isoforms for the downstream analysis. (Supplemental Figure 6-7 with different thresholds and supplemental tables 4-12).

1. HLA is known to accumulate considerable somatic variation. Of the many non-immunological genes determined to have multiple isoforms what are the isoform specific mutation rates in the same isoform molecule? Are the HLA genes unique in the number of mutations occurring in the same isoform?

Answer: We thank the reviewer for this important suggestion. We now show mutation expression patterns in isoforms of DOCK8 and STEAP4 in Figure 5. A new section is added to discuss the mutation expression of these two genes. As shown in supplemental figure 10, HLA-DQB1, HLA-DRB1, HLA-B, and HLA-C, have only one known isoform detected,

Editorial comments:

The present study pairs single-cell seq with LoopSeq synthetic long-read sequencing on samples of HCC and benign liver to identify mutations and fusion transcripts specific to cancer cells. The authors present a potentially important resource; however the overall support remains incomplete.

While the approach of evaluating isoform-specific changes at the cellular level to cancer seeks to address a timely and important topic, there is currently incomplete evidence in support of the major claims in the manuscript. In particular, major recommendations to provide stronger support for the combination of technologies and interpretation regarding cancer-associated genomic changes include: 1) systematic evaluation of UMAP-based clustering methods, to what subsets of data they are applied and subsequent interpretations, 2) direct comparisons of results with additional methods to quantify long-read sequencing data and those evaluating mutational consequences of HCC progression and 3) detailed expansion of the description of methods and rationale for selecting specific parameters and cell types for further analyses. Including these changes would significantly strengthen the support for utility of combining 10x single-cell with Loop-seq and provide compelling evidence for usage of this resource in dissecting HCC-associated molecular changes.

Answer: We appreciate the frank and constructive comments. The goal of UMAP is to obtain biological knowledge through unbiased data selection. Systematically, we select classifiers without any prior knowledge (blind to the samples). In our case, classifiers with high standard deviation across all the cells were chosen. We stressed this in the result section. The comparison among LOOPSeq, PacBio, and Oxford nanopore was made in our previous study. We cited that analysis in this paper. Analysis detail and pipelines were added in the revised manuscript to improve the reproducibility. The mutation expression analysis was quite clear-cut. The clustering classified the HCC and benign liver cells by itself and identified a few cancer cells in the benign liver sample. All these were accomplished without applying any knowledge.

Reviewer #1 (Recommendations For The Authors):

Overall, there are numerous problems with data presentation and insufficient description, which authors could fix.

1. Figure 4. A. It would be more clear if the figure showed the distribution of mutations in the molecule. Otherwise, it's hard to see if we see clusters of mutations or just 25 mutations spread uniformly across the transcript. B. It's unclear what the reader needs to take away from these columns of numbers.

Answer: The mutation positions are now presented as proportion to the location in a molecule. Column B is the distribution of mutation molecules from left panel in each cluster of cells (from Figure 3A) and their sample origin (HCC or benign liver). We clarify it a little more in the legend of Figure 4A.

1. As a reader, I did not understand how "mutated gene expression levels" and "mutated isoform expression levels" were calculated in terms of sequenced long reads

Answer: We defined the term and calculations in the methods section of the revised manuscript.

1. Page 6 "genes involving antigen presentation"

Answer: The full sentence of the subtitle is" Mutations of genes involving antigen presentation dominated the mutation expression landscape."

1. Page 6 "These unique mutational isoforms" - how are these isoforms unique?

Answer: We take away most of the "unique" adjectives to describe the non-redundant mutations.

1. Page 6. Unclear "All but one clusters contained cells co‐migrated with cells of their sources."

"Among 113 mutation isoforms, the major histocompatibility complex (HLA) was the most prominent with 68 iterations (60.2%) (Supplemental Table 3, Figure 3B)" There is nothing about HLA in Figure 3B.

Answer: We revised the sentence as "Cells in all but one clusters co-migrated with cells of their sources". The mutation isoform expressions were listed in supplemental Table 3. They are too small and become unreadable when put in the figure.

1. Page 10 "genes or isoforms that across all samples had with expression standard deviations less than" - probably "with" should not be there.

Answer: We correct the error and thank the reviewer for the comment.

1. Page 11 "UMAP analysis was performed using genes with standard deviations {greater than or equal to} 1.0 (182 wild‐type genes) and standard deviations >0.4 (282 mutated genes)". What do "wild-type" and "mutated" mean here?

Answer: We edited as "UMAP analysis was performed using gene expressions with standard deviations ≥ 1.0 (182 non-mutated genes) and gene mutation expression with standard deviations 0.4 (282 mutated genes)."

1. I could not find the description of Supplementary Tables.

Answer: The supplemental table legends are added in the revised manuscript.

1. In the Discussion section, the authors mention that mutations were mainly expressed in a specific isoform of a gene for a given cell. I suggest to emphasize this point in the Results section and illustrate it with a comparison of abundance of mutated and non-mutated isoforms

Answer: For HLA molecules, their expression appeared to be restricted to one known isoform, regardless of mutation status. This sentence is removed in the revision. A new section of DOCK8 and STEAP4 mutation expression is added to the result.

1. It is also mentioned that mutations may have an impact on the RNA splicing process. The authors should compare the observed isoform ratio to a prediction of the effect of variants on splicing by SpliceAI or similar tools

Answer: This sentence was removed from the discussion.

1. Figure 3c: triangles corresponding to HLA-positive cells are hard to distinguish

Answer: We provide a larger representation of the triangle and circle in figure 3c in the revision.

Reviewer #2 (Recommendations For The Authors):

Many of my comments could be addressed by spending time to provide the code/data and a walkthrough of analyses so that other users would be able to answer these questions on their own.

Answer: We have included a script section in the revision to ensure the reproducibility of the analysis. The raw data had been uploaded to GEO (see Methods).

Author Response

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

Reviewer #1:

1. The results that TF binding produces microdomains at medium and long linker DNA but not short linker is very interesting. Although the differences can be observed from the figure, it still lacks of quantitative comparison. It is not clear the exact definition of the microdomain observed from simulations and what numbers of microdomains can be identified under different conditions. A quantitative comparison of different conditions could also be provided.

We thank the reviewer for this suggestion. Our intent was to show qualitatively how TF binding locations that we design can direct fiber folding and create microdomains, which we define in the paper as high frequency contact regions in the contact maps, similar to the TADs observed in HiC maps. Together with the fiber configurations, contact maps allow us to identify formation of such microdomains, and to observe how these microdomains change depending on the conditions we build into the model, such as TF binding region or linker DNA length.

To address your point, we have added a clustering analysis of the contact matrices with nucleosome resolution and assign each contact along the genome position (nucleosome index) to a cluster. In Supporting Figure S6, we show how DBSCAN clustering provides a clustering distribution that quantitatively describes the microdomains observed in the matrices and estimates the number of microdomains. For example, in the 44 and 62 bp systems, the contacts along the genomic distance separate into 5, 2, and 1 nucleosome groups for topologies 1 to 3, and into 2 and 1 group for topology 4, respectively. In the 26 bp and Life-Like systems, where microdomains are more diffuse due to fiber rigidity or polymorphism, we see that the clustering results are not as TF-topology-dependent as in the 44 and 62 bp systems. We also decomposed the contact matrices into one dimensional plots that depict the magnitude of 𝑖, 𝑖 ± 𝑘 internucleosome interactions. We see that internucleosome patterns change with the TF binding topology, and that the 26 bp and Life-Like systems show the least changes.

1. When increasing TF concentration, from 0 to 100%, it seems that both packing ratio and sedimentation coefficients are not sensitive to the TF concentrations after 25%. Is it due to the saturation of TF binding? How many TF binding sites are considered at each concentration?

Yes, in most cases, at TF concentrations higher than 25%, the fiber compaction does not change due to saturation of TF binding. Although the TF concentrations are reached, such as 50%, 70%, or 100%, these do not influence the fiber architecture. A higher order folding and compaction cannot be reached due to excluded volume interactions that impede overlapping of beads in the model.<br /> We have clarified this in the manuscript.

As stated in the Methods section, the TF concentration refers to the number of linker DNA beads that can engage in a constraint compared to the total number of linker DNA beads. Thus, at 25% TF, 25% of linker DNA beads are engaged in TF constraints. We have added a comment on this in the Results section.

1. It is shown that the contact maps that reveal microdomains are ensemble-based maps and single trajectories do not show clear formation of microdomains. Does the formation of microdomains increase with the number of combined trajectories?

The formation of microdomains occurs in each single trajectory. However, the microdomains formed in each trajectory can be different. That is why ensemble-based maps show clearer trends of microdomains that might not be as visible in single-trajectory maps. If we increase the number of trajectories, the macrodomains will be more visible and there will be more macrodomains in the contact map, but the formation of microdomains will not increase in each single trajectory.

1. "As we see from Figure 4A, when the linker DNA is short, such as 26 and 35 bp, TF binding does not increase the packing ratio of the fiber." The results of 35bp cannot be found in Figure 4A. In addition, the color of 44 and 62 bp should be changed since they are very similar in the figure.

Thank you for catching this. The results corresponding to the 35 bp system are presented in the Supporting Figure 7. We have changed the text to read “As we see from Figure 4A and Figure S7..”.

We have changed the color of the 62 bp trace to blue in the plots of Figure 4. Consistently, we have also changed the color of the 62 bp fiber in Figure 2 and Figure 5.

1. For modelling of TF binding at increasing concentrations, it is mentioned that in these three conditions, TFs are allowed to bind to any region. Do you mean TF can also bind to nucleosomal DNA? Nucleosome structure prevents the binding of many TFs.

In our model, only linker DNA beads can engage in the constraints (bind TF).<br /> We have changed the text to read “TFs are allowed to bind to any linker DNA region”.

1. The details of the Mnase-seq dataset and how NFRs are identified should be provided, such as the coverage of the data and what read fragments are selected for NFR mapping.

MNase data in bedgraph format were downloaded from the Genome Expression Omnibus (GSM2083107) repository and loaded without further processing into the Genome Browser. NFRs were visually inspected and detected as genomic regions without peaks. As detailed in the GEO repository, the sequenced paired-end reads were mapped to the mm9 genome. Only uniquely mapped reads with no more than two mismatches were retained and reads with insert sizes less than 50 or larger than 500 bp were discarded.

We have clarified this in the manuscript.

1. The calculations of volume and area of the Eed promoter region should be further elucidated.

Thank you. We now elaborate upon these calculations. In particular, the Eed promoter region is defined between cores 123 and 129. The x,y or x,y,z coordinates of those cores are used to create the bounding area or volume by defining the shape’s vertices.

1. In Figure 3, it is not clear how different topology are identified.

In Figure 3 the topology, or TF binding regions, is the same for each of the 10 contact maps as these emerge from trajectory replicas of the same system which we named Topology 1. Different microdomains are formed in each individual trajectory as the high-frequency regions appear in different locations on each contact map. However, when these 10 maps are summed, the ensemble contact map clearly shows consensus microdomains in each region where TF binds.

Reviewer #2:

To further improve the manuscript, I have the following suggestions/comments.

1. While most of the conclusions in this paper follow from the evidence provided by the ximulations, the result in section 3.3 title "Gene locus repression is medicated by TF finding," may not follow from the results. In my opinion, repression is a more complex process, and many more factors (such as nucleosome positioning, nucleosome sliding, histone methylation, and other proteins such as PRC or HP1, etc) may be involved in repression. While compaction is often associated with repressed chromatin (heterochromatin), recent studies have shown that heterochromatin fibers are highly diverse, and compaction alone may not be the criteria for repression (eg. see Spracklin et al. Nat. Struct. Mol. Biol. 30, 38-51 (2023).). In this light, I would recommend slightly modifying the title to say, "TF binding-mediated compaction can help in gene locus repression" or something similar.

Yes! We completely agree that gene repression is a very complex phenomenon that involves many factors that we are approaching by modeling starting from the simplest strategy. Thus, we have changed the subtitle to read “TF binding-mediated compaction as possible mechanism of gene locus repression”.

1. Authors could also present the contact probability versus genomic distance. This may provide some generic features at nucleosome resolution, given the variability in linker length and LH density.

We thank the reviewer for this suggestion. We have now calculated the contact probability for the EED gene with and without TF binding (Supporting Figure 8). We see that the contact probability corresponding to short range interactions (i ± 2, 3, 4, 5, and 6) is slightly lower for the EED gene upon TF binding. However, a striking increase in the contact probability upon TF binding is seen in the genomic region between 3 and 5 kb, which corresponds to local loop interactions. Thus, TF binding slightly decreases local interactions but increases chromatin loops. Such changes are not observed for the EED system with LH density 0.8 (Supporting Figure 9), further supporting the idea that an increase in LH density hampers the effect of TF binding for the EED gene architecture. <br /> We have now added these results to the manuscript.

1. Write a short paragraph about the limitations of the model/study. For example, one of the limitations could be that, as of now, it has only the effect of a few proteins, but to predict repression, one may need to incorporate the effect of several proteins.

We agree with the reviewer that our model is a simple, first-step approach. Nonetheless, even the simplest mathematical model can be enlightening in helping dissect essential factors. Here, our model clearly shows how TF binding location modulates fiber architecture and the interplay between TF binding and other chromatin elements, like linker DNA length, LH density, and histone acetylation. We have now stated in the Discussion section that although limited due to being implicit and not considering other protein partners, our model can provide insights on the regulation of chromatin architecture by protein binding. Future modeling with explicit protein binding or combination of several proteins will further help us understand genome folding regulation.

1. The radius of gyration of 26 kb chromatin is around ~60nm in this paper. Is there any experimental measurement to compare (approximate order of magnitude)? While I do not know any measurement for Eed gene locus, I am aware of the results in the Boettiger et al. paper from Xiaowei Zhuang lab (Nature 2016). There, they find that the Rg of a 26 kb region is above 100nm. But that is for a different organism, a different set of genes. Also, see Sangram Kadam et al. Nature Communications 14 (1), 4108, 2023.

Thank you for this suggestion. To the best of our knowledge, there are no radius of gyration measurements for the EED gene. Regarding the two papers you cite, in the paper from Boettiger et al. (1) they determine by microscopy experiments that Rg ∝ 𝐿! where 𝐿 is the genomic length and 𝑐 is 0.37 ± 0.02 for active chromatin (Figure 1d of the paper). In such case, the Rg for a 26 kb region would be 43 ± 9 nm. Considering that these are Drosophila cells, our value of 62 nm is in good agreement with that estimate. Regarding the Kadam et al. paper (2), by coarse grained modeling they find an Rg of around 100 nm for different genes. Considering that the radius of gyration depends on cell type and fiber configuration (see for example (3) for the dependency of Rg on loop number and persistence length), we believe that our measurements in the same ball park as experimental results and other theoretical modeling studies are good indicators of our model’s reasonableness.

We have added this comparison to the manuscript.

1. The reason why it is useful to compare some distance measurements (physical dimension) with experiments is the following: The contact map in Hi-C only gives relative contact probabilities. It does not give absolute contact probabilities. To convert a Hi-C map into a physical distance, one requires comparison with some experimentally measured 3D distance. The radius of gyration is an ideal quantity to compare. From my experience, the contact probability is often much smaller than 1, suggesting that the chromatin is more expanded. But this could be due to the effect of many other proteins in vivo and the crowding, etc. I do not expect this work to incorporate all those effects. However, it may be useful to make a comment about it in the manuscript.

Thank you. We have added to the discussion a comment on our first-generation model of TF binding to chromatin and the neglect of many associated protein and RNA cofactors that certainly influence chromosome folding and domain formation on higher scales. Some distance measures are also added to the Results as mentioned above.

References

1. Boettiger,A.N., Bintu,B., Moffitt,J.R., Wang,S., Beliveau,B.J., Fudenberg,G., Imakaev,M., Mirny,L.A., Wu,C. and Zhuang,X. (2016) Super-resolution imaging reveals distinct chromatin folding for different epigenetic states. Nature, 529, 418–422.

2. Kadam,S., Kumari,K., Manivannan,V., Dutta,S., Mitra,M.K. and Padinhateeri,R. (2023) Predicting scale-dependent chromatin polymer properties from systematic coarsegraining. Nat. Commun., 14, 4108.

3. Wachsmuth,M., Knoch,T.A. and Rippe,K. (2016) Dynamic properties of independent chromatin domains measured by correlation spectroscopy in living cells. Epigenetics Chromatin, 9, 57.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

I have only a few very minor suggestions for improvement.

• the text repeatedly uses the terms "central nervous system" and "enteric nervous system", which are not in standard use in the field. These terms are not defined until the bottom of p. 12 even though they are used earlier. It would be useful for the authors to explicitly describe their definitions of these terms earlier in the paper.

Fixed.

• the inclusion of four pre-trained models is a powerful and useful aspect of WormPsyQi. Would it be possible to develop a simple tool that, when given the user's images, could recommend which of the four models would be most appropriate?

We appreciate the reviewer for bringing this up. To address this, we have now added an additional function in the pipeline to test all pre-trained models on representative input images. Before processing an entire dataset, users can view all segmentation results for images in Fiji to assess which model performed best, judged by the user. The GUI, running guide document, and manuscript have been modified accordingly.

In addition, we would like to emphasize that the pre-trained models were developed by iterative analyses of many reporters, often with multiple rounds of parameter tuning; the results were validated post hoc to choose the optimal model for each reporter, and we have listed this information in Supplemental Table 1 to inform the choice of the pre-trained model for commonly used reporter types.

• On p. 11 (and elsewhere), the differences in the performance of WormPsyQi and human experimenters are called "statistically insignificant". This statement is not particularly informative (absence of evidence is not evidence of absence). Can the authors provide a more rigorous analysis here - or provide an estimate of the typical effect size of the machine-vs-human difference?

To address this, we have included additional analysis in Figure 2 – figure supplement 3. For two reporters - I5 GFP::CLA-1 and M4 GFP::RAB-3 - we compare WormPsyQi vs. labelers and inter-labeler puncta quantification. A high Pearson correlation coefficient (r2) reflects greater correspondence between two independent scoring methods. We chose these two test cases to demonstrate that the machine-vs-human effect size is reporter-dependent. For I5, where the CLA-1 signal is very discrete and S/N ratio is high, the discrepancy between WormPsyQi, labeler 1, and labeler 2 is minimal (r2=0.735); moreover, scoring correspondence depends on the labeler (r2=0.642 and 0.942, respectively). In other words, WormPsyQi mimics some labelers better than others, which is to be expected. For M4, where the RAB-3 signal is diffuse and synapse density is high in the ROI, the inter-labeler discrepancy is high (r2=0.083) and WormPsyQi vs labeler (1 or 2) discrepancy is slightly reduced (r2=0.322 and 0.116, respectively). The problematic regions for the M4 RAB-3 reporter are emphasized in Figure 6 - figure supplement 1A. Overall, the additional analysis suggests that the effect size is contingent on the reporter type and image quality, and importantly for scoring difficult strains WormPsyQi may average out inter-labeler scoring variability.

• p. 12: "Again, relying on alternative reporters where possible..." This is an incomplete sentence - are some words missing?

Edited.

Reviewer #2 (Recommendations For The Authors):

1. The authors effectively validated the sexually dimorphic synaptic connectivity by comparing the synapse puncta numbers of PHB>AVA, PHA>AVG, PHB>AVG, and ADL>AVA. However, these differences appear to be quite robust. It would be beneficial for the authors to test whether WormPsyQi can detect more subtle changes at the synapses, such as 10-20% changes in puncta number and fluorescence intensity.

While the dimorphic strains were used to first validate WormPsyQi based on the ground truth of very well-characterized reporters, the reviewer reasonably asks whether our pipeline can pick up on more subtle differences. To address this, we have now included an additional figure (Figure 9 – figure supplement 2), where we performed pairwise comparisons between L4 and adult timepoints for the reporter M3 GFP::RAB-3. As reflected in panels A and C, although the difference between puncta number and mean intensity between L4 and adult is marginal (22% increase in puncta number and 13% increase in mean intensity from L4 to adult), WormPsyQi can pick it up as statistically significant.

1. On page 10, the authors mentioned that "cell-specific RAB-3 reporters have a more diffuse synaptic signal compared to the punctate signal in CLA-1 reporters for the same neuron, as shown for the neuron pair ASK (Figure 4 -figure supplement 1B, C)". It is important to note that in this case, the reporter gene expressing RAB-3 is part of an extrachromosomal array, whereas the reporter gene expressing CLA-1 is integrated into the chromosome. It's possible that the observed difference in pattern may arise from variations in the transgenic strategies employed.

To emphasize the difference in puncta features inherent to the reporter type, we have now added WormPsyQi segmentation results for ASK CLA-1 extrachromosomal reporter (otEx7455) next to the ASK CLA-1 integrant (otIs789) and ASK RAB-3 reporter (otEx7231) in Figure 4 – figure supplement 1C. Importantly, otEx7455 was integrated to generate otIs789, so they belong to the same transgenic line. Literature shows that RAB-3 and CLA-1 have different localization patterns and corresponding functions at presynaptic specializations, and this is qualitatively and quantitatively shown by the significant difference in puncta area size between RAB-3 and both CLA-1 reporters, i.e., both CLA-1 reporters have smaller, discrete puncta compared to RAB-3 (Figure 4 – figure supplement 1C). Quantitatively, in the case of ASK - where the synapse density is sparse enough that even diffuse RAB-3 puncta can be segmented without confounding adjacent puncta – overall puncta number between otEx7231 and otIs789 are similar. However, RAB-3 signal is diffuse and this poses quantification problems in cases where the synapse density is higher (e.g. AIB, SAA in Figure 4 – figure supplement 1D) and WormPsyQi fails to score puncta in these reporters since the signal is not punctate. As far as integrated vs. extrachromosomal reporters go, the reviewer is right in pointing out that some differences may be stemming from reporter type as our additional analysis between otIs789 and otEx7455 indeed shows fewer puncta in the latter owing to variable expressivity.

1. The authors mentioned that having a cytoplasmic reporter in the background of the synaptic reporter enhanced performance. It would be more informative to provide comparative results with and without cytoplasmic reporters, particularly for scenarios involving dim signals or densely distributed signals.

The presence of a cytoplasmic marker is critical in two specific scenarios: 1) images where the S/N ratio is poor, and 2) when the image S/N ratio is good, but the ROI is large, which would make the image processing computationally expensive.

To demonstrate the first scenario, we have included an additional panel in Figure 4 – figure supplement 1(B) to show how WormPsyQi performs on the PHB>AVA GRASP reporter with and without the channel having cytoplasmic marker. The original image was processed as-is in the former case with both the synaptic marker in green and cytoplasmic marker in red; for comparison, only the green channel having synaptic marker was used to simulate a situation where the strain does not have a cytoplasmic marker. As shown in the figure, in the presence of background autofluorescence signal from the gut (which can be easily confounded with GRASP puncta depending on the worm’s orientation), WormPsyQi quantified GRASP puncta much more robustly with the cytoplasmic label; without the cytoplasmic marker, gut puncta are incorrectly segmented as synapses (highlighted with red arrows) while some dim synaptic puncta are not picked up (highlighted with yellow arrows).

To demonstrate the second scenario, we now highlight the case of ASK CLA-1 in Figure 2 - figure supplement 4E. Additionally, we have emphasized in the manuscript that in cases where the S/N ratio is good and the image is restricted to a small ROI, WormPsyQi will perform well even in the absence of a cytoplasmic marker. This is equally important to note as having a specific cytoplasmic marker in the background may not always be feasible and, in fact, if the cytoplasmic marker is discontinuous or dim relative to puncta signal, using a suboptimal neurite mask for synapse segmentation would result in undercounting synapses.

1. On page 12, the author stated "We also note that in several cases, GRASP quantification differed from EM scoring". However, the EM scoring is primarily based on a single sample, making it challenging to conduct a statistical analysis for the purpose of comparison.

This is correct and is indeed a limitation of EM for this type of analysis. We have now reworded this sentence (page 14) to emphasize the reviewer’s point, and it is also elaborated further in the limitations section.

1. In Figure 6F, the discrepancy between WormPsyQi and human quantification in the analysis of RAB-3 is observed. The author stated that "the RAB-3 signal was too diffuse to resolve all puncta". To better illustrate this discrepancy, it would be beneficial to include images highlighting the puncta that WormPsyQi cannot score, providing direct evidence that diffusing signals are not able to automatically detectable.

To highlight puncta that were not segmented by WormPsyQi but were successfully scored manually, we have included arrows in Figure 6. In addition, for reporter M4p::GFP::RAB-3, we have included magnified insets in Figure 6 - figure supplement 1A to highlight the region where human annotator scores more puncta than WormPsyQi owing to the high synapse density. In future implementations, additional functionality can be built for separating these merged puncta into instances based on geometrical features such as shape and intensity contour.

1. In Figure 9 S1D, the results from WormPsyQi and the manual are totally different. To address this notable discrepancy, the authors should highlight and illustrate the areas of discrepancy in the images. This visual representation can assist future users in identifying signal types that may not be well-suited for WormPsyQi analysis and inspire the development of new strategies to tackle such challenges.

This is now addressed in additional figure panels in Figure 4 – figure supplement 1B and Figure 6 - figure supplement 1A.

Reviewer #3 (Recommendations For The Authors):

I found the comparison between manual quantification and WormPsyQi-based quantification to be very informative. In my opinion, quantifying the number of puncta is not the most tedious/difficult quantification even when done manually. Would the authors be able to include manual-WormPsyQi comparison for more time-consuming and potentially more prone to human error/bias quantifications such as puncta size or distribution patterns using a few markers with some inter/intra animal variabilities?

To address this point, we have now included an additional figure supplement to Figure 2 (Figure 2 – figure supplement 4). We focused on the ASK GFP::CLA-1 reporter and had two human annotators manually label the masks of puncta for each worm by scanning Z-stacks and drawing all pixels belonging to each puncta in Fiji, which were then processed by WormPsyQi’s quantification pipeline to score puncta number, volume, and distribution. We also included a comparison of overall image processing time for each annotator and WormPsyQi. For features analyzed, the difference between WormPsyQi and human annotators for ASK CLA-1 is not statistically significant for multiple puncta features. Importantly, WormPsyQi reduces overall processing time by at least an order of magnitude, and while this is already advantageous for counting puncta, it is especially useful for other important puncta features since a) they may not be easily discernible, and b) it is extremely laborious to quantify them manually in large datasets when pixel-wise labels are required.

The authors listed minimum human errors and biases as one of the benefits of WormPsyQi. For the markers with discrepancies in quantifications between human and WormPsyQi, have the authors encountered or considered human errors/biases as potential reasons for such discrepancies?

This is the same point brought up by reviewer 1. We added Figure 2- figure supplement 3 to compare WormPsyQi to different human labelers, and show that because human labels can introduce systematic bias, WormPsyQi reduces such bias by scoring images using the same metric.

The authors noted that WormPsyQi would be useful for comparing different genotypes/environments. Some mutants have known changes in synapse patterning/number. It would be helpful if the authors could validate WormPsyQi using some of the mutants with known synapse defects. For instance, zig-10 mutant increases the cholinergic synapse density just by a bit (Cherra and Jin, Neuron 2016), and nlr-1 mutant disrupts punctated localization of UNC-9 gap junction in the nerve ring (Meng and Yan, Neuron 2020), which could only be detectable by experts' eyes. It would be interesting to see if WormPsyQi picks up such subtle phenotypes.

We agree that our pipeline would need to be tested in multiple paradigms to test its performance on detecting additional subtle phenotypes. In the context of this paper, we note that the developmental analysis of puncta in Figure 8 was performed to validate the ground truth from previous EM-based analyses (Witvliet et al., 2021), albeit the latter was limited by sample size. We extended this developmental analysis to the pharyngeal reporters, and in some cases the difference across timepoints was marginal (as emphasized by additional Figure 9 - figure supplement 2), but still detected by WormPsyQi. Lastly, our synapse localization analysis in Figure 10 assigns the probability of finding a synapse at a particular location along a neurite, which is not easily discernible by manual scoring.

One of the benefits of the automated data analysis program is to be able to notice the differences you do not expect. For example, there are situations where you feel that in certain genotypes there is something different from wild type with their synapses but you can't tell what's different from wild type. In such cases, you may not know what to quantify. I think it would be beneficial if there were more parameters to be included in the default qualifications such as puncta number/size/intensity/distributions in the pipeline, so that the users may find unexpected phenotypes from one of the default quantifications.

We apologize if this was not clearer in the manuscript where we first describe the pipeline in detail. To clarify, the output of WormPsyQi is a CSV file which includes several quantitative features, such as mean/max/min fluorescence intensity, puncta volume, and position. While most of our analyses are focused on puncta count, the user can perform downstream statistical analyses on all additional features scored to infer which features are most significantly variable across conditions. To make this clearer, we have elaborated the text when we first describe our pipeline, and along with the new Figure 2 - figure supplement 4, we hope that this point is clearer now.

In addition, most proof-of-principle analysis we performed was focused on an ROI where we expect the synapses to localize. In practice, the user can input images and perform quantification across the entire image without biasing toward an ROI (this can be done in the GUI synapse corrector window) to also evaluate synaptic changes in regions outside the usual ROI.

The authors stated that WormPsyQi could mitigate the problems stemming from scoring images with low signal-to-noise ratio or in regions with high background autofluorescence, laboriousness of scoring large datasets, and inter-dataset variability. Other than the 'laboriousness of scoring large datasets' it appeared to me that WormPsyQi does not do better than manual quantifications, especially inter-dataset variability, as the authors noted variability among the transgenes as one of the limitations of the toolkits. If two datasets are taken with completely different setups such as two independent arrays taken with two distinct confocal microscopes, would WormPsyQi make these two datasets comparable?

We have included additional figure supplements to address the reviewer’s point. A significant advantage WormPsyQi offers over manual scoring is that it provides a standardized method of quantifying synapse features. As shown in Figure 2 – figure supplement 3, human labelers can introduce systematic bias (e.g. some over count puncta, while some undercount). In addition, while puncta number may be relatively easy to quantify, especially in a high-quality dataset, more subtle puncta features such as size, intensity, and distribution are much more laborious to quantify and require a priori knowledge of signal localization (Figure 2 – figure supplement 4, Figure 10). Altogether, our pipeline facilitates multiple measurements while also enabling robust quantification in hard-to-score cases such as the example shown for PHB>AVA reporter (Figure 4 - figure supplement 1B).

Minor comments:

Limitations are not quite specific to this work but those are general limitations to the concatemeric trans genes and fluorescently labeled synaptic proteins. I'd appreciate discussing specific limitations to WormPsyQi related to image acquisitions. For instance, for neurons with 3D structures would WormPsyQi be able to handle z-stacks closer to coverslip and stacks that are deeper side in a similar manner? Would the users need to be aware of such limitations when comparing different genotypes?

To address the reviewer’s comment, we have elaborated the last paragraph in the limitations section to explicitly discuss where the user should exercise caution. The reviewer reasonably points out that the fluorescent signal away from the cover slip is typically dimmer, and neurite masking in this case is indeed compromised if dim to start with. In such cases, we recommend that the user either performs some preprocessing such as deconvolution, denoising, or contrast enhancement to boost the neurite signal, or segment synapses without the neurite mask if the puncta signal is brighter than that of the cytoplasmic marker. We hope that our additional figure supplements will clarify that WormPsyQi’s performance is contingent on reporter type and image quality, thus making it easier for the user to discern where automated quantification falls short and alternative reporters should be explored. In general, if puncta are not discernible to the user due to very poor S/N ratio, for instance, we do not recommend using WormPsyQi to process such datasets; this will be manifest in the results of the new “test all models” feature we added in the revised version.

Some Rab-3 fusion proteins are described as RAB-3::GFP(BFP). Do these represent the C-terminal fusion of the fluorescent proteins? RAB-3 is a small GTPase with a lipid modification site at its C-terminus essential for its localization and function. Is it possible that the diffuse signal of some RAB-3 markers is caused by c-terminal fusion of the fluorescent protein?

While we do have reporters with N- and C-terminal RAB-3 fusions for different neurons, we do not have both for the same neuron to perform a fair comparison. However, as noted in response to a previous comment by reviewer 2, RAB-3 and CLA-1 have distinct localization patterns at the synapse and this aligns with their distinct functions: while RAB-3 localizes at synaptic vesicles, CLA-1 is an active zone protein required for synaptic vesicle clustering. Accordingly, we have observed diffuse RAB-3 signal in reporters irrespective of where the protein is tagged, and while this is not problematic for ROIs with a low synapse density, it confounds quantification in synapse-dense regions. In contrast, CLA-1 puncta are typically easier to quantify more discretely, which is particularly relevant for features such synapse distribution, size, and intensity.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

In this very strong and interesting paper the authors present a convincing series of experiments that reveal molecular mechanism of neuronal cell type diversification in the nervous system of Drosophila. The authors show that a homeodomain transcription factor, Bsh, fulfills several critical functions - repressing an alternative fate and inducing downstream homeodomain transcription factors with whom Bsh may collaborate to induce L4 and L5 fates (the author's accompanying paper reveals how Bsh can induce two distinct fates). The authors make elegant use of powerful genetic tools and an arsenal of satisfying cell identity markers.

Thanks!

I believe that this is an important study because it provides some fundamental insights into the conservation of neuronal diversification programs. It is very satisfying to see that similar organizational principles apply in different organisms to generate cell type diversity. The authors should also be commended for contextualizing their work very well, giving a broad, scholarly background to the problem of neuronal cell type diversification.

Thanks!

My one suggestion for the authors is to perhaps address in the Discussion (or experimentally address if they wish) how they reconcile that Bsh is on the one hand: (a) continuously expressed in L4/L4, (b) binding directly to a cohort of terminal effectors that are also continuously expressed but then, on the other hand, is not required for their maintaining L4 fate? A few questions: Is Bsh only NOT required for maintaining Ap expression or is it also NOT required for maintaining other terminal markers of L4? The former could be easily explained - Bsh simply kicks of Ap, Ap then autoregulates, but Bsh and Ap then continuously activate terminal effector genes. The second scenario would require a little more complex mechanism: Bsh binding of targets (with Notch) may open chromatin, but then once that's done, Bsh is no longer needed and Ap alone can continue to express genes. I feel that the authors should be at least discussing this. The postmitotic Bsh removal experiment in which they only checked Ap and depression of other markers is a little unsatisfying without further discussion (or experiments, such as testing terminal L4 markers). I hasten to add that this comment does not take away from my overall appreciation for the depth and quality of the data and the importance of their conclusions.

Great suggestions, we will discuss these two hypotheses as requested.

Bsh initiates Ap expression in L4 neurons which then maintain Ap expression independently of Bsh expression, likely through Ap autoregulation. During the synaptogenesis window, Ap expression becomes independent from Bsh expression, but Bsh and Ap are both still required to activate the synapse recognition molecule DIP-beta. Additionally, Bsh also shows putative binding to other L4 identity genes, e.g., those required for neurotransmitter choice, and electrophysiological properties, suggesting Bsh may initiate L4 identity genes as a suite of genes. The mechanism of maintaining identity features (e.g., morphology, synaptic connectivity, and functional properties) in the adult remains poorly understood. It is a great question whether primary HDTF Bsh maintains the expression of L4 identity genes in the adult. To test this, in our next project, we will specifically knock out Bsh in L4 neurons of the adult fly and examine the effect on L4 morphology, connectivity, and function properties.

Reviewer #2 (Public Review):

Summary:

In this paper, the authors explore the role of the Homeodomain Transcription Factor Bsh in the specification of Lamina neuronal types in the optic lobe of Drosophila. Using the framework of terminal selector genes and compelling data, they investigate whether the same factor that establishes early cell identity is responsible for the acquisition of terminal features of the neuron (i.e., cell connectivity and synaptogenesis).

Thanks for the positive words!

The authors convincingly describe the sequential expression and activity of Bsh, termed here as 'primary HDTF', and of Ap in L4 or Pdm3 in L5 as 'secondary HDTFs' during the specification of these two neurons. The study demonstrates the requirement of Bsh to activate either Ap and Pdm3, and therefore to generate the L4 and L5 fates. Moreover, the authors show that in the absence of Bsh, L4 and L5 fates are transformed into a L1 or L3-like fates.

Thanks!

Finally, the authors used DamID and Bsh:DamID to profile the open chromatin signature and the Bsh binding sites in L4 neurons at the synaptogenesis stage. This allows the identification of putative Bsh target genes in L4, many of which were also found to be upregulated in L4 in a previous single-cell transcriptomic analysis. Among these genes, the paper focuses on Dip-β, a known regulator of L4 connectivity. They demonstrate that both Bsh and Ap are required for Dip-β, forming a feed-forward loop. Indeed, the loss of Bsh causes abnormal L4 synaptogenesis and therefore defects in several visual behaviors. The authors also propose the intriguing hypothesis that the expression of Bsh expanded the diversity of Lamina neurons from a 3 cell-type state to the current 5 cell-type state in the optic lobe.

Thanks for the excellent summary of our findings!

Strengths:

Overall, this work presents a beautiful practical example of the framework of terminal selectors: Bsh acts hierarchically with Ap or Pdm3 to establish the L4 or L5 cell fates and, at least in L4, participates in the expression of terminal features of the neuron (i.e., synaptogenesis through Dip-β regulation).

Thanks!

The hierarchical interactions among Bsh and the activation of Ap and Pdm3 expression in L4 and L5, respectively, are well established experimentally. Using different genetic drivers, the authors show a window of competence during L4 neuron specification during which Bsh activates Ap expression. Later, as the neuron matures, Ap becomes independent of Bsh. This allows the authors to propose a coherent and well-supported model in which Bsh acts as a 'primary' selector that activates the expression of L4specific (Ap) and L5-specific (Pdm3) 'secondary' selector genes, that together establish neuronal fate.

Thanks again!

Importantly, the authors describe a striking cell fate change when Bsh is knocked down from L4/L5 progenitor cells. In such cases, L1 and L3 neurons are generated at the expense of L4 and L5. The paper demonstrates that Bsh in L4/L5 represses Zfh1, which in turn acts as the primary selector for L1/L3 fates. These results point to a model where the acquisition of Bsh during evolution might have provided the grounds for the generation of new cell types, L4 and L5, expanding lamina neuronal diversity for a more refined visual behaviors in flies. This is an intriguing and novel hypothesis that should be tested from an evo-devo standpoint, for instance by identifying a species when L4 and L5 do not exist and/or Bsh is not expressed in L neurons.

Thanks for the appreciation of our findings!

To gain insight into how Bsh regulates neuronal fate and terminal features, the authors have profiled the open chromatin landscape and Bsh binding sites in L4 neurons at mid-pupation using the DamID technique. The paper describes a number of genes that have Bsh binding peaks in their regulatory regions and that are differentially expressed in L4 neurons, based on available scRNAseq data. Although the manuscript does not explore this candidate list in depth, many of these genes belong to classes that might explain terminal features of L4 neurons, such as neurotransmitter identity, neuropeptides or cytoskeletal regulators. Interestingly, one of these upregulated genes with a Bsh peak is Dip-β, an immunoglobulin superfamily protein that has been described by previous work from the author's lab to be relevant to establish L4 proper connectivity. This work proves that Bsh and Ap work in a feed-forward loop to regulate Dip-β expression, and therefore to establish normal L4 synapses. Furthermore, Bsh loss of function in L4 causes impairs visual behaviors.<br /> Thanks for the excellent summary of our findings.

Weaknesses:

● The last paragraph of the introduction is written using rhetorical questions and does not read well. I suggest rewriting it in a more conventional direct style to improve readability.

We agree and have updated the text as suggested.

● A significant concern is the way in which information is conveyed in the Figures. Throughout the paper, understanding of the experimental results is hindered by the lack of information in the Figure headers. Specifically, the genetic driver used for each panel should be adequately noted, together with the age of the brain and the experimental condition. For example, R27G05-Gal4 drives early expression in LPCs and L4/L5, while the 31C06-AD, 34G07-DBD Split-Gal4 combination drives expression in older L4 neurons, and the use of one or the other to drive Bsh-KD has dramatic differences in Ap expression. The indication of the driver used in each panel will facilitate the reader's grasp of the experimental results.

We agree and have updated the figure annotation.

● Bsh role in L4/L5 cell fate: o It is not clear whether Tll+/Bsh+ LPCs are the precursors of L4/L5. Morphologically, these cells sit very close to L5, but are much more distant from L4.

Our current data show L4 and L5 neurons are generated by different LPCs. However, currently, we don’t have tools to demonstrate which subset of LPCs generate which lamina neuron type. We are currently working on a follow-up manuscript on LPC heterogeneity, but those experiments have just barely been started.

● Somatic CRISPR knockout of Bsh seems to have a weaker phenotype than the knockdown using RNAi. However, in several experiments down the line, the authors use CRISPR-KO rather than RNAi to knock down Bsh activity: it should be explained why the authors made this decision. Alternatively, a null mutant could be used to consolidate the loss of function phenotype, although this is not strictly necessary given that the RNAi is highly efficient and almost completely abolishes Bsh protein.

The reason we chose CRISPR-KO (L4-specific Gal4, uas-Cas9, and uas-Bsh-sgRNAs) is that it effectively removed Bsh expression from the majority of L4 neurons. However, it failed to knock down Bsh in L4 neurons using L4-split Gal4 and Bsh-RNAi because L4-split Gal4 expression depends on Bsh. We have updated this explanation in the text.

● Line 102: Rephrase "R27G05-Gal4 is expressed in all LPCs and turned off in lamina neurons" to "is turned off as lamina neurons mature", as it is kept on for a significant amount of time after the neurons have already been specified.

Thanks; we have made that change.

● Line 121: "(a) that all known lamina neuron markers become independent of Bsh regulation in neurons" is not an accurate statement, as the markers tested were not shown to be dependent on Bsh in the first place.

Good point. We have rephrased it as “that all known lamina neuron markers are independent of Bsh regulation in neurons”.

● Lines 129-134: Make explicit that the LPC-Gal4 was used in this experiment. This is especially important here, as these results are opposite to the Bsh Loss of Function in L4 neurons described in the previous section. This will help clarify the window of competence in which Bsh establishes L4/L5 neuronal identities through ap/pdm3 expression.

Thanks! We have updated Gal4 information in the text for every manipulation.

● DamID and Bsh binding profile:

● Figure 5 - figure supplement 1C-E: The genotype of the Control in (C) has to be described within the panel. As it is, it can be confused with a wild type brain, when it is in fact a Bsh-KO mutant.

Great point! Thank you for catching this and we have updated it.

● It Is not clear how L4-specific Differentially Expressed Genes were found. Are these genes DEG between Lamina neurons types, or are they upregulated genes with respect to all neuronal clusters? If the latter is the case, it could explain the discrepancy between scRNAseq DEGs and Bsh peaks in L4 neurons.

We did not use “L4-specific Differentially Expressed Genes”. Instead, we used all genes that are significantly transcribed in L4 neurons (line 209-213).

● Dip-β regulation:

● Line 234: It is not clear why CRISPR KO is used in this case, when Bsh-RNAi presents a stronger phenotype.

As we explained above, the reason we chose CRISPR-KO (L4-specific Gal4, uas-Cas9, and uas-BshsgRNAs) is that it effectively removed Bsh expression from the majority of L4 neurons. However, it failed to knock down Bsh in L4 neurons using L4-split Gal4 and Bsh-RNAi because L4-split Gal4 expression depends on Bsh. We have updated this explanation in the text.

● Figure 6N-R shows results using LPC-Gal4. It is not clear why this driver was used, as it makes a less accurate comparison with the other panels in the figure, which use L4-Split-Gal4. This discrepancy should be acknowledged and explained, or the experiment repeated with L4-Split-Gal4>Ap-RNAi.

I think you mean 6J-M shows results using LPC-Gal4. We first tried L4-Split-Gal4>Ap-RNAi but it failed to knock down Ap because L4-Split-Gal4 expression depends on Ap. We have added this to the text.

● Line 271: It is also possible that L4 activity is dispensable for motion detection and only L5 is required.

Thanks! Work from Tuthill et al, 2013 showed that L5 is not required for any motion detection. We have included this citation in the text.

● Discussion: It is necessary to de-emphasize the relevance of HDTFs, or at least acknowledge that other, non-homeodomain TFs, can act as selector genes to determine neuronal identity. By restricting the discussion to HDTFs, it is not mentioned that other classes of TFs could follow the same PrimarySecondary selector activation logic.

That is a great point, thank you! We have included this in the discussion.

Author Response:

We thank all reviewers for their comments and effort to improve our paper. We appreciate that the writing can be clarified overall, and some sections need more elaboration. We will provide these in the next revision within the coming months. Particularly, we will focus on some common themes identified by all reviewers:

1. We will clarify that the coarse-grained brain surfaces are an output of our algorithm alone and not to be directly/naively likened to actual brain surfaces, e.g. in terms of the location or shape of the folds. Our analysis purely focuses on the likeliness in terms of whole-brain morphometrics between actual brains and coarse-grained brains. Specifically on the point of “thickening” of the brain: this is anatomically well-founded, as less folded brains have a “thicker” cortex than more folded brains, when they are all normalised to the same size. This is fundamentally why the universal scaling law also applies to these coarse-grained brains. We will provide more detail to highlight this.

2. We will clarify the motivation behind our coarse-graining procedure better: mathematically, this is directly inspired by box-counting algorithms in fractal geometry; but this algorithm also has elegant parallels with other algorithms which we will highlight.

3. The age effects are demonstrated here in a small sample as a proof-of-principle, but we will update our latest results using ~100 subjects from the CamCAN data demonstrating the same effect. We have additionally described and verified these age effects in more detail in a separate preprint (https://arxiv.org/abs/2311.13501) with ~1500 subjects, and additionally showed that scale-dependent metrics substantially improve understanding and applications such as brain age prediction.

4. We have independently also received the feedback that we need to clarify how our method interacts with different resolution of the original MRI. We will add this as a new set of results, demonstrating that the MRI acquisition resolution (within a reasonable range) has a very small effect, as our method takes the reconstructed surfaces as a starting point.

5. We agree that it may be confusing to emphasise a constant K in the first set of results across species, and then later highlight a changing K in the human ageing results. We will clarify that in the first set of results, we find a “constant” K relative to a changing S: The range in K across melted primate brains is approx 0.1, whereas in S it is over 1.2. In other words, S changes are an order of magnitude higher than K changes. Hence, we described K as “constant” relative to S. Nevertheless, K shows subtle changes within individuals, which is what we are describing in the human ageing results. These changes are within the range of K values described in the across species results.

6. Finally, we will also make sure to summarise our specific contributions beyond existing work:

   (i) Showing for the first time that representative primate species follow the exact same fractal scaling – as opposed to previous work showing that they have a similar fractal dimension, i.e. slope, but not necessarily the same offset, as previous methods had no consistent way of comparing offsets.

   (ii) Previous work could also not show direct agreement in morphometrics between the coarse-grained brains of primate species and other non-primate mammalian species.

   (iii) Demonstrating in proof-of-principle that multiscale morphometrics, in practice, can have much larger effect sizes for classification applications. This moves beyond our previous work where we only showed the scaling law across and within species, but all on one (native) scale with comparable effect sizes for classification applications.

Author Response

Reviewer #2 (Public Review):

Weaknesses:

The paper contains multiple instances of non-scientific language, as indicated below. It would also benefit from additional details on the cryo-EM structure determination in the Methods and inclusion of commonly accepted requirements for cryo-EM structures, like examples of 2D class averages, raw micrographs, and FSC curves (between half-maps as well as between rigid-body fitted (or refined) atomic models of the different polymorphs and their corresponding maps). In addition, cryo-EM maps for the control experiments F1 and F2 should be presented in Figure 9.

We will include the suggested data on the Cryo-EM analyses in a revised version of the preprint. We did not collect data on the sample used for the seeds in the cross seeding experiments because we had already confirmed in multiple datasets that the conditions in F1 and F2 reproducibly produce fibrils of Type 1 and Type 3, respectively. In a revised version we will include the analyses of several more datasets at the F1 and F2 conditions to support this statement.

Reviewer #3 (Public Review):

Weaknesses:

1. The authors reveal that both Type 1 monofilament fibril polymorph (reminiscent of JOS-like polymorph) and Type 5 polymorph (akin to tissue-amplified-like polymorph) can both form under the same condition. Additionally, this condition also fosters the formation of flat ribbon-like fibril across different batches. Notably, at pH 5.8, variations in experimental groups yield disparate abundance ratios between polymorph 3B and 3C, indicating a degree of instability in fibrillar formation. The variability would potentially pose challenges for replicability in subsequent research. In light of these situations, I propose the following recommendations:

(1) An explicit elucidation of the factors contributing to these divergent outcomes under similar experimental conditions is warranted. This should include an exploration of whether variations in purified protein batches are contributing factors to the observed heterogeneity.

We are in complete agreement that understanding the factors that lead to polymorph variability is of utmost importance (and was the impetus for the manuscript itself). However the number of variables to explore is overwhelming and we will continue to investigate this in our future research. Regarding the variability between batches of purified protein, we also think that this could be a factor in the polymorph variability observed for otherwise “identical” aggregation conditions, particularly at pH 7 where the largest variety of polymorphs have been observed. While our data still indicates that Type 1,2 and 3 polymorphs are strongly selected by pH, the selection between interface variants 3B vs. 3C and 2A vs. 2B might also be affected by protein purity. Our standard purification protocol produces a single band by coomassie-stained SDS-PAGE however minor truncations and other impurities below a few percent would go undetected and, given the proposed roles of the N and C-termini in secondary nucleation, could have a large effect on polymorph selection and seeding. In line with the reviewer’s comments we now include a batch number for each EM dataset. While no new conclusions can be drawn from the inclusion of this additional data, we feel that it is important to acknowledge the possible role of batch to batch variability.

(2) To enhance the robustness of the conclusions, additional replicates of the experiments under the same condition should be conducted, ideally a minimum of three times.

The pH 5.8 conditions that yield Type 3 fibrils has already been repeated several times in the original manuscript. The pH 7.4 conditions were only mentioned twice, once as an unseeded and once as a cross-seeded fibrilization. We solved a second Type 1 structure from a second dataset from the same protein batch fibrillized under similar conditions at pH 7.4 but with the addition of inositol trisphosphate in the hopes that we could replicate one of the in vivo polymorphs. However only the Type 1 polymorphs were observed and so we will add this data point to the revised manuscript. We are currently screening more fibrils produced at pH 7.0 and will include any replicates of Type 5 or the Type 1M polymorphs or of new structures that are obtained at these conditions… however, as noted in the original manuscript, reproducibility at this pH might be difficult because there appears to be a wider range of accessible polymorphs. As will be mentioned in the revised version, the Type 5 structure was solved from a manually picked set of fibers that represented 10-20% of the observed fibrils. The remaining fibers in the sample comprised polymorphs that could not be analyzed due to their inhomogeneity or lack of twist.

(3) Further investigation into whether different polymorphs formed under the same buffer condition could lead to distinct toxicological and pathology effects would be a valuable addition to the study.

The correlation of toxicity with structure would in principle be interesting. However the Type 1 and Type 3 polymorphs formed at pH 5.8 and 7.4 are not likely to be biologically relevant. The pH 7 polymorphs (Type 5 and 1M) would be more interesting because they form under the same conditions and might be related to some disease relevant structures. Still, it is rare that a single polymorph appears at 7.0 (the Type 5 represented only 10-20% of the fibrils in the sample and the Type 1M also had unidentified double-filament fibrils in the sample). We plan to pursue this line of research and hope to include it in a future publication.

1. The cross-seeding study presented in the manuscript demonstrates the pivotal role of pH conditions in dictating conformation. However, an intriguing aspect that emerges is the potential role of seed concentration in determining the resultant product structure. This raises a critical question: at what specific seed concentration does the determining factor for polymorph selection shift from pH condition to seed concentration? A methodological robust approach to address this should be conducted through a series of experiments across a range of seed concentrations. Such an approach could delineate a clear boundary at which seed concentration begins to predominantly dictate the conformation, as opposed to pH conditions. Incorporating this aspect into the study would not only clarify the interplay between seed concentration and pH conditions, but also add a fascinating dimension to the understanding of polymorph selection mechanisms.

A more complete analysis of the mechanisms of aggregation, including the effect of seed concentration and the resulting polymorph specificity of the process, are all very important for our understanding of the aggregation pathways of alpha-synuclein and are currently the topic of ongoing investigations in our lab.

Furthermore, the study prompts additional queries regarding the behavior of cross-seeding production under the same pH conditions when employing seeds of distinct conformation. Evidence from various studies, such as those involving E46K and G51D cross-seeding, suggests that seed structure plays a crucial role in dictating polymorph selection. A key question is whether these products consistently mirror the structure of their respective seeds.

We thank the reviewer for reminding us to include a reference to these studies as a clear example of polymorph selection by cross-seeding which we will do in the revised version. Unfortunately, it is not 100% clear from the G51D cross seeding manuscript (https://doi.org/10.1038/s41467-021-26433-2) what conditions were used in the cross-seeding since different conditions were used for the seedless wild-type and mutant aggregations… however it appears that the wild-type without seeds was Tris pH 7.5 (although at 37C the pH could have dropped to 7-ish) and the cross-seeded wild-type was in Phosphate buffer at pH 7.0. In the E46K cross-seeding manuscript, it appears that pH 7.5 Tris was used for all fibrilizations (https://doi.org/10.1073/pnas.2012435118). In any event, both results point to the fact that at pH 7.0-7.5 under low-seed conditions (0.5%) the Type 4 polymorph can propagate in a seed specific manner.

1. In the Results section of "The buffer environment can dictate polymorph during seeded nucleation", the authors reference previous cell biological and biochemical assays to support the polymorph-specific seeding of MSA and PD patients under the same buffer conditions. This discussion is juxtaposed with recent research that compares the in vivo biological activities of hPFF, ampLB as well as LB, particularly in terms of seeding activity and pathology. Notably, this research suggests that ampLB, rather than hPFF, can accurately model the key aspects of Lewy Body Diseases (LBD) (refer to: https://doi.org/10.1038/s41467-023-42705-5). The critical issue here is the need to reconcile the phenomena observed in vitro with those in in-vivo or in-cell models. Given the low seed concentration reported in these studies, it is imperative for the authors to provide a more detailed explanation as to why the possible similar conformation could lead to divergent pathologies, including differences in cell-type preference and seeding capability.

We thank the reviewer for bring this recent report to our attention. The findings that ampLB and hPFF have different PK digestion patterns and that only the former is able to model key aspects of Lewy Body disease are in support of the seed-specific nature of some types of alpha-synuclein aggregation. We will add more discussion regarding the significant role that seed type and seed conditions likely play in polymorph selection.

1. In the Method section of "Image processing", the authors describe the helical reconstruction procedure, without mentioning much detail about the 3D reconstruction and refinement process. For the benefit of reproducibility and to facilitate a deeper understanding among readers, the authors should enrich this part to include more comprehensive information, akin to the level of detail found in similar studies (refer to: https://doi.org/10.1038/nature23002).

As suggested by reviewer #2, we will add more comprehensive information on the 3D reconstruction and refinement process to a revised version.

1. The abbreviation of amino acids should be unified. In the Results section "On the structural heterogeneity of Type 1 polymorphs", the amino acids are denoted using three-letter abbreviation. Conversely, in the same section under "On the structural heterogeneity of Type 2 and 3 structures", amino acids are abbreviated using the one-letter format. For clarity and consistency, it is essential that a standardized format for amino acid abbreviations be adopted throughout the manuscript.

That makes perfect sense and will be corrected in a revised version.

Reviewing Editor:

After discussion among the reviewers, it was decided that point 2 in Reviewer #3's Public Review (about the experiments with different concentrations of seeds) would probably lie outside the scope of a reasonable revision for this work.

We agree as stated above and will continue to work on this important point.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Strengths

This paper is well situated theoretically within the habit learning/OCD literature.

Daily training in a motor-learning task, delivered via smartphone, was innovative, ecologically valid and more likely to assay habitual behaviors specifically. Daily training is also more similar to studies with non-humans, making a better link with that literature. The use of a sequential-learning task (cf. tasks that require a single response) is also more ecologically valid.

The in-laboratory tests (after the 1 month of training) allowed the researchers to test if the OCD group preferred familiar, but more difficult, sequences over newer, simpler sequences.

The authors achieved their aims in that two groups of participants (patients with OCD and controls) engaged with the task over the course of 30 days. The repeated nature of the task meant that 'overtraining' was almost certainly established, and automaticity was demonstrated. This allowed the authors to test their hypotheses about habit learning. The results are supportive of the authors' conclusions.

Response: We truly appreciate the positive assessment of referee 1, particularly the consideration that our study is theoretically strong and that ‘the results are supportive of the authors' conclusions’. This is an important external endorsement of our conclusions, contrasting somewhat with the views of referee 2.

Weaknesses

The sample size was relatively small. Some potentially interesting individual differences within the OCD group could have been examined more thoroughly with a bigger sample (e.g., preference for familiar sequences). A larger sample may have allowed the statistical testing of any effects due to medication status. The authors were not able to test one criterion of habits, namely resistance to devaluation, due to the nature of the task

Response: We agree with the reviewer that the proof of principle established in our study opens new avenues for research into the psychological and behavioral determinants of the heterogeneity of this clinical population. However, considering the study timeline and the pandemic constraints, a bigger sample was not possible. Our sample can indeed be considered small if one compares it with current online studies, which do not require in-person/laboratory testing, thus being much easier to recruit and conduct. However, given the nature of our protocol (with 2 demanding test phases, 1-month engagement per participant and the inclusion of OCD patients without comorbidities only) and the fact that this study also involved laboratory testing, we consider our sample size reasonable and comparable to other laboratory studies (typically comprising on average between 30-50 participants in each group).

This article is likely to be impactful -- the delivery of a task across 30 days to a patient group is innovative and represents a new approach for the study of habit learning that is superior to an inlaboratory approach.

An interesting aspect of this manuscript is that it prompts a comparison with previous studies of goal-directed/habitual responding in OCD that used devaluation protocols, and which may have had their effects due to deficits in goal-directed behavior and not enhanced habit learning per se.

Response: Thank you for acknowledging the impact of our study, in particular the unique ability of our task to interrogate the habit system.

Reviewer #2 (Public Review):

In this study, the researchers employed a recently developed smartphone application to provide 30 days of training on action sequences to both OCD patients and healthy volunteers. The study tested learning and automaticity-related measures and investigated the effects of several factors on these measures. Upon training completion, the researchers conducted two preference tests comparing a learned and unlearned action sequences under different conditions. While the study provides some interesting findings, I have a few substantial concerns:

1. Throughout the entire paper, the authors' interpretations and claims revolve around the domain of habits and goal-directed behavior, despite the methods and evidence clearly focusing on motor sequence learning/procedural learning/skill learning. There is no evidence to support this framing and interpretation and thus I find them overreaching and hyperbolic, and I think they should be avoided. Although skills and habits share many characteristics, they are meaningfully distinguishable and should not be conflated or mixed up. Furthermore, if anything, the evidence in this study suggests that participants attained procedural learning, but these actions did not become habitual, as they remained deliberate actions that were not chosen to be performed when they were not in line with participants' current goals.

Response: We acknowledge that the research on habit learning is a topic of current controversy, especially when it comes to how to induce and measure habits in humans. Therefore, within this context referee’s 2 criticism could be expected. Across distinct fields of research, different methodologies have been used to measure habits, which represent relatively stereotyped and autonomous behavioral sequences enacted in response to a specific stimulus without consideration, at the time of initiation of the sequence, of the value of the outcome or any representation of the relationship that exists between the response and the outcome. Hence these are stimulus-bound responses which may or may not require the implementation of a skill during subsequent performance. Behavioral neuroscientists define habits similarly, as stimulus-response associations which are independent of reward or outcome, and use devaluation or contingency degradation strategies to probe habits (Dickinson and Weiskrantz, 1985; Tricomi et al., 2009). Others conceptualize habits as a form of procedural memory, along with skills, and use motor sequence learning paradigms to investigate and dissect different components of habit learning such as action selection, execution and consolidation (Abrahamse et al., 2013; Doyon et al., 2003; Squire et al., 1993). It is also generally agreed that the autonomous nature of habits and the fluid proficiency of skills are both usually achieved with many hours of training or practice, respectively (Haith and Krakauer, 2018).

We consider that Balleine and Dezfouli (2019) made an excellent attempt to bring all these different criteria within a single framework, which we have followed. We also consider that our discussion in fact followed a rather cautious approach to interpretation solely in terms of goaldirected versus habitual control.

Referee 2 does not actually specify criteria by which they define habits and skills, except for asserting that skilled behavior is goal-directed, without mentioning what the actual goal of the implantation of such skill is in the present study: the fulfillment of a habit? We assume that their definition of habit hinges on the effects of devaluation, as a single criterion of habit, but which according to Balleine and Dezfouli (2019) is only 1 of their 4 listed criteria. We carefully addressed this specific criterion in our manuscript: “We were not, however, able to test the fourth criterion, of resistance to devaluation. Therefore, we are unable to firmly conclude that the action sequences are habits rather than, for example, goal-directed skills. Regardless of whether the trained action sequences can be defined as habits or goal-directed motor skills, it has to be considered…”. Therefore, we took due care in our conclusions concerning habits and thus found the referee’s comment misleading and unfair.

We note that our trained motor sequences did in fact fulfil the other 3 criteria listed by Balleine and Dezfouli (2019), unlike many studies employing only devaluation (e.g. Tricomi et al 2009; Gillan et al 2011). Moreover, we cited a recent study using very similar methodology where the devaluation test was applied and shown to support the habit hypothesis (Gera et al., 2022).

Whether the initiation of the trained motor sequences in experiment 3 (arbitration) is underpinned by an action-outcome association (or not) has no bearing on whether those sequences were under stimulus-response control after training (experiment 1). Transitions between habitual and goal-directed control over behavior are quite well established in the experimental literature, especially when choice opportunities become available (Bouton et al (2021), Frölich et al (2023), or a new goal-directed schemata is recruited to fulfill a habit (Fouyssac et al, 2022). This switching between habits and goal-directed responding may reflect the coordination of these systems in producing effective behavior in the real world.

• Fouyssac M, Peña-Oliver Y, Puaud M, Lim NTY, Giuliano C, Everitt BJ, Belin D. (2021).Negative Urgency Exacerbates Relapse to Cocaine Seeking After Abstinence. Biological Psychiatry. doi: 10.1016/j.biopsych.2021.10.009

• Frölich S, Esmeyer M, Endrass T, Smolka MN and Kiebel SJ (2023) Interaction between habits as action sequences and goal-directed behavior under time pressure. Front. Neurosci. 16:996957. doi: 10.3389/fnins.2022.996957

• Bouton ME. 2021. Context, attention, and the switch between habit and goal-direction in behavior. Learn Behav 49:349– 362. doi:10.3758/s13420-021-00488-z

1. Some methodological aspects need more detail and clarification.

2. There are concerns regarding some of the analyses, which require addressing.

Response: We thank referee 2 for their detailed review of the methods and analyses of our study and for the helpful feedback, which clearly helps improve our manuscript. We will clarify the methodological aspects in detail and conduct the suggested analysis. Please see below our answers to the specific points raised.

Introduction:

1. It is stated that "extensive training of sequential actions would more rapidly engage the 'habit system' as compared to single-action instrumental learning". In an attempt to describe the rationale for this statement the authors describe the concept of action chunking, its benefits and relevance to habits but there is no explanation for why sequential actions would engage the habit system more rapidly than a single-action. Clarifying this would be helpful.

Response: We agree that there is no evidence that action sequences become habitual more readily than single actions, although action sequences clearly allow ‘chunking’ and thus likely engage neural networks including the putamen which are implicated in habit learning as well as skill. In our revised manuscript we will instead state: “we have recently postulated that extensive training of sequential actions could be a means for rapidly engaging the ‘habit system’ (Robbins et al., 2019)]”

DONE in page 2

1. In the Hypothesis section the authors state: “we expected that OCD patients... show enhanced habit attainment through a greater preference for performing familiar app sequences when given the choice to select any other, easier sequence”. I find it particularly difficult to interpret preference for familiar sequences as enhanced habit attainment.

Response: We agree that choice of the familiar response sequence should not be a necessary criterion for habitual control although choice for a familiar sequence is, in fact, not inconsistent with this hypothesis. In a recent study, Zmigrod et al (2022) found that 'aversion to novelty' was a relevant factor in the subjective measurement of habitual tendencies. It should also be noted that this preference was present in patients with OCD. If one assumes instead, like the referee, that the familiar sequence is goal-directed, then it contravenes the well-known 'egodystonia' of OCD which suggests that such tendencies are not goal-directed.

To clarify our hypothesis, we will amend the sentence to the following: “Finally, we expected that OCD patients would generally report greater habits, as well as attribute higher intrinsic value to the familiar app sequences manifested by a greater preference for performing them when given the choice to select any other, easier sequence”.

DONE in page 5. We have now rephrased it: “Additionally, we hypothesized that OCD patients would generally display stronger habits and assign greater intrinsic value to the familiar app sequences, evidenced by a marked preference for executing them even when presented with a simpler alternative sequence.”

A few notes on the task description and other task components:

1. It would be useful to give more details on the task. This includes more details on the time/condition of the gradual removal of visual and auditory stimuli and also on the within practice dynamic structure (i.e., different levels appear in the video).

Response: These details will be included in the revised manuscript. Thank you for pointing out the need for further clarification of the task design.

Done in page 7

1. Some more information on engagement-related exclusion criteria would be useful (what happened if participants did not use the app for more than one day, how many times were allowed to skip a day etc.).

Response: This additional information will be added to the revised manuscript. If participants omitted to train for more than 2 days, the researcher would send a reminder to the participant to request to catch up. If the participant would not react accordingly and a third day would be skipped, then the researcher would call to understand the reasons for the lack of engagement and gauge motivation. The participant would be excluded if more than 5 sequential days of training were missed. Only 2 participants were excluded given their lack of engagement.

Done in page 8

1. According to the (very useful) video demonstrating the task and the paper describing the task in detail (Banca et al., 2020), the task seems to include other relevant components that were not mentioned in this paper. I refer to the daily speed test, the daily random switch test, and daily ratings of each sequence's enjoyment and confidence of knowledge.

If these components were not included in this procedure, then the deviations from the procedure described in the video and Banca al. (2020) should be explicitly mentioned. If these components were included, at least some of them may be relevant, at least in part, to automaticity, habitual action control, formulation of participants' enjoyment from the app etc. I think these components should be mentioned and analyzed (or at least provide an explanation for why it has been decided not to analyze them).

This is also true for the reward removal (extinction) from the 21st day onwards which is potentially of particular relevance for the research questions.

Response: The task procedure was indeed the same as detailed in Banca et al., 2020. We did not include these extra components in this current manuscript for reasons of succinctness and because the manuscript was already rather longer than a common research article, given that we present three different, though highly inter-dependent, experiments in order to answer key interrelated questions in an optimal manner. However, since referee 2 considers this additional analysis to be important, we will be happy to include it in the supplementary material of the revised manuscript.

These additional components of the task as well as the respective analysis are now described in the Supplementary Materials.

Training engagement analysis:

1. I find referring to the number of trials including successful and unsuccessful trials as representing participants "commitment to training" (e.g. in Figure legend 2b) potentially inadequate. Given that participants need at least 20 successful trials to complete each practice, more errors would lead to more trials. Therefore, I think this measure may mostly represent weaker performance (of the OCD patients as shown in Figure 2b). Therefore, I find the number of performed practice runs, as used in Figure 2a (which should be perfectly aligned with the number of successful trials), a "clean" and proper measure of engagement/commitment to training.

Response: We acknowledge referee’s concern on this matter and agree to replace the y-axis variable of Figure 2b to the number of performed practices (thus aligning with Figure 2a). This amendment will remove any potential effect of weaker performance on the engagement measurement and will provide clearer results.

We have now decided to remove this figure as it does not add much to figure 2a. Instead, we replaced figure 2b and 2c for new plots, following new analysis linked to the next reviewer request (point 10)

1. Also, to provide stronger support for the claim about different diurnal training patterns (as presented in Figure 2c and the text) between patients and healthy individuals, it would be beneficial to conduct a statistical test comparing the two distributions. If the results of this test are not significant, I suggest emphasizing that this is a descriptive finding.

Response: Done, see revised Figure 2b and 2c. We have assessed the diurnal training patterns within each group using circular statistics, followed by independent-sample statistical testing of those circular distributions with the Watson’s U2 test ( Landler et al., 2021). While OCD participants have a group effect of practice with a significant peak at ~18:00, and HV participants have an earlier significant peak at ~15:00, the Watson’s U test did not find statistical betweengroup differences.

• Landler L, Ruxton GD, Malkemper EP. Advice on comparing two independent samples of circular data in biology. Scientific reports. 2021 Oct 13;11(1):20337.

Learning results:

1. When describing the Learning results (p10) I think it would be useful to provide the descriptive stats for the MT0 parameter (as done above for the other two parameters).

Response: Thank you for pointing this out. The descriptive stats for MT0 will be added to the revised version of the manuscript.

Done page 11

1. Sensitivity of sequence duration and IKI consistency (C) to reward:

I think it is important to add details on how incorrect trials were handled when calculating ∆MT (or C) and ∆R, specifically in cases where the trial preceding a successful trial was unsuccessful. If incorrect trials were simply ignored, this may not adequately represent trial-by-trial changes, particularly when testing the effect of a trial's outcome on performance change in the next trial.

Response: This is an important question. Our analysis protocol was designed to ensure that incorrect trials do not contaminate or confound the results. To estimate the trial-to-trial difference in ∆MT (or C) and ∆R, we exclusively included pairs of contiguous trials where participants achieved correct performance and received feedback scores for both trials. For example, if a participant made a performance error on trial 23, we did not include ∆R or ∆MT estimates for the pairs of trials 23-22 and 24-23. Instead of excluding incorrect trials from our analyses, we retained them in our time series but assigned them a NaN (not a number) value in Matlab. As a result, ∆R and ∆MT was not defined for those two pairs of trials. Similarly for C. This approach ensured that our analyses are not confounded by incremental or decremental feedback scores between noncontiguous trials. In the past, when assessing the timing of correct actions during skilled sequence performance, we also considered events that were preceded and followed by correct actions. This excluded effects such as post-error slowing from contaminating our results (Herrojo Ruiz et al., 2009, 2019). Therefore, we do not believe that any further reanalysis is required.

• Ruiz MH, Jabusch HC, Altenmüller E. Detecting wrong notes in advance: neuronal correlates of error monitoring in pianists. Cerebral cortex. 2009 Nov 1;19(11):2625-39.

• Bury G, García-Huéscar M, Bhattacharya J, Ruiz MH. Cardiac afferent activity modulates early neural signature of error detection during skilled performance. NeuroImage. 2019 Oct 1;199:704-17.

1. I have a serious concern with respect to how the sensitivity of sequence duration to reward is framed and analyzed. Since reward is proportional to performance, a reduction in reward essentially indicates a trial with poor performance, and thus even regression to the mean (along with a floor effect in performance [asymptote]) could explain the observed effects. It is possible that even occasional poor performance could lead to a participant demonstrating this effect, potentially regardless of the reward. Accordingly, the reduced improvement in performance following a reward decrease as a function of training length described in Figure 5b legend may reflect training-induced increased performance that leaves less room for improvement after poor trials, which are no longer as poor as before. To address this concern, controlling for performance (e.g., by taking into consideration the baseline MT for the previous trial) may be helpful. If the authors can conduct such an analysis and still show the observed effect, it would establish the validity of their findings."

Response: Thank you for raising this point. This has been done, see updated Figures 5 and 6. After normalizing the ∆MT(n+1) := MT(n+1) – MT(n) difference values by dividing them with the baseline MT(n) at trial n, we obtain the same results. Similar results are also obtained for IKI consistency (C).

See below our initial response from June 2023.

Thank you for raising this point. Figure 5b illustrates two distinct effects of reward changes on behavioral adaptation, which are expected based on previous research.

I. Practice effects: Firstly, we observe that as participants progress across bins of practice, the degree of improvement in behavior (reflected by faster movement time, MT) following a decrease in reward (∆R−) diminishes, consistent with our expectations based on previous work. Conversely, we found that ∆MT does not change across bins of practices following an increase in reward (∆R+).

We appreciate the reviewer’s suggestion regarding controlling for the reference movement time (MT) in the previous trial when examining the practice effect in the p(∆T|∆R−) and p(∆T|∆R+) distributions. In the revised manuscript, we will conduct the proposed control analysis to better understand whether the sensitivity of MT to score decrements changes across practice when normalising MT to the reference level on each trial. But see below for a preliminary control analysis.

II. Asymmetry of the effect of ∆R− and ∆R+ on performance: Figure 5b also depicts the distinct impact of score increments and decrements on behavioural changes. When aggregating data across practice bins, we consistently observed that the centre of the p(∆T|∆R−) distribution was smaller (more negative) than that of p(∆T|∆R+). This suggests that participants exhibited a greater acceleration following a drop in scores compared to a relative score increase, and this effect persisted throughout the practice sessions. Importantly, this enhanced sensitivity to losses or negative feedback (or relative drops in scores) aligns with previous research findings (Galea et al., 2015; Pekny et al., 2014; van Mastrigt et al., 2020).

We have conducted a preliminary control analysis to exclude the potential impact that reference movement time (MT) values could have on our analysis. We have assessed the asymmetry between behavioural responses to ∆R− and ∆R+ using the following analysis: We estimated the proportion of trials in which participants exhibited speed-up (∆T < 0) or slow-down (∆T > 0) behaviour following ∆R− and ∆R+ across different practice bins (bins 1 to 4). By discretising the series of behavioural changes (∆T) into binary values (+1 for slowing down, -1 for speeding up), we can assess the type of changes (speed-up, slow-down) without the absolute ∆T or T values contributing to our results. We obtained several key findings:

• Consistent with expectations (sanity check), participants exhibited more instances of speeding up than slowing down across all reward conditions.

• Participants demonstrated a higher frequency of speeding up following ∆R− compared to ∆R+, and this asymmetry persisted throughout the practice sessions (greater proportion of -1 events than +1 events). 53% events were speed-up events in the in the p(∆T|∆R+) distribution for the first bin of practices, and 55% for the last bin. Regarding p(∆T|∆R-), there were 63% speed-up events throughout each bin of practices, with this proportion exhibiting no change over time.

• Accordingly, the asymmetry of reward changes on behavioural adaptations, as revealed by this analysis, remained consistent across the practice bins.

Thus, these preliminary findings provide an initial response to referee 2 and offer valuable insights into the asymmetrical effects of positive/negative reward changes on behavioural adaptations. We plan to include these results in the revised manuscript, as well as the full control analysis suggested by the referee. We will further expand upon their interpretation and implications.

1. Another way to support the claim of reward change directionality effects on performance (rather than performance on performance), at least to some extent, would be to analyze the data from the last 10 days of the training, during which no rewards were given (pretending for analysis purposes that the reward was calculated and presented to participants). If the effect persists, it is less unlikely that the effect in question can be attributed to the reward dynamics.

Response: The reviewer’s concern is addressed in the previous quesQon. Also, this analysis would not be possible because our Gaussian fit analyses use the Qme series of conQnuous reward scores, in which ∆R− or ∆R+ are embedded. These events cannot be analyzed once reward feedback is removed because we do not have behavioral events following ∆R− or ∆R+ anymore.

Done

1. This concern is also relevant and should be considered with respect to the sensitivity of IKI consistency (C) to reward. While the relationship between previous reward/performance and future performance in terms of C is of a different structure, the similar potential confounding effects could still be present.

Response: We will conduct this analysis for the revised manuscript, similarly to the control analysis suggested by referee 2 on MT. Our preliminary control analysis, as explained above, suggests that the fundamental asymmetry in the effect of ∆R+ and ∆R+ on behavioral changes persists when excluding the impact of reference performance values in our Gaussian fit analysis.

Done. See updated Figure 6. The results are very similar once we normalize the IKI consistency index C with the IKI of the baseline performance at trial n.

1. Another related question (which is also of general interest) is whether the preferred app sequence (as indicated by the participants for Phase B) was consistently the one that yielded more reward? Was the continuous sequence the preferred one? This might tell something about the effectiveness of the reward in the task.

Response: We have now conducted this analysis. There is in fact no evidence to conclude that the continuously rewarded sequence was the preferred one. The result shows that 54.5% of HV and 29% of the OCD sample considered the continuous sequence to be their preferred one, a nonstatistically significant difference. Note that this preference may not necessarily be linked simply to programmed reward. The overall preference may be influenced by many other factors, such as, for example, the aesthetic appeal of particular combinations of finger movements.

Regarding both experiments 2 and 3:

1. The change in context in experiment 2 and 3 is substantial and include many different components. These changes should be mentioned in more detail in the Results section before describing the results of experiments 2 and 3.

Response: Following referee’s advice, we will move these details (currently written in the Methods section) to the Results section, when we introduce Phase B and before describing the results of experiments 2 and 3.

Done in page 21

Experiment 2:

1. In Experiment 2, the authors sometimes refer to the "explicit preference task" as testing for habitual and goal-seeking sequences. However, I do not think there is any justification for interpreting it as such. The other framings used by the authors - testing whether trained action sequences gain intrinsic/rewarding properties or value, and preference for familiar versus novel action sequences - are more suitable and justified. In support of the point I raised here, assigning intrinsic rewarding properties to the learned sequences and thereby preferring these sequences can be conceptually aligned with goal-directed behavior just as much as it could be with habit.

Response: We clearly defined the theoretical framing of experiment 2 as a test of whether trained action sequences gain intrinsic value and we are pleased to hear that the referee agrees with this framing. If the referee is referring to the paragraph below (in the Discussion), we actually do acknowledge within this paragraph that a preference for the trained sequences can either be conceptually aligned with a habit OR a goal-directed behavior.

“On the other hand, we are describing here two potential sources of evidence in favor of enhanced habit formation in OCD. First, OCD patients show a bias towards the previously trained, apparently disadvantageous, action sequences. In terms of the discussion above, this could possibly be reinterpreted as a narrowing of goals in OCD (Robbins et al., 2019) underlying compulsive behavior, in favor of its intrinsic outcomes”

This narrowing of goals model of OCD refers to a hypothetically transiQonal stage of compulsion development driven by behavior having an abnormally strong, goal-directed nature, typically linked to specific values and concerns.

If the referee is referring to the penulQmate sentence of hypothesis secQon, this has been amended in response to Q5. We cannot find any other possible instances in this manuscript stating that experiment 2 is a test of habitual or goal-directed behavior.

Experiment 3:

1. Similar to Experiment 2, I find the framing of arbitration between goal-directed/habitual behavior in Experiment 3 inadequate and unjustified. The results of the experiment suggest that participants were primarily goal-directed and there is no evidence to support the idea that this reevaluation led participants to switch from habitual to goal-directed behavior.

Also, given the explicit choice of the sequence to perform participants had to make prior to performing it, it is reasonable to assume that this experiment mainly tested bias towards familiar sequence/stimulus and/or towards intrinsic reward associated with the sequence in value-based decision making.

Response: This comment is aligned with (and follows) the referee’s criticism of experiment 1 not achieving automatic and habitual actions. We have addressed this matter above, in response 1 to Referee 2.

Mobile-app performance effect on symptomatology: exploratory analyses:

1. Maybe it would be worth testing if the patients with improved symptomatology (that contribute some of their symptom improvement to the app) also chose to play more during the training stage.

Response: We have conducted analysis to address this relevant question. There is no correlation between the YBOCS score change and the number of total practices, meaning that the patients who improved symptomatology post training did not necessarily chose to play the app more during the training stage (rs = 0.25, p = 0.15). Additionally, we have statistically compared the improvers (patients with reduced YBOCS scores post-training) and the non-improvers (patients with unchanged or increased YBOCS scores post-training) in their number of app completed practices during the training phase and no differences were observed (U = 169, p = 0.19).

The result from the correlational analysis has been added to the revised manuscript (page 28).

Discussion:

1. Based on my earlier comments highlighting the inadequacy and mis-framing of the work in terms of habit and goal-directed behavior, I suggest that the discussion section be substantially revised to reflect these concerns.

Response: We do not agree that the work is either "inadequate or mis-framed" and will not therefore be substantially revising the Discussion. We will however clarify further the interpretation we have made and make explicit the alternative viewpoint of the referee. For example, we will retitle experiment 3 as “Re-evaluation of the learned action sequence: possible test of goal/habit arbitration” to acknowledge the referee’s viewpoint as well as our own interpretation.

Done

1. In the sentence "Nevertheless, OCD patients disadvantageously preferred the previously trained/familiar action sequence under certain conditions" the term "disadvantageously" is not necessarily accurate. While there was potentially more effort required, considering the possible presence of intrinsic reward and chunking, this preference may not necessarily be disadvantageous. Therefore, a more cautious and accurate phrasing that better reflects the associated results would be useful.

Response: We recognize that the term "disadvantageously" may be semantically ambiguous for some readers and therefore we will remove it.

Done

Materials and Methods:

1. The authors mention: "The novel sequence (in condition 3) was a 6-move sequence of similar complexity and difficulty as the app sequences, but only learned on the day, before starting this task (therefore, not overtrained)." - for the sake of completeness, more details on the pre-training done on that day would be useful.

Response: Details of the learning procedure of the novel sequence (in condition 3, experiment 3) will be provided in the methods of the revised version of the manuscript.

Done in page 40

Minor comments:

1. In the section discussing the sensitivity of sequence duration to reward, the authors state that they only analyzed continuous reward trials because "a larger number of trials in each subsample were available to fit the Gaussian distributions, due to feedback being provided on all trials." However, feedback was also provided on all trials in the variable reward condition, even though the reward was not necessarily aligned with participants' performance. Therefore, it may be beneficial to rephrase this statement for clarity.

Response: We will follow this referee’s advice and will rephrase the sentence for clarity.

Done. See page 16.

1. With regard to experiment 2 (Preference for familiar versus novel action sequences) in the following statement "A positive correlation between COHS and the app sequence choice (Pearson r = 0.36, p = 0.005) further showed that those participants with greater habitual tendencies had a greater propensity to prefer the trained app sequence under this condition." I find the use of the word "further" here potentially misleading.

Response: The word "further" will be removed.

Done

Reviewer #1 (Recommendations For The Authors):

This is a very interesting manuscript, which was a pleasure to review. I have some minor comments you may wish to consider.

1. I believe that it is possible to include videos as elements in eLife articles - please consider if you can do this to demonstrate the action sequence on the smartphone. I followed the YouTube video, and it was very helpful to see exactly what participants did, but it would be better to attach the video directly, if possible.

Response: This is a great idea and we will definitely attach our video demonstrating the task to the revised manuscript (Version of Record) if the eLife editors allow.

We ask permission to the editor to add the video

1. The abstract states that the study uses a "novel smartphone app" but is the same one as described in Banca et al. Suggest writing simply "smartphone app".

Response: We will remove the word novel.

Done

1. Some of the hypotheses described in the second half of the Hypothesis section could be stated more explicitly. For example: "We also hypothesized that the acquisition of learning and automaticity would differ between the two action sequences based on their associated rewarded schedule (continuous versus variable) and reward valence (positive or negative)." The subsequent sentence explains the prediction for the schedule but what is the hypothesized direction for reward valence? More detail is subsequently given on p. 14, Results, but it would be better to bring these details up to the Introduction. "We additionally examined differential effects of positive and negative feedback changes on performance to build on previous work demonstrating enhanced sensitivity to negative feedback in patients with OCD (Apergis-Schoute et al 2023, Becker et al., 2014; Kanen et al., 2019)." In general, the second part of the Hypothesis section is a bit dense, sometimes with two predictions per sentence. It could be useful for the reader if hypotheses were enumerated and/or if a distinction was made among the hypotheses with respect to their importance.

We fully revised the hypothesis section, on page 5, following this reviewer’s suggestion. We think this section is much clearer now, in our revised manuscript.

Response: Thank you for pointing out the need for clarity in our hypothesis section. This is a very important point and we will carefully rewrite our hypothesis in the revised manuscript to make them as clear as possible.

1. Did medication status correlate with symptom severity in the OCD group (e.g., higher symptoms for the 6 participants on SSRI+antipsychotics?). Could this, or SSRI-only status, have impacted results in any way? I appreciate that there is no way to test medication status statistically but readers may be interested in your thoughts on this aspect.

Response: We have now conducted exploratory analysis to assess the potential effect of medication in the following output measures: app engagement (as measured by completed practices), explicit preference and YBOCS change post-training. The patients who were on combined therapy (SSRIs + antipsychotic) did not perform significantly different in these measures as compared to the remaining patients and no other effects of interest were observed. Their symptomatology was indeed slightly more severe but not statistically significant [Y-BOCS combined = 26.2 (6.5); Y-BOCS SSRI only = 23.8 (6.1); Y-BOCS No Med = 23.8 (2.2), mean(std)]. Only one patient showed symptom improvement after the app training, another became worse and the remaining patients on combined therapy remain stable during the month.

Palminteri et al (2011) found that unmedicated OCD patients exhibited instrumental learning deficits, which were fully alleviated with SSRI treatment. Therefore, it is possible that the SSRI medication (present in our sample) may have reduced habit formation and facilitated behavioral arbitration. However, since the effect goes against the habit hypothesis, it has is unlikely that it has confounded our measure of automaticity. If anything, medication rendered experiment 2 and 3 more goal-oriented. We agree that further studies are warranted to address the effect of SSRIs on these measures.

1. You could explain earlier why devaluation could not be tested here (it is only explained in the Limitations section near the end)

Response: The revised manuscript will be amended to account for this note.

Done in page 25.

1. Capitalize 'makey-makey', I didn't realize there was a product called Makey Makey until I Googled it.

Response: Sure. We will capitalize 'Makey-Makey'. Thank you for pointing this out!

Done

Reviewer #2 (Recommendations For The Authors):

Recommendations for the authors (ordered by the paper sections):

In the introduction

1. regarding this part "We used a period of 1-month's training to enable effective consolidation, required for habitual action control or skill retention to occur. This acknowledged previous studies showing that practice alone is insufficient for habit development as it also requires off-line consolidation computations, through longer periods of time (de Wit et al., 2018) and sleep (Nusbaum et al., 2018; Walker et al., 2003)." I advise the authors to re-check whether what is attributed here to de Wit et al. (2018) is indeed justified (if I remember correctly they have not mentioned anything about off-line consolidation computations).

Response: When we revise the manuscript, we will remove the de Wit et al. (2018) citation from this sentence.

Done

in the Outline paragraph

1. it stated: "We continuously collected data online, in real time, thus enabling measurements of procedural learning as well as automaticity development." I think this wording implies that the fact that the data was collected online in real time was advantageous in that it enabled to assess measurements of procedural learning and automaticity development, which in my understanding is not the case.

Response: To make this sentence clearer, we will change it to the following: ‘We continuously collected data online, to monitor engagement and performance in real time and to enable acquisition of sufficient data to analyze, à posteriori, procedural learning and automaticity development’.

Done in page 4: ‘We collected data online continuously to monitor engagement and performance in real-time. This approach ensured we acquired sufficient data for subsequent analysis of procedural learning and automaticity development’.

1. In the final sentence of this paragraph "or and" should be changed to "or/end".

Response: This was a typo. The word ‘and’ will be removed.

Done

1. In Figure 1c - Note that in the figure legend it says "Each sequence comprises 3 single press moves, 2 two-finger moves..." whereas in the example shown in the figure it's the other way around (2 single press moves and 3 two-finger moves).

Response: Thank you so much for spotting this! The example shown in the figure is incorrect. We apologize for the mistake. It should depict 3 single press moves, 2 two-finger moves and 1 three- finger move. The figure will be amended.

Done

In the results section:

1. Regarding the "were followed by a positive ring tone and the unsuccessful ones by a negative ring tone", I suggest mentioning that there was also a positive visual (rewarding) effect.

Response: Thank you. A mention to the visual effect will be added for both the positive (successful) and negative (unsuccessful) trials. Done in page 7

1. p 10. - Note a typo in the following sentence where the word "which" appears twice consecutively:

"Furthermore, both groups exhibited similar motor durations at asymptote which, which combined with the previous conclusion, indicates that OCD patients improved their motor learning more than controls, but to the same asymptote."

Response: Thank you for spotting this typo. The second word will be removed. Done

1. I have a few suggestions with respect to Figure 3:

2. keeping the y-axes scale similar in all subplots would be more visually informative.

Here we kept the y-axes scale similar in all subplots, except one of them, which was important to keep to capture all the data.

1. For the subplots in 3b I would recommend for the transparent regions, instead of the IQR, to use the median +/- 1.57 * IQR/sqrt(n) which is equivalent to how the notches are calculated in a box-plot figure (It is referred to as an approximate 95% confidence interval for the median). This should make the transparent area narrower and thus better communicate the results.

Done

1. I think the significant levels mentioned in figure legend 3b (which are referring to the group effect measured for each reward schedule type separately) is not mentioned in the text. While not crucial, maybe consider adding it in the text.

We don’t think this is necessary and may actually lead to confusion because in the text we report a Kruskal–Wallis H test (which is the most appropriate statistical test), including their H and p values for the group and reward effects. Since in the figure we separated the analysis and plots for variable and continuous reward schedules (for visual purposes) , we reported a U test separated for each reward schedule. Therefore, we consider that the correct statistics are reported in the appropriate places of the manuscript.

Response: Thank you for this very helpful suggestion. We will amend figure 3 accordingly.

1. In the Automaticity results (pp. 12 and 13) when describing the Descriptive stats the wrong parameter indicator are used (DL instead of CL and nD instead of nC.

Response: Thank you for noticing it. We will amend.

Done

1. In Sensitivity of IKI consistency (C) to reward results:

In Figure 6a legend: with respect to "... and for reward increments (∆R+, purple) and decrements (∆R-, green)" - note that there are also additional colors indicating these ∆Rs.

Response: Done. We had used a 2 x 2 color scheme: green hues for ∆R-, and purple hues for ∆R+. Then, OCD is denoted by dark colors, and HV by light colors. This represents all four colors used in the figure. For instance, OCD and ∆R- is dark green, whereas OCD and ∆R+ is denoted by dark purple.

1. p.21 - the YBOCS abbreviation appears before the full form is spelled out in the text.

Response: In the revised version, we will make sure the YBOCS abbreviation will be spelled out the first time it is mentioned.

Done in page 24

Experiments 2 and 3:

1. If there is a reason behind presenting the conditions sequentially rather than using intermixed trials in experiments 2 and 3, it would be useful to mention it in the text.

Response: Experiment 2 could have used intermixed trials. However, we were concerned that the use of intermixed trials in experiment 3 would increase excessively the memory load of the task, which could then be a confound.

Done in page 41

1. I wonder whether the presentation order of the conditions in experiments 2 and 3 affected participants' results? Maybe it is worth adding this factor to the analysis.

Response: As we mentioned both in the methods and results sections, we counterbalanced all the conditions across participants, in both experiments 2 and 3. This procedure ensures no order effects.

Experiment 2:

1. Regarding this sentence (pp. 21-22): "However, some participants still preferred the app sequence, specifically those with greater habitual tendencies, including patients who considered the app training beneficial." I think the part that mentions that there are "patients who considered the app training beneficial" appears below and it may confuse the reader. I suggest either providing a brief explanation or indicating that further details will be provided later in the text ("see below in...").

Response: We will clarify this section.

We added “see below exploratory analyses of “Mobile-app performance effect on symptomatology”” in the end of the sentence so that the reader knows this is further explained below. Page 25

1. Finally, in addition to subgrouping maybe it is worth testing whether there is a correlation between the YBOCS score change and the app-sequences preference (as to learn if the more they change their YBOCS the more they prefer the learned sequences and vice versa?)

Response: Thank you for suggesting this relevant correlational analysis, which we have now conducted. Indeed, there is a correlation between the YBOCS score change and the preference for the app-sequences, meaning that the higher the symptom improvement after the month training, the greater the preference for the familiar/learned sequence. This is particularly the case for the experimental condition 2, when subjects are required to choose between the trained app sequence and any 3-move sequence (rs = 0.35, p=0.04). A trend was observed for the correlation between the YBOCS score change and the preference for the app-sequences in experimental condition 1 (app preferred sequence versus any 6-move sequence): rs = 0.30, p=0.09.

This finding represents an additional corroboration of our conclusion that the app seems to be more beneficial to patients more prone to routine habits, who are somewhat more averse to novelty.

This analysis was added in page 24, 25 and page 35.

Experiment 3:

1. You mention "The task was conducted in a new context, which has been shown to promote reengagement of the goal system (Bouton, 2021)." In my understanding this observation is true also for experiment 2. In such case it should be stated earlier (probably under: "Phase B: Tests of actionsequence preference and goal/habit arbitration").

Response: As answered above in (Q17), we will follow this referee 2’s suggestion and describe the contextual details of experiments 2 and 3 in the Results section, when we introduce Phase B.

Done in page 21.

1. w.r.t this sentence - "...that sequence (Figure 8b, no group effects (p = 0.210 and BF = 0.742, anecdotal evidence)" I would add what the anecdotal evidence refers (as done in other parts of the paper), to prevent potential confusion.

Response: OK, this will be added.

Added on page 27

Discussion:

1. w.r.t. "Here we have trained a clinical population with moderately high baseline levels of stress and anxiety, with training sessions of a higher order of magnitude than in previous studies (de Wit et al., 2018, 2018; Gera et al., 2022) (30 days instead of 3 days)." The Gera et al. 2022 (was more than 3 days), you probably meant Gera et al. 2023 ("Characterizing habit learning in the human brain at the individual and group levels: a multi-modal MRI study", for which 3 days is true).

Response: Thank you for pointing this out. We will keep the citation to Gera et al 2022 given its relevance to the sentence but we will remove the information inside the parenthesis. This amendment will solve the issue raised here.

Done in page 32

1. w.r.t "to a simple 2-element sequence with less training (Gera et al., 2022)" - it's a 3-element sequence in practice.

Response: Thank you for this correction. We will amend this sentence accordingly.

Done in page 32

1. (p.30) w.r.t "and enhanced error-related negativity amplitudes in OCD" - a bit more context of what the negative amplitudes refer to would be useful (So the reader understands it refers to electrophysiology).

Response: We will add a sentence in our revised manuscript addressing this matter. This sentence has been removed in the revised manuscript

Supplementary materials:

1. under "Sample size for the reward sensitivity analysis":

It is stated "One practice corresponded to 20 correctly performed sequences. We therefore split the total number of correct sequences into four bins." I was not able to follow this reasoning here (20 correct trials in practice => splitting the data the 4 bins). More clarity here would be useful.

Response: We will clarify this procedure of our analysis in the revised version of the manuscript. Thanks.

Done. See Supplementary materials.

1. Also, maybe I am missing something, but I couldn't understand why the number of sequences available per bin is different for the calculation of ∆MT and C. Aren't any two consecutive sequences that are good for the calculation of one of these measures also good for the calculation of the other?

Response: Thank you for pointing this out. Indeed, the number of trials was the same for both analyses, ∆MT and C. We had saved an incorrect variable as number of trials. We will amend the text.

We have re-analyzed the trial number data. The average number of trials per bin both for the ∆MT and C analyses was 109 (9) in the HV and 127 (12) in OCD groups. Although the number was on average larger in the patient group, we did not find significant differences between groups (p = 0.47).

When assessing the p(∆T|∆R+) and p(∆T|∆R-) separately, more trials were available for p(∆T|∆R+), 107 (10) , than for p(∆T|∆R-), and 98 (8). These trial numbers differed significantly (p = 0.0046), but were identical for ∆MT and C analyses.

Done. Included in Supplementary materials.

Minor comments:

1. Not crucial, but maybe for the sake of consistency consider merging the "Self-reported habit tendencies" section and the "Other self-reported symptoms" section, preferably where the latter is currently placed.

Response: We fully understand the referee’s rationale underlying this suggestion. We indeed considered initially presenting the self-reported questionnaires all together, in a last, single section of the results, as suggested by the referee. However, we decided to report the higher habitual tendencies of OCD as an initial set of results, not only because it is a novel and important finding (which justifies it to be highlighted) but also because it is essential to the understanding of some of the remaining results presented.

1. In some figure legends the percentage of the interval of the mentioned confidence intervals (probably 95%) is missing. I suggest adding it.

Response: OK, this will be added.

Done

1. The NHS abbreviation appears without spelling out the full form.

Response: This will be amended accordingly.

I removed NHS as it is not relevant.

1. In p.38 the citation (Rouder et al., 2012) is duplicated (appears twice consecutively).

Response: Thank you for pointing this out. We will amend accordingly.

Done

In the results section:

1. The authors mention: "To promote motivation, the total points achieved on each daily training sessions were also shown, so participants could see how well they improved across days". Yet, if the score is based on the number of practices, it may not represent participants improvement in case in some days more practices are performed. I suggest to clarify this point.

Response: The goal of providing the scoring feedback was, as explained in the sentence, to gauge motivation and inform the subject about their performance. Having this goal in mind, it does not really matter if one day their scoring would be higher simply because they would have done more practice on that day. Participants could easily understand that the scoring reflected their performance on each practice so they would realize that the more practice, the greater their improvement and that the scoring would increase across days of practice. We will amend the sentence to the following: "To promote motivation, the total points achieved on each training session (i.e. practice) was also shown, so participants could see how well they improved across practice and across days".

Done in page 7 and 8.

Author Response

We thank the reviewers for their fair assessment of our work and will submit a revised version edited for clarity of presentation and precision of interpretations.

Author Response

Reviewer #3 (Public Review):

[...] Weaknesses:

The study produces a large amount of data that is in general cohesive and support the main conclusions, but more thorough considerations on some of their findings may be helpful, as exemplified by the following:

1) the effect of microglial ablation on chloral hydrate-induced RORR in Fig. 1B appears to be not the same as other anesthetics. what does this mean?

2) Macrophage ablation impedes anesthesia emergence from pentobarbital (Fig. 3C). how may this occur?

3) examination of the potential effect of microglial depletion on dendritic spine density is interesting but the experimental design does not seem to align well with the PPR and eEPSC data, which indicate a reduction in presynaptic release (Fig.10E) and increase of postsynaptic function (Fig. 10H), respectively. The PPR data seems to suggest a presynaptic effect of microglia; ablation.

This reviewer may confused the brain regions between our spine quantification (Figure 11) and patch-clamp recording (Figure 10). In our spine quantification, all evaluations were conducted in the mPFC. However, the patch-clamp recording were performed in SON (Figure 10 B-F) and LC (Figure 10 G-K), different brain regions from our spine quantification. As one of our conclusion, microglia differentially modulate the activity of neuronal network in a brain region-specific manner, neurons in different brain regions may exhibit different electrophysiological alterations upon microglial depletion. Therefore, this comment might be a factual error.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

This is an interesting, timely and informative article. The authors used publicly available data (made available by a funding agency) to examine some of the academic characteristics of the individuals recipients of the National Institutes of Health (NIH) k99/R00 award program during the entire history of this funding mechanism (17 years, total ~ 4 billion US dollars (annual investment of ~230 million USD)). The analysis focuses on the pedigree and the NIH funding portfolio of the institutions hosting the k99 awardees as postdoctoral researchers and the institutions hiring these individuals. The authors also analyze the data by gender, by whether the R00 portion of the awards eventually gets activated and based on whether the awardees stayed/were hired as faculty at their k99 (postdoctoral) host institution or moved elsewhere. The authors further sought to examine the rates of funding for those in systematically marginalized groups by analyzing the patterns of receiving k99 awards and hiring k99 awardees at historically black colleges and universities.

The goals and analysis are reasonable and the limitations of the data are described adequately. It is worth noting that some of the observed funding and hiring traits are in line with the Matthew effect in science (https://www.science.org/doi/10.1126/science.159.3810.56) and in science funding (https://www.pnas.org/doi/10.1073/pnas.1719557115). Overall, the article is a valuable addition to the research culture literature examining the academic funding and hiring traits in the United States. The findings can provide further insights for the leadership at funding and hiring institutions and science policy makers for individual and large-scale improvements that can benefit the scientific community.

Thank you for these comments. We have incorporated the articles referenced on the Matthew effect into the first paragraph of the Discussion our revised preprint.

Reviewer #2 (Public Review):

Early career funding success has an immense impact on later funding success and faculty persistence, as evidenced by well-documented "rich-get-richer" or "Matthew effect" phenomena in science (e.g., Bol et al. 2018, PNAS). Woitowich et al. examined publicly available data on the distribution of the National Institutes of Health's K99/R00 awards - an early career postdoc-to-faculty transition funding mechanism - and showed that although 85% of K99 awardees successfully transitioned into faculty, disparities in subsequent R01 grant obtainment emerged along three characteristics: researcher mobility, gender, and institution. Men who moved to a top-25 NIH funded institution in their postdoc-to-faculty transition experienced the shortest median time to receiving a R01 award, 4.6 years, in contrast to the median 7.4 years for women working at less well-funded schools who remained at their postdoc institutions. This result is consistent with prior evidence of funding disparities by gender and institution type. The finding that researcher mobility has the largest effect on subsequent funding success is key and novel, and enhances previous work showing the relationship between mobility and ones' access to resources, collaborators, or research objects (e.g., Sugimoto and Larivière, 2023, Equity for Women in Science (Harvard University Press)).

These results empirically demonstrate that even after receiving a prestigious early career grant, researchers with less mobility belonging to disadvantaged groups at less-resourced institutions continue to experience barriers that delay them from receiving their next major grant. This result has important policy implications aimed at reducing funding disparities - mainly that interventions that focus solely on early career or early stage investigator funding alone will not achieve the desired outcome of improving faculty diversity.

The authors also highlight two incredible facts: No postdoc at a historically Black college or university (HBCU) has been awarded a K99 since the program's launch. And out of all 2,847 R00 awards given thus far, only two have been made to faculty at HBCUs. Given the track record of HBCUs for improving diversity in STEM contexts, this distribution of awards is a massive oversight that demands attention.

At no fault of the authors, the analysis is limited to only examining K99 awardees and not those who applied but did not receive the award. This limitation is solely due to the lack of data made publicly available by the NIH. If this data were available, this study would have been able to compare the trajectory of winners versus losers and therefore could potentially quantify the impact of the award itself on later funding success, much like the landmark Bol et al. (2018) paper that followed the careers of winners of an early career grant scheme in the Netherlands. Such an analysis would also provide new insights that would inform policy.

Although data on applications versus awards for the K99/R00 mechanism are limited, there exists data for applicant race and ethnicity for the 2007-2017 period, which were made available by a Freedom of Information Act request through the now defunct Rescuing Biomedical Research Initiative: https://web.archive.org/web/20180723171128/http://rescuingbiomedicalresearch.org/blog/examining-distribution-k99r00-awards-race/. These results are not presently discussed in the paper, but are highly relevant given the discussion of K99 award impacts on the sociodemographic composition of U.S. biomedical faculty. From 2007 to 2017, the K99 award rate for white applicants was 31.0% compared to 26.7% for Asian applicants and 16.2% for Black applicants. In terms of award totals, these funding rates amount to 1,384 awards to white applicants, 610 to Asian applicants, and 25 to Black applicants for the entire 2007-2017 period. And in terms of R00 awards, or successful faculty transitions: whereas 77.0% of white K99 awardees received an R00 award, the conversion rate for Asian and Black K99 awardees was lower, at 76.1% and 60.0%, respectively. Regarding this K99-to-R00 transition rate, Woitowich et al. found no difference by gender (Table 2). These results are consistent with a growing body of literature that shows that while there have been improvements to equity in funding outcomes by gender, similar improvements for achieving racial equity are lagging.

The conclusions are well-supported by the data, and limitations of the data and the name-gender matching algorithm are described satisfactorily.

One aspect that the authors should expand or comment on is the change in the rate of K99 to R00 conversions. Since 2016, while the absolute number of K99 and R00 awards has been increasing, the percentage of R00 conversions appears to be decreasing, especially in 2020 and 2021. This observation is not clearly stated or shown in Figure 1 but is an important point - if the effectiveness of the K99/R00 mechanism for postdoc-to-faculty transitions has been decreasing lately, then something is undermining the purpose of this mechanism. This result bears emphasis and potentially discussion for possible reasons for why this is happening.

Thank you for these insightful comments. We now calculate a rolling conversion rate for K99 to R00 awards which shows there is not as much of a decline in conversion from K99 to R00 (Fig 1B). We still see a slight decline in 2021 and 2022. 468 K99 awards are from 2020 or later so they may still convert to the R00 phase. Thus it is difficult to draw conclusions about 2021/2022 yet. As more time passes, we may better be able to determine whether or not significant alteration from normal occurred in these years, presumably due to pressures from the Covid-19 pandemic. We also thank you for providing the details of the FOIA request. We have included a discussion of these data in the discussion.

Reviewer #3 (Public Review):

The researchers aim add to the literature on faculty career pathways with particular attention to how gender disparities persist in the career and funding opportunities of researchers. The researchers also examine aspects of institutional prestige that can further amplify funding and career disparities. While some factors about individuals' pathways to faculty lines are known, including the prospects of certain K award recipients, the current study provides the only known examination of the K99/R00 awardees and their pathways.

Strengths:

The authors establish a clear overview of the institutional locations of K99 and R00 awardees and the pathways for K99-to-R00 researchers and the gendered and institutional patterns of such pathways. For example, there's a clear institutional hierarchy of hiring for K99/R00 researchers that echo previous research on the rigid faculty hiring networks across fields, and a pivotal difference in the time between awards that can impact faculty careers. Moreover, there's regional clusters of hiring in certain parts of the US where multiple research universities are located. Moreover, documenting the pathways of HBCU faculty is an important extension of the Wapman et al. study (among others from that research group), and provides a more nuanced look at the pathways of faculty beyond the oft-discussed high status institutions. (However, there is a need for more refinement in this segment of the analyses as discussed further below.). Also, the authors provide important caveats throughout the manuscript about the study's findings that show careful attention to the complexity of these patterns and attempting to limit misinterpretations of readers.

Weaknesses:

The authors reference institutional prestige in relation to some of the findings, but there's no specific measure of institutional prestige included in the analyses. If being identified as a top 25 NIH-funded institution is the proximate measure for prestige in the study, then more justification of how that relates to previous studies' measures of institutional prestige and status are needed to further clarify the interpretations offered in the manuscript.

The identification of institutional funding disparities impacting HBCUs is an important finding and highlights another aspect of how faculty at these institutions are under resourced and arguably undervalued in their research contributions. However, a lingering question exists: why compare HBCUs with Harvard? What are the theoretical and/or methodological justifications for such comparisons? This comparison lends itself to reifying the status hierarchy of institutions that perpetuate funding and career inequalities at the heart of the current manuscript. If aggregating all HBCU faculty together, then a comparable grouping for comparison is needed, not just one institution. Perhaps looking at the top 25 NIH funded institutions could be one way of providing a clearer comparison. Related to this point is the confusing inclusion of Gallaudet in Figure 6 as it is not an officially identified HBCU. Was this institution also included in the HBCU-related calculations?

Thank you for this comment. We agree this comparison perpetuates the perception of the prestige hierarchy and is problematic. We now compare all institutions in the top 25 NIH funding category to all HBCUs. Thank you also for identifying our error in mis-coding Gallaudet as an HBCU. We have corrected this in the current version.

There is a clear connection that is missed in the current iteration of the manuscript derived from the work of Robert Merton and others about cumulative advantages in science and the "Matthew effect." While aspects of this connection are noted in the manuscript such as well-resourced institutions (those with the most NIH funding in this circumstance) hire each others' K99/R00 awardees, elaborating on these connections are important for readers to understand the central processes of how a rigid hierarchy of funding and career opportunities exist around these pathways. The work the authors build on from Daniel Larremore, Aaron Clauset, and their colleagues have also incorporated these important theoretical connections from the sociology of knowledge and science, and it would provide a more interdisciplinary lens and further depth to understanding the faculty career inequalities documented in the current study.

Reviewer #1 (Recommendations For The Authors):

Comments to authors:

1. For the benefit of general reader, it would be informative to mention the amount of annual NIH investment in the k99 funding mechanism in the text (230 awards representing a ~ 230 million US dollars investment).

Thank you for this suggestion. We have added that this is ~$25 million investment annually.

1. It is worth noting that some of the observed funding and hiring traits resemble the Matthew effect, discussed in: The Matthew effect in science: https://www.science.org/doi/10.1126/science.159.3810.56

The Matthew effect in science funding: https://www.pnas.org/doi/10.1073/pnas.1719557115

It would be of value to cite these for further context for the readers.

Thank you for this suggestion. We have included these references and briefly discussed the Matthew effect in the first paragraph of the Discussion.

1. Figs 3, 6 and Fig S1 are hard to read without zooming in due to their format and don't work great within a letter size page but can work if they are also linked to a zoomable web version. It would make sense to have an online navigable/searchable/selectable version. But when the reader zooms out, there are patterns that reflect what points the authors are making (though those could be illustrated differently). These figures are really made for online webapp visualization (such as Shiny in R).

We agree with this comment and have used the “googleVis()” package in R to put together interactive Sankey diagrams. These can be found at: https://dantyrr.github.io/K99-R00-analysis/ and they are referenced in the manuscript.

1. The abstract states 85% of awardees get R00 awards. That appears to come from 198/234 (page 6) though it's not explicitly stated, and other ratios give different answers (e.g., 1-304/3475 = 91%) but the 85% seems to be the right one. That first paragraph of the results could be clearer. Also, in the middle of page three the number given is 90% so something is inconsistent. For Figure 1A, given the methodology it should be possible to calculate a rolling conversion rate as "R00(t) / K99(t-1)" (and a similarly-calculated cumulative rate).

Thank you for catching these errors. These were introduced because there are R00 awardees that did not have extramural K99 awards. These are intramural NIH K99 awardees but there is no public data on these awardees. The correct number is 78% of K99 awardees that transitioned to the R00 phase. We have also calculated the rolling conversion rate which is 89% if you exclude the first 2 years of the program (when the first awardees were within the 2-yr K99 period) and final 2 years (when most recent K99 awardees were still within their first 2 years of the K99 period).

1. Assuming that 85% is the correct number, is there any information/insight into why ~1/6 of awardees do not continue to R00, which seems high given that only two years passes - that's a lot of awardees not getting R00 positions.

We are unsure of why these don’t convert. In the revised version of the manuscript, we speculate on this in the 4th paragraph of the discussion:

The factors that prevented the other 302 K99 awardees from 2019 and earlier unable to convert their K99-R00 grants is cause for concern within our greater academic community. Possible explanations include leaving the biomedical workforce, accepting tenure-track positions or other positions abroad, or by simply not successfully securing a tenable tenure-track offer.

1. It looks like perhaps a non-zero number of K99s are just one year and not two (e.g., see 2006 in Fig 1A, which should not appear if all 2006 awards were 2 years). What is the typical percentage of K99s not activated for a second year, and is this a sizable % of the 15% not converting to R00?

This is an interesting question. We didn’t originally look into this and the dataset that we originally downloaded from NIH reporter included a significant number of duplicates for the grants because year 1 of the K99 was listed on its own line and year 2 was listed on a different line. The first step in curating the data was to delete the duplicate values so we only had one entry per person. Unfortunately based on sorting of the data tables, sometimes the year 1 appeared above year 2 and at other times year 2 appeared before year 1. Because none of the data we were interested in are benchmarked to K99 start date, we removed the duplicate values non-specifically. With the dataset we currently have, we would not be able to tell which individuals dropped out (didn’t convert to R00) during the first or second year of the K99. In order to do this we would have to download the raw data from NIH reporter again and curate it again. We may do this in the future but for the purpose of publishing the current manuscript we prefer to focus our efforts on other aspects of the revision.

1. Further down page 3, the authors state that "men typically experience 2-3% greater funding success rates" is ambiguous, as rates are themselves a percentage. So, is it 2-3% greater as in 23% vs 20%, or is it 2-3% greater as in 20.6% vs 20%? Please clarify the language.

Thank you for asking for this clarification. We have updated the text here to reflect that we mean “23% vs 20%”.

1. Metrics such as time to first R01 are compared internally within the study set, which yields interesting insights, but more could be done to benchmark these metrics to non-K99 scientists.

We agree with the reviewer that this would be ideal; however, we feel that it is out of the scope of this manuscript. We may examine this in the future.

1. In the text, several times percentages are being referred to when the figures cited do not show percentages. For example (page 6) 'proportion of awardees that stayed at the same institution declined to about 20% where it has remained consistent (Fig 1B)' - Figure 1B does not show percentages, instead the reader would need to work out from the raw numbers what the pattern of percentages might look like. It's fine (great even) to provide the raw numbers, but would be great to show the percentages as well. This happened for multiple graphs.

Thank you for this comment. We agree that showing the percentage would be beneficial so we have included the percentages in Figure 1 for the conversion rate. We also added a standalone figure panel for the rolling conversion rate for Figure 1. For Figure 4, we have also included a right Y-axis to better indicate the % women.

1. Figure 4 - putting the %women on a 0-250 scale makes it difficult to see the changes in that curve. Please replot it as a separate graph with an appropriate scale (30-50%? 30-70%?)

Thank you for this comment. We have made this edit.

1. Figure 5 - The table appears inconsistent - the Moved/Stayed HR is 1.411 suggesting that moving is better for reducing time to R01, but then Woman/Man is 1.208, so one of these pairs needs to be written in the opposite order to have the table make sense (intended to be listed as 'better/worse'?)

Thank you for noticing this. In the revised manuscript we have re-run the cox proportional hazard model using the R package “survival” and the function “coxph()”. There were minor differences in the hazard ratios using this package instead of Graphpad prism; however, the R package is much more widely used compared to prism for these types of analysis. We present the new data in the table in Figure 5B in the revised manuscript. We now present the “detrimental” cox hazard value for each variable (i.e. 0.7095 for the mobility [moved/stayed]). We also underlined the variable which was detrimental to receiving an R01 award earlier.

1. Figure 5's graph appears strange. All the lines have an appearance of stochasticity but are actually multiples of each other, rising exactly in sync. Are these actually modeled lines? If so, why not instead actually draw the lines based on the real data from the real groups depicted, and give the n for each group?

Thank you for picking this up. The software we originally used to plot the graphs did plot modeled lines instead of the actual data. We have re-run the cox proportional hazard model using the R “survival” package v3.5-5 and the coxph() and survfit() functions. The updated data are in Figure 5 of the revised manuscript.

1. Table 1 should note that each column sums to 100%.

This is a good suggestion. In the revised manuscript, we have added a row to the table to indicate the column total N and %.

1. The authors discuss how k99/R00 grant reviewing process may have to change but the k99 awards also impact the faculty hiring ecosystem as well. There are faculty hiring job ads explicitly requesting or indicating preference towards k99 holders and the results described in this article show that k99 awarding is biased towards particular demographics at select wealthy institutions. Of course, collective/central action is almost always more effective/impactful (especially in shorter time line) than individual elective action. In other words, NIH changing granting patterns would likely work better than encouraging faculty searches to change the weight they give to K99s, because there are many searches and just one NIH. But these are not mutually exclusive and individual action can still help when central action isn't done (if the NIH does not change the k99/R00 grant review process for more inclusive funding and does not increase the number of annual k99 awards hence the annual budget for this award mechanism) and it would be good to have this discussed in the manuscript.

Thank you for this comment and thoughtful insights. We have included additional discussion on this in the final paragraph of the discussion.

Reviewer #2 (Recommendations For The Authors):

Thank you for conducting this important work. On top of some thoughts I have described in the public review (in particular, Chris Pickett's FOIA data on K99/R00 outcomes by applicant race and ethnicity), I only have a few comments for potential improvements to this paper:

1. The comparison of K99-R00 transition rates by gender was interesting. However, I missed the analysis on the K99-R00 transition rates by institution (by type or by top-25 NIH funded institution versus not). I think this analysis may be buried somewhere in the more nuanced descriptions about faculty flows from one institution type to another, but I was not able to locate it. I wonder if the authors could consider dedicating a subsection to specifically describing the transition rate by institution type, creating a table equivalent to Table 2. This section would probably fit best somewhere before the authors dive into the nuances of self-hires and faculty flows.

Said another way: As I was reading, I felt I was missing an answer to a simple question - are there differences in conversion rates by institution type (however you define institution type, as an MSI or non MSI, or top-25 NIH funded versus not)?

Thank you for this suggestion. We have created the table (Table 3 and Table 4) in the revised manuscript. We also made a new figure (now figure 5 in the revised manuscript). This was an interesting way to look at the data and it is very clear that the number of K99 and R00 awards is heavily concentrated within the institutions that have the highest NIH funding. We have added a paragraph in the results in a new section entitled “K99 and R00 awards are concentrated within the highest funded institutions”.

1. Regarding the comparison of HBCUs and Harvard: this analysis was elucidating, but I am not sure if the framing of this analysis as pertaining to "systematically marginalized groups" - see second sentence in the section, "Faculty doctorates differ between Harvard and HBCUs" is appropriate. While it is true that proportionally more faculty at HBCUs are from marginalized groups, there are also many faculty at HBCUs who are from privileged or advantaged backgrounds (e.g., white, men, educated at elite institutions). It would be more accurate to rephrase the second sentence to say something along the lines of, "We sought to examine the rates of funding for those at historically under-funded institutions." I recommend that the authors comb the paper for any other potential places in the text that conflate systemic marginalization with institution type, and rephrase as needed for accuracy.

Thank you for pointing this out. This is an extremely important point and we have removed any instances we could find where we conflate systemically marginalized groups with institution type.

1. I strongly recommend Sugimoto and Larivière (2023)'s new book, Equity for Women in Science, which has an entire section dedicated to previous work investigating how researcher mobility impacts access to resources, collaborations, et cetera (Chapter 5 on Mobility; other chapters on Funding are also relevant but I hone in on Mobility since this is such a key result of this work). I think this chapter would provide significant food-for-thought and background that could strengthen the Discussion section of the paper.

Thank you for this suggestion. We have added some discussion of mobility in the first paragraph of the Discussion.

1. I appreciated the subsection headings that described key results (e.g., "Institutions with the most NIH funding tend to hire K99/R00 awardees from other institutions with the most funding"; "K99/R00 awardee self-hires are more common at institutions with the top NIH funding.") This paper structure made it easier for me to ensure that I was getting the intended takeaway from a figure or section. But partway through the paper, the subheadings changed to being less declarative and therefore less informative (e.g., "Gender of K99/R00 awardees"; "Factors influencing K99/R00 awardee future funding success"). It would be great to rephrase these boilerplate subsection headers to be more declarative, like earlier subsection headings. For example, maybe say "Men receive the majority of K99 awards" or "No gender difference in the rate of conversion from K99 to R00" or something to that effect, depending on what result the authors wish to emphasize.

Thank you for this comment. This is a very good point. We have re-worded the more generic headings in the revised version.

1. Lastly, I would like to share a question that came to my mind that involves an additional analysis, but is work that is (probably) out-of-the-scope of this paper, but could instead be a separate paper or product. Circling back to Chris Pickett's FOIA-ed data on K99/R00 funding outcomes by applicant race and ethnicity (https://web.archive.org/web/20180723171128/http://rescuingbiomedicalresearch.org/blog/examining-distribution-k99r00-awards-race/): Given that Pickett's numbers provide incontrovertible information on the number of awards to various racial and ethnic groups, I wonder if it is possible to use this information as an "answer key" to (1) check the accuracy of an algorithm that assigns race based on name for applications in your analysis but for 2007-2017 period, and, (2) if the results are reasonable, then examine the dataset with race and ethnicity information. Some recent papers performing large-scale bibliometric analyses have applied such algorithms (e.g., see Kozlowski et al. 2022 PNAS Intersectional inequalities in science) and I wonder if they could be useful, or at least tested, here. Again, Pickett's data would serve as the benchmark to see if the algorithm produces numbers that are consistent with the actual funding outcomes; if they're not wildly off, or perhaps accurate for some groups but not others, there might be something here.

This is a really insightful comment. We have discussed whether we could assign ethnicity based on an algorithm and check based on Chris Pickett’s data. We agree that it is beyond the scope of this article, but has potential for future research.

Reviewer #3 (Recommendations For The Authors):

-In the methods section, it would be helpful to provide an overview of the number of universities, departments, and faculty represented in the data analyzed in the study.

Thank you for this comment. We agree with the reviewer. We have added a section to the results discussing the distribution of different types of institutions. We also added Table 3 and Table 4 and a new Figure 5 describing these. Regarding the faculty, we have discussed the demographics of the K99 and R00 awardees as best as we could. We do not have data on which faculty laboratories the K99 awardees were in when they received their awards. This information is not available through NIH reporter.

-I would consider incorporating, or at least citing, Jeff Lockhart and colleagues' recent paper Nature Human Behavior article "Name-based demographic inference and the unequal distribution of misrecognition" about to provide readers with an additional resource and more information about the likelihood of misattribution and general cautionary notes about using gender and race/ethnicity ascription/imputation approaches and tools for research.

Thank you for bringing this reference to our attention. We have incorporated this into the methods section describing our name-based gender determination.

-In the next to last sentence under the final paragraph of the methods section, there looks to be a typo as it should read "K99 or R00," not "K00" as currently written.

Thank you for catching this. We have now corrected it.

-Clarifying some of the data and measures used are necessary to limit confusion and misinterpretations of the study's findings.

Thank you. We have significantly updated the revised manuscript and hope that it is more clear.

-Elaborating more on the gender inequality notable in the Cox proportional hazard model would strengthen the authors' point about persistent gender inequalities within the K99/R00 funding mechanism and pathways. In its current iteration, the findings are somewhat buried by the discussion of institutional differences, but when we look at the findings and the plot associated with the model, we notice that men have more advantages than women in funding and institutional location.

Thank you for highlighting this. This is true and we have elaborated on the gender inequality in the revised version of the manuscript.

-Also for the Cox proportional hazard model, I would consider exploring the inclusion of data that can further clarify the biomedical research infrastructure of institutions. For example, in the conversation about the differences between Princeton and other universities including other Ivies, it's important to note that Princeton does not have a medical school. Moreover, other institutions do not operate or are affiliated with a hospital. Adding more data to the model that can better contextualize the research infrastructure around researchers with NIH awards beyond the size of the NIH portfolio can shed light on possibly other important institutional differences that undergird these inequalities.

Thank you for this comment. We have added additional details about the institutional type; however, to examine whether institutions are attached to a hospital (or are themselves as hospital like MGH etc.) or whether institutions include a medical school may be difficult. We would have to manually code these and then determine whether or not the award recipient was affiliated with a department within that entity or not. We believe that this is a fascinating question but that it is out of the scope of the present manuscript. This is something that we will look into for potential future publications.

-Throughout the manuscript there's usage of "elite" and "prestigious" that are somewhat ambiguous regarding what exactly they are referring to about institutional characteristics. This is a common issue in the literature, but trying to clarify what these terms specifically mean for the current study and checking for consistent usage with limited interchangeability that can add confusion for readers about what is being referred to would give added strength to the conversation provided by the authors.

Thank you for this suggestion. Based on these comments and those by the other reviewers, in the revised version of the manuscript, we have limited the use of “elite” and “prestigious” to describe institutions in order not to perpetuate biases toward certain institutions.

-In relation to the discussion at the end of the manuscript of the longer time to award noted for researchers who stay at the same institutions, another possibility for the disparity could be their reliance for service work (e.g., hiring committees, departmental committees, supporting graduate students through mentoring and/or dissertation committee work, etc.) in their institutions given their knowledge of and experience within it.

Thank you for this suggestion. We have added 2 sentences to the discussion reflecting this possibility.

-Engaging with how STEM professional cultures can perpetuate these funding disparities and related hiring and career outcomes could enhance the contributions of the study. In relation to STEM professional cultures, engaging with the work of Mary Blair-Loy and Erin Cech in their recent book, Misconceiving Merit, could help provide additional insights for readers.

Thank you for these comments. We have incorporated edits to the revised manuscript reflecting the work of Erin Cech and Mary Blair-Loy.

Author Response

Reviewer #1 (Public Review):

Summary:

In this study, the authors showed that activation of RelA and Stat3 in hepatocytes of DSS-treated mice induced CYPs and thereby produced primary bile acids, particularly CDCA, which exacerbated intestinal inflammation.

Strengths:

This study reveals the RelA/Stat3-dependent gene program in the liver influences intestinal homeostasis.

Weaknesses:

Additional evidence will strengthen the conclusion.

1) In Fig. 1C, photos show that phosphorylation of RelA and Stat3 was induced in only a few hepatocytes. The authors conclude that activation of both RelA and Stat3 induces inflammatory pathways. Therefore, the authors should show that phosphorylation of RelA and Stat3 is induced in the same hepatocytes during DSS treatment.

Experiments in progress and data will be submitted in the revised manuscript- Co-staining of pRela and pStat3(727) on treated liver sections.

2) In Fig. 5, the authors treated mice with CDCA intraperitoneally. In this experiment, the concentration of CDCA in the colon of CDCA-treated mice should be shown.

Experiments in progress and data will be submitted in the revised manuscript - Supplementation of CDCA to knockout animals and estimation of CDCA in the colon of DSS treated and untreated animals.

Reviewer #2 (Public Review):

Singh and colleagues employ a methodic approach to reveal the function of the transcription factors Rela and Stat3 in the regulation of the inflammatory response in the intestine.

Strengths of the manuscript include the focus on the function of these transcription factors in hepatocytes and the discovery of their role in the systemic response to experimental colitis. While the systemic response to induce colitis is appreciated, the cellular and molecular mechanisms that drive such systemic response, especially those involving other organs beyond the intestine are an active area of research. As such, this study contributes to this conceptual advance. Additional strengths are the complementary biochemical and metabolomics approaches to describe the activation of these transcription factors in the liver and their requirement - specifically in hepatocytes - for the production of bile acids in response to colitis.

Some weaknesses are noted in the presentation of the data, including a lack of comprehensive representation of findings in all conditions and genotypes tested.

These will be incorporated in the revised version.

Reviewer #3 (Public Review):

Summary:

The authors try to elucidate the molecular mechanisms underlying the intra-organ crosstalks that perpetuate intestinal permeability and inflammation.

Strengths:

This study identifies a hepatocyte-specific rela/stat3 network as a potential therapeutic target for intestinal diseases via the gut-liver axis using both murine models and human samples.

Weaknesses:

1) The mechanism by which DSS administration induces the activation of the Rela and Stat3 pathways and subsequent modification of the bile acid pathway remains clear. As the authors state, intestinal bacteria are one candidate, and this needs to be clarified. I recommend the authors investigate whether gut sterilization by administration of antibiotics or germ-free condition affects 1. the activation of the Rela and Stat3 pathway in the liver by DSS-treated WT mice and 2. the reduction of colitis in DSS-treated relaΔhepstat3Δhep mice.

Experiments in progress and data will be submitted in the revised manuscript - Antibiotic treatment for 2/4 weeks, subsequently mice will be treated with DSS and the Rela and Stat3 phosphorylation will be tested using western blotting.

2) It has not been shown whether DSS administration causes an increase in primary bile acids, represented by CDCA, in the colon of WT mice following activation of the Rela and Stat3 pathways, as demonstrated in Figure 6.

We have demonstrated a enhanced level of CDCA in the colon following DSS treatment in the wild type animals in figure 4B.

3) The implications of these results for IBD treatment, especially in what ways they may lead to therapeutic intervention, need to be discussed.

These will be incorporated in the revised version.

Author Response

We decided to address the comments of the reviewers with additional experiments and modification of the text with the aim of submitting a new version of the report.

We would like to underline that the current study is an extension of the work published in eLife (Atze et al., 2021). For this reason, and in agreement with eLife guidelines, we did not repeat all the background information on the method used to identify PG subunit isotopologues using mass spectrometry.

Reviewer #1 (Public Review):

Summary:

Liang et. al., uses a previously devised full isotope labeling of peptidoglycan followed by mass spec to study the kinetics of Lpp tethering to PG and the hydrolysis of this bond by YafK.

Strengths:

-The labeling and mass spec analysis technique works very well to discern differentially labelled Tri-KR muropeptide containing new and old Lpp and PG.

Weaknesses:

-Only one line of experimentation using mass spec based analysis of labeled PG-Lpp is used to make all conclusions in the paper. The evidence is also not enough to fully deleanate the role of YafK.

Our approach based on heavy isotope labelling and mass spectrometry has the power to identify and kinetically characterize the specific products of the reaction leading to the tethering of Lpp to PG and the hydrolysis of the corresponding bond. We therefore advocate that our experimentation is sufficient to obtain meaningful results without combining other lines of experimentation.

-Only one mutant (YafK) is used to make the conclusion.

The aim of the study is to determine the effect of the hydrolysis of the PG→Lpp bond on the dynamics of the tethering of Lpp to PG. Since YafK is the only enzyme catalyzing this reaction, it is appropriate to compare the wild-type strain to an isogenic yafK deletion mutant. Nonetheless, we carefully consider this comment and will investigate the dynamics of the tethering of Lpp to PG in mutants deficient in the production of the L,D-transpeptidases responsible for tethering Lpp to PG.

-The paper makes a lot of 'implications' with minimal proof to support their hypothesis. Other lines of experimentations must be added to fully delineate their claims.

See our answer to the first comment.

-Time points to analyse Tri-KR isotopologues in Wt (0,10,20,40,60 min) and yafK mutant (0,15, 25, 40, 60 min) are not the same.

The purpose of the experiments is to compare the kinetics of formation and hydrolysis of the PG→Lpp bond in the WT versus ΔyafK strains. Comparison of the kinetics is therefore possible even though the kinetics are not based on the exact same time points. Nonetheless, we will reproduce the kinetics experiment (see also answers to Reviewer 2) and use the same time points in these additional experiments.

-Experiments to define physiological role of YafK are also missing

We will investigate the effect of the yafK deletion on the formation of outer membrane vesicles.

Reviewer #2 (Public Review):

Summary:

The authors of this study have sought to better understand the timing and location of the attachment of the lpp lipoprotein to the peptidoglycan in E. coli, and to determine whether YafK is the hydrolase that cleaves lpp from the peptidoglycan.

Strengths:

The method is relatively straightforward. The authors are able to draw some clear conclusions from their results, that lpp molecules get cleaved from the peptidoglycan and then re-attached, and that YafK is important for that cleavage.

Weaknesses:

However, the authors make a few other conclusions from their data which are harder to understand the logic of, or to feel confident in based on the existing data. They claim that their 5-time point kinetic data indicates that new lpp is not substantially added to lipidII before it is added to the peptidoglycan, and that instead lpp is attached primarily to old peptidoglycan. I believe that this conclusion comes from the comparison of Fig.s 3A and 3C, where it appears that new lpp is added to old peptidoglycan a few minutes before new lpp is added to new peptidoglycan. However, the very small difference in the timing of this result, the minimal number of time points and the complete lack of any presentation of calculated error in any of the data make this conclusion very tenuous. In addition, the authors conclude that lpp is not significantly attached to septal peptidoglycan. The logic behind this conclusion appears to be based on the same data, but the authors do not provide a quantitative model to support this idea.

The reviewer is correct in stating that we claim that Lpp is not substantially added to lipid II before incorporation of the disaccharide-pentapeptide subunit into the expanding PG network. This conclusion is based on the paucity of PG-Lpp covalent adducts containing light PG and Lpp moieties at the earliest time points. To substantiate more thoroughly this finding, we will reproduce the kinetic experiments with more early time points. The paucity of the new→new PG-Lpp isotopologues also implies that Lpp might not be extensively tethered to septal peptidoglycan since the latter is assembled from newly synthesized PG (see our previous publication Atze et al. 2021 and references therein). Quantitatively, septal synthesis roughly accounts for one third of the total PG synthesis. It is therefore expected that tethering of Lpp to septal PG would represent one third of the total number of newly synthesized Lpp molecules tethered to PG. We therefore proposed that the paucity of new→new PG- Lpp isotopologues at early time points of the kinetics implies that Lpp is preferentially tethered to the side wall. This is only one of several conclusions that we reach in the present study and we were very careful in the wording of our results.

-This work will have a moderate impact on the field of research in which the connections between the OM and are being studied in E. coli. Since lpp is not widely conserved in gram negatives, the impact across species is not clear. The authors do not discuss the impact of their work in depth.

We respectfully disagree with this reviewer’s comment. The work reported in this article for E. coli opens the way to the analysis and comparison of the mechanisms of the tethering of proteins to PG in various bacteria. In addition, we would like to stress that the Gram-negative bacteria that produce Lpp-related proteins and tether them to the PG include other major pathogens such as Pseudomonas aeruginosa (DOI: 10.1128/spectrum.05217-22).

Author Response

eLife assessment

The manuscript presents valuable evidence of temporal correlations during specific oscillatory activity between the prefrontal cortex, thalamic nucleus reuniens, and the hippocampus, in naturally sleeping animals. Such correlations represent solid evidence to support the notion that the thalamic nucleus reuniens participates in the hippocampal and prefrontal cortex dialogue subserving memory processes.

Thank you for your assessment.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study, Basha and colleagues aim to test whether the thalamic nucleus reuniens can facilitate the hippocampus/prefrontal cortex coupling during sleep. Considering the importance of sleep in memory consolidation, this study is important to understand the functional interaction between these three majorly involved regions. This work suggests that the thalamic nucleus reuniens has a functional role in synchronizing the hippocampus and prefrontal cortex.

Strengths:

The authors performed recordings in naturally sleeping cats, and analysed the correlation between the main slow wave sleep oscillatory hallmarks: slow waves, spindles, and hippocampal ripples, and with reuniens' neurons firing. They also associated intracellular recordings to assess the reuniens-prefrontal connectivity, and computational models of large networks in which they determined that the coupling of oscillations is modulated by the strength of hippocampal-thalamic connections.

Thank you for your positive evaluation.

Weaknesses:

The authors' main claim is made on slow waves and spindle coupling, which are recorded both in the prefrontal cortex and surprisingly in reuniens. Known to be generated in the cortex by cortico-thalamic mechanisms, the slow waves and spindles recorded in reuniens show no evidence of local generation in the reuniens, which is not anatomically equipped to generate such activities. Until shown differently, these oscillations recorded in reuniens are most likely volume-conducted from nearby cortices. Therefore, such a caveat is a major obstacle to analysing their correlation (in time or frequency domains) with oscillations in other regions.

1. We fully agree with the reviewer that reuniens likely does not generate neither slow waves nor spindles. We do not make such claim, which we clearly stated in the discussion (lines 319-324). We propose that Reuniens neurons mediate different forms of activity. In the model, we introduced MD nucleus only because without MD we were unable to generate spindles. While the slow waves and spindles are generated in other thalamocortical regions, the REU neurons show these rhythms due to long-range projections from these regions to REU as has been shown in the model.

2. Definitely, we cannot exclude some influence of volume conductance on obtained LFP recordings in REU nucleus. However, we show modulation of spiking activity within REU by spindles. Spike modulation cannot be explained by volume conductance but can be explained by either synaptic drive (likely the case here) or some intrinsic neuronal processes (like T-current).

3. In our REU recordings for spike identification we used tetrode recordings. If slow waves and spindles are volume conducted, then slow waves and spindles recorded with tetrodes should have identical shape. Following reviewer comment, we took these recordings and subtracted one channel from another. The difference in signal during slow waves is in the order 0.1 mV. Considering that the distance between electrodes is in the order of 20 um, such a difference in voltage is major and can only be explained by local extracellular currents, likely due to synaptic activities originating in afferent structures.

Finally, the choice of the animal model (cats) is the best suited one, as too few data, particularly anatomical ones regarding reuniens connectivity, are available to support functional results.

1. Thalamus of majority of mammals (definitely primates and carnivores, including cats) contain local circuit interneurons (about 30 % of all neurons). A vast majority of studies in rodents (except LGN nucleus) report either absence or extremally low (i.e. Jager P, Moore G, Calpin P, et al. Dual midbrain and forebrain origins of thalamic inhibitory interneurons. eLife. 2021; 10: e59272.) number of thalamic interneurons. Therefore, studies on other species than rodents are necessary, and bring new information, which is impossible to obtain in rodents.

2. Cats’ brain is much larger than the brain of mice or rats, therefore, the effects of volume conductance from cortex to REU are much smaller, if not negligible. The distance between REU and closest cortical structure (ectosylvian gyrus) in cats is about 15 mm.

3. Indeed, there is much less anatomical data on cats as opposed to rodents. This is why, we performed experiments shown in the figure 1. This figure contains functional anatomy data. Antidromic responses show that recorded structure projects to stimulated structure. Orthodromic responses show that stimulated structure projects to recorded structure.

Reviewer #2 (Public Review):

Summary:

The interplay between the medial prefrontal cortex and ventral hippocampal system is critical for many cognitive processes, including memory and its consolidation over time. A prominent idea in recent research is that this relationship is mediated at least in part by the midline nucleus reuniens with respect to consolidation in particular. Whereas the bulk of evidence has focused on neuroanatomy and the effects of temproary or permanent lesions of the nucleus reuniens, the current work examined the electrophysiology of these three structures and how they inter-relate, especially during sleep, which is anticipated to be critical for consolidation. They provide evidence from intercellular recordings of the bi-directional functional connectivity among these structures. There is an emphasis on the interactions between these regions during sleep, especially slow-wave sleep. They provide evidence, in cats, that cortical slow waves precede reuniens slow waves and hippocampal sharp-wave ripples, which may reflect prefrontal control of the timing of thalamic and hippocampal events, They also find evidence that hippocampal sharp wave ripples trigger thalamic firing and precede the onset of reuniens and medial prefrontal cortex spindles. The authors suggest that the effectiveness of bidirectional connections between the reuniens and the (ventral) CA1 is particularly strong during non-rapid eye movement sleep in the cat. This is a very interesting, complex study on a highly topical subject.

Strengths:

An excellent array of different electrophysiological techniques and analyses are conducted. The temporal relationships described are novel findings that suggest mechanisms behind the interactions between the key regions of interest. These may be of value for future experimental studies to test more directly their association with memory consolidation.

We thank this reviewer for very positive evaluation of our study.

Weaknesses:

Given the complexity and number of findings provided, clearer explanation(s) and organisation that directed the specific value and importance of different findings would improve the paper. Most readers may then find it easier to follow the specific relevance of key approaches and findings and their emphasis. For example, the fact that bidirectional connections exist in the model system is not new per se. How and why the specific findings add to existing literature would have more impact if this information was addressed more directly in the written text and in the figure legends.

Thank you for this comment. In the revised version, we will do our best to simplify presentation and more clearly explain our findings.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Activity has effects on the development of neural circuitry during almost any step of differentiation. In particular during specific time periods of circuit development, so-called critical periods (CP), altered neural activity can induce permanent changes in network excitability. In complex neural networks, it is often difficult to pinpoint the specific network components that are permanently altered by activity, and it often remains unclear how activity is integrated during the CP to set mature network excitability. This study combines electrophysiology with pharmacological and optogenetic manipulation in the Drosophila genetic model system to pinpoint the neural substrate that is influenced by altered activity during a critical period (CP) of larval locomotor circuit development. Moreover, it is then tested whether and how different manipulations of synaptic input are integrated during the CP to tune network excitability.

Strengths:

Based on previous work, during the CP, network activity is increased by feeding the GABA-AR antagonist PTX. This results in permanent network activity changes, as highly convincingly assayed by a prolonged recovery period following induced seizure and by altered intersegmental locomotor network coordination. This is then used to provide two important findings: First, compelling electro- and optophysiological experiments track the site of network change down to the level of single neurons and pre- versus postsynaptic specializations. In short, increased activity during the CP increases both the magnitude of excitatory and inhibitory synaptic transmission to the aCC motoneuron, but excitation is affected more strongly. This results in altered excitation inhibition ratios. Fine electrophysiology shows that excitatory synapse strengthening occurs postsynaptically. High-quality anatomy shows that dendrite size and numbers of synaptic contacts remain unaltered. It is a major accomplishment to track the tuning of network excitability during the CP down to the physiology of specific synapses to identified neurons.

Second, additional experiments with single neuron resolution demonstrate that during the CP different forms of activity manipulation are integrated so that opposing manipulations can rescue altered setpoints. This provides novel insight into how developing neural network excitability is tuned, and it indicates that during the CP, training can rescue the effects of hyperactivity.

Weaknesses:

There are no major weaknesses to the findings presented, but the molecular cause that underlies increased motoneuron postsynaptic responsiveness as well as the mechanism that integrates different forms of activity during the CP remain unknown. It is clear that addressing these experimentally is beyond the scope of this study, but some discussion about different candidates would be helpful.

We discuss likely mechanisms that underpin the increase in postsynaptic responsiveness below (Reviewer #1 (Recommendations For The Authors):, point 2). To address possible mechanisms that integrate different forms of activity we now include a new paragraph in the discussion.

Reviewer #2 (Public Review):

Summary:

In this study, the authors use the tractable Drosophila embryonic/larval motor circuit to determine how manipulations of activity during a critical period (CP) modify the circuit in ways that persist into later developmental stages. Previously, this group demonstrated that manipulations to the aCC/MN-Ib neuron in embryonic stages enhance (or can rescue) susceptibility to seizures at later larval stages. Here, the authors demonstrate that following enhanced excitatory drive (by PTX feeding), the aCC neuron acquires increased sensitivity to cholinergic excitatory transmission, presumably due to increased postsynaptic receptor abundance and/or sensitivity, although this is not clarified. Although locomotion is not altered at later developmental larval stages, the authors suggest there is reduced "robustness" to induced seizures. The second part of the study then goes on to enhance inhibition during the CP in an attempt to counteract the enhanced excitation, and show that many aspects of the CP plasticity are rescued. The authors conclude that "average" E/I activity is integrated during the CP to determine the excitability of the mature locomotor network.

Overall, this study provides compelling mechanistic insight into how a final motor output neuron changes in response to enhanced excitatory drive during a CP to change the functionality of the circuit at later mature developmental stages. The first part of this study is strong, clearly showing the changes in the aCC neuron that result from enhanced excitatory input. This includes very nice electrophysiology and imaging data that assess synaptic function and structure onto aCC neurons from pre-motor inputs resulting from PTX exposure during development. However, the later experiments in Figures 6 and 7 designed to counteract the CP plasticity are somewhat difficult to interpret. In particular, the specificity of the manipulations of the ch neuron intended to counteract the CP plasticity is unclear, given the complexities of how these changes impact the excitability of all neurons during development. It is clear that CP plasticity is largely rescued in later stages, but it is hard to know if downstream or secondary adaptations may be masking the PTX-induced plasticity normally observed. Nonetheless, this study provides an important advance in our understanding of what parameters change during CPs to calibrate network dynamics at later developmental stages.

Reviewer #3 (Public Review):

Summary:

In Hunter, Coulson et al, the authors seek to expand our understanding of how neural activity during developmental critical periods might control the function of the nervous system later in life. To achieve increased excitation, the authors build on their previous results and apply picrotoxin 17-19 hours after egg-laying, which is a critical period of nervous system development. This early enhancement of excitation leads to multiple effects in third-instar larvae, including prolonged recovery from electroshock, increased synchronization of motor neuron networks, and increased AP firing frequency. Using optogenetics and whole-cell patch clamp electrophysiology, the authors elegantly show that picrotoxin-induced over-excitation leads to increased strength of excitatory inputs and not loss of inhibitory inputs. To enhance inhibition, the authors chose an approach that involved the stimulation of mechanosensory neurons; this counteracts picrotoxin-induced signs of increased excitation. This approach to enhancing inhibition requires further control experiments and validation.

Strengths:

• The authors confirm their previous results and show that 17-19 hours after egg laying is a critical period of nervous system development.

• Using Ca2+/Sr2+ substitutions, the authors demonstrate that synaptic connections between A18a  aCC show increased mEPSP amplitudes. The authors show that this aCC input is what is driving enhanced excitation.

• The authors demonstrate that the effects of over-excitation attributed to picrotoxin exposure are generalizable and also occur in bss mutant flies.

Weaknesses:

• The authors build on their previous work and argue that the critical period (17-19h after egg-laying) is a uniquely sensitive period of development. Have the authors already demonstrated that exposure to picrotoxin at L1 or L2 (and even early L3 if experimentally possible) does not lead to changes in induced seizure at L3? This would further the authors' hypothesis of the uniqueness of the 17-19h AEL period. If this has already been established in prior publications, then this needs to be further explained. I do note in Gaicehllo and Baines (2015) that Fig 2E shows the identification of the 17-19h window.

This is a pertinent comment. We now have evidence that activity manipulation (in this instance by increasing temperature, which recapitulates the effect of PTX) is not effective at larval stages (L1 to L3) but remains effective between 17-19hrs AEL. This observation forms part of a separate study where we explore the role of circadian activity on embryonic and larval neuronal development. We include a brief statement to address this comment in the revision (first paragraph of Results).

• Regarding experiments in Fig 2, authors only report changes in AP firing frequency. Can the authors also report other metrics of excitability, including measures of intrinsic excitability with and without picrotoxin exposure (including RMP, Rm)? Was a different amount of current injection needed to evoke stable 5-10 Hz firing with and without picrotoxin? In the representative figure (Fig. 2A), it appears that the baseline firing frequencies are different prior to optogenetic stimulation.

No differences in RM, Rin or capacitance were observed due to PTX. This is now included in the revision along with an explanation that different levels of current injection were used to measure effects to excitatory vs inhibitory synaptic drive. We did not specifically monitor the amount of current required to maintain stable firing.

• The ch-related experiments require further controls and explanation. Regarding experiments in Fig 6, what is the effect of ch neuron stimulation alone on time lag and AP frequency? Can the authors further clarify what is known about connections between aCC and ch neurons? It is difficult for this reviewer to conceptualize how enhancing ch-mediated inhibition would worsen seizures. While the cited study (Carreira-Rosario et al 2021) convincingly shows that inhibition of mechanosensory input leads to excessive spontaneous network activity, has it been shown that the converse - stimulation of ch neurons - indeed enhances network inhibition?

• The interpretation of ch-related experiments is further complicated by the explanation in the Discussion that ch neuron stimulation depolarizes aCC neurons; this seems to undercut the authors' previous explanation that the increased E:I ratio is corrected by enhanced inhibition from ch neurons. The idea that ch neurons are placing neurons in a depolarized refractory state is not substantiated by data in the paper or citations.

To respond to these two points combined: The reviewer is correct in stating that additional experiments will be required to fully understand mechanism. We believe that cholinergic (excitatory) chordotonal input to aCC may be an important component for setting the rhythm of the locomotor CPG. Indeed, it may be that CPG rhythm is a key factor during the CP. Our observations suggest optogenetic stimulation of Ch neurons alone is sufficient to induce large, ~400-, currents that resemble endogenous spontaneous rhythmic currents (SRCs) associated with CPG activity. SRCs occur with a characteristic frequency of ~1Hz, and we have some unpublished data that suggests it is possible to change this frequency using ch stimulation. This data therefore unifies prior work (Carreira-Rosario et al., 2021 description of a brake) with our own (observation that ch depolarize aCC). However, we do not include this speculation in the Discussion because the experiments we have conducted were pilots. They may be expanded upon and included in future work.

• In the Discussion, the authors suggest that enhanced proprioception leading to seizures is reminiscent of neurological conditions. This seems to be an oversimplification. Connecting abnormal proprioception to seizures is quite different from connecting abnormal proprioception to disorders of coordination. This should be revised.

Because this is peripheral to our main study, we have deleted this from the revision.

Reviewer #1 (Recommendations For The Authors):

1. Although the authors have to be commended for the scrutiny with which they pinpoint a site of circuit change, it cannot be excluded that other parts of the circuit also undergo adjustments in response to activity manipulation during the CP, e.g. the membrane properties of the interneurons. This is not a problem but should be discussed.

We agree with this comment and have added the following text to the discussion……’However, we recognise that other parts of the locomotor network may also undergo change due to CP manipulation. The advantage of this system is that most of these elements are now open to specific manipulation through cell-specific genetic drivers’. (Discussion paragraph 3)

1. It is surprising that there is no discussion of the potential molecular cause for the observed increases in postsynaptic responses to SV release from cholinergic neurons. Given that there are no differences in postsynaptic structure, puncta number etc., the subunit composition of the nAChR seems an obvious guess. What is known about the nAChRs subunit composition on aCC, and when during development do the receptors/different subunits become expressed? A paragraph in the discussion on this issue would be highly relevant to the manuscript.

Our own work (unpublished) together with a recent paper from the Littleton lab (https://www.sciencedirect.com/science/article/pii/S0896627323005810?via%3Dihub#mmc2) suggests that aCC expresses the majority, if not all, of the 7 alpha and 3 beta subunits that compromise nAChRs. The situation is further complicated by the fact that these receptors are pentameric and are composed of various subunits – the composition significantly altering channel kinetics. Less is known about expression timelines for each receptor subunit, and certainly not in aCC. We already include the following sentence in the results text……’ A change in the frequency of mini excitatory postsynaptic potentials (mEPSPs, a.k.a. minis) would suggest the adaptation is primarily presynaptic (e.g. increased probability of release), whilst a change in distribution and/or amplitude of minis is more consistent with a mechanism acting postsynaptically (e.g. increased or altered receptor subunits).’ Given that we know next to nothing about the nAChR subunit composition in aCC and how this might change due to CP manipulation, we feel it better not to speculate further. To help the reader, we include the following sentence in the discussion……’The precise mechanism contributing to increased mini amplitude remains to be determined, but a plausible scenario may involve change in cholinergic subunit composition.’ (Discussion paragraph 3)

1. It would be important to provide the p-values for Figures 1B and C, especially because it seems that the inhibition also becomes stronger upon PTX treatment during the CP. There is no statistical testing mentioned, was no test done or was it not significant? It is agreed that the effect size is clearly stronger for the increased excitation than for the increased inhibition, but looking at the data suggests that the effect on excitation is not much more significant than the effect on inhibition.

The reviewer is referring to Fig 2B&C. P values have been added to both main text and to the figure legend.

1. Associated with the point above, in the discussion line 407 and below the authors come back to this point and reason that it is surprising that increased excitation is not compensated for by homeostatic mechanisms. It is concluded that homeostatic compensation brings the system back to a setpoint that is defined during the critical period, but the setpoint is set higher in this case. However, an alternative explanation is that GABA administration during the critical period causes the excitation set point to be too high, but this is then partially counteracted in a homeostatic manner by increasing inhibition. If the p-values in Figures 2B and C are rather similar, this might even be the favorable interpretation.

We believe the reviewer means ‘PTX administration’ and not GABA. This is an interesting idea and one we had not really considered. We address this comment by adding the following text………. ‘Alternatively, whilst the increased inhibition we observe is not statistically significant (p = 0.15), it is close and has a medium effect size (Cohen’s d = 0.78), and thus may be indicative of an attempt by the locomotor network to rebalance activity back towards a genetically pre-determined level. In this regard, it may just not have sufficient range to be able to counter the increase in excitation due to CP manipulation.’ (Discussion paragraph 5)

1. To asses the magnitudes of A18a-mediated excitation and A31k-mediated inhibition to aCC, changes in aCC firing frequency were measured. For this aCC was injected with current to fire at all. However, the current injections were chosen to cause firing at 5-10 Hz. During a crawling burst, aCC fires well above 100Hz (Kadas et al., 2017). Are the effects also visible at such firing frequencies, or at least across different firing frequencies? I am not asking for additional experiments, but maybe the data are there and can be referred to?

Spiking in aCC occurs as burst firing, evoked by cholinergic synaptic drive, that lasts for ~300ms and achieving firing frequencies of between 50-100Hz (Kadas et al., 2017 and our own unpublished data). We did not test for effects to excitation or inhibition at these higher frequencies. We now make this explicit in the discussion by adding the following sentence……’The firing frequencies that we imposed (1-10Hz) are also lower than seen during fictive locomotion (Kadas et al., 2017), which shows burst firing lasting for ~300 ms and achieving spike frequencies of up to 100Hz.’ (Discussion paragraph 3)

1. In Figure 3B some minis are demarked by green arrows and others are not. Were the non-marked ones not included in the analysis, and what were the criteria to mark some and others not? This is particularly important because the cumulative distribution of minis is analyzed in Figure 3D, and this depends crucially on what qualifies as mini and what does not.

All mini’s are marked by green arrows. The events not marked are not mini’s. Drosophila neurons are small and have an unfavourable dendritic structure for recording minis. Thus, we carefully analyse traces by eye taking only events that show very rapid rise times and slower, exponential decay (the typical mini shape). There are, however, other events which are most likely single/multiple channel openings, which due to filtering are rounded. We now include this same trace, greatly expanded, as Fig S1D to show how we identified minis from non-minis.

1. The asynchronous release experiment under Sr2+ seems an elegant way to analyze minis upon optogenetic stimulation of an identified presynaptic cholinergic neuron. I suggest being a little more conservative with the term asynchronous release (or replacing it), which is usually the release of many single vesicles that follow AP-mediated synaptic transmission and has nicely been demonstrated at the Drosophila NMJ (Besse et al., 2007). Also, please show the trace in Figure S2A under Sr2+ at a higher pA magnification, it is really hard to see the minis there.

We have adopted a previously published technique that, in our view, correctly uses the term ‘asynchronous release’. This is not to say that all asynchronous release occurs via the same mechanism. Indeed, the papers that report the technique we use predate Besse 2007. We also expand the trace in Fig S1A (not S2A as wrongly indicated).

Reviewer #2 (Recommendations For The Authors):

1. Can the authors explain what they think is the parameter of "activity" being measured in the locomotor circuit (mainly aCC) during the CP? Is the aCC neuron simply summing (perhaps through a proxy like Ca2+) total excitation/inhibition over time during the CP?

Reviewer #1 also requests that we discuss how activity is ‘measured’ and thus we now include a dedicated paragraph in the discussion to address this concern. Whether aCC sums ‘average’ activity or perhaps is influenced by activity extremes remains uncertain. Our data is consistent with the former but further work is required to validate our conclusion. This work will be published in due course.

Related to understanding this concept, could the authors' silence activity (using Kir2.1, TNT, or BoNT) from each of the monosynaptic premotor inputs in otherwise wildtype and following PTX exposure to determine how the circuit responds when each of the monosynaptic inputs are silenced? This might inform the role they play in instructing how activity is measured over time during the CP.

This is an excellent suggestion and, indeed, we have planned such experiments. Silencing specific neurons, whilst manipulating the CP, may well result in more significant network instability due to the setting of multiple (and physiologically inappropriate) homeostatic set points. Such studies go beyond the scope of the present study and thus we prefer not to speculate at this early stage, but to wait for experimental data.

On a related note, the authors focus on just 2 premotor inputs, presumably due to the availability of specific drivers. But do the authors know how many other inputs (other ACh, Gaba, and glutamate) onto aCC there are, and to what extent do the authors think these are changed in similar or distinct ways? Is it implied that all neurons are similarly altered by the manipulations?

The connectome details the number and types of neurons that directly contact the aCC motoneuron (Zarin et al., 2019). In terms of cholinergic excitors, the results present in Figure 3 suggest that most (all?) inputs are strengthened following embryonic PTX exposure. However, to conclude this would be highly speculative and thus we refrain from doing so in the manuscript. As other single-neuron driver lines become available, such expts will hopefully be possible.

1. If PTX treatment does indeed increase CPG synchronicity, shouldn't there be a readout of this effect on larval locomotion? While the speed of locomotion wasn't significantly impacted, perhaps another parameter was altered.

It is quite possible that other aspects of locomotion are being altered (turning, rearing, etc), but we have not analysed for these more subtle behaviours. Indeed, although not statistically significant, there is a modest reduction in average velocity in larvae derived from PTX-exposed embryos. We see similar reductions in characterised seizure mutants which also show increased synchronicity (Streit et al., 2016).

1. In Figure 2 and elsewhere, what is the baseline level of AP firing rate in each aCC neuron, before optogenetic stimulation? Is this informative about how PTX exposure alters excitability to begin with, perhaps by changing intrinsic excitability.

We now include this data in the relevant results section. Interestingly, following exposure to PTX, basal firing was significantly increased in A18a (excitatory premotor) but not in A31k (inhibitory premotor). This reflects our experiment in which we conclude that excitatory drive to aCC is increased relative to inhibitory synaptic drive. Thus, this measure seemingly validates our conclusion that E:I balance has been altered following activity-manipulation during the CP.

1. Figure 3: The apparent increase in mini amplitude is very small (4.1 vs 4.5 pA); is this physiologically meaningful? Although the authors say the decrease in mini freq is not significant in Fig. 3B after PTX, it does appear rather large, a 40% reduction (5 vs 3 Hz).

We must be guided by statistics in drawing conclusions, but the reader can interpret our data as they wish. Minis measure quantal release and thus to appreciate how small change can, when combined over the many receptors present, influence cell physiology, one needs to compare spiking activity. We show in Fig 2 that such change is sufficient to increase the excitatory synaptic drive provided by the A18a neuron. The seemingly larger reduction in mini frequency is intriguing and may reflect additional change, but without further experiments we cannot draw firm conclusions.

1. The clever vibration assay is a good one to induce the activation of mechanosensory neurons, but the specificity of the changes induced by this is difficult to ascertain. One possibility would be to silence the output of the ch neurons (by expression to tetanus or botulinum toxin) and still put the larvae through the same vibration during the CP to see if the rescue is lost.

We agree that further experiments are required to fully understand underlying mechanism(s). However, we will not be able to complete such follow-on expts in a timely manner and thus, these must wait and form the basis of future studies.

Minor points 1. Typos - there are numerous areas where it seems a comma is used inappropriately (e.g. lines 28, 69, 77, 104, 348, 365, etc). Suggest line editing the final "version of record".

Checked and corrected.

1. It would be of benefit to show the genotypes of the larvae in the various experimental manipulations in the relevant figure legends. This reviewer could not follow exactly how each experiment was done as it was not always clear which driver was being used to express which transgene in what genetic background.

Done

Reviewer #3 (Recommendations For The Authors):

• Please provide sample videos of electroshock-induced seizures (e.g. Fig 1B). Is it clear that the period of immobility after electroshock is a seizure (perhaps defined as hyperactivity originating from the brain)? I acknowledge the Baines group is quite skilled in this technique and perhaps there is a straightforward answer or citation to include.

We refer the reader to Marley and Baines 2011 which contains videos of seizure activity (first paragraph of Results).

• Seizures are generated in the brain and travel to the periphery. Do the authors think it is possible that the peripheral manipulations in this manuscript might be controlling the behavioral readout of seizures without affecting hypersynchronous activity in the brain?

We include the following statement (in methods) to provide our best understanding for how peripheral electroshock induces seizure………. ‘Strong peripheral stimulation likely causes excessive and synchronous synaptic excitation within the CNS resulting in seizure. However, the precise mechanism of this effect remains to be determined.’ Moreover, we feel it unlikely that manipulation of Ch neurons, by vibration, would suppress the effects we observe via peripheral mechanisms. Indeed, the Ch manipulation is limited to the embryonic CP, whilst our seizure assays are recorded many days later at L3.

• How might enhancement of inhibition lead to worsened seizures? Is the enhancement of ch-related inhibition selectively affecting inhibitory circuits, thereby leading to a net increase in excitation?

This is a difficult point to respond to at present. Enhanced inhibition per se might similarly disturb the encoding of an appropriate homeostatic setpoint(s) thus leaving a network open to being destabilized by a strong stimulus. Indeed, we have previously shown that increased inhibition during the CP results in the same effect (seizure) as increasing excitation (Giachello and Baines, 2015). Thus, presuming activation of Ch neurons during the CP translates to increased inhibition, then worsened seizure behaviour is a predictable effect. How this is achieved remains unknown and we prefer not to speculate here.

Author Response

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

We are pleased that Reviewers 1 and 3 have recommended that the revised paper be published.

Reviewer #2

For point A: Their preliminary simulation in 3D looks also nice, although it’s referenced in the discussion but not actually included in manuscript - I would advise adding it even under the mention of preliminary.

We appreciate the reviewer for liking our 3D results and suggesting to include them in the manuscript. However, these are preliminary results of our ongoing work. We are yet to establish the corresponding viscosity results quantitatively in the 3D simulations. Because the relationship between viscosity and relaxation time is not (always) linear in glass forming systems, we hesitate to report our results for publication. We hope to report the new results as part of a separate work.

For point B/C: I see some of the points of the authors - although not all of it made it in the main text. I still have some points that puzzle me. For instance, the authors mention that a single value of viscosity (from Green-Kubo) is ”valid for all time scales and amplitude”. This sounds very surprising to me for a complex fluid even at equilibrium: doesn’t it for instance assume linear response (hence small amplitudes)? Fast vs slow probing of a complex medium should also matter (see refs previously mentioned). Related to this, it’s not clear how can self-propulsion not matter if one would shear the system at a finite time scale, given past work on motility-driven unjamming and the mechanism of the authors from facilitation ( wouldn’t shearing at time scales larger vs smaller than the typical time for given cells to spontaneously rearrange from self-propulsion change drastically the effective complex modulus of the system?)

There might be a slight misunderstanding between the reviewer and us when

we say ‘single value of viscosity is valid for all time-scales and amplitude’. Let us explain this point more carefully. In our problem, we are studying the dynamics of a many body system which is undergoing Brownian dynamics where the fluctuation-dissipation theorem need not be valid (as the friction and the selfpropulsion noise strength are not related via Fluctuation-Dissipation Theorem). Now, for us to use the concepts of linear-response (which in the present study are the Green-Kubo relations for the transport coefficients in terms of timecorrelations functions), we need to show that the within the simulation time, the system has reached state that could be described using an “equilibrium” probability measure. This is the precise reason we calculated the ergodicity measure, which is a way to show that all the phase-space have been sampled uniformly under the given Brownian dynamics. This suggests (does not prove) that the system has attained a stationary probability measure (i.e, near equilibrium) for the value of self-propulsion used. Now for this value of self-propulsion, the Green-Kubo relations hold for ‘any time-scale of the simulations’ so that we can perform a time average over the trajectories of the particles (which is an alias of the stationary probability measure under the values of self-propulsion used). If we change the amplitude of the self-propulsion, we need to again compute the ergodicity measure and show the stationarity of the probability measure. If the system is ergodic with respect to the new self-propulsion, we can again use Green-Kubo for the simulations. Note that we will definitely get a different value of viscosity under the new self-propulsion as the shear-stresses generated will be different but the Green-Kubo holds. If the system is not ergodic, for the self-propulsion with the new amplitude, we cannot use Green-Kubo relations. Also a priori, one cannot say what is a large/small amplitude of self-propulsion because it has to be compared with the intrinsic energy scale, which is encoded in the energy function, which is difficult to say without explicit calculations.

This is what we meant when we said, ‘single value of viscosity is valid for all time-scales and amplitude’. It is valid for time-scales of the simulations for a given amplitude of self-propulsion only if the system is ergodic. Note that if the system is not ergodic, then the results of Ref. [14] (in the main text) could be questioned on theoretical grounds, because they were analyzed using 3 the equilibrium rigidity percolation theory. Nevertheless, the authors of Ref. [14] showed that equilibrium phase transition theory works in tissues. For these reasons, we have been, just like the Reviewer, puzzled that equilibrium ideas appear to be valid in the cell system. Additional theoretical work has to be done to clarify these links in tissues. Although this is not the last word, we hope this clarifies our view point.

For point D: I agree with the simplicity argument, although the added sentence from the discussion “Furthermore, the physics of the dynamics in glass forming materials does not change in systems with and without attractive forces” seems a bit strong given works like Lois et al., PRL, 2008 or Koeze et al, PRL, 2018 finding fundamentally different physics of jamming with or without adhesion. In the two cited papers the authors only consider equilibrium transitions in systems with attraction using computer simulations. Apparently, jamming properties depend on the strength of attraction. There are no attempts to characterize the dynamics, the focus of our work.

What we meant is that any universal relations, such as the Vogel-FulcherTammann relation, would still be valid. Of course, non-universal quantities such as glass transition temperature Tg or fragility will change. In our case, changing the adhesion strength would change ϕS, and the parameters in the VFT. However, our contention is that the overall finding that increase in viscosity followed by saturation is unlikely to change. We have added some clarifying statements in the manuscript to make this clear.