eLife assessment

The authors made an important finding that CD29+/CD56+ progenitor cells isolated from human muscles have the potential to differentiate to tendons in vitro and in vivo. The author's approach to testing the tenogenesis of the CD29+/CD56+ progenitors is solid, and the conclusion is supported by enough evidence with minor flaws. This work will be of interest to the population who need tendon regeneration from their injury.

Reviewer #1 (Public Review):

Through a combination of in vitro and in vivo analyses, the authors demonstrate that CD56/CD29 positive progenitor cells from human muscle can be driven towards muscle or tendon fate in vitro and are able to contribute to muscle and tendon fates following transplantation in injured mice. This is in contrast to Pax7-lineage cells from mice which do not contribute to tendon repair in vivo. While the data strongly support that a subset of cells captured by this sorting strategy has tenogenic potential, their claims of progenitor bi-potency are not fully supported by the data as currently presented.

As discussed below, some aspects of the data analysis and sample preparation are incomplete and should be clarified to fully support the claims of the paper.

For the colony analysis, it is unclear from the methods and main text whether the initial individual sorted colonies were split and subject to different conditions to support the claim of bi-potency. The finding that 40% of colonies displayed tenogenic differentiation, may instead suggest heterogeneity of the sorted progenitor population. The methods as currently described, suggest that two different plates were subject to different induction conditions. It is therefore difficult to assess the strength of the claim of bi-potency.

This group uses the well-established CD56+/CD29+ sorting strategy to isolate muscle progenitor cells, however recent work has identified transcriptional heterogeneity within these human satellite cells (ie Barruet et al, eLife 2020). Given that they identify a tenocyte population in their human muscle biopsy in Figure 1a, it is critical to understand the heterogeneity contained within the population of human progenitors captured by the authors' FACS strategy and whether tenocytes contained within the muscle biopsy are also CD56+/CD29+.

The bulk RNA sequencing data presented in Figure 3 to contrast the expression of progenitor cells under different differentiation conditions are not sufficiently convincing. In particular, it is unclear whether more than one sample was used for the RNAseq analyses shown in Figure 3. The volcano plots have many genes aligned on distinct curves suggesting that there are few replicates or low expression. There is also a concern that the sorted cells may contain tenocytes as tendon genes SCX, MKX, and THBS4 were among the genes upregulated in the myogenic differentiation conditions (shown in Figure 3b).

eLife assessment

In this potentially important study, the authors conducted atomistic simulations to probe the salt-dependent phase separation of the low-complexity domain of hnRN-PA1 (A1-LCD). The authors have identified both direct and indirect mechanisms of salt modulation, provided explanations for four distinct classes of salt dependence, and proposed a model for predicting protein properties from amino acid composition. There is a range of opinions regarding the strength of evidence, with some considering the evidence as incomplete due to the limitations in the length and complexity of the atomistic MD simulations. The work should be put into a better context in relation to previous studies of salt effects on protein phase separation.

Reviewer #1 (Public Review):

Summary:

The authors examined the salt-dependent phase separation of the low-complexity domain of hnRN-PA1 (A1-LCD). Using all-atom molecular dynamics simulations, they identified four distinct classes of salt dependence in the phase separation of intrinsically disordered proteins (IDPs), which can be predicted based on their amino acid composition. However, the simulations and analysis, in their current form, are inadequate and incomplete.

Strengths:

The authors attempt to unravel the mechanistic insights into the interplay between salt and protein phase separation, which is important given the complex behavior of salt effects on this process. Their effort to correlate the influence of salt on the low-complexity domain of hnRNPA1 (A1-LCD) with a range of other proteins known to undergo salt-dependent phase separation is an interesting and valuable topic.

Weaknesses:

(1) The simulations performed are not sufficiently long (Figure 2A) to accurately comment on phase separation behavior. The simulations do not appear to have converged well, indicating that the system has not reached a steady state, rendering the analysis of the trajectories unreliable.

(2) The majority of the data presented shows no significant alteration with changes in salt concentration. However, the authors have based conclusions and made significant comments regarding salt activities. The absence of error bars in the data representation raises questions about its reliability. Additionally, the manuscript lacks sufficient scientific details of the calculations.

(3) In Figures 2B and 2C, the changes in the radius of gyration and the number of contacts do not display significant variations with changes in salt concentration. The change in the radius of gyration with salt concentration is less than 1 Å, and the number of contacts does not change by at least 1. The authors' conclusions based on these minor changes seem unfounded.

Reviewer #2 (Public Review):

This is an interesting computational study addressing how salt affects the assembly of biomolecular condensates. The simulation data are valuable as they provide a degree of atomistic details regarding how small salt ions modulate interactions among intrinsically disordered proteins with charged residues, namely via Debye-like screening that weakens the effective electrostatic interactions among the polymers, or through bridging interactions that allow interactions between like charges from different polymer chains to become effectively attractive (as illustrated, e.g., by the radial distribution functions in Supplementary Information). However, this manuscript has several shortcomings: (i) Connotations of the manuscript notwithstanding, many of the authors' concepts about salt effects on biomolecular condensates have been put forth by theoretical models, at least back in 2020 and even earlier. Those earlier works afford extensive information such as considerations of salt concentrations inside and outside the condensate (tie-lines). But the authors do not appear to be aware of this body of prior works and therefore missed the opportunity to build on these previous advances and put the present work with its complementary advantages in structural details in the proper context. (ii) There are significant experimental findings regarding salt effects on condensate formation [which have been modeled more recently] that predate the A1-LCD system (ref.19) addressed by the present manuscript. This information should be included, e.g., in Table 1, for sound scholarship and completeness. (iii) The strengths and limitations of the authors' approach vis-à-vis other theoretical approaches should be discussed with some degree of thoroughness (e.g., how the smallness of the authors' simulation system may affect the nature of the "phase transition" and the information that can be gathered regarding salt concentration inside vs. outside the "condensate" etc.). Accordingly, this manuscript should be revised to address the following. In particular, the discussion in the manuscript should be significantly expanded by including references mentioned below as well as other references pertinent to the issues raised.

(1) The ability to use atomistic models to address the questions at hand is a strength of the present work. However, presumably because of the computational cost of such models, the "phase-separated" "condensates" in this manuscript are extremely small (only 8 chains). An inspection of Fig.1 indicates that while the high-salt configuration (snapshot, bottom right) is more compact and droplet-like than the low-salt configuration (top right), it is not clear that the 50 mM NaCl configuration can reasonably correspond to a dilute or homogeneous phase (without phase separation) or just a condensate with a lower protein concentration because the chains are still highly associated. One may argue that they become two droplets touching each other (the chains are not fully dispersed throughout the simulation box, unlike in typical coarse-grained simulations of biomolecular phase separation). While it may not be unfair to argue from this observation that the condensed phase is less stable at low salt, this raises critical questions about the adequacy of the approach as a stand-alone source of theoretical information. Accordingly, an informative discussion of the limitation of the authors' approach and comparisons with results from complementary approaches such as analytical theories and coarse-grained molecular dynamics will be instructive-even imperative, especially since such results exist in the literature (please see below).

(2) The aforementioned limitation is reflected by the authors' choice of using Dmax as a sort of phase-separation order parameter. However, no evidence was shown to indicate that Dmax exhibits a two-state-like distribution expected of phase separation. It is also not clear whether a Dmax value corresponding to the linear dimension of the simulation box was ever encountered in the authors' simulated trajectories such that the chains can be reliably considered to be essentially fully dispersed as would be expected for the dilute phase. Moreover, as the authors have noted in the second paragraph of the Results, the variation of Dmax with simulation time does not show a monotonic rank order with salt concentration. The authors' explanation is equivalent to stipulating that the simulation system has not fully equilibrated, inevitably casting doubt on at least some of the conclusions drawn from the simulation data.

(3) With these limitations, is it realistic to estimate possible differences in salt concentration between the dilute and condensed phases in the present work? These features, including tie-lines, were shown to be amenable to analytical theory and coarse-grained molecular dynamics simulation (please see below).

(4) In the comparison in Fig.2B between experimental and simulated radius of gyration as a function of [NaCl], there is an outlier among the simulated radii of gyration at [NaCl] ~ 250 mM. An explanation should be offered.

(5) The phenomenon of no phase separation at zero and low salt and phase separation at higher salt has been observed for the IDP Caprin1 and several of its mutants [Wong et al., J Am Chem Soc 142, 2471-2489 (2020) [https://pubs.acs.org/doi/full/10.1021/jacs.9b12208], see especially Fig.9 of this reference]. This work should be included in the discussion and added to Table 1.

(6) The authors stated in the Introduction that "A unifying understanding of how salt affects the phase separation of IDPs is still lacking". While it is definitely true that much remains to be learned about salt effects on IDP phase separation, the advances that have already been made regarding salt effects on IDP phase separation is more abundant than that conveyed by this narrative. For instance, an analytical theory termed rG-RPA was put forth in 2020 to provide a uniform (unified) treatment of salt, pH, and sequence-charge-pattern effects on polyampholytes and polyelectrolytes (corresponding to the authors' low net charge and high net charge cases). This theory offers a means to predict salt-IDP tie-lines and a comprehensive account of salt effect on polyelectrolytes resulting in a lack of phase separation at extremely low salt and subsequent salt-enhanced phase separation (similar to the case the authors studied here) and in some cases re-entrant phase separation or dissolution [Lin et al., J Chem Phys 152. 045102 (2020) [https://doi.org/10.1063/1.5139661]]. This work is highly relevant and it already provided a conceptual framework for the authors' atomistic results and subsequent discussion. As such, it should definitely be a part of the authors' discussion.

(7) Bridging interactions by small ions resulting in effective attractive interactions among polyelectrolytes leading to their phase separation have been demonstrated computationally by Orkoulas et al., Phys Rev Lett 90, 048303 (2003) [https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.90.048303]. This result should also be included in the discussion.

(8) More recently, the salt-dependent phase separations of Caprin1, its RtoK variants and phosphorylated variant (see item #5 above) were modeled (and rationalized) quite comprehensively using rG-RPA, field-theoretic simulation, and coarse-grained molecular dynamics [Lin et al., arXiv:2401.04873 [https://arxiv.org/abs/2401.04873]], providing additional data supporting a conceptual perspective put forth in Lin et al. J Chem Phys 2020 (e.g., salt-IDP tie-lines, bridging interactions, re-entrance behaviors etc.) as well as in the authors' current manuscript. It will be very helpful to the readers of eLife to include this preprint in the authors' discussion, perhaps as per the authors' discretion-along the manner in which other preprints are referenced and discussed in the current version of the manuscript.

Reviewer #1 (Public Review):

Summary:

This study demonstrates a key role of oxLDL in enhancing Ang II-induced Gq signaling by promoting the AT1/LOX1 receptor complex formation. Importantly, Gq-mediated calcium influx was only observed in LOX1 and AT1 both expressing cells, and AT1-LOX1 interaction aggravated renal damage and dysfunction under the condition of a high-fat diet with Ang II infusion, so this study indicated a new therapeutic potential of AT1-LOX1 receptor complex in CKD patients with dyslipidemia and hypertension.

Strengths:

This study is very exciting and the work is also very detailed, especially regarding the mechanism of LOX1-AT1 receptor interaction and its impact on oxidative stress, fibrosis, and inflammation.

Weaknesses:

The direct evidence for the interaction between AT1 and LOX1 receptors in cell membrane localization is relatively weak. Here I raise some questions that may further improve the study.

Major points:

(1) The authors hypothesized that in the interaction of AT1/LOX1 receptor complex in response to ox-LDL and AngII, there should be strong evidence of fluorescence detection of colocalization for these two membrane receptors, both in vivo and in vitro. Although the video evidence for AT1 internalization upon complex activation is shown in Figure S1, the more important evidence should be membrane interaction and enhanced signal of intracellular calcium influx.

(2) Co-IP experiment should be provided to prove the AT1/LOX1 receptor interaction in response to ox-LDL and AngII in AT1 and LOX1 both expressing cells but not in AT1 only expressing cells.

(3) The authors mentioned that the Gq signaling-mediated calcium influx may change gene expression and cellular characteristics, including EMT and cell proliferation. They also provided evidence that oxidative stress, fibrosis, and inflammation were all enhanced after activating both receptors and inhibiting Gq was effective in reversing these changes. However, single stimulation with ox-LDL or AngII also has strong effects on ROS production, inflammation, and cell EMT, which has been extensively proved by previous studies. So, how to distinguish the biased effect of LOX1 or AT1r alone or the enhanced effect of receptor conformational changes mediated by their receptor interaction? Is there any better evidence to elucidate this point?

(4) How does the interaction between AT1 and LOX1 affect the RAS system and blood pressure? What about the serum levels of rennin, angiotensin, and aldosterone in ND-fed or HFD-fed mice?

Reviewer #2 (Public Review):

Individuals with chronic kidney disease often have dyslipidemia, with the latter both a risk factor for atherosclerotic heart disease and a contributor to progressive kidney disease. Prior studies suggest that oxidized LDL (oxLDL) may cause renal injury through the activation of the LOX1 receptor. The authors had previously reported that LOX1 and AT1 interact to form a complex at the cell surface. In this study, the authors hypothesize that oxLDL, in the setting of angiotensin II, is responsible for driving renal injury by inducing a more pronounced conformational change of the AT1 receptor which results in enhanced Gq signaling.

They go about testing the hypothesis in a set of three studies. In the first set, they engineered CHO cell lines to express AT1R alone, LOX1 in combination with AT1R, or LOX1 with an inactive form of AT1R and indirectly evaluated Gq activity using IP1 and calcium activity as read-outs. They assessed activity after treatment with AngII, oxLDL, or both in combination and found that treatment with both agents resulted in the greatest level of activity, which could be effectively blocked by a Gq inhibitor but not a Gi inhibitor nor a downstream Rho kinase inhibitor targeting G12/13 signaling. These results support their hypothesis, though variability in the level of activation was dramatically inconsistent from experiment to experiment, differing by as much as 20-fold. In contrast, within the experiment, differences between the AngII and AngII/oxLDL treatments, while nominally significant and consistent with their hypothesis, generally were only 10-20%. Another example of unexplained variability can be found in Figures 1g-1j. AngII, at a concentration of 10-12, has no effect on calcium flux in one set of studies (Figure 1g, h) yet has induced calcium activity to a level as great as AngII + oxLDL in another (Figure 1i). The inconsistency of results lessens confidence in the significance of these findings. In other studies with the LOX1-CHO line, they tested for conformational change by transducing AT1 biosensors previously shown to respond to AngII and found that one of them in fact showed enhanced BRET in the setting of oxLDL and AngII compared to AngII alone, which was blocked by an antibody to AT1R. The result is supportive of their conclusions. Limiting enthusiasm for these results is the fact that there isn't a good explanation as to why only 1 sensor showed a difference, and the study should have included a non-specific antibody to control for non-specific effects.

The authors then repeated similar studies using publicly available rat kidney epithelial and fibroblast cell lines that have an endogenous expression of AT1R and LOX1. In these studies, oxLDL in combination with AngiI also enhanced Gq signaling, while knocking down either AT1R or LOX1, and treatment with inhibitors of Gq and AT1R blocked the effects. Like the prior set of studies, however, the effects are very modest and there was significant inter-experimental variability, reducing confidence in the significance of the findings. The authors then tested for evidence that the enhanced Gq signaling could result in renal injury by comparing qPCR results for target genes. While the results show some changes, their significance is difficult to assess. A more global assessment of gene expression patterns would have been more appropriate. In parallel with the transcriptional studies, they tested for evidence of epithelial-mesenchymal transition (EMT) using a single protein marker (alpha-smooth muscle actin) and found that its expression increased significantly in cells treated with oxLDL and AngII, which was blocked by inhibition of Gq inhibition and AT1R. While the data are sound, their significance is also unclear since EMT is a highly controversial cell culture phenomenon. Compelling in vivo studies have shown that most if not all fibroblasts in the kidney are derived from interstitial cells and not a product of EMT. In the last set of studies using these cell lines, the authors examined the effects of AngII and oxLDL on cell proliferation as assayed using BrdU. These results are puzzling---while the two agents together enhanced proliferation which was effectively blocked by an inhibitor to either AT1R or Gq, silencing of LOX1 had no effect.

The final set of studies looked to test the hypothesis in mice by treating WT and Lox1-KO mice with different doses of AngII and either a normal or high-fat diet (to induce oxLDL formation). The authors found that the combination of high dose AngII and a high-fat diet (HFD) increased markers of renal injury (urinary 8-ohdg and urine albumin) in normal mice compared to mice treated with just AngII or HFD alone, which was blunted in Lox1-KO mice). These results are consistent with their hypothesis. However, there are other aspects of these studies that are either inconsistent or complicating factors that limit the strength of the conclusions. For example, Lox1- KO had no effect on renal injury marker expression in mice treated with low-dose AngII and HFD. It also should be noted that Lox1-KO mice had a lower BP response to AngII, which could have reduced renal injury independent of any effects mediated by the AT1R/LOX1 interaction. Another confounding factor was the significant effect the HFD diet had on body weight. While the groups did not differ based on AngII treatment status, the HFD consistently was associated with lower total body weight, which is unexplained. Next, the authors sought to find more direct evidence of renal injury using qPCR of candidate genes and renal histology. The transcriptional results are difficult to interpret; moreover, there were no significant histologic differences between groups. They conclude the study by showing the pattern of expression of LOX1 and AT1R in the kidney by immunofluorescence and conclude that the proteins overlap in renal tubules and are absent from the glomerulus. Unfortunately, they did not co-stain with any other markers to identify the specific cell types. However, these results are inconsistent with other studies that show AT1R is highly expressed in mesangial cells, renal interstitial cells, near the vascular pole, JG cells, and proximal tubules but generally absent from most other renal tubule segments.

In sum, this study tackles an important clinical issue and provides some in vitro evidence to support a mechanism whereby dyslipidemia could accelerate renal functional decline through activation of the AT1R/LOX1 complex by oxLDL and AngII.

However, a very high degree of variability in the results, modest within-experiment differences, some internal inconsistencies that aren't explained, and the lack of compelling and strongly supportive in vivo results suggest this is still more a hypothesis than an established likely mechanism.

eLife assessment

This manuscript reports important findings on the impact of maternal obesity on oocyte methylation and its transgenerational effects. The evidence presented to substantiate the major claims appears incomplete. This study would be of interest to biologists in the fields of epigenetics and metabolism.

Reviewer #1 (Public Review):

With socioeconomic development, more and more people are obese which is an important reason for sub-fertility and infertility. Maternal obesity reduces oocyte quality which may be a reason for the high risk of metabolic diseases for offspring in adulthood. Yet the underlying mechanisms are not well elucidated. Here the authors examined the effects of maternal obesity on oocyte methylation. Hyper-methylation in oocytes was reported by the authors, and the altered methylation in oocytes may be partially transmitted to F2. The authors further explored the association between the metabolome of serum and the altered methylation in oocytes. The authors identified decreased melatonin. Melatonin is involved in regulating the hyper-methylation of high-fat diet (HFD) oocytes, via increasing the expression of DNMTs which is mediated by the cAMP/PKA/CREB pathway.

Strengths:

This study is interesting and should have significant implications for the understanding of the transgenerational inheritance of GDM in humans.

Weaknesses:

The link between altered DNA methylation and offspring metabolic disorders is not well elucidated; how the altered DNA methylation in oocytes escapes reprogramming in transgenerational inheritance is also unclear.

Reviewer #3 (Public Review):

Summary:

Maternal obesity is a health problem for both pregnant women and their offspring. Previous works including work from this group have shown significant DNA methylation changes for offspring of obese pregnancies in mice. In this manuscript, Chao et al digested the potential mechanisms behind the DNA methylation changes. The major observations of the work include transgenerational DNA methylation changes in offspring of maternal obesity, and metabolites such as methionine and melatonin correlated with the above epigenetic changes. Exogenous melatonin treatment could reverse the effects of obesity. The authors further hypothesized that the linkage may be mediated by the cAMP/PKA/CREB pathway to regulate the expression of DNMTs.

Strengths:

The transgenerational change of DNA methylation following HFD is of great interest for future research to follow. The metabolic treatment that could change the DNA methylation in oocytes is also interesting and has potential relevance to future clinical practice.

Weaknesses:

The HFD oocytes have more 5mC signal based on staining and sequencing (Fig 1A-1F). However, the authors also identified almost equal numbers of hyper- and hypo-DMRs, which raises questions regarding where these hypo-DMRs were located and how to interpret their behaviors and functions. These questions are also critical to address in the following mechanistic dissections as the metabolic treatments may also induce bi-directional changes of DNA methylation. The authors should carefully assess these conflicts to make the conclusions solid.

The transgenerational epigenetic modifications are controversial. Even for F0 offspring under maternal obesity, there were different observations compared to this work (Hou, YJ., et al. Sci Rep, 2016). The authors should discuss the inconsistencies with previous works.

In addition to the above inconsistencies, the DNA methylation analysis in this work was not carefully evaluated. Several previous works were evaluating the DNA methylation in mice oocytes, which showed global methylation levels of around 50% (Shirane K, et al. PLoS Genet, 2013; Wang L., et al, Cell, 2014). In Figure 1E, the overall methylation level is about 23% in control, which is significantly different from previous works. The authors should provide more details regarding the WGBS procedure, including but not limited to sequencing coverage, bisulfite conversion rate, etc.

eLife assessment

This study reports the adhesion G-protein-coupled receptor A3 (ADGRA3) as a possible target for activating adaptive thermogenesis in white and brown adipose tissue. The study provides valuable insights for scientists who study metabolism, obesity, and adipose tissue biology. Meanwhile, the experimental evidence supporting the claim is incomplete, and more rigorous approaches are needed to demonstrate the relevance of this receptor in adipose tissue biology.

Reviewer #1 (Public Review):

Summary:

This article identifies ADGR3 as a candidate GPCR for mediating beige fat development. The authors use human expression data from the Human protein atlas and Gtex databases and combine this with experiments performed in mice and a murine cell line. They refer to a GPCR bioactivity screening tool PRESTO-Salsa, with which it was found that Hesperetin activates ADGR3. From their experiments, authors conclude that Hesperetin activates ADGR3, inducing a Gs-PKA-CREB axis resulting in adipose thermogenesis.

Strengths:

The authors analyze human data from public databases and perform functional studies in mouse models. They identify a new GPCR with a role in the thermogenic activation of adipocytes.

Weaknesses:

(1) Selection of ADGRA3 as a candidate GPCR relevant for mediating beiging in humans:

The authors identify genes upregulated in iBAT compared to iWAT in response to cold, and among these differentially expressed genes, they identify highly expressed GPCRs in human white adipocytes (visceral or subcutaneous). Finally, among these genes, they select a GPCR not previously studied in the literature.

If the authors are interested in beiging, why do they not focus on genes upregulated in iWAT (the depot where beiging is described to occur in mice), comparing thermoneutral to cold-induced genes? I would expect that genes induced in iWAT in response to cold would be extremely relevant targets for beiging. With their strategy, the authors exclude receptors that are induced in the tissue where beiging is actually described to occur.

Furthermore, the authors are comparing genes upregulated in cold in BAT (but not WAT) to highly expressed genes in human white adipocytes during thermoneutrality. Overall, the authors fail to discuss the logic behind their strategy and the obvious limitations of it.

(2) Relevance of ADGRA3 and comparison to established literature:

There has been a lot of literature and discussion about which receptor should be targeted in humans to recruit thermogenic fat. The current article unfortunately does not discuss this literature nor explain how it relates to their findings. For example, O'Mara et al (PMID: 31961826) demonstrated that chronic stimulation with the B3 adrenergic agonist, Mirabegron, resulted in the recruitment of thermogenic fat and improvement in insulin sensitivity and cholesterol. Later, Blondin et al (PMID: 32755608), highlighted the B2 adrenergic receptor as the main activation path of thermogenic fat in humans. There is also a recent report on an agonist activating B2 and B3 simultaneously (PMID: 38796310). Thus, to bring the literature forward, it would be beneficial if the current manuscript compared their identified activation path with the activation of these already established receptors and discussed their findings in relation to previous studies.

In Figures 1d and e, the authors show the expression of ADGRA3 in comparison to the expression of ADRB3. In human brown adipocytes, ADRB2 has been shown to be the main receptor through which adrenergic activation occurs (PMID: 32755608), thus authors should show the relative expression of this gene as well.

(3) Strategy to investigate the role of ADGRA3 in WAT beiging:

Having identified ADGRA3 as their candidate receptor, the authors proceed with investigations of this receptor in mouse models and the murine inguinal adipocyte cell line 3T3.

First of all, in Figure 1D, the authors show a substantially lower expression of ADGRA3 compared to ADRB3. It could thus be argued that a mouse would not be the best model system for studying this receptor. It would be interesting to see data from experiments in human adipocytes. Moreover, if the authors are interested in inducing beiging, why do they show expression in iBAT and not iWAT?

The authors perform in vivo experiments using intraperitoneal injections of shRNA or overexpression CMV-driven vectors and report effects on body temperature and glucose metabolism. It is here important to note that ADGRA3 is not uniquely expressed in adipocytes. A major advantage of databases like the Human Protein Atlas and Gtex, is that they give an overview of the gene expression across tissues and cell types. When looking up ADGRA3 in these databases, it is expressed in subcutaneous and visceral adipocytes. However, other cell types and tissues demonstrate an even higher expression. In the Human protein atlas, the enhanced cell types are astrocytes and hepatocytes. In the Gtex database tissues with the highest expression are Brain, Liver, and Thyroid.

With this information in mind, IP injections for modification of ADGRA3 receptor expression could be expected to affect any of these tissues and cells.

The manuscript report changes body temperature. However, temperature is regulated by the brain and also affected by thyroid activity. Did the authors measure the levels of circulating thyroid hormones? Gene expression changes in the brain? The authors report that Adgra3 overexpression decreased the TG level in serum and liver. The liver could be the primary targeted organ here, and the adipose effects might be secondary. The data would be easier to interpret if authors reported the effects on the liver, thyroid, and brain, and the gene expression across tissues should be discussed in the article.

Finally, the identification of Hesperetin using the PRESTO-Salsa tool, and how specific the effect of Hesperetin is on ADGRA3, is currently unclear. This should be better discussed, and authors should consider measuring the established effects of Hesperetin in their model systems, including apoptosis.

Reviewer #2 (Public Review):

Based on bioinformatics and expression analysis using mouse and human samples, the authors claim that the adhesion G-protein coupled receptor ADGRA3 may be a valuable target for increasing thermogenic activity and metabolic health. Genetic approaches to deplete ADGRA3 expression in vitro resulted in reduced expression of thermogenic genes including Ucp1, reduced basal respiration, and metabolic activity as reflected by reduced glucose uptake and triglyceride accumulation. In line, nanoparticle delivery of shAdgra3 constructs is associated with increased body weight, reduced thermogenic gene expression in white and brown adipose tissue (WAT, BAT), and impaired glucose and insulin tolerance. On the other hand, ADGRA3 overexpression is associated with an improved metabolic profile in vitro and in vivo, which can be explained by increasing the activity of the well-established Gs-PKA-CREB axis. Notably, a computational screen suggested that ADGRA3 is activated by hesperetin. This metabolite is a derivative of the major citrus flavonoid hesperidin and has been described to promote metabolic health. Using appropriate in vitro and in vivo studies, the authors show that hesperetin supplementation is associated with increased thermogenesis, UCP1 levels in WAT and BAT, and improved glucose tolerance, an effect that was attenuated in the absence of ADGRA3 expression.

Overall, the data suggest that ADGRA3 is a constitutively active Gs-coupled receptor that improves metabolism by activating adaptive thermogenesis in WAT and BAT. The conclusions of the paper are partly supported by the data, but some experimental approaches need further clarification.

(1) The in vivo approaches to modulate Adgra3 expression in mice are carried out using non-targeted nanoparticle-based approaches. The authors do not provide details of the composition of the nanomaterials, but it is highly likely that other metabolically active organs such as the liver are targeted. This is critical because Adgre3 is expressed in many organs, including the liver, adrenal glands, and gastrointestinal system. Therefore, many of the observed metabolic effects could be indirect, for example by modulating bile acids or corticosterone levels. Consistent with this, after digestion in the gastrointestinal tract, hesperetin is rapidly metabolized in intestinal and liver cells. Thus, hesperetin levels in the systemic circulation are likely to be insufficient to activate Adgra3 in thermogenic adipocytes/precursors. Overall, the authors need to repeat the key metabolic experiments in adipose-specific Adgra3 knockout/overexpression models to validate the reliability of the in vivo results. In addition, to validate the relevance of hesperetin supplementation for adaptive thermogenesis in BAT and WAT vivo, the levels of hesperetin present in the systemic circulation should be quantified.

(2) Standard measurements for energy balance are not presented. Quantitative data on energy expenditure, e.g. by indirect calorimetry, and food intake are missing and need to be included to validate the authors' claims.

(3) The thermographic images used to determine the BAT temperature are not very convincing. The distance and angle between the thermal camera and the BAT have a significant effect on the determination of the temperature, which is not taken into account, at least in the images presented.

(4) The 3T3-L1 cell line is not an adequate cell culture model to study thermogenic adipocyte differentiation. To validate their results, the key experiments showing that ADGRA3 expression modulates thermogenic marker expression in a hesperetin-dependent manner need to be performed in a reliable model, e.g. primary murine adipocytes.

(5) The experimental setup only allows the measurement of basal cellular respiration. More advanced approaches are needed to define the contribution of ADGRA3 versus classical adrenergic receptors to UCP1-dependent thermogenesis.

Reviewer #3 (Public Review):

Summary:

The manuscript by Zhao et al. explored the function of adhesion G protein-coupled receptor A3 (ADGRA3) in thermogenic fat biology.

Strengths:

Through both in vivo and in vitro studies, the authors found that the gain function of ADGRA3 leads to browning of white fat and ameliorates insulin resistance.

Weaknesses:

There are several lines of weak methodologies such as using 3T3-L1 adipocytes and intraperitoneal(i.p.) injection of virus. Moreover, as the authors stated that ADGRA3 is constitutively active, how could the authors then identify a chemical ligand?

Recommendations:

(1) Primary cultured cells should be used to perform gain and loss function analysis of ADGRA3, instead of using 3T3-L1. It is impossible to detect Ucp1 expression in 3T3-L1 cells.

(2) For virus treatment, the authors should consider performing local tissue injection, rather than IP injection. If it is IP injection, have the authors checked other tissues to validate whether the phenotype is fat-specific?

(3) The authors should clarify how constitutively active GPCR needs further ligands.

eLife assessment

In an important fMRI study with an elegant experimental design and rigorous cross-decoding analyses, this work shows a solid dissociation between two parietal regions in visually processing actions. Specifically, aIPL is found to be sensitive to the causal effects of observed actions, while SPL is sensitive to the patterns of body motion involved in those actions. Additional analysis and explanation would help to determine the strength of evidence and the mechanistic underpinnings would benefit from closer consideration. Nevertheless, the work will be of broad interest to cognitive neuroscientists, particularly vision and action researchers.

Reviewer #1 (Public Review):

Summary:

The authors report a study aimed at understanding the brain's representations of viewed actions, with a particular aim to distinguish regions that encode observed body movements, from those that encode the effects of actions on objects. They adopt a cross-decoding multivariate fMRI approach, scanning adult observers who viewed full-cue actions, pantomimes of those actions, minimal skeletal depictions of those actions, and abstract animations that captured analogous effects to those actions. Decoding across different pairs of these actions allowed the authors to pull out the contributions of different action features in a given region's representation. The main hypothesis, which was largely confirmed, was that the superior parietal lobe (SPL) more strongly encodes movements of the body, whereas the anterior inferior parietal lobe (aIPL) codes for action effects of outcomes. Specifically, region of interest analyses showed dissociations in the successful cross-decoding of action category across full-cue and skeletal or abstract depictions. Their analyses also highlight the importance of the lateral occipito-temporal cortex (LOTC) in coding action effects. They also find some preliminary evidence about the organisation of action kinds in the regions examined.

Strengths:

The paper is well-written, and it addresses a topic of emerging interest where social vision and intuitive physics intersect. The use of cross-decoding to examine actions and their effects across four different stimulus formats is a strength of the study. Likewise, the a priori identification of regions of interest (supplemented by additional full-brain analyses) is a strength.

Weaknesses:

I found that the main limitation of the article was in the underpinning theoretical reasoning. The authors appeal to the idea of "action effect structures (AES)", as an abstract representation of the consequences of an action that does not specify (as I understand it) the exact means by which that effect is caused, nor the specific objects involved. This concept has some face validity, but it is not developed very fully in the paper, rather simply asserted. The authors make the claim that "The identification of action effect structure representations in aIPL has implications for theories of action understanding" but it would have been nice to hear more about what those theoretical implications are. More generally, I was not very clear on the direction of the claim here. Is there independent evidence for AES (if so, what is it?) and this study tests the following prediction, that AES should be associated with a specific brain region that does not also code other action properties such as body movements? Or, is the idea that this finding -- that there is a brain region that is sensitive to outcomes more than movements -- is the key new evidence for AES?

On a more specific but still important point, I was not always clear that the significant, but numerically rather small, decoding effects are sufficient to support strong claims about what is encoded or represented in a region. This concern of course applies to many multivariate decoding neuroimaging studies. In this instance, I wondered specifically whether the decoding effects necessarily reflected fully five-way distinction amongst the action kinds, or instead (for example) a significantly different pattern evoked by one action compared to all of the other four (which in turn might be similar). This concern is partly increased by the confusion matrices that are presented in the supplementary materials, which don't necessarily convey a strong classification amongst action kinds. The cluster analyses are interesting and appear to be somewhat regular over the different regions, which helps. However: it is hard to assess these findings statistically, and it may be that similar clusters would be found in early visual areas too.

Reviewer #2 (Public Review):

Summary:

This study uses an elegant design, using cross-decoding of multivariate fMRI patterns across different types of stimuli, to convincingly show a functional dissociation between two sub-regions of the parietal cortex, the anterior inferior parietal lobe (aIPL) and superior parietal lobe (SPL) in visually processing actions. Specifically, aIPL is found to be sensitive to the causal effects of observed actions (e.g. whether an action causes an object to compress or to break into two parts), and SPL to the motion patterns of the body in executing those actions.

To show this, the authors assess how well linear classifiers trained to distinguish fMRI patterns of response to actions in one stimulus type can generalize to another stimulus type. They choose stimulus types that abstract away specific dimensions of interest. To reveal sensitivity to the causal effects of actions, regardless of low-level details or motion patterns, they use abstract animations that depict a particular kind of object manipulation: e.g. breaking, hitting, or squashing an object. To reveal sensitivity to motion patterns, independently of causal effects on objects, they use point-light displays (PLDs) of figures performing the same actions. Finally, full videos of actors performing actions are used as the stimuli providing the most complete, and naturalistic information. Pantomime videos, with actors mimicking the execution of an action without visible objects, are used as an intermediate condition providing more cues than PLDs but less than real action videos (e.g. the hands are visible, unlike in PLDs, but the object is absent and has to be inferred). By training classifiers on animations, and testing their generalization to full-action videos, the classifiers' sensitivity to the causal effect of actions, independently of visual appearance, can be assessed. By training them on PLDs and testing them on videos, their sensitivity to motion patterns, independent of the causal effect of actions, can be assessed, as PLDs contain no information about an action's effect on objects.

These analyses reveal that aIPL can generalize between animations and videos, indicating that it is sensitive to action effects. Conversely, SPL is found to generalize between PLDs and videos, showing that it is more sensitive to motion patterns. A searchlight analysis confirms this pattern of results, particularly showing that action-animation decoding is specific to right aIPL, and revealing an additional cluster in LOTC, which is included in subsequent analyses. Action-PLD decoding is more widespread across the whole action observation network.

This study provides a valuable contribution to the understanding of functional specialization in the action observation network. It uses an original and robust experimental design to provide convincing evidence that understanding the causal effects of actions is a meaningful component of visual action processing and that it is specifically localized in aIPL and LOTC.

Strengths:

The authors cleverly managed to isolate specific aspects of real-world actions (causal effects, motion patterns) in an elegant experimental design, and by testing generalization across different stimulus types rather than within-category decoding performance, they show results that are convincing and readily interpretable. Moreover, they clearly took great care to eliminate potential confounds in their experimental design (for example, by carefully ordering scanning sessions by increasing realism, such that the participants could not associate animation with the corresponding real-world action), and to increase stimulus diversity for different stimulus types. They also carefully examine their own analysis pipeline, and transparently expose it to the reader (for example, by showing asymmetries across decoding directions in Figure S3). Overall, this is an extremely careful and robust paper.

Weaknesses:

I list several ways in which the paper could be improved below. More than 'weaknesses', these are either ambiguities in the exact claims made, or points that could be strengthened by additional analyses. I don't believe any of the claims or analyses presented in the paper show any strong weaknesses, problematic confounds, or anything that requires revising the claims substantially.

(1) Functional specialization claims: throughout the paper, it is not clear what the exact claims of functional specialization are. While, as can be seen in Figure 3A, the difference between action-animation cross-decoding is significantly higher in aIPL, decoding performance is also above chance in right SPL, although this is not a strong effect. More importantly, action-PLD cross-decoding is robustly above chance in both right and left aIPL, implying that this region is sensitive to motion patterns as well as causal effects. I am not questioning that the difference between the two ROIs exists - that is very convincingly shown. But sentences such as "distinct neural systems for the processing of observed body movements in SPL and the effect they induce in aIPL" (lines 111-112, Introduction) and "aIPL encodes abstract representations of action effect structures independently of motion and object identity" (lines 127-128, Introduction) do not seem fully justified when action-PLD cross-decoding is overall stronger than action-animation cross-decoding in aIPL. Is the claim, then, that in addition to being sensitive to motion patterns, aIPL contains a neural code for abstracted causal effects, e.g. involving a separate neural subpopulation or a different coding scheme? Moreover, if sensitivity to motion patterns is not specific to SPL, but can be found in a broad network of areas (including aIPL itself), can it really be claimed that this area plays a specific role, similar to the specific role of aIPL in encoding causal effects? There is indeed, as can be seen in Figure 3A, a difference between action-PLD decoding in SPL and aIPL, but based on the searchlight map shown in Figure 3B I would guess that a similar difference would be found by comparing aIPL to several other regions. The authors should clarify these ambiguities.

(2) Causal effect information in PLDs: the reasoning behind the use of PLD stimuli is to have a condition that isolates motion patterns from the causal effects of actions. However, it is not clear whether PLDs really contain as little information about action effects as claimed. Cross-decoding between animations and PLDs is significant in both aIPL and LOTC, as shown in Figure 4. This indicates that PLDs do contain some information about action effects. This could also be tested behaviorally by asking participants to assign PLDs to the correct action category. In general, disentangling the roles of motion patterns and implied causal effects in driving action-PLD cross-decoding (which is the main dependent variable in the paper) would strengthen the paper's message. For example, it is possible that the strong action-PLD cross-decoding observed in aIPL relies on a substantially different encoding from, say, SPL, an encoding that perhaps reflects causal effects more than motion patterns. One way to exploratively assess this would be to integrate the clustering analysis shown in Figure S1 with a more complete picture, including animation-PLD and action-PLD decoding in aIPL.

(3) Nature of the motion representations: it is not clear what the nature of the putatively motion-driven representation driving action-PLD cross-decoding is. While, as you note in the Introduction, other regions such as the superior temporal sulcus have been extensively studied, with the understanding that they are part of a feedforward network of areas analyzing increasingly complex motion patterns (e.g. Riese & Poggio, Nature Reviews Neuroscience 2003), it doesn't seem like the way in which SPL represents these stimuli are similarly well-understood. While the action-PLD cross-decoding shown here is a convincing additional piece of evidence for a motion-based representation in SPL, an interesting additional analysis would be to compare, for example, RDMs of different actions in this region with explicit computational models. These could be, for example, classic motion energy models inspired by the response characteristics of regions such as V5/MT, which have been shown to predict cortical responses and psychophysical performance both for natural videos (e.g. Nishimoto et al., Current Biology 2011) and PLDs (Casile & Giese Journal of Vision 2005). A similar cross-decoding analysis between videos and PLDs as that conducted on the fMRI patterns could be done on these models' features, obtaining RDMs that could directly be compared with those from SPL. This would be a very informative analysis that could enrich our knowledge of a relatively unexplored region in action recognition. Please note, however, that action recognition is not my field of expertise, so it is possible that there are practical difficulties in conducting such an analysis that I am not aware of. In this case, I kindly ask the authors to explain what these difficulties could be.

(4) Clustering analysis: I found the clustering analysis shown in Figure S1 very clever and informative. However, there are two things that I think the authors should clarify. First, it's not clear whether the three categories of object change were inferred post-hoc from the data or determined beforehand. It is completely fine if these were just inferred post-hoc, I just believe this ambiguity should be clarified explicitly. Second, while action-anim decoding in aIPL and LOTC looks like it is consistently clustered, the clustering of action-PLD decoding in SPL and LOTC looks less reliable. The authors interpret this clustering as corresponding to the manual vs. bimanual distinction, but for example "drink" (a unimanual action) is grouped with "break" and "squash" (bimanual actions) in left SPL and grouped entirely separately from the unimanual and bimanual clusters in left LOTC. Statistically testing the robustness of these clusters would help clarify whether it is the case that action-PLD in SPL and LOTC has no semantically interpretable organizing principle, as might be the case for a representation based entirely on motion pattern, or rather that it is a different organizing principle from action-anim, such as the manual vs. bimanual distinction proposed by the authors. I don't have much experience with statistical testing of clustering analyses, but I think a permutation-based approach, wherein a measure of cluster robustness, such as the Silhouette score, is computed for the clusters found in the data and compared to a null distribution of such measures obtained by permuting the data labels, should be feasible. In a quick literature search, I have found several papers describing similar approaches: e.g. Hennig (2007), "Cluster-wise assessment of cluster stability"; Tibshirani et al. (2001) "Estimating the Number of Clusters in a Data Set Via the Gap Statistic". These are just pointers to potentially useful approaches, the authors are much better qualified to pick the most appropriate and convenient method. However, I do think such a statistical test would strengthen the clustering analysis shown here. With this statistical test, and the more exhaustive exposition of results I suggested in point 2 above (e.g. including animation-PLD and action-PLD decoding in aIPL), I believe the clustering analysis could even be moved to the main text and occupy a more prominent position in the paper.

(5) ROI selection: this is a minor point, related to the method used for assigning voxels to a specific ROI. In the description in the Methods (page 16, lines 514-24), the authors mention using the MNI coordinates of the center locations of Brodmann areas. Does this mean that then they extracted a sphere around this location, or did they use a mask based on the entire Brodmann area? The latter approach is what I'm most familiar with, so if the authors chose to use a sphere instead, could they clarify why? Or, if they did use the entire Brodmann area as a mask, and not just its center coordinates, this should be made clearer in the text.

eLife assessment

The authors have reported an important study in which they use a double-blind design to explore pharmacological manipulations in the context of a behavioral task. Despite a relatively small sample size, the findings are solid and motivate future explanations of the mechanism underlying their observations. The findings could be further strengthened by addressing some remaining concerns that relate to preprocessing, statistical details, and possible ocular artifacts.

Reviewer #1 (Public Review):

Summary:

The authors used a motivated saccade task with distractors to measure response vigor and reaction time (RT) in healthy human males under placebo or muscarinic antagonism. They also simultaneously recorded neural activity using EEG with event-related potential (ERP) focused analyses. This study provides evidence that the muscarinic antagonist Trihexyphenidyl (THP) modulates the motivational effects of reward on both saccade velocity and RT, and also increases the distractibility of participants. The study also examined the correlational relationships between reaction time and vigor and manipulations (THP, incentives) with components of the EEG-derived ERPs. While an interesting correlation structure emerged from the analyses relating the ERP biomarkers to behavior, it is unclear how these potentially epiphenomenal biomarkers relate to relevant underlying neurophysiology.

Strengths:

This study is a logical translational extension from preclinical findings of cholinergic modulation of motivation and vigor and the CNV biomarker to a normative human population, utilizing a placebo-controlled, double-blind approach.

While framed in the context of Parkinson's disease where cholinergic medications can be used, the authors do a good job in the discussion describing the limitations in generalizing their findings obtained in a normative and non-age-matched cohort to an aged PD patient population.

The exploratory analyses suggest alternative brain targets and/or ERP components that relate to the behavior and manipulations tested. These will need to be further validated in an adequately powered study. Once validated, the most relevant biomarkers could be assessed in a more clinically relevant population.

Weaknesses:

The relatively weak correlations between the main experimental outcomes provide unclear insight into the neural mechanisms by which the manipulations lead to behavioral manifestations outside the context of the ERP. It would have been interesting to evaluate how other quantifications of the EEG signal through time-frequency analyses relate to the behavioral outcomes and manipulations.

The ERP correlations to relevant behavioral outcomes were not consistent across manipulations demonstrating they are not reliable biomarkers to behavior but do suggest that multiple underlying mechanisms can give rise to the same changes in the ERP-based biomarkers and lead to different behavioral outcomes.

Reviewer #3 (Public Review):

Summary:

Grogan et al examine a role for muscarinic receptor activation in action vigor in a saccadic system. This work is motivated by a strong literature linking dopamine to vigor, and some animal studies suggesting that ACH might modulate these effects, and is important because patient populations with symptoms related to reduced vigor are prescribed muscarinic antagonists. The authors use a motivated saccade task with distractors to measure the speed and vigor of actions in humans under placebo or muscarinic antagonism. They show that muscarinic antagonism blunts the motivational effects of reward on both saccade velocity and RT, and also modulates the distractibility of participants, in particular by increasing the repulsion of saccades away from distractors. They show that preparatory EEG signals reflect both motivation and drug condition, and make a case that these EEG signals mediate the effects of the drug on behavior.

Strengths:

This manuscript addresses an interesting and timely question and does so using an impressive within-subject pharmacological design and a task well-designed to measure constructs of interest. The authors show clear causal evidence that ACH affects different metrics of saccade generation related to effort expenditure and their modulation by incentive manipulations. The authors link these behavioral effects to motor preparatory signatures, indexed with EEG, that relate to behavioral measures of interest and in at least one case statistically mediate the behavioral effects of ACH antagonism.

Weaknesses:

In full disclosure, I have previously reviewed this manuscript in another journal and the authors have done a considerable amount of work to address my previous concerns. However, I have a few remaining concerns that affect my interpretation of the current manuscript.

Some of the EEG signals (figures 4A&C) have profiles that look like they could have ocular, rather than central nervous, origins. Given that this is an eye movement task, it would be useful if the authors could provide some evidence that these signals are truly related to brain activity and not driven by ocular muscles, either in response to explicit motor effects (ie. Blinks) or in preparation for an upcoming saccade. For other EEG signals, in particular, the ones reported in Figure 3, it would be nice to see what the spatial profiles actually look like - does the scalp topography match that expected for the signal of interest?

A primary weakness of this paper is the sample size - since only 20 participants completed the study. The authors address the sample size in several places and I completely understand the reason for the reduced sample size (study halt due to COVID). That said, they only report the sample size in one place in the methods rather than through degrees of freedom in their statistical tests conducted throughout the results. In part because of this, I am not totally clear on whether the sample size for each analysis is the same - or whether participants were removed for specific analyses (ie. due to poor EEG recordings, for example). Beyond this point, but still related to the sample size, in some cases I worry that results are driven by a single subject. In particular, the interaction effect observed in Figure 1e seems like it would be highly sensitive to the single subject who shows a reverse incentive effect in the drug condition.

There are not sufficient details on the cluster-based permutation testing to understand what the authors did or whether it is reasonable. What channels were included? What metric was computed per cluster? How was null distribution generated?

The authors report that "muscarinic antagonism strengthened the P3a" - but I was unable to see this in the data plots. Perhaps it is because the variability related to individual differences obscures the conditional differences in the plots. In this case, event-related difference signals could be helpful to clarify the results.

For mediation analyses, it would be useful in the results section to have a much more detailed description of the regression results, rather than just reporting things in a binary did/did not mediate sort of way. Furthermore, the methods should also describe how mediation was tested statistically (ie. What is the null distribution that the difference in coefficients with/without moderator is tested against?).

eLife assessment

This study identified an innovative molecular mechanism linking diabetes to Alzheimer's disease (AD) risk, with important significance. The finding presents novel insights into AD pathogenesis and provides strong evidence about the mechanistic roles of Kallistatin, and the therapeutic potential of fenofibrate in AD. The experiments are well conducted, and the evidence is convincing.

Reviewer #1 (Public Review):

Summary:

Qi and colleagues investigated the role of the Kallistatin pathway in increasing hippocampal amyloid-β plaque accumulation and tau hyperphosphorylation in Alzheimer's disease, linking the increased Kallistatin level in diabetic patients with a higher risk of Alzheimer's disease development. A Kallistatin-overexpressing animal model was utilized, and memory impairment was assessed using Morris water maze and Y-maze. Kallistatin-related pathway protein levels were measured in the hippocampus, and phenotypes were rescued using fenofibrate and rosiglitazone. The current study provides evidence of a novel molecular mechanism linking diabetes and Alzheimer's disease and suggests the potential use of fenofibrate to alleviate memory impairment. However, several issues need to be addressed before further consideration.

Strengths:

The findings of this study are novel. The findings will have great impacts on diabetes and AD research. The studies were well conducted, and the results were convincing.

Weaknesses:

(1) The mechanism by which fenofibrate rescues memory loss in Kallistatin-transgenic mice is unclear. As a PPARalpha agonist, does fenofibrate target the Kallistatin pathway directly or indirectly? Please provide a discussion based on literature supporting either possibility.

(2) The current study exclusively investigated the hippocampus. What about other cognitive memory-related regions, such as the prefrontal cortex? Including data from these regions or discussing the possibility of their involvement could provide a more comprehensive understanding of the role of Kallistatin in memory impairment.

(3) Fenofibrate rescued phenotypes in Kallistatin-transgenic mice while rosiglitazone, a PPARgamma agonist, did not. This result contradicts the manuscript's emphasis on a PPARgamma-associated mechanism. Please address this inconsistency.

(4) Most of the immunohistochemistry images are unclear. Inserts have similar magnification to the original representative images, making judgments difficult. Please provide larger inserts with higher resolution.

(5) The immunohistochemistry images in different figures were taken from different hippocampal subregions with different magnifications. Please maintain consistency, or explain why CA1, CA3, or DG was analyzed in each experiment.

(6) Figure 5B is missing a title. Please add a title to maintain consistency with other graphs.

(7) Please list statistical methods used in the figure legends, such as t-test or One-way ANOVA with post-hoc tests.

Reviewer #2 (Public Review):

Summary:

The study links Alzheimer's disease (AD) with metabolic disorders through elevated Kallistatin levels in AD patients. Kallistatin-overexpressing mice show cognitive decline, increased Aβ and tau pathology, and impaired hippocampal function. Mechanistically, Kallistatin enhances Aβ production via Notch1 and promotes tau phosphorylation through GSK-3β activation. Fenofibrate improves cognitive deficits by reducing Aβ and tau phosphorylation in these mice, suggesting therapeutic potential in AD linked to metabolic syndromes.

Strengths:

This study presents novel insights into AD pathogenesis and provides strong evidence about the mechanistic roles of Kallistatin, and the therapeutic potential of fenofibrate in AD.

Weaknesses:

It was suggested that Kallistatin is primarily produced by the liver. The study demonstrates increased Kallistatin levels in the hippocampus tissue of AD mice. It would be valuable to clarify if Kallistatin is also increased in the liver of AD mice, providing a comprehensive understanding of its distribution in disease states.

Does Kallistatin interact directly with Notch1 ligands? Clarifying this interaction mechanism would enhance understanding of how Kallistatin influences Notch1 signaling in AD pathology.

Is there any observed difference in AD phenotype between male and female Kallistatin-transgenic (KAL-TG) mice? Including this information would address potential gender-specific effects on cognitive decline and pathology.

It is recommended to include molecular size markers in Western blots for clarity and accuracy in protein size determination.

The language should be revised for enhanced readability and clarity, ensuring that complex scientific concepts are communicated effectively to a broader audience.

Reviewer #3 (Public Review):

Summary:

The authors investigated the role of kallistatin in metabolic abnormalities associated with AD. They found that Kallistatin promotes Aβ production by binding to the Notch1 receptor and upregulating BACE1 expression. They identified that Kallistatin is a key player that mediates Aβ accumulation and tau hyperphosphorylation in AD.

Strengths:

This manuscript not only provides novel insights into the pathogenesis of AD, but also indicates that the hypolipidemic drug fenofibrate attenuates AD-like pathology in Kallistatin transgenic mice.

Weaknesses:

The authors did not illustrate whether the protective effect of fenofibrate against AD depends on kallistatin.

The conclusions are supported by the results, but the quality of some results should be improved.

eLife assessment

This valuable study investigates the selectivity of neuronal responses in the primary visual cortex and the dorsal lateral geniculate nucleus to stimuli presented far outside their receptive fields. The evidence supporting the claims is incomplete, due to lack of clarity. This paper should be of interest to neurophysiologists interested in vision and contextual modulations.

Reviewer #1 (Public Review):

Summary:

The authors report a study on how stimulation of receptive-field surround of V1 and LGN neurons affects their firing rates. Specifically, they examine stimuli in which a grey patch covers the classical RF of the cell and a stimulus appears in the surround. Using a number of different stimulus paradigms they find a long latency response in V1 (but not the LGN) which does not depend strongly on the characteristics of the surround grating (drifting vs static, continuous vs discontinuous, predictable grating vs unpredictable pink noise). They find that population responses to simple achromatic stimuli have a different structure that does not distinguish so clearly between the grey patch and other conditions and the latency of the response was similar regardless of whether the center or surround was stimulated by the achromatic surface. Taken together they propose that the surround-response is related to the representation of the grey surface itself. They relate their findings to previous studies that have put forward the concept of an 'inverse RF' based on strong responses to small grey patches on a full-screen grating. They also discuss their results in the context of studies that suggest that surround responses are related to predictions of the RF content or figure-ground segregation.

Strengths:

I find the study to be an interesting extension of the work on surround stimulation and the addition of the LGN data is useful showing that the surround-induced responses are not present in the feed-forward path. The conclusions appear solid, being based on large numbers of neurons obtained through Neuropixels recordings. The use of many different stimulus combinations provides a rich view of the nature of the surround-induced responses.

Weaknesses:

The statistics are pooled across animals, which is less appropriate for hierarchical data. There is no histological confirmation of placement of the electrode in the LGN and there is no analysis of eye or face movements which may have contributed to the surround-induced responses. There are also some missing statistics and methods details which make interpretation more difficult.

Reviewer #2 (Public Review):

Cuevas et al. investigate the stimulus selectivity of surround-induced responses in the mouse primary visual cortex (V1). While classical experiments in non-human primates and cats have generally demonstrated that stimuli in the surround receptive field (RF) of V1 neurons only modulate activity to stimuli presented in the center RF, without eliciting responses when presented in isolation, recent studies in mouse V1 have indicated the presence of purely surround-induced responses. These have been linked to prediction error signals. In this study, the authors build on these previous findings by systematically examining the stimulus selectivity of surround-induced responses.

Using neuropixels recordings in V1 and the dorsal lateral geniculate nucleus (dLGN) of head-fixed, awake mice, the authors presented various stimulus types (gratings, noise, surfaces) to the center and surround, as well as to the surround only, while also varying the size of the stimuli. Their results confirm the existence of surround-induced responses in mouse V1 neurons, demonstrating that these responses do not require spatial or temporal coherence across the surround, as would be expected if they were linked to prediction error signals. Instead, they suggest that surround-induced responses primarily reflect the representation of the achromatic surface itself.

The literature on center-surround effects in V1 is extensive and sometimes confusing, likely due to the use of different species, stimulus configurations, contrast levels, and stimulus sizes across different studies. It is plausible that surround modulation serves multiple functions depending on these parameters. Within this context, the study by Cuevas et al. makes a significant contribution by exploring the relationship between surround-induced responses in mouse V1 and stimulus statistics. The research is meticulously conducted and incorporates a wide range of experimental stimulus conditions, providing valuable new insights regarding center-surround interactions.

However, the current manuscript presents challenges in readability for both non-experts and experts. Some conclusions are difficult to follow or not clearly justified.

I recommend the following improvements to enhance clarity and comprehension:

(1) Clearly state the hypotheses being tested at the beginning of the manuscript.

(2) Always specify the species used in referenced studies to avoid confusion (esp. Introduction and Discussion).

(3) Briefly summarize the main findings at the beginning of each section to provide context.

(4) Clearly define important terms such as "surface stimulus" and "early vs. late stimulus period" to ensure understanding.

(5) Provide a rationale for each result section, explaining the significance of the findings.

(6) Offer a detailed explanation of why the results do not support the prediction error signal hypothesis but instead suggest an encoding of the achromatic surface.

These adjustments will help make the manuscript more accessible and its conclusions more compelling.

Reviewer #3 (Public Review):

Summary:

This paper explores the phenomenon whereby some V1 neurons can respond to stimuli presented far outside their receptive field. It introduces three possible explanations for this phenomenon and it presents experiments that it argues favor the third explanation, based on figure/ground segregation.

Strengths:

I found it useful to see that there are three possible interpretations of this finding (prediction error, interpolation, and figure/ground). I also found it useful to see a comparison with LGN responses and to see that the effect there is not only absent but actually the opposite: stimuli presented far outside the receptive field suppress rather than drive the neurons. Other experiments presented here may also be of interest to the field.

Weaknesses:

The paper is not particularly clear. I came out of it rather confused as to which hypotheses were still standing and which hypotheses were ruled out. There are numerous ways to make it clearer.

eLife assessment

This work presents important findings regarding the interaction of the monkeypox virus (MPXV) attachment H3 protein with the cellular receptor heparan sulfate and the use of this information to develop antivirals potentially effective against all orthopoxviruses. Using a combination of state-of-the art computational and wet experiments the authors present solid evidence to sustain their claims. These results will interest those working on basic orthopoxviruses biology and antiviral development.

Reviewer #1 (Public Review):

Summary:

The study aimed to better understand the role of the H3 protein of the Monkeypox virus (MPXV) in host cell adhesion, identifying a crucial α-helical domain for interaction with heparan sulfate (HS). Using a combination of advanced computational simulations and experimental validations, the authors discovered that this domain is essential for viral adhesion and potentially a new target for developing antiviral therapies.

Strengths:

The study's main strengths include the use of cutting-edge computational tools such as AlphaFold2 and molecular dynamics simulations, combined with robust experimental techniques like single-molecule force spectroscopy and flow cytometry. These methods provided a detailed and reliable view of the interactions between the H3 protein and HS. The study also highlighted the importance of the α-helical domain's electric charge and the influence of the Mg(II) ion in stabilizing this interaction. The work's impact on the field is significant, offering new perspectives for developing antiviral treatments for MPXV and potentially other viruses with similar adhesion mechanisms. The provided methods and data are highly useful for researchers working with viral proteins and protein-polysaccharide interactions, offering a solid foundation for future investigations and therapeutic innovations.

Weaknesses:

However, some limitations are notable. Despite the robust use of computational methodologies, the limitations of this approach are not discussed, such as potential sources of error, standard deviation rates, and known controls for the H3 protein to justify the claims. Additionally, validations with methodologies like X-ray crystallography would further benefit the visualization of the H3 and HS interaction.

Reviewer #2 (Public Review):

Summary:

The manuscript presenting the discovery of a heparan-sulfate (HS) binding domain in monkeypox virus (MPXV) H3 protein as a new anti-poxviral drug target, presented by Bin Zhen and co-workers, is of interest, given that it offers a potentially broad antiviral substance to be used against poxviruses. Using new computational biology techniques, the authors identified a new alpha-helical domain in the H3 protein, which interacts with cell surface HS, and this domain seems to be crucial for H3-HS interaction. Given that this domain is conserved across orthopoxviruses, authors designed protein inhibitors. One of these inhibitors, AI-PoxBlock723, effectively disrupted the H3-HS interaction and inhibited infection with Monkeypox virus and Vaccinia virus. The presented data should be of interest to a diverse audience, given the possibility of an effective anti-poxviral drug. 

Strengths:

In my opinion, the experiments done in this work were well-planned and executed. The authors put together several computational methods, to design poxvirus inhibitor molecules, and then they test these molecules for infection inhibition.

Weaknesses:

One thing that could be improved, is the presentation of results, to make them more easily understandable to readers, who may not be experts in protein modeling programs. For example, figures should be self-explanatory and understood on their own, without the need to revise text. Therefore, the figure legend should be more informative as to how the experiments were done.

Reviewer #3 (Public Review):

Summary:

The article is an interesting approach to determining the MPOX receptor using "in silico" tools. The results show the presence of two regions of the H3 protein with a high probability of being involved in the interaction with the HS cell receptor. However, the α-helical region seems to be the most probable, since modifications in this region affect the virus binding to the HS receptor.

Strengths:

In my opinion, it is an informative article with interesting results, generated by a combination of "in silico" and wet science to test the theoretical results. This is a strong point of the article.

Weaknesses:

Has a crystal structure of the H3 protein been reported?

The following text is in line 104: "which may represent a novel binding site for HS". It is unclear whether this means this "new binding site" is an alternative site to an old one or whether it is the true binding site that had not been previously elucidated.

eLife assessment

This study presents valuable framework and findings to our understanding of the brain cortex as a fractal object. Based on detailed methodology, the evidence provided on the stability of its shape property within 11 primate species is convincing, as well as the scale-specific effects of ageing on the human brain. This study will be of interest to neuroscientists interested in brain morphology, and to physicists and mathematicians interested in modeling the shapes of complex objects.

Reviewer #2 (Public Review):

In this manuscript, the authors analyze the shapes of cerebral cortices from several primate species, including subgroups of young and old humans, to characterize commonalities in patterns of gyrification, cortical thickness, and cortical surface area. The authors state that the observed scaling law shares properties with fractals, where shape properties are similar across several spatial scales. One way the authors assess this is to perform a "cortical melting" operation that they have devised on surface models obtained from several primate species. The authors also explore differences in shape properties between brains of young (~20 year old) and old (~80) humans. A challenge the authors acknowledge struggling with in reviewing the manuscript is merging "complex mathematical concepts and a perplexing biological phenomenon." This reviewer remains a bit skeptical about whether the complexity of the mathematical concepts being drawn from are justified by the advances made in our ability to infer new things about the shape of the cerebral cortex.

(1) The series of operations to coarse-grain the cortex illustrated in Figure 1 produces image segmentations that do not resemble real brains. The process to assign voxels in downsampled images to cortex and white matter is biased towards the former, as only 4 corners of a given voxel are needed to intersect the original pial surface, but all 8 corners are needed to be assigned a white matter voxel. The reason for introducing this bias (and to the extent that it is present in the authors' implementation) is not provided. The authors provide an intuitive explanation of why thickness relates to folding characteristics, but ultimately an issue for this reviewer is, e.g., for the right-most panel in Figure 2b, the cortex consists of several 4.9-sided voxels and thus a >2 cm thick cortex. A structure with these morphological properties is not consistent with the anatomical organization of typical mammalian neocortex.

(2) For the comparison between 20-year-old and 80-year-old brains, a well-documented difference is that the older age group possesses more cerebral spinal fluid due to tissue atrophy, and the distances between the walls of gyri becomes greater. This difference is born out in the left column of Figure 4b. It seems this additional spacing between gyri in 80 year olds requires more extensive down-sampling (larger scale values in Figure 4a) to achieve a similar shape parameter K as for the 20 year olds. The authors assert that K provides a more sensitive measure (associated with a large effect size) than currently used ones for distinguishing brains of young vs. old people. A more explicit, or elaborate, interpretation of the numbers produced in this manuscript, in terms of brain shape, might make this analysis more appealing to researchers in the aging field.

(3) In the Discussion, it is stated that self-similarity, operating on all length scales, should be used as a test for existing and future models of gyrification mechanisms. Given the lack of association between the abstract mathematical parameters described in this study and explicit properties of brain tissue and its constituents, it is difficult to envision how the coarse-graining operation can be used to guide development of "models of cortical gyrification."

(4) There are several who advocate for analyzing cortical mid-thickness surfaces, as the pial surface over-represents gyral tips compared to the bottoms of sulci in the surface area. The authors indicate that analyses of mid-thickness representations will be taken on in future work, but this seems to be a relevant control for accepting the conclusions of this manuscript.

Reviewer #3 (Public Review):

Summary: Through a rigorous methodology, the authors demonstrated that within 11 different primates, the shape of the brain followed a universal scaling law with fractal properties. They enhanced the universality of this result by showing the concordance of their results with a previous study investigating 70 mammalian brains, and the discordance of their results with other folded objects that are not brains. They incidentally illustrated potential applications of this fractal property of the brain by observing a scale-dependant effect of aging on the human brain.

Strengths:<br /> - New hierarchical way of expressing cortical shapes at different scales derived from previous report through implementation of a coarse-graining procedure<br /> - Investigation of 11 primate brains and contextualisation with other mammals based on prior literature<br /> - Proposition of tool to analyse cortical morphology requiring no fine tuning and computationally achievable<br /> - Positioning of results in comparison to previous works reinforcing the validity of the observation.<br /> - Illustration of scale-dependance of effects of brain aging in the human.

Weaknesses:<br /> - The notion of cortical shape, while being central to the article, is not really defined, leaving some interpretation to the reader<br /> - The organization of the manuscript is unconventional, leading to mixed contents in different sections (sections mixing introduction and method, methods and results, results and discussion...). As a result, the reader discovers the content of the article along the way, it is not obvious at what stages the methods are introduced, and the results are sometimes presented and argued in the same section, hindering objectivity.<br /> To improve the document, I would suggest a modification and restructuring of the article such that: 1) by the end of the introduction the reader understands clearly what question is addressed and the value it holds for the community, 2) by the end of the methods the reader understands clearly all the tools that will be used to answer that question (not just the new method), 3) by the end of the results the reader holds the objective results obtained by applying these tools on the available data (without subjective interpretations and justifications), and 4) by the end of the discussion the reader understands the interpretation and contextualisation of the study, and clearly grasps the potential of the method depicted for the better understanding of brain folding mechanisms and properties.

Author response:

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

eLife assessment:

This study presents valuable framework and findings to our understanding of the brain as a fractal object by observing the stability of its shape property within 11 primate species and by highlighting an application to the effects of aging on the human brain. The evidence provided is solid but the link between brain shape and the underlying anatomy remains unclear. This study will be of interest to neuroscientists interested in brain morphology, whether from an evolutionary, fundamental or pathological point of view, and to physicists and mathematicians interested in modeling the shapes of complex objects.

We now clarified the outstanding questions regarding if our model outputs can be related to actual primate brain anatomy, which we believe was mainly based on comments regarding the validity of our output of apparently thicker cortices than nature can produce.

We address this point in more detail in the point-by-point response below, but want to address this misunderstanding directly here: Our algorithm does not produce thicker cortices with increasing coarse-graining scales; in fact, the cortical thickness never exceeds the actual cortical thickness in our outputs, but rather thins with each coarse-graining scale. In other words, we believe that our outputs are fully in line with neuroanatomy across species.

Reviewer #2 (Public Review): 

In this manuscript, the authors analyze the shapes of cerebral cortices from several primate species, including subgroups of young and old humans, to characterize commonalities in patterns of gyrification, cortical thickness, and cortical surface area. The authors state that the observed scaling law shares properties with fractals, where shape properties are similar across several spatial scales. One way the authors assess this is to perform a "cortical melting" operation that they have devised on surface models obtained from several primate species. The authors also explore differences in shape properties between brains of young (~20 year old) and old (~80) humans. A challenge the authors acknowledge struggling with in reviewing the manuscript is merging "complex mathematical concepts and a perplexing biological phenomenon." This reviewer remains a bit skeptical about whether the complexity of the mathematical concepts being drawn from are justified by the advances made in our ability to infer new things about the shape of the cerebral cortex. 

To allow scientists from all backgrounds to adopt these complex ideas, we have made our code to “melt” the brains and for further downstream analysis publicly available. We have now also provided a graphical user interface, to allow users without substantial coding experience to run the analysis. We also believe that the algorithmic concepts are easy to understand due to the similarity to the coarse-graining procedures found in long-standing and well-accepted box-counting algorithms.

Beyond the theoretical insight of the fractal nature of cortices and providing an explicit and crucial link between vastly different brains that are gyrified and those that are not, we believe that the advance gained by our methods for future applications is clearly demonstrated in our proof-of-principle with a four-fold increase in effect size. For reference, an effect size of 8 would translate to an almost perfect separation of groups, i.e. an ideal biomarker with near 100% sensitivity and specificity.

(1) The series of operations to coarse-grain the cortex illustrated in Figure 1 produces image segmentations that do not resemble real brains.

As re-iterated in our Methods and Discussion: “Note, of course, that the coarse-grained brain surfaces are an output of our algorithm alone and are not to be directly/naively likened to actual brain surfaces, e.g. in terms of the location or shape of the folds. Our comparisons here between coarse-grained brains and actual brains is purely on the level of morphometrics across the whole cortex.”

Fig. 1 therefore serves as an explanation to the reader on the algorithmic outputs, but each melted brain is not supposed to be directly/visually compared to actual brains. Similar to algorithms measuring the fractal dimension, or the exposed surface area of a given brain, the intermediate outputs of these algorithms are not supposed to represent any biologically observed brain structures, but rather serve as an abstraction to obtain meaningful morphometrics.

We additionally added a note to the caption of Fig. 1 to clarify this point:

“Note that the actual size of the brains for analysis are rescaled (see Methods and Fig. 3); we display all brains scaled at an equal size here for the ease of visualisation of the method.”

Finally, we also edited the entire paper for terminology to clearly distinguish the terms of (1) the cortex as a 3D object, (2) coarse-grained and voxelised versions thereof, and (3) summary morphological measures derived from the former. When we invite comparisons in our paper between real brains and coarse-grained brains, this is always at the level of summary morphological measures, not at the level of the 3D objects/voxelisations themselves.

The process to assign voxels in downsampled images to cortex and white matter is biased towards the former, as only 4 corners of a given voxel are needed to intersect the original pial surface, but all 8 corners are needed to be assigned a white matter voxel. The reason for introducing this bias (and to the extent that it is present in the authors' implementation) is not provided.

This detail was in the Supplementary, and we have now added additional clarification on this specific point to our Supplementary:

“In detail, we assign all voxels in the grid with at least four corners inside the original pial surface to the pial voxelization. This process allows the exposed surface to remain approximately constant with increasing voxel sizes. A constant exposed surface is desirable, as we only want to gradually ‘melt’ and fuse the gyri, but not grow the bounding/exposed surface as well. We want the extrinsic area to remain approximately constant as we decrease the intrinsic area via coarse-graining; it is like generating iterates of a Koch curve in reverse, from more to less detailed, by increasing the length of smallest line segment.

We then assign voxels with all eight corners inside the original white matter surface to the white matter voxelization. This is to ensure integrity of the white matter, as otherwise white matter voxels in gyri may become detached from the core white matter, and thus artificially increase white matter surface area. Indeed, the main results of the paper are not very sensitive to this decision using all eight corners, vs. e.g. only four corners, as we do not directly use white matter surface area for the scaling law measurements. However, we still maintained this choice in case future work wants to make use of the white matter voxelisations or derivative measures.”

Note on the point of white matter integrity that if both grey and white matter voxelisations require all 8 corner to be inside the respective mesh, there will be voxels not assigned to either at the grey/white matter interface, causing potential downstream issues.

We further acknowledge:

“Of course, our proposed procedure is not the only conceivable way to erase shape details below a given scale; and we are actively working on related algorithms that are also computationally cheaper. Nevertheless, the current version requires no fine-tuning, is computationally feasible and conceptually simple, thus making it a natural choice for introducing the methodology and approach.”

The authors provide an intuitive explanation of why thickness relates to folding characteristics, but ultimately an issue for this reviewer is, e.g., for the right-most panel in Figure 2b, the cortex consists of several 4.9-sided voxels and thus a >2 cm thick cortex. A structure with these morphological properties is not consistent with the anatomical organization of typical mammalian neocortex. 

We assume the reviewer refers to Fig. 1B with the panel on scale=4.9mm. We would like to point out that Fig. 1 serves as an explanation of the voxelisation method. For the actual analysis and Results, we are using re-scaled brains (see Fig. 2 with the ever decreasing brain sizes). The rescaling procedure is now expanded as below:

“Morphological properties, such as cortical thicknesses measured in our ‘melted’ brains are to be understood as a thickness relative to the size of the brain. Therefore, to analyse the scaling behaviour of the different coarse-grained realisations of the same brain, we apply an isometric rescaling process that leaves all dimensionless shape properties unaffected (more details in Suppl. S3.1). Conceptually, this process fixes the voxel size, and instead resizes the surfaces relative to the voxel size, which ensures that we can compare the coarse-grained realisations to the original cortices, and test if the former, like the latter, also scale according to Eqn. (1). Resizing, or more precisely, shrinking the cortical surface is mathematically equivalent to increasing the box size in our coarse-graining method. Both achieved an erasure of folding details below a certain threshold. After rescaling, as an example, the cortical thickness also shrinks with increasing levels of coarse-graining, and never exceeds the thickness measured at native scale.”

We additionally added a note to the caption of Fig. 1 to clarify this point:

“Note that the actual size of the brains for analysis are rescaled (see Methods and Fig. 3); we display all brains scaled at an equal size here for the ease of visualisation of the method.”

Finally, we also edited the entire paper for terminology to clearly distinguish the terms of (1) the cortex as a 3D object, (2) coarse-grained versions thereof, and (3) summary morphological measures derived from the former. When we invite comparisons in our paper between real brains and coarse-grained brains, this is always at the level of summary morphological measures, not at the level of the 3D objects themselves and their detailed anatomical features.

(2) For the comparison between 20-year-old and 80-year-old brains, a well-documented difference is that the older age group possesses more cerebral spinal fluid due to tissue atrophy, and the distances between the walls of gyri becomes greater. This difference is born out in the left column of Figure 4b. It seems this additional spacing between gyri in 80 year olds requires more extensive down-sampling (larger scale values in Figure 4a) to achieve a similar shape parameter K as for the 20 year olds. The authors assert that K provides a more sensitive measure (associated with a large effect size) than currently used ones for distinguishing brains of young vs. old people. A more explicit, or elaborate, interpretation of the numbers produced in this manuscript, in terms of brain shape, might make this analysis more appealing to researchers in the aging field.

We have removed the main results relating to K and aging from our last revision already to avoid confusion. This is now only in the supplementary analysis, and our claim of K being a more sensitive measure for age and ageing – whilst still true – will be presented in more detail in a series of upcoming papers.

(3) In the Discussion, it is stated that self-similarity, operating on all length scales, should be used as a test for existing and future models of gyrification mechanisms. Given the lack of association between the abstract mathematical parameters described in this study and explicit properties of brain tissue and its constituents, it is difficult to envision how the coarse-graining operation can be used to guide development of "models of cortical gyrification."

We have clarified in more detail what we meant originally in Discussion:

“Finally, this dual universality is also a more stringent test for existing and future models of cortical gyrification mechanisms at relevant scales, and one that moreover is applicable to individual cortices. For example, any models that explicitly simulate a cortical surface as an output could be directly coarse-grained with our method and the morphological trajectories can be compared with those of actual human and primate cortices. The simulated cortices would only be ‘valid’ in terms of the dual universality, if it also produces the same morphological trajectories.”

However, we agree with the reviewer that our paper could be misread as demanding direct comparisons of each coarse-grained brain with an actual brain, and we have now added the following text to clarify that this is not our intention for the proposed method or outputs.

“Note, we do not suggest to directly compare coarse-grained brain surfaces with actual biological brain surfaces. As we noted earlier, 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 comparisons here between coarse-grained brains and actual brains is purely on the level of morphometrics across the whole cortex.”

Indeed, the dual universality imposes restrictive constraints on the possible shapes of real cortices, but do not fully specify them. Presumably, the location of individual folds in different individuals and species will depend on their respective evolutionary histories, so there is no reason to expect a match in fold location between the ‘melted’ cortices of more gyrified species, on one hand, and the cortex of a less-gyrified one, on the other,  even if their global morphological parameters and global mechanism of folding coincide.

(4) There are several who advocate for analyzing cortical mid-thickness surfaces, as the pial surface over-represents gyral tips compared to the bottoms of sulci in the surface area. The authors indicate that analyses of mid-thickness representations will be taken on in future work, but this seems to be a relevant control for accepting the conclusions of this manuscript.

In the context of some applications and methods, we agree that the mid-surface is a meaningful surface to analyse. However, in our work, the mid-surface is not. The fractal estimation rests on the assumption that the exposed area hugs the object of interest (hence convex hull of the pial surface), as the relationship between the extrinsic and intrinsic areas across scales determine the fractal relationship (Eq. 2). If we used the mid-surface instead of the pial surface for all estimation, this would not represent the actual object of interest, and it is separated from the convex hull. Estimating a new convex hull based on the mid surface would be the equivalent of asking for the fractal dimension of the mid-surface, not of the cortical ribbon. In other words, it would be a different question, bound to yield a different answer.

Hence, we indicated in our original response that we only have a provisional answer, but more work beyond the scope of this paper is required to answer this question, as it is a separate question. The mid-surface, as a morphological structure in its own right, will have its own scaling properties, and our provisional understanding is that these also yield a scaling law parallel to those of the cortical ribbon with the same or a similar fractal dimension. But more systematic work is required to investigate this question at native scale and across scales.

Reviewer #3 (Public Review):

Summary: Through a rigorous methodology, the authors demonstrated that within 11 different primates, the shape of the brain followed a universal scaling law with fractal properties. They enhanced the universality of this result by showing the concordance of their results with a previous study investigating 70 mammalian brains, and the discordance of their results with other folded objects that are not brains. They incidentally illustrated potential applications of this fractal property of the brain by observing a scale-dependant effect of aging on the human brain. 

Strengths: 

- New hierarchical way of expressing cortical shapes at different scales derived from previous report through implementation of a coarse-graining procedure 

- Investigation of 11 primate brains and contextualisation with other mammals based on prior literature 

- Proposition of tool to analyse cortical morphology requiring no fine tuning and computationally achievable 

- Positioning of results in comparison to previous works reinforcing the validity of the observation. 

- Illustration of scale-dependance of effects of brain aging in the human. 

Weaknesses: 

- The notion of cortical shape, while being central to the article, is not really defined, leaving some interpretation to the reader 

- The organization of the manuscript is unconventional, leading to mixed contents in different sections (sections mixing introduction and method, methods and results, results and discussion...). As a result, the reader discovers the content of the article along the way, it is not obvious at what stages the methods are introduced, and the results are sometimes presented and argued in the same section, hindering objectivity. 

To improve the document, I would suggest a modification and restructuring of the article such that: 1) by the end of the introduction the reader understands clearly what question is addressed and the value it holds for the community, 2) by the end of the methods the reader understands clearly all the tools that will be used to answer that question (not just the new method), 3) by the end of the results the reader holds the objective results obtained by applying these tools on the available data (without subjective interpretations and justifications), and 4) by the end of the discussion the reader understands the interpretation and contextualisation of the study, and clearly grasps the potential of the method depicted for the better understanding of brain folding mechanisms and properties. 

We thank this reviewer again for their attention to detail and constructive comments. We have followed the detailed suggestions provided by us in the Recommendations For The Authors, and summarise the main changes here:

- We have restructured all sections to be more clearly following Introduction, Methods, Results, and Discussion; by using subsections, we believe the structure is now more accessible to readers.

-  We have now clarified the concept of “cortical shape”, as we use it in our paper in several places, by distinguishing clearly the object of study, and the morphological properties measured from it.

Recommendations for the authors: 

Reviewer #2 (Recommendations For The Authors): None 

Reviewer #3 (Recommendations For The Authors): 

I once again compliment the authors for their elegant work. I am happy with the way they covered my first feedback. My second review takes into account some comments made by other reviewers with which I agree. 

We thank this reviewer again for their attention to detail and constructive comments.

Recommendations for clarifications: 

General comments: The purpose of the article could be made clearer in the introduction. When I differentiate results from discussion, I think of results as objective measures or observations, while discussion will relate to the interpretation of these results (including comparison with previous literature, in most cases). 

We have restructured all sections to be more clearly following Introduction, Methods, Results, and Discussion; by using subsection, we believe the structure is now more accessible to readers.

- l.39: define or discuss "cortical shape" 

We have gone through the entire paper and corrected for any ambiguities. We specifically distinguish between the cortex as a structure overall, shape measures derived from this structure, and coarse-grained versions of the structure.

- l.48-74: this would match either an introduction or a discussion rather than a methods section. 

Done

- l.98-106: this would match a discussion rather than a methods section. 

Done

- l.111: here could be a good spot to discuss the 4 vs 8 corners for inclusion of pial vs white matter voxelization 

We have discussed this in the more detailed Supplementary section now, as after restructuring, this appears to be the more suitable place.

- l.140-180: it feels that this section mixes methods, results and discussion of the results 

We agree and we have resolved this by removing sentences and re-arranging sections.

- l.183-217: mix of results and discussion 

We agree and we have resolved this by removing sentences and re-arranging sections.

Small cosmetic suggestions: 

- l.44: conservation of 'some' quantities: vague 

Changed to conservation of morphological relationships across evolution

- l.66: order of citations ([24, 22,23]) 

Will be fixed at proof stage depending on format of references.

- l.77: delete space between citation and period 

Done

- l.77: I would delete 'say' 

Done

- l.86: 'but to also analyse' -> 'to analyse' 

Done

- l.105: remove 'we are encouraged that' 

Done

- l.111: 'also see' -> 'see also' 

Done

- l.164: 'remarkable': subjective 

Done

- l.189: define approx. abbreviation 

Done

- l.190: 'approx' -> 'approx.' 

Revised

- l.195: 'dramatic': subjective 

removed

-l. 246: 'much' -> vague 

explained

eLife assessment

This study presents a valuable report on a machine-learning tool for predicting synergistic drug combinations for cancer treatment. However, the evidence supporting the claims of the authors is incomplete, as the reported model shows some evidence of overfitting, and the claims of the authors could be strengthened if additional validation experiments were performed. The work will be of interest to oncologists and medical biologists working on cancer.

Reviewer #1 (Public Review):

The authors introduce DIPx, a deep learning framework for predicting synergistic drug combinations for cancer treatment using the AstraZeneca-Sanger (AZS) DREAM Challenge dataset. While the approach is innovative, I have the following concerns and comments which hopefully will improve the study's rigor and applicability, making it a more powerful tool in the real clinical world.

(1) Test Set 1 comprises combinations already present in the training set, likely leading overfitting issue. The model might show inflated performance metrics on this test set due to prior exposure to these combinations, not accurately reflecting its true predictive power on unknown data, which is crucial for discovering new drug synergies. The testing approach reduces the generalizability of the model's findings to new, untested scenarios.

(2) The model struggles with predicting synergies for drug combinations not included in its training data (showing only a Spearman correlation of 0.26 in Test Set 2). This limits its potential for discovering new therapeutic strategies. Utilizing techniques such as transfer learning or expanding the training dataset to encompass a wider range of drug pairs could help to address this issue.

(3) The use of pan-cancer datasets, while offering broad applicability, may not be optimal for specific cancer subtypes with distinct biological mechanisms. Developing subtype-specific models or adjusting the current model to account for these differences could improve prediction accuracy for individual cancer types.

(4) Line 127, "Since DIPx uses only molecular data, to make a fair comparison, we trained TAJI using only molecular features and referred to it as TAJI-M.". TAJI was designed to use both monotherapy drug-response and molecular data, and likely won't be able to reach maximum potential if removing monotherapy drug-response from the training model. It would be critical to use the same training datasets and then compare the performances. From Figure 6 of TAJI's paper (Li et al., 2018, PMID: 30054332) , i.e., the mean Pearson correlation for breast cancer and lung cancer is around 0.5 - 0.6.

The following 2 concerns have been included in the Discussion section which is great:

(1) Training and validating the model using cell lines may not fully capture the heterogeneity and complexity of in vivo tumors. To increase clinical relevance, it would be beneficial to validate the model using primary tumor samples or patient-derived xenografts.

(2) The Pathway Activation Score (PAS) is derived exclusively from primary target genes, potentially overlooking critical interactions involving non-primary targets. Including these secondary effects could enhance the model's predictive accuracy and comprehensiveness.

Reviewer #2 (Public Review):

Trac, Huang, et al used the AZ Drug Combination Prediction DREAM challenge data to make a new random forest-based model for drug synergy. They make comparisons to the winning method and also show that their model has some predictive capacity for a completely different dataset. They highlight the ability of the model to be interpretable in terms of pathway and target interactions for synergistic effects. While the authors address an important question, more rigor is required to understand the full behavior of the model.

Major Points

(1) The authors compare DIPx to the winning method of the DREAm challenge, TAJI to show that from molecular features alone they retrain TAJI to create TAJI-M without the monotherapy data inputs. They mention that "of course, we could also use such data in DIPx...", but they never show the behaviour of DIPx with these data. The authors need to demonstrate that this statement holds true or else compare it to the full TAJI.

(2) It would be neat to see how the DIPx feature importance changes with monotherapy input. For most realistic scenarios in which these models are used robust monotherapy data do exist.

(3) In Figure 2, the authors compare DIPx and TAJI-M on various test sets. If I understood correctly, they also bootstrapped the training set with n=100 and reported all the model variants in many of the comparisons. While this is a nice way of showing model robustness, calculating p-values with bootstrapped data does not make sense in my opinion as by increasing the value of n, one can make the p-value arbitrarily small. The p-value should only be reported for the original models.

(4) From Figures 2 and 3, it appears DIPx is overfit on the training set with large gaps in Spearman correlations between Test Set 2/ONeil set and Test Set 1. It also features much better in cases where it has seen both compounds. Could the authors also compare TAJI on the ONeil dataset to show if it is as much overfit?

Reviewer #3 (Public Review):

Summary:

Predicting how two different drugs act together by looking at their specific gene targets and pathways is crucial for understanding the biological significance of drug combinations. Such combinations of drugs can lead to synergistic effects that enhance drug efficacy and decrease resistance. This study incorporates drug-specific pathway activation scores (PASs) to estimate synergy scores as one of the key advancements for synergy prediction. The new algorithm, Drug synergy Interaction Prediction (DIPx), developed in this study, uses gene expression, mutation profiles, and drug synergy data to train the model and predict synergy between two drugs and suggests the best combinations based on their functional relevance on the mechanism of action. Comprehensive validations using two different datasets and comparing them with another best-performing algorithm highlight the potential of its capabilities and broader applications. However, the study would benefit from including experimental validation of some predicted drug combinations to enhance its reliability.

Strengths:

The DIPx algorithm demonstrates the strengths listed below in its approach for personalized drug synergy prediction. One of its strengths lies in its utilization of biologically motivated cancer-specific (driver genes-based) and drug-specific (target genes-based) pathway activation scores (PASs) to predict drug synergy. This approach integrates gene expression, mutation profiles, and drug synergy data to capture information about the functional interactions between drug targets, thereby providing a potential biological explanation for the synergistic effects of combined drugs. Additionally, DIPx's performance was tested using the AstraZeneca-Sanger (AZS) DREAM Challenge dataset, especially in Test Set 1, where the Spearman correlation coefficient between predicted and observed drug synergy was 0.50 (95% CI: 0.47-0.53). This demonstrates the algorithm's effectiveness in handling combinations already in the training set. Furthermore, DIPx's ability to handle novel combinations, as evidenced by its performance in Test Set 2, indicates its potential for extrapolating predictions to new and untested drug combinations. This suggests that the algorithm can adapt to and make accurate predictions for previously unencountered combinations, which is crucial for its practical application in personalized medicine. Overall, DIPx's integration of pathway activation scores and its performance in predicting drug synergy for known and novel combinations underscore its potential as a valuable tool for personalized prediction of drug synergy and exploration of activated pathways related to the effects of combined drugs.

Weaknesses:

While the DIPx algorithm shows promise in predicting drug synergy based on pathway activation scores, it's essential to consider its limitations. One limitation is that the algorithm's performance was less accurate when predicting drug synergy for combinations absent from the training set. This suggests that its predictive capability may be influenced by the availability of training data for specific drug combinations. Additionally, further testing and validation across different datasets (more than the current two datasets) would be necessary to assess the algorithm's generalizability and robustness fully. It's also important to consider potential biases in the training data and ensure that DIPx predictions are validated through empirical studies including experimental testing of predicted combinations. Despite these limitations, DIPx represents a valuable step towards personalized prediction of drug synergy and warrants continued investigation and improvement. It would benefit if the algorithm's limitations are described with some examples and suggest future advancement steps.

eLife assessment

This study reports a comparative biochemical and structural analysis of two PLP decarboxylase enzymes from plants. The work is useful because of the potential application of these enzymes in industrial theanine production. The structure provides a solid basis for understanding substrate specificity but some aspects of the work are incomplete. The paper will be of interest to enzymologists studying PLP enzymes and those working on enzyme engineering in plants.

Reviewer #2 (Public Review):

Summary:

The manuscript focuses on comparison of two PLP-dependent enzyme classes that perform amino acyl decarboxylations. The goal of the work is to understand the substrate specificity and factors that influence catalytic rate in an enzyme linked to theanine production in tea plants.

Strengths:

The work includes x-ray crystal structures of modest resolution of the enzymes of interest. These structures provide the basis for design of mutagenesis experiments to test hypotheses about substrate specificity and the factors that control catalytic rate. These ideas are tested via mutagenesis and activity assays, in some cases both in vitro and in plants.

Weaknesses:

Although improved in a revision, the manuscript could be more clear in explaining the contents of the x-ray structures and how the complexes studied relate to the reactant and product complexes. Some of the figures lack sufficient clarity and description. Some of the claims about the health benefits of tea are not well supported by literature citations.

Reviewer #3 (Public Review):

In the manuscript titled "Structure and Evolution of Alanine/Serine Decarboxylases and the Engineering of Theanine Production," Wang et al. solved and compared the crystal structures of Alanine Decarboxylase (AlaDC) from Camellia sinensis and Serine Decarboxylase (SerDC) from Arabidopsis thaliana. Based on this structural information, the authors conducted both in vitro and in vivo functional studies to compare enzyme activities using site-directed mutagenesis and subsequent evolutionary analyses. This research has the potential to enhance our understanding of amino acid decarboxylase evolution and the biosynthetic pathway of the plant specialized metabolite theanine, as well as to further its potential applications in the tea industry.

Author response:

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

Response to reviewer’s comments

Reviewer #2 (Public Review):

Summary: 

The manuscript focuses on comparison of two PLP-dependent enzyme classes that perform amino acyl decarboxylations. The goal of the work is to understand the substrate specificity and factors that influence catalytic rate in an enzyme linked to theanine production in tea plants.

Strengths: 

The work includes x-ray crystal structures of modest resolution of the enzymes of interest. These structures provide the basis for design of mutagenesis experiments to test hypotheses about substrate specificity and the factors that control catalytic rate. These ideas are tested via mutagenesis and activity assays, in some cases both in vitro and in plants. 

Weaknesses:

Although improved in a revision, the manuscript could be more clear in explaining the contents of the x-ray structures and how the complexes studied relate to the reactant and product complexes. The manuscript could also be more concise, with a discussion section that is largely redundant with the results and lacking in providing scholarly context from the literature to help the reader understand how the current findings fit in with work to characterize other PLP-dependent enzymes or protein engineering efforts. Some of the figures lack sufficient clarity and description. Some of the claims about the health benefits of tea are not well supported by literature citations.

Thank you for your insightful comments on our manuscript and your recognition of the strengths of our study. We understand your concerns about the weaknesses mentioned, and we have addressed them appropriately in the revised manuscript. We acknowledge that the discussion section needs to be improved for conciseness and context. We have revised this part by removing the redundant content. We also acknowledge your comments concerning the clarity and description of some figures. We have revisited these figures and revised them, ensuring they are clear and adequately described. Lastly, concerning the claims about the health benefits of tea, we understand your concern about the lack of supporting citations. We ensure to back such claims with valid literature or, if necessary, omit these statements.

Reviewer #2 (Recommendations For The Authors):

(1) Line 21: Alanine Decarboxylase should not be capitalized.

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(2) Line 31: Grammatical error. Also not clear what "evolution analysis" means here. Revise to "Structural comparisons led us to..."

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(3) Line 34: Revise to "Combining a double mutant of CsAlaDC"

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(4) Line 35: Change word order to "increased theanine production 672%"

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(5) Line 37: meaning unclear. Revise to "provides a route to more efficient biosynthesis of theanine."

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(6) Line 44: I'm not sure that the "health effects" of tea have been proven in placebo controlled studies. And the references provided (2-4 and 5) do not describe original research articles supporting these claims. I would suggest removing these statements from the introduction and at later points in the manuscript.

Thank you for your thoughtful feedback and suggestions. Based on your suggestion, we have removed these statements: "The popularity of tea is determined by its favorable flavor and numerous health benefits (2-4). The flavor and health-beneficial effects of tea are conferred by the abundant secondary metabolites, including catechins, caffeine, theanine, volatiles, etc (5). " As for the subsequent statement: " It has also many health-promoting functions, including neuroprotective effects, enhancement of immune functions, and potential anti-obesity capabilities, among others. " the referenced literature cited can substantiate this conclusion.

(7) Line 58: insert "the" between provided and basis

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(8) Line 100: Not clear what this phrase means, "As expected, CsSerDC was closer to AtSerDC" Please clarify - closer to what?

We apologize for any confusion caused by the unclear phrasing. When referring to "CsSerDC was closer to AtSerDC," we intended to convey that CsSerDC exhibits a higher degree of sequence homology with AtSerDC than it does with the other enzymes evaluated in our investigation. However, a 1.29% difference between 86.21% and 84.92% in amino acid similarity is not statistically significant (Figure 1B and Supplementary table 1 in the original manuscript), we have deleted the relevant descriptions in the revised manuscript.

(9) Line 112: "were constructed into" makes no sense. It would be better to say the genes for the proteins of interest were inserted into the overexpression plasmid.

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(10) Line 115: missing the word "the" between generated and recombinant

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(11) Line 121: catalyze not catalyzed

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(12) Lines 129 and 130: The reported Km values are really large - in the mM range. Do these values make sense in terms of the available concentrations of the substrates inside the cell?

The content of alanine in tea plant roots ranges from 0.28 to 4.18 mg/g DW (Yu et al., 2021; Cheng et al., 2017). Correspondingly, the physiological concentration of alanine is 3.14 mM to 46.92 mM, in tea plant roots. The content of serine in plants ranges from 0.014 to 17.6 mg/g DW (Kumar et al., 2017). Correspondingly, the physiological concentration of serine is 0.13 mM to 167.48 mM in plants. Therefore, in this study, the Km values are within the range of available substrate concentrations inside the cell.

Yu, Y. et al. (2021) Glutamine synthetases play a vital role in high accumulation of theanine in tender shoots of albino tea germplasm "Huabai 1". J. Agric. Food Chem. 69 (46),13904-13915.

Cheng, S. et al. (2017) Studies on the biochemical formation pathway of the amino acid L-theanine in tea (Camellia sinensis) and other plants. J. Agric. Food Chem. 65 (33), 7210-7216.

Kumar, V. et al. (2017) Differential distribution of amino acids in plants. Amino Acids. 49(5), 821-869.

(13) Line 211: it is unclear what the phrase "as opposed to wild-type" means. Please clarify.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We intend to communicate that the wild-type CsAlaDC and AtSerDC demonstrate decarboxylase activity, while the mutated proteins have experienced a loss of decarboxylation activity. We have already modified this concern in the revised version of the manuscript.

(14) Line 222: residues not residue

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(15) Line 227 and Figure 4B: It is not clear what the different sequence logos mean in this part of the figure. The caption is too brief and not helpful. And the sentences describing this figure panel are also not sufficiently clear.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have provided a more detailed explanation of this section in the revised manuscript and added additional annotations in the figure caption to provide further clarity.

(16) Lines 233 and 234: "in the substrate specificity" is awkwardly worded. I would revise to "in selective binding of the appropriate substrate."

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have meticulously revised the description of this section.

(17) Line 243: a word is missing in this sentence - but I can't figure out the intended meaning or what the missing word is. Rephrase to improve clarity.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have revised this sentence to: " These findings indicate the essential role of Phe106 in the selective binding of alanine for CsAlaDC. "

(18) Line 255: The "expression system...was carried out" is not correct. I would say the expression system was used - but you probably also want to rearrange the sentences to more directly say what it was used for. Later, the word "the" is also missing.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have revised this sentence to: "To further verify that Phe106 of CsAlaDC and Tyr111 of AtSerDC were key amino acid residues determining its substrate recognition in planta, we employed the Nicotiana benthamiana transient expression system. "

(19) Line 273: use "understand" instead of "elucidate" and instead of "we proposed a prediction test:" say "we designed a test of the prediction that..."

Thank you very much for your careful reading of the manuscript. We have revised this sentence to: “In light of this observation, we postulated a hypothesis:”

(20) Line 301: I don't think "effectuate" is a word. Replace with something else.

Thank you very much for your careful reading of the manuscript. We have revised the sentence as: " The biosynthetic pathway of theanine in tea plants comprises two consecutive enzymatic steps: alanine decarboxylase facilitates the decarboxylation of alanine to generate EA, while theanine synthetase catalyzes the condensation reaction between EA and Glu to synthesize theanine. "

(21) Line 307: replace "activity" with "ability"

Thank you very much for your careful reading of the manuscript. We have corrected it in the revised manuscript.

(22) Line 322: I didn't find the discussion very useful. Much of it is simply a recap of the results - which is not necessary. The structural comparisons are overly descriptive without providing appropriate rationale or topic sentence structure so that the reader understands why certain details are emphasized. I think the manuscript would be much stronger if this section were not included or integreted more concisely into the results section where appropriate.

Thank you for your constructive comments. We understand your concerns about the discussion section of our manuscript. We acknowledge that the discussion section has redundancies with the result. In response to this, we have revised this section to eliminate unnecessary repetition of the results.

(23) Line 369: "an amino acid devoid of the hydroxyl moiety present in Lys" - what does this mean? Lys does not have a hydroxyl functional group. Please correct so that the sentence makes sense.

Thank you very much for your careful reading of the manuscript. This sentence states that the amino acid occupying the corresponding position in CsAlaDC is Phe, which lacks one hydroxyl functional group as compared to Lys. We have made modifications to the sentence as follows: "In contrast, the equivalent position in CsAlaDC is occupied by Phe, an amino acid lacking the hydroxyl group. This substitution enhances the hydrophobic nature of the substrate-binding pocket. "

(24) Line 370: "This structural nuance portends a predisposition for CsAlaDC to select the comparatively hydrophobic amino acid alanine as its suitable substrate." This sentence also makes no sense - please revise to use simpler language so the meaning is more clear.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have revised the sentence as follows: " Consequently, CsAlaDC demonstrates a unique predilection, selectively binding Ala (an amino acid with comparatively hydrophobic properties) as its preferred substrate."

(25) Lines 376-384: This section makes several references to "catalytic rings." I have no idea what this term means? If the authors mean a loop structure in the enzyme - please use the term "loop"

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have corrected it in the revised manuscript.

(26) Line 396-397: The authors reference data that is not shown in the manuscript. Either show the data in the results section or do not mention.

Thank you for your insightful comment regarding the unshown data referenced in the manuscript. We have included Supplementary figure 9 in the revised manuscript to display this data.

(27) Line 445-446: what is "mutation technology" - if the authors mean site-directed mutagenesis - please use the simpler and more recognizable terminology.

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have revised the sentence as follows: "Based on the findings of this study, site-directed mutagenesis can be employed to modify enzymes involved in theanine synthesis. This modification enhances the capacity of bacteria, yeast, model plants, and other organisms to synthesize theanine, thereby facilitating its application in industrial theanine production."

Reviewer #3 (Public Review):

In the manuscript titled "Structure and Evolution of Alanine/Serine Decarboxylases and the Engineering of Theanine Production," Wang et al. solved and compared the crystal structures of Alanine Decarboxylase (AlaDC) from Camellia sinensis and Serine Decarboxylase (SerDC) from Arabidopsis thaliana. Based on this structural information, the authors conducted both in vitro and in vivo functional studies to compare enzyme activities using site-directed mutagenesis and subsequent evolutionary analyses. This research has the potential to enhance our understanding of amino acid decarboxylase evolution and the biosynthetic pathway of the plant specialized metabolite theanine, as well as to further its potential applications in the tea industry.

Thank you very much for taking the time to review this manuscript. We appreciate all your insightful comments.

Reviewer #3 (Recommendations For The Authors):

The additional material added by the authors addresses some of the previously raised questions and enhances the manuscript's quality. However, certain critical issues we pointed out earlier remain unaddressed. Some of the new data also raises new questions. To provide readers with more comprehensive data, the authors should include additional quantitative data and convert the data presented in the reviewer's comments into supplemental figure format.

Thank you for acknowledging the improvements in the revised manuscript and providing further valuable feedback. We understand your concern about the critical issues that have not been fully addressed and the new questions raised by some of the newly added data. We have strived to address these issues with additional analysis and clarification in our subsequent revision. Regarding your suggestion for more quantitative data and converting the data mentioned in the reviewer's comments into a supplemental figure format, we agree that this would provide a more comprehensive view of the results. We have reformatted the relevant data into supplemental figures to enhance the clarity and accessibility of information. We are grateful for the time and effort you have dedicated to improving our manuscript.

* Page 5 & Figure 1B

"As expected, CsSerDC was most closed to AtSerDC, which implies that they shared similar functions. However, CsAlaDC is relatively distant from CsSerDC."

: In Figure 1B, CsSerDC and AtSerDC are in different clades, and this figure does not show that the two enzymes are closest. To provide another quantitative comparison, please provide a matrix table showing amino acid sequence similarities as a supplemental table. 

Comment: I don't believe that a 1.29% difference between 86.21% and 84.92% in amino acid similarity is statistically significant. Although the authors have rephrased the original sentence, it's improbable that this small 1.29% difference can explain the observed distinction.

Many thanks. We have carefully considered your comments. Indeed, the 1.29% difference in amino acid similarity cannot reflect the functional difference between the AlaDC and SerDC proteins. We have deleted the relevant descriptions in the revised manuscript.

* Page 6, Figure 2, Page 23 (Methods)

"The supernatants were purified with a Ni-Agarose resin column followed by size-exclusion chromatography."

: What kind of SEC column did the authors use? Can the authors provide the SEC elution profile comparison results and size standard curve?

Comment: The authors should include the SEC elution profiles as a supplemental figure or incorporate them as a panel in Figure 2. Furthermore, they should provide a description of the oligomeric state of each protein in this experiment. Additionally, there is a significant difference between CsSerDC (65.38 mL) and CsAlaDC (74.37 mL) elution volumes. Can this difference be explained structurally? In comparison to the standard curve of molecular weight provided by the authors, it appears that these proteins are at least homo-tetramers, which contradicts the description in the text. This should be re-evaluated and clarified.  

Thank you very much for your careful reading of the manuscript and valuable suggestions. We have included the SEC elution profile in Supplemental figure 1A and added descriptions of the oligomeric states of proteins in the revised manuscript. CsSerDC was eluted at 65.38 mL, corresponding to a molecular weight of 292 kDa, which is five times the monomeric protein (54.7 kDa). However, due to the absence of CsSerDC crystal structure, it remains uncertain whether the protein forms a pentamer. AtSerDC was eluted at 72.25 mL, with a corresponding molecular weight of 155 kDa, which is 3.3 times the monomer (47.3 kDa). CsAlaDC was eluted at 74.37 mL, with a corresponding molecular weight of 127 kDa, which is 2.7 times the monomer (47.3 kDa). The elution profiles suggest that AtSerDC and CsAlaDC potentially exist in homotrimeric form. This observation stands in contradiction to our subsequent findings where the protein manifests in a dimeric structure. A plausible explanation could be the non-ideal spherical shape of the protein. Under such circumstances, the hydrodynamic radius of the protein could supersede its actual size, potentially leading to an overestimation of the molecular weight on the size-exclusion chromatography [ref].

References:

Burgess, R. R. (2018) A brief practical review of size exclusion chromatography: Rules of thumb, limitations, and troubleshooting. Protein Expression and Purification. 150, 81-85.

Erdner J. M., et al. (2006) Size-Exclusion Chromatography Using Deuterated Mobile Phases. Journal of Chromatography A. 1129(1):41–46.

* Page 6 & Page 24 (Methods)

"The 100 μL reaction mixture, containing 20 mM substrate (Ala or Ser), 100 mM potassium phosphate, 0.1 mM PLP, and 0.025 mM purified enzyme, was prepared and incubated at standard conditions (45 {degree sign}C and pH 8.0 for CsAlaDC, 40 {degree sign}C and pH 8.0 for AtSerDC for 30 min)."

(1) The enzymatic activities of CsAldDC and AtSerDC were measured at two different temperatures (45 and 40 {degree sign}C), but their activities were directly compared. Is there a reason for experimenting at different temperatures?

(2) Enzyme activities were measured at temperatures above 40{degree sign}C, which is not a physiologically relevant temperature and may affect the stability or activity of the proteins. At the very least, the authors should provide temperature-dependent protein stability data (e.g., CD spectra analysis) or, if possible, temperature-dependent enzyme activities, to show that their experimental conditions are suitable for studying the activities of these enzymes.

Comment: I appreciate the authors for including temperature-dependent enzyme activity data in their study. However, it remains puzzling that plant enzymes were tested at a physiologically irrelevant temperature of 40 and 45 degrees Celsius. Additionally, it may not be appropriate to directly compare enzyme activity measurements at different temperatures. Furthermore, the data at 45 degrees in panel A appears to be an outlier, which contrasts with the overall trend observed in the graph.

We appreciate your point regarding the testing temperatures for plant enzymes. We fully appreciate the importance of conducting experiments under physiologically relevant conditions. But the intent behind operating at these elevated temperatures was to assess the thermal stability of the enzymes, which can be a valuable characteristic in certain applications, such as industrial production processes, and does not necessarily reflect their physiological conditions. Our findings indicate that CsAlaDC exhibits its peak activity at 45 °C. This result aligns with previously reported data in the literature [Bai, P. et al. (2021) figure 4e], thus bolstering our confidence in the reliability of our experimental outcomes.

Author response image 1.

Relative activity of CsAlaDC at different temperatures.

* Pages 6-7 & Table 1

(1) Use the correct notation for Km and Vmax. Also, the authors show kinetic parameters and use multiple units (e.g., mmol/L or mM for Km).

(2) When comparing the catalytic efficiency of enzymes, kcat/Km (or Vmax/Km) is generally used. The authors present a comparison of catalytic activity from results to conclusion. A clarification of what results are being compared is needed.

Comment: The authors are still comparing catalytic efficiency solely based on the Vmax values. As previously suggested, it would be advisable to calculate kcat/Km and employ it for comparing catalytic efficiencies. Furthermore, based on the data provided by the authors, I conducted a rough calculation of these catalytic efficiencies and did not observe a significant difference, which contrasts with the authors' statement, "These findings indicated that the catalytic efficiency of CsAlaDC is considerably lower than that of both CsSerDC and AtSerDC." This discrepancy requires clarification.  

We want to express our sincere appreciation for your meticulous review and constructive suggestions. We understand the importance of accurately comparing catalytic efficiencies using Kcat/Km values, rather than solely relying on Vmax values. Following your suggestion, we recalculated Kcat/Km to reanalyze our results. The computed Kcat/Km for CsSerDC and AtSerDC are 152.7 s-1 M-1 and 184.6 s-1 M-1, respectively. For CsAlaDC, the calculated Kcat/Km is 55.7 s-1 M-1. Therefore, the catalytic efficiency of CsSerDC and AtSerDC is approximately three times that of CsAlaDC.  What we intended to convey was that the Vmax of CsAlaDC is lower than that of CsSerDC and AtSerDC.  Our description in the manuscript was not accurate, and we have addressed this in the revised version.

* Pages 9 & 10

"This result suggested this Tyr is required for the catalytic activity of CsAlaDC and AtSerDC."

: The author's results are interesting, but it is recommended to perform the experiments in a specific order. First, experiments should determine whether mutagenesis affects the protein's stability (e.g., CD, as discussed earlier), and second, whether mutagenesis affects ligand binding (e.g., ITC, SPR, etc.), before describing how site-directed mutagenesis alters enzyme activity. In particular, the authors' hypothesis would be much more convincing if they could show that the ligand binding affinity is similar between WT and mutants.

Comments: While it is appreciated that you have included CD and UV-vis absorption spectra data, it would be more beneficial to provide quantitative data to address the previously proposed binding affinity. I also recommend presenting the data mentioned in the reviewer's comments as a supplementary figure for better clarity and reference.  

Thank you for your valuable feedback and suggestions. I agree that providing quantitative data would lend more support to our findings and better address the proposed binding affinity.

It is generally acknowledged that proteins complexed with PLP exhibit a yellow hue, and the ligand PLP forms a Schiff base structure with the ε-amino group of a lysine residue in the protein, with maximum absorbance around 420 nm. However, during our protein purification process, we observed that the purified protein retained its yellow coloration, even when PLP wasn't introduced into the purification buffer. Subsequent absorbance measurements revealed that the protein exhibited absorbance within the aforementioned wavelength (420 nm) (the experimental results are shown in the following figures), implying an inherent presence of the PLP ligand within the protein. This could have resulted from binding with PLP during the protein's expression in E. coli. Consequently, due to this inseparability between the protein and the ligand, obtaining quantitative data through experimental means becomes unfeasible.

Author response image 2.

(A) Absorption Spectra of CsAlaDC (WT) and CsAlaDC (Y336F). (B) Absorption Spectra of AtSerDC (WT) and AtSerDC (Y341F).

Regarding your suggestion about presenting the data mentioned in the reviewer's comments as a supplementary figure, we agree that it is an excellent idea. We have prepared supplementary figure 7 and supplementary figure 8 accordingly, ensuring that they present the required data.

* Page 10

"The results showed that 5 mM L-DTT reduced the relative activity of CsAlaDC and AtSerDC to 22.0% and 35.2%, respectively"

: The authors primarily use relative activity to compare WT and mutants. Can the authors specify the exact experiments, units, and experimental conditions? Is it Vmax or catalytic efficiency? If so, under what specific experimental conditions?

Response: "However, due to the unknown mechanism of DTT inhibition on protein activity, we have removed this part of the content in the revised manuscript."

Comment: I believe this requires a more comprehensive explanation rather than simply removing it from the text.  

Although we have observed that DTT is capable of inhibiting enzyme activity, at present, we are unable to offer a comprehensive explanation for the inhibitory effect of DTT on enzyme activity in terms of its structural and catalytic mechanisms. Further research is required to elucidate the mechanism of action of DTT. It is worth noting, however, that our study does not emphasize investigating the specific inhibitory mechanisms of DTT on enzyme activity. Furthermore, the existing findings do not provide an adequate explanation for the observed phenomenon, leading us to exclude this particular aspect from the content.

* Pages 10-12

: The identification of 'Phe106 in CsAlaDC' and 'Tyr111 in AtSerDC,' along with the subsequent mutagenesis and enzymatic activity assays, is intriguing. However, the current manuscript lacks an explanation and discussion of the underlying reasons for these results. As previously mentioned, it would be helpful to gain insights and analysis from WT-ligand and mutant-ligand binding studies (e.g., ITC, SPR, etc.). Furthermore, the authors' analysis would be more convincing with accompanying structural analysis, such as steric hindrance analysis.

Comment: While it is appreciated that you have included UV-vis absorption spectra data, it would be more beneficial to provide quantitative data to address the previously proposed binding affinity. I also recommend presenting the data mentioned in the reviewer's comments as a supplementary figure for better clarity and reference.  

Response: Thank you for your valuable feedback and suggestions. Given that the protein forms a complex with PLP during its expression in E. coli and cannot be dissociated from it, obtaining quantitative data via experimental protocols is rendered impracticable.

Author response image 3.

(A) Absorption Spectra of CsAlaDC (WT) and CsAlaDC (F106Y). (B) Absorption Spectra of AtSerDC (WT) and AtSerDC (Y111F).

Mutant proteins and wild-type proteins exhibited absorption bands at 420 nm, suggesting the formation of a Schiff base between PLP and the active-site lysine residue.

Regarding your suggestion about presenting the data mentioned in the reviewer's comments as a supplementary figure, we have prepared supplementary figure 7 and supplementary figure 8 accordingly, ensuring that they present the required data.

Author response:

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

Public Reviews:

Reviewer #1:  

Overall, the conclusions appear appropriately supported by the data, and the data appear of high quality.

Strengths:

The particular strengths of the paper include an impressive combination of genomic and imaging-based approaches and insightful genetically engineered cell systems. The manuscript reports interesting and potentially important findings. The text is generally very well written, the ideas are clearly explained, and the reasoning is easy to follow.

Weaknesses:

The main weakness seems to be that the heat and ethanol shock approaches likely elicit pleiotropic effects, and therefore it is a challenge to test the causal relationship between various observations. Nevertheless, even as indirect effects might contribute to some of the authors' observations, the results are definitively worth reporting.  

We agree that these two proteotoxic stresses can impact cell physiology in multiple ways and discuss this on lines 132-143 and 500-519. Moreover, in this revision we have more rigorously quantified the extent of proteotoxic stress elicited by the 39°C heat shock and 8.5% ethanol stress (Figure 1E; see response 1 to Reviewer 2). We have additionally added new Figure 2 that reveals an important difference in the way Hsf1 and its negative regulator, the Hsp70 co-chaperone Sis1, respond to HS and ES. This difference is evident at two different intensities for each stress as described in more detail below (see response 1 to Reviewer 2).

Presentation of some of the data could be improved.

We agree and have made improvements/data additions to multiple figures: Figure 1E; Figures 3A, B; Figures 4A, B; Figure 7 (data drawn from original Fig. 6 and Fig. 6 – fig. suppl. 1 and reorganized); Fig. 8B; Figure 9; Figure 10. Corresponding enhancements to the supplemental figures have been made as well. 

Reviewer #2:  

(1) The central finding of the study highlights the different dynamics of Hsf1, Pol II, and gene organization in response to heat shock versus ethanol stress. However, one important limitation to consider is that the two chosen conditions may not be directly comparable. For a balanced assessment, the authors should ideally expose yeast to various ethanol concentrations and different heat shock temperatures, ensuring the observed differences stem from the nature of the stressor rather than suboptimal stress intensity. At the very least, an additional single ethanol concentration point on each side of 8.5% should be investigated to ensure that 8.5% is near the optimum. In fact, comparing the number of Hsp104 foci in the two conditions in Fig. 1E and F suggests that the yeast is likely experiencing different intensities of stress for the chosen heat shock condition and ethanol concentration used in this study.

We thank the reviewer for this important suggestion. In this revision, we have included an enhanced analysis of the yeast cellular response to each of these stresses. As illustrated in revised Figure 1, the two stresses used throughout this study – 39°C heat shock and 8.5% ethanol stress – both elicit a proteotoxic response, as assayed by the de novo formation of Hsp104 clusters. While 10 min exposure to 8.5% ethanol results in the formation of multiple discrete (spherical) foci, a 10 min exposure to the elevated temperature leads the appearance of multiple, largely diffuse Hsp104 clusters, some of which are spherical (new Fig. 1D). The difference in morphology notwithstanding, we have attempted to quantify these clusters using Imaris v. 10.0.1 image analysis software; the results are depicted in Fig. 1E. Such quantification suggests that 8.5% ethanol elicits a more intense stress than exposure to 39°C. A caveat is that it is unclear whether diffuse Hsp104 clusters are comparable to compact Hsp104 foci (see response 3 below).

Beyond the apparent difference in intensity, a new analysis presented in new Figure 2 reveals that heat shock, elicited by temperature upshift to either 39°C or 42°C, induces relocalization of the J-protein Sis1 – a key negative regulator of Hsf1 – from the nucleoplasm to the nucleolar periphery. Sis1’s perinucleolar ring localization agrees with previous findings of 39°C heat-shocked cells (Feder et al., 2021). Ethanol stress, whether 5% or 8.5%, initially causes Sis1 to relocalize diffusely throughout the nucleus and cytosol. At 10 min, Sis1 localizes to the periphery of the nucleus, thereby providing a marked contrast to what is observed in response to heat shock. These new results are described on lines 174-191.

Taking these two observations together, we asked whether a less severe ethanol stress (5%) would induce Hsf1 puncta. It does, and as rapidly as 8.5% ethanol (data are presented in revised Figure 8-figure supplement 1). Interestingly, in the presence of 5% ethanol, Hsf1 puncta begin to dissolve at 30 min. This strongly contrasts with the case when cells are exposed to 8.5% ethanol (Figure 8; Figure 8-figure supplement 1). As we state in this revision (lines 414-424), the sustained presence of condensates that we originally observed is likely the consequence of the intensity of the proteotoxic stress elicited by exposure to 8.5% ethanol; analogous responses to these two stress conditions have been observed before (lines 495-501). 

(2) A second significant concern is the use of the term "Hsf1 condensate". Chowdhary et al.'s 2022 Molecular Cell study highlighted an inhomogeneous distribution and rapid dynamics of Hsf1 clustering upon heat shock, with sensitivity to 1,6-hexandiol, which is interpreted as evidence for condensation by LLPS. However this interpretation has been criticized severely by McSwiggen et al. Genes Dev 2019 and Mussacchio EMBO J 2022. It is important to mention that 1,6-hexandiol is known to affect chromatin organization (Itoh et al. Life Science Alliance 2021). Describing such clusters as 'condensates' without further experimental evidence is premature.  

While we appreciate and largely agree with the point made by this reviewer, we prefer to maintain the term “condensate”. Banani et al (2017) originally defined “biomolecular condensate” to mean selforganized membrane-free compartments that concentrate specific biomolecules. It was never meant to imply LLPS although its widespread use in the literature has led to that implication. We clarify our use of this term on lines 99-104.   

(3) Figure 1: Why does ethanol stress at 0 min display a larger number of Hsp104 foci per cell than heat shock at the same time? How are foci defined by the authors? In Fig. 1D, there are many smaller puncta. A comparative assessment of the number and size of foci for heat shock and ethanol stress would be beneficial.

We thank the reviewer for raising this point and have addressed it as follows.  First, we repeated the assay with a different strain (DPY1561) and increased the number of cells assayed from 40 to 200. This larger sample size created the same T=0 baseline for both stresses (Figure 1E). Second, we define Hsp104 foci as diffraction-limited structures with a diameter of ~0.4 µm (lines 747-749).  Third, employing Imaris v. 10.0.1, we quantified foci size (= volume) and a summary graph has been added to Figure 1E that also displays the number of foci per cell. In the legend to this figure, we point out that to conduct this analysis we assumed that the diffuse Hsp104 clusters seen in HS cells are comparable to the compact Hsp104 foci in ES cells (lines 1169-1171). 

(4) Figure 2: Selecting a housekeeping gene with consistent expression levels is crucial for meaningful qPCR analysis. Do SCR1 mRNA levels fluctuate during heat shock or ethanol stress?  

We thank the reviewer for this question. In revised Figure 3 – figure supplement 1C we provide a new graph (reproduced here) revealing that the levels of SCR1 do not significantly change under either heat shock or ethanol stress relative to the non-stressed control (0 min). One-way ANOVA analysis was performed for both HS and ES and p values were 0.094 and 0.083, respectively (calculated using GraphPad Prism 8).

(5) Additionally, certain genes, such as TMA10 and SSA4, lack visible bars at time 0. Are these levels undetectable? The varying y-axis scales are confusing; presenting data as relative fold changes could offer a clearer perspective.

Transcript levels for all genes evaluated here are detectable, even in the basal unstressed state. They are not visible on the histogram for certain genes at T= 0 due to the prodigious fold-increase in RNA elicited by heat shock.  However, to address this concern, we have added a bar graph inset displaying basal transcript levels for each gene in revised Figure 3. We reproduce data for SSA4 and TMA10 in the graphs below. In addition, we present transcript levels in new Figure 3 - figure supplement 1 for cells subjected to ethanol stress to allow a better appreciation of their increase over time. 

Author response image 1.

(6) Line 239: The evidence for chromatin compaction is unconvincing. An increase in H3 occupancy by ChIP might indicate a reduction in histone exchange dynamics but may not relate to overall chromatin compaction. The authors use H2A-mCherry to suggest a decrease in chromatin volume, but this data is not persuasive. Did the authors observe any changes in nuclear size? Perhaps quantifying chromatin compaction more directly, using signal intensity per volume, would be informative.

To address this concern, we attempted to quantify integrated density for H2A-mCherry using Image J software. While the volume decreased for both stresses, the integrated density only increased for ethanol stress. We speculate that this may be due to photobleaching which has been reported for heat shock. The combination of heat and acidic pH contribute to loss of fluorescence signal (Alkaabi et al., 2005). While the integrated density supports the idea of global chromatin compaction in the ethanol stress condition, given the above concerns with the HS sample we elected to not present these data.

(7) Line 340: The claim of a "strong spatiotemporal correlation" isn't evident from the data. Could correlation coefficients be provided? There is potential anti-correlation in Fig. 6 - Figure Supplement 1C.

We thank the reviewer for this excellent suggestion. We now present an analysis of the correlation between HSP104 – HSP12 coalescence and HSP104 transcription for both HS and ES time courses, using single cell data of Figures 7D, 7E and Figure 7- suppl. 1D.  This analysis is presented in new Figure 7F.

(8) Figure 8: The WT data in Fig 8 seem inconsistent with Fig. 4 (e.g. the interaction frequency for HSP104 and SSA2). Are these fluctuations between experiments, or are they side effects of IAA treatment? The use of ethanol as an IAA solvent vehicle raises concerns. It would be beneficial if the authors could demonstrate that 1.7% ethanol in the control does not induce ethanol stress.

We acknowledge that there existed an inconsistency in the magnitude of intergenic interaction frequencies reported in the two experiments for HSP104 and SSA2. Some of this might be attributed to the fact that different strains were used, W303-1B in Figure 4 and LRY016 (W303-1B; LEU2::pGPD1osTIR1) in Figure 8. Nonetheless, in each experiment there was a prodigious fold-increase in interaction frequency over the no stress (T= 0 min) control for both HS and ES conditions and moreover, in each experiment the magnitude of this interaction was greater for the 2.5 min HS sample vs. the 10 min ES sample. However, to obviate this concern, we have removed the HSP104-SSA2 analysis from Figure 9 (corresponds to original Fig. 8).

Regarding the second point, we cannot entirely rule out the concern that the 1.7% ethanol vehicle might impact 3C interaction frequencies. It is unlikely to be significant, however, given that most other pairwise tests evaluated in the two experiments (Figs. 5 and 9) resulted in similar 3C values. In particular, there was no consistent trend towards higher (or lower) interaction frequencies in the IAA experiment of Fig. 9.  

Reviewer #3:  

This is an interesting manuscript that builds off of this group's previous work focused on the interface between Hsf1, heat shock protein (HSP) mRNA production, and 3D genome topology. Here the group subjects the yeast Saccharomyces cerevisiae to either heat stress (HS) or ethanol stress (ES) and examines Hsf1 and Pol II chromatin binding, Histone occupancy, Hsf1 condensates, HSP gene coalescence (by 3C and live cell imaging), and HSP mRNA expression (by RT-qPCR and live cell imaging). The manuscript is well written, and the experiments seem well done, and generally rigorous, with orthogonal approaches performed to support conclusions…While identifying a mechanistic basis for the results [presented here] would be a tough task perhaps beyond the scope of this study, it would nevertheless be helpful to place these results in context with a series of other studies…importantly, this work left out PMID: 32015439 (HSF1 phase transition mediates stress adaptation and cell fate decisions) which is particularly relevant considering that it shows that it is human HSF1 condensate resolution rather than simple condensate formation that is associated with HSF1 transcriptional activity - which is similar to the findings here with this particular dose of HS resulting in resolution and high transcriptional activity versus ES resulting in resolution failure and lower activity. 

We thank the Reviewer for pointing out this oversight. In this revision, we cite Gaglia et al., 2020 and several others reporting HSF1 foci formation in human cells exposed to heat shock. The single cell analysis of Gaglia et al argued that dissolution of large HSF1 foci (aka “nuclear stress bodies”), typically several µm in diameter and localized over satellite III DNA repeats (Jolly et al., 1997, 2002), correlates with HSP gene activation. Importantly, these condensates are postulated to act as reservoirs of HSF1, sequestered away from HSP genes (Gaglia et al., 2020).  In contrast, Zhang et al., 2022 has shown that human HSF1 inducibly forms small condensates (~300 nm) that localize over HSP genes and whose formation directly correlates with HSP gene activation (we discuss the Jolly, Gaglia and Zhang findings on lines 382-394). Likewise, our work shows that in yeast, Hsf1 inducibly forms small, dynamic clusters that colocalize with HSR genes within 2.5 min of exposure to elevated temperature; these dissolve ~20-60 min later (Figure 8 and Figure 8-supp. 1). In concert with Hsf1 condensate formation, HSR gene repositioning and transcription/ Pol II recruitment are likewise evident within 2.5 min. Therefore, in HS cells there exists coordinate induction of condensate formation, Pol II recruitment, transcription and intergenic interactions (for a detailed kinetic analysis of HSR gene interactions, see Figures 5 and 6 of Chowdhary et al, 2017).  This tight temporal relationship is absent in ethanol stressed cells (Figures 3, 4, 5, 6, 7, 8; summarized in Figure 10 and Table 1).

It is also worth noting that the stresses themselves are quite different - ethanol can be used as a carbon source and so beyond inducing proteotoxic stress, the yeast are presumably adapting to this distinct metabolic state. Basically, it is not clear whether these differences are due to the dose of stress, versus we are looking at an early timepoint as ES initiates a genome-wide chromatin restructuring and gene expression reprogramming that goes beyond a response to proteotoxic stress. This reviewer is not suggesting a barrage of new experiments, but perhaps discussion points to contextualize results.

We thank the reviewer for this suggestion and in our revised manuscript discuss these issues (lines 414424 and 486-498 [5% vs. 8.5% ethanol]; lines 500-519 [ethanol as a metabolite]).

Recommendations for the authors:

Reviewer #1:

(1) In Figure 1E, the number of foci in control (0 min) cells is very different for the two conditions. Could the authors clarify/check this? Based on the mean numbers at time point 0, the control cells for the ethanol treatment already contain about 10-20 Hsp104 foci, compared to around 5 foci per cell in the control for heat shock.

We thank the reviewer for raising this point and have repeated the assay with a different strain (DPY1561).  And as shown in Figure 1E, have confirmed that the control samples have similar number of foci.  

(2) In the same Figure 1E, is the P-value relative to the control or the same time point in the other treatment? A comparison across treatments would be necessary to support the claim in lines 168-171 of the text.

The statistical analysis (Mann Whitney test) was performed by comparing each stress timepoint to the no stress control. We clarify this in the figure legend. 

(3) In Figure 1D, the heat-shock condition shows the same cells that are used in the control, but the cells in the ethanol-shock condition are different. This is a bit visually misleading compared to the experimental setup shown in panel 1C. The authors could show the control cells for the ethanol condition as well.

We thank the reviewer for this excellent suggestion and have added the 0 min image for the ethanol stress conditions.

(4) In Figure 7B adding images at 60min would help underscore the point that the condensates are stable in ethanol shocked cells.

We appreciate this suggestion as well and have included a 60 min timepoint for both stresses (Figure 8B). 

Reviewer #2:

(1) Line 113: Has it not been established that yeast Hsf1 is constitutively trimeric?

In yeast, only a fraction of Hsf1 is thought to be constitutively trimeric and it is this species that binds high-affinity HSEs even under non-stressful conditions (Giardina & Lis, 1995; Pincus et al., 2018). We have added this clarification to the text (lines 121-123). 

(2) Ethanol can precipitate proteins, especially in rich media like YPD. Did the authors notice any protein precipitation? If yes, how do they account for effects due to nutrient loss by precipitation?

This is an interesting point, but we did not notice any precipitates in either rich or synthetic liquid media containing 8.5% (v/v) ethanol for any of the time points used in the experiments.

(3) Figure 3: The figure appears incomplete. Can enhancer, promoter, coding region, and 3'UTR be shown consistently for all genes examined?

In response to this point, we have simplified this figure (new Fig. 4) by uniform presentation of factor occupancy at enhancer, promoter, and coding region loci for all but one of the genes evaluated. For HSP12 (330 bp), we were unable to distinguish promoter from coding region since the average sonicated chromatin fragment obtained using a Bioruptor is ~300 bp. Therefore, we evaluated only the HSP12 coding region for Pol II and histone H3 occupancy. 

(4) Figure 4: The comparison between heat shock at 2.5 min and ethanol stress at later points is puzzling. Why not use consistent time points as in Fig. 3?

Time points for the two stresses examined in this figure (new Fig. 5) were selected to represent times of peak intergenic interaction between HSR genes. These times were derived from our earlier analysis of 3C interactions during a heat shock time course (Figs. 5, 6 of Chowdhary et al., 2017) and ES data presented in this study, including Fig. 4 (Pol II ChIP time course) and Fig. 6 (3C time course). Data presented in Figs. 5 and 6 are consistent with the notion that intergenic interactions in cells subjected to ethanol stress are delayed relative to those observed in heat shocked cells, peaking in most cases at ~10 min (vs. ~2.5 min for heat stress (Chowdhary et al., 2017)).  

(5) Figure 5: Fig. 5B top panel seems to show color inconsistencies for bars at 0 and 120 min. Also, the xaxis on the top left panel seems to have a typo; should it read "10," not "0?"

We thank the reviewer for the observation. We changed the graphs in new Figure 6 to display the same color for all time points.  We also fixed the typo. 

(6) Line 302: The evidence presented supports maximal mRNA levels, but the claim of "maximal transcription" requires support from nascent RNA analysis.

We agree that RT-qPCR measures mRNA abundance, not nascent transcription. We have changed the text to refer to “transcript levels” where pertinent (lines 301-302; 1331-1332).

(7) How long do loci remain coalescent during heat shock versus ethanol stress? Both 3C and imaging analyses do not differentiate between frequency and duration, which seems essential for understanding interaction dynamics.

We thank the reviewer for this excellent question. In new Fig. 7D,E (data drawn from Fig. 6 – fig. suppl. 1), HSR gene coalescence detected in single cells over a HS or ES time course is charted.  Interpretable data exist for a small number of cells. Moreover, for both HS and ES states, in certain cells coalescence between the representative Hsf1 target genes HSP104 and HSP12 dissolves and then reappears. With this caveat in mind, the data suggest that HSP104-HSP12 coalescence can last at least 15 min in HS cells and up to 30 min in ES cells. We have not emphasized this point in the manuscript since a far more comprehensive analysis – beyond the scope of this study – is required.

(8) For longer analyses, how do the authors accommodate potential ethanol concentration changes due to evaporation?

For liquid cultures, we relied on maintaining minimal changes in the vapor pressure within the experimental vessel; to facilitate that, flasks were tightly covered to minimize evaporation and temperature was kept at 25°C. For most molecular analyses (RT-qPCR, ChIP, 3C), we confined our analysis to the first 60 min. For microscopy, the samples were encased within a concave slide, covered by a coverslip, as illustrated below. In addition, to tightly seal the coverslip on the slide we used petrolatum.  This arrangement minimized evaporation.

Author response image 2.

(9) Figure 9: This legend seems to have an incomplete sentence: "(represented using ...)."

We have substituted an entirely new model in this revised manuscript (new Figure 10) that omits the use of an ellipsis. (We had used it to symbolize a delay in the appearance of HSR gene transcription in ES cells.)

References  

Alkaabi, K. M., Yafea, A., & Ashraf, S. S. (2005). Effect of pH on thermal- and chemical-induced denaturation of GFP. Applied Biochemistry and Biotechnology, 126(2), 149–156. https://doi.org/10.1385/ABAB:126:2:149

Chowdhary, S., Kainth, A. S., & Gross, D. S. (2017). Heat Shock Protein Genes Undergo Dynamic Alteration in Their Three-Dimensional Structure and Genome Organization in Response to Thermal Stress. Molecular and Cellular Biology, 37(24), 1–23. https://doi.org/10.1128/mcb.00292-17

Feder, Z. A., Ali, A., Singh, A., Krakowiak, J., Zheng, X., Bindokas, V. P., Wolfgeher, D., Kron, S. J., & Pincus, D. (2021). Subcellular localization of the J-protein Sis1 regulates the heat shock response. Journal of Cell Biology, 220(1), e202005165. https://doi.org/10.1083/JCB.202005165

Gaglia, G., Rashid, R., Yapp, C., Joshi, G. N., Li, C. G., Lindquist, S. L., Sarosiek, K. A., Whitesell, L., Sorger, P. K., & Santagata, S. (2020). HSF1 phase transition mediates stress adaptation and cell fate decisions. Nature Cell Biology, 22(2), 151–158. https://doi.org/10.1038/s41556-019-0458-3

Giardina, C., & Lis, J. T. (1995). Dynamic protein-DNA architecture of a yeast heat shock promoter. Molecular and Cellular Biology, 15(5), 2737–2744. https://doi.org/10.1128/mcb.15.5.2737

Jolly, C., Konecny, L., Grady, D. L., Kutskova, Y. A., Cotto, J. J., Morimoto, R. I., & Vourc’h, C. (2002). In vivo binding of active heat shock transcription factor 1 to human chromosome 9 heterochromatin during stress. Journal of Cell Biology, 156(5), 775–781. https://doi.org/10.1083/jcb.200109018

Jolly, C., Morimoto, R. I., Robert-Nicoud, M., & Vourc’h, C. (1997). HSF1 transcription factor concentrates in nuclear foci during heat shock: Relationship with transcription sites. Journal of Cell Science, 110(23), 2935–2941. https://doi.org/10.1242/jcs.110.23.2935

Pincus, D., Anandhakumar, J., Thiru, P., Guertin, M. J., Erkine, A. M., & Gross, D. S. (2018). Genetic and epigenetic determinants establish a continuum of Hsf1 occupancy and activity across the yeast genome. Molecular Biology of the Cell, 29(26), 3168–3182. https://doi.org/10.1091/mbc.E18-060353

Zhang, H., Shao, S., Zeng, Y., Wang, X., Qin, Y., Ren, Q., Xiang, S., Wang, Y., Xiao, J., & Sun, Y. (2022). Reversible phase separation of HSF1 is required for an acute transcriptional response during heat shock. Nature Cell Biology, 24(3), 340–352. https://doi.org/10.1038/s41556-022-00846-7

eLife assessment

This is a valuable contribution to our understanding of how different cell stressors (ethanol or heat-shock) elicit unique responses at the genomic and topographical level under the regulation of yeast transcription factor Hsf1, providing solid evidence documenting the temporal coupling (or lack thereof) between Hsf1 aggregation and long-range communication among co-regulated heat-shock loci versus chromatin remodeling and transcriptional activation. A particular strength is the combination of genomic and imaging-based experimental approaches applied to genetically engineered in vivo systems.

Reviewer #2 (Public Review):

Significance:

Rubio et al. study the behavior of the transcription factor Hsf1 under ethanol stress, examining its distribution within the nucleus and the coalescence of heat shock response genes in budding yeast. In comparison to the heat shock response, the response to ethanol stress shows similar gene coalescence and Hsf1 binding. However, there is a notable delay in the transcriptional response to ethanol, and a disconnect between it and the appearance of irreversible Hsf1 condensates/puncta, highlighting important differences in how Hsf1 responds to these two related but distinct environmental stresses.

Overview and general concerns (from the original review):

The authors studied how yeast responds to ethanol stress (8.5%) and compared it to the heat shock response (from 25{degree sign}C to 39{degree sign}C). They observed a more gradual increase in the expression of heat shock response (HSR) genes during ethanol stress compared to heat shock. Additionally, the recruitment of Hsf1 and Pol II to HSR genes, and the inter- and intrachromosomal interactions among these genes, showed slower kinetics under ethanol stress. They attribute the delay in transcriptional response to chromatin compaction induced by ethanol. Despite this delay, these interactions persisted longer. Hsf1 clusters, previously documented during the heat shock response, were also observed during ethanol stress and persisted for an extended period. The conditional degradation of Hsf1 and Rpb1 eliminated most inter- and intrachromosomal interactions for heat shock responsive genes in both ethanol stress and heat shock conditions, indicating the importance of these factors for long distance interactions between HSR genes. Overall, this manuscript provides novel insights into the differential behavior of HSR genes under different stress conditions. This contributes to the broader understanding of how different stressors might elicit unique responses at the genomic and topographical level under the regulation of transcription factor Hsf1.

The central finding of the study highlights the different dynamics of Hsf1, Pol II, and gene organization in response to heat shock versus ethanol stress. However, one important limitation to consider is that the two chosen conditions may not be directly comparable. For a balanced assessment, the authors should ideally expose yeast to various ethanol concentrations and different heat shock temperatures, ensuring the observed differences stem from the nature of the stressor rather than suboptimal stress intensity. At the very least, an additional single ethanol concentration point on each side of 8.5% should be investigated to ensure that 8.5% is near the optimum. In fact, comparing the number of Hsp104 foci in the two conditions in Fig. 1E and F suggests that the yeast is likely experiencing different intensities of stress for the chosen heat shock condition and ethanol concentration used in this study.

A second significant concern is the use of the term "Hsf1 condensate". Chowdhary et al.'s 2022 Molecular Cell study highlighted an inhomogeneous distribution and rapid dynamics of Hsf1 clustering upon heat shock, with sensitivity to 1,6-hexandiol, which is interpreted as evidence for condensation by LLPS. But this interpretation has been criticized severely by McSwiggen at al. Genes Dev 2019 and Mussacchio EMBO J 2022. It is important to mention that 1,6-hexandiol is known to affect chromatin organization (Itoh et al. Life Science Alliance 2021). Describing such clusters as 'condensates' without further experimental evidence is premature. I encourage authors to settle on their neutral term 'puncta' which they use interchangeably with 'condensate' so as not to confuse the reader. The dynamic binding and unbinding of the low-abundance Hsf1 at coalescent chromatin target sites might explain the liquid-like properties of these clusters without the need for invoking the phase-separation hypothesis. While Hsf1 clusters exhibit features consistent with phase-separated condensates, other equally plausible alternative mechanisms, such as dynamic site-specific interactions (Musacchio, EMBO J, 2022), should also be considered. This is best left for the discussion where the underlying mechanism for puncta formation can be addressed.

Comments on revised version:

The authors have addressed the majority of my comments effectively. The new Sis1 experiment provides a clear illustration of a distinctive response to ethanol and heat. This work offers a comprehensive perspective on Hsf1 in stress response from multiple angles. I have two additional comments to improve the paper without re-review:

(Original point #3) Could the authors clarify the differences between DPY1561 and the original strain used? There appears to be missing statistical analysis for Figure 1E at the bottom.

(Original point #4) In the new Figure 7F, '% transcription' and '% coalescence' are presented. My understanding is that Figures 7D and 7E aim to demonstrate the correlation between HSP104 transcription (a continuous variable) and HSP104-HSP12 coalescence (a binary variable) at the single-cell level. However, averaging the data across cells masks individual variations and potential anti-correlations. The authors could explore statistical methods that handle correlations between a continuous variable and a binary variable. Alternatively, consider converting 'HSP104 transcription' to a binary variable and then performing a chi-square test to assess the association.

Reviewer #3 (Public Review):

This is an interesting manuscript that builds off of this group's previous work focused on the interface between Hsf1, heat shock protein (HSP) mRNA production, and 3D genome topology. Here the group subjects the yeast Saccharomyces cerevisiae to either heat stress (HS) or ethanol stress (ES) and examines Hsf1 and Pol II chromatin binding, Histone occupancy, Hsf1 condensates, HSP gene coalescence (by 3C and live cell imaging), and HSP mRNA expression (by RT-qPCR and live cell imaging). The manuscript is well written, and the experiments seem well done, and generally rigorous, with orthogonal approaches performed to support conclusions. The main findings are that both HS and ES result in Hsf1/Pol II-dependent intergenic interactions, along with formation of Hsf1 condensates. Yet, while HS results in rapid and strong induction of HSP gene expression and Hsf1 condensate resolution, ES result in slow and weak induction of HSP gene expression without Hsf1 condensate resolution. Thus, the conclusion is somewhat phenomenological - that the same transcription factor can drive distinct transcription, topologic, and phase-separation behavior in response to different types of stress. While identifying a mechanistic basis for these results would be a tough task perhaps beyond the scope of this study, it would nevertheless be helpful to place these results in context with a series of other studies demonstrating across various organisms showing Hsf1 driving distinct activities dependent on the context of activation. Perhaps even more importantly, this work left out PMID: 32015439 which is particularly relevant considering that it shows that it is human HSF1 condensate resolution rather than simple condensate formation that is associated with HSF1 transcriptional activity - which are similar to the findings here with this particular dose of HS resulting in resolution and high transcriptional activity versus ES resulting in resolution failure and lower activity. It is also worth noting that the stresses themselves are quite different - ethanol can be used as a carbon source and so beyond inducing proteotoxic stress, the yeast are presumably adapting to this distinct metabolic state. Basically, it is not clear whether these differences are due to the dose of stress, versus we are looking at an early timepoint as ES initiates a genome-wide chromatin restructuring and gene expression reprogramming that goes beyond a response to proteotoxic stress. This reviewer is not suggesting a barrage of new experiments, but perhaps discussion points to contextualize results.

Comments on latest version:

The authors have addressed my concerns.

Reviewer #1 (Public Review):

(1) Significance of the findings:

Cell-to-cell communication is essential for higher functions in bacterial biofilms. Electrical signals have proven effective in transmitting signals across biofilms. These signals are then used to coordinate cellular metabolisms or to increase antibiotic tolerance. Here, the authors have reported for the first time coordinated oscillation of membrane potential in E. coli biofilms that may have a functional role in photoprotection.

(2) Strengths of the manuscript:

- The authors report original data.<br /> - For the first time, they showed that coordinated oscillations in membrane potential occur in E. Coli biofilms.<br /> - The authors revealed a complex two-phase dynamic involving distinct molecular response mechanisms.<br /> - The authors developed two rigorous models inspired by 1) Hodgkin-Huxley model for the temporal dynamics of membrane potential and 2) Fire-Diffuse-Fire model for the propagation of the electric signal.<br /> - Since its discovery by comparative genomics, the Kch ion channel has not been associated with any specific phenotype in E. coli. Here, the authors proposed a functional role for the putative gated-voltage-gated K+ ion channel (Kch channel) : enhancing survival under photo-toxic conditions.

(3) Weakness:

- Contrarily to what is stated in the abstract, the group of B. Maier has already reported collective electrical oscillations in the Gram-negative bacterium Neisseria gonorrhoeae (Hennes et al., PLoS Biol, 2023).<br /> - The data presented in the manuscript are not sufficient to conclude on the photo-protective role of the Kch channel. The authors should perform the appropriate control experiments related to Fig4D,E, i.e. reproduce these experiments without ThT to rule out possible photo-conversion effects on ThT that would modify its toxicity. In addition, it looks like the data reported on Fig 4E are extracted from Fig 4D. If this is indeed the case, it would be more conclusive to report the percentage of PI-positive cells in the population for each condition. This percentage should be calculated independently for each replicate. The authors should then report the average value and standard deviation of the percentage of dead cells for each condition.<br /> - Although Fig 4A clearly shows that light stimulation has an influence on the dynamics of ThT signal in the biofilm, it is important to rule out possible contributions of other environmental variations that occur when the flow is stopped at the onset of light stimulation. I understand that for technical reasons, the flow of fresh medium must be stopped for the sake of imaging. Therefore, I suggest to perform control experiments consisting in stopping the flow at different time intervals before image acquisition (30min or 1h before). If there is no significant contribution from environmental variations due to medium perfusion arrest, the dynamics of ThT signal must be unchanged regardless of the delay between flow stop and the start of light stimulation.<br /> - To precise the role of K+ in the habituation response, I suggest using the ionophore valinomycin at sub-inhibitory concentrations (5 or 10µM). It should abolish the habituation response. In addition, the Kch complementation experiment exhibits a sharp drop after the first peak but on a single point. It would be more convincing to increase the temporal resolution (1min->10s) to show that there are indeed a first and a second peak. Finally, the high concentration (100µM) of CCCP used in this study completely inhibits cell activity. Therefore, it is not surprising that no ThT dynamics was observed upon light stimulation at such concentration of CCCP.<br /> - Since TMRM signal exhibits a linear increase after the first response peak (Supp Fig1D), I recommend to mitigate the statement at line 78.<br /> - Electrical signal propagation is an important aspect of the manuscript. However, a detailed quantitative analysis of the spatial dynamics within the biofilm is lacking. At minima, I recommend to plot the spatio-temporal diagram of ThT intensity profile averaged along the azimuthal direction in the biofilm. In addition, it is unclear if the electrical signal propagates within the biofilm during the second peak regime, which is mediated by the Kch channel: I have plotted the spatio-temporal diagram for Video S3 and no electrical propagation is evident at the second peak. In addition, the authors should provide technical details of how R^2(t) is measured in the first regime (Fig 7E).<br /> - In the series of images presented in supplementary Figure 4A, no wavefront is apparent. Although the microscopy technics used in this figure differs from other images (like in Fig2), the wavefront should be still present. In addition, there is no second peak in confocal images as well (Supp Fig4B) .<br /> - Many important technical details are missing (e.g. biofilm size, R^2, curvature and 445nm irradiance measurements). The description of how these quantitates are measured should be detailed in the Material & Methods section.<br /> - Fig 5C: The curve in Fig 5D seems to correspond to the biofilm case. Since the model is made for single cells, the curve obtained by the model should be compared with the average curve presented in Fig 1B (i.e. single cell experiments).<br /> - For clarity, I suggest to indicate on the panels if the experiments concern single cell or biofilm experiments. Finally, please provide bright-field images associated to ThT images to locate bacteria.<br /> - In Fig 7B, the plateau is higher in the simulations than in the biofilm experiments. The authors should add a comment in the paper to explain this discrepancy.

eLife assessment

This potentially valuable study presents claims of evidence for coordinated membrane potential oscillations in E. coli biofilms that can be linked to a putative K+ channel and that may serve to enhance photo-protection. The finding of waves of membrane potential would be of interest to a wide audience from molecular biology to microbiology and physical biology. Unfortunately, a major issue is that it is unclear whether the dye used can act as a Nernstian membrane potential dye in E. coli. The arguments of the authors, who largely ignore previously published contradictory evidence, are inadequate in that they do not engage with the fact that the dye behaves in their hands differently than in the hands of others. In addition, the lack of proper validation of the experimental method including key control experiments leaves the evidence incomplete.

Reviewer #2 (Public Review):

The authors use ThT dye as a Nernstian potential dye in E. coli. Quantitative measurements of membrane potential using any cationic indicator dye are based on the equilibration of the dye across the membrane according to Boltzmann's law.

Ideally, the dye should have high membrane permeability to ensure rapid equilibration. Others have demonstrated that E.coli cells in the presence of ThT do not load unless there is blue light present, that the loading profile does not look like it is expected for a cationic Nernstian dye. They also show that the loading profile of the dye is different for E.coli cells deleted for the TolC pump. I, therefore, objected to interpreting the signal from the ThT as a Vm signal when used in E.coli. Nothing the authors have said has suggested that I should be changing this assessment.

Specifically, the authors responded to my concerns as follows:

(1) 'We are aware of this study, but believe it to be scientifically flawed. We do not cite the article because we do not think it is a particularly useful contribution to the literature.' This seems to go against ethical practices when it comes to scientific literature citations. If the authors identified work that handles the same topic they do, which they believe is scientifically flawed, the discussion to reflect that should be included.

(2)'The Pilizota group invokes some elaborate artefacts to explain the lack of agreement with a simple Nernstian battery model. The model is incorrect not the fluorophore.'<br /> It seems the authors object to the basic principle behind the usage of Nernstian dyes. If the authors wish to use ThT according to some other model, and not as a Nernstian indicator, they need to explain and develop that model. Instead, they state 'ThT is a Nernstian voltage indicator' in their manuscript and expect the dye to behave like a passive voltage indicator throughout it.

(3)'We think the proton effect is a million times weaker than that due to potassium i.e. 0.2 M K+<br /> versus 10-7 M H+. We can comfortably neglect the influx of H+ in our experiments.'<br /> I agree with this statement by the authors. At near-neutral extracellular pH, E.coli keeps near-neutral intracellular pH, and the contribution from the chemical concentration gradient to the electrochemical potential of protons is negligible. The main contribution is from the membrane potential. However, this has nothing to do with the criticism to which this is the response of the authors. The criticism is that ThT has been observed not to permeate the cell without blue light. The blue light has been observed to influence the electrochemical potential of protons (and given that at near-neutral intracellular and extracellular pH this is mostly the membrane potential, as authors note themselves, we are talking about Vm effectively). Thus, two things are happening when one is loading the ThT, not just expected equilibration but also lowering of membrane potential. The electrochemical potential of protons is coupled via the membrane potential to all the other electrochemical potentials of ions, including the mentioned K+.

(4) 'The vast majority of cells continue to be viable. We do not think membrane damage is dominating.' In response to the question on how the authors demonstrated TMRM loading and in which conditions (and while reminding them that TMRM loading profile in E.coli has been demonstrated in Potassium Phosphate buffer). The request was to demonstrate TMRM loading profile in their condition as well as to show that it does not depend on light. Cells could still be viable, as membrane permeabilisation with light is gradual, but the loading of ThT dye is no longer based on simple electrochemical potential (of the dye) equilibration.

(5) On the comment on the action of CCCP with references included, authors include a comment that consists of phrases like 'our understanding of the literature' with no citations of such literature. Difficult to comment further without references.

(6) 'Shielding would provide the reverse effect, since hyperpolarization begins in the dense centres of the biofilms. For the initial 2 hours the cells receive negligible blue light. Neither of the referee's comments thus seem tenable.'<br /> The authors have misunderstood my comment. I am not advocating shielding (I agree that this is not it) but stating that this is not the only other explanation for what they see (apart from electrical signaling). The other I proposed is that the membrane has changed in composition and/or the effective light power the cells can tolerate. The authors comment only on the light power (not convincingly though, giving the number for that power would be more appropriate), not on the possible changes in the membrane permeability.

(7) 'The work that TolC provides a possible passive pathway for ThT to leave cells seems slightly niche. It just demonstrates another mechanism for the cells to equilibrate the concentrations of ThT in a Nernstian manner i.e. driven by the membrane voltage.' I am not sure what the authors mean by another mechanism. The mechanism of action of a Nernstian dye is passive equilibration according to the electrochemical potential (i.e. until the electrochemical potential of the dye is 0).

(8) 'In the 70 years since Hodgkin and Huxley first presented their model, a huge number of similar models have been proposed to describe cellular electrophysiology. We are not being hyperbolic when we state that the HH models for excitable cells are like the Schrödinger<br /> equation for molecules. We carefully adapted our HH model to reflect the currently understood electrophysiology of E. coli.'

I gave a very concrete comment on the fact that in the HH model conductivity and leakage are as they are because this was explicitly measured. The authors state that they have carefully adopted their model based on what is currently understood for E.coli electrophysiology. It is not clear how. HH uses gKn^4 based on Figure2 here https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1392413/pdf/jphysiol01442-0106.pdf, i.e. measured rise and fall of potassium conductance on msec time scales. I looked at the citation the authors have given and found a resistance of an entire biofilm of a given strain at 3 applied voltages. So why n^4 based on that? Why does unknown current have gqz^4 form? Sodium conductance in HH is described by m^3hgNa (again based on detailed conductance measurements), so why unknown current in E.coli by gQz^4? Why leakage is in the form that it is, based on what measurement?

Throughout their responses, the authors seem to think that collapsing the electrochemical gradient of protons is all about protons, and this is not the case. At near neutral inside and outside pH, the electrochemical potential of protons is simply membrane voltage. And membrane voltage acts on all ions in the cell.

Authors have started their response to concrete comments on the usage of ThT dye with comments on papers from my group that are not all directly relevant to this publication. I understand that their intention is to discredit a reviewer but given that my role here is to review this manuscript, I will only address their comments to the publications/part of publications that are relevant to this manuscript and mention what is not relevant.

Publications in the order these were commented on.

(1) In a comment on the paper that describes the usage of ThT dye as a Nernstian dye authors seem to talk about a model of an entire active cell.<br /> 'Huge oscillations occur in the membrane potentials of E. coli that cannot be described by the SNB model.' The two have nothing to do with each other. Nernstian dye equilibrates according to its electrochemical potential. Once that happens it can measure the potential (under the assumption that not too much dye has entered and thus lowered too much the membrane potential under measurement). The time scale of that is important, and the dye can only measure processes that are slower than that equilibration. If one wants to use a dye that acts under a different model, first that needs to be developed, and then coupled to any other active cell model.

(2) The part of this paper that is relevant is simply the usage of TMRM dye. It is used as Nernstian dye, so all the above said applies. The rest is a study of flagellar motor.

(3) The authors seem to not understand that the electrochemical potential of protons is coupled to the electrochemical potentials of all other ions, via the membrane potential. In the manuscript authors talk about, PMF~Vm, as DeltapH~0. Other than that this publication is not relevant to their current manuscript.

(4) The manuscript in fact states precisely that PMF cannot be generated by protons only and some other ions need to be moved out for the purpose. In near neutral environment it stated that these need to be cations (K+ e.g.). The model used in this manuscript is a pump-leak model. Neither is relevant for the usage of ThT dye.

Further comments include, along the lines of:

'The editors stress the main issue raised was a single referee questioning the use of ThT as an indicator of membrane potential. We are well aware of the articles by the Pilizota group and we believe them to be scientifically flawed. The authors assume there are no voltage-gated ion channels in E. coli and then attempt to explain motility data based on a simple Nernstian battery model (they assume E. coli are unexcitable<br /> matter). This in turn leads them to conclude the membrane dye ThT is faulty, when in fact it is a problem with their simple battery model.'

The only assumption made when using a cationic Nernstian dye is that it equilibrates passively across the membrane according to its electrochemical potential. As it does that, it does lower the membrane potential, which is why as little as possible is added so that this is negligible. The equilibration should be as fast as possible, but at the very least it should be known, as no change in membrane potential can be measured that is faster than that.

This behaviour should be orthogonal to what the cell is doing, it is a probe after all. If the cell is excitable, a Nernstian dye can be used, as long as it's still passively equilibrating and doing so faster than any changes in membrane potential due to excitations of the cells. There are absolutely no assumptions made on the active system that is about to be measured by this expected behaviour of a Nernstian dye. And there shouldn't be, it is a probe. If one wants to use a dye that is not purely Nernstian that behaviour needs to be described and a model proposed. As far as I can find, authors do no such thing.

There is a comment on the use of a flagellar motor as a readout of PMF, stating that the motor can be stopped by YcgR citing the work from 2023. Indeed, there is a range of references such as https://doi.org/10.1016/j.molcel.2010.03.001 that demonstrate this (from around 2000-2010 as far as I am aware). The timescale of such slowdown is hours (see here Figure 5 https://www.cell.com/cell/pdf/S0092-8674(10)00019-X.pdf). Needless to say, the flagellar motor when used as a probe, needs to stay that in the conditions used. Thus one should always be on the lookout at any other such proteins that could slow it down and we are not aware of yet or make the speed no longer proportional to the PMF. In the papers my group uses the motor the changes are fast, often reversible, and in the observation window of 30min. They are also the same with DeltaYcgR strain, which we have not included as it seemed given the time scales it's obvious, but certainly can in the future (as well as stay vigilant on any conditions that would render the motor a no longer suitable probe for PMF).

Reviewer #3 (Public Review):

This manuscript by Akabuogu et al. investigates membrane potential dynamics in E. coli. Membrane potential fluctuations have been observed in bacteria by several research groups in recent years, including in the context of bacterial biofilms where they have been proposed to play a role in cellular communication. Here, these authors investigate membrane potential in E. coli, in both single cells and biofilms. I have reviewed the revised manuscript provided by the authors, as well as their responses to the initial reviews; my opinion about the manuscript is largely unchanged. I have focused my public review on those issues that I believe to be most pressing, with additional comments included in the review to authors. Although these authors are working in an exciting research area, the evidence they provide for their claims is inadequate, and several key control experiments are still missing. In some cases, the authors allude to potentially relevant data in their responses to the initial reviews, but unfortunately these data are not shown. Furthermore, I cannot identify any traveling wavefronts in the data included in this manuscript. In addition to the challenges associated with the use of Thioflavin-T (ThT) raised by the second reviewer, these caveats make the work presented in this manuscript difficult to interpret.

First, some of the key experiments presented in the paper lack required controls:

(1) This paper asserts that the observed ThT fluorescence dynamics are induced by blue light. This is a fundamental claim in the paper, since the authors go on to argue that these dynamics are part of a blue light response. This claim must be supported by the appropriate negative control experiment measuring ThT fluorescence dynamics in the absence of blue light- if this idea is correct, these dynamics should not be observed in the absence of blue light exposure. If this experiment cannot be performed with ThT since blue light is used for its excitation, TMRM can be used instead.

In response to this, the authors wrote that "the fluorescent baseline is too weak to measure cleanly in this experiment." If they observe no ThT signal above noise in their time lapse data in the absence of blue light, this should be reported in the manuscript- this would be a satisfactory negative control. They then wrote that "It appears the collective response of all the bacteria hyperpolarization at the same time appears to dominate the signal." I am not sure what they mean by this- perhaps that ThT fluorescence changes strongly only in response to blue light? This is a fundamental control for this experiment that ought to be presented to the reader.

(2) The authors claim that a ∆kch mutant is more susceptible to blue light stress, as evidenced by PI staining. The premise that the cells are mounting a protective response to blue light via these channels rests on this claim. However, they do not perform the negative control experiment, conducting PI staining for WT the ∆kch mutant in the absence of blue light. In the absence of this control it is not possible to rule out effects of the ∆kch mutation on overall viability and/or PI uptake. The authors do include a growth curve for comparison, but planktonic growth is a very different context than surface-attached biofilm growth. Additionally, the ∆kch mutation may have impacts on PI permeability specifically that are not addressed by a growth curve. The negative control experiment is of key importance here.

Second, the ideas presented in this manuscript rely entirely on analysis of ThT fluorescence data, specifically a time course of cellular fluorescence following blue light treatment. However, alternate explanations for and potential confounders of the observed dynamics are not sufficiently addressed:

(1) Bacterial cells are autofluorescent, and this fluorescence can change significantly in response to stress (e.g. blue light exposure). To characterize and/or rule out autofluorescence contributions to the measurement, the authors should present time lapse fluorescence traces of unstained cells for comparison, acquired under the same imaging conditions in both wild type and ∆kch mutant cells. In their response to reviewers the authors suggested that they have conducted this experiment and found that the autofluorescence contribution is negligible, which is good, but these data should be included in the manuscript along with a description of how these controls were conducted.

(2) Similarly, in my initial review I raised a concern about the possible contributions of photobleaching to the observed fluorescence dynamics. This is particularly relevant for the interpretation of the experiment in which catalase appears to attenuate the decay of the ThT signal; this attenuation could alternatively be due to catalase decreasing ThT photobleaching. In their response, the authors indicated that photobleaching is negligible, which would be good, but they do not share any evidence to support this claim. Photobleaching can be assessed in this experiment by varying the light dosage (illumination power, frequency, and/or duration) and confirming that the observed fluorescence dynamics are unaffected.

Third, the paper claims in two instances that there are propagating waves of ThT fluorescence that move through biofilms, but I do not observe these waves in any case:

(1) The first wavefront claim relates to small cell clusters, in Fig. 2A and Video S2 and S3 (with Fig. 2A and Video S2 showing the same biofilm.) I simply do not see any evidence of propagation in either case- rather, all cells get brighter and dimmer in tandem. I downloaded and analyzed Video S3 in several ways (plotting intensity profiles for different regions at different distances from the cluster center, drawing a kymograph across the cluster, etc.) and in no case did I see any evidence of a propagating wavefront. (I attempted this same analysis on the biofilm shown in Fig. 2A and Video S2 with similar results, but the images shown in the figure panels and especially the video are still both so saturated that the quantification is difficult to interpret.) If there is evidence for wavefronts, it should be demonstrated explicitly by analysis of several clusters. For example, a figure of time-to-peak vs. position in the cluster demonstrating a propagating wave would satisfy this. Currently, I do not see any wavefronts in this data.

(2) The other wavefront claim relates to biofilms, and the relevant data is presented in Fig. S4 (and I believe also in what is now Video S8, but no supplemental video legends are provided, and this video is not cited in text.) As before, I cannot discern any wavefronts in the image and video provided; Reviewer 1 was also not able to detect wave propagation in this video by kymograph. Some mean squared displacements are shown in Fig. 7. As before, the methods for how these were obtained are not clearly documented either in this manuscript or in the BioRXiv preprint linked in the initial response to reviewers, and since wavefronts are not evident in the video it is hard to understand what is being measured here- radial distance from where? (The methods section mentions radial distance from the substrate, this should mean Z position above the imaging surface, and no wavefronts are evident in Z in the figure panels or movie.) Thus, clear demonstration of these wavefronts is still missing here as well.

Fourth, I have some specific questions about the study of blue light stress and the use of PI as a cell viability indicator:

(1) The logic of this paper includes the premise that blue light exposure is a stressor under the experimental conditions employed in the paper. Although it is of course generally true that blue light can be damaging to bacteria, this is dependent on light power and dosage. The control I recommended above, staining cells with PI in the presence and absence of blue light, will also allow the authors to confirm that this blue light treatment is indeed a stressor- the PI staining would be expected to increase in the presence of blue light if this is so.

(2) The presence of ThT may complicate the study of the blue light stress response, since ThT enhances the photodynamic effects of blue light in E. coli (Bondia et al. 2021 Chemical Communications). The authors could investigate ThT toxicity under these conditions by staining cells with PI after exposing them to blue light with or without ThT staining.

(3) In my initial review, I wrote the following: "In Figures 4D - E, the interpretation of this experiment can be confounded by the fact that PI uptake can sometimes be seen in bacterial cells with high membrane potential (Kirchhoff & Cypionka 2017 J Microbial Methods); the interpretation is that high membrane potential can lead to increased PI permeability. Because the membrane potential is largely higher throughout blue light treatment in the ∆kch mutant (Fig. 3[BC]), this complicates the interpretation of this experiment." In their response, the authors suggested that these results are not relevant in this case because "In our experiment methodology, cell death was not forced on the cells by introducing an extra burden or via anoxia." However, the logic of the paper is that the cells are in fact dying due to an imposed external stressor, which presumably also confers an increased burden as the cells try to deal with the stress. Instead, the authors should simply use a parallel method to confirm the results of PI staining. For example, the experiment could be repeated with other stains, or the viability of blue light-treated cells could be addressed more directly by outgrowth or colony-forming unit assays.

The CFU assay suggested above has the additional advantage that it can also be performed on planktonic cells in liquid culture that are exposed to blue light. If, as the paper suggests, a protective response to blue light is being coordinated at the biofilm level by these membrane potential fluctuations, the WT strain might be expected to lose its survival advantage vs. the ∆kch mutant in the absence of a biofilm.

Fifth, in several cases the data are presented in a way that are difficult to interpret, or the paper makes claims that are different to observe in the data:

(1) The authors suggest that the ThT and TMRM traces presented in Fig. S1D have similar shapes, but this is not obvious to me- the TMRM curve has very little decrease after the initial peak and only a modest, gradual rise thereafter. The authors suggest that this is due to increased TMRM photobleaching, but I would expect that photobleaching should exacerbate the signal decrease after the initial peak. Since this figure is used to support the use of ThT as a membrane potential indicator, and since this is the only alternative measurement of membrane potential presented in text, the authors should discuss this discrepancy in more detail.

(2) The comparison of single cells to microcolonies presented in figures 1B and D still needs revision:

First, both reviewer 1 and I commented in our initial reviews that the ThT traces, here and elsewhere, should not be normalized- this will help with the interpretation of some of the claims throughout the manuscript.

Second, the way these figures are shown with all traces overlaid at full opacity makes it very difficult to see what is being compared. Since the point of the comparison is the time to first peak (and the standard deviation thereof), histograms of the distributions of time to first peak in both cases should be plotted as a separate figure panel.<br /> Third, statistical significance tests ought to be used to evaluate the statistical strength of the comparisons between these curves. The authors compare both means and standard deviations of the time to first peak, and there are appropriate statistical tests for both types of comparisons.

(3) The authors claim that the curve shown in Fig. S4B is similar to the simulation result shown in Fig. 7B. I remain unconvinced that this is so, particularly with respect to the kinetics of the second peak- at least it seems to me that the differences should be acknowledged and discussed. In any case, the best thing to do would be to move Fig. S4B to the main text alongside Fig. 7B so that the readers can make the comparison more easily.

(4) As I wrote in my first review, in the discussion of voltage-gated calcium channels, the authors refer to "spiking events", but these are not obvious in Figure S3E. Although the fluorescence intensity changes over time, these fluctuations cannot be distinguished from measurement noise. A no-light control could help clarify this.

(5) In the lower irradiance conditions in Fig. 4A, the ThT dynamics are slower overall, and it looks like the ThT intensity is beginning to rise at the end of the measurement. The authors write that no second peak is observed below an irradiance threshold of 15.99 µW/mm2. However, could a more prominent second peak be observed in these cases if the measurement time was extended? Additionally, the end of these curves looks similar to the curve in Fig. S4B, in which the authors write that the slow rise is evidence of the presence of a second peak, in contrast to their interpretation here.

Additional considerations:

(1) The analysis and interpretation of the first peak, and particularly of the time-to-fire data is challenging throughout the manuscript the time resolution of the data set is quite limited. It seems that a large proportion of cells have already fired after a single acquisition frame. It would be ideal to increase the time resolution on this measurement to improve precision. This could be done by imaging more quickly, but that would perhaps necessitate more blue light exposure; an alternative is to do this experiment under lower blue light irradiance where the first spike time is increased (Figure 4A).

(2) The authors suggest in the manuscript that "E. coli biofilms use electrical signalling to coordinate long-range responses to light stress." In addition to the technical caveats discussed above, I am missing a discussion about what these responses might be. What constitutes a long-range response to light stress, and are there known examples of such responses in bacteria?

(3) The presence of long-range blue light responses can also be interrogated experimentally, for example, by repeating the Live/Dead experiment in planktonic culture or the single-cell condition. If the protection from blue light specifically emerges due to coordinated activity of the biofilm, the ∆kch mutant would not be expected to show a change in Live/Dead staining in non-biofilm conditions. The CFU experiment I mentioned above could also implicate coordinated long-range responses specifically, if biofilms and liquid culture experiments can be compared (although I know that recovering cells from biofilms is challenging.)

4. At the end of the results section, the authors suggest a critical biofilm size of only 4 μm for wavefront propagation (not much larger than a single cell!) The authors show responses for various biofilm sizes in Fig. 2C, but these are all substantially larger (and this figure also does not contain wavefront information.) Are there data for cell clusters above and below this size that could support this claim more directly?

(5) In Fig. 4C, the overall trajectories of extracellular potassium are indeed similar, but the kinetics of the second peak of potassium are different than those observed by ThT (it rises minutes earlier)- is this consistent with the idea that Kch is responsible for that peak? Additionally, the potassium dynamics also include the first ThT peak- is this surprising given that the Kch channel has no effect on this peak according to the model?

Detailed comments:

Why are Fig. 2A and Video S2 called a microcluster, whereas Video S3, which is smaller, is called a biofilm?

"We observed a spontaneous rapid rise in spikes within cells in the center of the biofilm" (Line 140): What does "spontaneous" mean here?

"This demonstrates that the ion-channel mediated membrane potential dynamics is a light stress relief process.", "E. coli cells employ ion-channel mediated dynamics to manage ROS-induced stress linked to light irradiation." (Line 268 and the second sentence of the Fig. 4F legend): This claim is not well-supported. There are several possible interpretations of the catalase experiment (which should be discussed); this experiment perhaps suggests that ROS impacts membrane potential but does not indicate that these membrane potential fluctuations help the cells respond to blue light stress. The loss of viability in the ∆kch mutant might indicate a link between these membrane potential experiments and viability, but it is hard to interpret without the no light controls I mention above.

"The model also predicts... the external light stress" (Lines 338-341): Please clarify this section. Where does this prediction arise from in the modeling work? Second, I am not sure what is meant by "modulates the light stress" or "keeps the cell dynamics robust to the intensity of external light stress" (especially since the dynamics clearly vary with irradiance, as seen in Figure 4A).

"We hypothesized that E. coli not only modulates the light-induced stress but also handles the increase of the ROS by adjusting the profile of the membrane potential dynamics" (Line 347): I am not sure what "handles the ROS by adjusting the profile of the membrane potential dynamics" means. What is meant by "handling" ROS? Is the hypothesis that membrane potential dynamics themselves are protective against ROS, or that they induce a ROS-protective response downstream, or something else? Later the authors write that changes in the response to ROS in the model agree with the hypothesis, but just showing that ROS impacts the membrane potential does not seem to demonstrate that this has a protective effect against ROS.

"Mechanosensitive ion channels (MS) are vital for the first hyperpolarization event in E. coli." (Line 391): This is misleading- mechanosensitive ion channels totally ablate membrane potential dynamics, they don't have a specific effect on the first hyperpolarization event. The claim that mechanonsensitive ion channels are specifically involved in the first event also appears in the abstract.

Also, the apparent membrane potential is much lower even at the start of the experiment in these mutants (Fig. 6C-D)- is this expected? This seems to imply that these ion channels also have a blue light-independent effect.

Throughout the paper, there are claims that the initial ThT spike is involved in "registering the presence of the light stress" and similar. What is the evidence for this claim?

"We have presented much better quantitative agreement of our model with the propagating wavefronts in E. coli biofilms..." (Line 619): It is not evident to me that the agreement between model and prediction is "much better" in this work than in the cited work (reference 57, Hennes et al. 2023). The model in Figure 4 of ref. 57 seems to capture the key features of their data.

In methods, "Only cells that are hyperpolarized were counted in the experiment as live" (Line 745): what percentage of cells did not hyperpolarize in these experiments?

Some indication of standard deviation (error bars or shading) should be added to all figures where mean traces are plotted.

Video S8 is very confusing- why does the video play first forwards and then backwards? It is easy to misinterpret this as a rise in the intensity at the end of the experiment.

Author response:

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

Critical synopsis of the articles cited by referee 2:

(1) ‘Generalized workflow for characterization of Nernstian dyes and their effects on bacterial physiology’, L.Mancini et al, Biophysical Journal, 2020, 118, 1, 4-14.

This is the central article used by referee 2 to argue that there are issues with the calibration of ThT for the measurement of membrane potentials. The authors use a simple Nernstian battery (SNB) model and unfortunately it is wrong when voltage-gated ion channels occur. Huge oscillations occur in the membrane potentials of E. coli that cannot be described by the SNB model. Instead a Hodgkin Huxley model is needed, as shown in our eLife manuscript and multiple other studies (see above). Arrhenius kinetics are assumed in the SNB model for pumping with no real evidence and the generalized workflow involves ripping the flagella off the bacteria! The authors construct an elaborate ‘work flow’ to insure their ThT results can be interpreted using their erroneous SNB model over a limited range of parameters.

(2) ‘Non-equivalence of membrane voltage and ion-gradient as driving forces for the bacterial flagellar motor at low load’, C.J.Lo, et al, Biophysical Journal, 2007, 93, 1, 294.

An odd de novo chimeric species is developed using an E. coli  chassis which uses Na+ instead of H+ for the motility of its flagellar motor. It is not clear the relevance to wild type E. coli, due to the massive physiological perturbations involved. A SNB model is using to fit the data over a very limited parameter range with all the concomitant errors.

(3) Single-cell bacterial electrophysiology reveals mechanisms of stress-induced damage’, E.Krasnopeeva, et al, Biophysical Journal, 2019, 116, 2390.

The abstract says ‘PMF defines the physiological state of the cell’. This statement is hyperbolic. An extremely wide range of molecules contribute to the physiological state of a cell. PMF does not even define the electrophysiology of the cell e.g. via the membrane potential. There are 0.2 M of K+ compared with 0.0000001 M of H+ in E. coli, so K+ is arguably a million times more important for the membrane potential than H+ and thus the electrophysiology!

Equation (1) in the manuscript assumes no other ions are exchanged during the experiments other than H+. This is a very bad approximation when voltage-gated potassium ion channels move the majority ion (K+) around!

In our model Figure 4A is better explained by depolarisation due to K+ channels closing than direct irreversible photodamage. Why does the THT fluorescence increase again for the second hyperpolarization event if the THT is supposed to be damaged? It does not make sense.

(4) ‘The proton motive force determines E. coli robustness to extracellular pH’, G.Terradot et al, 2024, preprint.

This article expounds the SNB model once more. It still ignores the voltage-gated ion channels. Furthermore, it ignores the effect of the dominant ion in E. coli, K+. The manuscript is incorrect as a result and I would not recommend publication.

In general, an important problem is being researched i.e. how the membrane potential of E. coli is related to motility, but there are serious flaws in the SNB approach and the experimental methodology appears tenuous.

Answers to specific questions raised by the referees

Reviewer #1 (Public Review):

Summary:

Cell-to-cell communication is essential for higher functions in bacterial biofilms. Electrical signals have proven effective in transmitting signals across biofilms. These signals are then used to coordinate cellular metabolisms or to increase antibiotic tolerance. Here, the authors have reported for the first time coordinated oscillation of membrane potential in E. coli biofilms that may have a functional role in photoprotection.

Strengths:

- The authors report original data.

- For the first time, they showed that coordinated oscillations in membrane potential occur in E. Coli biofilms.

- The authors revealed a complex two-phase dynamic involving distinct molecular response mechanisms.

- The authors developed two rigorous models inspired by 1) Hodgkin-Huxley model for the temporal dynamics of membrane potential and 2) Fire-Diffuse-Fire model for the propagation of the electric signal.

- Since its discovery by comparative genomics, the Kch ion channel has not been associated with any specific phenotype in E. coli. Here, the authors proposed a functional role for the putative K+ Kch channel : enhancing survival under photo-toxic conditions.

We thank the referee for their positive evaluations and agree with these statements.

Weaknesses:

- Since the flow of fresh medium is stopped at the beginning of the acquisition, environmental parameters such as pH and RedOx potential are likely to vary significantly during the experiment. It is therefore important to exclude the contributions of these variations to ensure that the electrical response is only induced by light stimulation. Unfortunately, no control experiments were carried out to address this issue.

The electrical responses occur almost instantaneously when the stimulation with blue light begins i.e. it is too fast to be a build of pH. We are not sure what the referee means by Redox potential since it is an attribute of all chemicals that are able to donate/receive electrons. The electrical response to stress appears to be caused by ROS, since when ROS scavengers are added the electrical response is removed i.e. pH plays a very small minority role if any.

- Furthermore, the control parameter of the experiment (light stimulation) is the same as that used to measure the electrical response, i.e. through fluorescence excitation. The use of the PROPS system could solve this problem.

>>We were enthusiastic at the start of the project to use the PROPs system in E. coli as presented by J.M.Krajl et al, ‘Electrical spiking in E. coli probed with a fluorescent voltage-indicating protein’, Science, 2011, 333, 6040, 345. However, the people we contacted in the microbiology community said that it had some technical issues and there have been no subsequent studies using PROPs in bacteria after the initial promising study. The fluorescent protein system recently presented in PNAS seems more promising, ‘Sensitive bacterial Vm sensors revealed the excitability of bacterial Vm and its role in antibiotic tolerance’, X.Jin et al, PNAS, 120, 3, e2208348120.

- Electrical signal propagation is an important aspect of the manuscript. However, a detailed quantitative analysis of the spatial dynamics within the biofilm is lacking. In addition, it is unclear if the electrical signal propagates within the biofilm during the second peak regime, which is mediated by the Kch channel. This is an important question, given that the fire-diffuse-fire model is presented with emphasis on the role of K+ ions.

We have presented a more detailed account of the electrical wavefront modelling work and it is currently under review in a physical journal, ‘Electrical signalling in three dimensional bacterial biofilms using an agent based fire-diffuse-fire model’, V.Martorelli, et al, 2024 https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

- Since deletion of the kch gene inhibits the long-term electrical response to light stimulation (regime II), the authors concluded that K+ ions play a role in the habituation response. However, Kch is a putative K+ ion channel. The use of specific drugs could help to clarify the role of K+ ions.

Our recent electrical impedance spectroscopy publication provides further evidence that Kch is associated with large changes in conductivity as expected for a voltage-gated ion channel (https://pubs.acs.org/doi/10.1021/acs.nanolett.3c04446, 'Electrical impedance spectroscopy with bacterial biofilms: neuronal-like behavior', E.Akabuogu et al, ACS Nanoletters, 2024, in print.

- The manuscript as such does not allow us to properly conclude on the photo-protective role of the Kch ion channel.

That Kch has a photoprotective role is our current working hypothesis. The hypothesis fits with the data, but we are not saying we have proven it beyond all possible doubt.

- The link between membrane potential dynamics and mechanosensitivity is not captured in the equation for the Q-channel opening dynamics in the Hodgkin-Huxley model (Supp Eq 2).

Our model is agnostic with respect to the mechanosensitivity of the ion channels, although we deduce that mechanosensitive ion channels contribute to ion channel Q.

- Given the large number of parameters used in the models, it is hard to distinguish between prediction and fitting.

This is always an issue with electrophysiological modelling (compared with most heart and brain modelling studies we are very conservative in the choice of parameters for the bacteria). In terms of predicting the different phenomena observed, we believe the model is very successful.

Reviewer #2 (Public Review):

Summary of what the authors were trying to achieve:

The authors thought they studied membrane potential dynamics in E.coli biofilms. They thought so because they were unaware that the dye they used to report that membrane potential in E.coli, has been previously shown not to report it. Because of this, the interpretation of the authors' results is not accurate.

We believe the Pilizota work is scientifically flawed.

Major strengths and weaknesses of the methods and results:

The strength of this work is that all the data is presented clearly, and accurately, as far as I can tell.

The major critical weakness of this paper is the use of ThT dye as a membrane potential dye in E.coli. The work is unaware of a publication from 2020 https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] that demonstrates that ThT is not a membrane potential dye in E. coli. Therefore I think the results of this paper are misinterpreted. The same publication I reference above presents a protocol on how to carefully calibrate any candidate membrane potential dye in any given condition.

We are aware of this study, but believe it to be scientifically flawed. We do not cite the article because we do not think it is a particularly useful contribution to the literature.

I now go over each results section in the manuscript.

Result section 1: Blue light triggers electrical spiking in single E. coli cells

I do not think the title of the result section is correct for the following reasons. The above-referenced work demonstrates the loading profile one should expect from a Nernstian dye (Figure 1). It also demonstrates that ThT does not show that profile and explains why is this so. ThT only permeates the membrane under light exposure (Figure 5). This finding is consistent with blue light peroxidising the membrane (see also following work Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com] on light-induced damage to the electrochemical gradient of protons-I am sure there are more references for this).

The Pilizota group invokes some elaborate artefacts to explain the lack of agreement with a simple Nernstian battery model. The model is incorrect not the fluorophore.

Please note that the loading profile (only observed under light) in the current manuscript in Figure 1B as well as in the video S1 is identical to that in Figure 3 from the above-referenced paper (i.e. https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com]), and corresponding videos S3 and S4. This kind of profile is exactly what one would expect theoretically if the light is simultaneously lowering the membrane potential as the ThT is equilibrating, see Figure S12 of that previous work. There, it is also demonstrated by the means of monitoring the speed of bacterial flagellar motor that the electrochemical gradient of protons is being lowered by the light. The authors state that applying the blue light for different time periods and over different time scales did not change the peak profile. This is expected if the light is lowering the electrochemical gradient of protons. But, in Figure S1, it is clear that it affected the timing of the peak, which is again expected, because the light affects the timing of the decay, and thus of the decay profile of the electrochemical gradient of protons (Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com]).

We think the proton effect is a million times weaker than that due to potasium i.e. 0.2 M K+ versus 10-7 M H+. We can comfortably neglect the influx of H+ in our experiments.

If find Figure S1D interesting. There authors load TMRM, which is a membrane voltage dye that has been used extensively (as far as I am aware this is the first reference for that and it has not been cited https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1914430 [ncbi.nlm.nih.gov]/). As visible from the last TMRM reference I give, TMRM will only load the cells in Potassium Phosphate buffer with NaCl (and often we used EDTA to permeabilise the membrane). It is not fully clear (to me) whether here TMRM was prepared in rich media (it explicitly says so for ThT in Methods but not for TMRM), but it seems so. If this is the case, it likely also loads because of the damage to the membrane done with light, and therefore I am not surprised that the profiles are similar.

The vast majority of cells continue to be viable. We do not think membrane damage is dominating.

The authors then use CCCP. First, a small correction, as the authors state that it quenches membrane potential. CCCP is a protonophore (https://pubmed.ncbi.nlm.nih.gov/4962086 [pubmed.ncbi.nlm.nih.gov]/), so it collapses electrochemical gradient of protons. This means that it is possible, and this will depend on the type of pumps present in the cell, that CCCP collapses electrochemical gradient of protons, but the membrane potential is equal and opposite in sign to the DeltapH. So using CCCP does not automatically mean membrane potential will collapse (e.g. in some mammalian cells it does not need to be the case, but in E.coli it is https://www.biorxiv.org/content/10.1101/2021.11.19.469321v2 [biorxiv.org]). CCCP has also been recently found to be a substrate for TolC (https://journals.asm.org/doi/10.1128/mbio.00676-21 [journals.asm.org]), but at the concentrations the authors are using CCCP (100uM) that should not affect the results. However, the authors then state because they observed, in Figure S1E, a fast efflux of ions in all cells and no spiking dynamics this confirms that observed dynamics are membrane potential related. I do not agree that it does. First, Figure S1E, does not appear to show transients, instead, it is visible that after 50min treatment with 100uM CCCP, ThT dye shows no dynamics. The action of a Nernstian dye is defined. It is not sufficient that a charged molecule is affected in some way by electrical potential, this needs to be in a very specific way to be a Nernstian dye. Part of the profile of ThT loading observed in https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] is membrane potential related, but not in a way that is characteristic of Nernstian dye.

Our understanding of the literature is CCCP poisons the whole metabolism of the bacterial cells. The ATP driven K+ channels will stop functioning and this is the dominant contributor to membrane potential.

Result section 2: Membrane potential dynamics depend on the intercellular distance

In this chapter, the authors report that the time to reach the first intensity peak during ThT loading is different when cells are in microclusters. They interpret this as electrical signalling in clusters because the peak is reached faster in microclusters (as opposed to slower because intuitively in these clusters cells could be shielded from light). However, shielding is one possibility. The other is that the membrane has changed in composition and/or the effective light power the cells can tolerate (with mechanisms to handle light-induced damage, some of which authors mention later in the paper) is lower. Given that these cells were left in a microfluidic chamber for 2h hours to attach in growth media according to Methods, there is sufficient time for that to happen. In Figure S12 C and D of that same paper from my group (https://ars.els-cdn.com/content/image/1-s2.0-S0006349519308793-mmc6.pdf [ars.els-cdn.com]) one can see the effects of peak intensity and timing of the peak on the permeability of the membrane. Therefore I do not think the distance is the explanation for what authors observe.

Shielding would provide the reverse effect, since hyperpolarization begins in the dense centres of the biofilms. For the initial 2 hours the cells receive negligible blue light. Neither of the referee’s comments thus seem tenable.

Result section 3: Emergence of synchronized global wavefronts in E. coli biofilms

In this section, the authors exposed a mature biofilm to blue light. They observe that the intensity peak is reached faster in the cells in the middle. They interpret this as the ion-channel-mediated wavefronts moved from the center of the biofilm. As above, cells in the middle can have different membrane permeability to those at the periphery, and probably even more importantly, there is no light profile shown anywhere in SI/Methods. I could be wrong, but the SI3 A profile is consistent with a potential Gaussian beam profile visible in the field of view. In Methods, I find the light source for the blue light and the type of microscope but no comments on how 'flat' the illumination is across their field of view. This is critical to assess what they are observing in this result section. I do find it interesting that the ThT intensity collapsed from the edges of the biofilms. In the publication I mentioned https://www.sciencedirect.com/science/article/pii/S0006349519308793#app2 [sciencedirect.com], the collapse of fluorescence was not understood (other than it is not membrane potential related). It was observed in Figure 5A, C, and F, that at the point of peak, electrochemical gradient of protons is already collapsed, and that at the point of peak cell expands and cytoplasmic content leaks out. This means that this part of the ThT curve is not membrane potential related. The authors see that after the first peak collapsed there is a period of time where ThT does not stain the cells and then it starts again. If after the first peak the cellular content leaks, as we have observed, then staining that occurs much later could be simply staining of cytoplasmic positively charged content, and the timing of that depends on the dynamics of cytoplasmic content leakage (we observed this to be happening over 2h in individual cells). ThT is also a non-specific amyloid dye, and in starving E. coli cells formation of protein clusters has been observed (https://pubmed.ncbi.nlm.nih.gov/30472191 [pubmed.ncbi.nlm.nih.gov]/), so such cytoplasmic staining seems possible.

>>It is very easy to see if the illumination is flat (Köhler illumination) by comparing the intensity of background pixels on the detector. It was flat in our case. Protons have little to do with our work for reasons highlighted before. Differential membrane permittivity is a speculative phenomenon not well supported by any evidence and with no clear molecular mechanism.

Finally, I note that authors observe biofilms of different shapes and sizes and state that they observe similar intensity profiles, which could mean that my comment on 'flatness' of the field of view above is not a concern. However, the scale bar in Figure 2A is not legible, so I can't compare it to the variation of sizes of the biofilms in Figure 2C (67 to 280um). Based on this, I think that the illumination profile is still a concern.

The referee now contradicts themselves and wants a scale bar to be more visible. We have changed the scale bar.

Result section 4: Voltage-gated Kch potassium channels mediate ion-channel electrical oscillations in E. coli

First I note at this point, given that I disagree that the data presented thus 'suggest that E. coli biofilms use electrical signaling to coordinate long-range responses to light stress' as the authors state, it gets harder to comment on the rest of the results.

In this result section the authors look at the effect of Kch, a putative voltage-gated potassium channel, on ThT profile in E. coli cells. And they see a difference. It is worth noting that in the publication https://www.sciencedirect.com/science/article/pii/S0006349519308793 [sciencedirect.com] it is found that ThT is also likely a substrate for TolC (Figure 4), but that scenario could not be distinguished from the one where TolC mutant has a different membrane permeability (and there is a publication that suggests the latter is happening https://onlinelibrary.wiley.com/doi/10.1111/j.1365-2958.2010.07245.x [onlinelibrary.wiley.com]). Given this, it is also possible that Kch deletion affects the membrane permeability. I do note that in video S4 I seem to see more of, what appear to be, plasmolysed cells. The authors do not see the ThT intensity with this mutant that appears long after the initial peak has disappeared, as they see in WT. It is not clear how long they waited for this, as from Figure S3C it could simply be that the dynamics of this is a lot slower, e.g. Kch deletion changes membrane permeability.

The work that TolC provides a possible passive pathway for ThT to leave cells seems slightly niche. It just demonstrates another mechanism for the cells to equilibriate the concentrations of ThT in a Nernstian manner i.e. driven by the membrane voltage.

The authors themselves state that the evidence for Kch being a voltage-gated channel is indirect (line 54). I do not think there is a need to claim function from a ThT profile of E. coli mutants (nor do I believe it's good practice), given how accurate single-channel recordings are currently. To know the exact dependency on the membrane potential, ion channel recordings on this protein are needed first.

We have good evidence form electrical impedance spectroscopy experiments that Kch increases the conductivity of biofilms  (https://pubs.acs.org/doi/10.1021/acs.nanolett.3c04446, 'Electrical impedance spectroscopy with bacterial biofilms: neuronal-like behavior', E.Akabuogu et al, ACS Nanoletters, 2024, in print.

Result section 5: Blue light influences ion-channel mediated membrane potential events in E. coli

In this chapter the authors vary the light intensity and stain the cells with PI (this dye gets into the cells when the membrane becomes very permeable), and the extracellular environment with K+ dye (I have not yet worked carefully with this dye). They find that different amounts of light influence ThT dynamics. This is in line with previous literature (both papers I have been mentioning: Figure 4 https://www.sciencedirect.com/science/article/pii/S0006349519303923 [sciencedirect.com] and https://ars.els-cdn.com/content/image/1-s2.0-S0006349519308793-mmc6.pdf [ars.els-cdn.com] especially SI12), but does not add anything new. I think the results presented here can be explained with previously published theory and do not indicate that the ion-channel mediated membrane potential dynamics is a light stress relief process.

The simple Nernstian battery model proposed by Pilizota et al is erroneous in our opinion for reasons outlined above. We believe it will prove to be a dead end for bacterial electrophysiology studies.

Result section 6: Development of a Hodgkin-Huxley model for the observed membrane potential dynamics

This results section starts with the authors stating: 'our data provide evidence that E. coli manages light stress through well-controlled modulation of its membrane potential dynamics'. As stated above, I think they are instead observing the process of ThT loading while the light is damaging the membrane and thus simultaneously collapsing the electrochemical gradient of protons. As stated above, this has been modelled before. And then, they observe a ThT staining that is independent from membrane potential.

This is an erroneous niche opinion. Protons have little say in the membrane potential since there are so few of them. The membrane potential is mostly determined by K+.

I will briefly comment on the Hodgkin Huxley (HH) based model. First, I think there is no evidence for two channels with different activation profiles as authors propose. But also, the HH model has been developed for neurons. There, the leakage and the pumping fluxes are both described by a constant representing conductivity, times the difference between the membrane potential and Nernst potential for the given ion. The conductivity in the model is given as gK*n^4 for potassium, gNa*m^3*h sodium, and gL for leakage, where gK, gNa and gL were measured experimentally for neurons. And, n, m, and h are variables that describe the experimentally observed voltage-gated mechanism of neuronal sodium and potassium channels. (Please see Hodgkin AL, Huxley AF. 1952. Currents carried by sodium and potassium ions through the membrane of the giant axon of Loligo. J. Physiol. 116:449-72 and Hodgkin AL, Huxley AF. 1952. A quantitative description of membrane current and its application to conduction and excitation in nerve. J. Physiol. 117:500-44).

In the 70 years since Hodgkin and Huxley first presented their model, a huge number of similar models have been proposed to describe cellular electrophysiology. We are not being hyperbolic when we state that the HH models for excitable cells are like the Schrödinger equation for molecules. We carefully adapted our HH model to reflect the currently understood electrophysiology of E. coli.

Thus, in applying the model to describe bacterial electrophysiology one should ensure near equilibrium requirement holds (so that (V-VQ) etc terms in authors' equation Figure 5 B hold), and potassium and other channels in a given bacterium have similar gating properties to those found in neurons. I am not aware of such measurements in any bacteria, and therefore think the pump leak model of the electrophysiology of bacteria needs to start with fluxes that are more general (for example Keener JP, Sneyd J. 2009. Mathematical physiology: I: Cellular physiology. New York: Springer or https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0000144 [journals.plos.org])

The reference is to a slightly more modern version of a simple Nernstian battery model. The model will not oscillate and thus will not help modelling membrane potentials in bacteria. We are unsure where the equilibrium requirement comes from (inadequate modelling of the dynamics?)

Result section 7: Mechanosensitive ion channels (MS) are vital for the first hyperpolarization event in E. coli.

The results that Mcs channels affect the profile of ThT dye are interesting. It is again possible that the membrane permeability of these mutants has changed and therefore the dynamics have changed, so this needs to be checked first. I also note that our results show that the peak of ThT coincides with cell expansion. For this to be understood a model is needed that also takes into account the link between maintenance of electrochemical gradients of ions in the cell and osmotic pressure.

The evidence for permeability changes in the membranes seems to be tenuous.

A side note is that the authors state that the Msc responds to stress-related voltage changes. I think this is an overstatement. Mscs respond to predominantly membrane tension and are mostly nonspecific (see how their action recovers cellular volume in this publication https://www.pnas.org/doi/full/10.1073/pnas.1522185113 [pnas.org]). Authors cite references 35-39 to support this statement. These publications still state that these channels are predominantly membrane tension-gated. Some of the references state that the presence of external ions is important for tension-related gating but sometimes they gate spontaneously in the presence of certain ions. Other publications cited don't really look at gating with respect to ions (39 is on clustering). This is why I think the statement is somewhat misleading.

We have reworded the discussion of Mscs since the literature appears to be ambiguous. We will try to run some electrical impedance spectroscopy experiments on the Msc mutants in the future to attempt to remove the ambiguity.

Result section 8: Anomalous ion-channel-mediated wavefronts propagate light stress signals in 3D E. coli biofilms.

I am not commenting on this result section, as it would only be applicable if ThT was membrane potential dye in E. coli.

Ok, but we disagree on the use of ThT.

Aims achieved/results support their conclusions:

The authors clearly present their data. I am convinced that they have accurately presented everything they observed. However, I think their interpretation of the data and conclusions is inaccurate in line with the discussion I provided above.

Likely impact of the work on the field, and the utility of the methods and data to the community:

I do not think this publication should be published in its current format. It should be revised in light of the previous literature as discussed in detail above. I believe presenting it in it's current form on eLife pages would create unnecessary confusion.

We believe many of the Pilizota group articles are scientifically flawed and are causing the confusion in the literature.

Any other comments:

I note, that while this work studies E. coli, it references papers in other bacteria using ThT. For example, in lines 35-36 authors state that bacteria (Bacillus subtilis in this case) in biofilms have been recently found to modulate membrane potential citing the relevant literature from 2015. It is worth noting that the most recent paper https://journals.asm.org/doi/10.1128/mbio.02220-23 [journals.asm.org] found that ThT binds to one or more proteins in the spore coat, suggesting that it does not act as a membrane potential in Bacillus spores. It is possible that it still reports membrane potential in Bacillus cells and the recent results are strictly spore-specific, but these should be kept in mind when using ThT with Bacillus.

>>ThT was used successfully in previous studies of normal B. subtilis cells (by our own group and A.Prindle, ‘Spatial propagation of electrical signal in circular biofilms’, J.A.Blee et al, Physical Review E, 2019, 100, 052401, J.A.Blee et al, ‘Membrane potentials, oxidative stress and the dispersal response of bacterial biofilms to 405 nm light’, Physical Biology, 2020, 17, 2, 036001, A.Prindle et al, ‘Ion channels enable electrical communication in bacterial communities’, Nature, 2015, 527, 59-63). The connection to low metabolism pore research seems speculative.

Reviewer #3 (Public Review):

It has recently been demonstrated that bacteria in biofilms show changes in membrane potential in response to changes in their environment, and that these can propagate signals through the biofilm to coordinate bacterial behavior. Akabuogu et al. contribute to this exciting research area with a study of blue light-induced membrane potential dynamics in E. coli biofilms. They demonstrate that Thioflavin-T (ThT) intensity (a proxy for membrane potential) displays multiphasic dynamics in response to blue light treatment. They additionally use genetic manipulations to implicate the potassium channel Kch in the latter part of these dynamics. Mechanosensitive ion channels may also be involved, although these channels seem to have blue light-independent effects on membrane potential as well. In addition, there are challenges to the quantitative interpretation of ThT microscopy data which require consideration. The authors then explore whether these dynamics are involved in signaling at the community level. The authors suggest that cell firing is both more coordinated when cells are clustered and happens in waves in larger, 3D biofilms; however, in both cases evidence for these claims is incomplete. The authors present two simulations to describe the ThT data. The first of these simulations, a Hodgkin-Huxley model, indicates that the data are consistent with the activity of two ion channels with different kinetics; the Kch channel mutant, which ablates a specific portion of the response curve, is consistent with this. The second model is a fire-diffuse-fire model to describe wavefront propagation of membrane potential changes in a 3D biofilm; because the wavefront data are not presented clearly, the results of this model are difficult to interpret. Finally, the authors discuss whether these membrane potential changes could be involved in generating a protective response to blue light exposure; increased death in a Kch ion channel mutant upon blue light exposure suggests that this may be the case, but a no-light control is needed to clarify this.

In a few instances, the paper is missing key control experiments that are important to the interpretation of the data. This makes it difficult to judge the meaning of some of the presented experiments.

(1) An additional control for the effects of autofluorescence is very important. The authors conduct an experiment where they treat cells with CCCP and see that Thioflavin-T (ThT) dynamics do not change over the course of the experiment. They suggest that this demonstrates that autofluorescence does not impact their measurements. However, cellular autofluorescence depends on the physiological state of the cell, which is impacted by CCCP treatment. A much simpler and more direct experiment would be to repeat the measurement in the absence of ThT or any other stain. This experiment should be performed both in the wild-type strain and in the ∆kch mutant.

ThT is a very bright fluorophore (much brighter than a GFP). It is clear from the images of non-stained samples that autofluorescence provides a negligible contribution to the fluorescence intensity in an image.

(2) The effects of photobleaching should be considered. Of course, the intensity varies a lot over the course of the experiment in a way that photobleaching alone cannot explain. However, photobleaching can still contribute to the kinetics observed. Photobleaching can be assessed by changing the intensity, duration, or frequency of exposure to excitation light during the experiment. Considerations about photobleaching become particularly important when considering the effect of catalase on ThT intensity. The authors find that the decrease in ThT signal after the initial "spike" is attenuated by the addition of catalase; this is what would be predicted by catalase protecting ThT from photobleaching (indeed, catalase can be used to reduce photobleaching in time lapse imaging).

Photobleaching was negligible over the course of the experiments. We employed techniques such as reducing sample exposure time and using the appropriate light intensity to minimize photobleaching.

(3) It would be helpful to have a baseline of membrane potential fluctuations in the absence of the proposed stimulus (in this case, blue light). Including traces of membrane potential recorded without light present would help support the claim that these changes in membrane potential represent a blue light-specific stress response, as the authors suggest. Of course, ThT is blue, so if the excitation light for ThT is problematic for this experiment the alternative dye tetramethylrhodamine methyl ester perchlorate (TMRM) can be used instead.

Unfortunately the fluorescent baseline is too weak to measure cleanly in this experiment. It appears the collective response of all the bacteria hyperpolarization at the same time appears to dominate the signal (measurements in the eLife article and new potentiometry measurements).

(4) The effects of ThT in combination with blue light should be more carefully considered. In mitochondria, a combination of high concentrations of blue light and ThT leads to disruption of the PMF (Skates et al. 2021 BioRXiv), and similarly, ThT treatment enhances the photodynamic effects of blue light in E. coli (Bondia et al. 2021 Chemical Communications). If present in this experiment, this effect could confound the interpretation of the PMF dynamics reported in the paper.

We think the PMF plays a minority role in determining the membrane potential in E. coli. For reasons outlined before (H+ is a minority ion in E. coli compared with K+).

(5) Figures 4D - E indicate that a ∆kch mutant has increased propidium iodide (PI) staining in the presence of blue light; this is interpreted to mean that Kch-mediated membrane potential dynamics help protect cells from blue light. However, Live/Dead staining results in these strains in the absence of blue light are not reported. This means that the possibility that the ∆kch mutant has a general decrease in survival (independent of any effects of blue light) cannot be ruled out.

>>Both strains of bacterial has similar growth curve and also engaged in membrane potential dynamics for the duration of the experiment. We were interested in bacterial cells that observed membrane potential dynamics in the presence of the stress. Bacterial cells need to be alive to engage in membrane potential  dynamics (hyperpolarize) under stress conditions. Cells that engaged in membrane potential dynamics and later stained red were only counted after the entire duration. We believe that the wildtype handles the light stress better than the ∆kch mutant as measured with the PI.

(6) Additionally in Figures 4D - E, the interpretation of this experiment can be confounded by the fact that PI uptake can sometimes be seen in bacterial cells with high membrane potential (Kirchhoff & Cypionka 2017 J Microbial Methods); the interpretation is that high membrane potential can lead to increased PI permeability. Because the membrane potential is largely higher throughout blue light treatment in the ∆kch mutant (Fig. 3AB), this complicates the interpretation of this experiment.

Kirchhoff & Cypionka 2017 J Microbial Methods, using fluorescence microscopy, suggested that changes in membrane potential dynamics can introduce experimental bias when propidium iodide is used to confirm the viability of tge bacterial strains, B subtilis (DSM-10) and Dinoroseobacter shibae, that are starved of oxygen (via N2 gassing) for 2 hours. They attempted to support their findings by using CCCP in stopping the membrane potential dynamics (but never showed any pictoral or plotted data for this confirmatory experiment). In our experiment methodology, cell death was not forced on the cells by introducing an extra burden or via anoxia. We believe that the accumulation of PI in ∆kch mutant is not due to high membrane potential dynamics but is attributed to the PI, unbiasedly showing damaged/dead cells. We think that propidium iodide is good for this experiment. Propidium iodide is a dye that is extensively used in life sciences. PI has also been used in the study of bacterial electrophysiology (https://pubmed.ncbi.nlm.nih.gov/32343961/, ) and no membrane potential related bias was reported.

Throughout the paper, many ThT intensity traces are compared, and described as "similar" or "dissimilar", without detailed discussion or a clear standard for comparison. For example, the two membrane potential curves in Fig. S1C are described as "similar" although they have very different shapes, whereas the curves in Fig. 1B and 1D are discussed in terms of their differences although they are evidently much more similar to one another. Without metrics or statistics to compare these curves, it is hard to interpret these claims. These comparative interpretations are additionally challenging because many of the figures in which average trace data are presented do not indicate standard deviation.

Comparison of small changes in the absolute intensities is problematic in such fluorescence experiments. We mean the shape of the traces is similar and they can be modelled using a HH model with similar parameters.

The differences between the TMRM and ThT curves that the authors show in Fig. S1C warrant further consideration. Some of the key features of the response in the ThT curve (on which much of the modeling work in the paper relies) are not very apparent in the TMRM data. It is not obvious to me which of these traces will be more representative of the actual underlying membrane potential dynamics.

In our experiment, TMRM was used to confirm the dynamics observed using ThT. However, ThT appear to be more photostable than TMRM (especially towars the 2nd peak). The most interesting observation is that with both dyes, all phases of the membrane potential dynamics were conspicuous (the first peak, the quiescent period and the second peak). The time periods for these three episodes were also similar.

A key claim in this paper (that dynamics of firing differ depending on whether cells are alone or in a colony) is underpinned by "time-to-first peak" analysis, but there are some challenges in interpreting these results. The authors report an average time-to-first peak of 7.34 min for the data in Figure 1B, but the average curve in Figure 1B peaks earlier than this. In Figure 1E, it appears that there are a handful of outliers in the "sparse cell" condition that likely explain this discrepancy. Either an outlier analysis should be done and the mean recomputed accordingly, or a more outlier-robust method like the median should be used instead. Then, a statistical comparison of these results will indicate whether there is a significant difference between them.

The key point is the comparison of standard errors on the standard deviation.

In two different 3D biofilm experiments, the authors report the propagation of wavefronts of membrane potential; I am unable to discern these wavefronts in the imaging data, and they are not clearly demonstrated by analysis.

The first data set is presented in Figures 2A, 2B, and Video S3. The images and video are very difficult to interpret because of how the images have been scaled: the center of the biofilm is highly saturated, and the zero value has also been set too high to consistently observe the single cells surrounding the biofilm. With the images scaled this way, it is very difficult to assess dynamics. The time stamps in Video S3 and on the panels in Figure 2A also do not correspond to one another although the same biofilm is shown (and the time course in 2B is also different from what is indicated in 2B). In either case, it appears that the center of the biofilm is consistently brighter than the edges, and the intensity of all cells in the biofilm increases in tandem; by eye, propagating wavefronts (either directed toward the edge or the center) are not evident to me. Increased brightness at the center of the biofilm could be explained by increased cell thickness there (as is typical in this type of biofilm). From the image legend, it is not clear whether the image presented is a single confocal slice or a projection. Even if this is a single confocal slice, in both Video S3 and Figure 2A there are regions of "haze" from out-of-focus light evident, suggesting that light from other focal planes is nonetheless present. This seems to me to be a simpler explanation for the fluorescence dynamics observed in this experiment: cells are all following the same trajectory that corresponds to that seen for single cells, and the center is brighter because of increased biofilm thickness.

We appreciate the reviewer for this important observation. We have made changes to the figures to address this confusion. The cell cover has no influence on the observed membrane potential dynamics. The entire biofilm was exposed to the same blue light at each time. Therefore all parts of the biofilm received equal amounts of the blue light intensity. The membrane potential dynamics was not influenced by cell density (see Fig 2C). 

The second data set is presented in Video S6B; I am similarly unable to see any wave propagation in this video. I observe only a consistent decrease in fluorescence intensity throughout the experiment that is spatially uniform (except for the bright, dynamic cells near the top; these presumably represent cells that are floating in the microfluidic and have newly arrived to the imaging region).

A visual inspection of Video S6B shows a fast rise, a decrease in fluorescence and a second rise (supplementary figure 4B). The data for the fluorescence was carefully obtained using the imaris software. We created a curved geometry on each slice of the confocal stack. We analyzed the surfaces of this curved plane along the z-axis. This was carried out in imaris.

3D imaging data can be difficult to interpret by eye, so it would perhaps be more helpful to demonstrate these propagating wavefronts by analysis; however, such analysis is not presented in a clear way. The legend in Figure 2B mentions a "wavefront trace", but there is no position information included - this trace instead seems to represent the average intensity trace of all cells. To demonstrate the propagation of a wavefront, this analysis should be shown for different subpopulations of cells at different positions from the center of the biofilm. Data is shown in Figure 8 that reflects the velocity of the wavefront as a function of biofilm position; however, because the wavefronts themselves are not evident in the data, it is difficult to interpret this analysis. The methods section additionally does not contain sufficient information about what these velocities represent and how they are calculated. Because of this, it is difficult for me to evaluate the section of the paper pertaining to wave propagation and the predicted biofilm critical size.

The analysis is considered in more detail in a more expansive modelling article, currently under peer review in a physics journal, ‘Electrical signalling in three dimensional bacterial biofilms using an agent based fire-diffuse-fire model’, V.Martorelli, et al, 2024 https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

There are some instances in the paper where claims are made that do not have data shown or are not evident in the cited data:

(1) In the first results section, "When CCCP was added, we observed a fast efflux of ions in all cells"- the data figure pertaining to this experiment is in Fig. S1E, which does not show any ion efflux. The methods section does not mention how ion efflux was measured during CCCP treatment.

We have worded this differently to properly convey our results.

(2) In the discussion of voltage-gated calcium channels, the authors refer to "spiking events", but these are not obvious in Figure S3E. Although the fluorescence intensity changes over time, it's hard to distinguish these fluctuations from measurement noise; a no-light control could help clarify this.

The calcium transients observed were not due to noise or artefacts.

(3) The authors state that the membrane potential dynamics simulated in Figure 7B are similar to those observed in 3D biofilms in Fig. S4B; however, the second peak is not clearly evident in Fig. S4B and it looks very different for the mature biofilm data reported in Fig. 2. I have some additional confusion about this data specifically: in the intensity trace shown in Fig. S4B, the intensity in the second frame is much higher than the first; this is not evident in Video S6B, in which the highest intensity is in the first frame at time 0. Similarly, the graph indicates that the intensity at 60 minutes is higher than the intensity at 4 minutes, but this is not the case in Fig. S4A or Video S6B.

The confusion stated here has now been addressed. Also it should be noted that while Fig 2.1 was obtained with LED light source, Fig S4A was obtained using a laser light source. While obtaining the confocal images (for Fig S4A ), the light intensity was controlled to further minimize photobleaching. Most importantly, there is an evidence of slow rise to the 2nd peak in Fig S4B. The first peak, quiescence and slow rise to second peak are evident.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Scientific recommendations:

- Although Fig 4A clearly shows that light stimulation has an influence on the dynamics of cell membrane potential in the biofilm, it is important to rule out the contribution of variations in environmental parameters. I understand that for technical reasons, the flow of fresh medium must be stopped during image acquisition. Therefore, I suggest performing control experiments, where the flow is stopped before image acquisition (15min, 30min, 45min, and 1h before). If there is no significant contribution from environmental variations (pH, RedOx), the dynamics of the electrical response should be superimposed whatever the delay between stopping the flow stop and switching on the light.

In this current research study, we were focused on studying how E. coli cells and biofilms react to blue light stress via their membrane potential dynamics. This involved growing the cells and biofilms, stopping the media flow and obtaining data immediately. We believe that stopping the flow not only helped us to manage data acquisition, it also helped us reduce the effect of environmental factors. In our future study we will expand the work to include how the membrane potential dynamics evolve in the presence of changing environmental factors for example such induced by stopping the flow at varied times.

- Since TMRM signal exhibits a linear increase after the first response peak (Supplementary Figure 1D), I recommend mitigating the statement at line 78.

- To improve the spatial analysis of the electrical response, I suggest plotting kymographs of the intensity profiles across the biofilm. I have plotted this kymograph for Video S3 and it appears that there is no electrical propagation for the second peak. In addition, the authors should provide technical details of how R^2(t) is measured in the first regime (Figure 7E).

See the dedicated simulation article for more details. https://www.biorxiv.org/content/10.1101/2023.11.17.567515v1

- Line 152: To assess the variability of the latency, the authors should consider measuring the variance divided by the mean instead of SD, which may depend on the average value.

We are happy with our current use of standard error on the standard deviation. It shows what we claim to be true.

- Line 154-155: To truly determine whether the amplitude of the "action potential" is independent of biofilm size, the authors should not normalise the signals.

Good point. We qualitatively compared both normalized and unnormalized data. Recent electrical impedance spectroscopy measurements (unpublished) indicate that the electrical activity is an extensive quantity i.e. it scales with the size of the biofilms.

- To precise the role of K+ in the habituation response, I suggest using valinomycin at sub-inhibitory concentrations (10µM). Besides, the high concentration of CCCP used in this study completely inhibits cell activity. Not surprisingly, no electrical response to light stimulation was observed in the presence of CCCP. Finally, the Kch complementation experiment exhibits a "drop after the first peak" on a single point. It would be more convincing to increase the temporal resolution (1min->10s) to show that there is indeed a first and a second peak.

An interesting experiment for the future.

- Line 237-238: There are only two points suggesting that the dynamics of hyperpolarization are faster at higher irradiance(Fig 4A). The authors should consider adding a third intermediate point at 17µW/mm^2 to confirm the statement made in this sentence.

Multiple repeats were performed. We are confident of the robustness of our data.

- Line 249 + Fig 4E: It seems that the data reported on Fig 4E are extracted from Fig 4D. If this is indeed the case, the data should be normalised by the total population size to compare survival probabilities under the two conditions. It would also be great to measure these probabilities (for WT and ∆kch) in the presence of ROS scavengers.

- To distinguish between model fitting and model predictions, the authors should clearly state which parameters are taken from the literature and which parameters are adjusted to fit the experimental data.

- Supplementary Figure 4A: why can't we see any wavefront in this series of images?

For the experimental data, the wavefront was analyzed by employing the imaris software. We systematically created a ROI with a curved geometry within the confocal stack (the biofilm). The fluorescence of ThT was traced along the surface of the curved geometry was analyzed along the z-axis.

- Fig 7B: Could the authors explain why the plateau is higher in the simulations than in the biofilm experiments? Could they add noise on the firing activities?

See the dedicated Martorelli modelling article. In general we would need to approach stochastic Hodgkin-Huxley modelling and the fluorescence data (and electrical impedance spectroscopy data) presented does not have extensive noise (due to collective averaging over many bacteria cells).

- Supplementary Figure 4B: Why can't we see the second peak in confocal images?

The second peak is present although not as robust as in Fig 2B. The confocal images were obtained with a laser source. Therefore we tried to create a balance between applying sufficient light stress on the bacterial cells and mitigating photobleaching.

Editing recommendations:

The editing recommendations below has been applied where appropriate

- Many important technical details are missing (e.g. R^2, curvature, and 445nm irradiance measurements). Error bars are missing from most graphs. The captions should clearly indicate if these are single-cell or biofilm experiments, strain name, illumination conditions, number of experiments, SD, or SE. Please indicate on all panels of all figures in the main text and in the supplements, which are the conditions: single cell vs. biofilm, strains, medium, centrifugal vs centripetal etc..., where relevant. Please also draw error bars everywhere.

We have now made appropriate changes. We specifically use cells when we were dealing with single cells and biofilms when we worked on biofilms. We decided to describe the strain name either on the panel or the image description.

- Line 47-51: The way the paragraph is written suggests that no coordinated electrical oscillations have been observed in Gram-negative biofilms. However, Hennes et al (referenced as 57 in this manuscript) have shown that a wave of hyperpolarized cells propagates in Neisseria gonorrhoea colony, which is a Gram-negative bacterium.

We are now aware of this work. It was not published when we first submitted our work and the authors claim the waves of activity are due to ROS diffusion NOT propagating waves of ions (coordinated electrical wavefronts).

- Line 59: "stressor" -> "stress" or "perturbation".

The correction has been made.

- Line 153: Please indicate in the Material&Methods how the size of the biofilm is measured.

The biofilm size was obtained using BiofilmQ and the step by step guide for using BiofilmQ were stated..

- Figure 2A: Please provide associated brightfield images to locate bacteria.

- Line 186: Please remove "wavefront" from the caption. Fig2B only shows the average signal as a function of time.

This correction has been implemented.

- Fig 3B,C: Please indicate single cell and biofilm on the panels and also WT and ∆kch.

- Line 289: I suggest adding "in single cell experiments" to the title of this section.

- Fig 5A: blue light is always present at regular time intervals during regime I and II. The presence of blue light only in regime I could be misleading.

- Fig 5C: The curve in Fig 5D seems to correspond to the biofilm case. The curve given by the model, should be compared with the average curve presented in Fig 1D.

- Fig 6A, B, and C: These figures could be moved to supplements.

- Line 392: Replace "turgidity" with "turgor pressure".

- Fig 7C,E: Please use a log-log scale to represent these data and indicate the line of slope 1.

- Fig 7E: The x-axis has been cropped.

- Please provide a supplementary movie for the data presented in Fig 7E.

- Line 455: E. Coli biofilms do not express ThT.

- Line 466: "\gamma is the anomalous exponent". Please remove anomalous (\gamma can equal 1 at this stage).

- Line 475: Please replace "section" with "projection".

- Line 476: Please replace "spatiotemporal" with "temporal". There is no spatial dependency in either figure.

- Line 500: Please define Eikonal approximation.

- Fig 8 could be moved to supplements.

- Line 553: "predicted" -> "predict".

- Line 593: Could the authors explain why their model offers much better quantitative agreement?

- Line 669: What does "universal" mean in that context?

- Line 671: A volume can be pipetted but not a concentration.

- Line 676: Are triplicates technical or biological replicates?

- Sup Fig1: Please use minutes instead of seconds in panel A.

- Model for membrane dynamics: "The fraction of time the Q+ channel is open" -> "The dynamics of Q+ channel activity can be written". Ditto for K+ channel...

- Model for membrane dynamics: "the term ... is a threshold-linear". This function is not linear at all. Why is it called linear? Also, please describe what \sigma is.

- ABFDF model: "releasing a given concentration" -> "releasing a local concentration" or "a given number" but it's not \sigma anymore. Besides, this \sigma is unlikely related to the previous \sigma used in the model of membrane potential dynamics in single cells. Please consider renaming one or the other. Also, ions are referred to as C+ in the text and C in equation 8. Am I missing something?

Reviewer #2 (Recommendations For The Authors):

I have included all my comments as one review. I have done so, despite the fact that some minor comments could have gone into this section, because I decided to review each Result section. I thus felt that not writing it as one review might be harder to follow. I have however highlighted which comments are minor suggestions or where I felt corrections.

However, while I am happy with all my comments being public, given their nature I think they should be shown to authors first. Perhaps the authors want to go over them and think about it before deciding if they are happy for their manuscript to be published along with these comments, or not. I will highlight this in an email to the editor. I question whether in this case, given that I am raising major issues, publishing both the manuscript and the comments is the way to go as I think it might just generate confusion among the audience.

Reviewer #3 (Recommendations For The Authors):

I was unable to find any legends for any of the supplemental videos in my review materials, and I could not open supplemental video 5.

I made some comments in the public review about the analysis and interpretation of the time-to-fire data. One of the other challenges in this data set is that the time resolution is limited- it seems that a large proportion of cells have already fired after a single acquisition frame. It would be ideal to increase the time resolution on this measurement to improve precision. This could be done by imaging more quickly, but that would perhaps necessitate more blue light exposure; an alternative is to do this experiment under lower blue light irradiance where the first spike time is increased (Figure 4A).

In the public review, I mentioned the possible impact of high membrane potential on PI permeability. To address this, the experiment could be repeated with other stains, or the viability of blue light-treated cells could be addressed more directly by outgrowth or colony-forming unit assays.

In the public review, I mentioned the possible combined toxicity of ThT and blue light. Live/dead experiments after blue light exposure with and without ThT could be used to test for such effects, and/or the growth curve experiment in Figure 1F could be repeated with blue light exposure at a comparable irradiance used in the experiment.

Throughout the paper and figure legends, it would help to have more methodological details in the main text, especially those that are critical for the interpretation of the experiment. The experimental details in the methods section are nicely described, but the data analysis section should be expanded significantly.

At the end of the results section, the authors suggest a critical biofilm size of only 4 µm for wavefront propagation (not much larger than a single cell!). The authors show responses for various biofilm sizes in Fig. 2C, but these are all substantially larger. Are there data for cell clusters above and below this size that could support this claim more directly?

The authors mention image registration as part of their analysis pipeline, but the 3D data sets in Video S6B and Fig. S4A do not appear to be registered- were these registered prior to the velocity analysis reported in Fig. 8?

One of the most challenging claims to demonstrate in this paper is that these membrane potential wavefronts are involved in coordinating a large, biofilm-scale response to blue light. One possible way to test this might be to repeat the Live/Dead experiment in planktonic culture or the single-cell condition. If the protection from blue light specifically emerges due to coordinated activity of the biofilm, the Kch mutant would not be expected to show a change in Live/Dead staining in non-biofilm conditions.

Line 140: How is "mature biofilm" defined? Also on this same line, what does "spontaneous" mean here?

Line 151: "much smaller": Given that the reported time for 3D biofilms is 2.73 {plus minus} 0.85 min and in microclusters is 3.27 {plus minus} 1.77 min, this seems overly strong.

Line 155: How is "biofilm density" characterized? Additionally, the data in Figure 2C are presented in distance units (µm), but the text refers to "areal coverage"- please define the meaning of these distance units in the legend and/or here in the text (is this the average radius?).

Lines 161-162: These claims seem strong given the data presented before, and the logic is not very explicit. For example, in the second sentence, the idea that this signaling is used to "coordinate long-range responses to light stress" does not seem strongly evidenced at this point in the paper. What is meant by a long-range response to light stress- are there processes to respond to light that occur at long-length scales (rather than on the single-cell scale)? If so, is there evidence that these membrane potential changes could induce these responses? Please clarify the logic behind these conclusions.

Lines 235-236: In the lower irradiance conditions, the responses are slower overall, and it looks like the ThT intensity is beginning to rise at the end of the measurement. Could a more prominent second peak be observed in these cases if the measurement time was extended?

Line 242-243: The overall trajectories of extracellular potassium are indeed similar, but the kinetics of the second peak of potassium are different than those observed by ThT (it rises some minutes earlier)- is this consistent with the idea that Kch is responsible for that peak? Additionally, the potassium dynamics also reflect the first peak- is this surprising given that the Kch channel has no effect on this peak?

Line 255-256: Again, this seems like a very strong claim. There are several possible interpretations of the catalase experiment (which should be discussed); this experiment perhaps suggests that ROS impacts membrane potential, but does not obviously indicate that these membrane potential fluctuations mitigate ROS levels or help the cells respond to ROS stress. The loss of viability in the ∆kch mutant might indicate a link between these membrane potential experiments and viability, but it is hard to interpret without the no-light control I mention in the public review.

Lines 313-315: "The model predicts... the external light stress". Please clarify this section. Where this prediction arises from in the modeling work? Second, I am not sure what is meant by "modulates the light stress" or "keeps the cell dynamics robust to the intensity of external light stress" (especially since the dynamics clearly vary with irradiance, as seen in Figure 4A).

Line 322: I am not sure what "handles the ROS by adjusting the profile of the membrane potential dynamics" means. What is meant by "handling" ROS? Is the hypothesis that membrane potential dynamics themselves are protective against ROS, or that they induce a ROS-protective response downstream, or something else? Later in lines 327-8 the authors write that changes in the response to ROS in the model agree with the hypothesis, but just showing that ROS impacts the membrane potential does not seem to demonstrate that this has a protective effect against ROS.

Line 365-366: This section title seems confusing- mechanosensitive ion channels totally ablate membrane potential dynamics, they don't have a specific effect on the first hyperpolarization event. The claim that mechanonsensitive ion channels are specifically involved in the first event also appears in the abstract.

Also, the apparent membrane potential is much lower even at the start of the experiment in these mutants- is this expected? This seems to imply that these ion channels also have a blue light independent effect.

Lines 368, 371: Should be VGCCs rather than VGGCs.

Line 477: I believe the figure reference here should be to Figure 7B, not 6B.

Line 567-568: "The initial spike is key to registering the presence of the light stress." What is the evidence for this claim?

Line 592-594: "We have presented much better quantitative agreement..." This is a strong claim; it is not immediately evident to me that the agreement between model and prediction is "much better" in this work than in the cited work. The model in Figure 4 of reference 57 seems to capture the key features of their data. Clarification is needed about this claim.

Line 613: "...strains did not have any additional mutations." This seems to imply that whole genome sequencing was performed- is this the case?

Line 627: I believe this should refer to Figure S2A-B rather than S1.

Line 719: What percentage of cells did not hyperpolarize in these experiments?

Lines 751-754: As I mentioned above, significant detail is missing here about how these measurements were made. How is "radius" defined in 3D biofilms like the one shown in Video S6B, which looks very flat? What is meant by the distance from the substrate to the core, since usually in this biofilm geometry, the core is directly on the substrate? Most importantly, this only describes the process of sectioning the data- how were these sections used to compute the velocity of ThT signal propagation?

I also have some comments specifically on the figure presentation:

Normalization from 0 to 1 has been done in some of the ThT traces in the paper, but not all. The claims in the paper would be easiest to evaluate if the non-normalized data were shown- this is important for the interpretation of some of the claims.

Some indication of standard deviation (error bars or shading) should be added to all figures where mean traces are plotted.

Throughout the paper, I am a bit confused by the time axis; the data consistently starts at 1 minute. This is not intuitive to me, because it seems that the blue light being applied to the cells is also the excitation laser for ThT- in that case, shouldn't the first imaging frame be at time 0 (when the blue light is first applied)? Or is there an additional exposure of blue light 1 minute before imaging starts? This is consequential because it impacts the measured time to the first spike. (Additionally, all of the video time stamps start at 0).

Please increase the size of the scale bars and bar labels throughout, especially in Figure 2A and S4A.

In Figure 1B and D, it would help to decrease the opacity on the individual traces so that more of them can be discerned. It would also improve clarity to have data from the different experiments shown with different colored lines, so that variability between experiments can be clearly visualized.

Results in Figure 1E would be easier to interpret if the frequency were normalized to total N. It is hard to tell from this graph whether the edges and bin widths are the same between the data sets, but if not, they should be. Also, it would help to reduce the opacity of the sparse cell data set so that the full microcluster data set can be seen as well.

Biofilm images are shown in Figures 2A, S3A, and Video S3- these are all of the same biofilm. Why not take the opportunity to show different experimental replicates in these different figures? The same goes for Figure S4A and Video S6B, which again are of the same biofilm.

Figure 2C would be much easier to read if the curves were colored in order of their size; the same is true for Figure 4A and irradiance.

The complementation data in Figure S3D should be moved to the main text figure 3 alongside the data about the corresponding knockout to make it easier to compare the curves.

Fig.ure S3E: Is the Y-axis in this graph mislabeled? It is labeled as ThT fluorescence, but it seems that it is reporting fluorescence from the calcium indicator?

Video S6B is very confusing- why does the video play first forwards and then backwards? Unless I am looking very carefully at the time stamps it is easy to misinterpret this as a rise in the intensity at the end of the experiment. Without a video legend, it's hard to understand this, but I think it would be much more straightforward to interpret if it only played forward. (Also, why is this video labeled 6B when there is no video 6A?)

Reviewer #1 (Public Review):

Summary:

In this study of metabolism using Xenopus, explanted porcine hearts and limbs, and human organs-on-chips, Sperry et al studied the ability of WB3 to slow metabolism and mobility. The group developed WB3, an analog of SNC80, void of SNC80's delta-opioid receptor binding capacity and studied its metabolic impact. The authors concluded that SNC80 and its analog WB3 can induce "biostasis" and produce a hypometabolic state which holds promise for prolonging organ viability in transplant surgery as well as other potential clinical benefits.

Strengths:

This study also opens new avenues for therapeutic possibilities in areas such as trauma, acute infection, and brain injuries. The overall methodology is acceptable, but certain concerns should be addressed.

Weaknesses:

Major comments:

(1) In cardiac and renal transplantation, cold preservation in ice remains a common practice for transporting explanted hearts to donors which remains a cheap and easily accessible way of preserving organs. While ex-vivo mechanical circulatory platforms have been developed and are increasingly being utilized to prolong organ viability, cold preservation remains widely used. The authors perfused explanted hearts with oxygenated perfusion preservation devices at subnormothermic temperatures (20-23C) which is even much lower than routinely used in clinical cardiopulmonary bypass scenarios (28-32C) (in the discussion, the authors allude to SNC80's possible "protective effect" in cardiac bypass). It is unclear how much of the hypometabolic state is related to WB3 administration versus hypothermia. The study will benefit from a comparison of WB3 administration and hypothermia in Xenopus, explanted porcine organs versus cold preservation alone to show distinction in biostasis parameters.

(2) The authors selected SNC80 based on a literature survey where it was identified based on its ability to induce hypothermia and protect against the effects of spinal cord ischemia in rodents. While this makes sense, were other drugs (eg. Puerarin) considered? The induction of hypothermia and spinal cord protective effect of SNC80 may be multifactorial and not necessarily related to its biostatic effects as the authors describe. Please provide some more context into the background of SNC80.

(3) In most of the models, the primary metric that the authors utilize to characterize metabolic activity is oxygen consumption, which is a somewhat limited indicator. For instance, this does not provide any information, however, on anaerobic metabolic activity. In addition, the ATP/ADP ratio was found to decrease in the organ chips where SNC80 was utilized, but similar findings were not presented for the other models.

(4) The authors should provide a more detailed explanation of SNC80's mechanisms of interaction with proteins related to transmembrane transport, mitochondrial activity, and metabolic processes. What is the impact of SNC80 on mitochondrial function, particularly ATP production and mitochondrial respiration? Are there changes in mitochondrial membrane potential, electron transport chain activity, or oxidative phosphorylation? In this context, authors discuss the potential role of NCX1 as a binding target for SNC80 and its various mechanisms in slowing metabolism. However, no experiments have been done to confirm this binding in the present study. Co-immunoprecipitation studies using appropriate antibodies against SNC80 and NCX1 should be considered to demonstrate their direct binding. Additionally, surface plasmon resonance (SPR) or isothermal titration calorimetry (ITC) experiments could be employed to quantify the binding affinity between SNC80 and NCX1, providing further evidence of their interaction. These experiments would elucidate the binding mechanism between SNC80 and NCX1 and reveal more information on the mechanism of action for SNC80.

(5) The manuscript notes that histological analysis was conducted, but it seems that only example images are provided, such as Fig 4f. Quantified histological data would provide a more thorough understanding of tissue integrity.

(6) Some of the points mentioned in the discussion and conclusion are rather strong and based on possible associations such as SNC80's potential vasodilatory capacity conferring a cardioprotective effect, ability to reversibly suppress metabolism across different temperatures and species. Please tone this down and stay limited to the organs studied. Further, the reversibility of the findings may be more objectively assessed by biomarkers with decreased immunofluorescence in response to ischemia such as troponin I for heart and albumin for liver. Additionally, an investigation of proteins involved in inflammation, hypoxia, and key cell death pathways using immunohistochemistry analysis can better describe the impact of treatment on apoptosis/necroptosis.

(7) What could be the underlying cause of the observed increase in intercellular spacing after SNC80 administration in porcine limbs which also seems to be evident in the heart histology samples? This seems to be more prominent in the SNC80 compared to the vehicle group.

(8) In the Discussion section, it would be valuable to provide a concise interpretation of the lipidomic data, particularly explaining how changes in acylcarnitine and cholesterol ester levels may relate to tadpole metabolism, hibernation, or other biological processes.

(9) What are the limitations or disadvantages of the study? Does SNC80 possess any immunomodulatory properties that might affect the outcomes of organ transplantation? Are there specific organs for which SNC80 may not be a suitable preservation agent, and if so, what are the reasons behind this?

Comments on revised version:

The authors have satisfactorily addressed our comments in the rebuttal letter. The limitations described by the authors in point #9, however, need to be incorporated in the revised manuscript in detail as they are important in guiding interpretation of the present data. Congratulations again on the important study.

eLife assessment

Pharmacological induction of physiological slowing combined with organ perfusion systems could provide a novel therapeutic strategy for tissue and organ preservation. Using a Xenopus model, the authors provide important findings on a use of drug to slow down metabolism for the purpose of organ preservation. The authors provide compelling evidence that SNC80 can rapidly and reversibly slow biochemical and metabolic activities while preserving cell and tissue viability. This approach may be beneficial for transplantation, trauma management, and improving organ survival in remote and low-resource settings

Reviewer #2 (Public Review):

Summary:

This manuscript titled "Identification of pharmacological inducers of a reversible hypometabolic state for whole organ preservation" reports the effects of delta opioid receptor activator SNC80 and its modified analog WB3 with ~1,000 times less delta opioid receptor binding activity on metabolic state.

Strengths:

This is an interesting study with potentially broad implications for organ preservation.

Weaknesses:

However, there are several limitations which raise concerns.

(1) The authors developed an analog of a known delta opioid receptor activator SNC80 with three orders of magnitude lesser binding with the delta opioid receptor WB3. This will likely reduce the undesirable effects of SNC80 while preserving metabolic slowing needed for organ preservation. Yet, most experiments were done with SNC80, not the superior modification, WB3, shown in only a limited set of experiments, Figure 3.

(2) The heart is one of the most challenging organs to preserve, and some experiments are done to establish the metabolic effects of SNC80. However, the biodistribution study, shown in Figure 2, conspicuously omitted the heart.

(3) I do not understand the design of the electrophysiology and contractility experiments with the porcine hearts. How did you defibrillate the hearts after removal and establishing perfusion? Lines 173-175 on Page 7 state: "After defibrillation with epinephrine, the P and QRS waveforms were visible in ECGs from 3 of 4 SNC80-treated hearts (Table S1), suggesting that those hearts regain atrial and ventricular polarization." Please clarify. Defibrillation is done with an electric shock. Also, please show the ECG recordings to support your conclusions about "polarization." What did you mean by "polarization"? Depolarization? Repolarization? Or resting potential. To establish a normal physiological state, please show ECG waveforms and present data on basic ECG characteristics: heart rate, PQ and QT intervals, and P and QRS durations. I recommend perfusion of the porcine heart with WB3, not only SNC80.

(4) Pathology data also raises concerns. The histology images shown in Figure 4f are not quantified, and they show apparently higher levels of tissue disruption in SNC80-treated tissue vs vehicle-treated. The test (lines 169-171) confirms this concern: "In some hearts treated with SNC80, greater waviness of muscle fibers was observed, possibly indicating a state of muscle contraction." It will be helpful to measure markers of apoptosis and necrosis and to apply TTC viability staining.

(5) The apparent state of contracture suggests a higher degree of myocardial damage and a high intracellular calcium level in SNC80-treated hearts. The authors suggested that the sodium-calcium exchanger NCX is a possible target of SNC80 and could be responsible for the "hypometabolic state." However, NCX1 is critically important in the extrusion of cytosolic Ca2+ during the diastolic phase. Failure to remove excessive calcium and restore ionic homeostasis would lead to calcium overload and heart failure.

(6) I am surprised the authors did not consider using the gold standard assay for measuring mitochondrial function in cells by the Seahorse Cell Mito Stress Test.

Comments on revised version:

I am satisfied with the revisions. The authors addressed major concerns with new data and/or provided satisfactory rebuttal.

Author response:

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

Public Reviews:

Reviewer #1:

(1) In cardiac and renal transplantation, cold preservation in ice remains a common practice for transporting explanted hearts to donors which remains a cheap and easily accessible way of preserving organs. While ex-vivo mechanical circulatory platforms have been developed and are increasingly being utilized to prolong organ viability, cold preservation remains widely used. The authors perfused explanted hearts with oxygenated perfusion preservation devices at subnormothermic temperatures (20-23C) which is even much lower than routinely used in clinical cardiopulmonary bypass scenarios (28-32C) (in the discussion, the authors allude to SNC80's possible "protective effect" in cardiac bypass). It is unclear how much of the hypometabolic state is related to WB3 administration versus hypothermia. The study will benefit from a comparison of WB3 administration and hypothermia in Xenopus, explanted porcine organs versus cold preservation alone to show distinction in biostasis parameters.

Indeed, we expect that both pharmaceutical interventions and cooling could contribute to a hypometabolic state. To assess this, the controls and the treated groups were exposed to the same temperatures for both the Xenopus (18C) and porcine heart experiments (20-23C). Therefore, we can conclude that any changes in the treatment group relative to control can be attributed to the introduction of SNC80 or WB3 and not from cooling alone.  

(2) The authors selected SNC80 based on a literature survey where it was identified based on its ability to induce hypothermia and protect against the effects of spinal cord ischemia in rodents. While this makes sense, were other drugs (eg. Puerarin) considered? The induction of hypothermia and spinal cord protective effect of SNC80 may be multifactorial and not necessarily related to its biostatic effects as the authors describe. Please provide some more context into the background of SNC80.

During our research program, we considered and tested other drugs (>100 existing compounds in Xenopus screens). Although the published hypothermic and tissue protective effects suggested to us that SNC80 should be included in screening, it was not until we observed effects across multiple test parameters, systems, and species that we honed in on SNC80 as a lead compound. We have added additional information to further clarify the background of SNC80 on pgs. 3-4. 

(3) In most of the models, the primary metric that the authors utilize to characterize metabolic activity is oxygen consumption, which is a somewhat limited indicator. For instance, this does not provide any information, however, on anaerobic metabolic activity. In addition, the ATP/ADP ratio was found to decrease in the organ chips where SNC80 was utilized, but similar findings were not presented for the other models. 

We thank reviewers for their important point. We have therefore added additional experiments, including the Seahorse Mitostress assay for the four human cell types (Caco-2, Huh7, LSEC and HUVEC) used in the Organ Chip systems. We have added a description and an interpretation of the results in the section, Stasis induction in cultured human cells and tissues and mention the role of glycolysis and cytosolic reductive carboxylation as compensatory mechanisms.  Although the ATP/ADP ratio gave us useful insight into Huh-7 cells and chips metabolic activity, this method requires transfection and live imaging which does not suit other models such as Xenopus, or whole organs. Additionally, in animal models there may be other confounding factors that might influence ATP/ADP.

(4) The authors should provide a more detailed explanation of SNC80's mechanisms of interaction with proteins related to transmembrane transport, mitochondrial activity, and metabolic processes. What is the impact of SNC80 on mitochondrial function, particularly ATP production and mitochondrial respiration? Are there changes in mitochondrial membrane potential, electron transport chain activity, or oxidative phosphorylation? In this context, the authors discuss the potential role of NCX1 as a binding target for SNC80 and its various mechanisms in slowing metabolism. However, no experiments have been done to confirm this binding in the present study. Coimmunoprecipitation studies using appropriate antibodies against SNC80 and NCX1 should be considered to demonstrate their direct binding. Additionally, surface plasmon resonance (SPR) or isothermal titration calorimetry (ITC) experiments could be employed to quantify the binding affinity between SNC80 and NCX1, providing further evidence of their interaction. These experiments would elucidate the binding mechanism between SNC80 and NCX1 and reveal more information on the mechanism of action for SNC80. 

We agree that further definition of the mechanism of action is an important next step for this work; however, it is far beyond the scope of the present study.

(5) The manuscript notes that histological analysis was conducted, but it seems that only example images are provided, such as Figure 4f. Quantified histological data would provide a more thorough understanding of tissue integrity. 

We have added quantified histological data to the manuscript that was performed by a clinician blinded to the groups and interventions (Figure 4f).

(6) Some of the points mentioned in the discussion and conclusion are rather strong and based on possible associations such as SNC80's potential vasodilatory capacity conferring a cardioprotective effect, and ability to reversibly suppress metabolism across different temperatures and species. Please tone this down and stay limited to the organs studied. Further, the reversibility of the findings may be more objectively assessed by biomarkers with decreased immunofluorescence in response to ischemia such as troponin I for the heart and albumin for the liver. Additionally, an investigation of proteins involved in inflammation, hypoxia, and key cell death pathways using immunohistochemistry analysis can better describe the impact of treatment on apoptosis/necroptosis. 

We have revised aspects of the Discussion and Conclusion to focus on the organs studied in the present work (pgs. 14-17). We agree that markers of inflammation, hypoxia, and cell death are critical for assessing tissue health post-treatment. We performed PCR to assess such markers (Figure 4e) and found reductions in inflammatory cytokine and injury biomarker levels. Although we agree that immunohistochemistry may be useful, such as for looking at any spatial patterns of injury, PCR offers broader dynamic range and higher sensitivity and therefore was chosen for this assay.

(7) What could be the underlying cause of the observed increase in intercellular spacing after SNC80 administration in porcine limbs which also seems to be evident in the heart histology samples? This seems to be more prominent in the SNC80 compared to the vehicle group. 

Since the muscle bundle areas of baseline and treated tissues were essentially the same, the increase in intracellular space in the SNC80-treated tissue suggests a compensatory reduction in muscle fiber diameter.  Intracellular metabolite concentrations have been shown to be quite stable over a large range of metabolic activities (Hochachka et al. 1998). As such, a reduction in metabolic activity induced by SNC80 may suggest reduction in the accumulation of intracellular metabolites. In order to maintain a stable intracellular metabolite concentration, water would have to be expelled accounting for the increased intracellular space.

P W Hochachka, G B McClelland, G P Burness, J F Staples, R K Suarez Comp Biochem Physiol B Biochem Mol Biol 120, 17–26 (1998).

(8) In the Discussion section, it would be valuable to provide a concise interpretation of the lipidomic data, particularly explaining how changes in acylcarnitine and cholesterol ester levels may relate to tadpole metabolism, hibernation, or other biological processes. 

An interpretation of the lipidomics data has been summarized in the Discussion (pg. 14).

(9) What are the limitations or disadvantages of the study? Does SNC80 possess any immunomodulatory properties that might affect the outcomes of organ transplantation? Are there specific organs for which SNC80 may not be a suitable preservation agent, and if so, what are the reasons behind this? 

This study is limited in two ways. The first is that we characterized the function of the donor pig heart outside of the body, and therefore future work will be required to verify the function and quality of the hearts after they have been transplanted. Secondly, SNC80 is not currently approved for use in clinical settings and during earlier pre-clinical trials of the drug, side effects including seizures were noted and its development was halted. It is hypothesized that these seizures are related to SNC80’s delta opioid activity, so we developed a new, non-opioid analog called WB3, which will be used in future work. We have added a description of the prior seizure findings to the text (pg. 5).

Based on assessment of tissue biomarkers by PCR, it seems that SNC80 does exhibit immunomodulating properties. Because organ transplant recipients are treated with strong immunosuppressants to prevent organ rejection, we anticipate that SNC80 would either further support this goal, have little additional effect, or reduce the amount of additional immunosuppressive drugs that would need to be administered. To date, our data does not suggest that there are specific organs for which SNC80 may not be a suitable preservation agent.     

Reviewer #2:

(1) The authors developed an analog of a known delta opioid receptor activator SNC80 with three orders of magnitude lesser binding with the delta opioid receptor WB3. This will likely reduce the undesirable effects of SNC80 while preserving the metabolic slowing needed for organ preservation. Yet, most experiments were done with SNC80, not the superior modification, WB3, shown in only a limited set of experiments, Figure 3.  

We included the WB3 studies in Xenopus to confirm that the biostatic activity is not mediated through the delta opioid receptor. We have only performed a limited number of experiments with WB3 because we are focused on improving its solubility so that it can be easily dissolved in common organ perfusates without DMSO, which we were able to use in the Xenopus experiments. 

(2) The heart is one of the most challenging organs to preserve, and some experiments are done to establish the metabolic effects of SNC80. However, the biodistribution study, shown in Figure 2, conspicuously omitted the heart. 

Thank you for this suggestion. We returned to the biodistribution study dataset and were able to measure uptake by the heart at the 1-hour time point. We observe an increase in uptake above levels observed for other tissues at 1 hour and at levels similar to the skeletal muscle at 2 hours (plot below). Unfortunately, the heart was not visible in a sufficient number of Xenopus tissue sections to reevaluate uptake at the 2-hour time point. We were also able to re-evaluate the lipidomics data for the heart. Acylcarnitine and cholesterol ester were not significantly different between vehicle and SNC80-treated groups. The lack of change in acylcarnitine is particularly important since its upregulation has been shown to be a marker for cardiovascular disease in humans (Deda et al. 2022). The expanded lipidomics data have been added to Figure 2.

Deda O, Panteris E, Meikopoulos T, Begou O, Mouskeftara T, Karagiannidis E, Papazoglou AS, Sianos G, Theodoridis G, Gika H. Correlation of serum acylcarnitines with clinical presentation and severity of coronary artery disease. Biomolecules. 2022 Feb 23;12(3):354.

Author response image 1.

(3) I do not understand the design of the electrophysiology and contractility experiments with the porcine hearts. How did you defibrillate the hearts after removal and establishing perfusion? Lines 173-175 on Page 7 state: "After defibrillation with epinephrine, the P and QRS waveforms were visible in ECGs from 3 of 4 SNC80-treated hearts (Table S1), suggesting that those hearts regain atrial and ventricular polarization." Please clarify.

Defibrillation is done with an electric shock. Also, please show the ECG recordings to support your conclusions about "polarization." What did you mean by "polarization"? Depolarization? Repolarization? Or resting potential. To establish a normal physiological state, please show ECG waveforms and present data on basic ECG characteristics: heart rate, PQ and QT intervals, and P and QRS durations. I recommend perfusion of the porcine heart with WB3, not only SNC80.  

Hearts were defibrillated by the application of a 10 to 30 Joule electrical shock delivered from internal paddles positioned at the right atrium (negative) across to the left ventricle (positive). Once rhythm was established, 0.5 ml of 1:1000 epinephrine was administered via the aortic inflow. Electrocardiogram (ECG) showed that both vehicle and SNC80-treated hearts exhibited irregular contractions after perfusate flush and during rewarming prior to defibrillation. After defibrillation (10-30 J electrical shock) followed by epinephrine, a regular heartbeat was established in 3 of 4 SNC80-treated hearts, exhibiting normal P and QRS waveforms (Table S1). That observation suggested that the intrinsic atrial and ventricular muscle fiber contractility was preserved, and the overall conduction system of the heart was viable. The pulse rates of SNC80-treated hearts were at or near normal for porcine hearts (70-120 beats/min) after defibrillation. Vehicle-treated hearts exhibited tachycardia following defibrillation, with all exhibiting pulse rates above the normal range for porcine hearts. We have added clarifying text and definitions (pg. 8). We have only performed a limited number of experiments with WB3 because we are focused on improving its solubility so that it can be easily dissolved in common organ perfusates without DMSO, which we were able to use in the Xenopus experiments.

(4) Pathology data also raises concerns. The histology images shown in Figure 4f are not quantified, and they show apparently higher levels of tissue disruption in SNC80-treated tissue vs vehicle-treated. The test (lines 169-171) confirms this concern: "In some hearts treated with SNC80, greater waviness of muscle fibers was observed, possibly indicating a state of muscle contraction."  

The histology images shown in Figure 4f were quantified and the myocardial injury score quantification show comparable histology between the groups.

(5) The apparent state of contracture suggests a higher degree of myocardial damage and a high intracellular calcium level in SNC80-treated hearts. 

The authors suggested that the sodium-calcium exchanger NCX is a possible target of SNC80 and could be responsible for the "hypometabolic state." However, NCX1 is critically important in the extrusion of cytosolic Ca2+ during the diastolic phase. Failure to remove excessive calcium and restore ionic homeostasis would lead to calcium overload and heart failure. 

The histological assessment doesn’t indicate a higher degree of myocardial damage in SNC80 treated hearts. Our data are not suggestive of high intracellular calcium buildup in SNC80treated hearts. If that were the case, we would have had challenges restoring the rhythm of the hearts on the Langendorff post-preservation, which was not observed.

(6) I am surprised the authors did not consider using the gold standard assay for measuring mitochondrial function in cells by the Seahorse Cell Mito Stress Test. 

Thank you for this important point. We have added data from the Seahorse Mitostress assay for the four human cell types (Caco-2, Huh7, LSEC and HUVEC) included in the Organ Chip experiments. We have added a description and an interpretation of the results in the section Stasis induction in cultured human cells and tissues. We now mention the role of glycolysis and cytosolic reductive carboxylation as compensatory mechanisms.   

Reviewer #3:

(1) The authors perform a literature search to identify SNC80 as a promising hit. However, the details of the literature search, a list of other potential hits, and the criteria for identification of SNC80 are not described. The hypometabolic effect of SNC80 exposure is well-characterized in the Xenopus model. Furthermore, the authors show that SNC80 localises to the brain, but do not discuss several studies that have pointed to convulsions induced by exposure to high doses of SCN80, and whether this would be apparent in the Xenopus studies. The authors have promising data on the WB3 morpholino that retains or even improves on the hypometabolism phenotype of SCN80 while likely not retaining delta opioid activity. However, this is not functionally demonstrated. Moreover, WB3 is not used in any of the other assays and models used in the study. In the setting of cardiac transplant surgery, co-administration of SNC80 reduces metabolic activity and inflammation, although it is unclear if there is an improvement in recovery of organ function due to SCN80.

Thank you for raising these important points. We have added details of the process to identify SNC80 (pgs. 3-4) and a discussion of the studies pointing to convulsions with high doses of SNC80 (pg. 5) (which were not observed in Xenopus studies). We have also incorporated measurements of oxygen consumption during WB3 treatment in Xenopus (Figure 3d).

(2) The reversible induction of hypometabolic status is also demonstrated in two different organ chips. These models could identify the differential response of epithelial cells and vascular cells to drug perfusion, but the authors have mostly focused on the former. Finally, the authors identify specific targets for the hypometabolic effect of SNC80, which is a valuable resource for other screening studies and can form the basis for future work. 

In the revised manuscript, we have also added data from the Seahorse Mitostress assay for the four human cell types (Caco-2, Huh7, LSEC and HUVEC). We have added a description and an interpretation of the results in the section Stasis induction in cultured human cells and tissues. We highlight the differences in metabolic response from the four cell types to SNC80 treatment. It is important to note that the metabolism-suppressing effects of SNC80 were most potent in the epithelial cells that were originally derived from highly metabolic tumors (Caco-2 and Huh7) versus primary normal endothelial cells (HUVEC and LSEC), which is also consistent with past work suggesting that targeting of the NCX1 channel might offer a way to slow tumor growth (Wan et al. 2022). Because we observed more prominent effects in epithelial cells in 2D assays, we decided to focus the 3D organ chips assays on epithelial cells.

Wan, H. et al. NCX1 coupled with TRPC1 to promote gastric cancer via Ca2+/AKT/β-catenin pathway. Oncogene (2022) doi:10.1038/s41388-022-02412-9.

Recommendations for the authors:

Reviewer #1:

(1) Line 136, "Based on these intriguing findings with human Organ Chips". No mention of human organ chips was made in the text at this point, suggest rewording.  

Thank you for identifying this error. We have revised this line (pg. 6).

(2) Please provide more information on previous studies that have explored other drugs for organ protection, the novelty of the findings of this study, and how the findings of this study compare to prior data. 

Building on the background of organ preservation drugs provided in the Introduction, we have added details to compare our outcomes to other drugs explored for organ protection (pg. 15).

(3) The dosing study in Supplemental Figure S1 provides some context on why the authors utilized the 100 uM SNC80 concentration. It would be helpful if the authors could elaborate in the Discussion on the mechanistic rationale for this concentration. 

This dose was chosen to maximize suppression of metabolic and activity parameters, while ensuring reversibility of biostasis. We have clarified this in the Discussion (pg. 14).

(4) In Supplement Figure S2a, the y-axis measures the relative metabolic rate. It seems from the text that this is a relative measure of oxygen consumption, but it should be clarified accordingly. 

We have clarified this point in the Methods section.  

(5) What is the specific time or time frame when the reversed effect of SNC80 is most pronounced or at its peak? 

When Xenopus are moved to fresh medium after SNC80 treatment, we observe a 15-minute period during which no reversal is evident from motion measurements. After that period, we observe a gradual, linear recovery over 2 hours. We cannot designate a specific period in which the reversal effect is most pronounced from these data.

(6) WB3 seems to show a faster and stronger impact on swimming in comparison to SNC80. What could be the potential reasons for this difference, and could this have any clinical implications? 

From our current data, we understand the key difference to be that SNC80 has greater affinity for the delta opioid receptor compared to WB3. Therefore, we hypothesize that by not interacting with the opioid system, WB3 induces faster and stronger impacts on swimming. In mice, it has been shown that SNC80 directly inhibits forebrain GABAergic neurons via activity at their delta opioid receptors, which leads to convulsions (Chung et al. 2015). Although we do not observe seizure-like behavior in Xenopus, drugs that inhibit GABAergic neurons can produce stimulant effects in vivo. Since WB3 has a lower affinity for the delta opioid receptor, it likely produces less stimulation, leading to faster and stronger suppression of swimming behaviors. Additionally, it is possible that WB3 interacts with additional targets we have not yet identified.

Chung PC, Boehrer A, Stephan A, Matifas A, Scherrer G, Darcq E, Befort K, Kieffer BL. Delta opioid receptors expressed in forebrain GABAergic neurons are responsible for SNC80-induced seizures. Behavioural brain research. 2015 Feb 1;278:429-34.

(7) Elaborate on the potential significance of SNC80's distribution in the GI tract, gill region, and skeletal muscle. How might this distribution relate to the observed physiological effects? 

In Xenopus tadpoles, we observe SNC80 uptake in the gill region and GI tract within 1 hour. The multiple possible routes of uptake in Xenopus (skin, gills, and mouth) may account for the relatively rapid physiological effects observed in our experiments. The uptake observed in the muscle may be specifically responsible for the slowed motion observed in Xenopus activity assays. This has been elaborated upon in the text (pg. 5).

(8) Please use italics where needed, e.g., in vitro, in vivo, etc. 

This has been updated throughout the article.

(9) Supplemental Figure S1 - Is there any reason for having 3 replicates for the 100uM compared to the 4 replicates in the other groups? 

Each group had 4 replicates; however, a review of the replicates for the 100 µM group suggested the presence of a leak or air bubble in one oxygen measurement vial, which, therefore, had to be excluded from the analysis.

(10) Figure 3 description - 'c' should be bold. 

Figure 3 has been updated.

Reviewer #3:

Title: The title suggests that several candidate compounds are identified but the study focuses primarily on SCN80. Please consider rephrasing to make it more specific to this molecule. Alternatively, the manuscript would be significantly strengthened if more data is provided for WB3. 

Although the study focuses on SNC80, we introduce an entirely novel molecule, WB3, and therefore, we feel it is more appropriate to indicate that multiple molecules were studied.

Line 58-59: please cite additional primary literature papers for the different therapeutics discussed. As an example, the authors do not cite or discuss Massen et al PMID: 31743376 which suggests that H2S is able to induce similar hypometabolic effects even at 37C. 

Thank you for this suggestion. We have expanded our discussion of primary literature paper for the therapeutics discussed (pg. 15).

Line 76 - 77: The authors do not provide any data on the other possible hits from their literature search or methods details on how this was done. No relevant literature has been cited. What criteria were used to finalise SNC80? 

During our research program, we considered and tested other drugs (>100 existing compounds in Xenopus screens). Although the published hypothermic and tissue-protective effects suggested that SNC80 should be included in screening, it was not until we observed effects across multiple test parameters, systems, and species that we honed in on SNC80 as a lead compound. We have added additional information to further clarify the background of SNC80 on pgs. 3-4.  

Line 85 and Lines 342-345 in the Discussion: SNC80 is reported to induce convulsions at high doses in rodents and primates - was this also evident in the Xenopus studies? How does the dose used in the Xenopus studies compare with the high dose (ca. 10 mg/kg) used in primate studies Danielson et al., PMID: 17112570? 

We did not observe convulsions in SNC80-treated Xenopus. However, we have updated the manuscript to include previous observations of convulsions in rodents and primates treated with SNC80 (pg. 5). Due to a number of differences, it is challenging to directly compare the dosing in Xenopus studies to those in the primate. In the present study, groups of 10 Xenopus were exposed to a 10 mL pool of 100 µM SNC80, which may be absorbed via oral, gill, and skin routes. Primates were dosed with 10 mg/kg delivered intramuscularly. Because these models may result in different drug biodistributions, any direct comparisons would be speculative. Further work in rodent models may help clarify the relevant dosing differences.

Line 117: what does 'double the concentration' mean? Is this with reference to the dose of SNC80? If so, is this sufficient to completely block opioid receptor activity? 

Yes, we meant that naltrindole was dosed at double the concentration of SNC80. We have clarified this in the text (pg. 5). Prior work in rodent brain tissue has shown that radiolabeled naltrindole binds to saturation at picomolar to nanomolar concentrations (Yamamura et al. 1992). To confirm our initial observations with naltrindole and SNC80, we also tested a SNC80 analog (WB3) with very low delta opioid activity (Figure 3), which showed similar effects.

Yamamura MS, Horvath R, Toth G, Otvos F, Malatynska E, Knapp RJ, Porreca F, Hruby VJ, Yamamura HI.

Characterization of [3H] naltrindole binding to delta opioid receptors in rat brain. Life sciences. 1992 Jan 1;50(16):PL119-24.

Figure 3c, d: It appears that WB3 is even more effective at rapidly reducing motion and inducing faster recovery which is an exciting result. However, in 3d it appears that longterm exposure of 8h has detrimental effects since the heart rate remains depressed. Please clarify. 

Yes, at 8 hours, we observe slow recovery and, in some cases, maintenance of depressed heart rates. This could be because the drug is more lipophilic and might remain in fat tissue for longer times. Although our current goal is to lengthen the time window for heart transplant surgery to 6 hours, we are working on formulating WB3 to optimize safety for longer applications (8+ hours).

Figure 4: the experiments with the heart transplants are well done, but do not demonstrate an additional protective effect over the current standard of care except for the reduced metabolism. Could the authors discuss this further in the discussion or provide data with WB83, which may show a stronger effect? Scale bars are missing in panel f.  

In addition to reduced metabolism, we also demonstrate reduced expression of inflammation, hypoxia, and cell death-related markers compared to machine perfusion alone (Figure 4e). The potential protective effect of the biostasis-inducing compounds will be further investigated in a planned orthotopic porcine transplant study where pigs will be followed up for 6 hours post weaning off a bypass machine allowing enough time to assess potential benefit of biostasisinducing drugs. Additionally, we have added scale bars (Figure 4f).

Order of manuscript: Line 136 already refers to the organ-chip data, which is only presented at the end. Please edit. I feel the manuscript would flow better with the organchip data presented before the heart transplant data. 

Organ-chip data: this is an important component of the story but is only shown in supplementary figures. Consider showing this data in the main figures, as eLife has no space restrictions. Furthermore, it is unclear if the effluent collected and analysed is from apical or vascular, or both. In any case, the analysis via microscopy-based methods appears restricted to the epithelium. The manuscript would be significantly strengthened by providing some data on the effect of SNC80 on vascular cells. 

As requested, we have moved the Organ Chips results to a main figure (new Fig. 5). We have added additional experiments, including the Seahorse Mitostress assay for the four human cell types (Caco-2, Huh7, LSEC and HUVEC). We have added a description and an interpretation of the results in the section Stasis induction in cultured human cells and tissues. The 2D assays showed that metabolism-suppressing effects of SNC80 were most potent in the epithelial cells that were originally derived from highly metabolic tumors (Caco-2 and Huh7) versus endothelial cells (HUVEC and LSEC). Based on these results, we decided to focus the 3D organ chips assays on epithelial cells only, and hence only analyzed effluents from the epithelial (apical) channel.

Methods section for fabrication of oxygen sensors: Please refer to prior papers from your lab (Grant et al., PMID: 35274118) with regards to details of the fabrication of the devices with inbuilt oxygen sensors. 

The methods used for the fabrication of oxygen sensors will be included in a separate manuscript currently in preparation.  

Figure S3 and Line 243-244: Please provide the data for untreated control organ chips in panels d and e a mean value for which is quoted in the main text. The images in panel f are too small for the reader to appreciate the point, please provide zooms. Scalebars are also missing from these images. Please increase the number of replicates for S3f - the liver-chip data has only two replicates which has very low power for statistical testing. In general, the number of organ chips used for the data for each panel is missing. 

As mentioned in the captions, Figure S3 (now Figure S5) panels d and e show average albumin production of Liver Chips at day 7-10 of culture. These measurements were performed before any treatment with SNC80 to characterize the chip’s functional metabolism. In panel g, although we only show biological N=2-3, each datapoint corresponds to an average of multiple fields of view (multiple technical replicates). We have now clarified this in the figure legend.

Figure S4 - I do not quite understand why the perfusion with the vehicle only also affects oxygen release in the liver chip. Is it possible to use a different vehicle? 

The liver and gut oxygen levels are not on the same y-axis (gut on the left and liver on the right). The oxygen fold change of the liver control chip is below 0.5, which is in the same range as the gut control chip (0 +/- 0.25). There is a natural variation in oxygen consumption over the lifetime of the chips (now Figure 5c), and untreated cells are metabolically active and consuming oxygen. The small drop observed suggests that liver chips may not have reached a stable oxygen consumption rate at the time of the experiment, whereas the gut chips have stabilized.  

Figure S5c-f: The units on the Y-axis are missing. 

Panels S5c-d (now Figure S6c-d) depict the percent cytotoxicity and are thus unitless. Panels S5e-h (now Figure S6e-h) show the effluent levels relative to baseline and are also unitless. We have updated the figure caption to clarify this.

eLife assessment

This important study highlights the importance of SLAM-SAP signaling in determining innate gamma-delta T cell sublineages and their T cell receptor repertoires. It uncovers the complex role of the SLAM-SAP pathway in developing specific gamma-delta T cell subsets. The evidence presented is compelling, backed by high-quality data obtained through advanced single cell proteogenomics techniques.This work will be of broad interest to immunologists.

Reviewer #1 (Public Review):

Summary:<br /> In this study the authors advance their previous findings on the role of the SLAM-SAP signaling pathway in the development and function of multiple innate-like gamma-delta T cell subsets. Using high throughput single cell proteogenomics approach, the authors uncover SAP-dependent developmental checkpoints, and the role of SAP signaling in regulating the diversion of γδ T cells into the αβ T cell developmental pathway. Finally, the authors define TRGV4/TRAV13-4(DV7)-expressing T cells as a novel, SAP-dependent Vγ4 γδT1 subset.

Strengths:<br /> This study furthers our understanding of the importance and complexity of the SLAM-SAP signaling pathway not only in the development of innate-like γδ T cells but also the how it potentially balances the γδ/αβ T cell lineage commitment. Additionally, this study reveals the role of SAP-dependent events in generation of γδ TCR repertoire.

Comments on revised version:

The conclusions of the study are supported by well thought-out experiments and compelling data.

Weaknesses:<br /> There are no major weakness in the study.

A few minor points:<br /> (1) In the subsets of the γδ T cells that exhibit reduced BLK expression in B6. SAP KO mice, have the authors examined the expression of Lck and/or Fyn?<br /> (2) Does BLK directly associate with SLAM F1 and or SLAM F6 receptors?<br /> (3) Given the emerging role of γδ T cells in host immunity, it will be useful if the authors add a discussion of how their findings are relevant in disease conditions such as in cancer.

The author has adequately addressed all the reviewers' comments.

Reviewer #2 (Public Review):

Summary:<br /> Mistri et al explore the role of SLAM-SAP signaling in the developmental programming of innate-like gd T cell subsets. Using proteo-genomics, they determined that abrogation of SLAM-SAP signaling altered that programming, reducing some IL-17 producing subsets, including a novel Vγ4 γδT1 subset, and diverting gdTCR-expressing precursors to the ab fate. Altogether, this is a very thorough, thoughtfully interpreted study that adds significantly to our understanding of the contribution of the SLAM-SAP pathway to lineage specification. A particularly interesting element is the role of SLAM-SAP in preventing gd17 progenitors from switching fates and adopting the ab fate.

Comments on revised version:

The authors have addressed the minor issues raised in the original submission.

Author response:

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

Reviewer #1:

(1)  In the subsets of the γδ T cells that exhibit reduced BLK expression in B6. SAP KO mice, have the authors examined the expression of Lck and/or Fyn? 

The reviewer raises an excellent point. We have included in the revised manuscript additional data on Lck and Fyn expression in our scRNAseq dataset in (new Suppl. Fig. 1 and new Suppl. Fig. 4). These data revealed that in contrast to Blk, which appears primarily restricted to the γδT17 clusters, Lck and Fyn exhibit a much broader distribution and lack restriction to specific clusters. We did note that, like Blk, Lck and Fyn transcripts were abundant in SAP-dependent C2 cluster cells. Pseudobulk analysis on the immature clusters revealed that, neither Fyn nor Lck expression level differences reached our cut-off of 0.5 log2 FC (log2 FC Blk = 1.06), leading us to conclude that Blk is particularly dependent on SAP. We did note, however, that the magnitude of Lck differential expression was close to the 0.5 log2 FC cut-off and that its expression was increased in B6.SAP-/- γδ T cells (Suppl. Fig. 4). These results have been added to lines 202-212 in the Results section and lines 491-499 in the Discussion section.

(2)  Does BLK directly associate with SLAM F1 and or SLAM F6 receptors? 

The reviewer raises an interesting question given previous reports that BLK, LCK, and FYN have all been implicated in γδ T cell development. While SAP has a well-known ability to recruit FYN to SLAMF1 and there is evidence of a similar SAP-mediated recruitment of LCK to SLAMF6, we are not aware of any evidence a SAP-BLK interaction or of a direct binding of BLK to SLAM family receptors. Future experiments to investigate this possiibility are certainly warranted. In the revised ms, we have included additional discussion of these possibilities (lines 491- 499).  

(3)  Given the emerging role of γδ T cells in host immunity, it would be useful if the authors could add a discussion of how their findings are relevant in disease conditions such as cancer. 

We agree and have included new text in the Introduction (lines 37-45). 

(4)  Delete repeated words in lines 546 and line 553. 

Thank you—this has been corrected in the revised manuscript.

Reviewer #2:

This is a very complete study and requires no additional experimentation. One thing to keep in mind in assessing the ultimate fate of the "ab wannabe cells" is that mechanisms exist to silence the gd TCR as cells differentiate to the DP stage and so their presence as diverted DP cells may not be evident by staining for gdTCR expression - and will only be evident transcriptomically. 

We appreciate this helpful comment from the reviewer which we will take into consideration in our future experimental design.

There are a couple of minor points to raise: 

(1)  Figure 3C is not called out in the text. 

Thank you—this has been corrected in the revised manuscript.

(2)  Line 546 - "dependent" is repeated.

Thank you—this has been corrected in the revised manuscript.

eLife assessment

This valuable study reports that while most plasmacytoid dendritic cells (pDCs) originate from common dendritic cell precursors, approximately 20% are derived from lymphoid progenitors shared with B cells. The methodology used and the evidence are solid, and further demonstrate the distinct transcription factor requirements and activities of this subset of pDCs, although the functional significance of this dendritic cell subset will require further elucidation. The findings will be of great interest for those interested in the developmental and functional biology of the immune system.

Reviewer #3 (Public Review):

Summary:

Plasmacytoid dendritic cells (pDCs) represent a specialized subset of dendritic cells (DCs) known for their role in producing type I interferons (IFN-I) in response to viral infections. It was believed that pDCs originated from common DC progenitors (CDP). However, recent studies by Rodrigues et al. (Nature Immunology, 2018) and Dress et al. (Nature Immunology, 2019) have challenged this perspective, proposing that pDCs predominantly develop from lymphoid progenitors expressing IL-7R and Ly6D. A minor subset of pDCs arising from CDP has also been identified as functionally distinct, exhibiting reduced IFN-I production but a strong capability to activate T cell responses. On the other hand, clonal lineage tracing experiments, as recently reported by Feng et al. (Immunity, 2022), have demonstrated a shared origin between pDCs and conventional DCs (cDCs), suggesting a contribution of common DC precursors to the pDC lineage.

In this context, Araujo et al. investigated the heterogeneity of pDCs in terms of both development and function. Their findings revealed that approximately 20% of pDCs originate from lymphoid progenitors common to B cells. Using Mb1-Cre x Bcl11a floxed mice, the authors demonstrated that the development of this subset of pDCs, referred to as "B-pDCs," relied on the transcription factor BCL11a. Functionally, B-pDCs exhibited a diminished capacity to produce IFN-I in response to TLR9 agonists but secreted more IL-12 compared to conventional pDCs. Moreover, B-pDCs, either spontaneously or upon activation, exhibited increased expression of activation markers (CD80/CD86/MHC-II) and a heightened ability to activate T cell responses in vitro compared to conventional pDCs. Finally, Araujo et al. characterized these B-pDCs at the transcriptomic level using bulk and single-cell RNA sequencing, revealing them as a unique subset of pDCs expressing certain B cell markers such as Mb1, as well as specific markers (Axl) associated with cells recently described as transitional DCs.<br /> Thus, in contrast to previous findings, this study posits that a small proportion of pDCs derive from B cell-committed lymphoid progenitors, and this subset of B-pDCs exhibits distinct functional characteristics, being less specialized in IFN-I production but rather in T cell activation.

Strengths:

Previously, the same research group delineated the significance of BCL11a as a critical transcription factor in pDC development (Ippolito et al., PNAS, 2014). This study elucidates the precise stage during hematopoiesis at which BCL11a expression becomes essential for the emergence of a distinct subset of pDCs, substantiated by robust genetic evidence in vivo. Furthermore, it underscores the shared developmental origin between pDCs and B cells, reinforcing prior research in the field that suggests a lymphoid origin of pDCs. Finally, this works attributes specific functional properties to pDCs originating from these lymphoid progenitors shared with B cells, emphasizing the early imprinting of functional heterogeneity during their development.

Weaknesses:

Using their Mb1-reporter mice, the authors demonstrate that YFP pDCs originating from lymphoid progenitors are functionally distinct from conventional pDCs, mostly in vitro, but their in vivo relevance remains unknown. As underlined by both reviewers I believe that it is crucial to investigate how Bcl11a conditional deficiency in Mb1 expressing cells affects the anti-viral immune response, for example, using the M-CoV infection model as described by Sulczewski et al. in Nature Immunology, 2023. The current in vivo data using TLR9 agonist and in vitro data using B-pDCs co-cultures with T cells insufficiently address what B-pDCs might be doing in infectious contexts.

Revisions:

I thank the authors for their responses to my questions and for addressing most of my comments clearly and thoroughly. However, one major question remains unanswered: What is the functional relevance of the subset of B-pDCs that they have characterized? This key question, also highlighted by the other reviewer, requires further investigation. The current in vivo data using TLR9 agonist and in vitro data using B-pDCs co-cultures with T cells insufficiently address what B-pDCs might be doing in infectious contexts.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Plasmacytoid dendritic cells (pDCs) represent a specialized subset of dendritic cells (DCs) known for their role in producing type I interferons (IFN-I) in response to viral infections. It was believed that pDCs originated from common DC progenitors (CDP). However, recent studies by Rodrigues et al. (Nature Immunology, 2018) and Dress et al. (Nature Immunology, 2019) have challenged this perspective, proposing that pDCs predominantly develop from lymphoid progenitors expressing IL-7R and Ly6D. A minor subset of pDCs arising from CDP has also been identified as functionally distinct, exhibiting reduced IFN-I production but a strong capability to activate T-cell responses. On the other hand, clonal lineage tracing experiments, as recently reported by Feng et al. (Immunity, 2022), have demonstrated a shared origin between pDCs and conventional DCs (cDCs), suggesting a contribution of common DC precursors to the pDC lineage.

In this context, Araujo et al. investigated the heterogeneity of pDCs in terms of both development and function. Their findings revealed that approximately 20% of pDCs originate from lymphoid progenitors common to B cells. Using Mb1-Cre x Bcl11a floxed mice, the authors demonstrated that the development of this subset of pDCs, referred to as "B-pDCs," relied on the transcription factor BCL11a. Functionally, B-pDCs exhibited a diminished capacity to produce IFN-I in response to TLR9 agonists but secreted more IL-12 compared to conventional pDCs. Moreover, B-pDCs, either spontaneously or upon activation, exhibited increased expression of activation markers (CD80/CD86/MHC-II) and a heightened ability to activate T-cell responses in vitro compared to conventional pDCs. Finally, Araujo et al. characterized these B-pDCs at the transcriptomic level using bulk and single-cell RNA sequencing, revealing them as a unique subset of pDCs expressing certain B cell markers such as Mb1, as well as specific markers (Axl) associated with cells recently described as transitional DCs.

Thus, in contrast to previous findings, this study posits that a small proportion of pDCs derive from B cell-committed lymphoid progenitors, and this subset of B-pDCs exhibits distinct functional characteristics, being less specialized in IFN-I production but rather in T cell activation.

Strengths:

Previously, the same research group delineated the significance of BCL11a as a critical transcription factor in pDC development (Ippolito et al., PNAS, 2014). This study elucidates the precise stage during hematopoiesis at which BCL11a expression becomes essential for the emergence of a distinct subset of pDCs, substantiated by robust genetic evidence in vivo. Furthermore, it underscores the shared developmental origin between pDCs and B cells, reinforcing prior research in the field that suggests a lymphoid origin of pDCs. Finally, this work attributes specific functional properties to pDCs originating from these lymphoid progenitors shared with B cells, emphasizing the early imprinting of functional heterogeneity during their development.

Weaknesses:

The authors delineate a subset of pDCs dependent on the BCL11a transcription factor, originating from lymphoid progenitors, and compare it to conventional pDCs, which they suggest differentiate from common DC progenitors of myeloid origin. However, this interpretation lacks support from the authors' data. Their single-cell RNA sequencing data identifies cells corresponding to progenitors (Prog2), from which the majority of pDCs, termed conventional pDCs, likely originate. This progenitor cell population expresses Il7r, Siglech, and Ly6D, but not Csfr1. The authors describe this progenitor as resembling a "pro-pDC myeloid precursor," yet these cells align more closely with lymphoid (Il7r+) progenitors described by Rodrigues et al. (Nature Immunology, 2018) and Dress et al. (Nature Immunology, 2019). Furthermore, analysis of their Mb1 reporter mice reveals that only a fraction of common lymphoid progenitors (CLP) express YFP, giving rise to a fraction of YFP+ pDCs. However, this does not exclude the possibility that YFP- CLP could also give rise to pDCs. The authors could address this caveat by attempting to differentiate pDCs from both YFP+ and YFP- CLPs in vitro in the presence of FLT3L. Additionally, transfer experiments using these lymphoid progenitors could be conducted in vivo to assess their differentiation potential in competitive settings.

Dear Reviewer 1, we appreciate your thoughtful comments. We made the decision to address the Prog2 cluster as “pro-pDC myeloid precursor” because despite its lack of CSFR-1, its CIPR similarity score showed highest transcriptional similarity to the population “SC.CDP.BM” (GEO accession number: GSM791114), which is shown to be Sca1- Flt3+ cKitlo.

A similar population identified as “common dendritic cell progenitor” is shown by Onai and colleagues (Onai et al. 2013, Immunity) to be capable of differentiating into pDCs by upregulating E2-2 and subsequently downregulating M-CSFR. In addition, we were unable to infer a developmental trajectory between Prog2 and B-pDCs using SimplePPT on Monocle3 (Figure 5B). Since we know our B-pDCs are CLP derived and most likely share a B cell progenitor population, we feel this lack of connectivity to the UMAP myeloid partition corroborates our assignment of Prog2 as a myeloid pDC progenitor (not CLP derived). Of note, recent work by Medina and colleagues has shown that while IL-7Rα knockout mice exhibit a block in B cell development at the all-lymphoid progenitor (ALP) stage, PDCA-1+ pDCs identified within the initially gated BLP population persisted (PLoS One, 2013), suggesting the IL7R chain is not required for the development of PDCA1+ cells. 

Using their Mb1-reporter mice, the authors demonstrate that YFP pDCs originating from lymphoid progenitors are functionally distinct from conventional pDCs, mostly in vitro, but their in vivo relevance remains unknown. It is crucial to investigate how Bcl11a conditional deficiency in Mb1-expressing cells affects the anti-viral immune response, for example, using the M-CoV infection model as described by Sulczewski et al. in Nature Immunology, 2023. Particularly, the authors suggest that their B-pDCs act as antigen-presenting cells involved in T-cell activation compared to conventional pDCs. However, these findings contrast with those of Rodrigues et al., who have shown that pDCs of myeloid origin are more effective than pDCs of lymphoid origin in activating T-cell responses. The authors should discuss these discrepancies in greater detail. It is also notable that B-PDCs acquire the expression of ID2 (Figure S3A), commonly a marker of conventional/myeloid DCs. The authors could analyze in more detail the acquisition of specific myeloid features (CD11c, CX3CR1) by this B-PDCs subset and discuss how the expression of ID2 may impair classical pDC features, as ID2 is a repressor of E2-2, a master regulator of pDC fate.

Both reviewers expressed the need to further investigate how Bcl11a conditional deficiency in Mb1-expressing cells affects anti-viral responses of B-pDCs. While the functional characterization of B-pDC in the context of infection could be highly informative, it is really outside the scope of the present study. Our discovery that B-pDCs expand robustly upon TLR-9 agonist challenges in vivo and can prime T cells in vitro efficiently, however, suggests that these cells might play an important role during viral infections or anti-cancer immunity.

Finally, through the analysis of their single-cell RNA sequencing data, the authors show that the subset of B-pDCs they identified expresses Axl, confirmed at the protein level. Given this specific expression profile, the authors suggest that B-pDCs are related to a previously described subset of transitional DCs, which were reported to share a common developmental path with pDCs, (Sulczewski et al. in Nature Immunology, 2023). While intriguing, this observation requires further phenotypic and functional characterization to substantiate this claim.

We agree with the reviewer’s comments. We are currently preparing a separate manuscript addressing the commonalities between human transitional DCs and murine non-conventional pDCs.

Reviewer #2 (Public Review):

Summary:

The origin of plasmatoid dendritic cells and their subclasses continues to be a debated field, akin to any immune cell field that is determined through the expression of surface markers (relative to clear subclass separation based on functional biology and experimentation). In this context, in this manuscript by Araujo et al, the authors attempt to demonstrate that a subtype of pDCs comes from lymphoid origin due to the presence of some B cell gene expression markers. They nomenclature these cells as B-pDCs. Strikingly, pDCs function via expression of IFNa where as B-pDCs do not express IFNa - thereby raising the question of what are their physiological or pathophysiological properties. B-pDCs also express AXL, a marker not seen in mouse pDCs but observed in human pDCs. Overall, using a combination of gene expression profiling of immune cells isolated from mice via RNA-seq and single-cell profiling the authors propose that B-pDCs are a novel subtype of pDCs in mice that were not previously identified and characterized.

Weaknesses:

My two points of discussion about this manuscript are as follows.

(1) How new are these observations that pDCs could also originate from common lymphoid progenitors. This fact has been previously outlined by many laboratories including Shigematsu et al, Immunity 2004. These studies in the manuscript can be considered new based on the single-cell profiling presented, only if the further characterization of the isolated B-pDCs is performed at the functional biology level. Overlapping gene expression profiles are often seen in developing immune cell types- especially when only evaluated at the RNA expression level- and can lead to cell type complexity (and identification of new cell types) that are not biologically and functionally relevant.

Dear reviewer 2, we appreciate your thoughtful comments. We believe our single cell seq analysis adds new information to the studies mentioned because of our broader approach to BM profiling. By using only one marker (PDCA1+), scRNA-seq allowed us to dissect not only several subpopulations of pDCs that to our knowledge were not previously dissected in mice, but also linked the transcriptional similarity of B-pDCs to myeloid derived pDCs (and even other myeloid cell types), as well as B cells.

(2) The authors hardly perform any experiments to interrogate the function of these B-pDCs. The discussion on this topic can be enhanced. Ideally, some biological experiments would confirm that B-pDCs are important.

Dear reviewer 2, we appreciate your thoughtful comment and agree about the need for further functional characterization of B-pDCs (please see comments directed to reviewer 1 above).

(1) Considering that Bcl11a conditional deficiency severely impacts the B cell lineage, there is a possibility that such an effect on B cells may indirectly influence pDC development. To address this, the authors could repeat their bone marrow transfer experiments in a competitive setting by mixing both Bcl11a WT and CKO BM cells (using congenic markers to track the origin of the BM cells) and then specifically assess whether BM cells originating from Bcl11a CKO donors have impaired pDC output.

Dear reviewer 2, while the comment above is valid (that the reduced number of mature B cells in our Bcl11a conditional knockout might indirectly impact B-pDC development), we and many others have previously shown that lack of transcriptional regulation of E2-2 and other pDC differentiation modulators by Bcl11a  (including ID2 and MTG16) intrinsically and selectively disrupts the pDC lineage. At the current stage, we feel rederiving Bcl11a cKOs and performing bone marrow transfers (which usually take several months) only to investigate indirect effects of B cells on pDC developments is outside the scope of this publication.

(2) As mentioned earlier, it is important to assess the potential of CLP, whether YFP- or YFP+, in their ability to give rise to pDCs both in vitro and in vivo. This is also crucial since the authors previously demonstrated that Bcl11a deficiency in all hematopoietic cells had a more drastic impact on pDC development than mb1-cre specific deficiency.

We agree the manuscript could be strengthened by differentiation experiments. However, in our previous publication (mentioned above by the reviewer), we specifically show that although fewer overall LSK progenitors were detected in Vav-Cre+ F/F mice, both MDP and CDP progenitor populations persisted within the Flt3+ compartment in cKO mice at percentages similar to controls. MDP (Lin– Flt3+ Sca-1− CD115+ c-kithi); CDP (Lin– Flt3+ Sca-1− CD115+ c-kitlo). This data confirms that CLPs give rise to a substantial pool of pDC subpopulations. Other works have shown this as well, both in vivo and in vitro (Wang et al. Immunity 2004;  Karsunky et al, JEM 2003, etc). We therefore feel that confirming the previous observations that CLPs can give rise to pDCs is unnecessary, as our main goal in this manuscript was to describe a new pDC subpopulation that emerges primarily from CD79a+ B cell biased progenitors.

(3) The authors show a more severe impact of Bcl11a CKO on pDC depletion in the spleen than in the BM. Is this effect specific to the spleen, or can it also be observed in lymph nodes? What is the overall impact of Bcl11a conditional deficiency on pDC distribution in tissues such as the liver and lung? These questions are important to address to understand whether the heterogeneity of pDCs is differentially affected by their localization.

We agree heterogeneity of pDCs can be affected by their microenvironment. Although phenotyping of lymph nodes in Bcl11a cKOs would greatly add to our manuscript, the genetically altered strains required are no longer being bred in our facility and resurrecting them from frozen sperm is outside the realm of this publication.

(4) Regarding the functional study of pDCs, as emphasized previously, it is important to assess the in vivo relevance of B-pDCs in infectious settings.

Dear reviewer 2, we appreciate your thoughtful comment. Please see our response directed to reviewer 1 above.

(5) The authors injected CpG-ODN into mice and analyzed pDC phenotype upon activation. It is important to note that upon activation, especially upon induction of IFN-I production in vivo, mPDCA1 expression is no longer specific to pDCs  (Blasius et al, Journal of Immunology, 2006). Therefore, to specifically characterize pDC phenotype upon activation, a differential gating strategy is required (CD11c, B220, Ly6C, and Siglec H) to ensure that bona fide pDCs are analyzed.

We agree with the reviewer that this would be a more appropriate characterization. Regarding PDCA1 promiscuity in activated states, we are not aware of any cell types that express very high levels of B220 and PDCA1 simultaneously other than pDCs. We therefore firmly believe that our assignment is valid. Interestingly, gating B220+ cells of Cpg challenged mice that show intermediate expression of PDCA1 results in an increase in the frequency of CD19+ B cells, which we were careful to avoid by gating only the cells that most strongly express PDCA1.

(6) How does pDC activation regulate their mb1 expression? Could conventional pDCs, upon activation, become B-PDCs? Could activation and induction of IFN-I production in vivo also affect CLP and increase the amount of YFP+ lymphoid progenitors and thus B-pDC output?

Dear reviewer, we agree with your concern, albeit beyond the scope of the present study. While changes in YFP MFI via flow cytometry upon vaccination was not substantial, we have included the following comment in the manuscript discussion, acknowledging the aforementioned possibility: “Of note, whether induction of IFN-I production in vivo could also affect CLP and increase the amount of YFP+ lymphoid progenitors and thus B-pDC output is unclear. Further research is required to answer this question.”

(7) If pDCs are preferentially expanding upon in vivo stimulation, it would be informative to assess their Ki67 profile. This is a surprising observation since pDCs are generally considered quiescent cells that were previously described to die in response to activation and IFN-I (Swiecki et al, Journal of Experimental Medicine, 2011).

We agree and have entered the following statement to address this concern: “Functionally, they expand more readily after TLR9 engagement than classical pDCs (either through increased proliferation or differentiation of other cell types) and excel at activating T cells in culture.”

(8) How does the conditional deficiency of BCL11a affect the production of IFN-I and IL-12 in vivo (serum) upon CpG-ODN stimulation?

Dear reviewer 2, we are currently unable to rederive the conditional knockout mouse strain in a timely fashion. However, our ELISA experiments performed under controlled in vitro activation conditions, along with the in vivo findings of Zhang et al.(PNAS 2017) warrants the hypothesis that B-pDCs most likely exhibit a similar cytokine secreting profile under inflammatory conditions.

(9) Given that B-PDCs show downregulation of pDC canonical markers, including IRF8 and TLR7, could the authors address how B-PDCs respond to TLR7 stimulation in vitro and assess a broader spectrum of cytokines produced by pDCs in response to such stimulation (IL-6, TNFa, CXCL10...)?

Dear reviewer 2, although expanding our findings to include B-pDC responses to TLR-7 stimulation would greatly enhance our manuscript, a technical deterrent stands in our way. As mentioned prior, sorting B-pDCs for new experiments using reporter YFP mice is currently not possible, as we have retired this mouse strain. Sorting of live CD79a+ BpDCs via FACS is also not feasible, as CD79a staining with most antibody clones requires permeabilization of cells for easier access to the intra-membrane portion of CD79a.

(10) It would be informative to compare scRNA sequencing data between control and Bcl11a CKO mice to ascertain their contribution to B-PDCs and whether this deficiency may affect other pDC clusters and/or progenitors.

We are unable to sort B-pDCs for new experiments, as we unfortunately retired the transgenic colony.

(11) Transitional DCs were reported to give rise to a subset of cDC2. Given that the authors claim that B-PDCs are related to this subset of transitional DCs, could the authors observe any YFP staining in cDC2 upon the generation of their BM chimeras?

We saw no YFP positivity in CD11c hi cells (cDCs) via flow or through scRNA-seq, indicating CD79a expression is unique in mature B cells and B-pDCs.

(12) Most of the statistical analysis is done with a student test. This requires a normal distribution of the sample which is highly unlikely given the size of the sample. Therefore, the authors shall rather use a non-parametric test (Mann Whitney) to compare their samples.

We agree and have redone our statistical analyses using non-parametric test (Mann Whitney).

eLife assessment

This important study examines E. coli growth and division, suggesting that inhomogeneous organization of ribosomes in the cytoplasm results in cell size-dependent growth rate perturbations. The work is conceptually appealing, but incomplete due to shortcomings in the experiments and modeling.

Reviewer #1 (Public Review):

Summary:

The manuscript by van Heerden et al. reports growth rate variations in the cell cycle of E. coli and links this variation to uneven ribosome concentrations in the cell at birth that arise from an uneven division of cell volumes between the daughter cells. The authors propose a model to explain the experimental data, whose main premises are the exclusion of ribosomes from the nucleoid volume and a linear dependence of the growth rate on ribosome concentration in the cell.

Strengths:

(1) The manuscript highlights an interesting aspect of growth rate variability in bacteria and proposes a mechanism for how this variation is homeostatically corrected.

(2) A sophisticated modeling to explain the experimental data.

Weaknesses:

(1) The experiments lack controls. A partially functional label (L9-mCherry) can make ribosomes much more limiting for growth than are not labeled ribosomes.

(2) The large variation of interdivision times 72-89 min in repeat experiments in Glc is problematic. Some parameters in the measurements related to cell growth appear not properly controlled. It is problematic for a work that aims to establish a new universal behavior related to cell growth.

(3) The authors have not provided convincing evidence that cells in their experiment grow in a steady state.

4) The findings are over-generalized. The existing data show the effects only at some growth rates, but the findings are presented as a new universal principle.

5) The model relies on many assumptions that are not clearly brought out and the choice of model parameters is questionable (in some cases, the parameters seem to contradict well-established experimental data, including the one from the experiments from the very same work). Small changes in parameters and various approximations can have large effects on the model's outcomes; without understanding these responses, the model has a rather limited value.

6) There appears to be a qualitative discrepancy between the model and the experimental data in Glc (the main condition studied). The model predicts that the cells born large have a specific elongation rate that is smaller than the average growth rate of cells, but it grows in time at the beginning of the cell cycle, while the experiments show a decreasing growth rate (Figure 1C, SI Figure S2).

Reviewer #2 (Public Review):

Summary:

This work demonstrates that when E. coli cells divide, and division is not quite symmetric, the smaller cell has a higher growth rate than the larger cell at the beginning, but not the end, of the cell cycle. The authors then demonstrate that smaller cells have a higher ribosome concentration than larger cells, which is consistent with the idea that the two cells receive roughly equal numbers of ribosomes at division because, as they also observe, ribosomes are excluded by the nucleoid from the middle of the mother cell. The experimental observations are reproduced by a mathematical model that assumes growth is driven by ribosome concentration, with contributions from metabolism and active feedback.

Strengths:

The work provides strong evidence in support of the growing consensus that cells correct size fluctuations by modulating growth rate, within a cell cycle and on a single-cell basis. It also offers a plausible explanation for the correction mechanism by showing that ribosomes are excluded from the middle of a mother cell and have a higher concentration in the smaller daughter cell. The work is clearly written and benefits from a strong coupling between the experimental and modeling results. It provides a solid contribution to the field of single-cell bacterial growth control.

Weaknesses:

Although the results strongly suggest it, the work does not explicitly demonstrate (e.g. by direct perturbation) that higher ribosome concentration is the cause of the higher growth rate. Also, it is unclear why an active compensation mechanism is needed in the model, i.e., why size-dependent growth (via ribosome concentration) does not correct growth rate perturbations within a cell cycle automatically.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

This article by Navratna et al. reports the first structure of human HGSNAT in an acetyl-CoA-bound state. Through careful structural analysis, the authors propose potential reasons why certain human mutations lead to lysosomal storage disorders and outline a catalytic mechanism. The structural data are of good quality, and the manuscript is clearly written. This study represents an important step toward understanding the mechanism of HGSNAT and is valuable to the field. I have the following suggestions:

(1) The authors should characterize whether the purified protein is active. Otherwise, how does one know if the detergent used maintains the protein in a biologically relevant state? The authors should at least attempt to do so. If these prove to be challenging, at the very least, the authors should try a cell-based assay to demonstrate that the GFP tag does not interfere with the function.

We have addressed these concerns in the revised version and mentioned these efforts in our previous response letter. We’re briefly mentioning them here again. We attempted measuring HGSNAT catalyzed reaction by monitoring the decrease in acetyl-CoA in the presence of D-glucosamine (acetyl group acceptor) using a coupled enzyme acetyl-CoA assay kit from SIGMA (MAK039) that converts acetyl-CoA to a fluorescent product measurable at Ex/Em of 535/587 nm. We noticed a decrease in the level of acetyl-CoA (gray) upon the addition of HGSNAT (red) (Rebuttal figure 1).

Author response image 1.

Acetyl-CoA levels in absence and presence of HGSNAT purified in digitonin. Decrease in the levels of 10 M acetyl-CoA was measured in presence of 10 M D-glucosamine and 30 nM HGSNAT at pH 7.5.

While optimizing the assay, Xu et al. (2024, Nat Struct Mol Biol) published structural and biochemical characterization of HGSNAT, showing that detergent-purified HGSNAT is active. In addition, we have shown by cryo-EM that GFP-tagged HGSNAT that we purified in detergent was already bound to the endogenous substrate ACO, an observation that has been observed by Xu et al., as well. Finally, we performed LC-MS on GFP-tagged HGSNAT purified in detergent to detect bound ACO, which could be further removed by dialysis. These results have been included in Figure S9. The endogenous binding of ACO to HGSNAT in detergent suggests that neither the tag nor detergent are detrimental to the function.

(2) In Figure 5, the authors present a detailed schematic of the catalytic cycle, which I find to be too speculative. There is no evidence to suggest that this enzyme undergoes isomerization, similar to a transporter, between open-to-lumen and open-to-cytosol states. Could it not simply involve some movements of side chains to complete the acetyl transfer?

We have already changed this figure in our latest submission. Perhaps the changes made were not obvious while reviewing. We agreed with this reviewer that the enzyme could likely achieve catalysis by simple side chain movements without undergoing extensive isomerization steps, as depicted in Figure 5. In the absence of data supporting large movements during the acetyl transfer reaction, old Figure 5 appeared speculative. Hence, we have edited Figure 5 in the revised version of the manuscript based on the observations we made in this study, and different states shown in the figure do not show any conformational changes and only depict acetyl transfer.

Reviewer #2 (Public Review):

Summary:

This work describes the structure of Heparan-alpha-glucosaminide N-acetyltransferase (HGSNAT), a lysosomal membrane protein that catalyzes the acetylation reaction of the terminal alpha-D-glucosamine group required for degradation of heparan sulfate (HS). HS degradation takes place during the degradation of the extracellular matrix, a process required for restructuring tissue architecture, regulation of cellular function and differentiation. During this process, HS is degraded into monosaccharides and free sulfate in lysosomes.

HGSNAT catalyzes the transfer of the acetyl group from acetyl-CoA to the terminal non-reducing amino group of alpha-D-glucosamine. The molecular mechanism by which this process occur has not been described so far. One of the main reasons to study the mechanism of HGSNAT is that multiple mutations spanning the entire sequence of the protein, such as, nonsense mutations, splice-site variants, and missense mutations lead to dysfunction that causes abnormal accumulation of HS within the lysosomes. This accumulation is a cause of mucopolysaccharidosis IIIC (MPS IIIC), an autosomal recessive neurodegenerative lysosomal storage disorder, for which there are no approved drugs or treatment strategies.

This paper provides a 3.26A structure of HGSNAT, determined by single-particle cryo-EM. The structure reveals that HGSNAT is a dimer in detergent micelles, and a density assigned to acetyl-CoA. The authors speculate about the molecular mechanism of the acetylation reaction, map the mutations known to cause MPS IIIC on the structure and speculate about the nature of the HGSNAT disfunction caused by such mutations.

Strengths:

The paper describes a structure of HGSNAT a member of the transmembrane acyl transferase (TmAT) superfamily. The high-resolution of a HGSNAT bound to acetyl-CoA is important for our understanding of HGSNAT mechanism. The density map is of high-quality, except for the luminal domain. The location of the acetyl-CoA allows speculation about the mechanistic role of multiple residues surrounding this molecule. The authors thoroughly describe the architecture of HGSNAT and map the mutations leading to MPS IIIC.

Reviewer #3 (Public Review):

Summary:

Navratna et al. have solved the first structure of a transmembrane N-acetyltransferase (TNAT), resolving the architecture of human heparan-alpha-glucosaminide N-acetyltransferase (HGSNAT) in the acetyl-CoA bound state using single particle cryo-electron microscopy (cryoEM). They show that the protein is a dimer, and define the architecture of the alpha- and beta-GSNAT fragments, as well as convincingly characterizing the binding site of acetyl-CoA.

Strengths:

This is the first structure of any member of the transmembrane acyl transferase superfamily, and as such it provides important insights into the architecture and acetyl-CoA binding site of this class of enzymes.

The structural data is of a high quality, with an isotropic cryoEM density map at 3.3Å facilitating building of a high-confidence atomic model. Importantly, the density for the acetyl-CoA ligand is particularly well-defined, as are the contacting residues within the transmembrane domain.

The structure of HSGNAT presented here will undoubtedly lay the groundwork for future structural and functional characterization of the reaction cycle of this class of enzymes.

Weaknesses:

While the structural data for the state presented in this work is very convincing, and clearly defines the binding site of acetyl-CoA, to get a complete picture of the enzymatic mechanism of this family, additional structures of other states will be required.

A weakness of the study is the lack of functional validation. The enzymatic activity of the enzyme characterized was not measured, and the enzyme lacks native proteolytic processing, so it is a little unclear whether the structure represents an active enzyme.

Recommendations for the authors:

Reviewer #3 (Recommendations For The Authors):

In the response to reviewers, the authors mention revised coordinates, but the revised coordinates provided to this reviewer do not reflect the stated changes (I assume a technical error somewhere)

Perhaps, the old coordinates in the deposition system were resubmitted with the revised draft. Nevertheless, we have made the changes suggested by this reviewer to structure in the previous round and have released the new coordinates (PDB ID: 8TU9).

Is there any evidence for the interprotomer disulfide except for the map? e.g. if it is a disulfide-linked dimer, one should see a shift in mobility on non-reducing vs reducing SDS-PAGE. Without this, the evidence from the map is not conclusive - while the symmetry-related cysteines are nearby to one another, based on the map I could argue that they could just as well be modeled with the cys sidechains reduced and pointing away from one another.

In addition to building the density based on cryo-EM maps, we have performed FSEC-based thermal melt analysis of the Ala mutation of C334 that is involved in disulfide at the dimer interface. C334A is still expressed as a dimer, suggesting that C334A is not the only residue stabilizing the dimer. Upon heating the detergent-solubilized protein, we noticed that the FSEC peak for C334A shows a monomeric HGSNAT (Figure 4-Figure supplement 1 in main manuscript). We hypothesize that in the absence of C334 disulfide, the extensive hydrophobic side-chain interaction network displayed in Figure 2C is responsible for maintaining the integrity of the dimer. Heating disturbs these non-disulfide interactions, thereby rendering the protein monomer. We have also performed PAGE analysis as suggested by this reviewer and noticed that reducing conditions result in a monomeric protein band (Rebuttal figure 2). While we were revising this manuscript, two other groups published structures of HGSNAT (Xu et al., 2024, Nat. Struct Mol Biol, and Zhao et al., 2024, Nat. Comm). These groups have also identified this disulfide at the dimer interface in their HGSNAT structures. Zhao et al. showed that this disulfide is not crucial for dimerization and also suggested that it can break depending on the conformation of HGSNAT. Our FSEC results agree with this observation.

Author response image 2.

Comparison of purified HGSNAT on native and reducing SDS-PAGE. The arrows on both the gels indicate N-GFP-HGSNAT. The two bands on the SDS PAGE are, perhaps, two differentially glycosylated forms of HGSNAT.

---

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

(1) The authors should characterize whether the purified protein is active. Otherwise, how does one know if the detergent used maintains the protein in a biologically relevant state? The authors should at least attempt to do so. If these prove to be challenging, at the very least, the authors should try a cell-based assay to demonstrate that the GFP tag does not interfere with the function. The authors would need to establish an in vitro assay using purified protein and assess the level of Acetyl-CoA in the reaction (there are commercial kits and a long list of literature showing how to measure this). They could also follow the HS acetylation reaction by e.g. HPLC-MS or NMR (among other methods).

The cryo-EM sample was prepared without the exogenous addition of ligand, as noted in the manuscript. However, we see that acetyl-CoA was intrinsically bound to the protein, indicating the ability of GFP-tagged HGSNAT protein to bind the ligand. Upon dialysis, we see release of acetyl-CoA from the protein, which we have confirmed by LC-MS analysis (Fig S9). We purified the protein at a pH optimal for acetyl-CoA binding, as suggested by Bame, K. J. and Rome, L. H. (1985) and Meikle, P. J. et al., (1995). Because we see acetyl-CoA in a structure obtained using a GFP fusion, we argue that GFP does not interfere with protein stability and ability to bind to the co-substrate. As demonstrated by existing literature HGSNAT catalyzed reaction is compartmentalized spatially and conditionally. The binding of acetyl-CoA happens towards the cytosol and is optimal at pH 7-0.8.0, while the transfer of the acetyl group to heparan sulfate occurs towards the luminal side and is optimal at pH 5.0-6.0. We attempted measuring HGSNAT catalyzed reaction by monitoring decrease in acetyl-CoA in presence of D-glucosamine (acetyl group acceptor) using a coupled enzyme acetyl-CoA assay kit from SIGMA (MAK039) that converts acetyl-CoA to a fluorescent product measurable at Ex/Em of 535/587 nm. We noticed a decrease in the level of acetyl-CoA in the presence of HGSNAT-ACO complex (blue) and apo HGSNAT (red); the difference compared to the ACO standard (gray) was not significant. While optimizing the assay, Xu et al. (2024, Nat Struct Mol Biol) published structural and biochemical characterization of HGSNAT, showing that detergent-purified HGSNAT is active.

Author response image 3.

Acetyl-CoA levels in absence and presence of HGSNAT purified in digitonin. Decrease in the levels of 10 mM acetyl-CoA was measured in presence of 10 mM D-glucosamine and 30 nM HGSNAT at pH 7.5.

(2) In Figure 5, the authors present a detailed schematic of the catalytic cycle, which I find to be too speculative. There is no evidence to suggest that this enzyme undergoes isomerization, similar to a transporter, between open-to-lumen and open-to-cytosol states. Could it not simply involve some movements of side chains to complete the acetyl transfer? The speculative nature of this assumption needs to be clearly acknowledged throughout the manuscript and discussed in more detail. The authors could use HDX-MS or introduce cysteine residues in the hypothetical inward- and outward-facing cavities and test accessibility by incubating the purified protein with maleimides or other agents reacting with free cysteine.

We thank the reviewers for this insightful critique. Yes, the enzyme could likely achieve catalysis by simple side chain movements without undergoing extensive isomerization steps, as depicted in Figure 5. We also agree with the reviewer that HDX-MS could be the best way to monitor the substrate-induced conformational dynamics within HGSNAT experimentally. In the absence of data supporting large movements during the acetyl transfer reaction, figure 5 is speculative. We have now edited Figure 5 in the revised version of the manuscript based on the observations we made in this study.

(3) The acetyl-CoA-bound state is described as the open-to-lumen state. Indeed, from Figure 1C, the lumen opening appears much larger than the cytosol opening. Is there any small tunnel that connects the substrate site to the cytosol? In other words, is this state accessible to both the lumen and the cytosol, albeit with a larger opening toward the lumen? This question arises because, in Figure S5, the tunnel calculated by MOLE seems to also connect to the cytosol.

Yes, it is likely that the ACOS is accessible via lumen and cytosol to varying degrees, as evidenced by MOLE prediction. However, binding of the bulky nucleoside head group of acetyl-CoA at ACOS blocks the cytosolic entrance in the confirmation discussed in this manuscript. MOLE prediction was performed on a structure devoid of acetyl-CoA, and it is possible that the protein doesn’t essentially undergo isomerization between open-to-lumen and open-to-cytosol confirmations during acetyl transfer. Likely, ACOS is always accessible from both the lumen and cytosol, but depending on the substrates or products bound, the accessibility could be limited to either the lysosomal lumen or cytosol. We have rewritten all the statements mentioning an open-to-lumen confirmation to reflect this argument.

(4) The authors state, "Interestingly, in most of the detergent conditions we tested, HGSNAT was predominantly dimeric (Fig S1C-H)," and also mention, "In all the detergents we tested, HGSNAT eluted as a dimer, a testament to the extensive side-chain interaction network." The dimerization is said to be mediated by a disulfide bond. I would be surprised if the detergents the authors tested could break a disulfide bond. Therefore, can this observation truly serve as a testament to an "extensive" side-chain interaction network?

We agree with the reviewer that detergents are unlikely to break a disulfide bond. To address this comment, we generated a C334A mutant of HGSNAT and extracted it from cells in 1% digitonin. It is still expressed as a dimer (Fig S8E). However, upon heating the detergent solubilized protein, we noticed that the FSEC peak for C334A shows a monomeric HGSNAT (Fig S8I and S8K). We hypothesize that in the absence of C334 disulfide, the extensive hydrophobic side-chain interaction network displayed in Figure 2C is responsible for maintaining the integrity of the dimer. Heating disturbs these non-disulfide interactions, thereby rendering the protein monomer.

(5) Apart from the cryo-EM structure, the article does not provide any other experimental evidence to support or explain a molecular mechanism. Due to the complete absence of functional assays, mutagenesis analysis, or other structures such as a ternary complex or an acetylated enzyme intermediate, the mechanistic model depicted in Figure 5 should be taken with caution. This uncertainty needs to be clearly described in the manuscript text. Performing additional mutagenesis experiments to test key hypotheses, or further discussing relevant data from the literature, would strengthen the manuscript.

We agree with the reviewer on the lack of supporting evidence for the mechanistic models proposed in Fig 5. They were made based on previously reported biochemical characterization of HGSNAT by Rome & Crain (1981), Rome et al. (1983), Miekle et al. (1995), and Fan et al. (2011). However, we agree with the reviewer that this schematic is not experimentally proven and is speculative at best. We have edited Figure 5 in the revised version of the manuscript. In addition, we have also performed mutagenesis analysis to study the stability of mutants (Fig S8) and performed LC-MS analysis to identify endogenously bound acetyl-CoA (Fig S9) to strengthen parts of the manuscript. We have discussed our findings in the results and modified the discussion according to these suggestions.

(6) It is discussed that H269 is an essential residue that participates in the acetylation reaction, possibly becoming acetylated during the process. However, there is no solid experimental evidence, e.g. mutagenesis analysis or structural analysis, in this or previous articles, that demonstrates this to be the case. Providing more information, ideally involving additional experimental work, would strengthen this aspect of the mechanism that is proposed. This would require establishing an in vitro assay, as described in 1).

H269, as a crucial catalytic residue, was suggested by monitoring the effect of chemical modifications of amino acids on acetylation of HGSNAT membranes by Bame, K. J. and Rome, L. H. (1986). We generated N258I and H269A mutants of HGSNAT and analyzed their stability. We noticed a greater destabilization in N258I compared to H269A (Fig S8). We believe this is because of the loss of ability to bind acetyl-CoA, as the TMs around a catalytic core of the protein in our cryo-EM structure were stabilized by interactions with acetyl-CoA. Recently, Xu et al. (2024, Nat Struct Mol Biol) suggested that they do not observe acetylated histidine in their structure. However, our structure and that reported by Xu et al. (2024) are obtained at cytosolic pH. Perhaps, acetylation of H269 occurs at acidic lysosomal pH. Extensive structural and catalytic investigation of HGSNAT at low pH is required to rule out H269 acetylation as a step in the HGSNAT catalyzed reaction.

(7) In the discussion part, the authors mention previous studies in which it was postulated that the catalytic reaction can be described by a random order mechanistic model or a Ping Pong Bi Bi model. However, the authors leave open the question of which of these mechanisms best describes the acetylation reaction. The structure presented here does not provide evidence that could support one mechanism or the other. The authors could explore if an in vitro experimental measurement of protein activity would provide any information in this regard.

We agree with the reviewer that a more detailed kinetic analysis is necessary to define the bisubstrate reaction mechanism of HGSNAT. All the existing structural data on two isoforms of HGSNAT is obtained at basic pH. As a result, the existing structures do not unambiguously demonstrate the bisusbtrate mechanism of HGSNAT. We believe low pH structural characterization and a detailed kinetic and structural characterization of HGSNAT in membrane mimetics like nanodiscs could provide more insights into the mechanism. However, these studies are a future undertaking and are not a part of this manuscript.

(8) Although the authors map the mutations leading to MPS IIIC on the structure and use FoldX software to predict the impact of these mutations on folding and fold stability, there is no experimental evidence to support FoldX's predictions. It would be ideal if an additional test for these predictions were included in the manuscript. The authors could follow the unfolding of purified mutants by SEC, FSEC, or changes in intrinsic fluorescence to assess protein stability.

As suggested here, we prepared HGSNAT MPSIIIC variants and tested their expression and stability (please see Fig S8). These results have been included in the revised version of the manuscript.

(9) Some sidechains that have quite strong sidechain density are missing atoms. I would be particularly careful with omitting sidechains that pack in the hydrophobic core, as this can tend to artificially reduce the clash score. Check F81, L62, P91 and V87, for example.

We have revisited the modeling of these regions and deposited new coordinates.

(10) W316 seems to have the wrong rotamer.

This has been corrected in the new coordinate file that has been released.

(11) N134 and N433 seem to have extra density. Are these known glycosylation sites?

As per Hrebicek M. et al., 2006 and Feldhammer M. et al., 2009, there are five predicted glycosylation sites: N66, N114, N134, N433, and N602. However, we see evidence for NAG density at N114, N134, and N433. These have now been modeled in the structure.

(12) At the C-terminal residue (Ile-635), the very C-terminal carboxylate is modeled pointing to a hydrophobic environment. It seems more likely to me that the Ile sidechain is packing here, with the C-terminal carboxylate facing the solvent.

Thank you for pointing this out. We have edited the orientation of the Ile sidechain accordingly.

Presentation and wording of results/methods:

- Figure S3 legend "At places with missing density, the side chains were trimmed to C- alpha" - this is incorrect, I think the authors mean C-beta.

We have corrected this error in the revised version of the manuscript.

- Figure S3 legend - the authors refer to a gray mesh, where a transparent surface is displayed.

Thanks for pointing this error out. We have corrected this in the revised version.

- Some colloquial/vague wording in the main text (a lot of sentences starting with "Interestingly, ...". Making the wording more specific would help the reader I think.

We have edited out ‘interestingly’ from the document and have re-written parts of the manuscript, per reviewers’ suggestion, for brevity.

- Figure S2 legend, "throughout the processing workflow the resolution of luminal domain was used as a guidepost" - it is not entirely clear to me what this means in this context, perhaps revise the wording?

We have rephrased this line in the revised draft of the manuscript.

- Figure S2 and methods, Local refinements of LD and TMD are mentioned, but not indicated on the processing workflow.

We have included a new Fig S2 & edited the legend, including these changes, per the reviewers’ suggestions.

Reviewer #1 (Public Review):

This article by Navratna et al. reports the first structure of human HGSNAT in an acetyl-CoA-bound state. Through careful structural analysis, the authors propose potential reasons why certain human mutations lead to lysosomal storage disorders and outline a catalytic mechanism. The structural data are of good quality, and the manuscript is clearly written. This study represents an important step toward understanding the mechanism of HGSNAT and is valuable to the field. I have the following suggestions:

(1) The authors should characterize whether the purified protein is active. Otherwise, how does one know if the detergent used maintains the protein in a biologically relevant state? The authors should at least attempt to do so. If these prove to be challenging, at the very least, the authors should try a cell-based assay to demonstrate that the GFP tag does not interfere with the function.

(2) In Figure 5, the authors present a detailed schematic of the catalytic cycle, which I find to be too speculative. There is no evidence to suggest that this enzyme undergoes isomerization, similar to a transporter, between open-to-lumen and open-to-cytosol states. Could it not simply involve some movements of side chains to complete the acetyl transfer?

Reviewer #2 (Public Review):

Summary:

This work describes the structure of Heparan-alpha-glucosaminide N-acetyltransferase (HGSNAT), a lysosomal membrane protein that catalyzes the acetylation reaction of the terminal alpha-D-glucosamine group required for degradation of heparan sulfate (HS). HS degradation takes place during the degradation of the extracellular matrix, a process required for restructuring tissue architecture, regulation of cellular function and differentiation. During this process, HS is degraded into monosaccharides and free sulfate in lysosomes.

HGSNAT catalyzes the transfer of the acetyl group from acetyl-CoA to the terminal non-reducing amino group of alpha-D-glucosamine. The molecular mechanism by which this process occur has not been described so far. One of the main reasons to study the mechanism of HGSNAT is that multiple mutations spanning the entire sequence of the protein, such as, nonsense mutations, splice-site variants, and missense mutations lead to dysfunction that causes abnormal accumulation of HS within the lysosomes. This accumulation is a cause of mucopolysaccharidosis IIIC (MPS IIIC), an autosomal recessive neurodegenerative lysosomal storage disorder, for which there are no approved drugs or treatment strategies.<br /> This paper provides a 3.26A structure of HGSNAT, determined by single-particle cryo-EM. The structure reveals that HGSNAT is a dimer in detergent micelles, and a density assigned to acetyl-CoA. The authors speculate about the molecular mechanism of the acetylation reaction, map the mutations known to cause MPS IIIC on the structure and speculate about the nature of the HGSNAT disfunction caused by such mutations.

Strengths:

The paper describes a structure of HGSNAT a member of the transmembrane acyl transferase (TmAT) superfamily. The high-resolution of a HGSNAT bound to acetyl-CoA is important for our understanding of HGSNAT mechanism. The density map is of high-quality, except for the luminal domain. The location of the acetyl-CoA allows speculation about the mechanistic role of multiple residues surrounding this molecule. The authors thoroughly describe the architecture of HGSNAT and map the mutations leading to MPS IIIC.

Reviewer #3 (Public Review):

Summary:

Navratna et al. have solved the first structure of a transmembrane N-acetyltransferase (TNAT), resolving the architecture of human heparan-alpha-glucosaminide N-acetyltransferase (HGSNAT) in the acetyl-CoA bound state using single particle cryo-electron microscopy (cryoEM). They show that the protein is a dimer, and define the architecture of the alpha- and beta-GSNAT fragments, as well as convincingly characterizing the binding site of acetyl-CoA.

Strengths:

This is the first structure of any member of the transmembrane acyl transferase superfamily, and as such it provides important insights into the architecture and acetyl-CoA binding site of this class of enzymes.

The structural data is of a high quality, with an isotropic cryoEM density map at 3.3Å facilitating building of a high-confidence atomic model. Importantly, the density for the acetyl-CoA ligand is particularly well-defined, as are the contacting residues within the transmembrane domain.

The structure of HSGNAT presented here will undoubtedly lay the groundwork for future structural and functional characterization of the reaction cycle of this class of enzymes.

Weaknesses:

While the structural data for the state presented in this work is very convincing, and clearly defines the binding site of acetyl-CoA, to get a complete picture of the enzymatic mechanism of this family, additional structures of other states will be required.

A weakness of the study is the lack of functional validation. The enzymatic activity of the enzyme characterized was not measured, and the enzyme lacks native proteolytic processing, so it is a little unclear whether the structure represents an active enzyme.

eLife assessment

Giamundo et al. present valuable data with new insights new insight into the role of Ezrin, a major membrane-actin linker that assembles signaling complexes, in the spatial regulation of EGF signaling mediators. The use of multiple state-of-the-art microscopy techniques, multiple cell lines and inhibitors, and in vivo models provides solid support for the majority of their conclusions. The findings are helpful for our understanding of EGF/mTOR signal transduction and support a critical role for the scaffolding protein Ezrin, in the upstream regulation of EGFR/AKT activity, TSC subcellular localization and mTORC1 signaling, thus contributing to our understanding of the regulation of endo-lysosomal signaling, alterations in which are implicated in many human diseases.

Reviewer #1 (Public Review):

Summary:<br /> The authors demonstrate that, while the loss of Ezrin increases lysosomal biogenesis and function, its presence is required for the specific endocytosis of EGFR. Upon further investigation, the authors reveal that Ezrin is a crucial intermediary protein that links EGFR to AKT, leading to the phosphorylation and inhibition of TSC. TSC is a critical negative regulator of the mTORC1 complex, which is dysregulated in various diseases, making their findings a valuable addition to multiple fields of study. Their cell signaling findings are translatable to an in vivo Medaka fish model and suggest that Ezrin may play a crucial role in retinal degeneration.

Strengths:<br /> Giamundo, Intartaglia, et al. utilized unbiased proteomic and transcriptomic screens in Ezrin KO cells to investigate the mechanistic function of Ezrin in lysosome and cell signaling pathways. The authors' findings are consistent with past literature demonstrating Ezrin's role in the EGFR and mTORC1 signaling pathways. They used several cell lines, small molecule inhibitors, and cellular and in vivo knockout models to validate signaling changes through biochemical and microscopy assays. Their use of multiple advanced microscopy techniques is also impressive.

Weaknesses:<br /> While the authors demonstrated activation of TSC1 (lysosomal accumulation) and inactivation of Akt (decreased phosphorylation in TSC1), as well as decreased mTORC1 signaling in Ezrin knockout cells, direct experiments showing the rescue of mTORC1 activity by AKT and TSC1 mutants are required to confirm the linear signaling pathway and establish Ezrin as a mediator of EGFR-AKT-TSC1-mTORC1 signaling. Although the authors presented representative images from advanced microscopy techniques to support their claims, there is insufficient quantification of these experiments. Additionally, several immunoblots in the manuscript lack vital loading controls, such as input lanes for immunoprecipitations and loading controls for western blots.

Reviewer #2 (Public Review):

Summary:<br /> The authors begin with the stated goal of gaining insight into the known repression of autophagy by Ezrin, a major membrane-actin linker that assembles signaling complexes on membranes. RNA and protein expression analysis is consistent with upregulation of lysosomal proteins in Ezrin-deficient MEFs, which the authors confirm by immunostaining and western blotting for lysosomal markers. Expression analysis also implicates EGF signaling as being altered downstream of Ezrin loss, and the authors demonstrate that Ezrin promotes relocalization of EGFR from the plasma membrane to endosomes. Ezrin loss impacts downstream MAPK/Akt/mTORC1 signaling, although the mechanistic links remain unclear. An Ezrin mutant Medaka fish line wa then generated to test Ezrin's role in retinal cells, which are known to be sensitive to changes in autophagy regulation. Phenotypes in this model appear generally consistent with observations made in cultured cells, though mild overall.

Strengths:<br /> Data on the impact of Ezrin-loss on relocalization of EGFR from the plasma membrane are extensive, and thoroughly demonstrate that Ezrin is required for EGFR internalization in response to EGF.

A new Ezrin-deficient in vivo model (Medaka fish) is generated.

Strong data demonstrates that Ezrin loss suppresses Akt signaling. Ezrin loss also clearly suppresses mTORC1 signaling in cell culture, although examination of mTORC1 activity is notably missing in Ezrin-deficient fish.

Weaknesses:<br /> LC3 is used as a readout of autophagy, however the lipidated/unlipidated LC3 ratio generally does not appear to change, thus there does not appear to be evidence that Ezrin loss is affecting autophagy in this study.

The conclusion is drawn that Ezrin loss suppresses EGF signaling, however this is complicated by a strong increase in phosphorylation of the p38 MAPK substrate MK2. Without additional characterization of MAPK and Erk signaling, the effect of Ezrin loss remains unclear.

Causative conclusions between effects on MAPK, Akt, and mTORC1 signaling are frequently drawn, but the data only demonstrate correlations. For example, many signaling pathways can activate mTORC1 including MAPK/Erk, thus reduced mTORC1 activity upon Ezrin-loss cannot currently be attributed to reduced Akt signaling. Similarly, other kinases can phosphorylate TSC2 at the sites examined here, so the conclusion cannot be drawn that Ezrin-loss causes a reduction in Akt-mediated TSC2 phosphorylation. In Figure 7, the conclusion cannot be drawn that retinal degeneration results from aberrant EGFR signaling.

It is unclear why TSC1 is highlighted in the title, as there does not appear to be any specific regulation of TSC1 here.

In Figure 1 the conclusion is drawn that there is an increase in lysosome number with Ezrin KO, however it does not appear that the current analysis can distinguish an increased number from increased lysosome size or activity. Similarly, conclusions about increased lysosome "biogenesis" could instead reflect decreased turnover.

Immunoprecipitation data for a role for Ezrin as a signaling scaffold appear minimal and seem to lack important controls.

In Figure 3A it seems difficult to conclude that EGFR dimerization is reduced since the whole blot, including the background between lanes, is lighter on that side.

In Figure 6C specificity controls for the TSC1 and TSC2 antibodies are not included, but seem necessary since their localization patterns appear very different from each other in WT cells.

In Figure 7 the signaling effects in Ezrin-deficient fish are mild compared to cultured cells, and effects on mTORC1 are not examined. Further data on the retinal cell phenotypes would strengthen the conclusions.

In Figure 7F there appears to be more EGFR throughout the cell, so it is difficult to conclude that more EGFR at the PM in Ezrin-/- fish means reduced internalization.

Reviewer #3 (Public Review):

Summary:<br /> In this study, the authors have attempted to demonstrate a critical role for the cytoskeletal scaffold protein Ezrin, in the upstream regulation of EGFR/AKT/MTOR signaling. They show that in the absence of Ezrin, ligand-induced EGFR trafficking and activation at the endosomes is perturbed, with decreased endosomal recruitment of the TSC complex, and a corresponding decrease in AKT/MTOR signaling.

Strengths:<br /> The authors have used a combination of novel imaging techniques, as well as conventional proteomic and biochemical assays to substantiate their findings. The findings expand our understanding of the upstream regulators of the EGFR/AKT MTOR signaling and lysosomal biogenesis, appear to be conserved in multiple species, and may have important implications for the pathogenesis and treatment of diseases involving endo-lysosomal function, such as diabetes and cancer, as well as neuro-degenerative diseases like macular degeneration. Furthermore, pharmacological targeting of Ezrin could potentially be utilized in diseases with defective TFEB/TFE3 functions like LSDs. While a majority of the findings appear to support the hypotheses, there are substantial gaps in the findings that could be better addressed. Since Ezrin appears to directly regulate MTOR activity, the effects of Ezrin KO on MTOR-regulated, TFEB/TFE3 -driven lysosomal function should be explored more thoroughly. Similarly, a more convincing analysis of autophagic flux should be carried out. Additionally, many immunoblots lack key controls (Control IgG in co-IPs) and many others merit repetition to either improve upon the quality of the existing data, validate the findings using orthogonal approaches, or provide a more rigorous quantitative assessment of the findings, as highlighted in the recommendation for authors.

Author response:

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

Response to reviewers (minor points):

We thank all reviewers for their very helpful suggestions and greatly appreciate their positive evaluation of our work.

Reviewer #1:

Ad 1) The reviewer states: Fig 5 While the data very nicely show that CPX and Syt1 have interdependent interactions in the chromaffin neurons, this seems to be not the case in neurons, where the loss of complexins and synaptotagmins have additive effects, suggesting independent mechanisms (eg Xue et al., 2010). This would be a good opportunity to discuss some possible differences between secretion in endocrine cells vs neurons.

We greatly appreciate the insightful suggestion by the reviewer. To accommodate the reviewer’s suggestion, we now discuss this issue on page 21, line 486-491: “In murine hippocampal neurons, loss of CpxI and Syt1 has additive effects on fast synchronous release, suggesting independent mechanisms (Xue et al., 2010). On the other hand, the same study also showed that Syt1 heterozygosity fails to reduce release probability in wild-type neurons, but does so in the absence of Cpx, again suggesting that Cpx and Syt1 may functionally interact in Ca2+-triggered release.”

Ad 2) The reviewer states: Fig 8 Shows an apparent shift in Ca sensitivity in N-terminal mutants suggesting a modification of Ca sensitivity of Syt1. Could there be also an alternative mechanism, that explains this phenotype which is based on a role of the n-term lowering the energy barrier for fusion, that in turn shifts corresponding fusion rates to take place at lower Ca saturation levels?

We fully agree with the reviewer. While our data indicate that Cpx and Syt1 act in a dependent manner in accelerating exocytosis, they do not provide decisive evidence that the NTD of CpxII directly modulates the Ca2+ affinity of Syt1, an issue that we discuss on page 23 , line 523529: ”The results favor a model wherein the CpxII NTD either directly regulates the biophysical properties of the Ca2+-sensor by increasing the apparent forward rate of Ca2+-binding or indirectly affects SytI-SNARE or SytI-membrane interactions, thereby, lowering the energy barrier of Ca2+triggered fusion.”

Reviewer #2:

Ad 1) The reviewer states: The authors provide a "chromaffin cell-centric" view of the function of mammalian Cplx in vesicle fusion. With the exception of mammalian renal ribbon synapses (and some earlier RNAi knockdown studies that had off-target effects), there is very little evidence for a "fusion-clamp"-like function of Cplxs in mammalian synapses. At conventional mammalian synapses, genetic loss of Cplx (i.e. KO) consistently decreases AP-evoked release, and generally either also decreases spontaneous release rates or does not affect spontaneous release, which is inconsistent with a "fusion-clamp" theory. This is in stark contrast to invertebrate (D. m. and C. e.) synapses where genetic Cplx loss is generally associated with strong upregulation of spontaneous release, providing support for Cplx acting as a "fusion-clamp".

We agree with the reviewer that it is difficult to reconcile contradictory findings regarding the role of Cpx in membrane fusion in vertebrates and invertebrates or between murine hippocampal neurons and neuroendocrine cells. On the other hand, we respectfully disagree with the statement of providing a "chromaffin cell-centric" view of the function of mammalian Cplx in vesicle fusion. In fact, a large number of model systems (in vitro and in vivo studies) support a scenario where complexin takes center stage in clamping of premature vesicle release. For example, in vitro analyses using a liposome fusion assay (Schaub et al., 2006, Nat Struct Mol Biol 13, 748; Schupp et al., 2016) or Hela cells that ectopically express “flipped” SNAREs on their cell surface (Giraudo et al., 2008, JBC 283, 21211) showed that complexin can inhibit the SNARE-driven fusion machinery. Likewise, several studies boosting complexin action by either genetic overexpression or peptide supplementation have provided evidence for the complexin clamp function in neuronal and nonneuronal cells (e.g. Itakura et al., 1999, BBRC 265, 691; Liu et al., 2007, Biochemistry 72, 439; Abderrahmani et al., 2004, J Cell Sci 117, 2239; Archer et al., 2002, JBC 277, 18249; Tang et al, 2006,

Cell 126, 1175; Vaithianathan et al., 2013, J Neurosci 33, 8216; Roggero et al., 2007, JBC, 282, 26335.)

In addition, chromaffin cells enable the investigation of secretion on the background of a well-defined intracellular calcium concentration. Indeed, CplxII knock-out in chromaffin cells demonstrated an enhanced tonic release which is evident at elevated levels of [Ca]i (>100nM), but absent at low resting [Ca]i (Dhara et al., 2014). Given this observation, it is tempting to speculate that variations in [Ca]i among the different preparations may contribute to the deviating expression of the complexin null phenotype in different preparations.

Ad 2) The reviewer states: The authors use a Semliki Forest virus-based approach to express mutant proteins in chromaffin cells. This strategy leads to a strong protein overexpression (~7-8 fold, Figure 3 Suppl. 1). Therefore, experimental findings under these conditions may not necessarily be identical to findings with normal protein expression levels.

As shown in Fig. 4, we use the secretion response of wt cells as a control so that we can assess the specificity and quality of the rescue approach in our experiments. In addition, the comparative analysis of the CpxII mutants was performed with respect to the equally overexpressed CpxII wt protein (Fig. 3 Suppl. 1), which we used as a control to determine the standard response under these conditions.

Ad 3) The reviewer states: Measurements of delta Cm in response to Ca2+ uncaging by ramping [Ca2+ ] from resting levels up to several µM over a me period of several seconds were used to establish changes in the release rate vs [Ca2+ ]i relationship. It is not clear to this reviewer if and how concurrently occurring vesicle endocytosis together with a possibly Ca2+-dependent kinetics of endocytosis may affect these measurements.

By infusing bovine chromaffin cells with 50µM free Ca2+, Smith and Betz have shown that the total capacitance increase is dominated by exocytosis and that significant endocytosis only sets in after 3 minutes (Smith and Betz, 1996, Nature, 380, 531). In the same line, we previously showed that mouse chromaffin cells (infused with 19µM free calcium over 2 minutes) responded with robust increase in membrane capacitance which strongly correlated with the number of simultaneously recorded amperometric events monitoring fusion of single vesicles (Dhara et al., 2014, Fig. 5B). Thus, capacitance alterations recorded under tonic intracellular Ca2+ increase in chromaffin cells are solely due to exocytosis and are not contaminated by significant endocytosis. As our Ca2+ ramp experiments were carried out for 6 seconds and the intracellular free [Ca]i did not exceed 19 µM the observed phenotypical differences between the experimental groups are most likely due to changes in exocytosis rather than endocytosis.

Ad 4) The reviewer states: It should be pointed out that an altered "apparent Ca2+ affinity" or "apparent Ca2+ binding rate" does not necessarily reflect changes at Ca2+-binding sites (e.g. Syt1).

We fully agree with the reviewer’s comment. As pointed out also in the response to reviewer 1, our experiments do not provide decisive evidence that the NTD of CpxII directly modulates the Ca2+ affinity of Syt1, an issue that we discuss on page 23 , line 523-529: ” The results favor a model wherein the CpxII NTD either directly regulates the biophysical properties of the Ca2+sensor by increasing the apparent forward rate of Ca2+-binding or indirectly affects SytI-SNARE or SytI-membrane interactions, thereby, lowering the energy barrier of Ca2+-triggered fusion.” 

AD 5) There are alternative models on how Cplx may "clamp" vesicle fusion (see Bera et al. 2022, eLife) or how Cplx may achieve its regulation of transmitter release without mechanistically "clamping" fusion (Neher 2010, Neuron). Since the data presented here cannot rule out such alternative models (in this reviewer's opinion), the authors may want to mention and briefly discuss such alternative models.

The study by Bara et al reiterates the model proposed by the Rothman group which attributes the clamping function of Cpx to its accessory alpha helix by hindering the progressive SNARE complex assembly. We have explicitly stated this issue in the original version of the manuscript (page 19, line 425) “As the accessory helix of Cpx has been found to bind to membrane proximal cytoplasmic regions of SNAP-25 and SybII (Malsam et al., 2012; Bykhovskaia et al., 2013; Vasin et al., 2016), an attractive scenario could be that both domains of CpxII, the CTD and the accessory helix, synergistically cooperate to stall final SNARE assembly”. In this context, we will now cite also the study by Bera et al.. 

A related view of the function of complexin suggested that it may act as an allosteric adaptor for sytI (Neher 2010, Neuron). Here, rather than postulang independent "clamp" and "trigger" functions for the dual action of complexin, these were explained as facets of a simple allosteric mechanism by which complexin modulates the Ca2+ dependence of release. Yet, this interpretation appears to be difficult to reconcile with the observation of our and other laboratories, showing that the fusion-promoting and clamping effects are separable (e.g. Dhara et al., 2014; Lai et al., 2014; Makke et al., 2018; Bera et al., 2022).  

Some parts of the Discussion are quite general and not specifically related to the results of the present study. The authors may want to consider shortening those parts.

Considering the contrary findings in the field of SNARE-regulating proteins, the authors hope that the reviewer will agree that it is necessary to discuss the new observations in a broader context, as also acknowledged by the first reviewer.

Last but not least, the presentation of the results could be improved to make the data more accessible to non-specialists, this concerns providing necessary background information, choice of colors, and labeling of diagrams.

Done

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors): 

Regarding figures: 

(1) Please use clearly distinct colors in diagrams. For example, in Figure 2 Suppl. 3, four different shades of red (or reddish) are used to color the traces and the respective bars. These different shades of red are difficult to discriminate. In Figure 5 Suppl. 1, the two greens are nearly indistinguishable.  

Done

(2) RRP size and SRP size on the one hand, and SR rate on the other represent different quantities which are measured in different units. Please use a separate y-axis for the SR (a rate measured in fF/s) and do not combine with RRP and SRP (pool sizes measured in fF). This would also automatically alleviate the need for axis breaks in the plots of RRP size and SRP size. In general, please do not use axis breaks which make interpretation of data unnecessarily more complicated.  

In order to clarify the display, we now define the different units together with the quantified parameter (e.g. RRP [fF], SRP [fF], SR [fF/s]) allowing us to omit a second axis in those subpanels.

(3) When plotting bar graphs showing mean tau_RRP, mean tau_SRP, and mean delay, please always use the correct y-axis labels, i.e. use "tau_RRP", "tau_SRP" and "delay" as y-axis labels as it was done for example in Figure 4D, and do not use "tau_RRP", "tau_SRP" and "delay" as x-axis labels as it was done for example in Figure 1D and many other figure panels.  

We have standardized the figure display. Yet, we would prefer to keep our way subpanel labelling which states the parameter underneath the bar graph and thereby makes the results more accessible.  

(4) Are the asterisks indicating statistical significance perhaps missing in Figure 4D, middle panel (tau_SRP)?

There was not a statistically significant difference (wt vs cpxIIko+CpxII EA, P=0.0826, Kruskal-Wallis with Dunn’ post hoc test).  

(5) According to the Results section (pages 12 to 13), I assume that in Figures 6 and 7 the labels "+Cplx XYZ" are used by the authors to identify an overexpression of Cplx XYZ in a Cplx WT background. The legend text reads however " ... cells expressing either Cplx2 wt or the mutant ...", which would not be correct. Please check.

We have changed the formulations to “overexpression” accordingly.

(6) The x-axis unit in Figure 8C is likely "µM" and not "M".

Done.

(7) The abbreviations "CplxII LL-EE" and "CplxII LL-WW", and "CplxII LLEE" and "CplxII LLWW" are very similar but refer to different mutants. Could you please think of a more specific and unambiguous abbreviation? Perhaps "CplxII L124E-L128E"?  

We have changed the abbreviations, accordingly (i.e. CpxII L124E-L128E).  

Regarding the manuscript text:  

Line 65: "prevents" instead of "impairs"? 

done

Line 67: why "in vivo"? 

We changed the formulation to ‘Several’

Line 83: "in addition to the clamping function ..." This is misleading. Many of the studies listed here did not provide evidence for enhanced spontaneous release following Cplx loss and often observed the opposite, reduced spontaneous release. The enhanced delayed release was observed by Strenzke et al 2009 J.Neurosci. and by Chang et al. 2015 J.Neurosci. (which the authors may want to cite). However, that enhanced delayed release occurred despite reduced spontaneous release indicating that it is not simply the result of a missing "fusion clamp". 

To accommodate the reviewer’s suggestion, we have changed the formulation to “Independent of the clamping function of Cpx….”

Line 104: "speeds up exocytosis that is controlled by the forward rate of Ca2+ binding" This is difficult to understand without context.  

We have now added the corresponding citations (Voets et al., 2001; Sorensen et al., 2003), which showed that exocytosis timing in chromaffin cells is largely determined by the kinetics of Ca2+-binding to SytI.

Line 116: "Cplx2 knock out ..." Please provide (here or earlier in the manuscript) information to the reader about which Cplx paralogs are expressed in chromaffin cells.  

We now state on line 111 that “CpxII is the only Cpx isoform expressed in chromaffin cells (Cai et al., 2008)”

Line 118: "=~" either "=" or "~". 

done

Line 120: "instead" seems superfluous.

done

Line 272: "calcium binding rates" should perhaps better read "apparent calcium binding rates". 

done

Line 290: "enhancing SytI's Ca2+ affinity" should perhaps better be "enhancing the apparent Ca2+ affinity of the release machinery". Ca2+ binding kinetics is never directly assayed here.

We agree and have phrased the sentence accordingly.

Line 300: "Expression of Cplx ... in Syt1 R233Q ki cells, ..." Perhaps better "Overexpression of Cplx ... in Syt1 R233Q ki/Cplx2 wt cells, ..." for clarification?

done

Lines 313ff: What is assayed here is the apparent Ca2+ binding kinetics and apparent KD values of the release machinery. Ca2+ binding to Syt1 is never directly measured!  

We agree and have changed the wording accordingly to “CpxII NTD supports the forward rate of calcium binding to SytI in accelerating exocytosis”

Line 347: "Complexin plays a dual role ..." This is partially misleading. It does so in chromaffin cells and D.m. and C.e. NMJs but not at conventional mammalian synapses. 

We agree and have changed the formulation to “In many secretory systems, Complexin plays a dual role in the regulation of SNARE-mediated vesicle fusion”

eLife assessment

This important work shows compelling data that significantly advances our understanding of the regulation of neurotransmitter and hormone secretion by exploring the mechanisms of how the protein complexin 2 (Cplx2) interacts with the calcium sensor synaptotagmin. The function of mammalian Cplx2 is studied using chromaffin cells derived from Cplx2 knock out mice as a system to overexpress and functionally characterize mutant Cplx2 forms and the interaction between Cplx2 and synaptotagmin. The authors identify structural requirements within the protein for Cplx's dual role in preventing premature vesicle exocytosis and enhancing evoked exocytosis. The findings are of broad interest to neuroscientists and cell biologists.

Reviewer #1 (Public Review):

Summary:

Using chromaffin cells as a powerful model system for studying secretion, the authors study the regulatory role of complexin in secretion. Complexin is still enigmatic in its regulatory role, as it both provides inhibitory and facilitatory functions in release. The authors perform an extensive structure-function analysis of both the C- and N-terminal regions of complexin. There are several interesting findings that significantly advances our understanding of cpx/SNARe interactions in regulating release. C-terminal amphipathic helix interferes with SNARE complex assembly and thus clamps fusion. There are acidic residues in the C-term that may be seen as putative interaction partners for Synaptotagmin. The N-terminus of Complexin promoting role may be associated with an interaction with Syt1. In particular the putative interaction with Syt1 is of high interest and supported by quite strong functional and biochemical evidence. The experimental approaches are state of the art, and the results are of the highest quality and convincing throughout. They are adequate and intelligently discussed in the rich context of the standing literature. Whilst there are some concerns about whether the facilitatory actions of complexion have to be tightly linked to Syt1 interactions, the proposed model will significantly advance the field by providing new directions in future research.

Reviewer #2 (Public Review):

Summary:

Complexin (Cplx) is expressed at nearly all chemical synapses. Mammalian Cplx comes in four different paralogs which are differentially expressed in different neurons or secretory cell types, either selectively or in combination with one or two other Cplx isoforms. Cplx binds with high affinity to assembled SNARE complexes and promotes evoked synchronous release. Cplx is assumed to preclude premature SV fusion by preventing full SNARE assembly, thereby arresting subsequent SNARE-driven fusion ("fusion-clamp" theory). The protein has multiple domains, the functions of which are controversially discussed. Cplx's function has been studied in a variety of model organisms including mouse, fly, worm, and fish with seemingly conflicting results which led to partly contradicting conclusions.<br /> Makee et al. study the function of mammalian Cplx2 in chromaffin cells by making use of Cplx2 ko mice to overexpress and functionally characterize mutant Cplx2 forms in cultured chromaffin cells. The main conclusion of the present study are:

The hydrophobic character of the amphipathic helix in Cplx's C-terminal domain is essential for inhibiting premature vesicle fusion at a [Ca2+]i of several hundreds of nM (pre-flash [Ca2+]i). The Cplx-mediated inhibition of fusion under these conditions does not rely on expression of either Syt1 or Syt7.

Slow-down of exocytosis by N-terminally truncated Cplx mutants in response to a [Ca2+]i of several µM (peak flash [Ca2+]i) occurs regardless of the presence or absence of Syt7 demonstrating that Cplx2 does not act as a switch favoring preferential assembly of the release machinery with Syt1,2 rather than the "slow" sensor Syt7.

Cplx's N-terminal domain is required for the Cplx2-mediated increase in the speed of exocytosis and faster onset of exocytosis which likely reflect an increased apparent Ca2+ sensitivity and faster Ca2+ binding of the release machinery.

Strengths:

The authors perform systematic truncation/mutational analyses of Cplx2. They analyze the impact of single and combined deficiencies for Cplx2 and Syt1 to establish interactions of both proteins.<br /> State-of-the-art methods are employed: Vesicle exocytosis is assayed directly and with high resolution using capacitance measurements. Intracellular [Ca2+] is controlled by loading via the patch-pipette and by UV-light induced flash-photolysis of caged [Ca2+]. The achieved [Ca2+ ] is measured with Ca2+ -sensitive dyes.<br /> The data is of high quality and the results are compelling.

Weaknesses:

With the exception of mammalian retinal ribbon synapses (and some earlier RNAi knock down studies which had off-target effects), there is little experimental evidence for a "fusion-clamp"-like function of Cplxs at mammalian synapses. At conventional mammalian synapses, genetic loss of Cplx (i.e. KO) consistently decreases AP-evoked release, and generally either also decreases spontaneous release rates or does not affect spontaneous release, which is inconsistent with a "fusion-clamp" theory. This is in stark contrast to invertebrate (D. m. and C. e.) synapses where genetic Cplx loss is generally associated with a strong upregulation of spontaneous release.

There are alternative scenarios explaining how Cplx may phenomenological "clamp" vesicle fusion rates without mechanistically assigning a "clamping" function to Cplx (Neher 2010, Neuron). In fact, changes in asynchronous release kinetics following conditioning AP trains observed at Cplx1 ko calyx of Held synapses do not favor a "fusion clamp" model (Chang et al., 2015, J.Neurosci.), while an alternative model, assigning Cplx the role of a "checkpoint" protein in SNARE assembly, quantitatively reproduces all experimental observations (Lopez et al., 2024, PNAS). It might be helpful for a reader to mention such alternative scenarios.

Reviewer #2 (Public Review):

Summary:

This paper utilizes a neural network model to investigate how the brain employs an adaptive chunking strategy to effectively enhance working memory capacity, which is a classical and significant question in cognitive neuroscience. By integrating perspectives from both the 'slot model' and 'limited resource models,' the authors adopted a neural network model encompassing the prefrontal cortex and basal ganglia, introduced an adaptive chunking strategy, and proposed a novel hybrid model. The study demonstrates that the brain can adaptively bind various visual stimuli into a single chunk based on the similarity of color features (a continuous variable) among items in visual working memory, thereby improving working memory efficiency. Additionally, it suggests that the limited capacity of working memory arises from the computational characteristics of the neural system, rather than anatomical constraints.

Strengths:

The neural network model utilized in this paper effectively integrates perspectives from both slot models and resource models (i.e., resource-like constraints within a slot-like system). This methodological innovation provides a better explanation for the limited capacity of working memory. By simulating the neural networks of the prefrontal cortex and basal ganglia, the model demonstrates how to optimize working memory storage and retrieval strategies through reinforcement learning (i.e., the efficient management of access to and from working memory). This biologically plausible simulation offers a novel perspective on human working memory and potentially provides a novel explanation for the working memory difficulties observed in patients with Parkinson's disease and other disorders. Furthermore, the effectiveness of the model is validated through computational simulation experiments, yielding reliable and robust predictions.

Weaknesses:

The model employs a spiking neural network, which is relatively complex. Additionally, while this paper validates the effectiveness of chunking strategies used by the brain to enhance working memory efficiency through computational simulations, further comparison with related phenomena observed in cognitive neuroscience experiments on limited working memory capacity, such as the recency effect, is necessary to verify its generalizability.

eLife assessment

This important work proposes a neural network model of interactions between the prefrontal cortex and basal ganglia to implement adaptive resource allocation in working memory, where the gating strategies for storage are adjusted by reinforcement learning. Numerical simulations provide convincing evidence for the superiority of the model in improving effective capacity, optimizing resource management, and reducing error rates, as well as solid evidence for its human-like performance. The paper could be strengthened further by a more thorough comparison of model predictions with human behavior and by improved clarity in presentation. This work will be of broad interest to computational and cognitive neuroscientists, and may also interest machine-learning researchers who seek to develop brain-inspired machine-learning algorithms for memory.

Reviewer #1 (Public Review):

Summary:

In this research, Soni and Frank investigate the network mechanisms underlying capacity limitations in working memory from a new perspective, with a focus on visual working memory (VWM). The authors have advanced beyond the classical neural network model, which incorporates the prefrontal cortex and basal ganglia (PBWM), by introducing an adaptive chunking variant. This model is trained using a biologically plausible, dopaminergic reinforcement learning framework. The adaptive chunking mechanism is particularly well-suited to the VWM tasks involving continuous stimuli and elegantly integrates the 'slot' and 'resource' theories of working memory constraints. The chunk-augmented PBWM operates as a slot-like system with resource-like limitations.

Through numerical simulations under various conditions, Soni and Frank demonstrate the performance of the chunk-augmented PBWM model surpasses the no-chunk control model. The improvements are evident in enhanced effective capacity, optimized resource management, and reduced error rates. The retention of these benefits, even with increased capacity allocation, suggests that working memory limitations are due to a combination of factors, including the efficient credit assignments that are learned flexibly through reinforcement learning. In essence, this work addresses fundamental questions related to a computational working memory limitation using a biologically-inspired neural network, and thus has implications for clinical conditions in which working memory is affected, such as Parkinson's disease, ADHD, and schizophrenia.

Strengths:

The integration of mechanistic flexibility, reconciling two theories for WM capacity into a single unified model, results in a neural network that is both more adaptive and human-like. Building on the PBWM framework ensures the robustness of the findings. The addition of the chunking mechanism tailors the original model for continuous visual stimuli. Chunk-stripe mechanisms contribute to the 'resource' aspect, while input-stripes contribute to the 'slot' aspect. This combined network architecture enables flexible and diverse computational functions, enhancing performance beyond that of the classical model.

Moreover, unlike previous studies that design networks for specific task demands, the proposed network model can dynamically adapt to varying task demands by optimizing the chunking gating policy through RL.

The implementation of a dopaminergic reinforcement learning protocol, as opposed to a hard-wired design, leads to the emergence of strategic gating mechanisms that enhance the network's computational flexibility and adaptability. These gating strategies are vital for VWM tasks and are developed in a manner consistent with ecological and evolutionary learning held by humans. Further examination of how reward prediction error signals, both positive and negative, collaborate to refine gating strategies reveals the crucial role of reward feedback in fine-tuning the working memory computations and the model's behavior, aligning with the current neuroscientific understanding that reward matters.

Furthermore, assessing the impact of a healthy balance of dopaminergic reward prediction error signals on information manipulation holds implications for patients with altered striatal dopaminergic signaling.

Weaknesses:

While I appreciate the novelty of the idea presented in this paper, which aligns with common interests in cognitive neuroscience, I believe there are several areas that require further clarification.

First, the method section appears somewhat challenging to follow. To enhance clarity, it might be beneficial to include a figure illustrating the overall model architecture. This visual aid could provide readers with a clearer understanding of the overall network model.

Additionally, the structure depicted in Figure 2 could be potentially confusing. Notably, the absence of an arrow pointing from the thalamus to the PFC and the apparent presence of two separate pathways, one from sensory input to the PFC and another from sensory input to the BG and then to the thalamus, may lead to confusion. While I recognize that Figure 2 aims to explain network gating, there is room for improvement in presenting the content accurately.

Still, for the method part, it would enhance clarity to explicitly differentiate between predesigned (fixed) components and trainable components. Specifically, does the supplementary material state that synaptic connection weights in striatal units (Go&NoGo) are trained using XCAL, while other components, such as those in the PFC and lateral inhibition, are not trained (I found some sentences in 'Limitations and Future Directions')?

I'm not sure about the training process shown in Figure 8. It appears that the training may not have been completed, given that the blue line representing the chunk stripe is still ascending at the endpoint. The weights depicted in panel d) seem to correspond with those shown in panels b) and c), no? Then, how is the optimization process determined to be finished? Alternatively, could it be stated that these weight differences approach a certain value asymptotically? It would be better to clarify the convergence criteria of the optimization process.

Author response:

We thank the reviewers for their constructive comments that will help us clarify and strengthen the paper. We will be happy to address all the comments and adjust the text accordingly. Regarding the suggestion in the assessment to include a “more thorough comparison with with human behavior”, we believe this comment reflects one of the reviewer’s comments to compare with order effects (primacy and recency); we did not see any other comments that would reflect this (our existing simulations do make contact with other human behavior regarding error distributions, including probability of recall, precision, sensitivity to reinforcement history, and dopamine manipulation effects on human WM). We thank the reviewers for this comment and we will conduct the appropriate simulations and analysis to compare with sequential effects in working memory.

eLife assessment

Shore et al. report important effects of a heterozygous mutation in the KCNT1 potassium channel on ion currents and firing behavior of excitatory and inhibitory neurons in the cortex of KCNT1-Y777H mice. The authors provide solid evidence of physiological differences between this heterozygous mutation and their previous work with homozygotes. The reviewers appreciated the inclusion of recordings in ex vivo slices and dissociated cortical neurons, as well as the additional evidence showing an increase in persistent sodium currents (INaP) in parvalbumin-positive interneurons in heterozygotes. However, they were unclear regarding the likelihood of the increased sodium influx through INaP channels increasing sodium-activated potassium currents in these neurons.

Reviewer #1 (Public Review):

Summary:

This manuscript reports the effects of a heterozygous mutation in the KCNT1 potassium channels on the properties of ion currents and firing behavior of excitatory and inhibitory neurons in the cortex of mice expressing KCNT1-Y777H. In humans, this mutation as well as multiple other heterozygotic mutations produce very severe early-onset seizures and produce a major disruption of all intellectual function. In contrast, in mice, this heterozygous mutation appears to have no behavioral phenotype or any increased propensity to seizures. A relevant phenotype is, however, evident in mice with the homozygous mutation, and the authors have previously published the results of similar experiments with the homozygotes. As perhaps expected, the neuronal effects of the heterozygous mutation presented in this manuscript are generally similar but markedly smaller than the previously published findings on homozygotes. There are, however, some interesting differences, particularly on PV+ interneurons, which appear to be more excitable than wild type in the heterozygotes but more excitable in the heterozygotes. This raises the interesting question, which has been explicitly discussed by the authors in the revised manuscript, as to whether the reported changes represent homeostatic events that suppress the seizure phenotype in the mouse heterozygotes or simply changes in excitability that do not reach the threshold for behavioral outcomes.

Strengths and Weaknesses:

(1) The authors find that the heterozygous mutation in PV+ interneurons increases their excitability, a result that is opposite from their previous observation in neurons with the corresponding homozygous mutation. They propose that this results from the selective upregulation of a persistent sodium current INaP in the PV+ interneurons. These observations are very interesting ones, and they raised some issues in the original submission:

A) The protocol for measuring the INaP current could potentially lead to results that could be (mis)interpreted in different ways in different cells. First, neither K currents nor Ca currents are blocked in these experiments. Instead, TTX is applied to the cells relatively rapidly (within 1 second) and the ramp protocol is applied immediately thereafter. It is stated that, at this time, Na currents and INaP are fully blocked but that any effects on Na-activated K currents are minimal. In theory this would allow the pre- to post- difference current to represent a relatively uncontaminated INaP. This would, however, only work if activation of KNa currents following Na entry is very slow, taking many seconds. A good deal of literature has suggested that the kinetics of activation of KNa currents by Na influx vary substantially between cell types, such that single action potentials and single excitatory synaptic events rapidly evoke KNa currents in some cell types. This is, of course, much faster than the time of TTX application. Most importantly, the kinetics of KNa activation may be different in different neuronal types, which would lead to errors that could produce different estimates of INaP in PV+ interneurons vs other cell types.

In their revised manuscript, the authors have provided good data demonstrating that, at least for the PV and SST neurons, loss of KNa currents after TTX application is slow relative to the time course of loss of INaP, justifying the use of this protocol for these neuronal types.

B) As the authors recognize, INaP current provides a major source of cytoplasmic sodium ions for the activation. An expected outcome of increased INaP is, therefore, further activation of KNa currents, rather than a compensatory increase in an inward current that counteracts the increase in KNa currents, as is suggested in the discussion.

The authors comment in the rebuttal that, despite the fact that sodium entry through INaP is known to activate KNa channels, an increase in INaP does not necessarily imply increased KNa current. This issue should be addressed directly somewhere in the text, perhaps most appropriately in the discussion.

C) The numerical simulations, in general, provide a very useful way to evaluate the significance of experimental findings. Nevertheless, while the in-silico modeling suggests that increases in INaP can increase firing rate in models of PV+ neurons, there is as yet insufficient information on the relative locations of the INaP channels and the kinetics of sodium transfer to KNa channels to evaluate the validity of this specific model.

The authors have now put in all of the appropriate caveats on this very nicely in the revised manuscript.

(2) The effects of the KCNT1 channel blocker VU170 on potassium currents are somewhat larger and different from those of TTX, suggesting that additional sources of sodium may contribute to activating KCNT1, as suggested by the authors. Because VU170 is, however, a novel pharmacological agent, it may be appropriate to make more careful statements on this. While the original published description of this compound reported no effect on a variety of other channels, there are many that were not tested, including Na and cation channels that are known to activate KCNT1, raising the possibility of off-target effects.

In the revised version, the authors have added more to the manuscript on this issue and have added a very clear discussion of this to the text (in the discussion section).

This is a very clear and thorough piece of work, and the authors are to be congratulated on this. My one remaining suggestion would be to make an explicit statement about whether increased sodium influx through INaP channels, which is thought to activate KNa channels, would be likely to increase KNa current in these neurons (see comment 1B).

Reviewer #2 (Public Review):

Summary:

In this manuscript, Shore et al. investigate the consequent changes in excitability and synaptic efficacy of diverse neuronal populations in an animal model of juvenile epilepsy. Using electrophysiological patch-clamp recordings from dissociated neuronal cultures, the authors find diverging changes in two major populations of inhibitory cell types, namely somatostatin (SST)- and parvalbumin (PV)-positive interneurons, in mice expressing a variant of the KCNT1 potassium channel. They further suggest that the differential effects are due to a compensatory increase in the persistent sodium current in PV interneurons in pharmacological and in silico experiments. It remains unclear why this current is selectively enhanced in PV-interneurons.

Strengths:

(1) Heterozygous KCNT1 gain of function variant was used which more accurately models the human disorder.

(2) The manuscript is clearly written, and the flow is easy to follow. The authors explicitly state the similarities and differences between the current findings and the previously published results in the homozygous KCNT1 gain of function variant.

(3) This study uses a variety of approaches including patch clamp recording, in silico modeling and pharmacology that together make the claims stronger.

(4) Pharmacological experiments are fraught with off-target effects and thus it bolsters the authors' claims when multiple channel blockers (TTX and VU170) are used to reconstruct the sodium-activated potassium current.

Weaknesses:

(1) This study mostly relies on recordings in dissociated cortical neurons. Although specific WT interneurons showed intrinsic membrane properties like those reported for acute brain slices, it is unclear whether the same will be true for those cells expressing KCNT1 variants, especially when the excitability changes are thought to arise from homeostatic compensatory mechanisms. The authors do confirm that mutant SST-interneurons are hypoexcitable using an ex vivo slice preparation which is consistent with work for other KCTN1 gain of function variants (e.g. Gertler et al., 2022). However, the key missing evidence is the excitability state of mutant PV-interneurons, given the discrepant result of reduced excitability of PV cells reported by Gertler et al in acute hippocampal slices.

Reviewer #3 (Public Review):

Summary:

The present manuscript by Shore et al. entitled Reduced GABAergic Neuron Excitability, Altered Synaptic Connectivity, and Seizures in a KCNT1 Gain-of-Function Mouse Model of Childhood Epilepsy" describes in vitro and in silico results obtained in cortical neurons from mice carrying the KCNT1-Y777H gain-of-function (GOF) variant in the KCNT1 gene encoding for a subunit of the Na+-activated K+ (KNa) channel. This variant corresponds to the human Y796H variant found in a family with Autosomal Dominant Nocturnal Frontal lobe epilepsy. The occurrence of GOF variants in potassium channel encoding genes is well known, and among potential pathophysiological mechanisms, impaired inhibition has been documented as responsible for KCNT1-related DEEs. Therefore, building on a previous study by the same group performed in homozygous KI animals, and considering that the largest majority of pathogenic KCNT1 variants in humans occur in heterozygosis, the Authors have investigated the effects of heterozygous Kcnt1-Y777H expression on KNa currents and neuronal physiology among cortical glutamatergic and the 3 main classes of GABAergic neurons, namely those expressing vasoactive intestinal polypeptide (VIP), somatostatin (SST), and parvalbumin (PV), crossing KCNT1-Y777H mice with PV-, SST- and PV-cre mouse lines, and recording from GABAergic neurons identified by their expression of mCherry (but negative for GFP used to mark excitatory neurons).

The results obtained revealed heterogeneous effects of the variant on KNa and action potential firing rates in distinct neuronal subpopulations, ranging from no change (glutamatergic and VIP GABAergic) to decreased excitability (SST GABAergic) to increased excitability (PV GABAergic). In particular, modelling and in vitro data revealed that an increase in persistent Na current occurring in PV neurons was sufficient to overcome the effects of KCNT1 GOF and cause an overall increase in AP generation.

Strengths:

The paper is very well written, the results clearly presented and interpreted, and the discussion focuses on the most relevant points.<br /> The recordings performed in distinct neuronal subpopulations (both in primary neuronal cultures and, for some subpopulations, in cortical slices, are a clear strength of the paper. The finding that the same variant can cause opposite effects and trigger specific homeostatic mechanisms in distinct neuronal populations is very relevant for the field, as it narrows the existing gap between experimental models and clinical evidence.

Weaknesses:

My main concern regarding the epileptic phenotype of the heterozygous mice investigated has been clarified in the revision, where the infrequent occurrence of seizures is more clearly stated. Also, a more detailed statistical analysis of the modeled neurons has been added in the revision.

Author response:

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

Introduction to the revised manuscript:

We thank all three reviewers for their time and insightful comments on our original submission. We are submitting a substantially revised manuscript that includes several new experiments, analyses, discussion points, and clarifications that we believe address all of the main concerns of the reviewers.

To address the request of Reviewers 2 and 3 to reinforce key findings in a more physiologically intact preparation, we performed recordings of YH-HET SST neurons in brain slices and found that these neurons show impairments in AP generation similar to those observed in YH-HET SST cultured neurons. These data are summarized in a new figure (Fig. 9). Along these lines, we performed additional recordings in cultured neurons at room temperature compared with physiological temperature and found that WT and YH-HET PV neuronal properties were similarly altered by temperature increases, suggesting that our YH variant-induced neuronal phenotypes are not temperature dependent. These data are shown in a new supplemental figure (Supplemental Fig. 4-3). To address concerns of Reviewer 1 regarding our KNa and NaP current recordings, we performed new experiments to further assess the specificity of the VU170 blocker in KNa KO neurons (summarized in Supplemental Fig. 5-2) and to better characterize the time course over which TTX blocks the persistent Na+ current and the KNa current (summarized in Supplemental Fig. 7-1). These latter two experiments provide further clarity and confidence in the accuracy of our measurements of both KNa and NaP currents. Lastly, to address the concern of Reviewer 3 regarding statistical analyses of the modeling data, we’ve added a new table with the results of a repeated measures ANOVA analysis (Supplemental Table 6), and two new figures illustrating the relative changes in each neuron group compared to their controls (Supplemental Figures 6-2 and 7-2). 

In addition to the new experiments and analyses, we’ve added three new paragraphs to the Discussion section. As the hyperexcitability phenotype in YH-HET PV neurons is somewhat unexpected, we’ve added a paragraph comparing our findings with those found in PV neurons in another KCNT1 GOF model. We’ve also added a paragraph to speculate on the contribution of YH-HET variant-induced alterations in SST and PV neurons to network behavior and seizure propensity. Lastly, we’ve added a paragraph to include the additional limitations and caveats of our study requested by the reviewers.  

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This manuscript reports the effects of a heterozygous mutation in the KCNT1 potassium channels on the properties of ion currents and the firing behavior of excitatory and inhibitory neurons in the cortex of mice expressing KCNT1-Y777H. In humans, this mutation as well as multiple other heterozygotic mutations produce very severe early-onset seizures and produce a major disruption of all intellectual function. In contrast, in mice, this heterozygous mutation appears to have no behavioral phenotype or any increased propensity to seizures.

Regarding the last sentence above, we wanted to clarify a point that we neglected to emphasize in the initial submission. In the Results section from our previous paper (Shore et al., 2020), we failed to observe seizures in 14 heterozygous mice, whereas 23/25 homozygous mice showed seizures by video-EEG. However, in the fifth paragraph of the Discussion section from that paper, we further stated that “during the preparation and review of [that] article, we observed seizures in two Kcnt1-Y777H heterozygous mice, one during a widefield Ca2+ imaging experiment and the other during a video-EEG experiment”. Thus, we concluded that “heterozygous expression can result in seizures in a rodent model, but apparently at a much lower frequency than that observed with homozygous expression”. To emphasize these findings, we’ve added a sentence to the Introduction in this manuscript about the occurrence of infrequent seizures in Kcnt1-Y777H heterozygous mice, along with a reference to the Discussion of our previous paper.

A relevant phenotype is, however, evident in mice with the homozygous mutation, and the authors have previously published the results of similar experiments with the homozygotes. As perhaps expected, the neuronal effects of the heterozygous mutation presented in this manuscript are generally similar but markedly smaller than the previously published findings on homozygotes. There are, however, some interesting differences, particularly on PV+ interneurons, which appear to be more excitable than wild type in the heterozygotes but more excitable in the heterozygotes. This raises the interesting question (which could be more explicitly discussed by the authors) as to whether the reported changes represent homeostatic events that suppress the seizure phenotype in the mouse heterozygotes or simply changes in excitability that do not reach the threshold for behavioral outcomes.

That is an interesting question. We have added a new paragraph to the Discussion speculating about whether the alterations in SST and PV excitability suppress seizures or do not reach the threshold for behavioral outcomes. This seems to be requested by the second reviewer as well in Weaknesses point #2.

Strengths and Weaknesses:

(1) The authors find that the heterozygous mutation in PV+ interneurons increases their excitability, a result that is opposite from their previous observation in neurons with the corresponding homozygous mutation.

We would like to provide a minor clarification to the above statement that, in this manuscript, we show that “the heterozygous mutation in PV+ interneurons increases their excitability, a result that is opposite from their previous observation in neurons with the corresponding homozygous mutation”. In our previous manuscript, we assessed YH-HOM phenotypes in NFS and FS GABAergic neurons, but did not specifically mark PV neurons. Although the YH-HOM FS neurons showed an increase in rheobase and a decrease in AP firing, the magnitudes of these effects were far less than those observed in the NFS population. More importantly, the FS GABAergic population likely consists of PV- and SST-expressing neurons; thus, we can not directly compare the results from the NFS and FS groups to the PV and SST groups, respectively (please see our response to Weaknesses point #3, Reviewer #2). We apologize for the confusion.

They propose that this results from the selective upregulation of a persistent sodium current INaP in the PV+ interneurons. While the observations are very interesting, there are three issues concerning this interpretation that should be addressed:

A) The protocol for measuring the INaP current could potentially lead to results that could be (mis)interpreted in different ways in different cells. First, neither K currents nor Ca currents are blocked in these experiments. Instead, TTX is applied to the cells relatively rapidly (within 1 second) and the ramp protocol is applied immediately thereafter. It is stated that, at this time, Na currents and INaP are fully blocked but that any effects on Na-activated K currents are minimal. In theory, this would allow the pre- to post-difference current to represent a relatively uncontaminated INaP. This would, however, only work if activation of KNa currents following Na entry is very slow, taking many seconds. A good deal of literature has suggested that the kinetics of activation of KNa currents by Na influx vary substantially between cell types, such that single action potentials and single excitatory synaptic events rapidly evoke KNa currents in some cell types. This is, of course, much faster than the time of TTX application. Most importantly, the kinetics of KNa activation may be different in different neuronal types, which would lead to errors that could produce different estimates of INaP in PV+ interneurons vs other cell types.

First, we’d like to point out that we did not want to block K+ currents (which would also block KNa) when measuring INaP for these experiments, because our hypothesis was that the increased KNa current in YH-HET PV neurons was somehow causing an increase in INaP, and it is possible that this increase depends on an intact KNa. Thus, we decided to use a method based on the observation in our experiments, and previously made by others (Budelli et al., 2009), that the reduction of outward current after TTX addition is slow relative to the rapid reduction in Na+ current. We understand and agree with the reviewer that, if KNa currents were blocked more quickly by TTX in some neuron types than others, then our estimate of INaP using this method would be contaminated in these neuron types, which would lead to inaccurate measurements. To assess this possibility among the main neuron types used in this study, we performed new experiments in which we monitored the time course of INaP block and subsequent IKNa loss following TTX application in PV and SST neurons during slow voltage ramps. We note that action potentials are not present in the slow voltage ramps due to inactivation of the transient Na+ current. These new experiments show that, in SST and PV (both WT and Het) neurons, the block of INaP is nearly complete at the 6s time point, whereas the decay in IKNa is far slower (V50 of ≈ 25s), and importantly, these results do not differ substantially by cell type or genotype. These data suggest that our measurements of INaP are not significantly contaminated by IKNa, and that this method allows for the effective separation of these two currents. These data have been added as a supplemental figure (Supplemental Fig. 7-1) and are briefly described and referenced in the Results section.

B) As the authors recognize, INaP current provides a major source of cytoplasmic sodium ions for the activation. An expected outcome of increased INaP is, therefore, further activation of KNa currents, rather than a compensatory increase in an inward current that counteracts the increase in KNa currents, as is suggested in the discussion.

We agree that the increase in INaP could theoretically further increase IKNa, as veratridine was previously shown to increase IKNa (Hage & Salkoff, 2012). However, we do not believe that this would necessarily be the case, because as the reviewer notes in their next comment, there is insufficient information on the relative locations of the INaP and KCNT1 channels, as well as the kinetics of sodium transfer to KCNT1 channels, and even less is known in the context of KCNT GOF neurons. Thus, there are a couple of plausible reasons that increased INaP may not alter KNa currents in YH-HET PV neurons: (1) In YH-HET PV neurons, the particular sodium channels that are responsible for the increased INaP may not be located within close proximity to the KCNT1 channels. (2) Homeostatic mechanisms that alter the AIS length, or move the AIS further from the soma, in response to altered neuronal excitability are well described (Grubb & Burrone, 2010; Kuba et al., 2010); thus, it is possible that in YH-HET PV neurons, the length or location of the AIS is altered, leading to uncoupling of the sodium channels that are responsible for the increased INaP to the KCNT1 channels.

C) Numerical simulations, in general, provide a very useful way to evaluate the significance of experimental findings. Nevertheless, while the in-silico modeling suggests that increases in INaP can increase firing rate in models of PV+ neurons, there is as yet insufficient information on the relative locations of the INaP channels and the kinetics of sodium transfer to KNa channels to evaluate the validity of this specific model.

We completely agree; thus, we have described each of these limitations in the Discussion. We state that the model neurons may “lack more detailed features of ion channels, such as post-translational modifications and subcellular localizations”, and that our KCNT1 model conductance is “hampered by an incomplete understanding of the relationship between Na+ influx, membrane voltage, and channel gating in neurons”.  

(2) The greatest effect of TTX application would be expected to be the elimination of large transient inward sodium currents. Why are no such currents visible in the control (pre-TTX) or the difference currents (Fig. 2)? Is it possible I missed something in the methods?

We apologize for the confusion and our mistake in failing to mention this important feature of the displayed traces. To include all of the representative traces in the figures, and prevent overlap of the traces, we removed the large inward sodium currents using the masking tool in Adobe Illustrator in Figure 2 and Supplemental Figure 5-1. We have added that information to the relevant figure legends. We have also provided unmasked images of the representative traces from Figure 2 and Supplemental Figure 5-1 to illustrate the large transient inward sodium currents, and the significant reduction of these currents with TTX treatment.

(3) As expected, the changes in many of the measured parameters are smaller in the present study with heterozygotes than those previously reported for the homozygous mutation. Some of the statements on the significance of some of the present findings need to be stated more clearly. For example, in the results section describing Fig. 2, it is stated that "In glutamatergic and NFS GABAergic YH-HET neurons, the overall KNa current was increased ...as measured by a significant effect of genotype ...." Later in the same paragraph it is stated that the increases in KNa current are not significant. Apparently, different tests lead to different conclusions. Both for the purpose of understanding the pathophysiological effects of changes in KNa current and for making further numerical simulations, more explicit clarifying statements should be made.

We apologize for the confusion on the description of these statistics. The results come from the same test, which is a Generalized Linear Mixed Model (GLMM). The factors in our GLMM were voltage step, genotype, and a voltage step x genotype interaction term. The overall effect of genotype is significant in glutamatergic neurons, but pairwise tests at each voltage step show no significant effect of genotype at any given voltage. This is somewhat analogous to running a traditional ANOVA on multiple groups and finding a significant ANOVA p-value but no significant post-hoc multiple comparisons tests, and is not uncommon. Our interpretation of this is that heterozygous expression of the YH variant in glutamatergic neurons likely increases KNa currents across positive potentials (as was seen with the YH-HOM glutamatergic neurons), but only a small amount at each positive step; thus, we lack the statistical power to determine any particular voltage step where this occurs.

(4) The effects of the KCNT1 channel blocker VU170 on potassium currents are somewhat larger and different from those of TTX, suggesting that additional sources of sodium may contribute to activating KCNT1, as suggested by the authors. Because VU170 is, however, a novel pharmacological agent, it may be appropriate to make more careful statements on this. While the original published description of this compound reported no effect on a variety of other channels, there are many that were not tested, including Na and cation channels that are known to activate KCNT1, raising the possibility of off-target effects.

We agree and thank the reviewer for making this point. To address this question, we measured KNa currents in WT vs. Kcnt1/Kcnt2-dKO neurons using VU170 to illustrate the extent of outward current due to off-target effects of the drug. These data have been included as a supplemental figure (Supplemental Fig. 5-2). We have also added several sentences to the Results section referencing this figure. Interestingly, in Kcnt1/Kcnt2-dKO neurons, VU170 seems to be quite specific across the negative potentials, as no outward currents are apparent until approximately -10 mV onward, whereas across positive potentials, there is a VU170-senstive outward current reaching ~1 nA by +50 mV. We have also included a note of caution in interpreting these data and added the possibility of off-target effects of VU170 as an alternative explanation for the differences observed on KNa currents between TTX and VU170 to the Discussion section.

(5) The experiments were carried out at room temperature. Is it possible that different effects on firing patterns in heterozygotes and homozygotes would be observed at more physiological temperatures?

Yes, it is reasonable to assume that an increased temperature would affect neuronal firing patterns in cultured neurons, as temperature differences have been shown to alter synaptic transmission and neuronal function, as assessed in both cultured neuron and slice recordings. All of our recordings were performed at room temperature in this study, and although they are valid with regard to between-group comparisons, this additional caveat is worth mentioning. We have added this to the paragraph describing study limitations in the Discussion section.

To better understand the effects of temperature in our recordings, we have now compared membrane and AP generation parameters at room temperature (~22°C) and at a more physiological temperature (35°C) in a before-after study of 16 WT neurons, including both glutamatergic and GABAergic neuron types. Not surprisingly, we found robust alterations in all parameters assessed, excluding resting membrane potential and capacitance. We further assessed the effect of temperature on WT and YH-HET PV neurons, as the PV neurons expressing the YH variant showed the most unexpected phenotypes in our study. In our room temperature recordings, we showed that the YH-HET variant decreased the rheobase current, increased the AP amplitude, and increased the AP firing. In our before-after comparison (22°C-35°C) of PV neurons (WT; n=11, YH-Het; n=10), the WT and YH-HET neurons showed the same temperature-dependent effects on these parameters, including increased rheobase, decreased AP amplitude, and a higher maximal firing rate, at 35°C compared to those at 22°C. These data have been added to the manuscript as a supplemental figure (Supplemental Fig. 4-3) and are briefly referenced and described in the Results section.     

Moreover, in our original manuscript, we showed that the effects of the homozygous YH variant on glutamatergic and NFS GABAergic neuron excitability were highly similar between cultured recordings at room temperature (~22°C) and slice recordings at 32°C. Taken together, these data suggest that the reported neurophysiological phenotypes downstream of the YH variant are likely not temperature dependent. 

Reviewer #2 (Public Review):

Summary:

In this manuscript, Shore et al. investigate the consequent changes in excitability and synaptic efficacy of diverse neuronal populations in an animal model of juvenile epilepsy. Using electrophysiological patch-clamp recordings from dissociated neuronal cultures, the authors find diverging changes in two major populations of inhibitory cell types, namely somatostatin (SST)- and parvalbumin (PV)-positive interneurons, in mice expressing a variant of the KCNT1 potassium channel. They further suggest that the differential effects are due to a compensatory increase in the persistent sodium current in PV interneurons in pharmacological and in silico experiments.

Strengths:

(1) Heterozygous KCNT1 gain of function variant was used which more accurately models the human disorder.

(2) The manuscript is clearly written, and the flow is easy to follow. The authors explicitly state the similarities and differences between the current findings and the previously published results in the homozygous KCNT1 gain of function variant.

(3) This study uses a variety of approaches including patch clamp recording, in silico modeling, and pharmacology that together make the claims stronger.

(4) Pharmacological experiments are fraught with off-target effects and thus it bolsters the authors' claims when multiple channel blockers (TTX and VU170) are used to reconstruct the sodium-activated potassium current. Having said that, it would be helpful to see the two drug manipulations used in the same experiment. Notably, does the more selective blocker VU170 mimic the results of TTX for NFS GABAergic cells in Figure 2? And does it unmask a genotype difference for FS GABAergic cells like the one seen in PV interneurons in Figure 5C3.

To illustrate the two drug manipulations in the same experiment, we recorded from WT SST and PV neurons (5 neurons/group) and blocked KNa currents first using TTX and then VU170, following wash out between the two drugs, in the same neurons. Below, we have plotted the points at each voltage step for each SST and PV neuron, and for each drug treatment, on the same graph to show how they vary directly. At each voltage step, lines connect the points representing the TTX-sensitive and VU-sensitive currents for each neuron to show the individual effects (left-most graphs). Summary data are shown across all voltages (middle graphs) and across negative voltages (right-most graphs).

Author response image 1.

We have not used VU170 on FS and NFS populations of GABAergic neurons. However, for reasons that are explained more extensively below in response to Weaknesses #3, we would not predict KNa currents recorded from SST- and PV-GABAergic neurons to mimic those of NFS- and FS-GABAergic neurons, respectively.

Weaknesses:

(1) This study relies on recordings in dissociated cortical neurons. Although specific WT interneurons showed intrinsic membrane properties like those reported for acute brain slices, it is unclear whether the same will be true for those cells expressing KCNT1 variants. This reviewer highly recommends confirming some of the key findings using an ex vivo slice preparation. This is especially important given the discrepant result of reduced excitability of PV cells reported by Gertler et al., 2022 (cited here in the manuscript but not discussed in this context) in acute hippocampal slices for a different KCNT1 gain of function variant.

We thank the reviewer for this suggestion. To test whether SST-expressing YH-HET neurons show similar impairments to those observed in culture, we crossed the FVB-Tg(GadGFP)45704Swn/J transgenic mouse line (Jackson Labs #003718), also known as the GIN line, to the Kcnt1-YH line. Mice from the GIN line express eGFP in a subpopulation of SST-expressing neurons in the hippocampus and cortex. We performed slice recordings of cortical layer 2/3, GFP-expressing neurons from P21-30, WT and YH-HET GIN mice. Although the input resistance was not significantly decreased, the rheobase was higher in the YH-HET neurons, and they fired fewer APs across increasing current steps, than WT neurons, supporting the main findings from the SST-expressing neurons in culture. These data have been added to the manuscript in a new figure (Fig. 9).

Regarding the previously published results on the effect of KCNT1 GOF on PV neuron excitability by Gertler et al., we have written a new paragraph in the Discussion section (last paragraph of the section, “Neuron-type-dependent KCNT1 GOF effects”) that discusses the differences between the findings by Gertler et al. and the current study. 

To further investigate the effects of heterozygous YH variant expression on SST- vs. PV-expressing neuron excitability in ex vivo slice recordings, we are now crossing a cre-inducible, Td-Tomato Red reporter line (Ai9) to the Kcnt1-YH line. After obtaining Ai9Tg/Tg; Kcnt1m/+ mice, we will cross these to Sst-Cre and Pvalb-Cre lines to be able to record from marked SST and PV, WT and YH-HET neurons in slice. We plan on submitting results from these recordings as an eLife Research Advances article linked to this article.

(2) It is unclear how different pieces of results fit together to form a story about the disease pathophysiology.

We have added a paragraph to the Discussion to speculate on how these various GABAergic subtype-specific effects downstream of the YH variant may contribute to overall network/brain pathology and seizure propensity in heterozygous mice.

For example, hyperexcitability of PV cells would suggest more inhibition which would counter seizure propensity. However, spontaneous inhibitory postsynaptic currents show no change in pyramidal neurons. Moreover, how do the authors reconcile that the reductions in synaptic inputs onto interneurons in Figure 3B with the increases in Figure 8? This should be discussed.

Generally, network and synaptic alterations downstream of the heterozygous variant were quite minimal compared with those of the homozygous variant. Although there were reductions in the frequency of synaptic inputs onto inhibitory neurons, the changes were relatively small. Thus, we concluded that the neuronal effects downstream of the heterozygous YH variant were below some threshold to result in broader network effects on synaptic activity and connectivity similar to those in the homozygous YH model. The discrepancies between our GABAergic vs. FS/NFS vs. VIP/SST/PV data will be discussed in more detail in response to Weakness #3.   

(3) Similarly, the results in this work are not entirely internally consistent. For example, given the good correspondence between FS and NFS GABAergic cells with PV and SST expression, why are FS GABAergic cells hyperexcitable in Figure 1? If anything, there is a tendency to show reduced excitability like the NFS GABAergic cells.

In our neuron cultures, 76-80% of Neu-N-expressing neurons are GFP+ (from the CamKII-eGFP virus used to mark glutamatergic neurons), and of the remaining ~20-24%, the majority are GABAergic (verified using the Dlx5/6-mRuby virus to mark GABAergic neurons and using electrophysiology to assess AP parameters and analyze evoked responses). In our original experiments, recordings sampled from this larger GABAergic population were used (Fig. 3), or this population was sorted almost equally into FS and NFS (Figs. 1 and 2).

In later experiments, we isolated and cultured neurons from VIP-Cre, SST-Cre, and PV-Cre mouse lines and marked these neuron types in vitro with a Cre-inducible mCherry virus. In the VIP-Cre cultures, ~6% of the GFP- population, or 1.2% of the Neu-N-population, was mCherry+. In the SST-Cre cultures, ~20.5% of the GFP- population, or 4.7% of the Neu-N-population, was mCherry+. In the PV-Cre cultures, less than 1% of the Neu-N-population was mCherry+, which is not surprising considering the relatively late onset of PV expression compared with those of VIP and SST. Thus, we would estimate that we are marking and recording from less than 30% of the total GABAergic population in these in vitro experiments, rather than the 80-90% that these three populations would sum to in vivo.  

Furthermore, using our original criteria for sorting GABAergic neurons into FS and NFS subtypes, all VIP recorded neurons were of the NFS type, PV of the FS type, whereas SST were of the FS (38%) and NFS (62%) types, which is not far off from the significant fraction of SST neurons that have been shown to be fast-spiking in slice experiments (Kvitsiani et al., 2013; Urban-Ciecko & Barth, 2016). Therefore, the FS group consists of both PV and SST neurons, and the NFS group consists of both VIP and SST neurons, and likely also contains immature PV neurons that have not yet developed a fast-spiking phenotype. Taken together, this suggests that the data from these two sets of experiments (FS/NFS vs. VIP/SST/PV) are not directly comparable.

Also, why do the WT I-V curves look so different between Figures 2 and 5? This reviewer suggests at least a brief explanation in the discussion.

As to the differences in appearance between the WT I-V curves in Figures 2 and 5, those plots are from different neuron types (Fig. 2: Glutamatergic, FS GABAergic, and NFS GABAergic vs. Fig. 5: VIP-, SST-, and PV-expressing), and the KNa currents are isolated using different methods (Fig. 2: TTX-subtraction vs. Fig. 5: VU170-subtraction). TTX blocks an inward Na+ current, which is apparent across subthreshold voltages in Fig. 2C1-3, whereas VU170 does not block this current, making it not apparent in Fig. 5C1-3. Also, the bottom three panels in Fig. 2C1-3 show the KNa current from -80 to 0 mV, whereas those in Fig. 5C1-3 show from -80 to -30 mV, to better illustrate the areas spanning KNa current increases, so their appearance is not directly comparable.

(4) Given the authors' claim that the KCNT1 activation curve is a major contributor to the observed excitability differences in specific GABA cell subtypes, it would be helpful to directly measure the activation curve in the variants experimentally as was done for WT KCNT1 in Figure 6A and use the derived kinetics in the compartmental model.

We apologize for the confusion. Although the activation curves among different GABAergic subtypes from WT KCNT1 are distinct, and we believe that these varying kinetics contribute to the neuron-type-specific phenotypes of KCNT1 GOF, we didn’t intend to suggest that the heterozygous Y777H variant itself causes neuron-type-specific alterations to the activation curves of the GABAergic subtypes. To clarify this point, below, we show the high similarity of the activation curves between WT KCNT1 and YH-HET KCNT1 in each of the GABAergic subtypes.

Author response image 2.

Reviewer #3 (Public Review):

Summary:

The present manuscript by Shore et al. entitled Reduced GABAergic Neuron Excitability, Altered Synaptic Connectivity, and Seizures in a KCNT1 Gain-of-Function Mouse Model of Childhood Epilepsy" describes in vitro and in silico results obtained in cortical neurons from mice carrying the KCNT1-Y777H gain-of-function (GOF) variant in the KCNT1 gene encoding for a subunit of the Na+-activated K+ (KNa) channel. This variant corresponds to the human Y796H variant found in a family with Autosomal Dominant Nocturnal Frontal lobe epilepsy. The occurrence of GOF variants in potassium channel encoding genes is well known, and among potential pathophysiological mechanisms, impaired inhibition has been documented as responsible for KCNT1-related DEEs. Therefore, building on a previous study by the same group performed in homozygous KI animals, and considering that the largest majority of pathogenic KCNT1 variants in humans occur in heterozygosis, the Authors have investigated the effects of heterozygous Kcnt1-Y777H expression on KNa currents and neuronal physiology among cortical glutamatergic and the 3 main classes of GABAergic neurons, namely those expressing vasoactive intestinal polypeptide (VIP), somatostatin (SST), and parvalbumin (PV), crossing KCNT1-Y777H mice with PV-, SST- and PV-cre mouse lines, and recording from GABAergic neurons identified by their expression of mCherry (but negative for GFP used to mark excitatory neurons).

The results obtained revealed heterogeneous effects of the variant on KNa and action potential firing rates in distinct neuronal subpopulations, ranging from no change (glutamatergic and VIP GABAergic) to decreased excitability (SST GABAergic) to increased excitability (PV GABAergic). In particular, modelling and in vitro data revealed that an increase in persistent Na current occurring in PV neurons was sufficient to overcome the effects of KCNT1 GOF and cause an overall increase in AP generation.

Strengths:

The paper is very well written, the results clearly presented and interpreted, and the discussion focuses on the most relevant points.

The recordings performed in distinct neuronal subpopulations are a clear strength of the paper. The finding that the same variant can cause opposite effects and trigger specific homeostatic mechanisms in distinct neuronal populations is very relevant for the field, as it narrows the existing gap between experimental models and clinical evidence.

Weaknesses:

My main concern is in the epileptic phenotype of the heterozygous mice investigated. In fact, in their previous paper the Authors state that "...Kcnt1-Y777H heterozygous mice did not exhibit any detectable epileptiform activity" (first sentence on page 4). However, in the present manuscript, they indicate twice in the discussion section that these mice exhibit "infrequent seizures". This relevant difference needs to be clarified to correctly attribute to the novel pathophysiological mechanism a role in seizure occurrence. Were such infrequent seizures clearly identified on the EEG, or were behavioral seizures? Could the authors quantify this "infrequent" value? This is crucial also to place in the proper perspective the Discussion statement regarding "... the increased INaP contribution to ... network hyperexcitability and seizures".

We apologize for the confusion. Indeed, in the Results section from our previous paper, we failed to observe seizures in 14 heterozygous mice, whereas 23/25 homozygous mice showed seizures by video-EEG. However, in the fifth paragraph of the Discussion section from that paper, we further stated that “during the preparation and review of [that] article, we observed seizures in two Kcnt1-Y777H heterozygous mice, one during a widefield Ca2+ imaging experiment and the other during a video-EEG experiment”. Thus, we concluded that “heterozygous expression can result in seizures in a rodent model, but apparently at a much lower frequency than that observed with homozygous expression”. To emphasize these findings, we’ve added a sentence to the Introduction in this manuscript about the occurrence of infrequent seizures in Kcnt1-Y777H heterozygous mice, along with a reference to the Discussion of our previous paper.

Of the two observed seizures, one seizure was captured in the Weston Lab at the University of Vermont from a Kcnt1-Y777H heterozygous mouse expressing a calcium indicator (after it was bred to the Snap25-GCaMP6s line) during a Ca2+ widefield imaging experiment, and it was accompanied by a time-locked video of the seizure event. The other seizure was recorded as a control during a drug study using video-EEG. This Kcnt1-Y777H heterozygous mouse had multiple tonic seizures, as evidenced by EEG traces and the accompanying video, which were recorded and analyzed in the Frankel Lab at Columbia University. The seizures from heterozygous mice have not been officially quantified, as they have only been rarely observed across multiple different experiments using heterozygous mice at multiple institutions, making quantification quite difficult.

Lastly, regarding attributing the role of the identified pathological mechanisms to seizure occurrence mentioned by the reviewer, we have added a paragraph to the Discussion to speculate on how the various GABAergic subtype-specific effects downstream of the YH variant may contribute to the general lack of network/brain pathology and seizure generation in heterozygous mice.  

Also, some statistical analysis seems to be missing. For example, I could not find any for the data shown in Fig. 6. Thus, the following statement: "the model PV neurons responded to KCNT1 GOF with decreased AP firing and an increased rheobase" requires proper statistical evaluation.

We thank the reviewer for this suggestion. We were initially hesitant to apply a formal statistical analysis to the modeling data because it differs in important ways from the experimental data. However, we have now provided statistical analyses of these data, with some caveats. Because we applied each KCNT1 GOF level (40, 35, and 30 mM) to the same set of neurons for each data set, we performed repeated measures ANOVA analyses to assess differences due to GOF in each subtype. We note that some changes are statistically significant, but may not be physiologically relevant. For example, there are changes in input resistance and rheobase in VIP neurons only at the higher GOF level (30 mM), but the magnitude of each change is quite small relative to those in SST neurons (Rin: 1.7 MΩ in VIP vs. 23.2 MΩ in SST, rheo: 1.7 pA in VIP vs. 52.5 pA in SST), and likely as a consequence, there are no downstream effects on the AP firing rate at either GOF level in VIP neurons. It is important to examine the magnitude of the effects and interpret them in the context of the changes in other neuron types and in the experimental data, thus, we’ve provided two new figures to better illustrate the relative changes in each neuron type (Supplemental Figures 6-2 and 7-2). We have also added these statistical results to Figures 6E2, 6F2, 6G2, and 7E, and Supplemental Fig. 6-1, and we have described them in the Results section. A summary of the statistics has also been added in Supplemental Table 6.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

In addition to addressing the weaknesses highlighted in the public review, this reviewer recommends using a KCNT1 agonist such as loxapine to see if activating the potassium channel mimics the KCNT1 GOF in SST and PV cells.

Although we appreciate this suggestion, we’re not sure whether treating GABAergic subtypes with loxapine would provide much clarity in the absence of many supporting experiments. First, the amount of channel activation and any changes in kinetics caused by loxapine would need to be measured and compared to the YH-HET GOF effects in order to interpret the results. In addition, the aforementioned caveat about off-target effects of small molecules would also have to be considered, as loxapine inhibits many other channels at nanomolar concentrations.

More importantly, we hypothesize that several of the GABAergic subtype-specific effects of KCNT1 GOF result from homeostatic or adaptive mechanisms due to long-term increases in KNa currents. For instance, PV-expressing YH-HET neurons had a lower rheobase, increased AP amplitude, and increased AP firing frequency, effects that we believe are due, not to increased KNa currents themselves, but to a compensatory increase in a persistent Na+ current. For the SST neurons, we hypothesize that the increased capacitance and soma size, together with the increased electrical coupling, exacerbate the hypoexcitability phenotype downstream of the YH variant. Thus, we would not necessarily expect that opening KCNT1 channels by acute loxapine treatment would mimic many of these effects.

Indeed, in a previous study using a different KCNT1 GOF mouse model, loxapine treatment mimics KCNT1 GOF effects in some neuron types (reduced AP firing frequency in loxapine-treated, WT PV neurons mimics that observed in heterozygous KCNT1 GOF PV neurons), but not in others (reduced AP firing frequency in loxapine-treated, WT pyramidal neurons does not mimic the unaltered AP firing frequency observed in heterozygous and homozygous KCNT1 GOF pyramidal neurons) (Gertler et al., 2022).  

Related to this suggestion by the reviewer, we are currently performing studies using a KCNT1 blocker in WT and Kcnt1-KO neurons to better understand the role of KCNT1 among cortical neuronal subtypes that will be published in a future manuscript.

Reviewer #3 (Recommendations For The Authors):

Though I realize that primary cultures allow for efficient identification of neuronal subclasses, it would have been useful to show that similar changes also occur in neurons with conserved in vivo connectivity, such as those recorded from brain slices.

We thank the reviewer for this suggestion. We have added an additional figure (Fig. 9) showing that the hypoexcitability phenotype observed in SST neurons in culture recordings is conserved in SST neurons in slice recordings from GIN mice, which express GFP predominately in SST-expressing neurons.

In addition, further experiments in PV neurons from Kcnt1-Y777H homozygous mice would provide evidence for a gene-dosage role in the changes found in heteros.

For this manuscript, we chose to focus our efforts on understanding the effects of heterozygous Kcnt1 variant expression in various neuronal subtypes with the goal of better modeling GOF variant effects in human disease. However, we’re very interested in investigating the effects of homozygous expression of the YH variant on each of the GABAergic subtypes to compare with those found in this study, but this requires more rounds of breeding to get homozygous mice with GABAergic subtype-specific expression of cre recombinase. We look forward to reporting the results from these experiments in a future manuscript.

Also, when addressing the issue regarding the different effects of the same GOF variant on the excitability of distinct neuronal populations in the Discussion or Introduction sections, the authors may want to cite the recent work on KCNQ2 and KCNQ3 by the Tzingounis group (https://pubmed.ncbi.nlm.nih.gov/37607817/).

We thank the reviewer for bringing this manuscript to our attention. We have added this citation to a new paragraph in the Discussion section regarding neuron-type specific effects of ion channel variants (the last paragraph focusing on the effects in PV neurons).

Budelli, G., Hage, T. A., Wei, A., Rojas, P., Jong, Y. J., O'Malley, K., & Salkoff, L. (2009). Na+-activated K+ channels express a large delayed outward current in neurons during normal physiology. Nat Neurosci, 12(6), 745-750. https://doi.org/10.1038/nn.2313

Gertler, T. S., Cherian, S., DeKeyser, J. M., Kearney, J. A., & George, A. L., Jr. (2022). K(Na)1.1 gain-of-function preferentially dampens excitability of murine parvalbumin-positive interneurons. Neurobiol Dis, 168, 105713. https://doi.org/10.1016/j.nbd.2022.105713

Grubb, M. S., & Burrone, J. (2010). Activity-dependent relocation of the axon initial segment fine-tunes neuronal excitability. Nature, 465(7301), 1070-1074. https://doi.org/10.1038/nature09160

Hage, T. A., & Salkoff, L. (2012). Sodium-activated potassium channels are functionally coupled to persistent sodium currents. J Neurosci, 32(8), 2714-2721. https://doi.org/10.1523/JNEUROSCI.5088-11.2012

Kuba, H., Oichi, Y., & Ohmori, H. (2010). Presynaptic activity regulates Na(+) channel distribution at the axon initial segment. Nature, 465(7301), 1075-1078. https://doi.org/10.1038/nature09087

Kvitsiani, D., Ranade, S., Hangya, B., Taniguchi, H., Huang, J. Z., & Kepecs, A. (2013). Distinct behavioural and network correlates of two interneuron types in prefrontal cortex. Nature, 498(7454), 363-366. https://doi.org/10.1038/nature12176

Shore, A. N., Colombo, S., Tobin, W. F., Petri, S., Cullen, E. R., Dominguez, S., Bostick, C. D., Beaumont, M. A., Williams, D., Khodagholy, D., Yang, M., Lutz, C. M., Peng, Y., Gelinas, J. N., Goldstein, D. B., Boland, M. J., Frankel, W. N., & Weston, M. C. (2020). Reduced GABAergic neuron excitability, altered synaptic connectivity, and seizures in a KCNT1 gain-of-function mouse model of childhood epilepsy. Cell Rep.

Urban-Ciecko, J., & Barth, A. L. (2016). Somatostatin-expressing neurons in cortical networks. Nat Rev Neurosci, 17(7), 401-409. https://doi.org/10.1038/nrn.2016.53

eLife assessment

Rachubinski and colleagues provide an important manuscript that includes two major advances in understanding immune dysregulation in a large cohort of individuals with Down syndrome. The work comprises compelling, comprehensive, and state-of-the-art clinical, immunological, and autoantibody assessment of autoimmune/inflammatory manifestations. Additionally, the authors report promising results from a clinical trial with the JAK inhibitor tofacitinib for individuals with dermatological autoimmune disease.

Reviewer #1 (Public Review):

Summary:

This paper represents a huge amount of work on a condition whose patients' health and well-being have not always been prioritized, and only relatively recently has the immune dysregulation seen in patients with Down Syndrome (DS) been garnering major research interest.

This paper provides an unparalleled examination of immune disorders in patients with DS. The authors also report the results from a clinical trial with the JAK inhibitor tofacitinib in DS patients.

Strengths:

This manuscript reports a herculean effort and provides an unparalleled examination of immune disorders in a large number of patients with DS.

Weaknesses:

Not a major weakness but, apart from finding an elevation of CD4 T central memory cells and more differentiated plasmablast, several of the alterations reported in this manuscript had already been suggested by a few case reports and a very small series. On the other hand, the number of patients (and controls) utilized for this study is remarkable and allows for drawing much firmer conclusions.

Reviewer #2 (Public Review):

In this manuscript, Rachubinski and colleagues provide a comprehensive clinical, immunological, and autoantibody assessment of autoimmune/inflammatory manifestations of patients with Down syndrome (DS) in a large number of patients with this disorder. These analyses confirm prior results of excess interferon and cytokine signals in DS patients and extend these observations to highlight early-onset immunological aberrancies, far before symptoms occur, as well as characterizing novel autoantibody reactivities in this patient population. Then, the authors report the interim analysis of an open-label, Phase II, clinical trial of the JAK1/3 inhibitor, tofacitinib, that aims to define the safety, clinical efficacy, and immunological outcomes of DS patients who suffer from inflammatory conditions of the skin. The clinical trial analysis indicates that the treatment is tolerated without serious adverse effects and that the majority of patients have experienced clinical improvement or remission in their corresponding clinical cutaneous manifestations as well as improvement or normalization of aberrant immunological signals such as cytokines.

The major strength of the study is the recruitment and uniform, systematic evaluation of an impressive number of DS patients. Moreover, the promising early results from the tofacitinib clinical trial pave the way for analysis of a larger number of patients within the Phase II trial and otherwise, which may lead to improved clinical outcomes for affected patients. An inherent weakness of such studies is the descriptive nature of several parameters and the relatively small size of tofacitinib-treated DS patients. However, the descriptive nature of some of the correlative research analyses is of scientific interest and is useful to generate hypotheses for future additional (including mechanistic) work, and treatment of 10 DS patients in a formal clinical trial at interim analysis is not a trivial task for a disease like this. The manuscript achieves the aims of the authors and the results support their conclusions. The authors appropriately acknowledge areas that require more research and areas that are not well understood. The results are represented in a useful manner and statistical methods and analyses appear sound.

Reviewer #3 (Public Review):

Summary:

Individuals with Down syndrome (DS) have high rates of autoimmunity and can have exaggerated immune responses to infection that can unfortunately cause significant medical complications. Prior studies from these authors and others have convincingly demonstrated that individuals with DS have immune dysregulation including increased Type I IFN activity, elevated production of inflammatory cytokines (hypercytokinemia), increased autoantibodies, and populations of dysregulated adaptive immune cells that pre-dispose to autoimmunity. Prior studies have demonstrated that using JAK inhibitors to treat patient samples in vitro, in small case series of patients, and in mouse models of DS leads to improvement of immune phenotype and/or clinical disease. This manuscript provides two major advances in our understanding of immune dysregulation and therapy for patients. First, they perform deep immune phenotyping on several hundred individuals with DS and demonstrate that immune dysregulation is present from infancy. Second, they report a promising interim analysis of a Phase II clinical trial of a JAK inhibitor in 10 people with DS and moderate to severe skin autoimmunity.

Strengths and weaknesses:

The relatively large cohort and careful clinical annotation here provide new insights into the immune phenotype of patients with DS. For example, it is interesting that regardless of autoimmune disease or autoantibody status, individuals with DS have elevated cytokines and CRP. Analysis of the cohorts by age demonstrated that some cytokines are significantly elevated in people with DS starting in infancy (e.g., IL-9 and IL-17C). Nearly all adults with DS in this study had autoantibodies (98%) and most had six or more autoantibodies (63%), which differed significantly from euploid study participants. This implies that all patients with DS might benefit from early intervention with therapy to reduce inflammation. However, it is also worth considering that an alternative interpretation that since hypercytokinemia does not vary based on disease state in individuals with DS, this may not be a key factor driving autoimmunity (although it may be relevant for other clinical symptoms such as neuroinflammation).

Small case series have suggested the benefit of JAK inhibitors to treat autoimmunity in DS. This is the first report of a prospective clinical trial to test a JAK inhibitor in this setting. The clinical trial entry criteria included moderate to severe autoimmune skin disease in patients aged 12-50 years with DS, and treatment was with the JAK1/3 inhibitor tofacitinib. This clinical trial is a critically important step for the field. The early results support that treatment is well tolerated with an improvement of interferon scores in patients and reduction of autoantibodies. Most patients experienced clinical improvement, with alopecia areata having the greatest response. Treatment may not affect all skin diseases equally, for example of the 5 patients with hidradenitis suppurativa, only 1 showed clinical improvement based on skin score. While very promising, the clinical trial results reported here are preliminary and based on an interim analysis of 10 patients at 16 weeks. Individuals with DS have a lifelong risk of immune dysregulation and thus it is unclear how long therapy, if of benefit, would need to be continued. The results of longer-term therapy will be informative when considering the risks/benefits of this therapy.

Author response:

Reviewer #1 (Recommendations For The Authors): 

This paper represents a huge amount of work on a condition whose patients' health and well-being have not always been prioritized, and only relatively recently has the immune dysregulation seen in patients with Down Syndrome (DS) been garnering major research interest. 

This paper provides an unparalleled examination of immune disorder in patients with DS. In a truly herculean effort, the authors provided the cumulative examination of over 440 patients with DS, confirmed the alterations in immune cell subsets (n=292, 96 controls) and multi-organ autoimmunity seen in these patients as they age, and identified autoantibody production that could contribute to conditions co-occurring in patients with DS. They also sought to look at whether the early immunosenescence seen in DS was due to the inflammatory profile by comparing age-associated markers in DS patients and euploid controls separately, finding that several markers are regulated with age regardless of group, while comparing the effect of age versus DS status on cytokine status identified inflammatory markers elevated in DS patients across the lifespan that do not increase with age or that increase with age only in the DS cohort. This is very interesting in the context of DS in particular, and immunity during aging in general. 

The second part of the manuscript presents the results from a clinical trial with the JAK inhibitor tofacitinib in DS patients. While the number of DS patients treated with tofacitinib was small, the results were often quite striking. Treatment was well-tolerated and the improvement of dermatological conditions was clear. The less responsive patients AA4 and AA2 provide a very clear illustration that these patients are sensitive to immune triggers during treatment. Additionally, the demonstration that patients' IFN scores and cytokine levels decreased without clear immunosuppression with tofacitinib treatment is encouraging, since treatment with this drug would need to be continuous. I would be curious to see if the patients added past the cutoff for interim analysis follow a similar trajectory. I would not ask the authors to add any data; the paper is well-written and logically constructed. 

I only have a small comment: I really did not like how Figure 2 a, d, and g tethered the coloring to the magnitude of fold change to show the effect of DS particularly for 2a and 2g. Given that these fold changes are quite modest, the coloring is very light and hard to distinguish. The clear takeaway is that the effect on T cells is greatest, but there must be a better way to illustrate this. Perhaps displaying this graph on a non-white background could help with contrast. 

We are grateful for the Reviewer’s very positive assessment of the manuscript and constructive feedback. We want to assure the Reviewer that similar analyses will be completed in the future for the entire cohort recruited into the trial to determine if similar trajectories and results are observed with the larger sample size. Additionally, following Reviewer’s guidance, we will explore alternative ways to present the data in Figure 2 for greater clarity in a revised version of the manuscript.

Reviewer #2 (Recommendations For The Authors): 

• Although the focus of the patients in the first part of the paper is on autoimmune/inflammatory conditions, it will be useful to also list the non-autoimmune infectious manifestations for reference with prevalence data. For example, otitis media, or lung infections (mentioned within the paper), or mucosal candidiasis. Same for other manifestations such as cardiac or malignant conditions. Given the impressive number of patients, it will be useful to the readers to have prevalence data for these as well, even in brief statements within the results. 

We appreciate this inquiry by the Reviewer and will present additional data on the co-occurring conditions mentioned by the Reviewer in a revised version of the manuscript.

• Have the authors looked at DN T cells and whether they may be enriched in DS patients, given their enrichment in some autoimmune conditions? 

Thanks for this inquiry. We did examine DN T cells (double negative T cells), which we referred to in our Figure 2 and Figure 2 – figure supplement 1 as non-CD4+ CD8+ T cells. Although this T cell subset is mildly elevated (in terms of frequency among T cells) in individuals with Down syndrome, the result did not reach statistical significance after multiple hypothesis correction. This negative result is shown in the heatmap in Figure 2 – figure supplement 1d.

• It would be useful to move the segment of the discussion that discusses the interim predefined analysis of the phase 2 trial to the corresponding segment of the results. As this reviewer was reading the paper, it was unclear why the interim analysis was done, whether it was predefined and it was not until the discussion that it became apparent. I believe it will help the readers to have a brief mention that this interim analysis was predefined and set to occur at the first 10 DS enrollees. Also, it would be helpful to state what is the total number of DS patients planned for enrollment in the Phase 2 trial which is continuing recruitment. 

We appreciate this comment and will modify the text following the Reviewer’s guidance in the revised manuscript. The trial will be considered complete once a total of 40 participants undergo 16-weeks of treatment with good medicine compliance (less that 15% missed doses).

• Although the authors present data on TPO autoantibodies before and after tofacitinib, it remains unclear whether the other non-TPO autoantibodies were altered during treatment or whether this was a TPO autoantibody-specific phenomenon. Was there an alteration in mature B cells or plasmablast populations after tofacitinib? If these data are available, they would further enhance the manuscript. If they are not available, it would be useful for the authors to discuss those in the discussion of the manuscript. 

We are grateful for this comment, which strongly aligns with our future research interests and plans for the analysis of the full cohort once the trial is completed. In the interim analysis, we analyzed only auto-antibodies related to autoimmune thyroid disease and celiac disease, as shown in the manuscript. However, we plan to complete a more comprehensive analysis of the effects of JAK inhibition on autoantibody production once the full sample set is available at the end of the trial. Likewise, the clinical trial protocol contemplates collection and processing of blood samples for immune mapping using mass cytometry, which will enable us to answer the question from the Reviewer about potential changes in B cells or plasmablasts populations. Following Reviewer’s guidance, we will discuss these planned analyses in the Discussion of the revised manuscript.

Reviewer #3 (Recommendations For The Authors): 

(1) Cellular immune phenotyping data in Figure 2 presents a large number of patients with DS versus euploid controls (292 and 96 respectively). Given the relatively large cohort there would seem to be an opportunity to determine whether age or sex alters the immune phenotype shown, for example, TEMRAs, etc. Was the data analyzed in this way? 

We welcome this comment, which clearly aligns with our research interests and planned additional analyses of these datasets generated by the Human Trisome Project. We can share with the Reviewer that although sex as a biological variable has minimal impacts on the strong immune dysregulation observed in Down syndrome, there are clear age-dependent effects, with some immune changes occurring early during childhood versus others taking place later in adult life. A manuscript describing a complete analysis of age-dependent effects on the multi-omics datasets in the Human Trisome Project is currently under preparation.

(2) The authors should strongly consider incorporating/discussing the findings from Gansa et al, Journal of Clinical Immunology May 2024 - where they reviewed the immune phenotype of 1299 patients with Down syndrome. 

Thanks for this suggestion, we will surely cite and discuss this recent paper in the revised manuscript.

(3) It is difficult to differentiate patients Hs2 and Ps1 in Figure 5d. 

Thanks for this observation, we will modify the labels for greater clarity in the revised manuscript.

(4) Given their finding of no correlation between cytokine levels/immune phenotype and autoimmunity, some additional discussion of the relevance of hypercytokinemia in the pathogenesis of autoimmunity would seem relevant (given that this was the basis for the clinical trial). The authors mention that cytokine levels may not be appropriate measures of disease in the patients. 

We welcome this opportunity to expand the discussion of the relevance of hypercytokinemia in the pathogenesis of autoimmunity and will do so in the revised manuscript.

(5) Data availability statement: appropriate.

eLife assessment

Based on analyses of retinae from genetically modified mice, and from wild-type ground squirrel and macaque, employing microscopic imaging, electrophysiology, and pharmacological manipulations, this valuable study on the role of Cav1.4 calcium channels in cone photoreceptor cells (i) shows that the expression of a Cav1.4 variant lacking calcium conductivity supports the development of cone synapses beyond what is observed in the complete absence of Cav1.4, and (ii) indicates that the cone pathway can partially operate even without calcium flux through Cav1.4 channels, thus preserving behavioral responses under bright light. The evidence for the function of Cav1.4 protein in synapse development is convincing and in agreement with a closely related earlier study by the same authors on rod photoreceptors. The mechanism of compensation of Cav1.4 loss by Cav3 remains unclear but appears to involve post-transcriptional processes. As congenital Cav1.4 dysfunction can cause stationary night blindness, this work relates to a wide range of neuroscience topics, from synapse biology to neuro-ophthalmology.

Joint Public Review

Cav1.4 calcium channels control voltage-dependent calcium influx at photoreceptor synapses, and congenital loss of Cav1.4 function causes stationary night blindness CSNB2. Based on a broad portfolio of methodological approaches - genetic mouse models, immunolabeling and microscopic imaging, serial block-face-SEM, ERGs, and electrophysiology - the authors show that cone photoreceptor synapse development is strongly perturbed in the absence of Cav1.4 protein, and that expression of a nonconducting Cav1.4 channel mitigates these perturbations. Further data indicate that Cav3 channels are present, which, according to the authors, may compensate for the loss of Cav1.4 calcium currents and thus maintain cone synaptic transmission. These data, which are in agreement with a similar study by the same authors on rod photoreceptor synapses, help to explain what functional defects exactly cause CSNB2 and why it is accompanied by only mild visual impairment.

The strengths of the present study are its conceptual and experimental soundness, the broad spectrum of cutting-edge methodological approaches pursued, and the convincing differential analysis of mutant phenotypes. Weaknesses mainly concern the mechanism by which Cav3 channels might partially compensate for the loss of Cav1.4 calcium currents.

Author response:

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

Reviewer #1 (Recommendations For The Authors): 

The authors should perform experiments to answer this question: does Cav3 transcription increase in the G369i-KI, or is there instead some post-transcriptional modulation that permits surface expression of functional Cav3-containing channels in the absence of typical HVA Ca conductances? Also, the authors should determine whether G369i-KI can mediate Ca2+ release from intracellular stores and whether release from stores is upregulated as Cav3-containing channel expression (or function) is increased. 

We performed transcriptomic (drop-seq) analysis to test whether a Cav3 subtype is upregulated in cones of G369i KI mice. These experiments show that, consistent with previous studies (PMID 35803735, 26000488), Cacna1h appears to be the primary Cav3 subtype expressed mouse cones. However, as shown in new Supp.Fig.S3, there was no significant difference in the levels of Cacna1h transcripts in WT and G369i KI cones. Therefore, we propose that there may be some post-transcriptional modification, or alteration in a pathway that regulates channel availability, that enables the contribution Cav3 channels to the whole-cell Ca2+ current in the absence of functional Cav1.4 channels cones.

We also performed Ca2+ imaging experiments in WT vs G369i KI cone terminals to assess whether the diminutive Cav3 current in G369i KI cone terminals may be compensated by upregulation of a Ca2+ signal such as from intracellular stores. Arguing against this possibility, depolarization-evoked Ca2+ signals in G369i KI cones were dramatically reduced compared to WT cones (new Fig.9). 

Reviewer #2 (Recommendations For The Authors): 

Major points- 

(1) It is stated in too many places that cone features in the Cav1.4 knock-in are "intact", preserved, or spared, but this representation is not accurate. There are two instances in this study that qualify as intact when comparing KI to WT: 1) the photopic a-waves in the Cav1.4 knock-in (also demonstrated in Maddox et al 2020) and 2) latency to the platform (current MS, Figure 7f). However, in the numerous instances listed below, the authors compared the Cav1.4 knock-in to the Cav1.4 knock-out, and then referred to the KI as exhibiting intact responses. The reference point for intactness needs to be wildtype, as appropriately done for Figures 2 and 3, and when comparing the KI to the KO the phrasing should be altered; for example: "the KI was spared from the extensive degeneration witnessed in the KO....". 

In most cases, we clearly note that there are key differences in the WT and the G369i KI cone synapses, which highlight the importance of Cav1.4-specific Ca2+ signals for certain aspects of the cone synapse. We disagree with the reviewer on the point that we did not often use the WT as a reference since most of our experiments involved comparisons of only WT and G369i KI (Figs. 3-6) or WT, G369i KI, and Cav1.4 KO (Figs.1,7—and in these cases comparisons specifically between WT and G369i KI mice were included). We used “intact” as a descriptor for G369i KI cone synapses since these are actually present, albeit abnormal in the G369i KI retina, whereas cone synapses are completely absent in the Cav1.4 KO retina. To avoid confusion, we modified our use of “intact” and “preserved” where appropriate.

A. Abstract, line 34 to 35: ".......preserved in KI but not in KO.". 

Abstract was rewritten and this line was removed.

B. Line 36: "....synaptogenesis remains intact". The MS documents many differences in the morphology of KI and WT cones (immunofluorescence and electron microscopy data), which is counter to an intact phenotype. 

The sentence was: “In CSNB2, we propose that Cav3 channels maintain cone synaptic output provided that the Ca2+-independent role of Cav1.4 in cone synaptogenesis remains intact.”

Here the meaning of “intact” refers to the Ca2+ -independent role of Cav1.4, not synapses. Thus, we have left the sentence unchanged.

C. This strikes the right balance, lines 67 to 68: "....although greatly impaired.....". 

D. Line 149, "Cone signaling to a postsynaptic partner is intact in G369i KI mice". This description is inaccurate. Here there is only WT and KI, and the text reads as follows in line 162: "terminals (Figure 6b). The ON and OFF components of EPSCs in G369i KI HCs were measurable, although lower in amplitude than in WT (Figure 6a,b)." Neither "measurable" nor "lower in amplitude" meet the definition of "intact", and actual numerical values are lacking in the text. 

We have added results showing that there are no light responses in the Cav1.4 KO horizontal cells and have modified the sentence to: “Cone synaptic responses are present in horizontal cells of G369i KI but not Cav1.4 KO mice”. 

We have modified discussion of these results as (line 210-213): “Consistent with the lack of mature ribbons and abnormal cone pedicles (Fig.1), HC light responses were negligible in Cav1.4 KO mice (Fig.8a,b). In contrast, the ON and OFF responses were present in G369i KI HCs although significantly lower in amplitude than in WT HCs (Fig. 8a,b).”

E. Please add a legend to Figure 6a to indicate the intensities. The shape of the KI responses is different from the control which is worthy of discussion: i) there is no clear cessation of HC EPSCs in the KI during the light ON period (when release stops, Im fluctuations should be minimal), and ii) the "peaked" appearances of the initial 500ms of the On and Off periods are very similar in shape for the KI (hard to interpret in the same fashion as a control response). How were the On and Off amplitudes analyzed? Furthermore, the OFF current is not summarized in Figure 6D, but should not this be when Cav3 should be opening and triggering release: Off response-EPSC? Lastly, Figure 6b,d shows a ~70% reduction in On-current in the KI, and the KI example of 6b an 80% reduction in Off current compared to WT. Yet, the only place asterisks are used to indicate sig diff is the DNQX data within each genotype in Fig 6d. These data cannot be described as showing "intact" KI responses, and the absence of numerical and statistical values needs to be addressed. 

New Fig.8a depicting the horizontal cell light responses has been modified to include the legend indicating light intensities. The ON and OFF amplitudes were analyzed as the peak current amplitudes. This information has been added to the legend.

The reviewer is correct in that the OFF response represents the EPSC whereas the ON response represents the decrease in the EPSC with light. To avoid confusion, we changed the y axis label for the averaged data to read ON or OFF “response” rather than “current” in new Fig.8b.

As the reviewer suggests, the more transient nature of the KI response during the light ON period could result from aberrant continuation of vesicular release during the light-induced hyperpolarization of cones in the KI mice, in contrast to the prolonged suppression of release by light which is evident in the WT responses. We speculated on this difference as follows (lines 237-241):

“In addition to its smaller amplitude, the transient nature of the ON response in G369i KI HCs suggested inadequate cessation of cone glutamate release by light (Fig.8b). Slow deactivation of Cav3 channels and/or their activation at negative voltages20 could give rise to Ca2+ signals that support release following light-induced hyperpolarization of G369i KI cones.”

We added astericks to new Fig.8b,d indicating statistical differences and description of the tests in the legend.

F. line 168 the section titled "Light responses of bipolar cells and visual behavior is spared in G369i KI but not Cav1.4 KO mice". 

Changed to: “Light responses of bipolar cells and visual behavior are present in G369i KI but not Cav1.4 KO mice”

Last sentence of erg results, 189-190: "These results suggest that cone-to-CBC signaling is intact in G369i KI mice.". "Spared and intact" are not accurate descriptions. The ERG data presented here shows massive differences between WT and the KI, except in the instance of awaves. 

This sentence was removed.

As for Figure 6, the results text related to Figure 7a-d does not present real numbers for ERG responses, and there is no indication of significant differences there or in the Figure panels. For instance, in Figure 7b, b-waves are KI are comparable to KO, except at the two highest-intensity flashes that show KI responses ~20% the amplitude of WT. Presentation of KI and KO data on a 6- to 10-fold expanded scale higher than WT can be misleading: a quick read of these Figure panels might make one incorrectly conclude that the KI is intact while the KO is impaired when compared to WT. The Methods section needs more details on the ERG analysis (e.g. any filtering out of oscillatory potentials when measuring b-wave, and what was the allowable range of time-to-peak for b-wave amplitude, etc..). 

The vertical scaling of the ERG results in new Fig.10c,d has been changed so as to reflect clearly diminished responses of the KO and KI vs the WT. Further details regarding the ERG analysis was added to the Methods section.

G. Can you point to other studies that have used the "visible platform swim test" used in Figure 7e, f, and specify further how mice were dark/light adapted prior to the recordings? 

As referenced in the Methods, original line 674, the methods we used for the swim test were described in our previous study (PMID 29875267). Other studies that have used this assay include PMIDs: 28262416, 26402607.

(2) The Maddox et al 2020 study does not safely address whether rods have a residual T-type Ca2+ current in the Cav 1.4 KO or KI. The study showed that membrane currents measured from rods in the KI and KO retina were distinct from WT, supporting their claim that L-type Ca2+ current is absent in the KI and KO. However, the recordings had shortcomings that challenge the analysis of Ca2+ currents: i) collected at room temp (22-24{degree sign}C), ii) at an unknown distance from the terminal (uncertain voltage clamp), iii) with a very slow voltage ramp rate that is not suitable for probing T-type currents (Figure 1d Maddox 2020, 140 mV over 1 sec: 7msec/1mV), and iv) at a signal-to-noise that does not allow to resolve a membrane current under 1 pA (avg wt rod Ca2+ current was -3.5 pA, and line noise ~1pA peak-to-peak in Maddox 2020). Suggestion: say T-type currents were not probed in Maddox et al 2020, but Davison et al 2022 did not find PCR signal for Cav3.2 in rods. 

We disagree that recordings in the Maddox 2020 study were not sufficient to uncover a T-type current. The voltage ramps in that study were not much slower than that of the Davison et al. 2022 study (they used 0.19 mV/ms). Moreover, in new Supp. Fig.S1, we show that like the slower voltage ramp (0.15 mV/ms) used in the prior study of G369i KI rods, the voltage ramps we used in the present study (0.5 mV/ms), which clearly evoke currents with T-type properties in G369i KI cones (Fig.2a,b, Fig.3a,b) do not evoke currents in WT or G369i KI rods.  

Minor comments. 

(1) Suggestion: add an overview panel to Figure 1 that shows the rod terminals in the KI. The problem is that cropping out the ribbon and active zone signals from rods, to highlight cones, can give the impression that the cones are partially spared in the KI, and the rods are not spared at all. (yet you nicely clarify this in Figure 4 and in the legend and text, etc.). 

We chose to modify the legend with this information as in Fig.4 rather than modify the figure.

(2) Mouse wt cone Ca2+ currents look like L-type currents, as do your monkey and squirrel cone recordings, and also much like those of mouse rods (see Figure S5, Hagiwara et al., 2018 or Grabner and Moser 2021). Your pharm data from mice and squirrels further supports your conclusion, and certainly took much effort. Davison et al 2022 J Neurosci showed PCR results that support their claim that a Cav3 current exists in wt cones. Questions: 1) have you tried PCR? 2) Can you offer more details on what Cav3 KO you tried and what antibodies failed to confirm the KO? As the authors know, one complication is that the deletion of one Cav can be compensated for by the expression of a new Cav. There are 3 types of Cav3s and removal of one type may be compensated for by another Cav3. 

We have included drop-seq data (new Supp.Fig.S3) implicating Cav3.2 as the main Cav3 subtype in cones and have modified our discussion of these results accordingly. These experiments did not reveal any changes in Cav3 subtype expression in G369i KI vs WT cones.

(3) Lines 95/96- onward, spend more time telling the story. When working out the biophysical and pharmacological behavior of the Ca2+ currents, you might want to initially refer to the membrane current as a membrane current, and then state how your voltage protocols, intra- and extra-cell solutions, and drugs helped you verify 1) L-type and 2) T-type Ca2+ currents. 

We have modified the text with more detail.

(4) If data is in hand, add a ramp I-V to Figure S2, which shows the response of the ground squirrel cone. The steps in S2a are excellent for making your point that a transient current is missing, and the bipolar is a great control to illustrate ML218 works. However, a comparison of a squirrel cone ramp to a bipolar ramp response could complete the figure. 

See Reponse to #5 below.

(5) Consider moving Supplementary Figures S2 and S3 to the main text; these are highly relevant to the story, novel, and well-executed. 

Fig.S2 and S3 were added as new Figs.4,5. The new Fig.4 includes voltage ramps in ground squirrel cones (panel a) to compare with the bipolar data (panel f).

(6) The nice electron microscopy reconstructions are not elaborated on in any detail, and there is no mention of ribbon size. Is the resolution sufficient to estimate ribbon size, the number of synaptic vesicles around the ribbon and in the adjacent cytosol? The images indicate major changes in the morphology of the terminals. Is the glial envelope similar in WT and KI? 

Since ribbons were quantified extensively in the confocal analyses in Fig.6, we felt it unnecessary to add this to the EM analysis which focused mainly on aspects of 3D structure (i.e., arrangement of ribbons, postsynaptic wiring, cone pedicle morphology). We added further discussion of the change in morphology of the G369i KI cone pedicle (lines 200-203): “Compared to WT, ribbons in G369i KI pedicles appeared disorganized and were often parallel rather than perpendicular to the presynaptic membrane (Fig.7a-c). Consistent with our confocal analyses (Fig.1), G369i KI cone pedicles extended telodendria in multiple directions rather than just apically (Fig. 7a).”

While we did not opt to characterize the glial envelope in WT cones, we did add an analysis of synaptic vesicles around ribbons to Table 2.

(7) Discussion line 250: "we found no evidence for a functional contribution of Cav3 in our recordings of cones in WT mice (Figures. 2,3), ground squirrels, or macaque (Supplementary Figures S2 and S3).". I would not use "functional" in this context because when comparing your work to Davison et al 2022, they defined functional as a separate response component driven by Cav3. For instance, they examined the influence of their T-type current on exocytosis (by membrane capacitance) and other features like spiking Ca2+ transients. Suggestion: substitute functional with "detectable", and say "we found no detectable Cav currents". Or if you had Ttype staining, but not T-type Ca2+ currents, then say "no functional current even though there is staining...". 

We have modified the text as (lines 336-338): “However, in contrast to recordings of WT mouse cone pedicles in a previous study21, we found no evidence for Cav3-mediated currents in somatic recordings of cones in WT mice (Figs.2,3).”

We propose an alternative interpretation of the results in the Davison et al study concerning the conclusion that Cav3 channels contribute to Ca2+ spikes and exocytosis. That study used 100 µM Ni2+ to block a “T-type” contribution to spike activity in cones. In their Figs.4,5, the spikes are suppressed by 100 µM Ni2+ and 10 µM nifedipine, a Cav1 antagonist, and spared by the T-type selective drug Z944. This is problematic for several reasons. First, as shown by the authors

(their Fig.2A1,A2) and others (PMID: 15541900), 100 µM Ni2+ inhibits Cav1-type currents in photoreceptors. Second, Z944 potentiates Cav1 current in their mouse cones (their Fig.2C1,C2). Thus, both reagents are suboptimal for dissecting the contribution of either Cav subtype to spiking activity. With respect to Cav3 channels and exocytosis, these authors interpreted a reduction in exocytosis upon holding at -39 mV compared to at -69 mV as indicating a loss of a T-type driven component of release. However, Cav1 channel inactivation (PMID: 12473074) could lead to the observed reduction in exocytosis at -30 mV.

(8) Additional literature related to your Intro and Discussion. Regarding CSNB2, related mutations of active zone proteins, and what happens to Ca2+ currents when ribbons are deleted, you might want to consider the following studies that measure Ca2+ currents from rods: conditional KO of RIM1/2 (Grabner et al 2015 JN), KO of ELKS1/2 (Hagiwara et al, 2018 JCB), and KO of Ribeye (Grabner and Moser eLife 2021). In these studies, the Cav currents were absent in rods of the ELKS1/2 DKO, strongly reduced (80%) in the RIM1/2DKO, but altered in more subtle ways (activation-inactivation) without significantly changing steady-state Ca2+ current in the Ribeye KO. This does not seem to support some of the arguments you have made in the Introduction and Discussion regarding ribbon size and Ca2+ currents, yet the suggested literature is related to the topic at hand. 

A description of these synaptic proteins as potential mediators of the effect of Cav1.4 on ribbon morphogenesis was added to the Discussion, lines 325-327.

(9) Line 129: "Along with the major constituents of the ribbon, CtBP2, and RIBEYE", for clarity Ribeye has two domains, one that is identical to CtBP2 (B-domain) and the unique Ribeye domain (A-domain) that is only expressed at ribbon synapses. And, Piccolino is also embedded in the ribbon (Brandstaetter lab, Wichmann/Moser labs). In other words, Ribeye and Piccolino are the major constituents of the ribbon. 

To avoid confusion, we simply mention Ctbp2 and RIBEYE in the context of the corresponding antibodies that were used to label ribbons.

(10) Abstract: consider to rephrase "Ca2+-independent role of Cav1.4" by "Ca2+-permeationindependent role of Cav1.4" or alike 

Sentence changed to: “In CSNB2, we propose that Cav3 channels maintain cone synaptic output provided that the nonconducting role of Cav1.4 in cone synaptogenesis remains intact.”

Reviewer #3 (Recommendations For The Authors): 

Cav1.4 voltage-gated calcium channels play an important role in neurotransmission at mammalian photoreceptor synapses. Mutations in the CACNA1f gene lead to congenital stationary night blindness that particularly affects the rod pathway. Mouse Cav1.4 knockout and Cav1.4 knockin models suggest that Cav1.4 is also important for the cone pathway. Deletion of Cav1.4 in the knockout models leads to signaling malfunctions and to abundant morphological re-arrangements of the synapse suggesting that the channel not only has a role in the influx of Ca2+ but also in the morphological organization of the photoreceptor synapse. Of note, also additional Cav-channels have been previously detected in cone synapses by different groups, including L-type Cav1.3 (Wu et al., 2007; pmid; Kersten et al., 2020; pmid), and also T-type Cav3.2 (Davison et al., 2021; pmid 35803735). 

In order to study a conductivity-independent role of Cav1.4 in the morphological organization of photoreceptor synapses, the authors generated the knockin (KI) mouse Cav1.4 G369i in a previous study (Maddox et al., eLife 2020; pmid 32940604). The Cav1.4 G369i KI channel no longer works as a Ca2+-conducting channel due to the insertion of a glycine in the pore-forming unit (Madox et al. elife 2020; pmid 32940604). In this previous study (Madox et al. elife 2020; pmid 32940604), the authors analyzed Cav1.4 G369i in rod photoreceptor synapses. In the present study, the authors analyzed cone synapses in this KI mouse. 

For this purpose, the authors performed a comprehensive set of experimental methods

including immunohistochemistry with antibodies (also with quantitative analyses), electrophysiological measurements of presynaptic Ca2+ currents from cone photoreceptors in the presence/absence of inhibitors of L-type- and T-type- calcium channels, electron microscopy (FIB-SEM), ERG recordings and visual behavior tests of the Cav G369i KI in comparison to the Cav1.4 knockout and wild-type control mice. 

The authors found that the non-conducting Cav channel is properly localized in cone synapses and demonstrated that there are no gross morphological alterations (e.g., sprouting of postsynaptic components that are typically observed in the Cav1.4 knockout). These findings demonstrate that cone synaptogenesis relies on the presence of Cav1.4 protein but not on its Ca2+ conductivity. This result, obtained at cone synapses in the present study, is similar to the previously reported results observed for rod synapses (Maddox et al., eLife 2020, pmid 32940604). No further mechanistic insights or molecular mechanisms were provided that demonstrated how the presence of the Cav channels could orchestrate the building of the cone synapse. 

We respectfully disagree regarding the mechanistic advance of our study. As indicated by Reviewer 2, a major advance of our study is in providing a mechanism that can explain the longstanding conundrum that congenital stationary night blindness type 2 mutations that would be expected to severely compromise Cav1.4 function do not produce complete blindness. Our study provides an important contrast to the Maddox et al 2020 study in showing that rods and cones respond differentially to loss of Cav1.4 function, which is also relevant to the visual phenotypes of CSNB2. How the presence of Cav1.4 orchestrates cone synaptogenesis is an important topic that is outside the scope of our present study.

In the present study, the authors also propose a homeostatic switch from L-type to (newly occurring) T-type calcium channels in the Cav1.4 G369i KI mouse as a consequence of the deficient calcium channel conductivity in the Cav1.4 G369i Cav1.4 KI mouse. In cones of the Cav1.4 G369i, the high-voltage activated, L-type Ca2+-entry was abolished, in agreement with their previous paper (Maddox et al., eLife 2020, pmid 32940604). The authors found a lowvoltage activated Ca2+ current instead that they assigned to T-type Ca2+-currents based on pharmacological inhibitor experiments. T-type Ca2+-currents/channels were already previously identified in other studies by independent groups and independent techniques

(electrophysiology, RT-PCR, single-cell sequencing) in cones of wild-type mice (Davison et al.,

2021, pmid 35803735; Macosko et al., 2015, pmid 26000488; Williams et al., 2022, pmid 35650675). In the present manuscript (Figures 3a/b), the authors also observed a low-voltage activated, T-type like current in cones of wild-type mice, that is isradipine-resistant and affected by the T-type inhibitor ML218. This finding appears compatible with a T-type-like current in wildtype cones and is consistent with the published data mentioned above, although the authors interpret this data in a different way in the discussion. 

Due to the noise inherent in whole cell voltage clamp measurements and some crossover effects in the pharmacology, we cannot completely exclude the presence of a T-type current in WT mouse cones. However, our results very clearly support a conclusion opposite to that stated by the reviewer. Namely, if WT mouse cones have T-type Ca currents, then they are far smaller than those in the Cav1.4 G369i KI and KO cones. In particular, while we identified message for Cav3.2 in WT mouse cones, we were unable to identify a functional T-type current by either voltage clamp measurements or pharmacology. See below for a detailed rebuttal.

This proposal of a homeostatic switch is not convincingly supported in this reviewer's opinion

(for further details, please see below). Furthermore, no data on possible molecular mechanisms were provided that would support such a proposal of a homeostatic switch of calcium channels. No mechanistic/molecular insights were provided for a proposed homeostatic switch between Ltype to T-type channels that the authors propose to occur between wild-type and Cav1.4 G369i as a consequence of conduction-deficient Cav1.4 G369i channels. Is this e.g. based on posttranslational modifications that switch on T-type channels or regulation at the transcriptional level inducing expression of T-type calcium channel or on other mechanisms? The authors remain descriptive with their central hypotheses. No molecular mechanisms/signaling pathways were provided that would support the idea of such a homeostatic switch. 

Homeostatic plasticity refers to the maintenance of neuronal function in response to some perturbation in neuronal activity and can result from changes in the expression of ion channel genes (PMID: 36377048, 32747440, 19778903) or regulatory pathways that modulate ion channels (PMID: 15051886, 32492405). We present multiple lines of evidence showing that Cav3 currents appear in cones upon genetically induced Cav1.4 loss of function and can support cone synaptic responses and visual behavior if cone synapse structure is maintained. Our new transcriptomic studies show no difference between levels of Cav3 channel transcripts in WT and G369i KI cones, suggesting that the appearance of the Cav3 currents in G369i KI cones does not result from an increase in Cav3 gene expression. We are currently investigating our transcriptomic dataset to determine if Cav3 regulatory pathways are upregulated in G369i KI cones and will present this in a follow-up study.

The authors show residual photopic signaling in the non-conducting Cav1.4 G369i KI mouse as judged by the recording of postsynaptic currents, ERG recordings and visual behavior tests though in a reduced manner. The residual cone-based signaling could be based on the nonaffected T-type Ca2+ channel conductivity in cone synapses. Given that the L-type current through Cav1.4 is gone in the Cav1.4 G369i KI as previously shown (Maddox et al., 2020, pmid 32940604), the T-type calcium current will remain. However as discussed above, this does not necessarily support the idea of a homeostatic switch. 

A major point which we highlighted with new results is that despite the expression of Cav3 transcripts in WT mouse cones, Cav3 channels do not contribute to the cone Ca2+ current. This is at odds with the Davison et al study (PMID: 35803735, see our response to Reviewer 2, pt 7 for caveats of this study), but our results convincingly show that the Cav3 current appears only when Cav1.4 is genetically inactivated. Pharmacological or electrophysiological methods that should reveal the presence of Cav3 currents do not change the properties of the Ca2+ current in cones of WT mice, ground squirrel, or macaque:

• Figs.2-4: Voltage steps to -40 mV (Fig 2e) that activate a sizeable T-current in G369i KI mouse cones produce a negligible transient at pulse onset in WT mouse cones. Similarly, transient currents that are obvious in G369i KI mouse cones during the final step to -30 mV are absent in WT cones.  When we block Cav1.4 with isradipine either in cones of WT mice or ground squirrel, the current that remains does not resemble a Cav3 current but rather a scaled down version of the L-type current. ML218, which readily blocks Cav3 channels in HEK293T cells and in G369i KI cones, has only minor effects in cones of WT mice and ground squirrel; these effects of ML218 can be attributed to non-specific actions on Cav1.4 (new Supp.Fig.S2). New Fig.4 (moved from the supplementary data to the main article) clearly shows that the ML218-sensitive current in ground squirrel cones exhibits properties of Cav1.4 not Cav3 channels. 

• Figs.2,5: Holding voltages that inactivate Cav3 channels have no effect on the Ca2+ current in cones of WT mice or macaque (recordings of macaque cones were moved from the supplement to the main article as new Fig.5).

In Figure 4 the authors measured an increase in the size of the active zone (as judged by the size of the bassoon cluster) and of the synaptic ribbons in the Cav1.4 G369i. A mechanistic explanation for this phenomenon was not provided and the underlying molecular mechanisms were not unraveled. 

The FIB-SEM data uncover some ultrastructural alteration/misalignments of the synaptic ribbons and misalignments of the regular arrangement of the postsynaptic dendrites in the G369i KI mice. Also concerning this observation, the study remains descriptive and does not reveal the underlying mechanisms as it would be expected for eLife. 

We respectfully disagree on the descriptive nature of our study and the need for a full characterization of the molecular mechanism underlying the cone synaptic defects in the G369i KI mouse.   

An important study in the field (Zanetti et al., Sci. Rep. 2021; pmid 33526839) should be also cited that used a gain-of-function mutation of Cav1.4 to analyze its functional and structural role in the cone pathway. 

We have added citation of this paper to the Discussion (lines 354-356).

In conclusion, the study has been expertly performed but remains descriptive without deciphering the underlying molecular mechanisms of the observed phenomena, including the proposed homeostatic switch of synaptic calcium channels. Furthermore, a relevant part of the data in the present paper (presence of T-type calcium channels in cone photoreceptors) has already been identified/presented by previous studies of different groups (Macosko et al., 2015; pmid 26000488; Davison et al., 2021; pmid 35803735; Williams et al., 2022; pmid 35650675). The degree of novelty of the present paper thus appears limited. I think that the study might be better suited in a more specialized journal than eLife. 

We thank the reviewer for acknowledging the rigor of our study but disagree with their evaluation regarding the novelty of our work as outlined in our responses above.

eLife assessment

The authors have presented an interesting set of results showing that female sex peptide signaling adversely affects late-life neurodegeneration after early-life exposure to repetitive mild head injury in Drosophila. This fundamental work substantially advances our understanding of how sex-dependent response to TBI occurs by identifying the Sex Peptide and the immune system as modulators of sex differences. The evidence supporting the conclusions is compelling with rigorous inclusion of controls and appropriate statistics.

Reviewer #1 (Public Review):

Summary:<br /> In this manuscript the authors use the model organism Drosophila to explore sex and age impacts of a TBI method. They find age and sex differences: older age is susceptible to mild TBI and females are also more susceptible. In particular, they pursue a finding that virgin vs mated females show different responses: virgins are protected but mated females succumb to TBI with climbing deficits. In fact, virgin females compared to mated females are largely protected. They discover this is associated with exposure of the females to Sex Peptide in the reproductive neurons of the female reproductive tract. When they extend to RNAseq of brains, they show that there are very few genes in common between males, mated females, virgins and females mated with males lacking sex peptide. But what the few chronic genes associated with mated females seem associated with the immune system. These findings suggest that mated females have a compromised immune system, which might make them more vulnerable. In a bigger context, these findings point to the idea that the life status of the animal/individual may have an important impact on the outcome of a TBI - here illustrated by the differential state of virgin vs mated females in Drosophila.

Strengths:<br /> This is an interesting paper that allows a detailed comparison of sex and age in TBI which is largely only possible in such a simple model, where large numbers and many variations can be addressed. Overall the findings are interesting.

Weaknesses:<br /> Although the findings beyond Sex peptide are observational, the work sets the stage for more detailed studies to pursue the role of the genes they find by RNAseq and whether for example, boosting the innate immune system would protect the mated females, among other experiments.

Reviewer #2 (Public Review):

Summary:<br /> In this manuscript, the authors use the Drosophila model system to study the impact of mild head trauma on sex-dependent brain deficits. They identify Sex Peptide as a modulator of greater negative outcomes in female flies. Additionally, they observe that increased age at the time of injury results in worse outcomes, especially in females, and that this is due to chronic suppression of innate immune defense networks in mated females. The results demonstrate a novel signaling pathway that promotes age- and sex-dependent outcomes after head injury.

Strengths:<br /> The authors have modified their previously reported TBI model in flies to mimic mild TBI, which is novel. Methods are explained in detail, allowing for reproducibility. Experiments are rigorous with appropriate statistics. A number if important controls are included. The work tells a complete mechanistic story and adds important data to increase our understanding of sex-dependent differences in recovery after TBI. The Discussion is comprehensive and puts the work in context of the field.

Weaknesses: None<br /> The authors answered the following concerns, and I have no other concerns:<br /> A very minor weakness is that exact n values should be included in the figure legends. There should also be confirmation of knockdown by RNAi in female flies either by immunohistochemistry or qRT-PCR if possible.

Reviewer #3 (Public Review):

Summary:<br /> In this manuscript, the authors used a Drosophila model to show that exposure to repetitive mild TBI causes neurodegenerative conditions that emerge late in life and disproportionately affect females. In addition to the well-known age-dependent impact, the authors identified Sex Peptide (SP) signaling as a key factor in female susceptibility to post-injury brain deficits.

Strengths:<br /> The authors have presented a compelling set of results showing that female sex peptide signaling adversely affects late-life neurodegeneration after early-life exposure to repetitive mild head injury in Drosophila. They have compared the phenotypes of adult male and female flies sustaining TBI at different ages, and the phenotypes of virgin females and mated females, 2) compared the phenotypes of eliminating SP signaling in mating females and introducing SP-signaling into virgin females, 3) compared transcriptomic changes of different groups in response to TBI. The results are generally consistent and robust.

Weaknesses:<br /> The authors have made their claims largely based on assaying climbing index and vacuole formation as the only indicators of late-life neurodegeneration after TBI. Furthermore, it is also really surprising to see so few DEGs even in wild-type males and mated females and to see that none of DEGs overlap among groups or are even related to the SP-signaling. The authors state that the reason is their TBI is very minor. It is critical to independently verify their RNA-sequencing results and to add some more molecular evidence to support their conclusion. Finally, since similar sex peptide signaling is not present in mammalians or humans, its implication in humans remains unclear.

Author response:

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

Reviewer #1 (Recommendations For The Authors):

My comments are largely limited to suggestions to make the manuscript easier to read and digest.

In the abstract they say RNA sequencing highlights changes in innate...

Could they be more specific? Innate immune system up or down? They do not indicate actual findings in the abstract.

We thank the reviewer for the comment and we have revised the abstract accordingly.  

Their use of non‐intuitive abbreviations is often confusing. Perhaps they can add a table in methods listing all the abbreviations so that the reader can follow the data better. mNGA, vmHT....etc.

As suggested, we have now included a list of the abbreviations used in the paper.

There are mis‐spellings in the manuscript.

We have gone through the manuscript and corrected the mis-spellings.   

Has the SPR RNAi line been validated?

The SPR RNAi line that we used has been extensively validated by Yapici et al., 2007 and several subsequent publications. Importantly, the effectiveness of SPR knockdown is evident in female flies as they exhibit dramatically reduced egg laying and, importantly, lack the typical post-mating behaviors (such as rejection of male flies after initial mating) observed in the wild type mated female flies. In fact, female flies with RNAi-mediated SPR knockdown behave identically to females mated with SP-null male flies, confirming the effective disruption of the SP-SPR signaling pathway. We have revised the manuscript and added these statements in the results section concerning SPR RNAi.  

In the figures showing the Climbing Index vs time, can they abbreviate seconds as sec vs s? At least I think it is seconds. At first, I thought it was Time or Times, and was confused about what they were indicating on those types of graphs (Figures 1D‐F).

We have revised the figure as suggested by the reviewer.

In Figure 3F, they have a significance indicated in an unclear manner. It looks like they are comparing neuropil to the cortex, but I think they really mean to compare the cortex of sham to cortex of D31?

The reviewer was correct. We have revised figure 3F to make this clear.     

In Figure 4B, what is the y‐axis? Percentage of what? Is that percentage of total flies?

The reviewer was correct. We have revised the figure to make this clear. 

In a figure like SF3 B, what is the y‐axis? "Norm. Accum. CI" Can they explain the abbreviation?

We have revised the Y-axis label to be “Normalized accumulative CI”.  We have also made this clear in the legend.   

In the methods, what does this mean: "Regions devoid of Hoechst and phalloidin signal in non‐physiologically appropriate areas were considered vacuoles"? What are non‐physiologically appropriate areas? To me, that would mean outside of the brain. I would have thought the areas should be physiologically appropriate (aka neuropil and cortex)? This is confusing.

We have revised the method section to be more specific.  In the Drosophila brain, there are structures such as esophagus that are devoid of both Hoechst and phalloidin staining, which were excluded from our vacuole quantification.    

Reviewer #2 (Recommendations For The Authors):

Since I use mammalian systems, my comment about the confirmation of siRNA should be removed if this is not possible in the Drosophila system.

We have revised the figures to include total N values when appropriate. Including individual n values for each experimental assay and condition will inevitably crowd the figure legends, so specific values are available upon request. 

Regarding RNAi knockdown of sex peptide receptors (SPRs), we agree that confirmation of the knockdown by IHC or qRT-PCR will further strengthen our findings. It should be noted, however, that the RNAi line we used has been extensively validated by Yapici et al., 2007 and several subsequent publications. Importantly, the effectiveness of SPR knockdown is evident in female flies as they exhibit dramatically reduced egg laying and, importantly, lack the typical post-mating behaviors (such as rejection of male flies after initial mating) observed in the wild type mated female flies. In fact, female flies with RNAi-mediated SPR knockdown behave identically to females mated with SP-null male flies, confirming the effective disruption of the SP-SPR signaling pathway. We have revised the manuscript to include these statements in the results concerning the SPR RNAi knockdown.    

Reviewer #3 (Recommendations For The Authors):

(1) In Figures 1 and 2, the authors found that females have a lower climbing index in the acute phase in D17 injury, not due to neurodegeneration as shown no significant changes of brain vacuolation and other markers. However, in Figure 3, the authors found that female flies have a lower climbing index, more brain vacuolation, and neurodegeneration in the late phase. It's not very convincing that having a lower climbing index at the late phase is due to neurodegeneration. Is it possible that females suffered from more severe acute effects, at least in D17 injury?

We thank the reviewer for this point. Female flies injured on D17 displayed acute climbing deficits at 90 minutes post-injury. Since we did not observe significant structural changes in the brain at this time, we believe that this short-term functional deficit is not due to acute neuronal death. Here it is important to note that males did not display any acute climbing deficits when injured on D17, which suggests that the females suffered from more severe acute effects than males. However, these injured female flies recovered fully at 24 hours post-injury and displayed no climbing deficits. At two weeks post-injury, we observe climbing deficits and increased vacuole formation as a direct result of the injuries on D17 (see Supplemental Figure 3). When we assessed sensorimotor behavior and brain vacuolation on D45, we found that the injured females had significantly lower climbing indices and more brain vacuolation than the non-injured females of the same age. In this case, the concurrent observance of decreased climbing ability and increased brain vacuolation suggests chronic neurodegeneration in aged, injured females. This is not to be confused with the acute neuronal death observed by other groups using injury models of stronger severity. Overall, our data are consistent with the current view that in many neurodegenerative diseases, functional deficits often precede observable brain degeneration, which may take years to manifest.

(2) The authors determined late‐life brain deficits and neurodegeneration purely based on climbing index and vacuole formation. These phenotypes are not really specific to TBI‐related neurodegeneration and the significance and mechanisms of vacuole formation are not clear. Indeed, in Figures 3 A and B, male flies especially D31inj tend to have a much larger variation than any other groups. What could be the reasons? The authors should perform additional analyses on TBI‐related neurodegeneration in flies, which have been shown before, such as retinal degeneration and loss, neuronal degeneration, and loss, neuromuscular junction abnormalities, etc (Genetics. 2015 Oct; 201(2): 377‐402).

We thank the reviewer for the thorough evaluation of our manuscript. The reviewer raised a very important question: whether the neurodegeneration observed in our model is specific to TBI. As the reviewer rightly pointed out, the neurodegenerative phenotypes are unlikely to be specific to TBI-related neurodegeneration. Throughout the manuscript, we have tried to convey the notion that the mild physical impacts to the head represent one form of environmental insults, which in combination with other risk factors such as aging can lead to the emergence of neurodegenerative conditions. It should be noted that the negative geotaxis assay and vacuolation quantification are two well-established approaches to assess sensorimotor deficits and frank brain degeneration in fly brains. 

It is important to emphasize that the head-specific impacts delivered to the flies in our study are much milder than those used in previous studies. As we showed in our figure 1, this very mild form of head trauma (referred to as vmHT) did not cause any death, nor affected the lifespan of the injured flies. Our supplemental data also show very minimal structural neuronal damage and no acute and chronic apoptosis induced by vmHT exposure. Consistently, we did not observe any exoskeletal or eye damage immediately following injuries, nor did we observe any retinal degeneration and pseudopupil loss at the chronic stage of these flies. We have incorporated these important points in the revised manuscript.  

(3) In Figure 4, it would be important to perform the behavior test fly speed and directional movement in the acute phase as well to determine whether the females have reduced performance at the acute phase.

We thank the reviewer for this suggestion. Please note that our modified NGA has already improved the spatiotemporal resolution over the classic NGA.  The data presented in Fig.3 show that there are no acute deficits for young cohorts.  Therefore, we do not believe that the detailed analysis of the direction and speed of these flies is essential.  

Unfortunately, the current setup for the AI-based analysis requires manual corrections of tracking errors, which are time-consuming and tedious.  We are building a newly designed AI-based NGA (NGA.ai) that will allow automatic tracking and quantification with minimal manual interventions. Once it is completed, we will perform some of the analyses that the reviewer suggested.  

(4) In Figure 8, the authors performed an RNA‐seq analysis and identified some dysregulated gene expressions. However, it is really surprising to see so few DEGs even in wild‐type males and mated females, and to see that none of DEGs overlap among groups or related to the SP‐signaling. This raises questions about the validity of the RNAseq analysis. It is critical to independently verify their RNA‐sequencing results and to add some more molecular evidence to support their conclusion.

We agree that future studies are needed to independently validate our RNA sequencing results. We believe that the small number of DEGs are likely due to two unique features of our study: (1) the very mild nature of our injury paradigm and (2) the chronic examination timepoint that was long after the head injury and SP exposure, which distinguish our study from previous fly TBI studies.  As pointed out in the manuscript, our study was aimed to understand how early life exposure to repetitive head traumatic insults could lead to the latelife onset of neurodegenerative conditions. We hope to further validate our results in our next phase of experiments using single-cell RNA sequencing and RT-qPCR. 

(5) The current results raise a series of interesting questions: what implication of female fly mating and its associated Sex Peptide signaling would be to mammalians or humans? Would mammalian female animals mating with wild‐type or sex hormone‐null male animals have different effects on their post‐injury behavior tests or neuropathological changes? What are the mechanisms underlying the sexual dimorphism?

As the reviewer pointed out, it would be very interesting to explore the possible roles of sex peptide-signaling in other animals and humans. As far as we know, there is no known mammalian ortholog to the insect sex peptide, so it would be difficult to study SP or an SPlike molecule in mammalian models. However, we believe that prolonged post-mating changes associated with reproduction in female fruit flies contribute to their elevated vulnerability to neurodegeneration.  In this regard, drastic changes within the biology of female mammals associated with reproduction can potentially lead to vulnerability to neurodegeneration. We agree that this demands further study, which may be done with future collaborators using rodent or large animal models.  We have discussed this point in the manuscript.

eLife assessment

These valuable findings develop a mouse model with trackable fusion Fshr protein, which will be of use to the field. The animal model helps to elucidate the expression and function of the FSH receptor in extra-gonadal tissues. The strength of the evidence is solid in most parts, although additional validation of the localization data would strengthen the study considerably.

Reviewer #1 (Public Review):

The manuscript by Hong-Qian Chen and collaborators describes the development of a mouse model that co-expresses a fluorescent protein ZsGreen marker in gene fusion with the Fshr gene.

The authors are correct in that there is a lack of reliable antibodies against many of the GPCR family members. The approach is novel and interesting, with a potential to help understand the expression pattern of gonadotropin receptors. There has been a very long debate about the expression of gonadotropin receptors in other tissues other than gonads. While their expression of the Fshr in some of those tissues has been detected by a variety of methods, their physiological, or pathophysiological, function(s) remain elusive.

The authors in this manuscript assume that the expression of ZsGren and the Fshr are equal. While this is correct genetically (transcription->translation) it does not go hand in hand to other posttranslational processes.

One of the shocking observations in this manuscript is the expression of Fshr in Leydig cells. Other observations are in the osteoblasts and endothelial cells as well as epithelial cells in different organs. The expression of ZsGreen in these tissues seems high and one shall start questioning if there are other mechanisms at play here.

First, the turnover of fluorescent proteins is long, longer than 48h, which means that they accumulate at a different speed than the endogenous Fshr. This means that ZsGreen will accumulate in time while the Fshr receptor might be degraded almost immediately. This correlated with mRNA expression (by the authors) but does not with the results of other studies in single-cell sequencing (see below).

Then, the expression of ZsGreen in Leydig cells seems much higher than in Sertoli cells, this is "disturbing" to put it mildly. This is visible in both, the ZsGreen expression and the FISH assay (Fig 2 B-D).

The expression in WAT and BAT is also questionable as the expression of ZsGreen is high everywhere. What makes it difficult to actually believe that the images are truly informative? For example, the stainings of the aorta show the ZsGreen expression where elastin and collagen fibres are - these are not "cells" and therefore are not expressing ZsGreen.

FISH expression (for Fshr) in WT mice is missing.

Also, the tissue sections were stained with the IgG only (neg control) but in practice both the KI and the WT tissues should be stained with the primary and secondary antibodies.

The only control that I could think of to truly get a sense of this would be a tagged receptor (N-terminal) that could then be analysed by immunohistochemistry.

The authors also claim:<br /> To functionally prove the presence of Fshr in osteoblasts/osteocytes, we also deleted Fshr in osteocytes using an inducible model. The conditional knockout of Fshr triggered a much more profound increase in bone mass and decrease in fat mass than blockade by Fsh antibodies (unpublished data)

This would be a good control for all their images. I think it is necessary to make the large claim of extragonadal expression, as well as intragonadal such as Leydig cells.

Claiming that the under-developed Leydig cells in FSHR KO animals is due to a direct effect of the FSHR, and not via a cross-talk between Sertoli and Leydig cells, is too much of a claim. It might be speculated to some degree but as written at the moment is suggests this is "proven".

We also do not know if this Fshr expressed is a spliced form that would also result in the expression of ZsGreen but in a non-functional Fshr, or whether the Fshr is immediately degraded after expression. The insertion of the ZsGreen might have disturbed the epigenetics, transcription or biosynthesis of the mRNA regulation.

The authors should go through single-cell data of WT mice to show the existence of the Fshr transcript(s). For example here:<br /> https://www.nature.com/articles/sdata2018192

Comments after revision:

The response by the authors does not seem sufficient or adequate, by any length, for what one would expect for a work having such a large claim as the expression of the Fshr in multiple cell types and organs. It is not the fact that Fshr might be expressed extragonadally or even by other cells in the gonads, but the surprising images where virtually every cell in the provided tissues, and not only cells but structures, glows green.

It is not possible to know, as a reviewer, whether the excitation intensity and exposure for all images is equal. We believe that they cannot be, as control organs such as fat, testes, ovaries, and vasculature have a natural fluorescence background.

Leydig cells cannot simply express more Fshr than Sertoli cells, that would go against what we have known for >50 years in physiology. While it is scientific to question 'old' data, to make extraordinary claims there is a need for "extraordinary evidence". There is very low expression in Sertoli cells (Fig 2) while Leydig cells and spermatozoa glow vividly.

Moreover, even the tails of spermatozoa glow! This is not cytoplasm and cannot contain a soluble fluorescent protein.

The controls should be shown side-by-side to the experimental images. It would be a lot more credible if the WT and the KI tissues were placed on the same slide, with images taken from them side by side not only for ZsGreen but antibody immunofluorescence staining.

Moreover, I noticed that the entire manuscript is based on a single founder mouse, which is not acceptable as an error - either multiple integrations other than in the correct locus or genetic instability created by the KI integration would result in promiscuous expression. The founder mouse is not well enough characterised as it is only performed by Southern blots and PCR, while additional integrations cannot be detected by such. Other methods should be used such as FISH or even whole genome sequencing. Yet, several lines should be used to ensure no other effects exist.

In Fig 5, the section of aorta shows low staining in the elastin/collagen fibres, while there is clearly in Suppl Data 2. In the same figure, the 2nd lung images show green fluorescence in the mucosa (centre) which should not be as there is no cells there.

The additional single-cell data does not truly support their claims, in the sense that while some of the data might go in line e.g. Leydig cells showing as high expression as "tubules", there are many other cell types that show no expression such as hepatocytes and skeletal muscle, where the authors claim to have high expression of Fshr. Moreover, in the datasets presented organs like "ovary" have almost no Fshr expression, which should question the validity of such.

The authors use an Fshr antibody without enough validation. The Fshr KO animals should be used for this. In fact, one of the very first statements in the manuscript is that antibodies against GPCRs in general, and gonadotropin receptors more specifically, are unreliable. The fact that controls show the same pattern as transgenic animals questions the validity, as no single acceptable antibody against FSHR recognises Leydig cells.

The detection of Fshr in e.g. adipocytes of B6 mice is as questionable as many other claims of gonadotropin receptors in extragonadal tissues, which has been questioned a number of times by many researchers.

One question we should ask is, is there any tissue on these mice that does not 'express' (Fshr)-ZsGreen? Because from what I see every single tissue analysed has 'Fshr". Which might be the problem why it is so difficult to find.

Some images seem to be duplicated such as in Fig 2C where the first row and the 3rd row seem to be the same image.

Author response:

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

We would like to thank you very much for reviewing our manuscript and express our sincere appreciation for the valuable and thoughtful comments that led us to significantly improve the manuscript on Fshr-ZsGreen reporter mice. We have seriously taken your comments to make a major revision of the manuscript, and here is a summary of the revision:

(1) New data on Fshr expression are input to the revised Manuscript:

a. Fshr expression in the testis and adipose tissues (WAT and BAT) of B6 mice;

b. Fshr expression in the testis of B6 by RNA-smFISH;

c. Comparison of Fshr expression in the testis and ovary between Fshr-ZsGreen and B8 mice by ddRT-PCR to prove Fshr expression without interruptions by insertion of P2A-ZsGreen vector;

d. Reduction of Fshr expression in osteocytes within the femoral sections from DMP1-CreERT2:Fshrfl/fl mice;

e. Fshr expression in an established Leydig cell line-TM3 by immunofluorescence and ddRT-PCR, also show Fshr located in the nuclei of TM3 cells;

f. Fshr expression at scRNA-seq level from 5 public single cell portals as Supplementary Data 3 to support our findings of the widespread expression pattern of Fshr, particularly in Leydig cells.

(2) Re-organization of Figure 2 with a new legend.

(3) A new paragraph is added to the Discussion Section of the revised MS to explain the function of P2A peptide in generation of GFP reporter mice and why Fshr express is not interrupted by the P2A-ZsGreen insertion in Fshr-ZsGreen reporter.

(4) Deletion of Figure 1-D-c, as it is not necessary.

(5) Replace of Figure 8-A (the left panel) with a reduced exposure time image.

(6) Amended parts of the revised MS are labeled in red.

A point by point response to the Reviewers’ comments:

Reviewer 1:

One of the shocking observations in this manuscript is the expression of FSHR in Leydig cells. Other observations are in the osteoblasts and endothelial cells as well as epithelial cells in different organs. The expression of ZsGreen in these tissues seems high and one shall start questioning if there are other mechanisms at play here.

First, the turnover of fluorescent proteins is long, longer than 48h, which means that they accumulate at a different speed than the endogenous FSHR This means that ZsGreen will accumulate in time while the FSHR receptor might be degraded almost immediately. This correlated with mRNA expression (by the authors) but does not with the results of other studies in single-cell sequencing (see below).

The expression of ZsGreen in Leydig cells seems much higher than in Sertoli cells, this is "disturbing" to put it mildly. This is visible in both the ZsGreen expression and the FISH assay (Figure 2 B-D).

Thank you for this valuable comments. We added new data on Fshr expression to prove the presence of Fshr in Leydig cells in B6 detected by immunofluorescence staining, RNA-smFISH and ddRT-PCR, as well as in TM3 cells-isolated Leydig cells from a male mice in the revise MS (Fig 2E, F and G), that demonstrate no interruptions of normal Fshr expression by insertion of P2A-ZsGreen vector into a locus located between exon10 and stop code. We use ZsGreen as an indicator for active Fshr promoter status, rather than a method to measure Fshr expression, which is done by ddRT-PCR. These data are shown in Figure 2G of the revised MS

In addition, we provide scRNA-seq based evidence on Fshr expression in human Leydig cells from two single cell portals (DISCO and BioGPS) as shown in Supplementary Data 3 in the revised MS. We also cited a recent report on scRNA-seq analysis of Fshr expression in Hu sheep in the revised MS as Reference 65 (PMID: 37541020) 1, which also clearly showed Fshr expression in Leydig cells at single cell level in Hu Sheep.

We believe that the lack of Fshr expression in some single cell databases may be due to the degradation of Fshr transcript in cells during the process of single cell populations. In our laboratory, we spent more than 6 months to optimize methods and reagents to perverse mRNA integrity more than 8 for RAN-seq.

The expression in WAT and BAT is also questionable as the expression of ZsGreen is high everywhere. That makes it difficult to believe that the images are truly informative. For example, the stainings of aorta show the ZsGreen expression where elastin and collagen fibres are - these are not "cells" and therefore are not expressing ZsGreen.

FISH expression (for FSHR) in WT mice is missing.

Also, the tissue sections were stained with the IgG only (neg control) but in practice both the KI and the WT tissues should be stained with the primary and secondary antibodies. The only control that I could think of to truly get a sense of this would be a tagged receptor (N-terminal) that could then be analysed by immunohistochemistry.

Reply 2 and 3: Thank you for these comments. New data on Fshr expression in WAT and BAT of B6 mice by immunofluorescence staining and in the testis of B6 mice by immunofluorescence staining and RNA-smFISH are added to the revised MS (Fig.2D and E, and Fig. 4G), showing similar patterns to that of Fshr-ZsGreen mice. Furthermore, we provide more evidences as Supplementary Data 3 on Fshr expression obtained from 4 public single cell portables, showing FSHR expression in a widespread organs and tissues (including different fractions of adipose cells) of human, mice and rat at single cell levels. Please also check Fshr expression pattern in adipose tissues by immunostaining for Fshr in previous reports (Fig. 3a of PMID: 28538730 and Fig. 2 of PMID: 25754247) 2 3, which showed a similar expression pattern to our finding. These data should address your concerns on Fshr expression in WAT and BAT and other organs/tissues.

Regard of “For example, the stainings of aorta show the ZsGreen expression where elastin and collagen fibres are - these are not "cells" and therefore are not expressing ZsGreen.” We believe that you referred to the image of the aorta in Supplementary Data2. However, Please take a look at the images of the aorta in Figure 5-C, which shows positively stained the layer of ‘elastin and collagen fibres’ for EMCN and a-SMA colocalized with Fshr expression with stained DAPI at a 1000X magnification, indicating endothelial cells and the cellular membrane presented in this layer, not just ‘elastin and collagen’.

The authors also claim:

To functionally prove the presence of FSHR in osteoblasts/osteocytes, we also deleted FSHR in osteocytes using an inducible model. The conditional knockout of FSHR triggered a much more profound increase in bone mass and decrease in fat mass than blockade by FSHR antibodies (unpublished data).

This would be a good control for all their images. I think it is necessary to make the large claim of extragonadal expression, as well as intragonadal such as Leydig cells.

Thank you for this very encouraging comment. As you suggested, we did add a result of reduced Fshr expression in osteocytes from DMP1-CreERT2+:Fshrfl/fl mice treated with tamoxifen to the revise MS, as shown in Figure 3D, demonstrating Fshr present in osteocytes and the specificity of Fshr antibody. Furthermore, we incorporated your advice on making ‘ large claim of extrogonadal and intragonadal expression of Fshr’ into the revised MS in red.

Claiming that the under-developed Leydig cells in FSHR KO animals are due to a direct effect of the FSHR, and not via a cross-talk between Sertoli and Leydig cells, is too much of a claim. It might be speculated to some degree but as written at the moment it suggests this is "proven".

Thank you for pointing out this incorrect claim and we apologized for it. In the revised MS, we deleted this claim.

We also do not know if this FSHR expressed is a spliced form that would also result in the expression of ZsGreen but in a non-functional FSHR, or whether the FSHR is immediately degraded after expression. The insertion of the ZsGreen might have disturbed the epigenetics, transcription, or biosynthesis of the mRNA regulation.

Thanks for this comment. In the revised MS, we added a new section to explain the function of P2A peptide in generation of a GFP reporter by sgRNA-guilded site specific knockin of P2A ZsGreen vector through CRISPRA/cas9 and provided a new result on comparison of Fshr expression in the testes and ovaries from Fshr-ZsGreen and B6 mice, showing equivalent Fshr expression between Fshr-ZsGreen and B6 mice (Figure 2G), which indicates no interruptions of Fshr expression by the insertion of P2A vector.

The authors should go through single-cell data of WT mice to show the existence of the FSHR transcript(s).<br /> For example here:<br /> https://www.nature.com/articles/sdata2018192

Thank you so much for the valuable comment. Yes, we took you critical advice to check Fshr expression through 4 single cell portals, including DISCO, GTEx, BioGPS and Human single cell portal, and present the collected data as Supplementary Data 3 in the revised MS, that strongly support our findings of the wider Fshr expression. Particularly, Fshr expression in Leydig cells is proved by scRNA-seq studies of human cells from DISCO and BioGPS, as well as a recent study in Hu sheep (PMID: 37541020) 1 and we cited it in the revised MS.

Reviewer 2:

Is the FSHR expression pattern affected by the knockin mice (no side-by-side comparison between wt and GSGreen mice, using in situ hybridization and ddRTPCR, at least in the gonads, is provided)?

Thanks for the comment. In the revised MS, we provided a set of new data on Fshr expression in the testis, ovary, WAT and BAT of B6 mice by immunofluorescence staining and by RNA-smFISH for Fshr expression, showing similar expression patterns. Additionally, we also performed ddRT-PCT to compare Fshr expression in the testes and ovaries between Fshr-ZsGreen and B6 mice, demonstrating equivalent expression of Fshr expression between Fshr-ZsGreen and B6 mice. Interestingly, we also observed an significantly higher Fshr expression in the testis than that in the ovary (more than 30 folds).

Is the splicing pattern of the FSHR affected in the knockin compared to wt mice, at least in the gonads?

Thanks for the question. Please see our reply to the Reviewer 1 for the function of P2A peptide used for generation of GFP reporters.  Although we didn’t directly assess the splicing pattern, we provide a result of comparison of Fshr expression in Figure 2F in the revised MS, indirectly showing no changes of the splicing pattern. We will assess the splicing pattern of Fshr in the future that has been neglected in the field.

Are there any additional off-target insertions of GSGreen in these mice?” and “Are similar results observed in separate founder mice?

Thanks for the questions. As we describe it in the method section  in detail in the MS, Fshr-ZsGreen reporter was produced by the a site-specific long ssDNA recombination of the P2A-ZsGreen targeting vector to the locus between Exon10 and stop code by CRIPRA/cas9, which was guided by site-specific single guide RNA (sgRNA). We showed the results of Southern blot, DNA sequencing and site-specific PCR, proving the site-specific insertion of P2A-ZsGreen as shown in Figure 1. Because of the site-specific recombination, professionally, only one funder line is required for the study and there are no additional off-target insertions.

How long is GSGreen half-life? Could a very long half-life be a major reason for the extremely large expression pattern observed?

Thanks for the question. The half life of ZsGreen, also called ZsGreen1, is at least 26 h in mammalian cells or slightly longer due to its tetrameric structure, in contrast with the monomeric configuration of other well-known fluorescent proteins (PMID: 17510373) 4. The rationale for using this GFP protein is that ZsGreen is an exceptionally bright green fluorescent protein, which is up to 4X brighter than EGFP—and is ideally suited for whole-cell labelling, promoter-reporter studies, considering of the higher turnover and rapid degradation of Fshr transcript. In this study, we used ZsGreen as a monitor or an indicator of the active Fshr endogenous promoter, rather than a means for measuring the promoter activity. Therefore, regardless of its accumulation or not, ZsGreen driven by Fshr promoter, indicates the presence of active Fshr promoter in the defined cells. In stead, we used ddRT-PCR to measure Fshr expression degrees in this study. In addition, we also provide single cell sequence-based evidence from 4 public single cell portables to support our findings of the wide Fshr expression. Please see Supplementary Data 3 in the revised MS.

References:

(1) Su J, Song Y, Yang Y, et al. Study on the changes of LHR, FSHR and AR with the development of testis cells in Hu sheep. Anim Reprod Sci. Sep 2023;256:107306. doi:10.1016/j.anireprosci.2023.107306

(2) Liu P, Ji Y, Yuen T, et al. Blocking FSH induces thermogenic adipose tissue and reduces body fat. Nature. Jun 1 2017;546(7656):107-112. doi:10.1038/nature22342

(3) Liu XM, Chan HC, Ding GL, et al. FSH regulates fat accumulation and redistribution in aging through the Galphai/Ca(2+)/CREB pathway. Aging Cell. Jun 2015;14(3):409-20. doi:10.1111/acel.12331

(4) Bell P, Vandenberghe LH, Wu D, Johnston J, Limberis M, Wilson JM. A comparative analysis of novel fluorescent proteins as reporters for gene transfer studies. J Histochem Cytochem. Sep 2007;55(9):931-9. doi:10.1369/jhc.7A7180.2007

eLife assessment

This important study identifies a novel gastrointestinal enhancer of Ctnnb1. The authors present convincing evidence to support their claim that the dosage of Wnt/β-catenin signaling controlled by this enhancer is critical to intestinal epithelia homeostasis and the progression of colorectal cancers. The study will be of interest to biomedical researchers interested in Wnt signaling, tissue-specific enhancers, intestinal homeostasis, and colon cancer.

Reviewer #1 (Public Review):

Summary:

Ctnnb1 encodes β-catenin, an essential component of the canonical Wnt signaling pathway. In this study, the authors identify an upstream enhancer of Ctnnb1 responsible for the specific expression level of β-catenin in the gastrointestinal tract. Deletion of this promoter in mice and analyses of its association with human colorectal tumors support that it controls the dosage of Wnt signaling critical to the homeostasis in intestinal epithelia and colorectal cancers.

Strengths:

This study has provided convincing evidence to demonstrate the functions of a gastrointestinal enhancer of Ctnnb1 using combined approaches of bioinformatics, genomics, in vitro cell culture models, mouse genetics, and human genetics. The results support the idea that the dosage of Wnt/β-catenin signaling plays an important role in the pathophysiological functions of intestinal epithelia. The experimental designs are solid and the data presented are of high quality. This study significantly contributes to the research fields of Wnt signaling, tissue-specific enhancers, and intestinal homeostasis.

Weaknesses:

One weakness of this manuscript is an insufficient discussion on the Ctnnb1 enhancers for different tissues. For example, do specific DNA motifs and transcriptional factors contribute to the tissue-specificity of the neocortical and gastrointestinal enhancers? It is also worth discussing the potential molecular mechanisms controlling the gastrointestinal expression of Ctnnb1 in different species since the identified human and mouse enhancers don't seem to share significant similarities in primary sequences.

Reviewer #2 (Public Review):

Wnt signaling is the name given to a cell-communication mechanism that cells employ to inform on each other's position and identity during development. In cells that receive the Wnt signal from the extracellular environment, intracellular changes are triggered that cause the stabilization and nuclear translocation of β-catenin, a protein that can turn on groups of genes referred to as Wnt targets. Typically these are genes involved in cell proliferation. Genetic mutations that affect Wnt signaling components can therefore affect tissue expansion. Loss of function of APC is a drastic example: APC is part of the β-catenin destruction complex, and in its absence, β-catenin protein is not degraded and constitutively turns on proliferation genes, causing cancers in the colon and rectum. And here lies the importance of the finding: β-catenin has for long been considered to be regulated almost exclusively by tuning its protein turnover. In this article, a new aspect is revealed: Ctnnb1, the gene encoding for β-catenin, possesses tissue-specific regulation with transcriptional enhancers in its vicinity that drive its upregulation in intestinal stem cells. The observation that there is more active β-catenin in colorectal tumors not only because the broken APC cannot degrade it, but also because transcription of the Ctnnb1 gene occurs at higher rates, is novel and potentially game-changing. As genomic regulatory regions can be targeted, one could now envision that mutational approaches aimed at dampening Ctnnb1 transcription could be a viable additional strategy to treat Wnt-driven tumors.

Reviewer #3 (Public Review):

The authors of this paper identify an enhancer upstream of the Ctnnb1 gene that selectively enhances expression in intestinal cells. This enhancer sequence drives expression of a reporter gene in the intestine and knockout of this enhancer attenuates Ctnnb1 expression in the intestine while protecting mice from intestinal cancers. The human counterpart of this enhancer sequence is functional and involved in tumorigenesis. Overall, this is an excellent example of how to fully characterize a cell-specific enhancer. The strength of the study is the thorough nature of the analysis and the relevance of the data to the development of intestinal tumors in both mice and humans. A minor weakness is that the loss of this enhancer does not completely compromise the expression of the Ctnnb1 gene in the intestine, suggesting that other elements are likely involved. Adding some discussion on that point would be helpful.

Author response:

eLife assessment 

This important study identifies a novel gastrointestinal enhancer of Ctnnb1. The authors present convincing evidence to support their claim that the dosage of Wnt/β-catenin signaling controlled by this enhancer is critical to intestinal epithelia homeostasis and the progression of colorectal cancers. The study will be of interest to biomedical researchers interested in Wnt signaling, tissue-specific enhancers, intestinal homeostasis, and colon cancer. 

We greatly appreciate editors’ and reviewers’ extensive and constructive comments and suggestions. We will do our utmost to revise the manuscript accordingly.

Public Reviews: 

Reviewer #1 (Public Review): 

Summary: 

Ctnnb1 encodes β-catenin, an essential component of the canonical Wnt signaling pathway. In this study, the authors identify an upstream enhancer of Ctnnb1 responsible for the specific expression level of β-catenin in the gastrointestinal tract. Deletion of this promoter in mice and analyses of its association with human colorectal tumors support that it controls the dosage of Wnt signaling critical to the homeostasis in intestinal epithelia and colorectal cancers. 

Strengths: 

This study has provided convincing evidence to demonstrate the functions of a gastrointestinal enhancer of Ctnnb1 using combined approaches of bioinformatics, genomics, in vitro cell culture models, mouse genetics, and human genetics. The results support the idea that the dosage of Wnt/β-catenin signaling plays an important role in the pathophysiological functions of intestinal epithelia. The experimental designs are solid and the data presented are of high quality. This study significantly contributes to the research fields of Wnt signaling, tissue-specific enhancers, and intestinal homeostasis. 

Weaknesses: 

One weakness of this manuscript is an insufficient discussion on the Ctnnb1 enhancers for different tissues. For example, do specific DNA motifs and transcriptional factors contribute to the tissue-specificity of the neocortical and gastrointestinal enhancers? It is also worth discussing the potential molecular mechanisms controlling the gastrointestinal expression of Ctnnb1 in different species since the identified human and mouse enhancers don't seem to share significant similarities in primary sequences. 

We agree with the reviewer that the manuscript lacks sufficient discussions on how enhancers control cell-type-specific expressions of target genes, which is one of the most important questions in the field of transcription regulation. Equally important are the common and species-specific features of this regulation. In general, motif composition, location, order, and affinity with trans-factors within enhancers are four key elements. We will elaborate the point in follow-up revision.

Reviewer #2 (Public Review): 

Wnt signaling is the name given to a cell-communication mechanism that cells employ to inform on each other's position and identity during development. In cells that receive the Wnt signal from the extracellular environment, intracellular changes are triggered that cause the stabilization and nuclear translocation of β-catenin, a protein that can turn on groups of genes referred to as Wnt targets. Typically these are genes involved in cell proliferation. Genetic mutations that affect Wnt signaling components can therefore affect tissue expansion. Loss of function of APC is a drastic example: APC is part of the β-catenin destruction complex, and in its absence, β-catenin protein is not degraded and constitutively turns on proliferation genes, causing cancers in the colon and rectum. And here lies the importance of the finding: β-catenin has for long been considered to be regulated almost exclusively by tuning its protein turnover. In this article, a new aspect is revealed: Ctnnb1, the gene encoding for β-catenin, possesses tissue-specific regulation with transcriptional enhancers in its vicinity that drive its upregulation in intestinal stem cells. The observation that there is more active β-catenin in colorectal tumors not only because the broken APC cannot degrade it, but also because transcription of the Ctnnb1 gene occurs at higher rates, is novel and potentially game-changing. As genomic regulatory regions can be targeted, one could now envision that mutational approaches aimed at dampening Ctnnb1 transcription could be a viable additional strategy to treat Wnt-driven tumors. 

We appreciate the reviewer for acknowledging the potential significance represented by the manuscript. We also recognize that targeting genomic regulatory regions to dampen Ctnnb1 transcription could be a promising strategy for treating Wnt-driven tumors, including many colorectal carcinomas. However, we would like to point out that three are significant technical challenges associated with AAV delivery to the GI epithelium, including the hostile environment, immune response, and low delivery efficiency.

Reviewer #3 (Public Review): 

The authors of this paper identify an enhancer upstream of the Ctnnb1 gene that selectively enhances expression in intestinal cells. This enhancer sequence drives expression of a reporter gene in the intestine and knockout of this enhancer attenuates Ctnnb1 expression in the intestine while protecting mice from intestinal cancers. The human counterpart of this enhancer sequence is functional and involved in tumorigenesis. Overall, this is an excellent example of how to fully characterize a cell-specific enhancer. The strength of the study is the thorough nature of the analysis and the relevance of the data to the development of intestinal tumors in both mice and humans. A minor weakness is that the loss of this enhancer does not completely compromise the expression of the Ctnnb1 gene in the intestine, suggesting that other elements are likely involved. Adding some discussion on that point would be helpful.

We are quite encouraged by the reviewer’s positive comments. We agree with the reviewer that other cis-regulatory elements may be involved in the transcription of Ctnnb1 within the GI epithelium. It is also possible that the basal transcription of Ctnnb1 within the GI epithelium is relatively high, and that enhancers can only boost transcription within a certain range. We will discuss these possibilities in the revision.

eLife assessment

The valuable work by authors improves our understanding on the effects of tree diversity on host-parasitoid communities of insects in forests in an experimental setting. Most of the analyses used are solid, but some of the conclusions seem a bit too strong and stretched.

eLife assessment

The manuscript presents a machine-learning method to predict protein hotspot residues. The validation is incomplete, along with the misinterpretation of the results with other current methods like FTMap.

The paper describes a program developed to identify PPI-hot spots using the free protein structure and compares it to FTMap and SPOTONE, two webservers that they consider as competitive approaches to the problem. We appreciate the effort in providing a new webserver that can be tested by the community but we continue to have major concerns:

(1) The comparison to the FTMap program is problematic. The authors misinterpret the article they refer to, i.e., Zerbe et al. "Relationship between hot spot residues and ligand binding hot spots in protein-protein interfaces" J. Chem. Inf. Model. 52, 2236-2244, (2012). FTMap identifies hot spots that bind small molecular ligands. The Zerbe et al. article shows that such hot spots tend to interact with hot spot residues on the partner protein in a protein-protein complex (emphasis on "partner"). Thus, the hot spots identified by FTMap are not the hot spots defined by the authors. In fact, because the Zerbe paper considers the partner protein in a complex, the results cannot be compared to the results of Chen et al. This difference is missed by the authors, and hence the comparison of the FTMap is invalid.

(2) Chen et al. use a number of usual features in a variety of simple machine-learning methods to identify hot spot residues. This approach has been used in the literature for more than a decade. Although the authors say that they were able to find only FTMap and SPOTONE as servers, there are dozens of papers that describe such a methodology. Some examples are given here: (Higa and Tozzi, 2009; Keskin, et al., 2005; Lise, et al., 2011; Tuncbag, et al., 2009; Xia, et al., 2010). There are certainly more papers. Thus, while the web server is a potentially useful contribution, the paper does not provide a fundamentally novel approach.

Author response:

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

eLife assessment

The manuscript presents a machine-learning method to predict protein hotspot residues. The validation is incomplete, along with the misinterpretation of the results with other current methods like FTMap.

We believe that validation is complete: The two most common techniques for testing and validating machine-learning methods are to split the dataset into either (1) a training set and a test set with a fixed ratio (e.g., 70% for training and 30% for testing) or (2) multiple subsets/folds; i.e., cross-validation. We did not employ a training set to train the model and a separate test set to evaluate its performance, as Reviewer 2 assumed. Instead, we employed cross-validation, as it helps reduce the variability in performance estimates compared to a single training/test split, and utilizes the entire dataset for training and testing, making efficient use of the limited data. Each fold was used once as a test set and the remaining folds as the training set - this process was repeated for each fold and the model's performance was measured using the F1 score. We had listed the mean validation F1 score in Table 1.

We have clarified our comparison with FTMAP  - see reply to point 1 of reviewer 1 below. 

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The paper describes a program developed to identify PPI-hot spots using the free protein structure and compares it to FTMap and SPOTONE, two webservers that they consider as competitive approaches to the problem. On the positive side, I appreciate the effort in providing a new webserver that can be tested by the community but have two major concerns as follows.

(1) The comparison to the FTMap program is wrong. The authors misinterpret the article they refer to, i.e., Zerbe et al. "Relationship between hot spot residues and ligand binding hot spots in protein-protein interfaces" J. Chem. Inf. Model. 52, 2236-2244, (2012). FTMap identifies hot spots that bind small molecular ligands. The Zerbe et al. article shows that such hot spots tend to interact with hot spot residues on the partner protein in a protein-protein complex (emphasis on "partner"). Thus, the hot spots identified by FTMap are not the hot spots defined by the authors. In fact, because the Zerbe paper considers the partner protein in a complex, the results cannot be compared to the results of Chen et al. This difference is missed by the authors, and hence the comparison of the FTMap is invalid. I did not investigate the comparison to SPOTONE, and hence have no opinion.

Brenke et al. (Bioinformatics 2009 25: 621-627), who developed FTMAP, defined hot spots as regions of the binding surface that “contribute a disproportionate amount to the binding free energy”. Kozakov et al. (Proc. Natl. Acad. Sci. 2011:108, 13528-1353) used unbound protein structures as input to FTMap to predict binding hot spots for protein-protein interactions (PPIs), which are defined as regions (so-called consensus sites) on a protein surface that bind multiple probe clusters − the main hot spot is the largest consensus site binding the largest number of probe clusters. 

Zerbe et al. (J. Chem. Inf. Model. 2012:52, 2236) noted that a consensus “site is expected to be important in any interaction that involves that region of the target independent of any partner protein.” They showed that for hot spot residues found by Ala scanning not only overlapped with the probe ligands but also form consensus sites, as shown in Figure 4. They stated that “A residue can also be identified as a hot spot by alanine scanning if it contributes to creating such a favorable binding environment by being among the residues forming a consensus site on the protein to which it belongs.”

To clarify the comparison with FTmap in the revised version, we have added the following sentence in the Abstract on p. 3:

“We explored the possibility of detecting PPI-hot spots using (i) FTMap in the PPI mode, which identifies hot spots on protein-protein interfaces from the free protein structure, and (ii) the interface residues predicted by AlphaFold-Multimer.”

We have added the following sentences in the Introduction section on p. 4:

“We explored the possibility of detecting PPI-hot spots using the FTMap server in the PPI mode, which identifies hot spots on protein-protein interfaces from free protein structures.45 These hot spots are identified by consensus sites − regions that bind multiple probe clusters.42,45,59 Such regions are deemed to be important for any interaction involving that region of the target, independent of partner protein.42 PPIhot spots were identified as residues in van der Waals (vdW) contact with probe ligands within the largest consensus site containing the most probe clusters.”

and in the Results section on p. 5:

“Given the free protein structure, PPI-HotspotID and SPOTONE53 predict PPI-hot spots based on a probability threshold (> 0.5). FTMap, in the PPI mode, detects PPIhot spots as consensus sites/regions on the protein surface that bind multiple probe clusters.59 Residues in vdW contact with probe molecules within the largest consensus site were compared with PPI-hotspotID/SPOTONE predictions.”

(2) Chen et al. use a number of usual features in a variety of simple machine-learning methods to identify hot spot residues. This approach has been used in the literature for more than a decade. Although the authors say that they were able to find only FTMap and SPOTONE as servers, there are dozens of papers that describe such a methodology. Some examples are given here: (Higa and Tozzi, 2009; Keskin, et al., 2005; Lise, et al., 2011; Tuncbag, et al., 2009; Xia, et al., 2010). There are certainly more papers. Thus, while I consider the web server as a potentially useful contribution, the paper does not provide a fundamentally novel approach.

Our paper introduces several novel elements in our approach: 

(1) Most PPI-hot spot prediction methods employ PPI-hotspots where mutations decrease protein binding free energy by > 2 kcal/mol (J. Chem. Inf. Model. 2022, 62, 1052). In contrast, our method incorporates not only PPI-hot spots with such binding free energy changes, but also those whose mutations have been curated in UniProtKB to significantly impair/disrupt PPIs. Because our method employs the largest collection of experimentally determined PPI-hot spots, it could uncover elusive PPI-hot spots not within binding interfaces, as well as potential PPI-hot spots for other protein partners (see point 3 below). 

(2) Whereas most machine-learning methods for PPI-hot spot prediction focus on features derived from (i) primary sequences or (ii) protein-protein complexes, we introduce novel features such as per-residue free energy contributions derived from unbound protein structures. We further revealed the importance of one of our novel features, namely, the gas-phase energy of the target protein relative to its unfolded state and provided the physical basis for its importance. For example, PPI-hot spots can enhance favorable enthalpic contributions to the binding free energy through hydrogen bonds or van der Waals contacts across the protein’s interface. This makes them energetically unstable in the absence of the protein’s binding partner and solvent; hence providing a rationale for the importance of the gas-phase energy of the target protein relative to its unfolded state.

(3) As a result of these novel elements, our approach, PPI-HotspotID,  could identify many true positives that were not detected by FTMap or SPOTONE (see Results and Figure 1). Previous methods generally predict residues that make multiple contacts across the proteinprotein interface as PPI-hot spots. In contrast, PPI-HotspotID can detect not only PPI-hot spots that make multiple contacts across the protein-protein interface, but also those lacking direct contact with the partner protein (see Discussion).

(4) Unlike most machine-learning methods which require feature customization, data preprocessing, and model optimization, our use of AutoGluon’s AutoTabular module automates data preprocessing, model selection, hyperparameter optimization, and model evaluation. This automation reduces the need for manual intervention.

We have revised and added the following sentences on p. 9 in the Discussion section to highlight the novelty of our approach: 

“Here, we have introduced two novel elements that have helped to identify PPI-hot spots using the unbound structure. First, we have constructed a dataset comprising 414 experimentally known PPI-hot spots and 504 nonhot spots, and carefully checked that PPI-hot spots have no mutations resulting in ΔΔGbind < 0.5 kcal/mol, whereas nonhot spots have no mutations resulting in ΔΔGbind ≥ 0.5 kcal/mol or impact binding in immunoprecipitation or GST pull-down assays (see Methods). In contrast, SPOTONE53 employed nonhot spots defined as residues that upon alanine mutation resulted in ΔΔGbind < 2.0 kcal/mol. Notably, previous PPI-hot spot prediction methods did not employ PPIhot spots whose mutations have been curated to significantly impair/disrupt PPIs in UniProtKB (see Introduction). Second, we have introduced novel features derived from unbound protein structures such as the gas-phase energy of the target protein relative to its unfolded state.”

Strengths:

A new web server was developed for detecting protein-protein interaction hot spots.

Weaknesses:

The comparison to FTMap results is wrong. The method is not novel.

See reply to points 1 and 2 above.

Reviewer #2 (Public Review):

Summary:

The paper presents PPI-hotspot a method to predict PPI-hotspots. Overall, it could be useful but serious concerns about the validation and benchmarking of the methodology make it difficult to predict its reliability.

Strengths:

Develops an extended benchmark of hot-spots.

Weaknesses:

(1) Novelty seems to be just in the extended training set. Features and approaches have been used before.

The novelty of our approach extends beyond just the expanded training set, as summarized in our reply to Reviewer #1, point 2 above. To our knowledge, previous studies did not leverage the gas-phase energy of the target protein relative to its unfolded state for detecting PPI-hot spots from unbound structures. Previous studies did not automate the training and validation process. In contrast, we used AutoGluon’s AutoTabular module to automate the training  of (i) individual “base” models, including LightGBM, CatBoost, XGBoost, random forests, extremely randomized trees, neural networks, and K-nearest neighbours, then (ii) multiple “stacker” models. The predictions of multiple “stacker” models were fed as inputs to additional higher layer stacker models in an iterative process called multi-layer stacking. The output layer used ensemble selection to aggregate the predictions of the stacker models. To improve stacking performance, AutoGluon used all the data for both training and validation through repeated k-fold bagging of all models at all layers of the stack, where k is determined by best precision. This comprehensive approach, including repeated k-fold bagging of all models at all layers of the stack, sets our methodology apart from previous studies, including SPOTONE (see Methods). 

(2) As far as I can tell the training and testing sets are the same. If I am correct, it is a fatal flaw.

The two most common techniques for testing and validating machine-learning methods are to split the dataset into either (1) a training set and a test set with a fixed ratio (e.g., 70% for training and 30% for testing) or (2) multiple subsets/folds; i.e., cross-validation. We did not employ a training set to train the model and a separate test set to evaluate its performance. Instead, we employed cross-validation, where the model was trained and evaluated multiple times. Each fold was used once as a test set and the remaining folds serve as the training set - this process was repeated for each fold. For each test set, we assessed  the model's performance using the F1 score. We had listed the mean validation F1 score in Table 1 in the original manuscript. Cross-validation helps reduce the variability in performance estimates compared to a single training/test split. It also utilizes the entire dataset for training and testing, making efficient use of the limited data. We have clarified this on p. 14 in the revised version:

“AutoGluon was chosen for model training and validation due to its robustness and userfriendly interface, allowing for the simultaneous and automated exploration of various machine-learning approaches and their combinations. Instead of using a single training set to train the model and a separate test set to evaluate its performance, we employed cross-validation, as it utilizes the entire dataset for both training and testing, making efficient use of the limited data on PPI-hot spots and PPI-nonhot spots. AutoGluonTabular automatically chose a random partitioning of our dataset into multiple subsets/folds for training and validation. Notably, the training and validation data share insignificant homology, as the average pairwise sequence identity in our dataset is 26%. Each fold was used once as a test set, while the remaining folds served as the training set. For each test set, the model's performance was measured using the F1 score.”

(3) Comparisons should state that: SPOTONE is a sequence (only) based ML method that uses similar features but is trained on a smaller dataset. FTmap I think predicts binding sites, I don't understand how it can be compared with hot spots. Suggesting superiority by comparing with these methods is an overreach.

In the Introduction on page 3, we had already stated that:

“SPOTONE53 predicts PPI-hot spots from the protein sequence using residue-specific features such as atom type, amino acid (aa) properties, secondary structure propensity, and mass-associated values to train an ensemble of extremely randomized trees. The PPIhot spot prediction methods have mostly been trained, validated, and tested on data from the Alanine Scanning Energetics database (ASEdb)55 and/or the Structural Kinetic and Energetic database of Mutant Protein Interactions (SKEMPI) 2.0 database.56”

On p. 4, we have clarified how we used FTMAP to detect hot spots - see reply to Reviewer #1, point 1. 

“We explored the possibility of detecting PPI-hot spots using the FTMap server in the PPI mode, which identifies hot spots on protein-protein interfaces from free protein structures.45 These hot spots are identified by consensus sites − regions that bind multiple probe clusters.42,45,59 Such regions are deemed to be important for any interaction involving that region of the target, independent of partner protein.42 PPI-hot spots were identified as residues in van der Waals (vdW) contact with probe ligands within the largest consensus site containing the most probe clusters.”

(4) Training in the same dataset as SPOTONE, and then comparing results in targets without structure could be valuable.

We think that the dataset used by SPOTONE is not as “clean” as ours since SPOTONE employed nonhot spots defined as aa residues that upon alanine mutation resulted in ΔΔGbind < 2.0 kcal/mol.  In contrast, we define nonhot spots as residues whose mutations resulted in protein  ΔΔGbind changes < 0.5 kcal/mol. Moreover, we carefully checked that the nonhot spots have no mutations resulting in ΔΔGbind changes ≥ 0.5 kcal/mol or impact binding in immunoprecipitation or GST pull-down assays (see Methods). We cannot compare results in targets without structure because we require the free protein structure to compute the perresidue free energy contributions. 

(5) The paper presents as validation of the prediction and experimental validation of hotspots in human eEF2. Several predictions were made but only one was confirmed, what was the overall success rate of this exercise?

We did not test all predicted PPI-hot spots but only the PPI-hot spot with the highest probability of 0.67 (F794) and 7 other predicted PPI-hot spots that were > 12 Å from F794 as well as 4 predicted PPI-nonhot spots. Among the 13 predictions tested, F794 and the 4 predicted nonhot spots were confirmed to be correct. 

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Remove the comparison to FTMap, and find a more appropriate reference method, even if it requires installing programs rather than using the available web servers.

We have clarified comparison to FTMap in the revised ms - see our reply above.

eLife assessment

This useful study examines the neural activity in the motor cortex as a monkey reaches to intercept moving targets, focusing on how tuned single neurons contribute to an interesting overall population geometry. The presented results and analyses are solid, though the investigation of this novel task could be strengthened by clarifying the assumptions behind the single neuron analyses, and further analyses of the neural population activity and its relation to different features of behaviour.

Reviewer #1 (Public Review):

Summary:

This study addresses the question of how task-relevant sensory information affects activity in the motor cortex. The authors use various approaches to address this question, looking at single units and population activity. They find that there are three subtypes of modulation by sensory information at the single unit level. Population analyses reveal that sensory information affects the neural activity orthogonally to motor output. The authors then compare both single unit and population activity to computational models to investigate how encoding of sensory information at the single unit level is coordinated in a network. They find that an RNN that displays similar orbital dynamics and sensory modulation to the motor cortex also contains nodes that are modulated similarly to the three subtypes identified by the single unit analysis.

Strengths:

The strengths of this study lie in the population analyses and the approach of comparing single-unit encoding to population dynamics. In particular, the analysis in Figure 3 is very elegant and informative about the effect of sensory information on motor cortical activity. The task is also well designed to suit the questions being asked and well controlled.

It is commendable that the authors compare single units to population modulation. The addition of the RNN model and perturbations strengthen the conclusion that the subtypes of individual units all contribute to the population dynamics. However, the subtypes (PD shift, gain, and addition) are not sufficiently justified. The authors also do not address that single units exhibit mixed modulation, but RNN units are not treated as such.

Weaknesses:

The main weaknesses of the study lie in the categorization of the single units into PD shift, gain, and addition types. The single units exhibit clear mixed selectivity, as the authors highlight. Therefore, the subsequent analyses looking only at the individual classes in the RNN are a little limited. Another weakness of the paper is that the choice of windows for analyses is not properly justified and the dependence of the results on the time windows chosen for single-unit analyses is not assessed. This is particularly pertinent because tuning curves are known to rotate during movements (Sergio et al. 2005 Journal of Neurophysiology).

This paper shows sensory information can affect motor cortical activity whilst not affecting motor output. However, it is not the first to do so and fails to cite other papers that have investigated sensory modulation of the motor cortex (Stavinksy et al. 2017 Neuron, Pruszynski et al. 2011 Nature, Omrani et al. 2016 eLife). These studies should be mentioned in the Introduction to capture better the context around the present study. It would also be beneficial to add a discussion of how the results compare to the findings from these other works.

This study also uses insights from single-unit analysis to inform mechanistic models of these population dynamics, which is a powerful approach, but is dependent on the validity of the single-cell analysis, which I have expanded on below.

I have clarified some of the areas that would benefit from further analysis below:

(1) Task:<br /> The task is well designed, although it would have benefited from perhaps one more target speed (for each direction). One monkey appears to have experienced one more target speed than the others (seen in Figure 3C). It would have been nice to have this data for all monkeys.

(2) Single unit analyses:<br /> In some analyses, the effects of target speed look more driven by target movement direction (e.g. Figures 1D and E). To confirm target speed is the main modulator, it would be good to compare how much more variance is explained by models including speed rather than just direction. More target speeds may have been helpful here too.

The choice of the three categories (PD shift, gain addition) is not completely justified in a satisfactory way. It would be nice to see whether these three main categories are confirmed by unsupervised methods.

The decoder analyses in Figure 2 provide evidence that target speed modulation may change over the trial. Therefore, it is important to see how the window considered for the firing rate in Figure 1 (currently 100ms pre - 100ms post movement onset) affects the results.

(3) Decoder:<br /> One feature of the task is that the reach endpoints tile the entire perimeter of the target circle (Figure 1B). However, this feature is not exploited for much of the single-unit analyses. This is most notable in Figure 2, where the use of a SVM limits the decoding to discrete values (the endpoints are divided into 8 categories). Using continuous decoding of hand kinematics would be more appropriate for this task.

(4) RNN:<br /> Mixed selectivity is not analysed in the RNN, which would help to compare the model to the real data where mixed selectivity is common. Furthermore, it would be informative to compare the neural data to the RNN activity using canonical correlation or Procrustes analyses. These would help validate the claim of similarity between RNN and neural dynamics, rather than allowing comparisons to be dominated by geometric similarities that may be features of the task. There is also an absence of alternate models to compare the perturbation model results to.

Reviewer #2 (Public Review):

Summary:

In this manuscript, Zhang et al. examine neural activity in the motor cortex as monkeys make reaches in a novel target interception task. Zhang et al. begin by examining the single neuron tuning properties across different moving target conditions, finding several classes of neurons: those that shift their preferred direction, those that change their modulation gain, and those that shift their baseline firing rates. The authors go on to find an interesting, tilted ring structure of the neural population activity, depending on the target speed, and find that (1) the reach direction has consistent positioning around the ring, and (2) the tilt of the ring is highly predictive of the target movement speed. The authors then model the neural activity with a single neuron representational model and a recurrent neural network model, concluding that this population structure requires a mixture of the three types of single neurons described at the beginning of the manuscript.

Strengths:

I find the task the authors present here to be novel and exciting. It slots nicely into an overall trend to break away from a simple reach-to-static-target task to better characterize the breadth of how the motor cortex generates movements. I also appreciate the movement from single neuron characterization to population activity exploration, which generally serves to anchor the results and make them concrete. Further, the orbital ring structure of population activity is fascinating, and the modeling work at the end serves as a useful baseline control to see how it might arise.

Weaknesses:

While I find the behavioral task presented here to be excitingly novel, I find the presented analyses and results to be far less interesting than they could be. Key to this, I think, is that the authors are examining this task and related neural activity primarily with a single-neuron representational lens. This would be fine as an initial analysis since the population activity is of course composed of individual neurons, but the field seems to have largely moved towards a more abstract "computation through dynamics" framework that has, in the last several years, provided much more understanding of motor control than the representational framework has. As the manuscript stands now, I'm not entirely sure what interpretation to take away from the representational conclusions the authors made (i.e. the fact that the orbital population geometry arises from a mixture of different tuning types). As such, by the end of the manuscript, I'm not sure I understand any better how the motor cortex or its neural geometry might be contributing to the execution of this novel task.

Main Comments:

My main suggestions to the authors revolve around bringing in the computation through a dynamics framework to strengthen their population results. The authors cite the Vyas et al. review paper on the subject, so I believe they are aware of this framework. I have three suggestions for improving or adding to the population results:

(1) Examination of delay period activity: one of the most interesting aspects of the task was the fact that the monkey had a random-length delay period before he could move to intercept the target. Presumably, the monkey had to prepare to intercept at any time between 400 and 800 ms, which means that there may be some interesting preparatory activity dynamics during this period. For example, after 400ms, does the preparatory activity rotate with the target such that once the go cue happens, the correct interception can be executed? There is some analysis of the delay period population activity in the supplement, but it doesn't quite get at the question of how the interception movement is prepared. This is perhaps the most interesting question that can be asked with this experiment, and it's one that I think may be quite novel for the field--it is a shame that it isn't discussed.

(2) Supervised examination of population structure via potent and null spaces: simply examining the first three principal components revealed an orbital structure, with a seemingly conserved motor output space and a dimension orthogonal to it that relates to the visual input. However, the authors don't push this insight any further. One way to do that would be to find the "potent space" of motor cortical activity by regression to the arm movement and examine how the tilted rings look in that space (this is actually fairly easy to see in the reach direction components of the dPCA plot in the supplement--the rings will be highly aligned in this space). Presumably, then, the null space should contain information about the target movement. dPCA shows that there's not a single dimension that clearly delineates target speed, but the ring tilt is likely evident if the authors look at the highest variance neural dimension orthogonal to the potent space (the "null space")--this is akin to PC3 in the current figures, but it would be nice to see what comes out when you look in the data for it.

(3) RNN perturbations: as it's currently written, the RNN modeling has promise, but the perturbations performed don't provide me with much insight. I think this is because the authors are trying to use the RNN to interpret the single neuron tuning, but it's unclear to me what was learned from perturbing the connectivity between what seems to me almost arbitrary groups of neurons (especially considering that 43% of nodes were unclassifiable). It seems to me that a better perturbation might be to move the neural state before the movement onset to see how it changes the output. For example, the authors could move the neural state from one tilted ring to another to see if the virtual hand then reaches a completely different (yet predictable) target. Moreover, if the authors can more clearly characterize the preparatory movement, perhaps perturbations in the delay period would provide even more insight into how the interception might be prepared.

Reviewer #3 (Public Review):

Summary:

This experimental study investigates the influence of sensory information on neural population activity in M1 during a delayed reaching task. In the experiment, monkeys are trained to perform a delayed interception reach task, in which the goal is to intercept a potentially moving target.

This paradigm allows the authors to investigate how, given a fixed reach endpoint (which is assumed to correspond to a fixed motor output), the sensory information regarding the target motion is encoded in neural activity.

At the level of single neurons, the authors found that target motion modulates the activity in three main ways: gain modulation (scaling of the neural activity depending on the target direction), shift (shift of the preferred direction of neurons tuned to reach direction), or addition (offset to the neural activity).

At the level of the neural population, target motion information was largely encoded along the 3rd PC of the neural activity, leading to a tilt of the manifold along which reach direction was encoded that was proportional to the target speed. The tilt of the neural manifold was found to be largely driven by the variation of activity of the population of gain-modulated neurons.<br /> Finally, the authors studied the behaviour of an RNN trained to generate the correct hand velocity given the sensory input and reach direction. The RNN units were found to similarly exhibit mixed selectivity to the sensory information, and the geometry of the « neural population » resembled that observed in the monkeys.

Strengths:

- The experiment is well set up to address the question of how sensory information that is directly relevant to the behaviour but does not lead to a direct change in behavioural output modulates motor cortical activity.

- The finding that sensory information modulates the neural activity in M1 during motor preparation and execution is non trivial, given that this modulation of the activity must occur in the nullspace of the movement.

- The paper gives a complete picture of the effect of the target motion on neural activity, by including analyses at the single neuron level as well as at the population level. Additionally, the authors link those two levels of representation by highlighting how gain modulation contributes to shaping the population representation.

Weaknesses:

- One of the main premises of the paper is the fact that the motor output for a given reach point is preserved across different target motions. However, as the authors briefly mention in the conclusion, they did not record muscle activity during the task, but only hand velocity, making it impossible to directly verify how preserved muscle patterns were across movements. While the authors highlight that they did not see any difference in their results when resampling the data to control for similar hand velocities across conditions, this seems like an important potential caveat of the paper whose implications should be discussed further or highlighted earlier in the paper.

- The main takeaway of the RNN analysis is not fully clear. The authors find that an RNN trained given a sensory input representing a moving target displays modulation to target motion that resembles what is seen in real data. This is interesting, but the authors do not dissect why this representation arises, and how robust it is to various task design choices. For instance, it appears that the network should be able to solve the task using only the motion intention input, which contains the reach endpoint information. If the target motion input is not used for the task, it is not obvious why the RNN units would be modulated by this input (especially as this modulation must lie in the nullspace of the movement hand velocity if the velocity depends only on the reach endpoint). It would thus be important to see alternative models compared to true neural activity, in addition to the model currently included in the paper. Besides, for the model in the paper, it would therefore be interesting to study further how the details of the network setup (eg initial spectral radius of the connectivity, weight regularization, or using only the target position input) affect the modulation by the motion input, as well as the trained population geometry and the relative ratios of modulated cells after training.

- Additionally, it is unclear what insights are gained from the perturbations to the network connectivity the authors perform, as it is generally expected that modulating the connectivity will degrade task performance and the geometry of the responses. If the authors wish the make claims about the role of the subpopulations, it could be interesting to test whether similar connectivity patterns develop in networks that are not initialized with an all-to-all random connectivity or to use ablation experiments to investigate whether the presence of multiple types of modulations confers any sort of robustness to the network.

- The results suggest that the observed changes in motor cortical activity with target velocity result from M1 activity receiving an input that encodes the velocity information. This also appears to be the assumption in the RNN model. However, even though the input shown to the animal during preparation is indeed a continuously moving target, it appears that the only relevant quantity to the actual movement is the final endpoint of the reach. While this would have to be a function of the target velocity, one could imagine that the computation of where the monkeys should reach might be performed upstream of the motor cortex, in which case the actual target velocity would become irrelevant to the final motor output. This makes the results of the paper very interesting, but it would be nice if the authors could discuss further when one might expect to see modulation by sensory information that does not directly affect motor output in M1, and where those inputs may come from. It may also be interesting to discuss how the findings relate to previous work that has found behaviourally irrelevant information is being filtered out from M1 (for instance, Russo et al, Neuron 2020 found that in monkeys performing a cycling task, context can be decoded from SMA but not from M1, and Wang et al, Nature Communications 2019 found that perceptual information could not be decoded from PMd)?

Author response:

eLife assessment

This useful study examines the neural activity in the motor cortex as a monkey reaches to intercept moving targets, focusing on how tuned single neurons contribute to an interesting overall population geometry. The presented results and analyses are solid, though the investigation of this novel task could be strengthened by clarifying the assumptions behind the single neuron analyses, and further analyses of the neural population activity and its relation to different features of behaviour.

Thanks for recognizing the content of our research, and please stay tuned for our follow-up studies on neural dynamics during interception.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study addresses the question of how task-relevant sensory information affects activity in the motor cortex. The authors use various approaches to address this question, looking at single units and population activity. They find that there are three subtypes of modulation by sensory information at the single unit level. Population analyses reveal that sensory information affects the neural activity orthogonally to motor output. The authors then compare both single unit and population activity to computational models to investigate how encoding of sensory information at the single unit level is coordinated in a network. They find that an RNN that displays similar orbital dynamics and sensory modulation to the motor cortex also contains nodes that are modulated similarly to the three subtypes identified by the single unit analysis.

Strengths:

The strengths of this study lie in the population analyses and the approach of comparing single-unit encoding to population dynamics. In particular, the analysis in Figure 3 is very elegant and informative about the effect of sensory information on motor cortical activity. The task is also well designed to suit the questions being asked and well controlled.

We appreciate these kind comments.

It is commendable that the authors compare single units to population modulation. The addition of the RNN model and perturbations strengthen the conclusion that the subtypes of individual units all contribute to the population dynamics. However, the subtypes (PD shift, gain, and addition) are not sufficiently justified. The authors also do not address that single units exhibit mixed modulation, but RNN units are not treated as such.

We’re sorry for not providing sufficient grounds to introduce the subtypes. We determined the PD shift, gain, and addition as pertinent subtypes based on classical cosine tuning model (Georgopoulos et al., 1982) and referred to some gain modulation studies (e.g. Pesaran et al. 2010, Bremner and Andersen, 2012). Here, we applied this subtype analysis as a criteria to identify the modulation in neuronal population rather than to sort neuron into distinct cell types. We will update Methods in the revised version of manuscript.

Weaknesses:

The main weaknesses of the study lie in the categorization of the single units into PD shift, gain, and addition types. The single units exhibit clear mixed selectivity, as the authors highlight. Therefore, the subsequent analyses looking only at the individual classes in the RNN are a little limited. Another weakness of the paper is that the choice of windows for analyses is not properly justified and the dependence of the results on the time windows chosen for single-unit analyses is not assessed. This is particularly pertinent because tuning curves are known to rotate during movements (Sergio et al. 2005 Journal of Neurophysiology).

The mixed selectivity or precisely the mixed modulation is indeed a significant feature of neuronal population in the present study. The purpose of the subtype analysis was to serve as a criterion for the potential modulation mechanisms. However, the results appear to be a spectrum than clusters. It still through some insights to understand the modulation distribution and we will refine the description in the next version. In the current version, we observed single-unit tuning and population neural state with sliding windows, focusing on the period around movement onset (MO) due to the emergence of a ring-like structure. We will clarify the choice of windows and the dependence assessment in the next version. It’s a great suggestion to consider the role of rotating tuning curves in neural dynamics during interception.

This paper shows sensory information can affect motor cortical activity whilst not affecting motor output. However, it is not the first to do so and fails to cite other papers that have investigated sensory modulation of the motor cortex (Stavinksy et al. 2017 Neuron, Pruszynski et al. 2011 Nature, Omrani et al. 2016 eLife). These studies should be mentioned in the Introduction to capture better the context around the present study. It would also be beneficial to add a discussion of how the results compare to the findings from these other works.

Thanks for the reminder. We will introduce the relevant research in the next version of manuscript.

This study also uses insights from single-unit analysis to inform mechanistic models of these population dynamics, which is a powerful approach, but is dependent on the validity of the single-cell analysis, which I have expanded on below.

I have clarified some of the areas that would benefit from further analysis below:

(1) Task:

The task is well designed, although it would have benefited from perhaps one more target speed (for each direction). One monkey appears to have experienced one more target speed than the others (seen in Figure 3C). It would have been nice to have this data for all monkeys.

Great suggestion! However, it’s hard to implement as the implanted arrays have been removed.

(2) Single unit analyses:

In some analyses, the effects of target speed look more driven by target movement direction (e.g. Figures 1D and E). To confirm target speed is the main modulator, it would be good to compare how much more variance is explained by models including speed rather than just direction. More target speeds may have been helpful here too.

Nice suggestion! The fitting goodness of the simple model (just motor direction) is much less than the complex model (including target speed). We will update the results in the next version.

The choice of the three categories (PD shift, gain addition) is not completely justified in a satisfactory way. It would be nice to see whether these three main categories are confirmed by unsupervised methods.

A good point. We will have a try with unsupervised methods. 

The decoder analyses in Figure 2 provide evidence that target speed modulation may change over the trial. Therefore, it is important to see how the window considered for the firing rate in Figure 1 (currently 100ms pre - 100ms post movement onset) affects the results.

Thanks for the suggestion and close reading. We will test the decoder in other epochs.

(3) Decoder:

One feature of the task is that the reach endpoints tile the entire perimeter of the target circle (Figure 1B). However, this feature is not exploited for much of the single-unit analyses. This is most notable in Figure 2, where the use of a SVM limits the decoding to discrete values (the endpoints are divided into 8 categories). Using continuous decoding of hand kinematics would be more appropriate for this task.

This is a very reasonable suggestion. In this study, we discrete the reach-direction as the previous studies (Li et al., 2018&2022) and thought that the discrete decoding was already enough to show the interaction of sensory and motor variables. In future studies, we will try continuous decoding of hand kinematics.

(4) RNN:

Mixed selectivity is not analysed in the RNN, which would help to compare the model to the real data where mixed selectivity is common. Furthermore, it would be informative to compare the neural data to the RNN activity using canonical correlation or Procrustes analyses. These would help validate the claim of similarity between RNN and neural dynamics, rather than allowing comparisons to be dominated by geometric similarities that may be features of the task. There is also an absence of alternate models to compare the perturbation model results to.

Thank you for these helpful suggestions. We will perform decoding analysis on RNN units to verify if there is interaction of sensory and motor variables as in real data, as well as the canonical correlation or Procrustes analysis.

Reviewer #2 (Public Review):

Summary:

In this manuscript, Zhang et al. examine neural activity in the motor cortex as monkeys make reaches in a novel target interception task. Zhang et al. begin by examining the single neuron tuning properties across different moving target conditions, finding several classes of neurons: those that shift their preferred direction, those that change their modulation gain, and those that shift their baseline firing rates. The authors go on to find an interesting, tilted ring structure of the neural population activity, depending on the target speed, and find that (1) the reach direction has consistent positioning around the ring, and (2) the tilt of the ring is highly predictive of the target movement speed. The authors then model the neural activity with a single neuron representational model and a recurrent neural network model, concluding that this population structure requires a mixture of the three types of single neurons described at the beginning of the manuscript.

Strengths:

I find the task the authors present here to be novel and exciting. It slots nicely into an overall trend to break away from a simple reach-to-static-target task to better characterize the breadth of how the motor cortex generates movements. I also appreciate the movement from single neuron characterization to population activity exploration, which generally serves to anchor the results and make them concrete. Further, the orbital ring structure of population activity is fascinating, and the modeling work at the end serves as a useful baseline control to see how it might arise.

Thank you for recognizing our work.

Weaknesses:

While I find the behavioral task presented here to be excitingly novel, I find the presented analyses and results to be far less interesting than they could be. Key to this, I think, is that the authors are examining this task and related neural activity primarily with a single-neuron representational lens. This would be fine as an initial analysis since the population activity is of course composed of individual neurons, but the field seems to have largely moved towards a more abstract "computation through dynamics" framework that has, in the last several years, provided much more understanding of motor control than the representational framework has. As the manuscript stands now, I'm not entirely sure what interpretation to take away from the representational conclusions the authors made (i.e. the fact that the orbital population geometry arises from a mixture of different tuning types). As such, by the end of the manuscript, I'm not sure I understand any better how the motor cortex or its neural geometry might be contributing to the execution of this novel task.

The present study shows the sensory modulation on motor tuning in single units and neural state during motor execution period. It’s a pity that the findings were constrained in certain time windows. We are still working this topic, and hopefully will address related questions in our follow-up studies.

Main Comments:

My main suggestions to the authors revolve around bringing in the computation through a dynamics framework to strengthen their population results. The authors cite the Vyas et al. review paper on the subject, so I believe they are aware of this framework. I have three suggestions for improving or adding to the population results:

(1) Examination of delay period activity: one of the most interesting aspects of the task was the fact that the monkey had a random-length delay period before he could move to intercept the target. Presumably, the monkey had to prepare to intercept at any time between 400 and 800 ms, which means that there may be some interesting preparatory activity dynamics during this period. For example, after 400ms, does the preparatory activity rotate with the target such that once the go cue happens, the correct interception can be executed? There is some analysis of the delay period population activity in the supplement, but it doesn't quite get at the question of how the interception movement is prepared. This is perhaps the most interesting question that can be asked with this experiment, and it's one that I think may be quite novel for the field--it is a shame that it isn't discussed.

Great idea! We are on the way, and close to complete the puzzle.

(2) Supervised examination of population structure via potent and null spaces: simply examining the first three principal components revealed an orbital structure, with a seemingly conserved motor output space and a dimension orthogonal to it that relates to the visual input. However, the authors don't push this insight any further. One way to do that would be to find the "potent space" of motor cortical activity by regression to the arm movement and examine how the tilted rings look in that space (this is actually fairly easy to see in the reach direction components of the dPCA plot in the supplement--the rings will be highly aligned in this space). Presumably, then, the null space should contain information about the target movement. dPCA shows that there's not a single dimension that clearly delineates target speed, but the ring tilt is likely evident if the authors look at the highest variance neural dimension orthogonal to the potent space (the "null space")--this is akin to PC3 in the current figures, but it would be nice to see what comes out when you look in the data for it.

Nice suggestion. Target-speed modulation mainly influences PC3, which is consistent with ‘null space’ hypothesis. We will try other methods of dimensionality reduction (e.g. dPCA, Manopt) to determine the potent and null space.

(3) RNN perturbations: as it's currently written, the RNN modeling has promise, but the perturbations performed don't provide me with much insight. I think this is because the authors are trying to use the RNN to interpret the single neuron tuning, but it's unclear to me what was learned from perturbing the connectivity between what seems to me almost arbitrary groups of neurons (especially considering that 43% of nodes were unclassifiable). It seems to me that a better perturbation might be to move the neural state before the movement onset to see how it changes the output. For example, the authors could move the neural state from one tilted ring to another to see if the virtual hand then reaches a completely different (yet predictable) target. Moreover, if the authors can more clearly characterize the preparatory movement, perhaps perturbations in the delay period would provide even more insight into how the interception might be prepared.

We are sorry that we didn’t clarify the definition of “none” type, which can be misleading. The 43% unclassified nodes include those inactive ones, when only activate (task-related) nodes included, the ratio of unclassified nodes would be much lower. By perturbing the connectivity, we intended to explore the interaction between different modulations.

Thank you for the great advice. We tried moving neural states from one ring to another without changing the directional cluster, but this perturbation didn’t have a significant influence on network performance as expected. We will check this result again and try perturbations in the delay period.

Reviewer #3 (Public Review):

Summary:

This experimental study investigates the influence of sensory information on neural population activity in M1 during a delayed reaching task. In the experiment, monkeys are trained to perform a delayed interception reach task, in which the goal is to intercept a potentially moving target.

This paradigm allows the authors to investigate how, given a fixed reach endpoint (which is assumed to correspond to a fixed motor output), the sensory information regarding the target motion is encoded in neural activity.

At the level of single neurons, the authors found that target motion modulates the activity in three main ways: gain modulation (scaling of the neural activity depending on the target direction), shift (shift of the preferred direction of neurons tuned to reach direction), or addition (offset to the neural activity).

At the level of the neural population, target motion information was largely encoded along the 3rd PC of the neural activity, leading to a tilt of the manifold along which reach direction was encoded that was proportional to the target speed. The tilt of the neural manifold was found to be largely driven by the variation of activity of the population of gain-modulated neurons.

Finally, the authors studied the behaviour of an RNN trained to generate the correct hand velocity given the sensory input and reach direction. The RNN units were found to similarly exhibit mixed selectivity to the sensory information, and the geometry of the « neural population » resembled that observed in the monkeys.

Strengths:

- The experiment is well set up to address the question of how sensory information that is directly relevant to the behaviour but does not lead to a direct change in behavioural output modulates motor cortical activity.

- The finding that sensory information modulates the neural activity in M1 during motor preparation and execution is non trivial, given that this modulation of the activity must occur in the nullspace of the movement.

- The paper gives a complete picture of the effect of the target motion on neural activity, by including analyses at the single neuron level as well as at the population level. Additionally, the authors link those two levels of representation by highlighting how gain modulation contributes to shaping the population representation.

Thanks for your recognition.

Weaknesses:

- One of the main premises of the paper is the fact that the motor output for a given reach point is preserved across different target motions. However, as the authors briefly mention in the conclusion, they did not record muscle activity during the task, but only hand velocity, making it impossible to directly verify how preserved muscle patterns were across movements. While the authors highlight that they did not see any difference in their results when resampling the data to control for similar hand velocities across conditions, this seems like an important potential caveat of the paper whose implications should be discussed further or highlighted earlier in the paper.

Thanks for the suggestion. We will highlight the resampling results as important control in the next version of manuscript.

- The main takeaway of the RNN analysis is not fully clear. The authors find that an RNN trained given a sensory input representing a moving target displays modulation to target motion that resembles what is seen in real data. This is interesting, but the authors do not dissect why this representation arises, and how robust it is to various task design choices. For instance, it appears that the network should be able to solve the task using only the motion intention input, which contains the reach endpoint information. If the target motion input is not used for the task, it is not obvious why the RNN units would be modulated by this input (especially as this modulation must lie in the nullspace of the movement hand velocity if the velocity depends only on the reach endpoint). It would thus be important to see alternative models compared to true neural activity, in addition to the model currently included in the paper. Besides, for the model in the paper, it would therefore be interesting to study further how the details of the network setup (eg initial spectral radius of the connectivity, weight regularization, or using only the target position input) affect the modulation by the motion input, as well as the trained population geometry and the relative ratios of modulated cells after training.

Great suggestions. It’s a considerable pity that we didn’t dissect the formation reason and influence factor of the representation in the current version. We’ve tried several combinations of inputs before: in the network which received only motor intention and GO inputs, there were rings but not tilting related to target-speed; in the network which received only target location and GO inputs, there were ring-like structures but not clear directional clusters. We will check these results and try alternative models in the next version. In future studies, we will examine the influence of network setup details.

- Additionally, it is unclear what insights are gained from the perturbations to the network connectivity the authors perform, as it is generally expected that modulating the connectivity will degrade task performance and the geometry of the responses. If the authors wish the make claims about the role of the subpopulations, it could be interesting to test whether similar connectivity patterns develop in networks that are not initialized with an all-to-all random connectivity or to use ablation experiments to investigate whether the presence of multiple types of modulations confers any sort of robustness to the network.

Thank you for the great suggestions. By perturbations, we intended to explore the contribution of interaction between certain subpopulations. We tried ablation experiments, but the result was not significant. Probably because the most units were of mixed selectivity, the units of only modulations were not enough for bootstrapping, or the random sampling from single subpopulation (bearing mixed selectivity) could be repeated. We will consider these suggestions carefully in the revised version.

- The results suggest that the observed changes in motor cortical activity with target velocity result from M1 activity receiving an input that encodes the velocity information. This also appears to be the assumption in the RNN model. However, even though the input shown to the animal during preparation is indeed a continuously moving target, it appears that the only relevant quantity to the actual movement is the final endpoint of the reach. While this would have to be a function of the target velocity, one could imagine that the computation of where the monkeys should reach might be performed upstream of the motor cortex, in which case the actual target velocity would become irrelevant to the final motor output. This makes the results of the paper very interesting, but it would be nice if the authors could discuss further when one might expect to see modulation by sensory information that does not directly affect motor output in M1, and where those inputs may come from. It may also be interesting to discuss how the findings relate to previous work that has found behaviourally irrelevant information is being filtered out from M1 (for instance, Russo et al, Neuron 2020 found that in monkeys performing a cycling task, context can be decoded from SMA but not from M1, and Wang et al, Nature Communications 2019 found that perceptual information could not be decoded from PMd)?

How and where sensory information modulates M1 are very interesting and open questions. We will discuss further about this topic in the next version.

Reviewer #3 (Public Review):

Summary:

In this manuscript, the authors show that impairment of hind limb muscle contraction by cast immobilization suppresses skeletal muscle thermogenesis and activates thermogenesis in brown fat. They also propose that free BCAAs derived from skeletal muscle are used for BAT thermogenesis, and identify IL-6 as a potential regulator.

Strengths:

The data support the conclusions for the most part.

Weaknesses:

The data provided in this manuscript are largely descriptive. It is therefore difficult to assess the potential significance of the work. Moreover, many of the described effects are modest in magnitude, questioning the overall functional relevance of this pathway. There are no experiments that directly test whether BCAAs derived from adipose tissue are used for thermogenesis, which would require more robust tracing experiments. In addition, the rigor of the work should be improved. It is also recommended to put the current work in the context of the literature.

eLife assessment

This is a useful paper regarding the roles of brown adipose tissue and skeletal muscle in thermogenesis in mice, with potential significance for the field. The overall approach is innovative but on balance the evidence for the claim is incomplete, as cast immobilization, while innovative, is likely stressful, may impact muscle and BAT directly, and imposes an energetic cost of motion on the animal that is not accounted for. Further experiments are also needed to directly assess the role of adipose-derived BCAAs in thermogenesis.

Reviewer #1 (Public Review):

Summary:

Heat production mechanisms are flexible, depending on a wide variety of genetic, dietary, and environmental factors. The physiology associated with each mechanism is important to understand since loss of flexibility is associated with metabolic decline and disease.

The phenomenon of compensatory heat production has been described in some detail in publications and reviews, notably by modifying BAT-dependent thermogenesis (for example by deleting UCP1 or impairing lipolysis, cited in this paper).

These authors chose to eliminate exercise as an alternative means of maintaining body temperature. To do this, they cast either one or both mouse hindlimbs.

This paper is set up as an evaluation of a loss of function of muscle on the functionality of BAT.

Strengths:

The study is supported by a variety of modern techniques and procedures.

Weaknesses:

The authors show that cast immobilization (CI) does not work as a (passive) loss of function, instead, this procedure produces a dramatic gain of function, putting the animal under considerable stress, inducing b-adrenergic effectors, increased oxygen consumption, and IL6 expression in a variety of tissues, together with commensurate cachectic effects on muscle and fat. The BAT is put under considerable stress, super-induced but relatively poor functioning.

Thus within hours and days of CI, there is massive muscle loss (leading to high circulating BCAAs), and loss of lipid reserves in adipose and liver. The lipid cycle that maintains BAT thermogenesis is depleted and the mouse is unable to maintain body temperature.

I cannot agree with these statements in the Discussion:

"We have here shown that cast immobilization suppressed skeletal muscle thermogenesis, resulting in failure to maintain core body temperature in a cold environment."<br /> • This result could also be attributed to high stress and decreased calorie reserves. Note also: CI suppresses 50% of locomotor activity, but the actual work done by the mouse carrying bilateral casts is not taken into account.

"Thermoregulatory system in endotherms cannot be explained by thermogenesis based on muscle contraction alone, with nonshivering thermogenesis being required as a component of the ability to tolerate cold temperatures in the long term."<br /> • This statement is correct, and it clearly showcases how difficult it is to interpret results using this CI strategy. The question to the author is- which components of muscle thermogenesis are actually inhibited by CI, and what is the relative heat contribution?

This conclusion is overinterpreted:

"In conclusion, we have shown that cast immobilization induced thermogenesis in BAT that was dependent on the utilization of free amino acids derived from skeletal muscle, and that muscle-derived IL-6 stimulated BCAA metabolism in skeletal muscle. Our findings may provide new insights into the significance of skeletal muscle as a large reservoir of amino acids in the regulation of body temperature".

In terms of the production of the article - the data shown in the heat maps has oddly obscure log10 dimensions. The changes are minimal, approx. 1.5x increase/decrease and therefore significance would be key to reporting these data. Fig.3C heatmap is not suitable. What are the 6 lanes to each condition? Overall, this has little/no information.

Rather than cherry-picking for a few genes, the results could be made more rigorous using RNA-seq profiling of BAT and muscle tissues.

Reviewer #2 (Public Review):

Summary:

In this study, the authors identified a previously unrecognized organ interaction where limb immobilization induces thermogenesis in BAT. They showed that limb immobilization by cast fixation enhances the expression of UCP1 as well as amino acid transporters in BAT, and amino acids are supplied from skeletal muscle to BAT during this process, likely contributing to increased thermogenesis in BAT. Furthermore, the experiments with IL-6 knockout mice and IL-6 administration to these mice suggest that this cytokine is likely involved in the supply of amino acids from skeletal muscle to BAT during limb immobilization.

Strengths:

The function of BAT plays a crucial role in the regulation of an individual's energy and body weight. Therefore, identifying new interventions that can control BAT function is not only scientifically significant but also holds substantial promise for medical applications. The authors have thoroughly and comprehensively examined the changes in skeletal muscle and BAT under these conditions, convincingly demonstrating the significance of this organ interaction.

Weaknesses:

Through considerable effort, the authors have demonstrated that limb-immobilized mice exhibit changes in thermogenesis and energy metabolism dynamics at their steady state. However, The impact of immobilization on the function of skeletal muscle and BAT during cold exposure has not been thoroughly analyzed.

eLife assessment

Semenova et al. have studied a large cross-sectional cohort of people living with HIV on suppressive antiretroviral therapy and performed high dimensional flow-cytometry for analysis with data science/machine learning approaches to investigate associations of immunological and clinical parameters and intact/total HIV DNA levels (and categorizations). The study is useful in introducing these new methods and large data set and appears mostly solid, though some of the claims were incompletely supported by the modeling results. The authors have revised the text to fairly reflect their results, yet open questions remain about utility, particularly as to the value of categorical classification (vs continuous measurement) of reservoir size.

On responding to the first round of reviews, the authors have nicely adjusted their wording and fairly describe the results of their study. Certain markers were identified for further investigation. Yet, an overall non-obvious relationship between immune markers and HIV reservoirs has been shown previously, and despite the attempt to leverage powerful ML algorithms, they are not magical and cannot reveal strong relationships that fundamentally do not exist. In addition, categorical classification is for now hard to interpret and the more powerful ML algorithms do not seem to outperform more classic regression methods. Therefore, it remains relatively hard to evaluate the utility of this kind of study.

Initial summary:

Semenova et al. have studied a large cross-sectional cohort of people living with HIV on suppressive ART, N=115, and performed high dimensional flow-cytometry to then search for associations between immunological and clinical parameters and intact/total HIV DNA levels.

A number of interesting data science/ML approaches were explored on the data and the project seems a serious undertaking. However, like many other studies that have looked for these kinds of associations, there was not a very strong signal. Of course the goal of unsupervised learning is to find new hypotheses that aren't obvious to human eyes, but I felt in that context, there were (1) results slightly oversold, (2) some questions about methodology in terms mostly of reservoir levels, and (3) results were not sufficiently translated back into meaning in terms of clinical outcomes.

Strengths:

The study is evidently a large and impressive undertaking and combines many cutting edge statistical techniques with a comprehensive experimental cohort of people living with HIV, notably inclusive of populations underrepresented in HIV science. A number of intriguing hypotheses are put forward that could be explored further. Data will be shared and could be a useful repository for more specific analyses.

Weaknesses:

Despite the detailed experiments and methods, there was not a very strong signal for variable(s) predicting HIV reservoir size. The spearman coefficients are ~0.3, (somewhat weak, and acknowledged as such) and predictive models reach 70-80% prediction levels, though of sometimes categorical variables that are challenging to interpret.

There are some questions about methodology, as well as some conclusions that are not completely supported by results, or at minimum not sufficiently contextualized in terms of clinical significance. Edit, authors have substantially revised the text.

On associations: the false discovery rate correction was set at 5%, but data appear underdetermined with fewer observations than variables (144vars > 115ppts), and it isn't always clear if/when variables are related (e.g inverses of one another, for instance %CD4 and %CD8).

The modeling of reservoir size was unusual, typically intact and defective HIV DNA are analyzed on a log10 scale (both for decays and predicting rebound). Also sometimes in this analysis levels are normalized (presumably to max/min?, e.g. S5), and given the large within-host variation of level we see in other works, it is not trivial to predict any downstream impact of normalization across population vs within person. Edit, fixed.

Also, the qualitative characterization of low/high reservoir is not standard, and naturally will split by early/later ART if done as above/below median. Given the continuous nature of these data it seems throughout that predicting above/below median is a little hard to translate into clinical meaning.

Lastly, work is comprehensive and appears solid, but the code was not shared to see how calculations were performed. Edit, fixed.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Semenova et al. have studied a large cross-sectional cohort of people living with HIV on suppressive ART, N=115, and performed high dimensional flow cytometry to then search for associations between immunological and clinical parameters and intact/total HIV DNA levels.

A number of interesting data science/ML approaches were explored on the data and the project seems a serious undertaking. However, like many other studies that have looked for these kinds of associations, there was not a very strong signal. Of course, the goal of unsupervised learning is to find new hypotheses that aren't obvious to human eyes, but I felt in that context, there were (1) results slightly oversold, (2) some questions about methodology in terms mostly of reservoir levels, and (3) results were not sufficiently translated back into meaning in terms of clinical outcomes.

We appreciate the reviewer’s perspective.  In our revised version of the manuscript, we have attempted to address these concerns by more adequately explaining the limitations of the study and by more thoroughly discussing the context of the findings.  We are not able to associate the findings with specific clinical outcomes for individual study participants but we speculate about the overall biological meaning of these associations across the cohort.  We cannot disagree with the reviewer, but we find the associations statistically significant, potentially reflecting real biological associations, and forming the basis for future hypothesis testing research. 

Strengths:

The study is evidently a large and impressive undertaking and combines many cutting-edge statistical techniques with a comprehensive experimental cohort of people living with HIV, notably inclusive of populations underrepresented in HIV science. A number of intriguing hypotheses are put forward that could be explored further. Sharing the data could create a useful repository for more specific analyses.

We thank the reviewer for this assessment.

Weaknesses:

Despite the detailed experiments and methods, there was not a very strong signal for the variable(s) predicting HIV reservoir size. The Spearman coefficients are ~0.3, (somewhat weak, and acknowledged as such) and predictive models reach 70-80% prediction levels, though sometimes categorical variables are challenging to interpret.

We agree with the reviewer that individual parameters are only weakly correlated with the HIV reservoir, likely reflecting the complex and multi-factorial nature of reservoir/immune cell interactions.  Nevertheless, these associations are statistically significant and form the basis for functional testing in viral persistence.

There are some questions about methodology, as well as some conclusions that are not completely supported by results, or at minimum not sufficiently contextualized in terms of clinical significance.  On associations: the false discovery rate correction was set at 5%, but data appear underdetermined with fewer observations than variables (144vars > 115ppts), and it isn't always clear if/when variables are related (e.g inverses of one another, for instance, %CD4 and %CD8).

When deriving a list of cell populations whose frequency would be correlated with the reservoir, we focused on well-defined cell types for which functional validation exists in the literature to consider them as distinct cell types.  For many of the populations, gating based on combinations of multiple markers leads to recovery of very few cells, and so we excluded some potential combinations from the analysis.  We are also making our raw data available for others to examine and find associations not considered by our manuscript.

The modeling of reservoir size was unusual, typically intact and defective HIV DNA are analyzed on a log10 scale (both for decays and predicting rebound). Also, sometimes in this analysis levels are normalized (presumably to max/min?, e.g. S5), and given the large within-host variation of level we see in other works, it is not trivial to predict any downstream impact of normalization across population vs within-person.

We have repeated the analysis using log10 transformed data and the new figures are shown in Figure 1 and S2-S5.

Also, the qualitative characterization of low/high reservoir is not standard and naturally will split by early/later ART if done as above/below median. Given the continuous nature of these data, it seems throughout that predicting above/below median is a little hard to translate into clinical meaning.

Our ML models included time before ART as a variable in the analysis, and this was not found to be a significant driver of the reservoir size associations, except for the percentage of intact proviruses (see Figure 2C). Furthermore, we analyzed whether any of the reservoir correlated immune variables were associated with time on ART and found that, although some immune variables are associated with time on therapy, this was not the case for most of them (Table S4). We agree that it is challenging to translate above or below median into clinical meaning for this cohort, but we emphasize that this study is primarily a hypothesis generating approach requiring additional validation for the associations observed.  We attempted to predict reservoir size as a continuous variable using the data and this approach was not successful (Figure S13). We believe that a significantly larger cohort will likely be required to generate a ML model that can accurately predict the reservoir as a continuous variable.  We have added additional discussion of this to the manuscript.

Lastly, the work is comprehensive and appears solid, but the code was not shared to see how calculations were performed.

We now provide a link to the code used to perform the analyses in the manuscript, https://github.com/lesiasemenova/ML_HIV_reservoir.

Reviewer #2 (Public Review):

Summary:

Semenova et. al., performed a cross-sectional analysis of host immunophenotypes (using flow cytometry) and the peripheral CD4+ T cell HIV reservoir size (using the Intact Proviral DNA Assay, IPDA) from 115 people with HIV (PWH) on ART. The study mostly highlights the machine learning methods applied to these host and viral reservoir datasets but fails to interpret these complex analyses into (clinically, biologically) interpretable findings. For these reasons, the direct translational take-home message from this work is lost amidst a large list of findings (shown as clusters of associated markers) and sentences such as "this study highlights the utility of machine learning approaches to identify otherwise imperceptible global patterns" - lead to overinterpretation of their data.

We have addressed the reviewer’s concern by modifications to the manuscript that enhance the interpretation of the findings in a clinical and biological context.

Strengths:

Measurement of host immunophenotyping measures (multiparameter flow cytometry) and peripheral HIV reservoir size (IPDA) from 115 PWH on ART.

Major Weaknesses:

(1) Overall, there is little to no interpretability of their machine learning analyses; findings appear as a "laundry list" of parameters with no interpretation of the estimated effect size and directionality of the observed associations. For example, Figure 2 might actually give an interpretation of each X increase in immunophenotyping parameter, we saw a Y increase/decrease in HIV reservoir measure.

We have added additional text to the manuscript in which we attempt to provide more immunological and clinical interpretation of the associations.  We also have emphasized that these associations are still speculative and will require additional validation.  Nevertheless, our data should provide a rich source of new hypotheses regarding immune system/reservoir interaction that could be tested in future work.

(2) The correlations all appear to be relatively weak, with most Spearman R in the 0.30 range or so.

We agree with the review that the associations are mostly weak, consistent with previous studies in this area.  This likely is an inherent feature of the underlying biology – the reservoir is likely associated with the immune system in complex ways and involves stochastic processes that will limit the predictability of reservoir size using any single immune parameter. We have added additional text to the manuscript to make this point clearer.

(3) The Discussion needs further work to help guide the reader. The sentence: "The correlative results from this present study corroborate many of these studies, and provide additional insights" is broad. The authors should spend some time here to clearly describe the prior literature (e.g., describe the strength and direction of the association observed in prior work linking PD-1 and HIV reservoir size, as well as specify which type of HIV reservoir measures were analyzed in these earlier studies, etc.) and how the current findings add to or are in contrast to those prior findings.

We have added additional text to the manuscript to help guide the readers through the possible biological significance of the findings and the context with respect to prior literature.

(4) The most interesting finding is buried on page 12 in the Discussion: "Uniquely, however, CD127 expression on CD4 T cells was significantly inversely associated with intact reservoir frequency." The authors should highlight this in the abstract, and title, and move this up in the Discussion. The paper describes a very high dimensional analysis and the key takeaways are not clear; the more the author can point the reader to the take-home points, the better their findings can have translatability to future follow-up mechanistic and/or validation studies.

We appreciate the reviewer’s comment.  We have increased the emphasis on this finding in the revised version of the manuscript.

(5) The authors should avoid overinterpretation of these results. For example in the Discussion on page 13 "The existence of two distinct clusters of PWH with different immune features and reservoir characteristics could have important implications for HIV cure strategies - these two groups may respond differently to a given approach, and cluster membership may need to be considered to optimize a given strategy." It is highly unlikely that future studies will be performing the breadth of parameters resulting here and then use these directly for optimizing therapy.

Our analyses indicate that membership of study participants in cluster1 or cluster 2 can be fairly accurately determined by a small number of individual parameters (KLRG1 etc, Figure 4F), and measuring the cells of PWH with the degree of breadth used in this paper would not be necessary to classify PWH into these clusters.  As such, we feel that it is not unrealistic to speculate that this finding could turn out to be clinically useful, if it becomes clear that the clusters are biologically meaningful.

(6) There are only TWO limitations listed here: cross-sectional study design and the use of peripheral blood samples. (The subsequent paragraph notes an additional weakness which is misclassification of intact sequences by IPDA). This is a very limited discussion and highlights the need to more critically evaluate their study for potential weaknesses.

We have expanded on the list of limitations discussed in the manuscript. In particular, we now address the size of the cohort, the composition with respect to different genders and demographics, lack of information for the timing of ART and the lack of information regarding intracellular transcriptional pathways.

(7) A major clinical predictor of HIV reservoir size and decay is the timing of ART initiation. The authors should include these (as well as other clinical covariate data - see #12 below) in their analyses and/or describe as limitations of their study.

All of the participants that make up our cohort were treated during chronic infection, and the precise timing of ART initiation is unclear in most of these cases.  We have added additional information to explain this in the manuscript and include this in the list of limitations.

Reviewer #3 (Public Review):

Summary:

This valuable study by Semenova and colleagues describes a large cross-sectional cohort of 115 individuals on ART. Participants contributed a single blood sample which underwent IPDA, and 25-color flow with various markers (pre and post-stimulation). The authors then used clustering, decision tree analyses, and machine learning to look for correlations between these immunophenotypic markers and several measures of HIV reservoir volume. They identified two distinct clusters that can be somewhat differentiated based on total HIV DNA level, intact HIV DNA level, and multiple T cell cellular markers of activation and exhaustion.

The conclusions of the paper are supported by the data but the relationships between independent and dependent variables in the models are correlative with no mechanistic work to determine causality. It is unclear in most cases whether confounding variables could explain these correlations. If there is causality, then the data is not sufficient to infer directionality (ie does the immune environment impact the HIV reservoir or vice versa or both?). In addition, even with sophisticated and appropriate machine learning approaches, the models are not terribly predictive or highly correlated. For these reasons, the study is very much hypothesis-generating and will not impact cure strategies or HIV reservoir measurement strategies in the short term.

We appreciate the reviewer’s comments regarding the value of our study.  We fully acknowledge that the causal nature and directionality of these associations are not yet clear and agree that the study is primarily hypothesis generating in nature.  Nevertheless, we feel that the hypotheses generated will be valuable to the field.  We have added additional text to the manuscript to emphasize the hypothesis generating nature of this paper.

Strengths:

The study cohort is large and diverse in terms of key input variables such as age, gender, and duration of ART. Selection of immune assays is appropriate. The authors used a wide array of bioinformatic approaches to examine correlations in the data. The paper was generally well-written and appropriately referenced.

Weaknesses:

(1) The major limitation of this work is that it is highly exploratory and not hypothesis-driven. While some interesting correlations are identified, these are clearly hypothesis-generating based on the observational study design.

We agree that the major goal of this study was hypothesis generating and that our work is exploratory in nature. Performing experiments with mechanism testing goals in human participants with HIV is challenging.  Additionally, before such mechanistic studies can be undertaken, one must have hypotheses to test. As such we feel our study will be useful for the field in helping to identify hypotheses that could potentially be tested.

(2) The study's cross-sectional nature limits the ability to make mechanistic inferences about reservoir persistence. For instance, it would be very interesting to know whether the reservoir cluster is a feature of an individual throughout ART, or whether this outcome is dynamic over time.

We agree with the reviewer’s comment. Longitudinal studies are challenging to carry out with a study cohort of this size, and addressing questions such as the one raised by the reviewer would be of great interest. We believe our study nevertheless has value in identifying hypotheses that could be tested in a longitudinal study.

(3) A fundamental issue is that I am concerned that binarizing the 3 reservoir metrics in a 50/50 fashion is for statistical convenience. First, by converting a continuous outcome into a simple binary outcome, the authors lose significant amounts of quantitative information. Second, the low and high reservoir outcomes are not actually demonstrated to be clinically meaningful: I presume that both contain many (?all) data points above levels where rebound would be expected soon after interruption of ART. Reservoir levels would also have no apparent outcome on the selection of cure approaches. Overall, dividing at the median seems biologically arbitrary to me.

The reviewer raises a valid point that the clinical significance of above or below median reservoir metrics is unclear, and that the size of the reservoir has potentially little relation to rebound and cure approaches.  In the manuscript, we attempted to generate models that can predict reservoir size as a continuous variable in Figure S13 and find that this approach performs poorly, while a binarized approach was more successful. As such we have included both approaches in the manuscript.  It is possible that future studies with larger sample sizes and more detailed measurements will perform better for continuous variable prediction.  While this is a fairly large study (n=115) by the standards of HIV reservoir analyses, it is a small study by the standards of the machine learning field, and accurate predictive ML models for reservoir size as a continuous variable will likely require a much larger set of samples/participants.  Nevertheless, we feel our work has value as a template for ML approaches that may be informative for understanding HIV/immune interactions and generates novel hypotheses that could be validated by subsequent studies.

(4) The two reservoir clusters are of potential interest as high total and intact with low % intact are discriminated somewhat by immune activation and exhaustion. This was the most interesting finding to me, but it is difficult to know whether this clustering is due to age, time on ART, other co-morbidity, ART adherence, or other possible unmeasured confounding variables.

We agree that this finding is one of the more interesting outcomes of the study. We examined a number of these variables for association with cluster membership, and these data are reported in Figure S8A-D.  Age, years of ART and CD4 Nadir were all clearly different between the clusters.   The striking feature of this clustering, however, is the clear separation between the two groups of participants, as opposed to a continuous gradient of phenotypes.  This could reflect a bifurcation of outcomes for people with HIV, dynamic changes in the reservoir immune interactions over time, or different levels of untreated infection.  It is certainly possible that some other unmeasured confounding variables contribute to this outcome and we have attempted to make this limitation clearer.

(5) At the individual level, there is substantial overlap between clusters according to total, intact, and % intact between the clusters. Therefore, the claim in the discussion that these 2 cluster phenotypes may require different therapeutic approaches seems rather speculative. That said, the discussion is very thoughtful about how these 2 clusters may develop with consideration of the initial insult of untreated infection and / or differences in immune recovery.

We agree with the reviewer that this claim is speculative, and we have attempted to moderate the language of the text in the revised version.

(6) The authors state that the machine learning algorithms allow for reasonable prediction of reservoir volume. It is subjective, but to me, 70% accuracy is very low. This is not a disappointing finding per se. The authors did their best with the available data. It is informative that the machine learning algorithms cannot reliably discriminate reservoir volume despite substantial amounts of input data. This implies that either key explanatory variables were not included in the models (such as viral genotype, host immune phenotype, and comorbidities) or that the outcome for testing the models is not meaningful (which may be possible with an arbitrary 50/50 split in the data relative to median HIV DNA volumes: see above).

We acknowledge that the predictive power of the models generated from these data is modest and we have clarified this point in the revised manuscript. As the reviewer indicates, this may result from the influence of unmeasured variables and possible stochastic processes.  The data may thus demonstrate a limit to the predictability of reservoir size which may be inherent to the underlying biology.  As we mention above, this study size (n-115) is fairly small for the application of ML methods, and an increased sample size will likely improve the accuracy of the models. At this stage, the models we describe are not yet useful as predictive clinical tools, but are still nonetheless useful as tools to describe the structure of the data and identify reservoir associated immune cell types.

(7) The decision tree is innovative and a useful addition, but does not provide enough discriminatory information to imply causality, mechanism, or directionality in terms of whether the immune phenotype is impacting the reservoir or vice versa or both. Tree accuracy of 80% is marginal for a decision tool.

The reviewer is correct about these points.  In the revised manuscript, we have attempted to make it clear that we are not yet advocating using this approach as a decision tool, but simply a way to visualize the data and understand the structure of the dataset.  As we discuss above, the models will likely need to be trained on a larger dataset and achieve higher accuracy before use as a decision tool.

(8) Figure 2: this is not a weakness of the analysis but I have a question about interpretation. If total HIV DNA is more predictive of immune phenotype than intact HIV DNA, does this potentially implicate a prior high burden of viral replication (high viral load &/or more prolonged time off ART) rather than ongoing reservoir stimulation as a contributor to immune phenotype? A similar thought could be applied to the fact that clustering could only be detected when applied to total HIV DNA-associated features. Many investigators do not consider defective HIV DNA to be "part of the reservoir" so it is interesting to speculate why these defective viruses appear to have more correlation with immunophenotype than intact viruses.

We agree with the reviewer that this observation could reflect prior viral burden and we have added additional text to make this clearer.  Even so, we cannot rule out a model in which defective viral DNA is engaged in ongoing stimulation of the immune system during ART, leading to the stronger association between total DNA and the immune cell phenotypes. We hypothesize that the defective proviruses could potentially be triggering innate immune pattern recognition receptors via viral RNA or DNA, and a higher burden of the total reservoir leads to a stronger apparent association with the immune phenotype.  We have included text in the discussion about this hypothesis.

(9) Overall, the authors need to do an even more careful job of emphasizing that these are all just correlations. For instance, HIV DNA cannot be proven to have a causal effect on the immunophenotype of the host with this study design. Similarly, immunophenotype may be affecting HIV DNA or the correlations between the two variables could be entirely due to a separate confounding variable

We have revised the text of the manuscript to emphasize this point, and we acknowledge that any causal relationships are, at this point, simply speculation. 

(10) In general, in the intro, when the authors refer to the immune system, they do not consistently differentiate whether they are referring to the anti-HIV immune response, the reservoir itself, or both. More specifically, the sentence in the introduction listing various causes of immune activation should have citations. (To my knowledge, there is no study to date that definitively links proviral expression from reservoir cells in vivo to immune activation as it is next to impossible to remove the confounding possible imprint of previous HIV replication.) Similarly, it is worth mentioning that the depletion of intact proviruses is quite slow such that provial expression can only be stimulating the immune system at a low level. Similarly, the statement "Viral protein expression during therapy likely maintains antigen-specific cells of the adaptive immune system" seems hard to dissociate from the persistence of immune cells that were reactive to viremia.

We updated the text of the manuscript to address these points and have added additional citations as per the reviewer’s suggestion.

(11) Given the many limitations of the study design and the inability of the models to discriminate reservoir volume and phenotype, the limitations section of the discussion seems rather brief.

We have now expanded the limitations section of the discussion and added additional considerations. We now include a discussion of the study cohort size, composition and the detail provided by the assays.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

A few specific comments:

"This pattern is likely indicative of a more profound association of total HIV DNA with host immunophenotype relative to intact HIV DNA."

Most studies I have seen (e.g. single cell from Lictherfeld/Yu group) show intact proviruses are generally more activated/detectable/susceptible to immune selection, so I have a hard time thinking defective proviruses are actually more affected by immunotype.

We hypothesize that this association is actually occurring in the opposite direction – that the defective provirus are having a greater impact on the immune phenotype, due to their greater number and potential ability to engage innate or adaptive immune receptors. We have clarified this point in the manuscript

"The existence of two distinct clusters of PWH with different immune features and reservoir characteristics could have important implications for HIV cure strategies - these two groups may respond differently to a given approach, and cluster membership may need to be considered to optimize a given strategy."

I find this a bit of a reach, given that the definition of 2 categories depended on the total size.

We have modified the language of this section to reduce the level of speculation.

"This study is cross-sectional in nature and is primarily observational, so caution should be used interpreting findings associated with time on therapy".

I found this an interesting statement because ultimately time on ART shows up throughout the analysis as a significant predictor, do you mean something about how time on ART could indicate other confounding variables like ART regimen or something?

We have rephrased this comment to avoid confusion.  We were simply trying to make the point that we should avoid speculating about longitudinal dynamics from cross sectional data.

"As expected, the plots showed no significant correlation for intact HIV DNA versus years of ART (Figure 1B), while total reservoir size was positively correlated with the time of ART (Figure 1A, Spearman r = 0.31)."<br />  Is this expected? Studies with longitudinal data almost uniformly show intact decay, at least for the first 10 or so years of ART, and defective/total stability (or slight decay). Also probably "time on ART" to not confuse with the duration of infection before ART.

We have updated the language of this section to address this comment.  We have avoided comparing our data with respect to time on ART to longitudinal studies for reasons given above.

On dimensionality reduction, as this PaCMAP seems a relatively new technique (vs tSNE and UMAP which are more standard, but absolutely have their weaknesses), it does seem important to contextualize. I think it would still be useful to show PCA and asses the % variance of each additional dimension to assess the effective dimensionality, it would be helpful to show a plot of % variance by # components to see if there is a cutoff somewhere, and if PaCMAP is really picking this up to determine the 2 dimensions/2 clusters is ideal. Figure 4B ultimately shows a lot of low/high across those clusters, and since low/high is defined categorically it's hard to know which of those dots are very close to the other categories.

We have added this analysis to the manuscript – found in Figure S9. The PCA plot indicates that members of the two clusters also separate on PCA although this separation is not as clear as for the PaCMAP plot.

Minor comments on writing etc:

Intro

-Needs some references on immune activation sequelae paragraph.

We have added some additional references to this section.

-"promote the entry of recently infected cells into the reservoir" -- that is only one possible mechanistic explanation, it's not unreasonable but it seems important to keep options open until we have more precise data that can illuminate the mechanism of the overabundance.

We have modified the text to discuss additional hypotheses.

-You might also reference Pankau et al Ppath for viral seeding near the time of ART.

We have added this reference.

-"Viral protein expression during therapy likely maintains antigen-specific cells of the adaptive immune system" - this was unclear to me, do you mean HIV-specific cells that act against HIV during ART? I think most studies show immunity against HIV (CD8 and CD4) wanes over time during ART.

The Goonetilleke lab has recently generated data indicating that antiviral T cell responses are remarkably stable over time on ART, but we agree with the reviewer that the idea that ongoing antigen expression in the reservoir maintains these cells is speculative.  We have modified the text to make this point clearer.

-Overall I think the introduction lacked a little bit of definitional precision: i.e. is the reservoir intact vs replication competent vs all HIV DNA and whether we are talking about PWH on long-term ART and how long we should be imagining? The first years of ART are certainly different than later, in terms of dynamics. The ultimate implications are likely specific for some of these categorizations.

-"persistent sequelae of the massive disruptions to T cell homeostasis and lymphoid structures that occur during untreated HIV infection" needs a lot more context/referencing. For instance, Peter Hunt showed a decrease in activation after ART a long time ago.

-Heather Best et al show T cell clonality stays perturbed after ART.

We have updated the text of the introduction and added references to address the reviewer’s comments.

Results

-It would be important to mention the race of participants and any information about expected clades of acquired viruses, this gets mentioned eventually with reference to the Table but the breakdown would be helpful right away.

We have added this information to the results section.

-"performed Spearman correlations", may be calculated or tested?

We have corrected the language for this sentence.

Comments on figures:

-Figure 1 data on linear scale (re discussion above) -- hard to even tell if there is a decay (to match with all we know from various long-term ART studies).

-Figure 4 data is shown on ln (log_e) scale, which is hard to interpret for most people.

-Figures 4 C,D, and E should have box plots to visually assess the significance.

-Figure 4B legend says purple/pink but I think the colors are different in the plot, could be about transparency

-Figure 5 it is now not clear if log_e(?).

-Figure 6 "HIV reservoir characteristics" might be better to make this more explicit. Do you mean for instance in the 6B title Total HIV DNA per million CD4+ T cells I think?

We have made these modifications.

Reviewer #2 (Recommendations For The Authors):

Minor Weaknesses:

(1) The Introduction is too long and much of the text is not directly related to the study's research question and design.

We have streamlined the introduction in the revised manuscript.

(2) While no differences were seen by age or race, according to the authors, this is unlikely to be useful since the numbers are so small in some of these subcategories. Results from sensitivity analyses (e.g., excluding these individuals) may be more informative/useful.

We agree that the lower numbers of participants for some subgroupings makes it challenging to know for sure if there are any differences based on these variables.  Have added text to clarify this. We have added age, race and gender to the LOCO analysis and to the variable inflation importance analysis (Table S5).

(3) For Figure 4, based on what was described in the Results section of the manuscript, the authors should clarify that the figures show results for TOTAL HIV DNA only (not intact DNA): "Dimension reduction machine learning approaches identified two robust clusters of PWH when using total HIV DNA reservoir-associated immune cell frequencies (Figure 4A), but not for intact or percentage intact HIV DNA (Figure 4B and 4C)".

We have added this information.

(4) The statement on page 5, first paragraph, "Interestingly, when we examined a plot of percent intact proviruses versus time on therapy (Figure 1C), we observed a biphasic decay pattern," is not new (Peluso JCI Insight 2020, Gandhi JID 2023, McMyn JCI 2023). Prior studies have clearly demonstrated this biphasic pattern and should be cited here, and the sentence should be reworded with something like "consistent with prior work", etc.

We have added citations to these studies and rephrased this comment.

(5) The Cohort and sample collection sections are somewhat thin. Further details on the cohort details should include at the very minimum some description of the timing of ART initiation (is this mostly a chronic-treated cohort?) and important covariate data such as nadir CD4+ T cell count, pre-ART viral load, duration of ART suppression, etc.

The cohort was treated during chronic infection, and we have clarified this in the manuscript.  Information regarding CD4 nadir and years on ART are included in Table 1.  Unfortunately, pre-ART viral load was not available for most members of this cohort, so we did not use it for analyses. The partial pre-ART viral load data is included with the dataset we are making publicly available.

Reviewer #3 (Recommendations For The Authors):

Minor points:

(1) What is meant by CD4 nadir? Is this during primary infection or the time before ART initiation?

We have clarified this description in the manuscript.  This term refers to the lowest CD4 count recorded during untreated infection.

(2) The authors claim that determinants of reservoir size are starting to emerge but other than the timing of ART, I am not sure what studies they are referring to.

We have updated the language of this section.  We intended to refer to studies looking at correlates of reservoir size, and feel that this is a more appropriate term that ‘determinants’

(3) The discussion does not tie in the model-generated hypotheses with the known mechanisms that sustain the reservoir: clonal proliferation balanced by death and subset differentiation. It would be interesting to tie in the proposed reservoir clusters with these known mechanisms.

We have added additional text to the manuscript to address these mechanisms.

(4) Figure 1: Total should be listed as total HIV DNA.

We have updated this in the manuscript.

(5) Figure 1C: Worth mentioning the paper by Reeves et al which raises the possibility that the flattening of intact HIV DNA at 9 years may be spurious due to small levels of misclassification of defective as intact.

We have added this reference.

(6) "Total reservoir frequency" should be "total HIV DNA concentration"

We respectfully feel that “frequency” is a more accurate term than “concentration”, since we are expressing the reservoir as a fraction of the CD4 T cells, while “concentration” suggests a denominator of volume.

(7) Figure S2-5: label y-axis total HIV DNA.

We have updated this figure.

eLife assessment

The present work provides new insights into detailed brain morphology. Using state-of-the-art methods, it provides compelling evidence for the relevance of sucal morphology for the precise localization of brain function. The fundamental findings have great relevance for the fields of imaging neuroscience and individualized medicine as ever-improving techniques improve precision to the point where individual brain anatomy is taking centre stage.

Reviewer #1 (Public Review):

Summary:<br /> Ever-improving techniques allow the detailed capture of brain morphology and function to the point where individual brain anatomy becomes an important factor. This study investigated detailed sulcal morphology in the parieto-occipital junction. Using cutting-edge methods, it provides important insights into local anatomy, individual variability, and local brain function. The presented work advances the field and will stimulate future research into this important area.

Strengths:<br /> Detailed, very thorough methodology. Multiple raters mapped detailed sulci in a large cohort. The identified sulcal features and their functional and behavioural relevance are then studied using various complementary methods. The results provide compelling evidence for the importance of the described sulcal features and their proposed relationship to cortical brain function.

Comments on revised version:

The revised manuscript addresses all my concerns.

Reviewer #2 (Public Review):

Summary:<br /> After manually labelling 144 human adult hemispheres in the lateral parieto-occipital junction (LPOJ), the authors 1) propose a nomenclature for 4 previously unnamed highly variable sulci located between the temporal and parietal or occipital lobes, 2) focus on one of these newly named sulci, namely the ventral supralateral occipital sulcus (slocs-v) and compare it to neighbouring sulci to demonstrate its specificity (in terms of depth, surface area, gray matter thickness, myelination, and connectivity), 3) relate the morphology of a subgroup of sulci from the region including the slocs-v to the performance in a spatial orientation task, demonstrating behavioural and morphological specificity. In addition to these results, the authors propose an extended reflection on the relationship between these newly named landmarks and previous anatomical studies, a reflection about the slocs-v related to functional and cytoarchitectonic parcellations as well as anatomic connectivity and an insight about potential anatomical mechanisms relating sulcation and behaviour.

Strengths:<br /> - To my knowledge, this is the first study addressing the variable tertiary sulci located between the superior temporal sulcus (STS) and intra-parietal sulcus (IPS).<br /> - This is a very comprehensive study addressing altogether anatomical, architectural, functional and cognitive aspects.<br /> - The definition of highly variable yet highly reproductible sulci such as the slocs-v feeds the community with new anatomo-functional landmarks (which is emphasized by the provision of a probability map in supp. mat., which in my opinion should be proposed in the main body).<br /> - The comparison of different features between the slocs-v and similar sulci is useful to demonstrate their difference.<br /> - The detailed comparison of the present study with state of the art contextualises and strengthens the novel findings.<br /> - The functional study complements the anatomical description and points towards cognitive specificity related to a subset of sulci from the LPOJ<br /> - The discussion offers a proposition of theoretical interpretation of the findings<br /> - The data and code are mostly available online (raw data made available upon request).

Weaknesses:<br /> - While the identification of the sulci has been done thoroughly with expert validation, the sulci have not been labelled in a way that enables the demonstration of the reproducibility of the labelling.

The proposed methodology is convincing in identifying and studying the relationship between highly variable sulci and cognition. This improves our refined understanding of the general anatomical variability in the LPOJ and its potential functional/cognitive correlates. This work is important to the understanding of sulcal variability and its implications on functional and cognitive aspects.

Comments on revised version:

Thank you for the elegant and informative work.

Reviewer #3 (Public Review):

Summary:<br /> 72 subjects, and 144 hemispheres, from the Human Connectome Project had their parietal sulci manually traced. This identified the presence of previous undescribed shallow sulci. One of these sulci, the ventral supralateral occipital sulcus (slocs-v), was then demonstrated to have functional specificity in spatial orientation. The discussion furthermore provides an eloquent overview of our understanding of the anatomy of the parietal cortex, situating their new work into the broader field. Finally, this paper stimulates further debate about the relative value of detailed manual anatomy, inherently limited in participant numbers and areas of the brain covered, against fully automated processing that can cover thousands of participants but easily misses the kinds of anatomical details described here.

Strengths:<br /> - This is the first paper describing the tertiary sulci of the parietal cortex with this level of detail, identifying novel shallow sulci and mapping them to behaviour and function.<br /> - It is a very elegantly written paper, situating the current work into the broader field.<br /> - The combination of detailed anatomy and function and behaviour is superb.

Weaknesses:<br /> - the numbers of subjects are inherently limited both in number as well as in being typically developing young adults.<br /> - while the paper begins by describing four new sulci, only one is explored further in greater detail.<br /> - there is some tension between calling the discovered sulci new vs acknowledging they have already been reported, but not named.<br /> - the anatomy of the sulci, as opposed to their relation to other sulci, could be described in greater detail.

Overall, to summarize, I greatly enjoyed this paper and believe it to be a highly valued contribution to the field.

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors report that activation of excitatory DREADDs in the mid-cervical spinal cord can increase inspiratory activity in mice and rats. This is an important first step toward an ultimate goal of using this, or similar, technology to drive breathing in disorders associated with decreased respiratory motor output, such as spinal injury or neurodegenerative disease.

Strengths:

Strengths of this study include a comparison of non-specific DREADD expression in the mid-cervical spinal cord versus specific targeting to ChAT-positive neurons, and the measurement of multiple respiratory-related outcomes, including phrenic inspiratory output, diaphragm EMG activity, and ventilation. The data show convincingly that DREADDs can be used to drive phrenic inspiratory activity, which in turn increases diaphragm EMG activity and ventilation.

Weaknesses:

The main limitation is that the ligand, J60, was not given to control animals without spinal DREADD expression. Since J60 may have off-target effects (PMID: 37530882), a discussion of this limitation is warranted, particularly in light of the one rat that was reported to not have detectible mCherry expression in the mid-cervical spinal cord, yet had robust increases in diaphragm output after J60 administration.

In experiments in ChAT-Cre animals, several neuronal types will express DREADDs, including non-phrenic motor neurons and some interneurons. As such, these experiments do not specifically "target" phrenic motor neurons any more so than experiments in WT animals. Experiments in ChAT-Cre animals also do not avoid inducing "non-specific expression in the vicinity of the phrenic motor nucleus". This is not a study design flaw per se, but an overinterpretation of findings.

Reviewer #2 (Public Review):

Summary:

This study shows that when excitatory DREADD receptors are expressed in the ventral area of the cervical spinal cord containing phrenic motoneurons, the systemic administration of the DREADD ligand J60 increases diaphragm EMG activity without altering respiratory rate. The authors took a non-selective expression approach in wild-type mice, as well as a more selective Cre-dependent approach in Chat-Cre mice and Chat-Cre rats to stimulate cervical motoneurons in the spinal cord. This is a proof of principle study that supports the use of DREADD technology to stimulate the motor output to the diaphragm.

Strengths:

The strengths of the study lie in the use of both mice and rats and testing activation of diaphragmatic activity with multiple experimental approaches to show that diaphragm EMG and tidal volume are increased.

Weaknesses:

Weaknesses of the study consist in the lack of some important control experiments to consolidate the findings: a test of DREADD ligand effects in the absence of viral construct; repeated respiratory challenges within the same recording session in whole body plethysmographs that could compromise the behaving experiments; and lastly, a limited qualitative analysis of the histological data that does not allow for confirmation of expression in phrenic motoneurons.

eLife assessment

The authors report that chemogenetic methods targeting the ventral cervical spinal cord can be used to increase phrenic inspiratory motor output and subsequent diaphragm EMG activity and ventilation in rodents. These findings are important because they are a necessary first step towards using chemogenetic methods to drive inspiratory activity in disorders in which motor neurons are compromised, such as spinal injury and degenerative disease. The data are convincing, with rigorous assessments of phrenic inspiratory activity and its ability to drive the diaphragm and subsequent ventilation, as well as assessments of DREADD expression.

eLife assessment

This is an important study that connects the polymerase-associated factor 1 complex (Paf1C) with Histone 2B monoubiquitination and the expression of genes key to virulence in Cryptococcus neoformans. The provided information is convincing and has the potential to open several opportunities to further understand the basic biology of this significant human fungal pathogen.

Reviewer #1 (Public Review):

Summary:

In the manuscript entitled "Rtf1 HMD domain facilitates global histone H2B monoubiquitination and regulates morphogenesis and virulence in the meningitis-causing pathogen Cryptococcus neoformans" by Jiang et al., the authors employ a combination of molecular genetics and biochemical approaches, along with phenotypic evaluations and animal models, to identify the conserved subunit of the Paf1 complex (Paf1C), Rtf1, and functionally characterize its critical roles in mediating H2B monoubiquitination (H2Bub1) and the consequent regulation of gene expression, fungal development, and virulence traits in C. deneoformans or C. neoformans. Specially, the authors found that the histone modification domain (HMD) of Rtf1 is sufficient to promote H2B monoubiquitination (H2Bub1) and the expression of genes related to fungal mating and filamentation, and restores the fungal morphogenesis and pathogenicity defects caused by RTF1 deletion.

Strengths:

The manuscript is well-written and presents the findings in a clear manner. The findings are interesting and contribute to a better understanding of Rtf1-mediated epigenetic regulation of fungal morphogenesis and pathogenicity in a major human fungal pathogen, and potentially in other fungal species, as well.

Weaknesses:

A major limitation of this study is the absence of genome-wide information on Rtf1-mediated H2B monoubiquitination (H2Bub1), as well as a lack of detail regarding the function of the Plus3 domain. Although overexpression of HMD in the rtf1Δ mutant restored global H2Bub1 levels, it did not rescue certain critical biological functions, such as growth at 39{degree sign}C and melanin production (Figure 4C-D). This suggests that the precise positioning of H2Bub1 is essential for Rtf1's function. A comprehensive epigenetic landscape of H2Bub1 in the presence of HMD or full-length Rtf1 would elucidate potential mechanisms and shed light on the function of the Plus3 domain.

Reviewer #2 (Public Review):

Summary:

The authors set out to determine the role of Rtf1 in Cryptococcal biology, and demonstrate that Rtf1 acts independently of the Paf1 complex to exert regulation of Histone H2B monoubiquitylation (H2Bub1). The biological impact of the loss of H2Bub1 was observed in defects in morphogenesis, reduced production of virulence factors, and reduced pathogenic potential in animal models of cryptococcal infection.

Strengths:

The molecular data is quite compelling, demonstrating that the Rtf1-depednent functions require only this histone modifying domain of Rtf1, and are dependent on nuclear localization. A specific point mutation in a residue conserved with the Rtf1 protein in the model yeast demonstrates the conservation of that residue in H2Bub1 modification. Interestingly, whereas expression of the HMD alone suppressed the virulence defect of the rtf1 deletion mutant, it did not suppress defects in virulence factor production.

Weaknesses:

The authors use two different species of Cryptococcus to investigate the biological effect of Rtf1 deletion. The work on morphogenesis utilized C. deneoformans, which is well-known to be a robust mating strain. The virulence work was performed in the C. neoformans H99 background, which is a highly pathogenic isolate. The study would be more complete if each of these processes were assessed in the other strain to understand if these biological effects are conserved across the two species of Cryptococcus. H99 is not as robust in morphogenesis, but reproducible results assessing mating and filamentation in this strain have been performed. Similarly, C. deneoformans does produce capsule and melanin.

There are some concerns with the conclusions related to capsule induction. The images reported in Figure B are purported to be grown under capsule-inducing conditions, yet the H99 panel is not representative of the induced capsule for this strain. Given the lack of a baseline of induction, it is difficult to determine if any of the strains may be defective in capsule induction. Quantification of a population of cells with replicates will also help to visualize the capsular diversity in each strain population.

The authors demonstrate that for specific mating-related genes, the expression of the HMD recapitulated the wild-type expression pattern. The RNA-seq experiments were performed under mating conditions, suggesting specificity under this condition. The authors raise the point in the discussion that there may be differences in Rtf1 deposition on chromatin in H99, and under conditions of pathogenesis. The data that overexpression of HMD restores H2Bub1 by western is quite compelling, but does not address at which promoters H2Bub1 is modulating expression under pathogenesis conditions, and when full-length Rtf1 is present vs. only the HMD.

Reviewer #3 (Public Review):

Summary:

In this very comprehensive study, the authors examine the effects of deletion and mutation of the Paf1C protein Rtf1 gene on chromatin structure, filamentation, and virulence in Cryptococcus.

Strengths:

The experiments are well presented and the interpretation of the data is convincing.

Weaknesses:

Yet, one can be frustrated by the lack of experiments that attempt to directly correlate the change in chromatin structure with the expression of a particular gene and the observed phenotype. For example, the authors observed a strong defect in the expression of ZNF2, a known regulator of filamentation, mating, and virulence, in the rtf1 mutant. Can this defect explain the observed phenotypes associated with the RTF1 mutation? Is the observed defect in melanin production associated with altered expression of laccase genes and altered chromatin structure at this locus?

Author response:

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In the manuscript entitled "Rtf1 HMD domain facilitates global histone H2B monoubiquitination and regulates morphogenesis and virulence in the meningitis-causing pathogen Cryptococcus neoformans" by Jiang et al., the authors employ a combination of molecular genetics and biochemical approaches, along with phenotypic evaluations and animal models, to identify the conserved subunit of the Paf1 complex (Paf1C), Rtf1, and functionally characterize its critical roles in mediating H2B monoubiquitination (H2Bub1) and the consequent regulation of gene expression, fungal development, and virulence traits in C. deneoformans or C. neoformans. Specially, the authors found that the histone modification domain (HMD) of Rtf1 is sufficient to promote H2B monoubiquitination (H2Bub1) and the expression of genes related to fungal mating and filamentation, and restores the fungal morphogenesis and pathogenicity defects caused by RTF1 deletion.

Strengths:

The manuscript is well-written and presents the findings in a clear manner. The findings are interesting and contribute to a better understanding of Rtf1-mediated epigenetic regulation of fungal morphogenesis and pathogenicity in a major human fungal pathogen, and potentially in other fungal species, as well.

Weaknesses:

A major limitation of this study is the absence of genome-wide information on Rtf1-mediated H2B monoubiquitination (H2Bub1), as well as a lack of detail regarding the function of the Plus3 domain. Although overexpression of HMD in the rtf1Δ mutant restored global H2Bub1 levels, it did not rescue certain critical biological functions, such as growth at 39 °C and melanin production (Figure 4C-D). This suggests that the precise positioning of H2Bub1 is essential for Rtf1's function. A comprehensive epigenetic landscape of H2Bub1 in the presence of HMD or full-length Rtf1 would elucidate potential mechanisms and shed light on the function of the Plus3 domain.

We thank the reviewer (and other reviewers) for this excellent suggestion. We have planned to carry out CUT&Tag assay to gain a comprehensive epigenetic landscape of H2Bub1 in the presence of HMD or full-length Rtf1 under conditions, where overexpression of HMD failed to rescue the phenotypes in the _rtf1_Δ mutant, such as growth at 39 °C.

Reviewer #2 (Public Review):

Summary:

The authors set out to determine the role of Rtf1 in Cryptococcal biology, and demonstrate that Rtf1 acts independently of the Paf1 complex to exert regulation of Histone H2B monoubiquitylation (H2Bub1). The biological impact of the loss of H2Bub1 was observed in defects in morphogenesis, reduced production of virulence factors, and reduced pathogenic potential in animal models of cryptococcal infection.

Strengths:

The molecular data is quite compelling, demonstrating that the Rtf1-depednent functions require only this histone modifying domain of Rtf1, and are dependent on nuclear localization. A specific point mutation in a residue conserved with the Rtf1 protein in the model yeast demonstrates the conservation of that residue in H2Bub1 modification. Interestingly, whereas expression of the HMD alone suppressed the virulence defect of the rtf1 deletion mutant, it did not suppress defects in virulence factor production.

Weaknesses:

The authors use two different species of Cryptococcus to investigate the biological effect of Rtf1 deletion. The work on morphogenesis utilized C. deneoformans, which is well-known to be a robust mating strain. The virulence work was performed in the C. neoformans H99 background, which is a highly pathogenic isolate. The study would be more complete if each of these processes were assessed in the other strain to understand if these biological effects are conserved across the two species of Cryptococcus. H99 is not as robust in morphogenesis, but reproducible results assessing mating and filamentation in this strain have been performed. Similarly, C. deneoformans does produce capsule and melanin.

This is a fair point raised by the reviewer, and we are going to test whether these biological effects are conserved across the two species. We will access effects of RTF1 deletion on bisexual mating hyphal formation in C. neoformans H99 background and capsule and melanin productions in C. deneoformans XL280 background.

There are some concerns with the conclusions related to capsule induction. The images reported in Figure B are purported to be grown under capsule-inducing conditions, yet the H99 panel is not representative of the induced capsule for this strain. Given the lack of a baseline of induction, it is difficult to determine if any of the strains may be defective in capsule induction. Quantification of a population of cells with replicates will also help to visualize the capsular diversity in each strain population.

We thank the reviewer for raising this concern. We are going to confirm the conclusions related to capsule induction under multiple capsule-inducing conditions, including Dulbecco’s Modified Eagle’s Medium (DMEM), Littman’s medium, and 10% fetal bovine serum (FBS) agar medium [1].

The authors demonstrate that for specific mating-related genes, the expression of the HMD recapitulated the wild-type expression pattern. The RNA-seq experiments were performed under mating conditions, suggesting specificity under this condition. The authors raise the point in the discussion that there may be differences in Rtf1 deposition on chromatin in H99, and under conditions of pathogenesis. The data that overexpression of HMD restores H2Bub1 by western is quite compelling, but does not address at which promoters H2Bub1 is modulating expression under pathogenesis conditions, and when full-length Rtf1 is present vs. only the HMD.

We thank the reviewer for raising these concerns. As mentioned in the response to Reviewer 1, our CUT&Tag assay will provide evidence to address these questions.

Reviewer #3 (Public Review):

Summary:

In this very comprehensive study, the authors examine the effects of deletion and mutation of the Paf1C protein Rtf1 gene on chromatin structure, filamentation, and virulence in Cryptococcus.

Strengths:

The experiments are well presented and the interpretation of the data is convincing.

Weaknesses:

Yet, one can be frustrated by the lack of experiments that attempt to directly correlate the change in chromatin structure with the expression of a particular gene and the observed phenotype. For example, the authors observed a strong defect in the expression of ZNF2, a known regulator of filamentation, mating, and virulence, in the rtf1 mutant. Can this defect explain the observed phenotypes associated with the RTF1 mutation? Is the observed defect in melanin production associated with altered expression of laccase genes and altered chromatin structure at this locus?

We completely agree with the reviewer, and as mentioned in our response to Reviewer 1 and 2, we are going to conduct CUT&Tag assay to investigate the genetic relationship between Rtf1-mediated H2Bub1 and the expression of particular genes.

(1) Jang, E.-H., et al., Unraveling Capsule Biosynthesis and Signaling Networks in Cryptococcus neoformans. Microbiology Spectrum, 2022. 10(6): p. e02866-22.

Reviewer #1 (Public Review):

Summary:

In this manuscript, Jiao D et al reported the induction of synthetic lethal by combined inhibition of anti-apoptotic BCL-2 family proteins and WSB2, a substrate receptor in CRL5 ubiquitin ligase complex. Mechanistically, WSB2 interacts with NOXA to promote its ubiquitylation and degradation. Cancer cells deficient in WSB2, as well as heart and liver tissues from Wsb2-/- mice exhibit high susceptibility to apoptosis induced by inhibitors of BCL-2 family proteins. The anti-apoptotic activity of WSB2 is partially dependent on NOXA.

Overall, the finding, that WSB2 disruption triggers synthetic lethality to BCL-2 family protein inhibitors by destabilizing NOXA, is rather novel. The manuscript is largely hypothesis-driven, with experiments that are adequately designed and executed. However, there are quite a few issues for the authors to address, including those listed below.

Specific comments:

(1) At the beginning of the Results section, a clear statement is needed as to why the authors are interested in WSB2 and what brought them to analyze "the genetic co-dependency between WSB2 and other proteins".

(2) In general, the biochemical evidence supporting the role of WSB2 as a SOCS box-containing substrate-binding receptor of CRL5 E3 in promoting NOXA ubiquitylation and degradation is relatively weak. First, since NOXA2 binds to WSB2 on its SOCS box, which consists of a BC box for Elongin B/C binding and a CUL5 box for CUL5 binding, it is crucial to determine whether the binding of NOXA on the SOCS box affects the formation of CRL5WSB2 complex. The authors should demonstrate the endogenous binding between NOXA and the CRL5WSB2 complex. Additionally, the authors may also consider manipulating CUL5, SAG, or ElonginB/C to assess if it would affect NOXA protein turnover in two independent cell lines. Second, in all the experiments designed to detect NOXA ubiquitylation in cells, the authors utilized immunoprecipitation (IP) with FLAG-NOXA/NOXA, followed by immunoblotting (IB) with HA-Ub. However, it is possible that the observed poly-Ub bands could be partly attributed to the ubiquitylation of other NOXA binding proteins. Therefore, the authors need to consider performing IP with HA-Ub and subsequently IB with NOXA. Alternatively, they could use Ni-beads to pull down all His-Ub-tagged proteins under denaturing conditions, followed by the detection of FLAG-tagged NOXA using anti-FLAG Ab. The authors are encouraged to perform one of these suggested experiments to exclude the possibility of this concern. Furthermore, an in vitro ubiquitylation assay is crucial to conclusively demonstrate that the polyubiquitylation of NOXA is indeed mediated by the CRL5WSB2 complex.

(3) In their attempt to map the binding regions between NOXA and WSB2, the authors utilized exogenous proteins of both WSB2 and NOXA. To strengthen their findings, it would be more convincing to perform IP with exogenous wt/mutant WSB2 or NOXA and subsequently perform IB to detect endogenous NOXA or WSB2, respectively. Additionally, an in vitro binding assay using purified proteins would provide further evidence of a direct binding between NOXA and WSB2.

eLife assessment

This study reports a fundamental observation concerning cell death regulation by the anti-apoptotic BCL2 family NOXA. The authors convincingly demonstrate that NOXA is destabilized through the interaction with WSB2, a substrate receptor in CRL5 ubiquitin ligase complex, sensitizing the cells to treatments. These are key findings for cell biologists and cancer researchers as they identified a new target impacting drug responsiveness in cancer therapies.

Reviewer #2 (Public Review):

Summary:

Exploring the DEP-MAP database and two drug-screen databases, the authors identify WSB2 as an interactor of several BCL2 proteins. In follow-up experiments, they show that CRL5/WSB2 controls NOXA protein levels via K48 ubiquitination following direct protein-protein interaction, and cell death sensitivity in the context of BH3 mimetic treatment, where WSB2 depletion synergizes with drug treatment.

Strengths:

The authors use a set of orthogonal methods across different model cell lines and a new WSB2 KO mouse model to confirm their findings. They also manage to correlate WSB2 expression with poor prognosis in prostate and liver cancer, supporting the idea that targeting WSB2 may sensitize cancers for treatment with BH3 mimetics.

Weaknesses:

The conclusions drawn based on the findings in cancer patients are very speculative, as regulation of NOXA cannot be the sole function of CRL5/WSB2 and it is hence unclear what causes correlation with patient survival. Moreover, the authors do not provide a clear mechanistic explanation of how exactly higher levels of NOXA promote apoptosis in the absence of WSB2. This would be important knowledge, as usually high NOXA levels correlate with high MCL1, as they are turned over together, but in situations like this, or loss of other E3 ligases, such as MARCH, the buffering capacity of MCL1 is outrun, allowing excess NOXA to kill (likely by neutralizing other BCL2 proteins it usually does not bind to, such as BCLX). Moreover, a necroptosis-inducing role of NOXA has been postulated. Neither of these options is interrogated here.

eLife assessment

This study presents an important computational tool for the quantification of the cellular composition of human tissues profiled with ATAC-seq. The methodology is solid and its application results on breast cancer tumor tissues are convincing. It advances existing methods by utilizing a comprehensive reference profile for major cancer-relevant cell types, compatible with a widely-used cell type deconvolution tool.

Reviewer #1 (Public Review):

Summary:

Building upon their famous tool for the deconvolution of human transcriptomics data (EPIC), Gabriel et al. implemented a new methodology for the quantification of the cellular composition of samples profiled with Assay for Transposase-Accessible Chromatin sequencing (ATAC-seq). To build a signature for ATAC-seq deconvolution, they first created a compendium of ATAC-seq data and derived chromatin accessibility marker peaks and reference profiles for 12 cell types, encompassing immune cells, endothelial cells, and fibroblasts. Then, they coupled this novel signature with the EPIC deconvolution framework based on constrained least-square regression to derive a dedicated tool called EPIC-ATAC. The method was then assessed using real and pseudo-bulk ATAC-seq data from human peripheral blood mononuclear cells (PBMC) and, finally, applied to ATAC-seq data from breast cancer tumors to show it accurately quantifies their immune contexture.

Strengths:

Overall, the work is of very high quality. The proposed tool is timely; its implementation, characterization, and validation are based on rigorous methodologies and results in robust estimates. The newly-generated, validation data and the code are publicly available and well-documented. Therefore, I believe this work and the associated resources will greatly benefit the scientific community.

Weaknesses:

In the benchmarking analysis, EPIC-ATAC was compared also to deconvolution methods that were originally developed for transcriptomics and not for ATAC-seq data. However, the authors described in detail the specific settings used to analyze this different data modality as robustly as possible, and they discussed possible limitations and ideas for future improvement.

Reviewer #2 (Public Review):

Summary:

The manuscript expands the current bulk sequencing data deconvolution toolkit to include ATAC-seq. The EPIC-ATAC tool successfully predicts accurate proportions of immune cells in bulk tumour samples and EPIC-ATAC seems to perform well in benchmarking analyses. The authors achieve their aim of developing a new bulk ATAC-seq deconvolution tool.

Strengths:

The manuscript describes simple and understandable experiments to demonstrate the accuracy of EPIC-ATAC. They have also been incredibly thorough with their reference dataset collections and have been robust in their benchmarking endeavours and measured EPIC-ATAC against multiple datasets and tools. This tool will be valuable to the community it serves.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Building upon their famous tool for the deconvolution of human transcriptomics data (EPIC), Gabriel et al. implemented a new methodology for the quantification of the cellular composition of samples profiled with Assay for Transposase-Accessible Chromatin sequencing (ATAC-Seq). To build a signature for ATAC-seq deconvolution, they first created a compendium of ATAC-seq data and derived chromatin accessibility marker peaks and reference profiles for 21 cell types, encompassing immune cells, endothelial cells, and fibroblasts. They then coupled this novel signature with the EPIC deconvolution framework based on constrained least-square regression to derive a dedicated tool called EPIC-ATAC. The method was then assessed using real and pseudo-bulk RNA-seq data from human peripheral blood mononuclear cells (PBMC) and, finally, applied to ATAC-seq data from breast cancer tumors to show it accurately quantifies their immune contexture.

Strengths:

Overall, the work is of very high quality. The proposed tool is timely; its implementation, characterization, and validation are based on rigorous methodologies and resulted in robust results. The newly-generated, validation data and the code are publicly available and well-documented. Therefore, I believe this work and the associated resources will greatly benefit the scientific community.

Weaknesses:

CA few aspects can be improved to clarify the value and applicability of the EPIC-ATAC and the transparency of the benchmarking analysis.

(1) Most of the validation results in the main text assess the methods on all cell types together, by showing the correlation, RMSE, and scatterplots of the estimated vs. true cell fractions. This approach is valuable for showing the overall method performance and for detecting systematic biases and noisy estimates. However, it provides very limited insights regarding the capability of the methods to estimate the individual cell types, which is the ultimate aim of deconvolution analysis. This limitation is exacerbated for rare cell types, which could even have a negative correlation with the ground truth fractions, but not weigh much on the overall RMSE and correlation. I would suggest integrating into the main text and figures an in-depth assessment of the individual cell types. In particular, it should be shown and discussed which cell types can be accurately quantified and which ones are less reliable.

We thank the reviewer for raising this important point. Discussing the accuracy of EPIC-ATAC in predicting individual cell-type proportions would indeed be valuable in the main text. We have updated the text as follows.

In the first version of our manuscript, we had a section called “T cell subtypes quantification reveals the ATAC-Seq deconvolution limits for closely related cell types” which highlighted that EPIC-ATAC shows low performances when predicting the proportions of cell types that are closely related, e.g., CD4+ T cell or CD8+ T cell subtypes. The section is now named “Accuracy of ATAC-Seq deconvolution is determined by the abundance and specificity of each cell type” and has been expanded to discuss the accuracy of EPIC-ATAC predictions within each major cell type.

To do so, we represented in Figure 5A the performances of EPIC-ATAC in each cell type present in the benchmarking datasets from Figures 3 and 4. Additionally, we have kept in the supplementary figures the details of the correlation values and RMSE values within each cell type and for each tool (Supplementary Figures 9 and 10). The following text has been added in the main text to describe these analyses:

“Accuracy of ATAC-Seq deconvolution is determined by the abundance and specificity of each cell type

To investigate the impact of cell type abundance on the accuracy of ATAC-Seq deconvolution, we evaluated EPIC-ATAC predictions in each major cell type separately in the different benchmarking datasets (Figure 5A). NK cells, endothelial cells, neutrophils or dendritic cells showed lower correlation values. These values can be explained by the fact that these cell types are low-abundant in our benchmarking datasets (Figure 5A). For the endothelial cells and dendritic cells, the RMSE values associated to these cell types remain low. This suggests that while the predictions of EPIC-ATAC might not be precise enough to compare these cell-type proportions between different samples, the cell-type quantification within each sample is reliable. For the NK cells and the neutrophils, we observed more variability with higher RMSE values in some datasets which suggests that the markers and profiles for these cell types might be improved. Supplementary Figures 9 and 10 detail the performances of each tool when considering each cell type separately in the PBMC and the cancer datasets. As for EPIC-ATAC, the predictions from the other deconvolution tools are more reliable for the frequent cell types.”

(2) In the benchmarking analysis, EPIC-ATAC is compared to several deconvolution methods, most of which were originally developed for transcriptomics data. This comparison is not completely fair unless their peculiarities and the limitations of tweaking them to work with ATAC-seq data are discussed. For instance, some methods (including the original EPIC) correct for cell-type-specific mRNA bias, which is not present in ATAC-seq data and might, thus, result in systematic errors.

We thank the reviewer for this comment and have updated the results and methods sections as follows:

We provide in the Materials and methods section, the paragraph “Benchmarking of the EPIC-ATAC framework against other existing deconvolution tools” which describes how each tool included in the benchmark was used in the ATAC-Seq context. We have added a reference to this section in the main text when introducing the first benchmarking analysis.

For each tool, the main changes consisted in: (i) replacing the initial RNA-Seq profiles and markers by the EPIC-ATAC reference profiles and markers and (ii) providing as input a bulk ATAC-Seq dataset with matched ATAC-Seq features (the same approach as the one used in EPIC-ATAC was considered, see answer to the next comment). Having reference profiles/markers and an ATAC-Seq bulk query with matched features was the only requirement of the different deconvolution models to be able to run on ATAC-Seq data with the default methods parameters, except for quanTIseq. Indeed, this method, like EPIC, corrects its estimations for cell-type-specific mRNA content bias. We have disabled this option for the bulk ATAC-Seq deconvolution.

We can however not exclude that a hyper parametrization of each tool could have helped to improve their current performances. Also, for RNA-Seq data deconvolution, some of the methods followed specific features filtering, e.g., the quanTIseq framework removes a manually curated list of noisy genes as well as aberrant immune genes identified in the TCGA data and ABIS uses immune-specific housekeeping genes. We can hypothesize that additional filtering could be explored for the ATAC-Seq deconvolution to improve the performance of the tools.

We have clarified these points in the results section when introducing the benchmarking, in the methods and in the discussion section.

(3) On a similar note, it could be made more explicit which adaptations were introduced in EPIC, besides the ad-hoc ATAC-seq signature, to make it applicable to this type of data.

In the first version of the manuscript, we described the changes brought to EPIC to perform bulk ATAC-Seq deconvolution in the Material and methods section in the paragraph “Running EPIC-ATAC on bulk ATAC-Seq data”.  We have moved and completed this paragraph in the results section before the description of the evaluation of EPIC-ATAC in different datasets. The paragraph is the following:

“EPIC-ATAC integrates the marker peaks and profiles into EPIC to perform bulk ATAC-Seq data deconvolution

The cell-type specific marker peaks and profiles derived from the reference samples were integrated into the EPIC deconvolution tool (Racle et al., 2017; Racle and Gfeller, 2020). We will refer to this ATAC-Seq deconvolution framework as EPIC-ATAC. To ensure the compatibility of any input bulk ATAC-Seq dataset with the EPIC-ATAC marker peaks and reference profiles, we provide an option to lift over hg19 datasets to hg38 (using the liftOver R package) as the reference profiles are based on the hg38 reference genome. Subsequently, the features of the input bulk matrix are matched to our reference profiles’ features. To match both sets of features, we determine for each peak of the input bulk matrix the distance to the nearest peak in the reference profiles peaks. Overlapping regions are retained and the feature IDs are matched to their associated nearest peaks. If multiple features are matched to the same reference peak, the counts are summed. Before the estimation of the cell-type proportions, we transform the data following an approach similar to the transcripts per million (TPM) transformation which has been shown to be appropriate to estimate cell fractions from bulk mixtures in RNA-Seq data (Racle et al., 2017; Sturm et al., 2019). We normalize the ATAC-Seq counts by dividing counts by the peak lengths as well as samples depth and rescaling counts so that the counts of each sample sum to 106. In RNA-Seq based deconvolution, EPIC uses an estimation of the amount of mRNA in each reference cell type to derive cell proportions while correcting for cell-type-specific mRNA bias. For the ATAC-Seq based deconvolution these values were set to 1 to give similar weights to all cell-types quantifications. Indeed ATAC-Seq measures signal at the DNA level, hence the quantity of DNA within each reference cell type is similar.”

(4) Given that the final applicability of EPIC-ATAC is on real bulk RNA-seq data, whose characteristics might not be completely recapitulated by pseudo-bulk samples, it would be interesting to see EPIC and EPIC-ATAC compared on a dataset with matched, real bulk RNA-seq and ATAC-seq, respectively. It would nicely complement the analysis of Figure 7 and could be used to dissect the commonalities and peculiarities of these two approaches.

We thank the reviewer for raising this important point. EPIC-ATAC will be applied to real bulk ATAC-Seq data and pseudobulk data cannot indeed fully recapitulate the bulk signals.  Recently, a dataset composed of more than 100 samples with matched bulk RNA-Seq, bulk ATAC-Seq as well as matched flow cytometry data has been published by Morandini and colleagues in GeroScience in November 2023. We thus retrieved these data to compare the predictions obtained by EPIC-ATAC on the bulk ATAC-Seq data and the predictions of the original version of EPIC on the bulk RNA-Seq data to the cell-type quantification obtained by flow cytometry. We also assessed whether both modalities could be complementary using a simple approach averaging the predictions obtained from both modalities. The results of these analyzes have been summarized in the Figure 7C and are described in the main text in the last paragraph of the paper:

“We compared the predictions obtained using each modality to the flow cytometry cell-type quantifications. EPIC-ATAC predictions were better correlated with the flow cytometry measures for some cell types (e.g., CD8+, CD4+ T cells, NK cells) while this trend was observed with the EPIC-RNA predictions in other cell types (B cells, neutrophils, monocytes) (Figure 7C). We then tested whether the predictions obtained from both modalities could be combined to improve the accuracy of each cell-type quantification. Averaging the predictions obtained from both modalities shows a moderate improvement (Figure 7C), suggesting that the two modalities can complement each other.”

Reviewer #2 (Public Review):

Summary:

The manuscript expands the current bulk sequencing data deconvolution toolkit to include ATAC-seq. The EPIC-ATAC tool successfully predicts accurate proportions of immune cells in bulk tumour samples and EPIC-ATAC seems to perform well in benchmarking analyses. The authors achieve their aim of developing a new bulk ATAC-seq deconvolution tool.

Strengths:

The manuscript describes simple and understandable experiments to demonstrate the accuracy of EPIC-ATAC. They have also been incredibly thorough with their reference dataset collections. The authors have been robust in their benchmarking endeavours and measured EPIC-ATAC against multiple datasets and tools.

Weaknesses:

Currently, the tool has a narrow applicability in that it estimates the percentage of immune cells in a bulk ATAC-seq experiment.

Comments:

(1) Has any benchmarking been done on the runtime of the tool? Although EPIC-ATAC seems to "win" in benchmarking metrics, sometimes the differences are quite small. If EPIC-ATAC takes forever to run, compared to another tool that is a lot quicker, might some people prefer to sacrifice 0.01 in correlation for a quicker running tool?

We thank the reviewer for raising this point that was not addressed in the manuscript. We have added a supplementary figure (Supplementary Figure 8) which represents the CPU time used by each tool. The figure shows that all the tools could be run in less than 20 seconds in average. This figure has been mentioned at the end of the benchmarking paragraphs.

(2) In Figure 3B the data points look a bit squashed in the bottom-left corner. Could the plot be replotted with the data point spread out? There also seems to be some inter-patient variability. Could the authors comment on that?

We have updated Figure 3B to increase the visibility of the dots in the bottom-left corner. To do so, we have limited the x and y axes to the maximum of the predicted proportions for the y axis and true proportions for the x axis.

We also acknowledge that the accuracy of the predictions varies across samples. In particular, one sample (Sample4, star shape on Figure 3B) exhibits larger discrepancies between EPIC-ATAC predictions and the ground truth. To understand the lower performance, we have visualized our marker peaks in the five PBMC samples (Figure below). Based on this visualization, we can see that Sample4 might be an outlier sample considering that its cellular composition is similar to that of Sample2 and Sample5, however this sample shows particularly high ATAC-Seq accessibility at the monocytes and dendritic markers. This can explain why EPIC-ATAC overestimates the proportions of the two populations in this case. We have added the previously mentioned figures as a Supplementary Figure (Supplementary Figure 2) and have described it in the results section in the paragraph “EPIC-ATAC accurately estimates immune cell fractions in PBMC ATAC-Seq samples”.

(3) Could the authors comment on the possibility of expanding EPIC-ATAC into more than a percentage prediction tool? Perhaps EPIC-ATAC could remove the immune cell signal from the bulk ATAC-seq data to "purify" the uncharacterised cells in silico, or generate pseudo-ATAC-seq tracks of the identified cell types.

We thank the reviewer for this interesting question. As suggested by the reviewer, one approach to purify bulk genomics data using the cell-type proportions estimated by a cell-type deconvolution tool is to subtract the weighted sum of the signal observed in the reference data, weights corresponding to the predicted proportions. We used this approach on the EPIC-ATAC predictions obtained from pseudobulks built from scATAC-Seq data from diverse cancer types coming from the Human Tumor Atlas Network (HTAN) (See also the answer of the first recommendation of Reviewer 1). This dataset allows us to compare for a relatively large number of samples (a maximum of 25  samples in a cancer type cohort) the purified signal to the true signal derived from the single-cell data. The results are presented in the figure below which shows that the correlations between the predicted and true signals are relatively good in most of the cancer types (blue boxplots). However, these correlation levels are lower than the ones obtained when comparing the signal obtained from the entire pseudobulk (red boxplots) with the true signal. This suggests that this purification approach leads to a signal that is less precise and accurate than the signal resulting from all cells mixtures.

Author response image 1.

Boxplots of the correlation values obtained from the comparison of the bulk signal and the ground truth signal from the uncharacterized cells in each sample (red) and from the comparison of the predicted signal and the ground truth signal from the uncharacterized cells in each sample (blue).

Also, note that in our simple approach, negative values can be obtained. The predicted signal will thus be difficult to interpret and to use in downstream analyses. Methods claiming to perform purification of bulk samples use more complex and dedicated algorithms. For example, Symphony (Burdziak et al., 2019) (cited in our introduction) uses single-cell RNA-Seq data in addition to the bulk chromatin accessibility data to infer cluster-specific accessibility profiles. Considering that EPIC was not designed for purification purposes, we decided not to include this analysis in the updated version of the manuscript.

Recommendations For The Authors:

Reviewer #1 (Recommendations For The Authors):

(1) The original EPIC had two different signatures for application to blood or tumor RNA-seq. It is not clear instead if EPIC-ATAC applies with the same signature and framework to any tissue and disease context. This aspect should be clarified in the text.

We thank the reviewer for raising this point which was not clear in the previous version of the manuscript. As in the original version of EPIC, in EPIC-ATAC two reference profiles and sets of markers are available, the PBMC reference and the TME reference. We used the PBMC reference profiles and markers to deconvolve the PBMC samples and the TME reference profiles and markers to deconvolve the cancer samples. We have clarified this point in the result section of the main text in the paragraph “ATAC-Seq data from sorted cell populations reveal cell-type specific marker peaks and reference profiles” as follows (added text underlined):

“The resulting marker peaks specific only to the immune cell types were considered for the deconvolution of PBMC samples (PBMC markers). For the deconvolution of tumor bulk samples, the lists of marker peaks specific to fibroblasts and endothelial cells were added to the PBMC markers. This extended set of markers was further refined based on the correlation patterns of the markers in tumor bulk samples from the diverse solid cancer types from The Cancer Genome Atlas (TCGA) (Corces et al., 2018), i.e., markers exhibiting the highest correlation patterns in the tumor bulk samples were selected using the findCorrelation function from the caret R package (Kuhn, 2008) (Figure 1, box 4, see the Material and methods, section 2). The latter filtering ensures the relevance of the markers in the TME context since cell-type specific TME markers are expected to be correlated in tumor bulk ATAC-Seq measurements (Qiu et al., 2021). 716 markers of immune, fibroblasts and endothelial cell types remained after the last filtering (defined as TME markers). Considering the difference in cell types and the different filtering steps applied on the PBMC and TME markers, we recommend to use the TME markers and profiles to deconvolve bulk samples from tumor samples and the PBMC markers and profiles to deconvolve PBMC samples.”

We also note that when running EPIC-ATAC using the PBMC markers and the TME markers independently to perform the deconvolution of the cancer datasets, we see that overall the use of the TME markers leads to a better performance (Figure below).

Figure legend: Correlation and RMSE values obtained when running EPIC-ATAC on each cancer dataset (points) using the PBMC (red) and the TME (blue) markers.

To demonstrate that the TME markers can be applied to different cancer types, we have completed the evaluation of EPIC-ATAC on tumor samples by considering an additional dataset: the Human Tumor Atlas Network (HTAN) single-cell multiomic (scRNA-Seq and scATAC-Seq) dataset. We have processed this dataset and built scATAC-Seq pseudobulks for 7 cancer types on which EPIC-ATAC was applied to. This analysis has been summarized in Figure 4 and Supplementary Figure 4 and shows that EPIC-ATAC is applicable in a diverse set of tissues.

(2) EPIC and EPIC-ATAC have a valuable feature, which is absent from most deconvolution methods: the estimation of unknown content. It would be informative for the users to understand from the benchmarking analysis whether this feature gives an advantage to EPIC-ATAC with respect to the other approaches.

Indeed, among the tools that we included in our benchmarking analysis, only EPIC-ATAC and quanTIseq enable users to predict the proportions of cells that are not present in the reference profiles, i.e., the uncharacterized cells. For the other tools we thus fixed the estimated proportions of uncharacterized cells to 0. This approach provides a clear and significant advantage to EPIC-ATAC and to quanTIseq. For this reason, we also provide a version of the benchmarking in which we exclude the uncharacterized cells and rescale the true and estimated cell-type proportions to sum to 1. In this second benchmarking approach, EPIC-ATAC still outperforms some of the other deconvolution tools.

We have clarified this point in the results section, in the paragraph “EPIC-ATAC accurately predicts fractions of cancer and non-malignant cells in tumor samples”.

(3) The selection of the most discriminative markers is very well described in the text and beautifully illustrated in Figure 2. However, it is unclear why UMAP plots are used to represent cell-type similarities and dissimilarities. Would a linear dimensionality reduction approach like PCA be already sufficient to show these groups, especially considering the not-so-extreme dimensionality of the underlying data? In addition, a statistic that could be also considered to compare clusters to the cell type labels in the two scenarios is the Adjusted Rand Index (ARI).

We thank the reviewer for this relevant comment. We initially used UMAP to facilitate the visualization of the different cell-type groups. However, it is true that the three first axes of the principal component analyses performed based on each set of marker peaks already capture most of the structure in the data and that the use of UMAP can lead to an artificial enhancement of separation between the different groups of cells. We have updated Figure 2B by replacing the UMAP scatter plots by 3D representations of the first three principal components of the PCA and have added in Supplementary Figure 1B the pairwise scatter plots of these first 3 principal components. On the main figures, we have also added the ARI metric comparing the cell-type annotation and the clustering obtained using the first 10 axes of the PCA and model based clustering.

(4) In the introduction, it is stated that "the reasonable cost and technical advantages of these protocols foreshadow an increased usage of ATAC-Seq in cancer studies". I would suggest adding a reference to justify this trend. Also, it should be discussed how ATAC-seq deconvolution compares to other types of deconvolution approaches applied to cheaper epigenetic data like methylation one (e.g. epidish, methylcc, tca, minfi).

We have complemented this sentence with two references to justify the assertion: (i) a review published by Luo, Gribskov and Wang in 2022 showing the increasing number of ATAC-Seq studies in the field of cancer research, and (ii) a protocol paper from Grandi et al. published in 2022 on the state-of-the-art Omni-ATAC protocol for ATAC-sequencing which discusses the broad applicability and the technical advantages of ATAC-sequencing. Also in the preceding sentence, a recent ATAC-Seq protocol that can be applied to FFPE samples has been mentioned, FFPE samples being the most common samples in clinical cancer research.

We agree with the reviewer on the fact that other epigenetic assays such as methylation assays are cost effective. However, ATAC-sequencing provides additional information on the epigenetic landscape of a sample’s genome and some questions regarding regulatory regions and transcription factor activity cannot be answered with methylation data. Methods that can be applied on ATAC-Seq data specifically are thus needed. Most of the cell-type deconvolution algorithms existing so far are applicable on RNA-Seq or methylation data. These algorithms often use similar methodological concepts, e.g., linear combination of the reference profiles for reference-based methods, which could be used in different modalities. However, methylation-based deconvolution tools often take as input a data format that is specific to methylation data, e.g., two color micro array data (RGChannelSet R object) for the minfi deconvolution function (estimatesCellCounts) or leverage methylation-specific information to perform the deconvolution. For example, methylCC uses a model based on latent variables representing a binarized measures of the methylation status of cell-type specific regions (1 or 0 for clearly methylated or unmethylated regions). Such methods are more difficult to adapt than tools  based on RNA-Seq data where the signal is quantified using read counts similarly to ATAC-Seq data.

Nevertheless, some methods such as EPIdish or MethylCIBERSORT have proposed new methylation reference profiles and have used existing models that are not specific to methylation data to deconvolve the bulk data. In our work, we followed a similar approach where we propose new reference profiles specific to chromatin accessibility data, integrate them to an existing method EPIC as well as test them in other existing tools. Note that methylation reference profiles cannot be directly used for ATAC-Seq data deconvolution considering that methylation measures methylation status at CpG sites (dinucleotides) and ATAC-Seq measures the accessibility of regions of hundreds base pairs.

An analysis comparing the performance of methylation-based deconvolution and ATAC-Seq based deconvolution would be informative. However, such analysis is beyond the scope of our paper considering that none of the datasets used for our benchmarking provide these two modalities for the same samples.

In the manuscript, we have completed the references associated to the methylation-based deconvolution tools with the ones mentioned in the previous paragraphs and by the reviewer and have completed the discussion as follows:

“The comparison of EPIC-ATAC applied on ATAC-Seq data with EPIC applied on RNA-Seq data has shown that both modalities led to similar performances and that they could complement each other. Another modality that has been frequently used in the context of bulk sample deconvolution is methylation. Methylation profiling techniques such as methylation arrays are cost effective (Kaur et al., 2023) and DNA methylation signal is highly cell-type specific (Kaur et al., 2023; Loyfer et al., 2023). Considering that methylation and chromatin accessibility measure different features of the epigenome, additional analyses comparing and/or complementing ATAC-seq based deconvolution with methylation-based deconvolution could be of interest as future datasets profiling both modalities in the same samples become available.”

(5) In the Results section, some methodological steps could be phrased in a bit more extensive way to let the reader understand the rationale and the actual approach. I recognize there is also a reference to the Methods section, where all methodologies are reported in detail, but some of the sentences are hard to understand due to their synthetic format, e.g.: "markers with potential residual accessibility in human tissues were then filtered out".

We thank the reviewer for this comment and we have followed his recommendation to expand sentences with a synthetic format. Text changes and additions are underlined below:

“To limit batch effects, the collected samples were homogeneously processed from read alignment to peak calling. For each cell type, we derived a set of stable peaks observed across samples and studies, i.e. for each study, peaks detected in at least half of the samples were considered, and for each cell type, only peaks detected jointly in all studies were kept (see Materials and Methods, section 1).”

“To filter out markers that could be accessible in other human cell-types than those included in our reference profiles, we used the human atlas study (K. Zhang et al., 2021), which identified modules of open chromatin regions accessible in a comprehensive set of human tissues, and we excluded from our marker list the markers overlapping these modules (Figure 1, box 3, see Materials and Methods section 2).”

“For the deconvolution of tumor bulk samples, the lists of marker peaks specific to fibroblasts and endothelial cells were added to the PBMC markers. This extended set of markers was further refined based on the correlation patterns of the markers in tumor bulk samples from the diverse solid cancer types from The Cancer Genome Atlas (TCGA) (Corces et al., 2018), i.e., markers exhibiting the highest correlation patterns in the tumor bulk samples were selected using the findCorrelation function from the caret R package (Kuhn, 2008)  (Figure 1, box 4, see the Material and methods, section 2).”

Also, following the comments and recommendations of the Reviewer 1, we have: (i) moved the method section describing the adaptation of EPIC to ATACseq data to provide more details in the results section (see answer to the third comment of Reviewer 1), (ii) clarified how the existing tools used in the benchmarking analyses were adapted for ATAC-Seq deconvolution (see answer to the second comment of Reviewer 1), and (iii) detailed how the comparison between our estimations of the infiltration levels in the samples from Kumegawa et al. and the estimations from the original study was performed (see answer to the seventh recommendation of Reviewer 1).

(6) In the main text, it is stated that "the list of markers was further refined based on the correlation patterns of the markers in tumor bulk samples from diverse cancer types from The Cancer Genome Atlas". It should be clarified if these are only solid cancers, or if blood cancers were also used.

We have considered only the solid cancers and have clarified this point in the results section: “This extended set of markers was further refined based on the correlation patterns of the markers in tumor bulk samples from the diverse solid cancer types from The Cancer Genome Atlas”.

(7) When reporting that "these predictions are consistent with the infiltration level estimations reported in the original publication", it should be mentioned how the infiltration levels were quantified in this publication and how this agreement was quantified. This would be important also to claim in the abstract that "EPIC-ATAC accurately infers the immune contexture of the main breast cancer subtypes".

We thank the reviewer for this comment, we acknowledge that the agreement between the EPIC-ATAC predictions and the infiltration levels quantified in the original publication should be further described in the paper. We have expanded the text in the results section in the paragraph “EPIC-ATAC accurately infers the immune contexture in a bulk ATAC-Seq breast cancer cohort” to clarify this point. Additionally, we have added a panel in Figure 6 (panel A) which shows a good agreement between EPIC-ATAC predictions and the metric used in the original paper to evaluate the infiltration levels of different cell types.

The added text is underlined below:

“We applied EPIC-ATAC to a breast cancer cohort of 42 breast ATAC-Seq samples including samples from two breast cancer subtypes, i.e., 35 oestrogen receptor (ER)-positive human epidermal growth factor receptor 2 (HER2)-negative (ER+/HER2-) samples and 7 triple negative (TNBC) tumors (Kumegawa et al., 2023). No cell sorting was performed in parallel to the chromatin accessibility sequencing. For this reason, the authors used a set of cell-type-specific cis-regulatory elements (CREs) identified in scATAC-Seq data from similar breast cancer samples (Kumegawa et al., 2022) and estimated the amount of infiltration of each cell type by averaging the ATAC-Seq signal of each set of cell-type-specific CREs in their samples. We used EPIC-ATAC to estimate the proportions of different cell types of the TME. These predictions were then compared to the metric used by Kumegawa and colleagues in their study to infer levels of infiltration. A high correlation between the two metrics was observed for each cell type (Pearson’s correlation coefficient from 0.5 for myeloid cells to 0.94 for T cells, Figure 6A).”  

(8) It should be made explicit if EPIC-ATAC quantifies mDC, pDC, or their sum.

In our collection of reference ATAC-Seq samples from which the markers and profiles have been derived, mDCs and pDCs were both included in the dendritic cells.  EPIC-ATAC thus quantifies the total amount of dendritic cells, i.e., mDCs and pDCs included. We have added a sentence in the main text to clarify this point:

To identify robust chromatin accessibility marker peaks of cancer relevant cell types, we collected 564 samples of sorted cell populations from twelve studies including eight immune cell types (B cells […] dendritic cells (DCs) (mDCs and pDCs are grouped in this cell-type category) […] and  endothelial (Liu et al., 2020; Xin et al., 2020) cells (Figure 1 box 1, Figure 2A, Supplementary Table 1).

Reviewer #2 (Recommendations For The Authors):

The authors should double-check the naming of tools is done correctly e.g. ChIPSeeker has been spelled incorrectly in some instances throughout the manuscript.

We thank the reviewer for pointing out this mistake and have corrected the mistake in the main text.

eLife assessment

This important study combines experiment and theory to examine how the intrinsic physiological properties of neurons involved in orchestrating birdsong are related to the temporal structure of song. Intrinsic properties determine how neurons respond to inputs, and in this manuscript, the authors describe rules that connect these intrinsic properties to a learned behaviour, the learned song of an adult songbird. The experimental data are convincing and the computational model builds on a robust and well-validated biophysical framework. Although some key points of the model could be established more strongly, the evidence supporting the idea that song temporal structure is related to intrinsic physiology is solid and this research will be of general interest to researchers in the field and neurophysiologists.

Reviewer #1 (Public Review):

Summary:

Previous research from the Margoliash laboratory has demonstrated that the intrinsic electrophysiological properties of one class of projection neurons in the song nucleus HVC, HVCX neurons, are similar within birds and differ between birds in a manner that relates to the bird's song. The current study builds on this research by addressing how intrinsic properties may relate to the temporal structure of the bird's song and by developing a computational model for how this can influence sequence propagation of activity within HVC during singing.

First, the authors identify that the duration of the song motif is correlated with the duration of song syllables and particularly the length of harmonic stacks within the song. They next found positive correlations between some of the intrinsic properties, including firing frequency, sag ratio, and rebound excitation area with the duration of the birds' longest harmonic syllable and some other measure of motif duration. These results were extended by examining measures of firing frequency and sag ratio between two groups of birds that were experimentally raised to learn songs that only differed by the addition of a long terminal harmonic stack in one of the groups. Lastly, the authors present an HH-based model elucidating how the timing and magnitude of rebound excitation of HVCX neurons can function to support previously reported physiological network properties of these neurons during singing.

Strengths:

By trying to describe how intrinsic properties (IPs) may relate to the structure of learned behavior and providing a potentially plausible model (see below for more on this) for how differences in IPs can relate to sequence propagation in this neural network, this research is addressing an important and challenging issue. An understanding of how cell types develop IPs and how those IPs relate to the function and output of a network is a fundamental issue. Tackling this in the zebra finch HVC is an elegant approach because it provides a quantifiable and reliable behavior that is explicitly tied to the neurons that the authors are studying. Nonetheless, this is a difficult problem, and kudos to the authors for trying to unravel this.

Correlations between harmonic stack durations and song durations are well-supported and interesting. This provides a new insight that can and will likely be used by other research groups in correlating neuronal activity patterns to song behavior and motif duration. Additionally, correlations between IPs associated with rebound excitation are also well supported in this study.

The HH-model presented is important because it meaningfully relates how high or low rebound excitation can set the integration time window for HVCX neurons. Further, the synaptic connectivity of this model provides at least one plausible way in how this functions to permit the bursting activity of HVCX neurons during singing (and potentially during song playback experiments in sleeping birds). Thus, this model will be useful to the field for understanding how this network activity intersects with 'learned' IPs in an important class of neurons in this circuit.

Weaknesses:

The main weakness of the study is that there is somewhat of a disconnect between the physiological measurements described and the key components of the circuit model presented at the end of the paper. Thus, better support could be provided to link the magnitude of rebound excitation with song temporal structure. The rebound excitation area is shown to be positively correlated with the longest harmonic stack. Does this correlation hold when the four birds with unusually long stacks (>150ms) are excluded? Is rebound excitation area positively correlated with motif duration? Additionally, rebound excitation was not considered when examining experimentally tutored birds. Further analysis of these correlations can better link this research to the model presented.

The HH model is of general interest, but I am concerned about the plausibility of some of this circuitry, particularly because synaptic connectivity underlying information flow is a central component of the model. At several steps in the model, excitatory drive onto HVCX neurons is coming from another HVCX neuron. Although disynaptic inhibition between HVCX neurons and between HVCRA and HVCX neurons is well established, I am not aware of any data indicating direct synaptic connections between HVCX neurons.

Thus, how does the model change if all excitatory drive onto HVCX neurons are coming from HVCRA neurons? Currently, the model is realized through neurons being active at syllable or gesture transitions. What does the model predict about the distribution of HVCRA neurons activity across songs if they are the exclusive excitatory input to HVCX neurons? A better consideration of these issues can improve the suitability of the model in the context of known connectivity.

If I understand the model and ideas correctly, birds with longer motifs should exhibit longer pauses in the activity of tonically active HVC interneurons during singing and they should exhibit longer post-rebound integration windows. Experimental evidence supporting either of these ideas is not provided and would strengthen this research.

Reviewer #2 (Public Review):

Intrinsic properties of a neuron refer to the ion channels that a neuron expresses. These ion channels determine how a neuron responds to its inputs. How intrinsic properties link to behavior remains poorly understood. Medina and Margoliash address this question using the zebra finch, a well-studied songbird. Previous studies from their lab and other labs have shown that the intrinsic properties of adult songbird basal-ganglia projecting premotor neurons, are more similar within a bird than across birds. Across birds, this similarity is related to the extent of similarity in the songs; the more similar the song between two birds, the more similar the intrinsic properties between the neurons of these two birds. Finally, the intrinsic properties of these neurons change over the course of development and are sensitive to intact auditory feedback. However, the song features that relate to these intrinsic properties and the function of the within-bird homogeneity of intrinsic properties are unclear.

In this manuscript, the authors address these two questions by examining the intrinsic properties of basal-ganglia projecting premotor neurons in zebra finch brain slices. Specifically, they focus on the Ih current (as this is related to rhythmic activity in many pattern-generating circuits) and correlate the properties of the Ih current with song features. They find that the sag ratio (a measure of the driving force of the Ih current) and the rebound area (a measure of the post-inhibitory depolarisation) are both correlated with the temporal features of the song. First, they show the presence of correlations between the length of the song motif and the length of the longest syllable (most often a harmonic stack syllable). Based on this, they conclude that longer song motifs are composed of longer syllables. Second, they show that HVCX neurons within a bird have more similar sag ratios and rebound areas than across birds. Third, the mean sag ratio and mean rebound areas across birds were correlated with the duration of the longest harmonic stack within the song. These two results suggest that IPs are correlated with the temporal structure of the song. To further test this, the authors used natural and experimental tutoring procedures to have birds that learned two different types of songs that only differed in length; the longer song had an extra harmonic stack at the end. Using these two sets of birds, the authors find larger sag ratios and higher firing frequencies in birds with longer songs. Fifth, they show that the post-inhibitory rebound area allows neurons to respond to excitatory inputs and produce spikes. Neurons with a larger rebound area have a larger time window for responding to excitatory inputs. Based on this, they speculate that HVCX neurons with larger rebound areas integrate over larger time windows. Finally, they make a network model of HVC and show that one specific model could explain sequence-specific bursting of HVCX neurons.

Strengths

The question being addressed is an interesting question and the authors use appropriate techniques. The authors find a new temporal structure within the song, specifically, they find that longer songs typically have more syllables and longer syllables. As far as I know, this has not been shown earlier. The authors build on existing literature to suggest that IPs of HVCX neurons are correlated with the temporal structure of songs.

Weaknesses

I have a number of concerns with the statistics and interpretation of the results, insufficient controls for one experiment, and the specifics of the model that affect the implications of these results. These concerns are listed in the recommendations for the authors.

Reviewer #3 (Public Review):

It is rare to find systems in neuroscience where a detailed mechanistic link can be made between the biophysical properties of individual neurons and observable behaviors. In this study, Medina and Margoliash examined how the intrinsic physiological properties of a subclass of neurons in HVC, the main nucleus orchestrating the production of birdsong, might have an effect on the temporal structure of a song. This builds on prior work from this lab demonstrating that intrinsic properties of these neurons are highly consistent within individual animals and more similar between animals with similar songs, by identifying specific acoustic features of the song that covary with intrinsic properties and by setting forth a detailed biophysical network model to explain the relationship.

The main experimental finding is that excitability, hyperpolarization-evoked sag, and rebound depolarization are correlated with song duration and the duration of long harmonic elements. This motivates the hypothesis that rebound depolarization acts as a coincidence detector for the offset of inhibition associated with the previous song element and excitation associated with the start of the next element, with the delay and other characteristics of the window determined primarily by Ih. The idea is then that the temporal sensitivity of coincidence detection, which is common to all HVCx neurons, sets a global tempo that relates to the temporal characteristics of a song. This model is supported by some experimental data showing variation in the temporal integration of rebound spiking and by a Hodgkin-Huxley-based computational model that demonstrates proof of principle, including the emergence of a narrow (~50 ms) post-inhibitory window when excitatory input from other principal neurons can effectively evoke spiking.

Overall, the data are convincing and the model is compelling. The manuscript plays to the strengths of zebra finch song learning and the well-characterized microcircuitry and network dynamics of HVC. Of particular note, the design for the electrophysiology experiments employed both a correlational approach exploiting the natural variation in zebra finch song and a more controlled approach comparing birds that were tutored to produce songs that differed primarily along a single acoustical dimension. The modeling is based on Hodgkin-Huxley ionic conductances that have been pharmacologically validated, and the connections and functional properties of the network are consistent with prior work. This makes for a level of mechanistic detail that will likely be fruitful for future work.

There are some minor to moderate weaknesses. A minor weakness in the analysis of the experimental data relates to the handling of multiple correlations. There are several physiological variables that covary and several acoustical variables that covary, which makes it difficult to interpret standard Pearson correlation coefficients between any two individual variables. This is a minor concern because the results of the correlational analysis were confirmed in separate experiments with controlled tutoring, but a partial correlation analysis or latent factor analysis would be a more rigorous way of analyzing the natural live tutoring data.

eLife assessment

This preprint explores the involvement of cyclic di-GMP in genome stability and antibiotic persistence regulation in bacterial biofilms. The authors proposed a novel mechanism that, due to bacterial adhesion, increases c-di-GMP levels and influences persister formation through interaction with HipH. While the work may provide useful insights that could attract researchers in biofilm studies and persistence mechanisms, the main findings are inadequately supported and require further validation and refinement in experimental design.

Reviewer #1 (Public Review):

The authors propose a UPEC TA system in which a metabolite, c-di-GMP, acts as the AT with the toxin HipH. The idea is novel, but several key ideas are missing in regard to the relevant literature, and the experimental design is flawed. Moreover, they are absolutely not studying persister cells as Figure 1b clearly shows they are merely studying dying cells since no plateau in killing (or anything close to a plateau) was reached. So in no way has persistence been linked to c-di-GMP. Moreover, I do not think the authors have shown how the c-di-GMP sensor works. Also, there is no evidence that c-di-GMP is an antitoxin as no binding to HipH has been shown. So at best, this is an indirect effect, not a new toxin/antitoxin system as for all 7 TAs, a direct link to the toxin has been demonstrated for antitoxins.

Weaknesses:

(1) L 53: biofilm persisters are no different than any other persisters (there is no credible evidence of any different persister cells) so this reviewer suggests changing 'biofilm persisters' to 'persisters' throughout the text.

(2) L 51: persister cells do not mutate and, once resuscitated, mutate like any other growing cell so this sentence should be deleted as it promotes an unnecessary myth about persistence.

(3) L 69: please include the only metabolic model for persister cell formation and resuscitation here based on single cells (e.g., doi.org/10.1016/j.bbrc.2020.01.102 , https://doi.org/10.1016/j.isci.2019.100792 ); otherwise, you write as if there are no molecular mechanisms for persistence/resuscitation.

(4) The authors should cite in the Intro or Discussion that others have proposed similar novel TAs including a ppGpp metabolic toxin paired with an enzymatic antitoxin SpoT that hydrolyzes the toxin (http://dx.doi.org/10.1016/j.molcel.2013.04.002).

(5) Figure 1b: there are no results in this paper related to persister cells. Figure 1b simply shows dying cells were enumerated. Hence, the population of stressed cells increased, not 'persister cells' (Figure 1f), in the course of these experiments.

(6) Figure S1: I see no evidence that the authors have shown this c-di-GMP detects different c-di-GMP levels since there appears to be no data related to varying c-di-GMP concentrations with a consistent decrease. Instead, there is a maximum. What are the concentration of c-di-GMP on the X-axis for panels C, D, and E? How were c-di-GMP levels varied such that you know the c-di-GMP concentration?

(7) The viable portion of the VBNC population are persister cells so there is no reason to use VBNC as a separate term. Please see the reported errors often made with nucleic acid staining dyes in regard to VBNCs.

Reviewer #2 (Public Review):

Summary:

Hebin et al reported a fascinating story about antibiotic persistence in the biofilms. First, they set up a model to identify the increased persisters in the biofilm status. They found that the adhesion of bacteria to the surface leads to increased c-di-GMP levels, which might lead to the formation of persisters. To figure out the molecular mechanism, they screened the E.coli Keio Knockout Collection and identified the HipH. Finally, the authors used a lot of data to prove that c-di-GMP not only controls HipH over-expression but also inhibits HipH activity, though the inhibition might be weak.

Strengths:

They used a lot of state-of-the-art technologies, such as single-cell technologies as well as classical genetic and biochemistry approaches to prove the concept, which makes the conclusions very solid. Overall, it is a very interesting and solid story that might attract diverse readers working with c-di-GMP, persisters, and biofilm.

Weaknesses:

(1) Is HipH the only target identified by screening the E.coli Keio Knockout Collection?

(2) Since the story is complicated, a diagrammatic picture might be needed to illustrate the whole story. And the title does not accurately summarize the novelty of this study.

(3) The ratio of mVenus NB to mScarlet-I (R) negatively correlates with the concentration of c-di-GMP. Therefore, R -1 demonstrates a positive correlation with the concentration of c-di-GMP. Is this method validated with other methods to quantify c-di-GMP， or used in other studies？

(4) References are missing throughout the manuscript. Please add enough references for every conclusion.

(5) The novelty of this study should be clearly written and compared with previous references. For example, is it the first study to report the mechanism that the adhesion of bacteria to the surface leads to increased persister formation?

(6) in vitro DNA cleavage assay. Why not use bacterial genomic DNA to test the cleaving of HipH on the bacterial genome?

(7) C-di-gmp -HipH is not a TA, it does not fit in the definition of the TA systems. You can say C-di-gmp is an antitoxin based on your study, but C-di-gmp -HipH is not a TA pair.