Author Response

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

Reviewer #1 (recommendations for the authors):

The following are comments that the authors may wish to address or clarify:

(1) The claim that respiration and fermentation occur concurrently in the agr mutant during aerobic growth is not strongly supported by the evidence presented…. However, since neither lactate production nor a difference in the NAD+/NADH ratio between the wild type and agr mutant was observed, it is challenging to assert that fermentation is occurring. Relying solely on a gene expression signature indicative of fermentation is, in my view, inadequate to conclusively establish that aerobic fermentation is taking place.

Lactate production. The data we provide in Figure 5-E of the original manuscript (Figure 5-C in the revised manuscript) indicates that lactate production is lower in the wild-type compared to the Δagr mutant.

The exact focus of Reviewer 1’s concern is not clearly specified, but may have been referring to how the result was described in the text:

“Although the stimulatory effect of the agr deletion on production of the fermentation product lactate was not observed in optimally aerated broth cultures after growth to late exponential growth phase, it was confirmed for organisms grown in broth under more metabolically demanding, suboptimal aeration conditions (Figure 5E). Overall, these results are consistent with transcription-level up-regulation of respiratory and fermentative pathways in agr-deficient strains.”

The greater sensitivity of suboptimal aeration conditions is unsurprising and relates to a low rate of fermentation during the vigorous aeration (shaking at 250 rpm) conditions commonly used to grow S. aureus. To clarify the point, we modified the text to provide additional context as follows:

Line 271: “Although the stimulatory effect of the agr deletion on production of the fermentation product lactate was not observed in optimally aerated broth cultures after growth to late exponential growth phase, it was confirmed for organisms grown in broth under more metabolically demanding, suboptimal aeration conditions (limitations in the rate of respiration when oxygen is limiting are expected to increase overall levels of fermentation) (Figure 5C). Overall, these results are consistent with transcription-level up-regulation of respiratory and fermentative pathways in agr-deficient strains.” NAD+/NADH ratio. Extended studies of the NAD+/NADH ratio, requested by Reviewer 1 under Comments 12 and 13, document an effect of the Δagr mutant not seen in Figure 5F in the original submission. Our responses to Comments 12 and 13 below address this issue.

(2) The mechanisms through which the ΔagrΔrot double mutant resists H2O2 are not clearly elucidated. While the authors suggest that the ΔagrΔrot double mutant expresses several genes involved in combating oxidative stress, essential genetic studies that would validate this hypothesis have not been conducted.

The data we provide indicate 1) that wild-type strains are tolerant to peroxide and 2) that wild-type strains are able to render inducible several known reactive oxygen species (ROS)-protective genes in the presence of peroxide in a rot-dependent manner. Δagr strains, which do not demonstrate this response, are more readily killed by peroxide. Additional data indicate that increased respiration caused by deletion of agr is associated with increased endogenous ROS. Higher levels of endogenous ROS can modulate tolerance to subsequent challenge by ROS (1). Collectively, these observations support a model of Δagr-induced hyper-susceptibility in which elevation of endogenous ROS results in a suboptimal ROS-defense response that plays a role in increased peroxide lethality.

We prefer to test this model in future studies directed at understanding the complexities of the interaction among agr-mediated tolerance, endogenous ROS levels, and induction of protective responses in S. aureus. Culprit protective genes, alone and in various combinations, will be inactivated in Δagr mutant and wild-type strains, tested in killing assays with and without agents that mitigate endogenous ROS, and subjected to RNAseq, proteomic, and metabolomic analyses, as part of a larger program to identify factors involved in S. aureus tolerance to lethal stress.

To clarify the issue raised by the reviewer we altered the wording in the following sentences as follows:

Line 335: “Elevated expression of protective genes suggests that the double mutant survives damage from H2O2 better because protective genes are rendered inducible (loss of Rot-mediated repression).”

Line 440: “Details of agr-mediated protection are sketched in Figure 10. At low levels of ROS, agr is activated by a redox sensor in AgrA, RNAIII is expressed and represses the Rot repressor, thereby rendering protective genes (e.g., clpB/C, dps) inducible via an unknown mechanism (induction, candidate protective gene(s), and their connection to endogenous ROS levels are being pursued, independent of the current report).

(3) The reason behind the agr mutant's low metabolic efficiency, as evidenced by low levels (Fig 5A) despite enhanced respiration and acetate production, is not clearly explained. Could insights from the modeling shed light on why the ATP levels are low in the agr mutant?

Comparative modeling of central metabolic pathways, in combination with in vitro metabolic analyses of Δagr and wild-type strains, revealed the metabolic inefficiency but cannot explain it. The basis for the metabolic inefficiency conferred by agr inactivation is unknown. The possibility that aberrant sorting of cell wall surface proteins could lead to metabolic inefficiency was raised in the Discussion where we wrote:

“Our work supports this idea by showing that increased respiration caused by deletion of agr is associated with increased ROS-mediated lethality. The basis for the metabolic inefficiency conferred by agr inactivation is unknown. Given that Δagr mutants are unable to downregulate surface proteins during stationary phase (2, 3), it is possible that deletion of agr perturbs the cytoplasmic membrane or the machinery that sorts proteins across the cell wall. In support of this notion, jamming SecY translocation machinery of E. coli results in downstream events shared with antibiotic lethality, including accelerated respiration and accumulation of ROS (4). In this scenario, the formation of a futile macromolecular cycle may accelerate cellular respiration to meet the metabolic demand of unresolvable problems caused by elevated surface sorting.”

For clarification, we modified the text as follows:

Line 461: “Our work supports this idea by showing that increased respiration caused by deletion of agr is associated with increased ROS-mediated lethality. How agr deficiency is connected to the corruption of downstream processes that result in metabolic inefficiency and increased endogenous ROS levels is unknown. Given that Δagr mutants are unable to downregulate surface proteins during stationary phase (2, 3), it is possible that deletion of agr perturbs the cytoplasmic membrane or the machinery that sorts proteins across the cell wall.”

agr has been linked to defects in peptidoglycan autolysis (5). Cho et al. (2019) found that β-lactam treatment can induce a futile cycle of peptidoglycan synthesis and degradation that has been linked to increased production of endogenous ROS (6). Thus, an alternative, nonmutually exclusive route to a futile cycle and elevated endogenous ROS levels in agr-deficient cells other than surface protein dysregulation may be via decreased cell wall cross-linking. We prefer not to include this and other speculations, because they are not necessary or revealing and because they would detract from the manuscript by disrupting its sense of narrative and brevity.

(4) The observation that menadione can protect the agr mutant from H2O2 is perplexing. The authors propose that even though menadione generates superoxide through redox cycling, this superoxide might inhibit the TCA cycle, thereby restricting respiration, which could be advantageous for the agr mutant. To substantiate this hypothesis, it would be imperative to demonstrate that a double mutant ΔagrΔacnA exhibits long-lived protection against H2O2.

Rowe et al. (2020) definitively showed that a burst of menadioneassociated ROS inactivates the TCA cycle in S. aureus, leading to reduced respiration and ATP production (7). Both aconitase activity and ATP levels in menadione-treated cultures were complemented by the antioxidant N-acetyl cysteine. In the present work we demonstrate, using the same experimental conditions as Rowe et al., that menadione protected the Δagr mutant from peroxide killing but had little effect on the wild-type strain. Addition of N-acetyl cysteine in the presence of menadione restored H2O2 susceptibility to the Δagr mutant and had no effect on the wild-type. Collectively, these observations support the idea that menadione inactivates the TCA cycle, leading to reduced respiration, and increased protection of the Δagr mutant from peroxide killing.

As requested, we tested whether the ΔagrΔacnA double mutant exhibits protection against H2O2. The new data we now provide (Figure 8—figure supplement 2A) show that a ΔacnA mutation completely protected the Δagr mutant from peroxide killing after growth to late exponential growth phase, but it had little if any effect on the wild-type strain. To evaluate long-lived protection, we compared survival rates of ΔagrΔacnA mutant and Δagr cells following dilution of overnight cultures and regrowth prior to challenge with H2O2, which revealed partial protection of the Δagr mutant (Figure 8— figure supplement 2B).

We explained these results with the following:

Line 351: “Rowe et al. (2020) showed that menadione exerts its effects on endogenous ROS by inactivating the TCA cycle in S. aureus. To determine whether this mechanism can also induce protection in the Δagr mutant, we inactivated the TCA cycle gene acnA in wild-type and Δagr strains (Figure 8—figure supplement 2). We found that ΔacnA mutation completely protected the Δagr mutant from peroxide killing after growth to late exponential growth phase but had little effect on the wild-type strain. This finding supports the idea that TCA cycle activity contributes to an imbalance in endogenous ROS homeostasis in the Δagr mutant, and that this shift is a critical factor for Δagr hyperlethality. When we evaluated long-lived protection by comparing survival rates of ΔagrΔacnA mutant and Δagr cells following dilution of overnight cultures and regrowth prior to challenge with H2O2, ΔacnA remained protective, but less so (Figure 8—figure supplement 2). These partial effects of an ΔacnA deficiency suggest that Δagr stimulates long-lived lethality for peroxide through both TCA-dependent and TCA-independent pathways.”

(5) Figure 10 presents a model suggesting that Rot-mediated repression of respiration is essential for long-lasting resistance to H2O2 lethality. However, the connection between decreased respiration and long-lived resistance to ROS is not evident, especially considering that the respiration rate varies over the growth phase and does not seem to align with the long-lived and steady protection provided by agr. However, the authors could investigate this by examining whether inactivating qox in the agr mutant restores its resistance to H2O2. The experiments with menadione are not particularly persuasive, as menadione could have additional effects on the cells that are not accounted for.

As requested, we tested whether the ΔagrΔqoxC double mutant exhibits protection against H2O2. qox deficiency was hyperlethal in wild-type and Δagr strains, even with the lowest concentration of H2O2 used in our assay. Indeed, surviving cells were undetectable, precluding comparison of survival differences between wild-type and Δagr mutant strains. This striking finding can be explained by prior work highlighting the profound and pleotropic effects of qox deficiency on metabolism that involve not only control of respiration but also participation in other physiological processes such as cell growth and morphological differences. For example, in Bacillus, qox deficiency decreases TCA cycle flux and increases overflow metabolism (8). Additionally, we confirmed prior work in S. aureus showing that qox deficiency decreases growth rate and yield (9, 10), dramatically increases production of pigment that functions as an oxidation shield, and decreases hemolytic activity (11). Moreover, we found that that qox deficiency results in a striking increase (~150%) in endogenous ROS in both wild-type and agr mutant cells, likely explaining the hyperlethality phenotype. Thus, interpretation of killing assay results must account for the complex and likely reciprocal interactions among Δqox-mediated metabolic changes, agrA-mediated redox sensing, and Δagrmediated changes in metabolism. Since killing data are not necessary or revealing without this information, we prefer to address the role of qox in future studies directed at understanding the complexities of the interaction among agr-mediated tolerance, endogenous ROS levels, and induction of protective responses in S. aureus.

(6) The repeated use of the term 'agr wild type' throughout the text is somewhat distracting. It might be clearer to simply use 'wild type,' as it is implied that this refers to the agr+ genotype.

We modified the text by replacing 'agr wild-type' with “wild-type” as suggested by the Reviewer.

(7) In the text, the authors imply that the extended lag phase of the agr mutant is observed solely in nutrient-limited CDM. However, Figure 1 and Figure Supplement 3A reveal that the strains were actually cultivated in CDM supplemented with glucose and Casamino acids, which makes the medium rich in both carbon and nitrogen, in addition to other nutrients present in CDM. The authors should clarify the composition of the media used and assess whether the term 'nutrient-limited CDM' is accurate in this context.

The extended lag phase of the Δagr mutant is observable in TSB but it is more easily appreciated in CDM, perhaps owing to a larger range of carbohydrates and other nutrient types (TSB a rich and complex medium for which the composition is unknown) and a higher concentration of glucose (2.5 mM versus 2.2 mM).

For clarification, we modified line 135 as follows:

Line 184: “Lag-time differences between strains were more obvious in experiments using less complex, chemically defined medium (CDM)…”

(8) Figure 1 - Figure Supplement 3C represents the growth rate in terms of [OD/min]. However, it would be more accurate to calculate the growth rate (μ) based on the change in the natural logarithm of optical density (OD) relative to the corresponding change in time, using appropriate units (preferably, h⁻¹). Additionally, the method employed for measuring growth rates should be detailed in the Materials and Methods section.

Our responses to Reviewer 2 Minor Comment 1 below address this issue.

(9) The resolution of the inset charts in Figure 4B is poor, and the Y-axis lacks labels. The figure legend should also specify whether the flux distribution (represented by thick black arrows in Fig 4B) is predicted for the wild type or the mutant.

We modified Figure 4B and the legend accordingly.

(10) On Page 9, the term "RT-PCR" should be corrected to "RT-qPCR."

We thank the Reviewer for their attention to detail in picking up our error. We modified text accordingly.

(11) It is ambiguous whether the agr mutant is producing more acetate, based on the information provided in Figure 5B. Since the cells might have entered the post-exponential phase at 5 hours, they could start consuming acetate. Consequently, the elevated acetate concentration in the agr mutant might result from a delay in acetate consumption rather than increased production. To discern between the production and consumption of acetate, it is essential to measure acetate concentrations at earlier time points as well as the corresponding glucose concentrations in the media. This will help ascertain when the agr mutant enters the post-exponential phase. A similar concern also exists in the case of lactate (Fig 5E) since it is not clear when lactate was measured.

As requested, we measured acetate levels at earlier time points (1, 2, 3, 4, h of growth). New Figure 5B shows that the Δagr mutant accumulated more acetate than the wild-type strain during exponential growth at 3 h, well before entry into postexponential phase (see growth curves in Figure 1—figure supplement 1).

In the original report, lactate levels were measured at 4 h for organisms grown under suboptimal aeration conditions (see Reviewer 1, Comment 1). When we measured lactate accumulation at 3 h it remained higher in the Δagr mutant compared to the wildtype. Likewise, acetate levels at 3 h under suboptimal aeration conditions remained elevated in the Δagr mutant compared to the wild-type. These results support the idea that inactivation of agr promotes production rather than decreased consumption of acetate and lactate in the culture medium.

(12) In Figure 5G-H, presenting the actual NAD+ and NADH values side-by-side would facilitate a more straightforward interpretation of the data by the readers.

(13) On Page 9, the text states that respiration and fermentation lower the NAD+/NADH ratio. However, this seems contradictory as these processes would typically increase the NAD+/NADH ratio. Furthermore, it would be beneficial for the authors to provide supporting evidence for the statement made at the beginning of Page 10, which claims that there is greater consumption of NADH in the agr mutant.

Responses to Comments 12 and 13 were grouped together.

We thank the Reviewer for their attention to detail in picking up our error about the NAD+/NADH ratio. The ratio is expected to be elevated by increases in respiration and fermentation, not lowered, owing to increased consumption of NADH.

Figure 5I in the submitted manuscript indicated a small but insignificant decrease in the NAD+/NADH ratio of the Δagr mutant. Thus, the NAD+/NADH ratio remained tightly bounded, but if anything was decreased, not increased.

We explained this finding as follows:

Line 284: “Collectively, these observations suggest that a surge in NADH production and reductive stress in the Δagr strain induces a burst in respiration and fermentation.”

The NAD+/NADH ratio in Figure 5F of the submitted manuscript was calculated from NADH and total (NAD+/NADH) levels. As requested, we measured individual NAD+ and NADH concentrations. We found that the decrease in the NAD+/NADH ratio of the Δagr mutant was now large, significant, and largely due to a relative increase in NADH.

We have included these new data in a revised Figure 5 in the revised version of the manuscript and clarify the relationship among the NAD+/NADH ratio, respiration, and fermentation in the Δagr mutant by modifying the wording of the text as follows:

Line 280: “Since respiration and fermentation generally increase NAD+/NADH ratios and since these activities are increased in Δagr strains (Figure 5C and 5E-F), we expected a higher NAD+/NADH ratio relative to wild-type cells. However, we observed an increase decrease in the NAD+/NADH ratio due to a large surge in NADH accompanied by a modest drop in NAD+ compared to wild-type. Collectively, these observations suggest that a surge in NADH production and reductive stress in the Δagr strain induces a burst in respiration, but levels of NADH are saturating, thereby driving fermentation in the presence of oxygen.

Reviewer #2 (Recommendations For The Authors):

(1) The RNA-seq analysis revealed that the Δagr strain exhibited increased expression of genes involved in respiration and fermentation, suggesting enhanced energy generation. However, metabolic modeling based on transcriptomic data indicated a decrease in tricarboxylic acid (TCA) cycle and lactate flux per unit of glucose uptake in the Δagr mutant. Additionally, intracellular ATP levels were significantly lower in the Δagr mutant compared to the wild-type strain, despite the carbon being directed into an acetate-generating, ATP-yielding carbon "overflow" pathway. Furthermore, growth analysis in nutrient-constrained medium demonstrated a decrease in the growth rate and yield of the Δagr mutant. Given that S. aureus actively utilizes the electron transport chain (ETC) to replenish NAD pools during aerobic growth on glucose, supporting glycolytic flux and pyruvate dehydrogenase complex (PDHC) activity while restricting TCA cycle activity through carbon catabolite repression (CCR), it is suggested that the authors analyze glucose consumption rates in conjunction with the determination of intracellular levels of pyruvate, AcCoA, and TCA cycle intermediates such as citrate and fumarate. These additional experiments will provide valuable insights into the metabolic fate of glucose and pyruvate and their subsequent impact on cellular respiration and fermentation in the Δagr mutant.

(2) The authors highlighted the importance of redox balance in Δagr cells by emphasizing the tendency of these cells to prioritize NAD+-generating lactate production over generating additional ATP from acetate. However, the results regarding acetate and lactate production in Δagr cells during aerobic growth suggest that carbon is directed towards acetate generation rather than lactate.

Responses to Comments 1 and 2 have been combined.

As requested, we measured glucose consumption and intracellular levels of several different metabolites in the wild-type and Δagr mutant strain. The results are consistent with the idea that increased acetogenesis and fermentation in Δagr mutant cells contribute to increased ATP production and NAD+ recycling, respectively. These two processes appear to be relatively favored over the flux of pyruvate carbon into the TCA cycle of the Δagr mutant.

We explained our finding as follows:

Line 288: “To help determine the metabolic fate of glucose, we measured glucose consumption and intracellular levels of pyruvate and TCA-cycle metabolites fumarate and citrate in the wild-type and Δagr mutant strains. At 4 h of growth to late-exponential phase, intracellular pyruvate and acetyl-CoA levels were increased in the Δagr mutant compared to wild-type strain, but levels of fumarate and citrate were similar (Figure 5— figure supplement 1D-E). Glucose was depleted after 4 h of growth, but glucose consumption after 3 h of growth (exponential phase) was increased in the Δagr mutant compared to the wild-type strain (Figure 5—figure supplement 1A). These observations, together with the decrease in the NAD+/NADH ratio and increase in acetate and lactate production described above, are consistent with a model in which respiration in Δagr mutants is inadequate for 1) energy production, resulting in an increase in acetogenesis, and 2) maintenance of redox balance, resulting in an increase in fermentative metabolism, lactate production, and conversion of NADH to NAD+. Increased levels of acetate compared to lactate under optimal aeration conditions suggests that demand for ATP is in excess of demand for NAD+.”

Future work will compare additional extracellular and intracellular (e.g., formate, ethanol, acetoin) metabolites to test these and other models using a combination of approaches (e.g., mass spectrometry, nuclear magnetic resonance, genetic deletion studies, transcriptomics) and will determine the mechanisms underlying metabolic differences in wild-type and Δagr mutant strains.

To maintain a sense of narrative we added a new subheading after the explanation of our findings:

Line 311: “Transcriptional changes due to Δagr mutation are long-lived and result in down-regulation of H2O2-stimulated genes relative to those in an agr wild-type.”

(3) The authors mentioned that respiration and fermentation typically reduce the NAD+/NADH ratios, and since these activities are elevated in Δagr strains (Figure 5F-G), they initially anticipated a lower NAD+/NADH ratio compared to wild-type agr cells. However, the increase in respiration and activation of fermentative pathways leads to a decrease in NADH levels, therefore resulting in an increase in the NAD+/NADH ratio.

We have clarified the issue with new experiments and by modifying the wording as shown in the response to Reviewer 1 Comment 13.

(4) To improve the clarity and completeness of this work, it would be advantageous for the authors to provide specific details regarding the glucose concentration in the TSB media and the aeration conditions during growth, including the flask-tomedium ratio. These additional experimental parameters are essential for ensuring the reproducibility and comprehensiveness of the study, allowing for a more precise understanding and interpretation of the observed metabolic changes in the Δagr strain.

We modified the Methods as suggested.

Minor comments:

(1) The growth rate in Figure 1-figure supplement 3 should not be presented as a simple calculation of OD/min and needs to be recalculated.

We recalculated the growth rate and modified Figure 1 as suggested. The exponential phase was used to determine growth rate (µ) from two datapoints, OD1 and OD2 flanking the linear portion of the curve, following the equation lnOD2-lnOD1/t2-t1, as described (12).

(2) Δrot (BS1301) should be removed from Figure 2 (A) legend as it is not presented in the panel A.

We modified Figure 2 as suggested.

(3) The authors should specify in the Figure 3 (D) legend that the kinetics of killing by H2O2 was performed for ΔrnaIII and ΔagrBD mixtures.

We modified Figure 3 as suggested.

(4) In the Figure 4 legend for (C), the statement "See Supplementary file 2 for supporting information" should be changed to "See Supplementary file 3 for supporting information."

We modified Supplementary file name as suggested.

References cited in responses

(1) Brynildsen MP, Winkler JA, Spina CS, MacDonald IC, Collins JJ. 2013. Potentiating antibacterial activity by predictably enhancing endogenous microbial ROS production. Nature biotechnology 31:160-165.

(2) Morfeldt E, Taylor D, von Gabain A, Arvidson S. 1995. Activation of alphatoxin translation in Staphylococcus aureus by the trans-encoded antisense RNA, RNAIII. EMBO J 14:4569-4577.

(3) Novick RP, Ross HF, Projan SJ, Kornblum J, Kreiswirth B, Moghazeh S. 1993. Synthesis of staphylococcal virulence factors is controlled by a regulatory RNA molecule. EMBO J 12:3967-3975.

(4) Takahashi N, Gruber CC, Yang JH, Liu X, Braff D, Yashaswini CN, Bhubhanil S, Furuta Y, Andreescu S, Collins JJ, Walker GC. 2017. Lethality of MalE-LacZ hybrid protein shares mechanistic attributes with oxidative component of antibiotic lethality. Proc Natl Acad Sci U S A 114:9164-9169.

(5) Fujimoto DF, Bayles KW. 1998. Opposing roles of the Staphylococcus aureus virulence regulators, Agr and Sar, in Triton X-100- and penicillin-induced autolysis. J Bacteriol 180:3724-3726.

(6) Cho H, Uehara T, Bernhardt TG. 2014. Beta-lactam antibiotics induce a lethal malfunctioning of the bacterial cell wall synthesis machinery. Cell 159:13001311.

(7) Rowe SE, Wagner NJ, Li L, Beam JE, Wilkinson AD, Radlinski LC, Zhang Q, Miao EA, Conlon BP. 2020. Reactive oxygen species induce antibiotic tolerance during systemic Staphylococcus aureus infection. Nat Microbiol 5:282-290.

(8) Zamboni N, Sauer U. 2003. Knockout of the high-coupling cytochrome aa3 oxidase reduces TCA cycle fluxes in Bacillus subtilis. FEMS Microbiol Lett 226:121-126.

(9) Halsey CR, Lei S, Wax JK, Lehman MK, Nuxoll AS, Steinke L, Sadykov M, Powers R, Fey PD. 2017. Amino acid catabolism in Staphylococcus aureus and the runction of carbon catabolite repression. mBio 8.

(10) Hammer ND, Reniere ML, Cassat JE, Zhang Y, Hirsch AO, Indriati Hood M, Skaar EP. 2013. Two heme-dependent terminal oxidases power Staphylococcus aureus organ-specific colonization of the vertebrate host. mBio 4.

(11) Lan L, Cheng A, Dunman PM, Missiakas D, He C. 2010. Golden pigment production and virulence gene expression are affected by metabolisms in Staphylococcus aureus. J Bacteriol 192:3068-3077.

(12) Grosser MR, Weiss A, Shaw LN, Richardson AR. 2016. Regulatory requirements for Staphylococcus aureus nitric oxide resistance. J Bacteriol 198:2043-2055.

eLife assessment

This important study outlines how the agr quorum sensing system in Staphylococcus aureus confers long-lived protection against oxidative stress, thereby linking bacterial metabolism to virulence in this pathogen. While the findings, which are supported by solid data, seem at first glance to contradict earlier findings that show increased fitness of agr mutants under oxidative stress, the core conclusions of the study are well-substantiated. The authors should, however, re-evaluate their interpretations related to the impact of agr inactivation on bacterial metabolic and redox status following oxidative stress. The topic of the paper holds broad relevance to microbiologists, especially those focusing on host-pathogen interactions and bacterial responses to ROS.

Reviewer #1 (Public Review):

As a pathogen, S. aureus has evolved strategies to evade the host's immune system. It effectively remains 'under the radar' in the host until it reaches high population densities, at which point it triggers virulence mechanisms, enabling it to spread within the host. The agr quorum sensing system is central to this process, as it coordinates the pathogen's virulence in response to its cell density.

In this study, Podkowik and colleagues suggest that cells activating agr signaling also benefit from protection against H2O2 stress, whereas inactivation of agr increases cell death. The underlying cause of this lack of protection is tied to an ATP deficit in the agr mutant, leading to increased glucose consumption and NADH production, ultimately resulting in a redox imbalance. In response to this imbalance, the agr mutant increases respiration, resulting in the endogenous production of ROS which synergizes with H2O2 to mediate killing of the agr mutant. Suppressing respiration in the agr mutant restored protection against H2O2 stress.

Additionally, the authors establish that agr-dependent protection against oxidative stress is also linked to RNAIII activation, and the subsequent block of Rot translation. However, the specific protective genes regulated by Rot remain unidentified. Thus, according to the evidence provided, agr triggers intrinsic mechanisms that not only decrease harmful ROS production within the cell but also alleviate its detrimental effects.

Interestingly, these protective mechanisms are long-lived, and guard the cells against external oxidative stressors such as H2O2, even after the agr system has been 'turned off' in the population.

While the study offers valuable insight into how agr signaling protects cells against H2O2 stress, a reevaluation of the interpretation of redox imbalance is warranted.

Reviewer #2 (Public Review):

In their study, Podkowik et al. elucidate the protective role of the accessory gene regulator (agr) system in Staphylococcus aureus against hydrogen peroxide (H2O2) stress. Their findings demonstrate that agr safeguards the bacterium by controlling the accumulation of reactive oxygen species (ROS), independent of agr activation kinetics. This protection is facilitated through a regulatory interaction between RNAIII and Rot, impacting virulence factor production and metabolism, thereby influencing ROS levels. Notably, the study highlights the remarkable adaptive capabilities of S. aureus conferred by agr. The protective effects of agr extend beyond the peak of agr transcription at high cell density, persisting even during the early log-phase. This indicates the significance of agr-mediated protection throughout the infection process. The absence of agr has profound consequences, as observed by the upregulation of respiration and fermentation genes, leading to increased ROS generation and subsequent cellular demise. Interestingly, the study also reveals divergent effects of agr deficiency on susceptibility to hydrogen peroxide compared to ciprofloxacin. While agr deficiency heightens vulnerability to H2O2, it also upregulates the expression of bsaA, countering the endogenous ROS induced by ciprofloxacin. These findings underscore the complex and context-dependent nature of agr-mediated protection. Furthermore, in vivo investigations using murine models provide valuable insights into the importance of agr in promoting S. aureus fitness, particularly in the context of neutrophil-mediated clearance, with notable emphasis on the pulmonary milieu. Overall, this study significantly advances our understanding of agr-mediated protection in S. aureus and sheds light on the sophisticated adaptive mechanisms employed by the bacterium to fortify itself against oxidative stress encountered during infection.

The conclusions drawn in this paper are generally well-supported by the data. To enhance the clarity of the study, it is recommended that the authors consider refraining from combining the data for lactate production during microaerobic growth with the remaining data obtained for aerobic growth. Different aeration conditions can significantly impact the metabolic status of the cells.

In this regard, the statement, "Collectively, these data suggest that Δagr increases respiration and aerobic fermentation to compensate for low metabolic efficiency," might be potentially misleading and could benefit from a revision to accurately reflect the nuances of the experimental conditions.

Additionally, the authors' statement, 'The tendency of Δagr cells to forgo the additional ATP yield from acetate production in favor of NAD+-generating lactate (23, 24) underscores the importance of redox balance in Δagr cells,' appears contradictory to the data presented in Fig 5, where the Δagr mutant demonstrates an approximately threefold increase in acetate production during exponential growth compared to the wild-type strain. A clarification or adjustment in the manuscript may be necessary to ensure consistency and accurate interpretation.

Furthermore, the authors' statement, 'Collectively, these observations suggest that a surge in NADH consumption and reductive stress in the Δagr strain induces a burst in respiration, but levels of NADH are saturating, thereby driving fermentation in the presence of oxygen,' may need revision. Data presented in Figure 5 suggest the opposite - a surge in NADH accumulation leading to a decrease in the NAD/NADH ratio, rather than a surge in the 'consumption' of NADH. Clarifying this point in the manuscript would ensure accurate representation of the findings.

The authors attention to these matters would greatly contribute to the precision and clarity of the findings.

Author Response

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

Response to reviewer’s comments

Reviewer #1 (Public Review):

In this study, the structural characteristics of plant AlaDC and SerDC were analyzed to understand the mechanism of functional differentiation, deepen the understanding of substrate specificity and catalytic activity evolution, and explore effective ways to improve the initial efficiency of theanine synthesis.

On the basis of previous solid work, the authors successfully obtained the X-ray crystal structures of the precursors of theanine synthesis-CsAlaDC and AtSerDC, which are key proteins related to ethylamine synthesis, and found a unique zinc finger structure on these two crystal structures that are not found in other Group II PLP-dependent amino acid decarboxylases. Through a series of experiments, it is pointed out that this characteristic zinc finger motif may be the key to the folding of CsAlaDC and AtSerDC proteins, and this discovery is novel and prospective in the study of theine synthesis.

In addition, the authors identified Phe106 of CsAlaDC and Tyr111 of AtSerDC as key sites of substrate specificity by comparing substrate binding regions and identified amino acids that inhibit catalytic activity through mutation screening based on protein structure. It was found that the catalytic activity of CsAlaDCL110F/P114A was 2.3 times higher than that of CsAlaDC. At the same time, CsAlaDC and AtSerDC substrate recognition key motifs were used to carry out evolutionary analysis of the protein sequences that are highly homologous to CsAlaDC in embryos, and 13 potential alanine decarboxylases were found, which laid a solid foundation for subsequent studies related to theanine synthesis.

In general, this study has a solid foundation, the whole research idea is clear, the experimental design is reasonable, and the experimental results provide strong evidence for the author's point of view. Through a large number of experiments, the key links in the theanine synthesis pathway are deeply studied, and an effective way to improve the initial efficiency of theanine synthesis is found, and the molecular mechanism of this way is expounded. The whole study has good novelty and prospectivity, and sheds light on a new direction for the efficient industrial synthesis of theanine

Response: Thank you very much for taking time to review this manuscript. We appreciate all your insightful comments and constructive suggestions.

Reviewer #1 (Recommendations For The Authors):

(1) If some test methods are not original, references or method basis should be indicated.

Response: Thank you very much for your careful reading of the manuscript and valuable suggestions. We have added references for the enzymatic activity experiments performed to measure the synthesis of theanine in the revised manuscript.

(2) The conclusion is a little lengthy, and the summary of the whole study is not well condensed.

Response: Thank you very much for your valuable suggestions. We have refined the conclusion in the revised manuscript, and it is as follows:

In conclusion, our structural and functional analyses have significantly advanced understanding of the substrate-specific activities of alanine and serine decarboxylases, typified by CsAlaDC and AtSerDC. Critical amino acid residues responsible for substrate selection were identified—Tyr111 in AtSerDC and Phe106 in CsAlaDC—highlighting pivotal roles in enzyme specificity. The engineered CsAlaDC mutant (L110F/P114A) not only displayed enhanced catalytic efficiency but also substantially improved L-theanine yield in a synthetic biosynthesis setup with PsGS or GMAS. Our research expanded the repertoire of potential alanine decarboxylases through the discovery of 13 homologous enzyme candidates across embryophytic species and uncovered a special motif present in serine protease-like proteins within Fabale, suggesting a potential divergence in substrate specificity and catalytic functions. These insights lay the groundwork for the development of industrial biocatalytic processes, promising to elevate the production of L-theanine and supporting innovation within the tea industry.

Reviewer #2 (Public Review)

Summary:

The manuscript focuses on the 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 the 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 the 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:

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 structure and mechanism section would also be strengthened by including a diagram of the reaction mechanism and including context about reactivity. As it stands, much of the structural results section consists of lists of amino acids interacting with certain ligands without any explanation of why these interactions are important or the role they play in catalysis. The experiments testing the function of a novel Zn(II)-binding domain also have serious flaws. I don't think anything can be said at this point about the function of the Zn(II) due to a lack of key controls and problems with experimental design.

Response: Thank you very much for your thoughtful comments and feedback on our manuscript. We are pleased to hear that the work's strengths, such as the X-ray crystal structures and the mutagenesis experiments tied to the catalytic rate and substrate specificity, align with the goals of our research.

We recognize the areas identified for improvement and appreciate the suggestions provided. We have emphasized how we use the structural information obtained to infer the roles of key amino acid residues in the reaction. Additionally, we have added a diagram of the reaction mechanism in the Supplementary figure to provide clearer context on reactivity and improve the overall understanding of the catalytic process. Regarding the structural results section, we have included a discussion that contextualizes the list of amino acids and their interactions with the ligands by explaining their significance and roles in catalysis. We acknowledge the weaknesses you've pointed out in the experiments concerning the novel Zn(II)-binding domain, but we would like to clarify that the focus of our study was not primarily on the zinc structure. While we agree that there may be limitations in the experimental design and controls for the zinc binding domain, we believe that these flaws do not significantly impact the overall findings of the study. The experiment served as a preliminary exploration of the potential functionality of the domain, and further studies are required to fully understand its role and mechanism.

Reviewer #2 (Recommendations For The Authors):

(1) In addition to the points raised in the public review, it would be ideal to provide some context for the enzymatic characterization. Why are the differences in kinetic parameters for AlaDC and SerDC significant?

Response: Thank you for your comments and suggestions. The Km values for CsAlaDC and SerDCs are comparable, suggesting similar substrate affinities. However, CsAlaDC exhibits a significantly lower Vmax compared to AtSerDC and CsSerDC. This discrepancy implies that CsAlaDC and SerDCs may differ in the rates at which they convert substrate to product when saturated with substrate. SerDCs may have a faster turnover rate, meaning they convert substrate to product and release the enzyme more quickly, resulting in a higher Vmax. Differences in the stability or correct folding of the enzymes under assay conditions can also affect their Vmax. If SerDCs are more stable, they might maintain their catalytic activity better at higher substrate concentrations, contributing to a higher Vmax. We have added these to the part of “Enzymatic properties of CsAlaDC, AtSerDC, and CsSerDC” in our revised manuscript.

(2) Why is Phe106/Tyr111 pair critical for substrate specificity? Does the amino acid contact the side chain? It might be helpful to a reader to formulate a hypothesis for this interaction.

Response: Thank you for the question and comments. We conducted a comparison between the active sites of CsAlaDC and AtSerDC and observed a distinct difference in only two amino acids: F106 in CsAlaDC and Y111 in AtSerDC. The remaining amino acids were found to be identical. Expanding on previous research concerning Group II PLP-dependent amino acid decarboxylases, it was postulated and subsequently confirmed that these specific amino acids play a crucial role in substrate recognition. However, since we lack the structure of the enzyme-substrate complex, we are unable to elucidate the precise interactions occurring between the substrate and the amino acids at this particular site based solely on structural information.

(3) Line 55 - Define EA again.

Response: Thank you very much for your careful reading of the manuscript and valuable suggestions. We have redefined “EA” as the abbreviation for ethylamine in the revised manuscript.

(4) Line 58 - The meaning of "determined by the quality formation of tea" is not clear.

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

(5) Line 65 - Missing words between "despite they".

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

(6) Line 67 - Need a reference for the statement about lower activity?

Response: Thank you for the question and comments. We have provided the following reference to support this statement in the revised manuscript.

Reference: Bai, P. et al. (2021) Biochemical characterization of specific Alanine Decarboxylase (ADC) and its ancestral enzyme Serine Decarboxylase (SDC) in tea plants (Camellia sinensis). BMC Biotechnol. 21,17.

(7) Line 100-101 - The meaning of "its closer relationship was Dicots plants." is not clear.

Response: We have revised the sentence in the revised manuscript, as follows: “Phylogenetic analysis indicated that CsAlaDC is homologous with SerDCs in Dicots plants.”

(8) Line 139 - Missing a word between "as well as" and "of".

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

(9) Line 142 - The usage of comprised here is not correct. It would be more correct to say "The overall architecture of CsAlaDC and AtSerDC is homodimeric with the two subunits...".

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

(10) Line 148-149 - I didn't understand the statement about the "N-terminal structures" Are these structures obtained from protein samples that have a truncated N-terminus?

Response: Group II PLP-dependent amino acid decarboxylases are comprised of three distinct structural domains: the N-terminal domain, the large domain, and the C-terminal domain. Each of these domains possesses unique structural features. Similarly, CsAlaDC and AtSerDC can also be classified into three structural domains based on their specific characteristics. To achieve more stable proteins for further experiments, we conducted truncation on both of these proteins. The truncated section pertains to a subsection of the N-terminal domain and is truncated from the protein's N-terminus.

(11) Line 153 - Say "is composed of" instead of "composes of".

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

(12) Line 156 - I didn't understand the statement about the cofactor binding process. What is the cofactor observed? And how can we say anything about the binding process from a single static structure of the enzyme? It might be better to say that the cofactor binding site is located at the subunit junction - but the identity of the cofactor still needs to be defined first.

Response: Thank you for your comments and suggestions. The cofactor mentioned here is PLP. We aim to elucidate the binding state of PLP at the active site, excluding the binding process. The description has been revised in the revised manuscript.

(13) Lines 157-158 - I didn't understand the conclusion about the roles of each monomer. In the images in Figure 3 - both monomers appear to bind PLP but the substrate is not present - so it's not clear how conclusions can be drawn about differential substrate binding in the two subunits.

Response: Thank you very much for your careful reading and valuable suggestions. The main idea we want to convey is that this protein possesses two active sites. At each active site, the two monomers carry out distinct functions. Of course, our previous conclusion is inaccurate due to the non-existence of the substrate. So, we have made the necessary amendments in the revised manuscript.

(14) Line 161 - I would say loop instead of ring.

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

(15) Line 165 - Please provide some references for this statement. It would also be ideal to state the proximity of the Zn-binding motif to the active site or otherwise provide some information about the role of the motif based on its location.

Response: Thank you for your comments and suggestions. We have provided the following references to support this statement in the revised manuscript.

Author response image 1.

(A) Structure of histidine decarboxylase. (B) Structure of glutamate decarboxylase.

Reference:

30 Komori, H. et al. (2012) Structural study reveals that Ser-354 determines substrate specificity on human Histidine Decarboxylase. J Biol Chem. 287, 29175-83.

31 Huang, J. et al. (2018) Lactobacillus brevis CGMCC 1306 glutamate decarboxylase: Crystal structure and functional analysis. Biochem Biophys Res Co. 503, 1703-1709

In CsAlaDC, the zinc is positioned at a distance of 29.6 Å from the active center, whereas in AtSerDC, the zinc is situated 29 Å away from the active center. Hence, we hypothesize that this structure does not impact the enzyme's catalytic activity but might be correlated with its stability.

(16) Lines 166-178 - This paragraph appears to be a list of all of the interactions between the protein, PLP, and the EA product. It would be ideal to provide some text to explain why these interactions are important and what we can learn from them.

Response: Thank you very much for your careful reading of the manuscript and valuable suggestions. We have been conducting additional analysis on the functional roles of amino acid residues involved in the interaction between the active site and PLP. This analysis focuses on aiding PLP binding, determining its orientation, and understanding enzyme catalytic mechanisms. These details are mentioned in the revised manuscript.

(17) Line 192 - Bond not bound.

Response: Thank you very much for your careful reading of the manuscript and valuable suggestions. We have made corrections in the revised manuscript.

(18) Lines 201-207 - It would be ideal to verify that the inclusion of 5 mM DTT affects Zn binding. It's not clear to me that this reagent would necessarily disrupt Zn binding. Under certain circumstances, it could instead promote Zn association. For example, if the Cys ligands are oxidized initially but then become reduced? I don't think the current experiment really provides any insight into the role of the Zn.

Response: Thank you for your valuable insights regarding the role of DTT and its potential effects on Zn binding in our experiments. The main function of DTT is to protect or restore the reduced state of proteins and other biological molecules, particularly by disrupting the crosslinking formed by thiol (-SH) groups and disulfide bonds to maintain the function and structure of proteins. Therefore, the reason for DTT's inhibition of enzyme activity is unknown, and we cannot provide a reasonable explanation for this phenomenon. As a result, we have removed the section discussing the inhibition of enzyme activity by DTT in our revised manuscript.

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

Reviewer #3 (Recommendations For The Authors):

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?

Response: We use a Superdex 200 (Hiload 16/600) column for size exclusion chromatography. The comparison results of SEC elution profiles for AtSerDC and CsAlaDC, along with the standard curve of SEC column, are presented below.

Author response image 2.

(A) Comparison of elution profiles of CsAlaDC and AtSerDC. (B) Elution profile of Blue Dextron 2000. (C) Elution profile of mixed protein (Aldolase, 158000 Da,71.765ml; Conalbumin, 75000 Da,79.391ml; Ovalbumin, 44000 Da,83.767ml; Carbonic anhydrase, 29000 Da,90.019ml; Ribonuclease A, 13700 Da,98.145ml). (D) Size standard curves of Superdex 200 (Hiload 16/600) column.

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 ℃ and pH 8.0 for CsAlaDC, 40 ℃ 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 ℃, but their activities were directly compared. Is there a reason for experimenting at different temperatures?

Response: We determined that the optimal reaction temperature for AtSerDC is 40°C and for CsAlaDC is 45°C through our verification process. Consequently, all subsequent experiments were performed at these specific temperatures.

Author response image 3.

(A) Relative activity of CsAlaDC at different temperatures. (B) Relative activity of AtSerDC at different temperatures.

(2) Enzyme activities were measured at temperatures above 40℃, 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.

Response: Thank you very much for your careful reading. We have already validated that the experimental temperature we used did not significantly affect the stability of the protein before experimenting. The results are shown in the figure below:

Author response image 4.

Place the two proteins individually into water baths set at temperatures of 25°C, 37°C, 45°C, 60°C, and 80°C for 15 minutes. Subsequently, carry out enzymatic reactions utilizing a standard reaction system, with untreated enzymes serving as the experimental control within the said system. The experimental results suggest that the temperature at which we experimented does not have a significant impact on the stability of the enzyme.

(3) The authors used 20 mM of substrate. What are the physiological concentrations of alanine and serine typically found in plants?

Response: 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. In this study, the substrate concentration of 20 mM was close to the actual concentrations of alanine and serine in plants.

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.

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).

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

(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.

Response: Thank you for your comments and suggestions. The catalytic activity is assessed by comparing reaction rates.

Page 7 & Figure 3

In Figure 3A, the authors describe the overall structure, but a simple explanation or labeling within the figure should be added.

Response: Thank you very much for your suggestions, we have made modifications to Figure 3A as follows:

Author response image 5.

Crystal structures of CsAlaDC and AtSerDC. (A) Dimer structure of CsAlaDC. The color display of the N-terminal domain, large domain, and C-terminal domains of chain A is shown in light pink, khaki and sky blue, respectively. Chain B is shown in spring green. The PLP molecule is shown as a sphere model. The zinc finger structure at the C-terminus of CsAlaDC is indicated by the red box. The gray spheres represent zinc ions, while the red dotted line depicts the coordination bonds formed by zinc ions with cysteine and histidine.

Figures 3F & 4A

In these figures, the two structures are overlaid and compared, but the colors are very similar to see the differences. The authors should use a different color scheme.

Response: Thank you very much for your suggestions, we have made modifications to the Figure 3F & 4A as follows:

Author response image 6.

(Figure 3F) The monomers of CsAlaDC and AtSerDC are superimposed. CsAlaDC is depicted in spring green, while AtSerDC is shown in plum. The conserved amino acid catalytic ring is indicated by the red box.

(Figure 4A) Superposition of substrate binding pocket amino acid residues in CsAlaDC and AtSerDC. The amino acid residues of CsAlaDC are shown in spring green, the amino acid residues of AtSerDC are shown in plum, with the substrate specificity-related amino acid residue highlighted in a red ellipse.

Pages 7 & 8

Figures 3 and 4 do not include illustrations of what the authors describe in the text. The reader will not be able to understand the descriptions until they download and view the structures themselves. The authors should create additional figures to make it easier for readers to understand the structures.

Response: Thank you very much for your suggestions, we have included supplementary figure 1 in the revised manuscript, which presents more elaborate structural depictions of the two proteins.

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.

Response: Thank you for your insightful feedback on our manuscript, which we greatly appreciate. Your suggestion to methodically sequence the experiments provides a clear pathway to bolster the strength and conclusiveness of our results.

We agree that it is crucial to first assess the stability of the mutant proteins, as changes therein could inadvertently affect catalytic activity. To this end, we have employed circular dichroism (CD) to study the potential structural alterations in the proteins induced by mutations. The experimental results are shown in the following figure:

Author response image 7.

(A) Circular Dichroism Spectra of CsAlaDC (WT). (B) Circular Dichroism Spectra of CsAlaDC (Y336F). (C) Circular Dichroism Spectra of CD of AtSerDC (WT). (D) Circular Dichroism Spectra of AtSerDC (Y341F).

The experimental results indicate that the secondary structure of the mutant proteins remains unchanged, which means the mutations do not alter the protein's stability.

The ligand PLP forms a Schiff base structure with the ε-amino group of a lysine residue in the protein, with maximum absorbance around 420-430 nm. Since we have already added PLP during the protein purification process, as long as the absorbance of mutant proteins and wild-type proteins is the same at 420-430 nm at equivalent concentrations, it indicates that the mutant proteins do not affect the binding of the ligand PLP. Therefore, we scanned the UV-visible absorption spectra of both the wild-type and mutant proteins, and the results are as presented in the following figure:

Author response image 8.

(A) UV-Visible Absorption Spectra of CsAlaDC (WT) compared to CsAlaDC (Y336F). (B) UV-Visible Absorption Spectra of AtSerDC (WT) compared to AtSerDC (Y341F).

The mutant protein and the wild-type protein exhibit similar absorbance at 420-430 nm, indicating that the mutation does not affect the binding of PLP to the protein.

The above experiments have confirmed that the mutations do not significantly affect the stability of the protein or the affinity for the ligand, so we can more confidently attribute changes in enzyme activity to the specific role of the tyrosine residue in question. We believe this comprehensive approach will substantiate our hypothesis and illustrate the necessity of this Tyr residue for the catalytic activity of CsAlaDC and AtSerDC enzymes.

Figure 3

In the 3D structure figure provided by the authors, the proposed reaction mechanism of the enzyme and the involved amino acids are not included. Can the authors add a supplementary figure with a schematic drawing that includes more information, such as distances?

Response: Thank you for your valuable feedback on our manuscript. We completely agree that a schematic drawing with additional details, including distances, would enhance the clarity and understanding of the enzymatic mechanism. In response to your suggestion, we have added a supplementary figure 2 in the revised manuscript that accurately illustrates the proposed reaction pathway, highlighting the key amino acids involved.

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: Thank you for your attention and review of our research paper, we appreciate your suggestions and feedback. The experimental protocol employed to evaluate the influence of DTT on protein catalytic efficiency is outlined as follows:

The 100 μL reaction mixture, containing 20 mM substrate (Ala or Ser), 100 mM potassium phosphate, 0.1 mM PLP, 5 mM L-DTT, and 0.025 mM purified enzyme, was prepared and incubated at standard conditions (45 °C and pH 8.0 for CsAlaDC for 5 min, 40 °C and pH 8.0 for AtSerDC for 2 min). DTT is absent as a control in the reaction system. Then the reaction was stopped with 20 μL of 10% trichloroacetic acid. The product was derivatized with 6-aminoquinolyl-N-hydroxy-succinimidyl carbamate (AQC) and subjected to analysis by UPLC. All enzymatic assays were performed in triplicate.

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

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.

Response: Thank you for your insightful comments and constructive feedback on our manuscript. We appreciate the interest you have expressed in the identification of 'Phe106 in CsAlaDC' and 'Tyr111 in AtSerDC' and their functional implications based on mutagenesis and enzymatic assays.

In order to investigate the binding status of the mutant protein and the ligand PLP,we scanned the UV-visible absorption spectra of both the wild-type and mutant proteins, and the results are as presented in the following figure:

Author response image 9.

(A) UV-Visible Absorption Spectra of CsAlaDC (WT) compared to CsAlaDC (F106Y). (B) UV-Visible Absorption Spectra of AtSerDC (WT) compared to AtSerDC (Y111F).

The mutant protein and the wild-type protein exhibit similar absorbance at 420-430 nm, indicating that the mutation does not affect the binding of PLP to the protein. Therefore, we can conclude that the change in activity of the mutant protein is caused by the substitution of the amino acid at that site, i.e., the amino acid at that site affects substrate specificity. By combining the structure of the two proteins, we can see that the Lys at position 111 of AtSerDC is a hydrophilic amino acid, which increases the hydrophilicity of the active site, and thus the substrate is the hydrophilic amino acid Ser. In contrast, the amino acid at the corresponding site in CsAlaDC is Phe, which, lacking a hydroxyl group compared to Lys, increases the hydrophobicity of the active site, making the substrate lean towards the hydrophobic amino acid Ala. We have added a discussion of the potential reasons for this result to the revised manuscript's discussion section.

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.

Response: Many thanks for your constructive suggestion. We added a matrix table showing amino acid sequence similarities in the supplemental materials. The results showed that the similarity of amino acid sequences between CsSerDC and AtSerDC is 86.21%, which is higher than that between CsAlaDC and CsSerDC (84.92%). This data exactly supports the description of Figure 1B. We added the description of the amino acid sequence similarities analysis in the revised manuscript. The description of "As expected, CsSerDC was most closed to AtSerDC, which implies that they shared similar functions. " is not accurate enough, so we revised it to "As expected, CsSerDC was closer to AtSerDC, which implies that they shared similar functions.", in the revised manuscript.

Page 5 & Figure 1C

Figure 1C, which shows a multiple sequence alignment with the amino acid sequences of the 6 SerDCs and CsAlaDC, clearly shows the differences between the sequences of AlaDC and other SerDCs. However, the authors' hypothesis would be more convincing if they showed that this difference is also conserved in AlaDCs from other plants. Can the authors show a new multiple-sequence alignment by adding more amino acid sequences of other AlaDCs?

Response: Thank you for your comments and suggestions. We aim to discover additional alanine decarboxylase. However, at present, the only experimentally confirmed alanine decarboxylase is CsAlaDC. No experimentally verified alanine decarboxylases have been found in other plant species.

Figure 5A

Figure 5A is missing the error bar.

Response: Figure 5A serves as a preliminary screening for these mutants, without conducting repeated experiments. Subsequently, only the L110F and P114A mutants, which exhibited significantly improved activity, underwent further experimental verification to confirm their enhanced functionality.

eLife assessment

This study reports 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 results, particularly the x-ray crystal structures, provide a solid basis for understanding substrate specificity. 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. 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.

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 original reviews.

eLife assessment

In this valuable study, the discovery and subsequent design of the AF03-NL chimeric antibody yielded a tool for studying filoviruses and provides a possible blueprint for future therapeutics. However, the data are incomplete and not presented clearly, which obscures flaws in the analyses and leaves unexplained phenomena. The work will be of interest to virologists studying antibodies.

Author response: Thank for your very valuable comments. The ms has been revised substantially and some new data have been added to further support the conclusions.

Public Reviews:

Reviewer #1 (Public Review):

Summary and Strengths:

Zhang et al. conducted a study in which they isolated and characterized a Marburg virus (MARV) glycoprotein-specific antibody, AF-03. The antibody was obtained from a phage-display library. The study shows that AF-03 competes with the previously characterized MARV-neutralizing antibody MR78, which binds to the virus's receptor binding site. The authors also performed GP mutagenesis experiments to confirm that AF-03 binds near the receptor binding site. In addition, the study confirmed that AF-03, like MR78, can neutralize Ebola viruses with cleaved glycoproteins. Finally, the authors demonstrated that NPC2-fused AF-03 was effective in neutralizing several filovirus species.

Weaknesses:

(1) The main premise of this study is unclear. Flyak et al. in 2015 described the isolation and characterization of a large panel of neutralizing antibodies from a Marburg survivor (Flyak et al., Cell, 2015). Based on biochemical and structural characterization, Flyak proposed that the Marburg neutralizing antibodies bind to the NPC1 receptor binding side. In the same study, it has been shown that several MARV-neutralizing antibodies can bind to cleaved Ebola glycoproteins that were enzymatically treated to remove the mucin-like domain and glycan cap. In the following study, it has been shown that the bispecific-antibody strategy can be used to deliver Marburg-specific antibodies into the endosome, where they can neutralize Ebola viruses (Wec et al., Science 2016). Finally, the use of lysosome-resident protein NPC2 to deliver antibody cargos to late endosomes has been previously described (Wirchnianski et al., Front. Immunol, 2021). The above-mentioned studies are not referenced in the introduction. The authors state that "there is no licensed treatment or vaccine for Marburg [virus] infection." While this is true, there are human antibodies that recognize neutralizing epitopes - that information can't be excluded while providing the rationale for the study. Furthermore, the authors use the word "novel" to describe the AF-03 antibody. How novel is AF-03 if multiple Marburg-neutralizing antibodies were previously characterized in multiple studies? Since AF-03 competes with previously characterized MR78, it binds to the same antigenic region as MR78. AF-03 also has comparable neutralization potency as MR78.

Author response: Thank for your valuable advice. In terms of the novelty of AF-03, the inhibition assay indicates that Q128/N129/C226 functions as key amino acids responsible for AF-03 neutralization given that the neutralizing capacity of AF-03 to pesudotyped virus harboring these mutants is impaired (see revised Fig. 2A left panel). Furthermore, ELISA assays show that mutation of Q128S-N129S or C226Y significantly disrupts the binding of GP to AF-03, while the neutralizing and binding capacity of MR78 to mutant GP and pseudovirus harboring C226Y instead of Q128S-N129S is not almost affected (see revised Fig. 2A right panel and 2B). Considering the fact that AF-03 and MR78 could compete with each other to bind to MARV GP (Fig. 2D). we thus make a conclusion that the epitopes of these two mAbs overlapped partially. Therefore, AF-03 is not a clone of MR78 and is a novel neutralizing mAb to MARV.

The work from Wirchnianski and colleagues has been referenced actually in the ms (see Ref. 38). Although our strategy for the design of broad-spectrum neutralizing antibody refers to their work, we further expand the species being evaluated including RAVN and mutated EBOV strains. The results show that NPC2-fused AF-03 exhibits neutralizing activity to 10 filovirus species and 17 EBOV mutants (Fig. 6A and B). The work by Flyak et al. in 2015 that described the isolation and characterization of a large panel of neutralizing antibodies from a Marburg survivor has been cited in Introduction section accordingly.

(2) Without the AF-03-MARV GP crystal structure, it's unclear how van der Waals interactions, H-bonds, and polar and electrostatic interactions can be evaluated. While authors use computer-guided homology modeling, this technique can't be used to determine critical interactions. Furthermore, Flyak et al. reported that binding to the NPC1 receptor binding site is the main mechanism of Marburg virus neutralization by human monoclonal antibodies. Since both AF-03 (this study) and MR78 (Flyak study) competed with each other, that information alone was sufficient for GP mutagenesis experiments that identified the NPC1 receptor binding site as the main region for mutagenesis.

Author response: Computer-guided homology modeling has been exploited successfully in our lab to determine key residues responsible for the interaction between antigen and mAbs (Immunol Res. 2015, 62:377; Scand J Immunol. 2019, 90:e12777; Sci Rep. 2022, 12:8469; Front Immunol. 2022, 13:831536). We refer to the crystal structure of MARV GP and the complex of MR78 and GP reported previously (Cell 2015, 160:904) and then model the complex of MARV GP and AF-03. Although AF-03 and MR78 compete with each other, we show that the epitopes of these two mAbs just overlap partially (Fig. 2A-D).

(3) The AF-03-GP affinity measurements were performed using bivalent IgG molecules and trimeric GP molecules. This format does not allow accurate measurements of affinity due to the avidity effect. The reported KD value is abnormally low due to avidity effects. The authors need to repeat the affinity experiments by immobilizing trimeric GPs and then adding monovalent AF-03 Fab.

Author response: As shown in Fig. 1A, GP protein used in this work is not trimer but largely monomer composed of MLD-deleted GP1 and GP2, which may at a certain extent weaken the engagement between GP and AF-03. It is noteworthy that we re-done the SPR assays for the binding of AF-03 to GP and show that KD value is 4.71x10-11M (see revised Fig. 1C). This GP protein is thus available to the evaluation of mAb affinity. In addition, it is reasonable to utilize bivalent IgG to detect the affinity of mAb to monomeric GP since the affinity likely decreases significantly when monovalent Fab is used.

Reviewer #2 (Public Review):

Summary:

The authors describe the discovery of a filovirus neutralizing antibody, AF03, by phage display, and its subsequent improvements to include NPC2 that resulted in a greater breadth of neutralization. Overall, the manuscript would benefit from considerable grammatical review, which would improve the communication of each point to the reader. The authors do not convincingly map the AF03 epitope, nor do they provide any strong support for their assumption that AF03 targets the NPC1 binding site. However, the authors do show that AF03 competes for MR78 binding to its epitope, and provides good support for the internalization of AF03-NL as the mechanism for improved breadth over the original AF03 antibody.

Strengths:

This study shows convincing binding to Marburgvirus GP and neutralization of Marburg viruses by AF03, as well as convincing neutralization of Ebolaviruses by AF03-NL. While there are no distinct populations of PE-stained cells shown by FACS in Figure 5A, the cell staining data in Figure 5C are compelling to a non-expert in endosomal staining like me. The control experiments in Figure 7 are compelling showing neutralization by AF03-NL but not AF03 or NPC2 alone or in combination. Altogether these data support the internalisation and stabilisation mechanism that is proposed for the gain in neutralization breadth observed for Ebolaviruses by AF03-NL over AF03 alone.

Weaknesses:

Overall, this reviewer is of the opinion that this paper is constructed haphazardly. For instance, the neutralization of mutant pseudoviruses is shown in Figure 2 before the concept of pseudovirus neutralization by AF03 is introduced in Figure 3. Similarly, the control experiments for AF03+NPC2 are described in Figure 7 after the data for breadth of neutralization are shown in Figure 6. GP quality controls are shown in Figure 2 after GP ELISAs / BLI experiments are done in Figure 1. This is disorienting for the reader.

Author response: AF-03 production and its binding capacity to GP is determined in Fig. 1. The epitopes of AF-03 is identified in Fig. 2. The neutralizing activity of AF-03 to pseudotyped MARV in vitro and in vivo is detected in Fig. 3. The neutralizing activity of AF-03 to pseudotyped ebolavirus harboring cleaved GP is detected in Fig. 4. The endosome-delivering ability of AF03-NL is examined in Fig. 5. The neutralization of filovirus species and EBOV mutants by AF03-NL is detected in Fig. 6. The requirement of CI-MPR for neutralization activity of AF03-NL is determined in Fig. 7. We think that this arrangement is suitable.

Figure 1: The visualisation of AF03 modelling and docking endeavours is extremely difficult to interpret. Firstly, there is no effort to orient the non-specialist reader with respect to the Marburgvirus GP model. Secondly, from the figures presented it is impossible to tell if the Fv docks perfectly onto the GP surface, or if there are violent clashes between the deeply penetrating AF03 CDRs and GP. This information would be better presented on a white background, perhaps showing GP in surface view from multiple angles and slices. The authors attempt to label potential interactions, but these are impossible to read, and labels should be added separately to appropriately oriented zoomed-in views.

Author response: To be readily understood the rationale of computer-guided modeling, the descriptions in the Methods and Results section have been refined accordingly. In addition, the information of the theoretical structure was presented on white background (see revised Fig. 1D-F).

Figure 2: The neutralization of mutant pseudoviruses cannot be properly assessed using bar graphs. These data should be plotted as neutralization curves as they were done for the wild-type neutralization data in Figure 3. The authors conclude that Q128 & N129 are contact residues, but the neutralization data for this mutant appear odd as the lowest two concentrations of AF03 show higher neutralization than the second highest AF03 concentration. Neutralization of T204/Q205/T206 (green), Y218 (orange), K222 (blue), or C226 (purple) appears to be better than neutralization of the wild-type MARV. The authors do not discuss this oddity. What are the IC50's? The omission of antibody concentrations on the x-axis and missing IC50 values give a sense of obscuring the data, and the manuscript would benefit from greater transparency, and be much easier to interpret if these were included. I am intrigued that the Q128S/N129S mutant is reported as having little effect on the neutralization of MR78. The bar graph appears to show some effect (difficult to interpret without neutralization curves and IC50 data), and indeed PDB:5UQY seems to suggest that these amino acids form a central component of the MR78 epitope (Q128 forms potential hydrogen bonds with CDRH1 Y35 and CDRL3 Y91, while N129 packs against the MR78 CDRH3 and potentially makes additional polar contact with the backbone). Lastly, since neutralization was tested in both HEK293T cells and Huh7 cells in Figure 3, the authors should clarify which cells were used for neutralization in Figure 2.

Author response: Thank for your advice. Accordingly, in the revised ms, the neutralization curve of AF-03 and MR78 is presented in revised Fig. 2A. The neutralization of AF-03 to pseudotyped MARV harboring Q128S/N129S or C226Y is impaired significantly compared with WT MARV and those bearing other indicated mutations, while Q128S/N129S instead of C226Y mutation affect the neutralizing capacity of MR78 at a certain extent. This is consistent with the data on the binding of AF-03 or MR78 to MARV GP protein assayed by ELISA (see revised Fig. 2B). Overall, these results show that Q128/N129/C226 functions as key amino acids responsible for AF-03 neutralization.

Figure 3: The first two images in Figure 3C showing bioluminescent intensity from pseudovirus-injected mice pretreated with either 10mg/kg or 3mg/kg AF03 are identical images. This is apparent from the location, shape, and intensity of the bioluminescence, as well as the identical foot placement of each mouse in these two panels. Currently, this figure is incomplete and should be corrected to show the different mice treated with either 10mg/kg or 3mg/kg of AF03.

Author response: Thank for your carefulness. Indeed, it is our mistake. In the revised ms, this fault has been corrected. The correct images have been added (see revised Fig. 3C).

Figure 4 would benefit from a control experiment without antibodies comparing infection with GP-cleaved and GP-uncleaved pseudoviruses. The paragraph describing these data was also difficult to read and would benefit from additional grammatical review.

Author response: Accordingly, a control experiment comparing the infection of GP-cleaved with GP-uncleaved pseudoviruses is performed. The results show that The infection of pseudotyped ebolavirus harboring cleaved GP to host cells is comparable or stronger than those containing intact GP（see revised Fig. s1）. Therefore, the data in Fig. 4 support the inhibition of cell entry of ebolavirus species harboring cleaved GP by AF-03, which is not attributed to the possible impairment of cell entry capacity of GPcl-containing ebolavirus. In addition, the sentences have been modified to be read smoothly.

Figure 5: The authors should clarify in the methods section that the "mock" experiment included the PE anti-human IgG Fc antibody. Without this clarification, the lack of a distinct negative population in the FACS data could be interpreted as non-specific staining with PE. If the PE antibody was added at an equivalent concentration to all panels, what does the directionality of the arrowheads in Figure 5A (labelled PE) and 5B (labelled pHrodo Red) indicate?

Author response: Thank for your advice. In the revised version, we denote that the mock is actually a human IgG isotype in the figure legend. The arrowheads denote the fluorescence intensity of PE or pHrodo on the lateral axis of the plots. Of course, herein the percentage of PE or pHrodo-positive cells is shown.

Figure 6B: These data would benefit from the inclusion of IC50, transparency of antibody concentrations used, and consistency in the direction of antibody concentrations (increasing to the right or left of the x-axis) when compared to Figure 2.

Author response: The concentration of antibody titrated is shown in figure legends. The direction of antibody concentrations is unified throughout the paper. Although IC50 is not included, these data clearly show that AF03-NL rather than AF-03 prominently inhibits the cell entry of EBOV mutants.

Reviewer #1 (Recommendations For The Authors):

Line 143: anti-human should be anti-human.

Line 223: From the SDS-PAGE results, it's not clear that the AF-03 was expressed in the eukaryotic cell line. Please, rephrase the sentence.

Line 263: ELISA experiments can't be used to determine affinity.

Line 394: Flyak et al. generated human antibodies from PBMC samples of Marburg survivors, not plasma samples.

Author response: According to reviewer's advice, the sentences have been modified or corrected to more accurately describe the results. As well, the grammatic errors in the ms have been corrected carefully.

eLife assessment

In this valuable study, the discovery and subsequent design of the AF03-NL chimeric antibody led to a tool for studying filoviruses and provides a possible blueprint for future therapeutics. In general, the data presented are solid, although further improvements can be made in the overall presentation of the results. The work will be of interest to virologists studying antibodies.

Reviewer #2 (Public Review):

Summary:<br /> The authors describe the discovery of a filovirus neutralizing antibody, AF03, by phage display, and its subsequent improvements to include NPC2 that resulted in greater breadth of neutralization. Overall, the manuscript is much improved from first review.

While the authors only use docking studies and this does not convincingly map the AF03 epitope, they do provide compelling evidence that residues Q128, N129, and possibly C226 are part of the epitope or at least close enough to affect binding and neutralisation. This is not conclusive support for their assumption that AF03 targets the NPC1 binding site. However, the authors do show that AF03 competes for MR78 binding to its epitope (in the NPC1 binding site), and this is enough to roughly place the epitope in this region (barring the possibility of an adjacent binding site with steric occlusion of the MR78 epitope).<br /> The authors provide evidence for broad neutralisation, and also provide good support for the internalization of AF03-NL as the mechanism for improved breadth over the original AF03 antibody.

Strengths:<br /> This study shows convincing binding to Marburgvirus GP and neutralization of Marburg viruses by AF03, as well as convincing neutralization of Ebolaviruses by AF03-NL. While there is not good separation of PE-stained populations by FACS in figure 5A, the cell staining data in Figure 5C are compelling to a non-expert in endosomal staining like myself. The control experiments in Figure 7 are compelling showing neutralization by AF03-NL but not AF03 or NPC2 alone or in combination. Altogether these data support the internalisation and stabilisation mechanism that is proposed for the gain in neutralization breadth observed for Ebolaviruses by AF03-NL over AF03 alone.

Weaknesses:<br /> To support their affinity measurements, the authors argue that they show GP is a monomer in Figure 1A by SDS-PAGE. SDS-PAGE cannot be used to assess oligomerisation of GP. Native PAGE or size exclusion profiles would have been better suited to this purpose. If affinity was calculated on a 1GP:2IgG binding sites as the authors imply, then the affinity data are incorrect due to avidity effects. As suggested by a previous reviewer, using monomeric Fab would solve this problem.

The information for figure 2 states: "we investigated if this mutated MARV species was STILL sensitive o AF-03 treatment". But, "we sought to determine whether AF-03 could impede pseudotyped MARV viral entry" only happens in Figure 3. This information for figure 3 has now already been determined in Figure 2 where wildtype MARV is neutralised (black curves) introducing redundancy. The authors should first show that AF-03 can neutralise MARV pseudotyped virus, and then assess whether mutants are STILL sensitive to AF-03.

Figure 1: The visualisation of AF03 modelling and docking is better on a white background, but still difficult to interpret as currently presented. The labels of predicted contact residues are still impossible to read, and the yellow text does not show. As suggested previously, a zoom-in showing predicted co-location with Q128 and N129 would show these data better. It would also be useful to orient the reader with respect to trimeric membrane bound GP.

Figure 2: The presentation of these data is much improved and support the text.

Figure 3: The presentation of these data is much improved and support the text.

Figure 4: The presentation of these data is much improved and support the text.

Figure 5: The presentation of these data is much improved and support the text.

Author Response

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

Reviewer #1 (Public Review):

Major Concerns:

(1) An important point that the authors should clarify in this study is whether mice are detecting qualitative or quantitative differences between fresh and old cat saliva. Do the environmental conditions in which the old saliva was maintained cause degradation of Fel d 4, the main protein known for inducing a defensive response in rodents? (see Papes et al, 2010 again). If that is the case, one would expect that a lower concentration of Fel d 4 in the old saliva after protein degradation would result in reduced antipredator responses. Alternatively, if the authors believe that different proteins that are absent in the old saliva are contributing to the increased defensive responses observed with the fresh saliva, further protein quantification experiments should be performed. An important experiment to differentiate qualitative versus quantitative differences between the two types of saliva would be diluting the fresh saliva to verify if the amount of protein, rather than the type of protein, is the main factor regulating the behavioral differences.

We thank the reviewer for their important suggestions. We agree that both the quality and quantity of molecular components in saliva undergo changes after the saliva is kept at room temperature for 4 hours. Our findings indicate that mice detect these changes through the VNO and adjust their defensive response patterns accordingly. For instance, freezing behavior is reduced in response to 4-hour-old saliva compared to fresh saliva. On the other hand, the duration of interaction with saliva (investigation behavior) remains low, and the stress hormone ACTH level is upregulated in both cases. A future study ought to identify the specific molecules—most likely proteins or peptides—in cat saliva responsible for these distinct defensive responses in mice. While Fel d 4 stands as one of the potential candidates as it has been shown to induce a form of defensive behavior in mice (Papes et al., 2010), there exists a possibility of a different molecule or a combination of multiple molecules playing a role. Once the molecules are identified, it is imperative to investigate how their quantity and quality change over time and how these factors correlate with freezing behavior in mice. Such an exploration will provide answers to this ethologically significant question raised by the reviewer. We added a paragraph in Discussion under the “The VNO as the sensor of predator cues that induce fear-related behavior” section to clarify this.

(2) The authors claim that fresh saliva is recognized as an immediate danger by rodents, whereas old saliva is recognized as a trace of danger. However, the study lacks empirical tests to support this interpretation. With the current experimental tests, the behavioral differences between animals exposed to fresh vs. old saliva could be uniquely due to the reduced amount of the exact same protein (e.g., Fel d 4) in the two samples of saliva.

As mentioned in response to comment 1, we agree with the alterations in both the quality and quantity of molecules within saliva after 4 hours. What we would like to emphasize in our current study is that mice detect these time-dependent changes through the VNO and subsequently adjust their defensive response patterns. Identifying the specific molecules responsible for inducing behavioral changes and investigating their time-dependent alterations is crucial in the next step. We added a paragraph in the Discussion under the 'The VNO as the sensor of predator cues that induce fear-related behavior' section to clarify this.

(3) In Figure 4H, the authors state that there were no significant differences in the number of cFos-positive cells between the two saliva-exposed groups. However, this result disagrees with the next result section showing that fresh and old saliva differentially activate the VMH. It is unclear why cFos quantification and behavioral correlations were not performed in other upstream areas that connect the VNO to the VMH (e.g., BNST, MeA, and PMCo). That would provide a better understanding of how brain activity correlates with the different types of behaviors reported with the fresh vs. old saliva.

We greatly appreciate this valuable advice. We added c-Fos immunoreactivity (IR) data in the BNST, MeApv, and PAG, together with the data for VMH as shown in new Figure 4G-J. Upon exposure to both fresh and old saliva, we observed an upregulation trend of cFos in the MeApv, VMH, and dPAG, but not in the BNST, compared to the control stimulus.

Moreover, we conducted correlation analyses between the numbers of cFos-positive neurons and the duration of freezing behavior in those neural substrates, which have been added to new Figure 5. The numbers of cFos-IR signals in neurons in the BNST and dPAG did not correlate with the duration of freezing behavior in any of the exposure groups (Figure 5C, F). However, in addition to a significant positive correlation in the VMH for the fresh saliva-exposed group (R2 = 0.5708, 95% CI [-0.1449, 0.9714], p = 0.0412) (Figure 5E), we observed a similar positive correlation trend in the MeApv (R2 = 0.3854, 95% CI [0.3845, 0.9525], p = 0.0942), although it was not statistically significant possibly due to low sample numbers (Figure 5D).

Based on these results, our current circuit model is as follows: different numbers of the VNO sensory neurons activated by fresh and old saliva result in differential excitation levels in mitral cells in the AOB. This, in turn, leads to the differential activation of targeting neural substrates, possibly MeApv, resulting in the differential activation of VMH neurons. This model is depicted in Figure 7 and discussed under the section of 'Differential processing of fresh and old saliva signals in the VNO-to-VMH pathway' in the Discussion."

(4) The interpretation that fresh and old saliva activates different subpopulations of neurons in the VMH based on the observation that cFos positively correlates with freezing responses only with the fresh saliva lacks empirical evidence. To address this question, the authors should use two neuronal activity markers to track the response of the same population of VHM cells within the same animals during exposure to fresh vs. old saliva. Alternatively, they could use single-cell electrophysiology or imaging tools to demonstrate that cat saliva of distinct freshness activates different subpopulations of cells in the VMH. Any interpretation without a direct within-subject comparison or the use of cell-type markers would become merely speculative. Furthermore, the authors assume that differential activations of mitral cells between fresh and old saliva result in the differential activation of VMH subpopulations (page 13, line 3). However, there are intermediate structures between the mitral cells and the VMH, which are completely ignored in this study (e.g., BNST, medial amygdala).

We appreciate this important feedback. We agree that performing a same-animal comparison for fresh and old saliva exposure will offer direct evidence of the differential activation of a sub-population of VMH neurons. However, there is technical difficulties. We have stimulated the same animal with the same or different types of swabs (e.g., Freshcontrol, fresh-fresh, fresh-old, or old-fresh) and observed that once mice were exposed to a saliva-containing swab and exhibited freezing behavior, they no longer made contact with the second swab within the timeframe when two different types of neuroactivity markers can be analyzed. As shown in Figure 2A, direct contact with the saliva swab is necessary for triggering saliva-elicited freezing behavior. Therefore, we concur that conducting further investigations into real-time neural activation responses to both fresh and old saliva within the same subjects, using an appropriate stimulus delivery method into the VNO, as demonstrated in (Bansal et al., 2021; Ben-Shaul et al., 2010; Bergan et al., 2014), would be useful to strengthen our argument.

For the second part of the comment regarding the intermediate structures between the mitral cells and the VMH, please refer to our comment above in response to comment 3.

(5) The authors incorrectly cited the Papes et al., 2010 article on several occasions across the manuscript. In the introduction, the authors cited the Papes et al 2010 study to make reference to the response of rodents to chemical cues, but the Papes et al. study did not use any of the chemical cues listed by the authors (e.g., fox feces, snake skin, cat fur, and cat collars). Instead, the Papes et al. 2010 article used the same chemical cue as the present study: cat saliva. The Papes et al. 2010 article was miscited again in the results section where the authors cited the study to make reference to other sources of cat odor that differ from the cat saliva such as cat fur and cat collars. Because the Papes et al. 2010 article has previously shown the involvement of Trpc2 receptors in the VNO for the detection of cat saliva and the subsequent expression of defensive behaviors by using Trpc2-KO mice, the authors should properly cite this study in the introduction and across the manuscript when making reference to their findings.

The study conducted by Papes et al. in 2010 (Papes et al., 2010) explored mouse defensive responses triggered by native odors derived from three natural mouse predator species: cat, snake, and rat. These odors were derived from neck fur swabs, shed skin, and urine, respectively. Notably, all three types of samples induced defensive risk assessment and avoidance behaviors in mice. These responses were significantly diminished in Trpc2 knock-out (KO) mice, which lack the Trpc2 transduction channel in their vomeronasal sensory neurons, resulting in an impairment in transmitting sensory signals to the brain. Moreover, Papes et al. (2010) mentioned that, 'we did find cat saliva, a potential source of fur chemosignals, sufficient to induce c-Fos expression in the AOB and initiate defensive behavior.' While Papes et al. reported c-Fos expression in the AOB as well as behavioral responses induced by cat saliva in C57BL/6 mice, they did not provide information regarding the c-Fos expression or the defensive behavioral responses to cat saliva in Trpc2KO mice. Overall, we highly value these findings and explicitly state in the results section of our study that ‘Cat saliva has been considered as a source of predator cues found on cat fur and collars, which induce defensive behaviors in rodents (Engelke et al., 2021; Papes et al., 2010),’ providing the rationale for our utilization of cat saliva in our experimental design.

(6) In the introduction, the authors hypothesized that the VNO detects predator cues and sends sensory signals to the VMH to trigger defensive behavioral decisions and stated that direct evidence to support this hypothesis is still missing. However, the evidence that cat saliva activates the VMH and that activity in the VMH is necessary for the expression of antipredator defensive response in rodents has been previously demonstrated in a study by Engelke et al., 2021 (PMID: 33947849), which was entirely omitted by the authors.

We appreciate this insightful comment. Our original sentence meant that the direct evidence was missing for the hypothesis that the mouse VNO detects predator cues and sends sensory signals to the VMH, triggering appropriate defensive behavioral decisions. To clarify this, we altered the sentence (the last sentence of the second last paragraph in Introduction) to “However, how the sensory signals detected through the VNO-to-VMH circuitry modulate behavioral decisions in specific contexts remains elusive.

The study in Engelke et al., 2021(Engelke et al., 2021) has shown that cat saliva activates the VMH and that activity in the VMH is necessary for the expression of antipredator defensive response, including freezing behavior, in rats. This important paper is now cited at multiple locations; page 4 line 16, page 9 line 8, and page 14 line 17. Interestingly, the vomeronasal receptor genes expressed in cat saliva-responsive VNO neurons, V2R-A4 subfamily genes, seem to have expanded independently within mice and rats, lacking direct V2R-A4 orthologues between mice and rats (Rocha et al. submitted). Therefore, exploring the sensory mechanism behind the induction of defensive behavioral responses in rats by cat saliva would be highly intriguing. Comparing the mechanism operating in rats with that observed in mice could offer valuable insights into understanding how the divergent sensory signaling pathways lead to the VMH-mediated defensive behavioral responses across different species.

(7) In the discussion, the authors stated that their findings suggest that the induction of robust freezing behavior is mediated by a distinct subpopulation of VMH neurons. The authors should cite the study by Kennedy et al., 2020 (PMID: 32939094) that shows the involvement of VMH in the regulation of persistent internal states of fear, which may provide an alternative explanation for why distinct concentrations of saliva could result in different behavioral outcomes.

We appreciate this valuable advice to cite this important paper. It is now cited at page 14 line 17 in the Discussion under “Differential activation of VMH neurons potentially underlying distinct intensities of freezing behavior.” We agree that it is intriguing to hypothesize that different freshness of cat saliva induces different degree of persistence of neural activity in a subpopulation of VMH neurons, which regulates the freezing behavior intensity.

(8) The anatomical connectivity between the olfactory system and the ventromedial hypothalamus (VMH) in the abstract is unclear. The authors should clarify that the VMH does not receive direct inputs from the vomeronasal organ (VNO) nor the accessory olfactory bulb (AOB) as it seems in the current text.

We apologize for the confusion caused by our statement in the abstract. The reviewer is correct that the VMH does not receive direct inputs from the VNO and AOB. The abstract now states: 'The vomeronasal organ (VNO) is one of the major sensory input channels through which predator cues are detected with ascending inputs to the medial hypothalamic nuclei, especially to the ventromedial hypothalamus (VMH), through the medial amygdala (MeA) and bed nucleus of the stria terminalis (BNST).’

Reviewer #2 (Public Review):

Weakness:

The findings are relatively preliminary. The identities of the receptor and the ligand in the cat saliva that induces the behavior remain unclear. The identity of VMH cells that are activated by the cat saliva remains unclear. There is a lack of targeted functional manipulation to demonstrate the role of V2R-A4 or VMH cells in the behavioral response to cat saliva.

We concur with the reviewer’s comments and agree with the necessity to explore the behavioral response to cat saliva in mice with V2R-A4 receptor(s) knocked out, alongside those with targeted functional manipulations in the VMH. These future studies will allow us to further elucidate the molecular and neural mechanisms underlying this sensory-tohypothalamic circuit.

Reviewer #3 (Public Review):

Weaknesses:

(1) It is unclear if fresh and old saliva indeed alter the perceived imminence predation, as claimed by the authors. Prior work indicates that lower imminence induces anxiety-related actions, such as re-organization of meal patterns and avoidance of open spaces, while slightly higher imminence produces freezing. Here, the authors show that fresh and old predator saliva only provoke different amounts of freezing, rather than changing the topography of defensive behaviors, as explained above. Another prediction of predatory imminence theory would be that lower imminence induced by old saliva should produce stronger cortical activation, while fresh saliva would activate the amygdala, if these stimuli indeed correspond to significantly different levels of predation imminence.

We thank the reviewer for this valuable insight. In our current study, we exclusively compared defensive behavioral responses to 15-minute-old and 4-hour-old cat saliva in mice within their home cages. In future studies, it would be intriguing to expand this investigation by examining behavioral changes in response to saliva collected at additional time points across diverse behavioral settings. Additionally, exploring neural activity in various brain regions in future studies would complement our understanding of these responses.

(2) It is known that predator odors activate and require AOB, VNO, and VMH, thus replications of these findings are not novel, decreasing the impact of this work.

We acknowledge the previous findings mentioned by the reviewer. Our finding in this paper is that cat saliva samples with different freshness predominantly activate different numbers of VNO sensory neurons expressing the same subfamily of sensory receptors, which results in differential activation of the downstream circuit to modulate behavioral outputs.

(3) There is a lack of standard circuit dissection methods, such as characterizing the behavioral effects of increasing and decreasing the neural activity of relevant cell bodies and axonal projections, significantly decreasing the mechanistic insights generated by this work.

We thank the reviewer for the valuable comments. We acknowledge that exploring the behavioral effects through the manipulation of specific cell types within defined neural substrates, along with characterizing circuit connectivity, is crucial to understand this circuit more thoroughly in future studies.

(4) The correlation shown in Figure 5c may be spurious. It appears that the correlation is primarily driven by a single point (the green square point near the bottom left corner). All correlations should be calculated using Spearman correlation, which is non-parametric and less likely to show a large correlation due to a small number of outliers. Regardless of the correlation method used, there are too few points in Figure 5c to establish a reliable correlation. Please add more points to 5c.

We thank the reviewer for this important suggestion. We assessed normality of the data using the Shapiro-Wilk and Kolmogorov-Smirnov tests, confirming that the dataset is parametric. We anticipate employing a larger sample size in future studies to further examine rigorous correlation patterns.

(5) Some of the findings are disconnected from the story. For example, the authors show that V2R-A4-expressing cells are activated by predator odors. Are these cells more likely to be connected to the rest of the predatory defense circuit than other VNO cells?

Yes, our hypothesis posits that V2R-A4-expressing VNO sensory neurons serve as receptor neurons for predator cues present in cat saliva. Additionally, we assume that these specific sensory neurons have stronger anatomical connections with the defensive circuit compared to VNO sensory neurons expressing other receptor subfamilies. In our modified Discussion section, we discussed this point under “V2R-A4 subfamily as the receptor for predator cues in cat saliva.”

(6) Were there other behavioral differences induced by fresh compared to old saliva? Do they provoke differences in stretch-attend risk evaluation postures, number of approaches, the average distance to odor stimulus, the velocity of movements towards and away from the odor stimulus, etc?

We appreciate the reviewer's valuable comments. We have now incorporated an analysis of stretch-sniff risk assessment behavior, presented in new Figure 1F (graph) and Supplemental Figure 1B (raster plot). Mice exhibited stretch-sniff risk assessment behavior, which remained consistent across control, fresh saliva, and old saliva swabs. Additionally, we have also included a raster plot for direct investigation, previously noted as ‘interaction’ in the original manuscript (Supplemental Figure 1C). Mice exposed to a swab containing either fresh or old saliva significantly avoided directly investigating the swab. In contrast, mice exposed to a clean control swab spent a significant amount of time directly investigating the swab, engaging in behaviors such as sniffing and chewing (Figure 1G). A comparison of temporal behavioral patterns revealed a slightly higher frequency of direct investigation behavior toward old saliva compared to fresh saliva at the beginning of the exposure period (Supplemental Figure 1C).

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(A) In the discussion (page 13, line 13), the authors proposed approaches to isolate receptors among the V2R-A4 subfamily that could be responsible for the detection of predator cues in cat saliva such as mRNA profiling from cells isolated from VNO GCaMP imaging. However, the authors argue that this method can lead to false positive results. The authors should clarify what they mean by this exactly.

We meant that pairing of kairomones and their cognate vomeronasal receptors is overall challenging, and subsequent confirmations by performing loss-of-function, as well as gainof-function studies, are necessary to avoid false positive receptor-ligand pairings. We modified the sentence in the discussion as follows: “…. as well as receptor mRNA profiling from isolated single cells activated by cat saliva in GcaMP imaging using the VNO slices in vitro (Haga-Yamanaka et al., 2014; Wong et al., 2020). Receptor candidates identified using either of the methods can be further confirmed by examining necessity and sufficiency for detecting cat saliva using genetically modified mouse lines.”

(B) In the discussion, the authors mention that imminent predator cues present in the cat saliva activate a specific population of VMN neurons. However, the authors have not demonstrated that imminent predator cues exist and the differences between fresh and old saliva are not simply a matter of concentration and integrity of the same protein (see a similar concern in item 2 above).

In alignment with our responses to the reviewer’s public comments 1 and 2, we acknowledge the changes in both the quality and quantity of molecules in cat saliva when kept at room temperature for 4 hours. Our findings demonstrate that mice detect this timedependent alteration through the VNO, leading to subsequent adjustments in their defensive response patterns. The identification of specific molecules responsible for inducing behavioral changes and an exploration of their time-dependent alterations are crucial steps in our ongoing research. To provide further clarification, we have added a paragraph in the discussion section under 'The VNO as the sensor of predator cues that induce fear-related behavior.’

(C) In the introduction, the authors cite several studies and reviews that investigated sensory neural circuits that mediate behavioral responses to chemical predator cues in mice. However, the majority of these studies used rats. Therefore, it is recommended to instead indicate that these studies focus on using rodent models.

We appreciate this insightful comment. We have now replaced the term 'mice/mouse' with 'rodents' in corresponding parts of the manuscript.

(D)The description of the extended amygdala is unclear and gives the impression that the posteroventral part of the medial amygdala is also part of the extended amygdala (page 3, line 25).

We appreciate the reviewer’s important feedback. We have removed the phrase 'the extended amygdala consisting of' from the text.

(E) The authors should justify why they have focused on the role of V2R-A4 in cat saliva detection. As shown in the Figure 3A schematic, many other receptors within the V2R family could have been evaluated. Additionally, the authors should indicate how many mice were used for calculating the ratio for each receptor in Figure 3C, and a group comparison should be performed.

As shown in Supplemental Figure 2 and Figure 3C, our initial investigation involved assessing the co-localization of pS6 signals with signals derived from in situ hybridization probes for all V2R subfamilies. Each probe was designed to recognize all the receptor genes within the subfamily under the tested conditions. This examination led to the identification of V2R-A4, whose probe signals overlap with pS6 signals induced by exposure to cat saliva. In Figure 3C, the percentage of total overlap between the in situ probe and pS6 signals in VNO sections was examined from n=3-6 animals, which is now mentioned in the modified figure legend.

(F) The authors should make it clear to readers at the very beginning of the manuscript that the behavioral differences between fresh and old saliva are not caused by the inefficiency of the old cat saliva to induce defensive responses. Thus, other antipredator behavioral responses should be also quantified (e.g., avoidance time, number and time of investigations to the cat saliva source, risk-assessment, etc.)

We appreciate this valuable comment from the reviewer. In the original version of our manuscript, we used the term 'interaction' to indicate 'direct interaction with the swab for investigation.' We have now replaced the term 'interaction' with 'direct investigation' and added the temporal patterns of these behavioral episodes in Supplemental Figure 1C. Our observations indicate that mice avoid directly investigating both fresh and old saliva compared to the control (Figure 1G). However, there is a slight increase in investigation behavior toward old saliva at the beginning of exposure compared to fresh saliva (Supplemental Figure 1C). Furthermore, we have included the duration (Figure 1F) and temporal patterns (Supplemental Figure 1B) of stretch-sniff risk assessment behavior. Notably, stretch-sniff behavior did not differ towards control, fresh, and old saliva swabs.

(G) The selected representative images for Gαo- and pS6-labeled neurons in Figure 2 should have similar levels of DAPI labeling. Further, the plot depicting the duration of freezing as a function of pS6-IR signals in the VNO (Figure 2H) is difficult to follow. The authors should indicate on the graph which data points represent fresh or old cat saliva exposure, similar to the style used in Figure 5 plots.

We have replaced the representative image in Figure 2E to align the DAPI intensity. Additionally, we updated the data points in Figure 2H and introduced a color code to indicate saliva types.

(H) The schematic in Figure 4 is misleading because the AOB does not directly project to the VMH. The authors should explain which regions are conveying indirect predator information from AOB to VMH (see a similar concern in item 7 above).

We thank the reviewer’s important feedback. We modified the image in Figure 4A to show the entire defensive behavior circuit initiated from the VNO.

Reviewer #2 (Recommendations For The Authors):

(1) This result suggests that V2R-A4 may be the dominant VR for mice to detect cat saliva.

Future studies should determine the identity of the receptor and the ligand in the cat saliva. Additionally, the functional importance of V2R-A4 remains unclear. It is important to knockout the receptor and test changes in cat saliva-induced freezing.

We concur with the reviewer’s comments and recognize the necessity of exploring the behavioral response to cat saliva in mice with V2R-A4 receptor(s) knocked out. Moreover, the identification of the ligand in cat saliva is critical for a deeper understanding of the molecular mechanisms in future studies.

(2) AOB does not project to VMH directly. Other known important nodes for the predator defense circuit include MeApv, BNST, PMd, AHN, and PAG. It will be helpful to provide c-Fos data in those regions (especially MEA and BNST as they are between AOB and VMH) to provide a complete picture of how the brain processes cat saliva to induce the behavior change.

We appreciate this important feedback by the reviewer. We have now added c-Fos expression analysis data in the BNST, MeApv, and PAG, in addition to the VMH. Upon exposure to fresh and old saliva, we observed the upregulation of cFos in the MeApv, VMH, and dPAG, but not in the BNST, compared to the control stimulus. The data are now shown in Figure 4G-J. Moreover, we also added correlation analyses between the numbers of cFospositive neurons and the duration of freezing behavior in those neural substrates to Figure 5. The numbers of cFos-IR signals in neurons in the BNST and dPAG, did not correlate with the duration of freezing behavior in any of the exposure groups (Figure 5C, F). However, in addition to a significant positive correlation in the fresh saliva-exposed group in the VMH (R2 = 0.5708, 95% CI [-0.1449, 0.9714], p = 0.0412) (Figure 5E), we observed a similar positive correlation trend in the MeApv (R2 = 0.3854, 95% CI [-0.3845, 0.9525], p = 0.0942), although it was not statistically significant possibly due to low sample numbers (Figure 5D). Based on these results, our current circuit model is as follows: different numbers of the VNO sensory neurons activated by fresh and old saliva result in differential excitation levels in mitral cells in the AOB. Differential excitation of mitral cells leads to the differential activation of targeting neural substrates, possibly MeApv, which results in differential activation of VMH neurons. This model is depicted in Figure 7 and discussed under the section of “Differential processing of fresh and old saliva signals in the VNO-toVMH pathway” in Discussion.

(3) It is interesting that activation level difference in the VNO by old and fresh cat saliva does not transfer to AOB. It could be informative to examine the correlation between VNO and AOB p6/c-Fos cell number and AOB and VMH c-Fos cell number across animals to understand whether the activation levels across those regions are related. If they are not correlated, it could be helpful to add a discussion regarding potential reasons, e.g. neuromodulatory inputs to the AOB.

We agree that analyzing the number of pS6/cFos-positive cells from all the regions in the same animals are ideal; however, due to technical difficulties, we were unable to collect the entire set of neural substrates from the same animals.

(4) Please indicate n in all figure plots and specify what individual dots mean. In Figure 4h, there are 7 dots in the old saliva group, presumably indicating 7 animals. In Figure 6b, there appear to be more than 7 dots for the old cat saliva group. Are there more than 7 animals? If so, why are they not included in Figure 4h? If not, what does each dot mean? Note that each dot should represent an independent sample. One animal should not contribute more than one dot.

We apologize for the confusion about Figure 6b. Each of these dots indicates the number of cFos signals in a single VMH hemisphere sample. The data used for this analysis were the same as the ones for the VMH used in Figure 4. This is now clarified in the figure legends.

(5) The identification of a cluster of VMHdm cells uniquely activated by fresh cat saliva urine is interesting. It will be important to identify the molecular handle of the cells to facilitate further investigation. This could be achieved using either activity-dependent RNAseq or double in situ of saliva-induced c-Fos and candidate genes (candidate gene may be identified based on the known gene expression pattern).

We agree that these experiments are very valuable. We would like to perform those experiments in future studies.

Reviewer #3 (Recommendations For The Authors):

(1) Please cite recent relevant papers showing VMH activity induced by predators, such as https://pubmed.ncbi.nlm.nih.gov/33115925/ and https://pubmed.ncbi.nlm.nih.gov/36788059/

We thank the reviewer’s suggestion to cite these important papers. https://pubmed.ncbi.nlm.nih.gov/33115925/ (Esteban Masferrer et al., 2020) and https://pubmed.ncbi.nlm.nih.gov/36788059/ (Tobias et al., 2023) are now cited at page 14 line 17 in the Discussion under “Differential activation of VMH neurons potentially underlying distinct intensities of freezing behavior.”

(2) Add complete statistical information in the figure legends of all figures, which should include n, name of test used, and exact p values.

We included statistical analysis results in figure legends; for Figure 6B, we provided statistical analysis results in Supplemental Table 1.

(3) Please paste all figure legends directly below their corresponding figure to make the manuscript easier to read.

We have added figure legends directly below their corresponding figures.

Editor's note:

Should you choose to revise your manuscript, please include full statistical reporting including exact p-values wherever possible alongside the summary statistics (test statistic and df) and 95% confidence intervals. These should be reported for all key questions and not only when the p-value is less than 0.05.

Statistics analysis results have been included in figure legends and supplemental table 1.

References

Bansal R, Nagel M, Stopkova R, Sofer Y, Kimchi T, Stopka P, Spehr M, Ben-Shaul Y. 2021. Do all mice smell the same? Chemosensory cues from inbred and wild mouse strains elicit stereotypic sensory representations in the accessory olfactory bulb. BMC Biol 19:133.

Ben-Shaul Y, Katz LC, Mooney R, Dulac C. 2010. In vivo vomeronasal stimulation reveals sensory encoding of conspeciic and allospeciic cues by the mouse accessory olfactory bulb. Proc Natl Acad Sci U S A 107:5172‒5177.

Bergan JF, Ben-Shaul Y, Dulac C. 2014. Sex-speciic processing of social cues in the medial amygdala. Elife 3:e02743.

Engelke DS, Zhang XO, OʼMalley JJ, Fernandez-Leon JA, Li S, Kirouac GJ, Beierlein M, Do-Monte FH. 2021. A hypothalamic-thalamostriatal circuit that controls approachavoidance conlict in rats. Nat Commun 12:2517.

Esteban Masferrer M, Silva BA, Nomoto K, Lima SQ, Gross CT. 2020. Differential Encoding of Predator Fear in the Ventromedial Hypothalamus and Periaqueductal Grey. J Neurosci 40:9283‒9292.

Papes F, Logan DW, Stowers L. 2010. The vomeronasal organ mediates interspecies defensive behaviors through detection of protein pheromone homologs. Cell 141:692‒703.

Tobias BC, Schuette PJ, Maesta-Pereira S, Torossian A, Wang W, Sethi E, Adhikari A. 2023. Characterization of ventromedial hypothalamus activity during exposure to innate and conditioned threats. Eur J Neurosci 57:1053‒1067.

eLife assessment

This valuable study addresses one way in which animals identify predator-associated cues and respond in a manner that reflects the imminence of the potential threat. The report shows that, in mice, fresh saliva from a natural predator (cat) elicits a greater defensive response compared to old cat saliva and implicates the vomeronasal organ and ventromedial hypothalamus as part of a circuit that underlies this process. While the study has potential, the results are somewhat preliminary, and as such support for the primary conclusions is incomplete.

Reviewer #1 (Public Review):

Summary:

Animals in natural environments need to identify predator-associated cues and respond with the appropriate behavioral response to survive. In rodents, some chemical cues produced by predators (e.g., cat saliva) are detected by chemosensory neurons in the vomeronasal organ (VNO). The VNO transmits predator-associated information to the accessory olfactory bulb, which in turn projects to the medial amygdala and the bed nucleus of the stria terminalis, two regions implicated in the initiation of antipredator defensive behaviors. A downstream area to these two regions is the ventromedial hypothalamus (VMH), which has been shown to control both active (i.e., flight) and passive (i.e, freezing) antipredator defensive responses via distinct efferent projections to the anterior hypothalamic nucleus or the periaqueductal gray, respectively. However, whether differences in predator-associated sensory information initially processed in the VNO and further conveyed to the VMH can trigger different types of behavioral responses remained unexplored. To address this question, here the authors investigated the behavioral responses of mice exposed to either fresh or old cat saliva, and further compared the underlying neural circuits that are activated by cat saliva with different freshness.

The scientific question of the study is valid, the experiments were well-performed, and the statistical analyses are appropriate. However, there are some concerns that may directly affect the main interpretation of the results.

Major Concerns:

(1) An important point that the authors should clarify in this study is whether mice are detecting qualitative or quantitative differences between the fresh and old cat saliva. Do the environmental conditions in which the old saliva was maintained cause a degradation of Fel d 4, the main protein known for inducing a defensive response in rodents? (see Papes et al, 2010 again). If that is the case, one would expect that a lower concentration of Fel d 4 in the old saliva after protein degradation would result in reduced antipredator responses. Alternatively, if the authors believe that different proteins that are absent in the old saliva are contributing to the increased defensive responses observed with the fresh saliva, further protein quantification experiments should be performed. An important experiment to differentiate qualitative versus quantitative differences between the two types of saliva would be diluting the fresh saliva to verify if the amount of protein, rather than the type of protein, is the main factor regulating the behavioral differences.

(2) The authors claim that fresh saliva is recognized as an immediate danger by rodents, whereas old saliva is recognized as a trace of danger. However, the study lacks empirical tests to support this interpretation. With the current experimental tests, the behavioral differences between animals exposed to fresh vs. old saliva could be uniquely due to the reduced amount of the exact same protein (e.g., Fel d 4) in the two samples of saliva.

(3) In Figure 4H, the authors state that there were no significant differences in the number of cFos-positive cells between the two saliva-exposed groups. However, this result disagrees with the next result section showing that fresh and old saliva differentially activate the VMH. It is unclear why cFos quantification and behavioral correlations were not performed in other upstream areas that connect the VNO to the VMH (e.g., BNST, MeA, and PMCo). That would provide a better understanding of how brain activity correlates with the different types of behaviors reported with the fresh vs. old saliva.

(4) The interpretation that fresh and old saliva activates different subpopulations of neurons in the VMH based on the observation that cFos positively correlates with freezing responses only with the fresh saliva lacks empirical evidence. To address this question, the authors should use two neuronal activity markers to track the response of the same population of VHM cells within the same animals during exposure to fresh vs. old saliva. Alternatively, they could use single cell electrophysiology or imaging tools to demonstrate that cat saliva of distinct freshness activates different subpopulations of cells in the VMH. Any interpretation without a direct within-subject comparison or the use of cell-type markers would become merely speculative. Furthermore, the authors assume that differential activations of mitral cells between fresh and old saliva result in the differential activation of VMH subpopulations (page 13, line 3). However, there are intermediate structures between the mitral cells and the VMH, which are completely ignored in this study (e.g., BNST, medial amygdala).

(5) The authors incorrectly cited the Papes et al., 2010 article on several occasions across the manuscript. In the introduction, the authors cited the Papes et al 2010 study to make reference to the response of rodents to chemical cues, but the Papes et al. study did not use any of the chemical cues listed by the authors (e.g., fox feces, snake skin, cat fur, and cat collars). Instead, the Papes et al. 2010 article used the same chemical cue as the present study: cat saliva. The Papes et al. 2010 article was miscited again in the results section where the authors cited the study to make reference to other sources of cat odor that differ from the cat saliva such as cat fur and cat collars. Because the Papes et al. 2010 article has previously shown the involvement of Trpc2 receptors in the VNO for the detection of cat saliva and the subsequent expression of defensive behaviors by using Trpc2-KO mice, the authors should properly cite this study in the introduction and across the manuscript when making reference to their findings.

(6) In the introduction, the authors hypothesized that the VNO detects predator cues and sends sensory signals to the VMH to trigger defensive behavioral decisions and stated that direct evidence to support this hypothesis is still missing. However, the evidence that cat saliva activates the VMH and that activity in the VMH is necessary for the expression of antipredator defensive response in rodents has been previously demonstrated in a study by Engelke et al., 2021 (PMID: 33947849), which was entirely omitted by the authors.

(7) In the discussion, the authors stated that their findings suggest that the induction of robust freezing behavior is mediated by a distinct subpopulation of VMH neurons. The authors should cite the study by Kennedy et al., 2020 (PMID: 32939094) that shows the involvement of VMH in the regulation of persistent internal states of fear, which may provide an alternative explanation for why distinct concentrations of saliva could result in different behavioral outcomes.

(8) The anatomical connectivity between the olfactory system and the ventromedial hypothalamus (VMH) in the abstract is unclear. The authors should clarify that the VMH does not receive direct inputs from the vomeronasal organ (VNO) nor the accessory olfactory bulb (AOB) as it seems in the current text.

UNADDRESSED AND ADDITIONAL CONCERNS (RE-SUBMISSION)

In this revised version of the manuscript, the authors have made important modifications in the text, inserted new references, and incorporated additional quantifications of cFos immunolabeling in three brain regions, as recommended by the reviewers. While these modifications have significantly improved the quality of the manuscript, other critical concerns raised during the initial submission of the manuscript (Major concerns 1, 2, and 4; some of them also raised by the other reviewers) were not properly addressed by the authors. On several occasions, the authors recognize the importance of clarifying the points for the correct interpretation of the results but opt for leaving the open questions to be addressed during future studies. Therefore, the authors might consider adding a new section at the end of the manuscript to include all the caveats and future directions.

In addition to these unaddressed concerns, some new issues have emerged in the new version of the manuscript. For example, the following paragraph introduced in the discussion section is not supported by the experimental findings.

"We assume that such differential activations of the mitral cells between fresh and old saliva result in the differential activation of targeting neural substrates, possibly MeApv, which results in differential activation of VMH neurons (Figure 7)."

Although the authors did not observe statistical differences in cFos expression in the pvMeA among groups, they claim that the differences in cFos expression in the VMH between fresh vs. old saliva are mediated by differential activation of upstream neurons in the MeApv. The lack of statistical differences may be caused by the reduced number of subjects in each group, as recognized in the text by the authors. Moreover, the authors propose that in addition to fel d 4, multiple molecules present in the cat saliva can be inducing distinct defensive responses in the animals, but they do not provide any reference to support their claim.

Reviewer #2 (Public Review):

In this study, Nguyen et al. showed that cat saliva can robustly induce freezing behavior in mice. This effect is mediated through the accessory olfactory system as it requires physical contact and is abolished in Trp2 KO mice. The authors further showed that V2R-A4 cluster is responsive to cat saliva. Lastly, they demonstrated c-Fos induction in AOB and VMHdm/c by the cat saliva. The c-Fos level in the VMHdm/c is correlated with the freezing response.

Strength:

The study opens an interesting direction. It reveals the potential neural circuit for detecting cat saliva and driving defense behavior in mice. The behavior results and the critical role of the accessory olfactory system in detecting cat saliva are clear and convincing.

Weakness:

The findings are relatively preliminary. The identities of the receptor and the ligand in the cat saliva that induces the behavior remain unclear. The identity of VMH cells that are activated by the cat saliva remains unclear. There is a lack of targeted functional manipulation to demonstrate the role of V2R-A4 or VMH cells in the behavioral response to the cat saliva.

Reviewer #3 (Public Review):

Summary:

Nguyen et al show data indicating that the vomeronasal organ (VNO) and ventromedial hypothalamus (VMH) are part of a circuit that elicits defensive responses induced by predator odors. They also show that using fresh or old predator saliva may be a method to change the perceived imminence of predation. The authors also identify a family of VNO receptors that are activated by cat saliva. Next, the authors show how different components of this defensive circuit are activated by saliva, as measured by fos expression. Though interesting, the findings are not all integrated into a single narrative, and some of the results are only replications of earlier findings using modern methods. Overall, these findings provide incremental advance.

Strengths:

(1) Predator saliva is a stimulus of high ethological relevance

(2) The authors performed a careful quantification of fos induction across the anterior-posterior axis in figure 6

Weaknesses:

(1) It is unclear if fresh and old saliva indeed alter the perceived imminence of predation, as claimed by the authors. Prior work indicates that lower imminence induces anxiety-related actions, such as re-organization of meal patterns and avoidance of open spaces, while slightly higher imminence produces freezing. Here, the authors show that fresh and old predator saliva only provoke different amounts of freezing, rather than changing the topography of defensive behaviors, as explained above. Another prediction of predatory imminence theory would be that lower imminence induced by old saliva should produce stronger cortical activation, while fresh saliva would activate amygdala, if these stimuli indeed correspond to significantly different levels of predation imminence.

(2) It is known that predator odors activate and require AOB, VNO and VMH, thus replications of these findings are not novel, decreasing the impact of this work.

(3) There is a lack of standard circuit dissection methods, such as characterizing the behavioral effects of increasing and decreasing neural activity of relevant cell bodies and axonal projections, significantly decreasing the mechanistic insights generated by this work

(4) The correlation shown in Figure 5c may be spurious. It appears that the correlation is primarily driven by a single point (the green square point near the bottom left corner). All correlations should be calculated using Spearman correlation, which is non-parametric and less likely to show a large correlation due to a small number of outliers. Regardless of the correlation method used, there are too few points in Figure 5c to establish a reliable correlation. Please add more points to 5c.

(5) Please cite recent relevant papers showing VMH activity induced by predators, such as https://pubmed.ncbi.nlm.nih.gov/33115925/ and https://pubmed.ncbi.nlm.nih.gov/36788059/

(6) Add complete statistical information in the figure legends of all figures, which should include n, name of test used and exact p values.

(7) Some of the findings are disconnected from the story. For example, the authors show V2R-A4-expressing cells are activated by predator odors. Are these cells more likely to be connected to the rest of the predatory defense circuit than other VNO cells?

(8) Please paste all figure legends directly below their corresponding figure to make the manuscript easier to read

(9) Were there other behavioral differences induced by fresh compared to old saliva? Do they provoke differences in stretch-attend risk evaluation postures, number of approaches, average distance to odor stimulus, velocity of movements towards and away the odor stimulus, etc?

Reviewer #1 (Public Review):

Summary:

The paper combines phenotypic and genomic analyses of the "sheltered load" (i.e. the accumulation of deleterious mutations linked to S-Loci that are hidden from selection in the homozygous state) in Arabidopsis. The authors compare results to previous theoretical predictions concerning the extent of the load in dominant vs recessive S-alleles, and further develop exciting theory to reconcile differences between previous theory and observed results.

Strengths:

This is a very nice combination of theory and data to address a classical question in the field.

Weaknesses:

The "genetic load" is a poorly defined concept in general, and its quantification via the number of putatively deleterious mutations is quite difficult. Furthermore counting up the number of derived mutations at fully constrained nucleotides may not be a great estimate of the load, and certainly does not allow for evaluation of recessivity -- a concept critical to ideas concerning the sheltered load. Alternative approaches - including estimating the severity of mutations - could be helpful as well. This imperfection in available approaches to test theory must be acknowledged more strongly by the authors.

eLife assessment

This study presents valuable empirical work and simulations that are relevant for the evolution of genetic load linked to self-incompatibility alleles in Arabidopsis. The evidence supporting the findings is solid but could be improved by a more detailed quantitative assessment of the impacts of deleterious alleles and their dominance. The simulation results are somewhat incomplete, as details of the approach and code do not appear to be available, but this could be easily remedied. The work will be of relevance to geneticists interested in the evolution of allelic diversity in similar systems.

Reviewer #2 (Public Review):

Summary:

This study looks into the complex dominance patterns of S-allele incompatibilities in Brassicaceae, through which it attempts to learn more about the sheltering of deleterious load. I found several weak points in the analyses that diminished my excitement about the results. In particular, the way in which deleterious mutations were classified lacked the ability to distinguish the severity of the mutations and thus their expected associated dominance. Furthermore, the simulation approach could have provided this exact sort of insight but was not designed to do so, making this comparison to the empirical data also less than exciting for me.

Major and minor comments:

I think the introduction (or somewhere before we dive into it in the results) of the dominance hierarchy for the S-alleles needs a more in-depth explanation. Not being familiar with this beforehand really made this paper inaccessible to me until I then went to find out more before continuing. I would expect this paper to be broad enough that self-contained information makes it accessible to all readers. For example, lines 110-115 could be in the Introduction.

Along with my above comment, perhaps it is not my place to comment, but I find the paper not of a broad enough scope to be of interest to a broad readership. This S-allele dominance system is more than simple balancing selection, it is a very complex and specific form of dominance between several haplotypes, and the mechanism of dominance does not seem to be genetic. I am not sure that it thus extrapolates to broad comments on general dominance and balancing selection, e.g. it would not be the same as considering inversions and this form of balancing selection where we also expect recessive deleterious mutations to accumulate.

It would have been particularly interesting, or a nice addition, to see deleterious mutations classed by something like SNPeff or GERP where you can have different classes of moderate to severe deleterious variants, which we would expect also to be more recessive the more deleterious they are. In line with my next comment on the simulations, I think relative differences between mutations expected to be more or less dominant may be even more insightful into the process of sheltering which may or may not be going on here.

In the simulations, h=0 and s=0.01 (as in Figure 5) for all deleterious mutations seems overly simplistic, and at the convenient end for realistic dominance. I think besides recessive lethals which we expect to be close to h=0 would have a much larger selection coefficient, and other deleterious mutations would only be partially recessive at such an s value. I expect this would change some of the simulation results seen, though to what degree I am not certain. It would be nice to at least check the same exact results for h=0.3 or 0.2 (or additionally also for recessive lethals, e.g. h=0 and s=-0.9). I would also disagree with the statement in line 677, many studies have shown, particularly those on balancing selection, that partially recessive deleterious mutations are not eliminated by natural selection and do play a role in population genetic dynamics. I am also not surprised that extinction was found for higher s values when the mutation rate for such mutations was very high and the distribution of s values was constant. An influx of such highly deleterious mutations is unlikely to ever let a population survive, yet that does NOT mean that in nature, the rare influx of such mutations does lead to them being sheltered. I find overall that the simulation results contribute very little, to none, to this paper, as without something more realistic, like a simultaneous distribution of s and h values, you cannot say which, if any class of these mutations are the ones expected to accumulate because of S-allele dominance. Rather they only show the disappointing or less exciting result that fully recessive, weakly deleterious mutations (which I again think do not even exist in nature as I said above) have minor, to no effect across the classes of S-allele dominance. They provide no insight into whether any type of recessive deleterious mutation can accumulate under the S-allele dominance hierarchy, and that is the interesting question at hand. I would either remove these simulations or redo them in another approach. The authors never mention what simulation approach was used, so I can only assume this is custom, in-house code. Yet I do not find that code provided on the github page. I do not know if the lack of a distribution for h and s values is then a choice or a programming limitation, but I see it as one that should be overcome if these simulations are meant to be meaningful to the results of the study.

Author Response

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

PUBLIC REVIEWS

Reviewer #1 (Public Review):

In this study, the authors investigate the role of triglycerides in spermatogenesis. This work is based on their previous study (PMID: 31961851) on triglyceride sex differences in which they showed that somatic testicular cells play a role in whole body triglyceride homeostasis. In the current study, they show that lipid droplets (LDs) are significantly higher in the stem and progenitor cell (pre-meiotic) zone of the adult testis than in the meiotic spermatocyte stages. The distribution of LDs anti-correlates with the expression of the triglyceride lipase Brummer (Bmm), which has higher expression in spermatocytes than early germline stages. Analysis of a bmm mutant (bmm[1]) - a P-element insertion that is likely a hypomorphic - and its revertant (bmm[rev]) as a control shows that bmm acts autonomously in the germline to regulate LDs. In particular, the number of LDs is significantly higher in spermatocytes from bmm[1] mutants than from bmm[rev] controls. Testes from males with global loss of bmm (bmm[1]) are shorter than controls and have fewer differentiated spermatids. The zone of bam expression, typically close to the niche/hub in WT, is now many cell diameters away from the hub in bmm[1] mutants. There is an increase in the number of GSCs in bmm[1] homozygotes, but this phenotype is probably due to the enlarged hub. However, clonal analyses of GSCs lacking bmm indicate that a greater percentage of the GSC pool is composed of bmm[1]-mutant clones than of bmm[rev]-clones. This suggests that loss of bmm could impart a competitive advantage to GSCs, but this is not explored in greater detail. Despite the increase in number of GSCs that are bmm[1]-mutant clones, there is a significant reduction in the number of bmm[1]-mutant spermatocyte and post-meiotic clones. This suggests that fewer bmm[1]mutant germ cells differentiate than controls. To gain insights into triglyceride homeostasis in the absence of bmm, they perform mass spec-based lipidomic profiling. Analyses of these data support their model that triglycerides are the class of lipid most affected by loss of bmm, supporting their model that excess triglycerides are the cause of spermatogenetic defects in bmm[1]. Consistent with their model, a double mutant of bmm[1] and a diacylglycerol Oacyltransferase 1 called midway (mdy) reverts the bmm-mutant germline phenotypes.

There are numerous strengths of this paper. First, the authors report rigorous measurements and statistical analyses throughout the study. Second, the authors utilize robust genetic analyses with loss-of-function mutants and lineage-specific knockdown. Third, they demonstrate the appropriate use of controls and markers. Fourth, they show rigorous lipidomic profiling. Lastly, their conclusions are appropriate for the results. In other words, they don't over-state the results. Overall, the rigorously quantified results support the major aim that appropriate regulation of triglycerides are needed in a germline cell-autonomous manner for spermatogenesis.

This paper should have a positive impact on the field. First and foremost, there is limited knowledge about the role of lipid metabolism in spermatogenesis. The lipidomic data will be useful to researchers in the field who study various lipid species. Going forward, it will be very interesting to determine what triglycerides regulate in germline biology. In other words, what functions/pathways/processes in germ cells are negatively impacted by elevated triglycerides. And as the authors point out in the discussion, it will be important to determine what regulates bmm expression such that bmm is higher in later stages of germline differentiation.

We thank the Reviewer for their positive assessment of our revised manuscript!

Reviewer #2 (Public Review):

Summary:

Here, the authors show that neutral lipids play a role in spermatogenesis. Neutral lipids are components of lipid droplets, which are known to maintain lipid homeostasis, and to be involved in non-gonadal differentiation, survival, and energy. Lipid droplets are present in the testis in mice and Drosophila, but not much is known about the role of lipid droplets during spermatogenesis. The authors show that lipid droplets are present in early differentiating germ cells, and absent in spermatocytes. They further show a cell autonomous role for the lipase brummer in regulating lipid droplets and, in turn, spermatogenesis in the Drosophila testis. The data presented show that a relationship between lipid metabolism and spermatogenesis is congruous in mammals and flies, supporting Drosophila spermatogenesis as an effective model to uncover the role lipid droplets play in the testis.

Strengths and weaknesses:

The authors do a commendably thorough characterization of where lipid droplets are detected in normal testes: located in young somatic cells, and early differentiating germ cells. They use multiple control backgrounds in their analysis, including w[1118], Canton S, and Oregon R, which adds rigor to their interpretations. The authors employ markers that identify which lipid droplets are in somatic cells, and which are in germ cells. The authors use these markers to present measured distances of somatic and germ cell-derived lipid droplets from the hub. Because they can also measure the distance of somatic and germ cells with age-specific markers from the hub, these results allow the authors to correlate position of lipid droplets with the age of cells in which they are present. This analysis is clearly shown and well quantified.

The quantification of lipid droplet distance from the hub is applied well in comparing brummer mutant testes to wild type controls. The authors measure the number of lipid droplets of specific diameters, and the spatial distribution of lipid droplets as a function of distance from the hub. These measurements quantitatively support their findings that lipid droplets are present in an expanded population of cells further from the hub in brummer mutants. The authors further quantify lipid droplets in germline clones of specified ages; the quantitative analysis here is displayed clearly and supports a cell autonomous role for brummer in regulating lipid droplets in spermatocytes.

Data examining testis size and number of spermatids in brummer mutants clearly indicates the importance of regulating lipid droplets to spermatogenesis. The authors show beautiful images supported by rigorous quantification supporting their findings that brummer mutants have both smaller testes with fewer spermatids at both 29 and 25C. There is also significant data supporting defects in testis size, but not spermatid number, in 14-day-old brummer mutant animals compared to controls. Their analysis clearly shows an expanded region beyond the testis apex that includes younger germ cells, supporting a role for lipid droplets influencing germ cell differentiation during spermatogenesis.

The authors present a series of data exploring a cell autonomous role for brummer in the germline, including clonal analysis and tissue specific manipulations. The clonal data indicating increased lipid droplets in spermatocyte clones, and a higher proportion of brummer mutant GSCs at the hub are convincing and supported by quantitation. The authors also show a tissue specific rescue of the brummer testis size phenotype by knocking down mdy specifically in germ cells, which is also supported by statistically significant quantitation. The authors present data examining the number of spermatocyte and post-meiotic clones 14 days after clonal induction. Their finding is significant with a p-value of 0.0496, which they acknowledge is less robust than their other data reported in this study, and could be a result of a low sample size. They indicate that future studies might validate these results with additional samples.

The authors do a beautiful job of validating where they detect brummer-GFP by presenting their own pseudotime analysis of publicly available single cell RNA sequencing data. Their data is presented very clearly, and supports expression of brummer in older somatic and germline cells of the age when lipid droplets are normally not detected. The authors also present a thorough lipidomic analysis of animals lacking brummer to identify triglycerides as an important lipid droplet component regulating spermatogenesis.

Impact:

The authors present data supporting the broad significance of their findings across phyla. This data represents a key strength of this manuscript. The authors show that loss of a conserved triglyceride lipase impacts testis development and spermatogenesis, and that these impacts can be rescued by supplementing diet with medium-chain triglycerides. The authors point out that these findings represent a biological similarity between Drosophila and mice, supporting the relevance of the Drosophila testis as a model for understanding the role of lipid droplets in spermatogenesis. The connection buttresses the relevance of these findings and this model to a broad scientific community.

We thank the Reviewer for their positive assessment of our revised paper!

RECOMMENDATIONS FOR THE AUTHORS

Reviewer #2 (Recommendations For The Authors):

The authors addressed most of my recommendations in a way that is satisfactory to me. I would like a bit more information added to the methods section about how hub area was quantified. For example, did the authors measure area within a defined region in a single Z plane (perhaps the Z plane at the center of the hub, or the Z plane with the largest area)? Alternatively, did they authors measure area in a more 3 dimensional way, i.e. volume. Adding this information to the methods would satisfy all of my previous recommendations.

We thank the Reviewer for pointing out that this information was not clear in the revised manuscript. We changed the methods section to clarify our methods as follows:

“The hub was identified as the FasIII-positive area of the testis. Hub size was estimated by measuring the FasIII-positive area in a Z-projected image of the hub in each testis. Zprojections were made using the ‘sum slices’ function in Fiji.”

eLife assessment

This important study identifies a role for triglycerides and lipid droplets in spermatogenesis, with data supporting relevance of this finding across phyla. The work shows with convincing data that a triglyceride lipase is required cell-autonomously for germline differentiation into meiotic stages and haploid spermatids and that an increase in triglycerides is detrimental to spermatogenesis. This paper would be of interest to developmental and cell biologists working on gametogenesis.

Reviewer #1 (Public Review):

In this study, the authors investigate the role of triglycerides in spermatogenesis. This work is based on their previous study (PMID: 31961851) on triglyceride sex differences in which they showed that somatic testicular cells play a role in whole body triglyceride homeostasis. In the current study, they show that lipid droplets (LDs) are significantly higher in the stem and progenitor cell (pre-meiotic) zone of the adult testis than in the meiotic spermatocyte stages. The distribution of LDs anti-correlates with the expression of the triglyceride lipase Brummer (Bmm), which has higher expression in spermatocytes than early germline stages. Analysis of a bmm mutant (bmm[1]) - a P-element insertion that is likely a hypomorphic - and its revertant (bmm[rev]) as a control shows that bmm acts autonomously in the germline to regulate LDs. In particular, the number of LDs is significantly higher in spermatocytes from bmm[1] mutants than from bmm[rev] controls. Testes from males with global loss of bmm (bmm[1]) are shorter than controls and have fewer differentiated spermatids. The zone of bam expression, typically close to the niche/hub in WT, is now many cell diameters away from the hub in bmm[1] mutants. There is an increase in the number of GSCs in bmm[1] homozygotes, but this phenotype is probably due to the enlarged hub. However, clonal analyses of GSCs lacking bmm indicate that a greater percentage of the GSC pool is composed of bmm[1]-mutant clones than of bmm[rev]-clones. This suggests that loss of bmm could impart a competitive advantage to GSCs, but this is not explored in greater detail. Despite the increase in number of GSCs that are bmm[1]-mutant clones, there is a significant reduction in the number of bmm[1]-mutant spermatocyte and post-meiotic clones. This suggests that fewer bmm[1]-mutant germ cells differentiate than controls. To gain insights into triglyceride homeostasis in the absence of bmm, they perform mass spec-based lipidomic profiling. Analyses of these data support their model that triglycerides are the class of lipid most affected by loss of bmm, supporting their model that excess triglycerides are the cause of spermatogenetic defects in bmm[1]. Consistent with their model, a double mutant of bmm[1] and a diacylglycerol O-acyltransferase 1 called midway (mdy) reverts the bmm-mutant germline phenotypes.

There are numerous strengths of this paper. First, the authors report rigorous measurements and statistical analyses throughout the study. Second, the authors utilize robust genetic analyses with loss-of-function mutants and lineage-specific knockdown. Third, they demonstrate the appropriate use of controls and markers. Fourth, they show rigorous lipidomic profiling. Lastly, their conclusions are appropriate for the results. In other words, they don't over-state the results. Overall, the rigorously quantified results support the major aim that appropriate regulation of triglycerides are needed in a germline cell-autonomous manner for spermatogenesis.

This paper should have a positive impact on the field. First and foremost, there is limited knowledge about the role of lipid metabolism in spermatogenesis. The lipidomic data will be useful to researchers in the field who study various lipid species. Going forward, it will be very interesting to determine what triglycerides regulate in germline biology. In other words, what functions/pathways/processes in germ cells are negatively impacted by elevated triglycerides. And as the authors point out in the discussion, it will be important to determine what regulates bmm expression such that bmm is higher in later stages of germline differentiation.

Reviewer #2 (Public Review):

Summary:

Here, the authors show that neutral lipids play a role in spermatogenesis. Neutral lipids are components of lipid droplets, which are known to maintain lipid homeostasis, and to be involved in non-gonadal differentiation, survival, and energy. Lipid droplets are present in the testis in mice and Drosophila, but not much is known about the role of lipid droplets during spermatogenesis. The authors show that lipid droplets are present in early differentiating germ cells, and absent in spermatocytes. They further show a cell autonomous role for the lipase brummer in regulating lipid droplets and, in turn, spermatogenesis in the Drosophila testis. The data presented show that a relationship between lipid metabolism and spermatogenesis is congruous in mammals and flies, supporting Drosophila spermatogenesis as an effective model to uncover the role lipid droplets play in the testis.

Strengths and weaknesses:

The authors do a commendably thorough characterization of where lipid droplets are detected in normal testes: located in young somatic cells, and early differentiating germ cells. They use multiple control backgrounds in their analysis, including w[1118], Canton S, and Oregon R, which adds rigor to their interpretations. The authors employ markers that identify which lipid droplets are in somatic cells, and which are in germ cells. The authors use these markers to present measured distances of somatic and germ cell-derived lipid droplets from the hub. Because they can also measure the distance of somatic and germ cells with age-specific markers from the hub, these results allow the authors to correlate position of lipid droplets with the age of cells in which they are present. This analysis is clearly shown and well quantified.

The quantification of lipid droplet distance from the hub is applied well in comparing brummer mutant testes to wild type controls. The authors measure the number of lipid droplets of specific diameters, and the spatial distribution of lipid droplets as a function of distance from the hub. These measurements quantitatively support their findings that lipid droplets are present in an expanded population of cells further from the hub in brummer mutants. The authors further quantify lipid droplets in germline clones of specified ages; the quantitative analysis here is displayed clearly and supports a cell autonomous role for brummer in regulating lipid droplets in spermatocytes.

Data examining testis size and number of spermatids in brummer mutants clearly indicates the importance of regulating lipid droplets to spermatogenesis. The authors show beautiful images supported by rigorous quantification supporting their findings that brummer mutants have both smaller testes with fewer spermatids at both 29 and 25C. There is also significant data supporting defects in testis size, but not spermatid number, in 14-day-old brummer mutant animals compared to controls. Their analysis clearly shows an expanded region beyond the testis apex that includes younger germ cells, supporting a role for lipid droplets influencing germ cell differentiation during spermatogenesis.

The authors present a series of data exploring a cell autonomous role for brummer in the germline, including clonal analysis and tissue specific manipulations. The clonal data indicating increased lipid droplets in spermatocyte clones, and a higher proportion of brummer mutant GSCs at the hub are convincing and supported by quantitation. The authors also show a tissue specific rescue of the brummer testis size phenotype by knocking down mdy specifically in germ cells, which is also supported by statistically significant quantitation. The authors present data examining the number of spermatocyte and post-meiotic clones 14 days after clonal induction. Their finding is significant with a p-value of 0.0496, which they acknowledge is less robust than their other data reported in this study, and could be a result of a low sample size. They indicate that future studies might validate these results with additional samples.

The authors do a beautiful job of validating where they detect brummer-GFP by presenting their own pseudotime analysis of publicly available single cell RNA sequencing data. Their data is presented very clearly, and supports expression of brummer in older somatic and germline cells of the age when lipid droplets are normally not detected. The authors also present a thorough lipidomic analysis of animals lacking brummer to identify triglycerides as an important lipid droplet component regulating spermatogenesis.

Impact:

The authors present data supporting the broad significance of their findings across phyla. This data represents a key strength of this manuscript. The authors show that loss of a conserved triglyceride lipase impacts testis development and spermatogenesis, and that these impacts can be rescued by supplementing diet with medium-chain triglycerides. The authors point out that these findings represent a biological similarity between Drosophila and mice, supporting the relevance of the Drosophila testis as a model for understanding the role of lipid droplets in spermatogenesis. The connection buttresses the relevance of these findings and this model to a broad scientific community.

Author Response

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

Reviewer #1 (Public Review):

In this manuscript, the authors demonstrated that YAP/TAZ promotes P-body formation in a series of cancer cell lines. YAP/TAZ modulates the transcription of multiple P-body-related genes, especially repressing the transcription of the tumor suppressor proline-rich nuclear receptor coactivator 1 (PNRC1) through cooperation with the NuRD complex. PNRC1 functions as a critical repressor in YAP-induced biogenesis of P-bodies and tumorigenesis in colorectal cancer (CRC). Reexpression of PNRC1 or disruption of P-bodies attenuated the protumorigenic effects of YAP. Overall, these findings are interesting and the study was well conducted.

We thank the reviewer for the positive comments for our work.

Major concerns:

(1) RNAseq data indicated that Yap has the capacity to suppress the expression of numerous genes. In addition to PNRC1, could there be additional Yap targeting factors involved in Yap-mediated the formation of P-bodies?

Yes, indeed. Additional YAP target genes, such as AJUBA, SAMD4A, are also involved in YAP-mediated the formation of P-bodies (Fig. 1B-D). Knockdown of either SMAD4A or AJUBA attenuated the P-body formation induced by overexpression of YAP5SA (Fig. 3A).

(2) It is still not clear how PNRC1 regulates P-bodies. Knockdown of PNRC1 prevented the reduction of P-bodies caused by Yap knockdown. How do the genes related to P-bodies that are positively regulated by Yap, such as SAMD4A, AJUBA, and WTIP, change in this scenario? Given that the expression of Yap can differ considerably among various cell types, is it possible for P-bodies to be present in tumor cells lacking Yap expression?

The detail mechanism of PNRC1’s suppressive effect on P-body formation was well explored in Gaviraghi et al.’s paper, in which PNRC1 was first identified as a tumor suppressor gene (EMBO, 2018, PMID: 30373810). Gaviraghi et al. revealed that overexpression of PNRC1 leads to translocation of cytoplasmic DCP1A/DCP2 into the nucleolus, which subsequently attenuates rRNA transcription and ribosome biogenesis. Since DCP1A and DCP2 are essential for formation of P-bodies, loss of cytoplasmic DCP1A/DCP2 also disrupts P-body formation. This background information has been included in the Results and Discussion sections in the manuscript:

Previously, we have performed the RNA-seq analysis of HCT116 cells with overexpression of PNRC1. Compared with YAP5SA overexpression (520 differentially expressed genes), overexpression of PNRC1 showed less effect on the gene expression profile (147 differentially expressed genes) and expression of SAMD4A, AJUBA and WTIP were not affected by PNRC1 overexpression.

In this study, we found that YAP could promote P-body formation in a series of cancer cell lines. During the exploration, we observed that P-bodies hardly existed in the RKO colorectal cancer cell line (Figure 1 for the reviewer). However, the regulatory effect of YAP/TAZ on SAMD4A, AJUBA, and WTIP was still observed (Figure 2 for the reviewer). These data suggest that YAP’s activity could be sufficient but not required for the P-body formation. So, we agree that P-bodies could be present in tumor cells lacking Yap expression.

Author response image 1.

Author response image 2.

(3) The authors demonstrated that CHD4 can bind to Yap target genes, such as CTGF, AJUBA, SAMD4A (Figure 4 - Figure Supplement 1D). Does the NuRD complex repress the expression of these genes? the NuRD complex could prevent the formation of P-bodies?

Good point! Following the reviewer’s suggestions, we detected the mRNA levels of AJUBA, WTIP and SAMD4A, and the P-body formation the CHD4 knockdown cells. Interestingly, knockdown of CHD4 induced mild downregulation of AJUBA, WTIP and SAMD4A in HCT116 cells (Figure 3 for the reviewer). Of note, NuRD complex is involved in both transcriptional repression and activation (PNAS 2011, PMID: 21490301; Stem Cell Reports. 2021, PMID: 33961790). As expected, knockdown CHD4 induced decreased number of P-bodies in HCT116 cell (new Figure 4-Supplement 1E), which is consistent to the enhanced expression of PNRC1 (Figure 4F).

Author response image 3.

Author response image 4.

(4) YAP/TAZ promotes the formation of P-bodies which contradicts the previous study's conclusion (PMID: 34516278). Please address these inconsistent findings.

The contradictory observations between our and the previous studies could be due to the different cell lines (HUVEC vs cancer cell lines) and different stimuli (KHSV infection vs normal culture condition or serum stimulation, cell density and stiffness). Actually, we have discussed the contradictory observation in the previous study in the Discussion section as followed:

“In contrast, a recent study, which provided the first link between YAP and P-bodies, implicated YAP as a negative regulator of P-bodies in KHSV-infected HUVECs (Castle et al, 2021). Elizabeth L. Castle et al. reported that virus-encoded Kaposin B (KapB) induces actin stress fiber formation and disassembly of P-bodies, which requires RhoA activity and the YAP transcriptional program (Castle et al, 2021). YAP-enhanced autophagic flux was proposed to participate in KapB-induced P-body disassembly, consistent with the concept that stress granules and P-bodies are cleared by autophagy (Buchan et al, 2013; Castle et al, 2021). However, an increasing number of studies have reported the contradictory role of YAP in autophagy regulation, which suggests that YAP-mediated autophagy regulation is cell type- and context-dependent (Jin et al, 2021; Pei et al, 2022; Totaro et al, 2019; Wang et al, 2020). Furthermore, though YAP is required for the cell proliferation in HUVEC, transformed cell lines often display elevated baseline YAP/TAZ activity compared to normal cells and possess many alterations in growth signaling pathways including autophagy signaling (Nguyen & Yi, 2019; Shen & Stanger, 2015; Zanconato et al, 2016). Thus, the contradictory observations regarding the role of YAP in modulating P-body formation between Elizabeth L. Castle et al.’s study and our study could be due to the different cell contexts and different cell conditions (baseline vs. KHSV infection).”

Reviewer #2 (Public Review):

In a study by Shen et al., the authors investigated YAP/TAZ target genes that play a role in the formation of processing bodies (P-bodies). P-bodies are membraneless cytoplasmic granules that contain translationally repressed mRNAs and components of mRNA turnover. GO enrichment analysis of the RNA-Seq data of colorectal cancer cells (HCT116) after YAP/TAZ knockdown showed that the downregulated genes were enriched in P-body resident proteins. Overexpression, knockdown, and ChIP-qPCR analyses showed that SAMD4A, PNRC1, AJUBA, and WTIP are YAP-TEAD target genes that also play a role in P-body biogenesis. Using P-body markers such as DDX6 and DCP1A, the authors showed that the knockdown of YAP in the HCT116 cell line causes a reduction in the number of P-bodies. Similarly, overexpression of constitutively active YAP (YAP 5SA) increased the P-body number. The YAP-TEAD target genes SAMD4A and AJUBA positively regulate P-body formation, because lowering their expression levels using siRNA reduces the number of P-bodies. The other YAP target gene, PNRC1, is a negative regulator of P-body biogenesis and consistently YAP suppresses its expression through the recruitment of the NuRD complex. YAP target genes that modulate P-body formation play prominent roles in oncogenesis. PNRC1 suppression is key to YAP-mediated proliferation, colony formation, and tumorigenesis in HCT116 xenografts. Similarly, SAMD4 and AJUBA knockdown abrogated cell viability. In summary, this study demonstrated that SAMD4, AJUBA, WTIP, and PNRC1 are bona fide YAP-TEAD target genes that play a role in P-body formation, which is also linked to the oncogenesis of colon cancer cells.

We thank the reviewer for the positive comments for our work.

Major Strengths:

The majority of the experiments were appropriately planned so that the generated data could support the conclusions drawn by the authors. The phenotype observed with YAP/TAZ knockdown correlated inversely with YAP5SA overexpression, which is complementary. Where possible, the authors also used point mutations that selectively disrupt protein-protein interactions, such as YAP S94A and PNRC1 W300A. The CRC cell line HCT116 was used throughout the study; additionally, data from other cancer cell lines were used to support the generality of the findings.

We thank the reviewer for the positive comments regarding the strength and significance of our work.

Weaknesses:

The authors did not elucidate the mechanistic link between P-body formation and oncogenesis; therefore, it is unclear why an increase in the number of P-bodies is pro-tumorigenic. AJUBA and SAMD4 may have housekeeping functions and reduce the proliferation of YAP-independent cell lines. Figure 6 - Figure Supplement 4 shows a reduction in cell viability and migration in control HCT116 cell lines upon AJUBA/SAMD4 knockdown. Therefore, it is unclear whether their tumor suppressive role is YAP-dependent. The authors extrapolated and suggested that their findings could be exploited therapeutically, without providing much detail. How do they plan to stimulate the expression of PNRC1? It is not necessary for every scientific finding to lead to a therapeutic benefit; therefore, they can tone down such statements if therapeutic exploitation is not realistic. The authors elucidated a mechanism for PNRC1 repression and one wonders why no attempts were made to understand the mechanism of activation of SAMD4, AJUBA, and WTIP expression.

We thank the reviewer for pointing out these issues to further improve the quality of our study. As mentioned in the Abstract section, the role of P-bodies in tumorigenesis and tumor progression is not well studied. In this study, we revealed that disruption of P-body formation by knockdown of essential P-body-related genes attenuates YAP-driven oncogenic function in CRC, which provides evidence implicating the pro-tumorigenic role of P-bodies. We agree with the reviewer that the mechanism of P-body formation promoting tumorigenesis is an important scientific question warranting exploration and plan to investigate this fancy question in next study.

AJUBA has been known to act as a signal transducer in oncogenesis and promote CRC cell survival (Pharmacol Res. 2020, PMID: 31740385; Oncogene. 2017, PMID: 27893714). Furthermore, as the reviewer suggested, we found that knockdown of both AJUBA and SAMD4A suppressed the cell proliferation in the YAP-deficient cell line, SHP-77, which further implicates the oncogenic role of AJUBA and SAMD4A (Figure 4 for the reviewer). Numerous studies have shown that YAP/TAZ knockdown suppressed the cell proliferation of HCT116 cells. Thus, not surprisingly, knockdown of AJUBA and SAMD4A also repressed the cell proliferation of the “parental” control HCT116 cells. Since the molecular mechanistic studies identified the AJUBA and SAMD4A were bona fide YAP-TEAD target genes, the co-dependencies of YAP and AJUBA/SAMD4A in the HCT116 cells imply that the pro-tumorigenic function of YAP could be dependent on activation of AJUBA/SAMD4A, in some extent (due to the large amount of YAP target genes).

Author response image 5.

Tumor suppressor genes are frequently epigenetically silenced in cancer cells, so is PNRC1. In our preliminary study, we found that the DNA methyltransferase inhibitor 5-Azacytidine dramatically increased the mRNA level of PNRC1 in HCT116 cells (Figure 5 for the reviewer), which suggests that PNRC1 is epigenetically suppressed by DNA methylation in CRC cells and could be re-activated or re-expressed by DNA methyltransferase inhibitor for the cancer treatment.

Author response image 6.

YAP/TAZ are well-known as transcriptional co-activators and the mechanism of transcriptional activation of target genes has been well-studied (Cell Stress. 2021, PMID: 34782888). However, years later, the function of YAP/TAZ as the transcriptional co-repressors was brought to the forefront. Both NuRD and Polycomb repressive complex 2 (PRC2) are involved in the transcriptional repressor function of YAP (Cell Rep. 2015, PMID: 25843714; Cancer Res. 2020, PMID: 32409309). Thus, we focused on exploring mechanism for PNRC1 repression in this study, but not the mechanism of activation of SAMD4A, AJUBA, and WTIP expression.

Reviewer #2 (Recommendations For The Authors):

Suggested experiments: The suggested experiments were aimed at minimizing the weaknesses of the manuscript. The roles of AJUBA and SAMD4 can be elucidated in a YAP-independent cell line. After knockdown of AJUBA or SAMD4 in a YAP-independent cell line, the effects on proliferation and migration should be determined.

Following the reviewer’s suggestions, we explored the role of AJUBA and SAMD4A in the YAP-independent cell line, SHP-77 (Cancer Cell. 2021, PMID: 34270926). Unfortunately, SHP-77 cells are suspension cells mixed with some loosely adherent cells, and we found that SHP-77 cells are not available for cell migration assay. By CCK8 assay, we found that knockdown of both AJUBA and SAMD4A suppressed the cell proliferation in SHP-77 cells, which further implicates the oncogenic role of AJUBA and SAMD4A.

Author response image 7.

Experiments directed at elucidating whether the mRNAs of tumor suppressor genes undergo sequestration and decay in P-bodies that ultimately promote tumorigenesis will provide a mechanistic link between P-body formation and tumorigenesis. The enrichment of P-bodies through biochemical methods has been employed in other studies. RNA-seq after P-body enrichment may provide opportunities to unravel the link between P-body formation and tumorigenesis.

We thank the reviewer for the constructive suggestions to further improve the significance of our study. We do have plans to purify the P-bodies to further elucidate underlying mechanisms of pro-tumorigenic role of P-bodies tumor cells. However, we are newcomers in the P-body field and encountered a lot of issues to establish the biochemical assays of P-bodies. Hopefully, we can solve these technical issues soon and present our new data in the next paper.

eLife assessment

This valuable study advances our understanding that YAP/TAZ, as well as their target genes, play a prominent role in the formation of processing bodies (P-bodies). The evidence supporting the conclusions is convincing. The article could be improved through further analysis to elucidate the mechanistic link between P-body formation and oncogenesis. The work will be of broad interest to scientists working in the field of Hippo signaling and cancer biology.

Reviewer #2 (Public Review):

In a study by Shen et al.. al., the authors investigated YAP/TAZ target genes that play a role in the formation of processing bodies (P-bodies). P-bodies are membraneless cytoplasmic granules that contain translationally repressed mRNAs and components of mRNA turnover. GO enrichment analysis of the RNA-Seq data of colorectal cancer cells (HCT116) after YAP/TAZ knockdown showed that the downregulated genes were enriched in P-body resident proteins. Overexpression, knockdown, and ChIP-qPCR analyses showed that SAMD4A, PNRC1, AJUBA, and WTIP are YAP-TEAD target genes that also play a role in P-body biogenesis. Using P-body markers such as DDX6 and DCP1A, the authors showed that knockdown of YAP in the HCT116 cell line causes a reduction in the number of P-bodies. Similarly, overexpression of constitutively active YAP (YAP 5SA) increased the P-body number. The YAP-TEAD target genes SAMD4A and AJUBA positively regulate P-body formation, because lowering their expression levels using siRNA reduces the number of P-bodies. The other YAP target gene, PNRC1, is a negative regulator of P-body biogenesis and consistently YAP suppresses its expression through the recruitment of the NuRD complex. YAP target genes that modulate P-body formation play prominent roles in oncogenesis. PNRC1 suppression is key to YAP-mediated proliferation, colony formation, and tumorigenesis in HCT116 xenografts. Similarly, SAMD4 and AJUBA knockdown abrogated cell viability. In summary, this study demonstrated that SAMD4, AJUBA, WTIP, and PNRC1 are bona fide YAP-TEAD target genes that play a role in P-body formation, which is also linked to the oncogenesis of colon cancer cells.

Major Strengths:

The majority of the experiments were appropriately planned so that the generated data could support the conclusions drawn by the authors. The phenotype observed with YAP/TAZ knockdown correlated inversely with YAP5SA overexpression, which is complementary. Where possible, the authors also used point mutations that selectively disrupt protein-protein interactions, such as YAP S94A and PNRC1 W300A. The CRC cell line HCT116 was used throughout the study; additionally, data from other cancer cell lines were used to support the generality of the findings.

Weaknesses:

The authors did not elucidate the mechanistic link between P-body formation and oncogenesis; therefore, it is unclear why an increase in the number of P-bodies is pro-tumorigenic. The authors extrapolated and suggested that PNRC1 expression could be exploited therapeutically, without providing much detail. How do they plan to stimulate the expression of PNRC1? It is not necessary for every scientific finding to lead to a therapeutic benefit; therefore, they can tone down such statements if therapeutic exploitation is not realistic. The authors elucidated a mechanism for PNRC1 repression and one wonders why no attempts were made to understand the mechanism of activation of SAMD4, AJUBA, and WTIP expression.

Author Response

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

eLife assessment

This important study combines psychophysics, fMRI, and TMS to reveal a causal role of FEF in generating an attention-induced ocular dominance shift, with potential relevance for clinical applications. The evidence supporting the claims of the authors is solid, but the theoretical and mechanistic interpretation of results and experimental approaches need to be strengthened. The work will be of broad interest to perceptual and cognitive neuroscience.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Based on a "dichoptic-background-movie" paradigm that modulates ocular dominance, the present study combines fMRI and TMS to examine the role of the frontoparietal attentional network in ocular dominance shifts. The authors claimed a causal role of FEF in generating the attention-induced ocular dominance shift.

Strengths:

A combination of fMRI, TMS, and "dichoptic-background-movie" paradigm techniques is used to reveal the causal role of the frontoparietal attentional network in ocular dominance shifts. The conclusions of this paper are mostly well supported by data.

Weaknesses:

(1) The relationship between eye dominance, eye-based attention shift, and cortical functions remains unclear and merits further delineation. The rationale of the experimental design related to the hemispheric asymmetry in the FEF and other regions should be clarified.

Thanks for the reviewer’s comments! We have further clarified the relationship between eye dominance shift, eye-based attention, and cortical functions in the Introduction and Discussion. In the Introduction, we introduce the modulating effects of eye-based attention on eye dominance. On one hand, eye-based attention can enhance eye dominance of the attended eye in real time (see page 3 first paragraph or below):

”For instance, presenting top-down attentional cues to one eye can intensify the competition strength of input signals in the attended eye during binocular rivalry (Choe & Kim, 2022; Zhang et al., 2012) and shift the eye balance towards the attended eye (Wong et al., 2021).”

On the other hand, prolonged eye-based attention can induce a shift of eye dominance to the unattended eye (see page 3 second paragraph or below):

“In Song et al. (2023)’s “dichoptic-backward-movie” adaptation paradigm (see Figure 1B), participants are presented with regular movie images in one eye (i.e., attended eye) while the other eye (i.e., unattended eye) received the backward movie images of the same episode. They were also instructed to try their best to follow the logic of the regular movie and ignore the superimposed backward movie. Therefore, the goal-directed eye-based attention was predominantly focused on the attended eye. Song et al. (2023) found that the predominance of the unattended eye in binocular rivalry increased after one hour of adaptation to the “dichoptic-backward-movie”, indicating a shift of perceptual ocular dominance towards the unattended eye. Since the overall energy of visual input from the two eyes was balanced throughout the adaptation period, the change of ocular dominance after adaptation is thought to result from unbalanced eye-based attention rather than unbalanced input energy as in typical short-term monocular deprivation (Bai et al., 2017; Lunghi et al., 2011; Zhou et al., 2014).”

Moreover, we discussed how FEF regulates attention-induced ocular dominance shift (see page 21 second paragraph to page 23 first paragraph or below, which also respond to this reviewer’s comment of Weakness #2):

“Then how does FEF regulate the attention-induced ocular dominance shift? Our previous work has found that the aftereffect (for simplicity, hereafter we use aftereffect to denote the attention-induced ocular dominance shift) can be produced only when the adapting stimuli involve adequate interocular competition, and is measurable only when the testing stimuli are not binocularly fused (Song et al., 2023). Given the indispensability of interocular competition, we explained those findings in the framework of the ocular-opponency-neuron model of binocular rivalry (Said & Heeger, 2013). The model suggests that there are some opponency neurons which receive excitatory inputs from monocular neurons for one eye and inhibitory inputs from monocular neurons for the other eye (e.g. AE-UAE opponency neurons receive excitatory inputs from the attended eye (AE) and inhibitory inputs from the unattended eye (UAE)). Then a difference signal is computed so that the opponency neurons fire if the excitatory inputs surpass the inhibitory inputs. Upon activation, the opponency neurons will in turn suppress the monocular neurons which send inhibitory signals to them.

Based on this model, we proposed an ocular-opponency-neuron adaptation account to explain the aftereffect, and pointed out that the attentional system likely modulated the AE-UAE ocular opponency neurons (Song et al., 2023). So why would FEF modulate the AE-UAE opponency neurons? The reason may be two fold. Firstly, understanding the logic during the dichoptic-backward-movie viewing may require filtering out the distracting information (from the unattended eye) and sustaining attention (to the attended eye), which is exactly the role of FEF (Esterman et al., 2015; Lega et al., 2019).

Secondly, due to the special characteristics of binocular vision system, filtering the distracting input from the unattended eye may have to rely on the interocular suppression mechanism. According to the ocular-opponency-neuron model, this is achieved by the firing of the AE-UAE opponency neurons that send inhibitory signals to the UAE monocular neurons.

As mentioned previously, the firing of the AE-UAE opponency neurons requires stronger activity for the AE monocular neurons than for the UAE monocular neurons. This is confirmed by the results shown in Figure 8 of Song et al. (2023) that monocular response for the attended eye during the entire adaptation phase was slightly stronger than that for the unattended eye. Accordingly, during adaptation the AE-UAE opponency neurons were able to activate for a longer period thus adapted to a larger extent than the UAE-AE opponency neurons. This would cause the monocular neurons for the unattended eye to receive less inhibition from the AE-UAE opponency neurons in the post-test as compared with the pre-test, leading to a shift of ocular dominance towards the unattended eye. In this vein, the magnitude of this aftereffect should be proportional to the extent of adaptation of the AE-UAE relative to UAE-AE opponency neurons. Attentional enhancement on the AE-UAE opponency neurons is believed to strengthen this aftereffect, as it has been found that attention can enhance adaptation (Dong et al., 2016; Rezec et al., 2004). Inhibition of FEF likely led such attentional modulation to be much less effective. Consequently, the AE-UAE opponency neurons might not have the chance to adapt to a sufficiently larger extent than the UAE-AE opponency neurons, leading to a statistically non-detectable aftereffect in Experiment 2. Therefore, the results of Experiments 2-4 in the present study suggest that within the context of the ocular-opponency-neuron adaptation account, FEF might be the core area to fulfill the attentional modulations on the AE-UAE opponency neurons.”

We used the experimental design with hemispheric asymmetry in the FEF and other regions for two reasons. First, many studies have shown that the dorsal attentional network has a functional right-hemisphere dominance (Duecker et al., 2013; Mayrhofer et al., 2019; Sack, 2010). This was also indicated by the results of Experiment 1 (Figure 3). Second, we found that a recent research applying TMS to FEF and IPS stimulated only the right hemisphere (Gallotto et al., 2022). Therefore, we selected the right FEF and right IPS as the target regions for cTBS. In the Methods section of Experiment 2, we have elucidated the reasons for the selection of cTBS target regions (see page 35, first paragraph or below):

“Given that the dorsal attentional network primarily consists of the FEF and the IPS (Corbetta & Shulman, 2002; Mayrhofer et al., 2019), with a functional right-hemisphere dominance (Duecker et al., 2013; Mayrhofer et al., 2019; Sack, 2010), we selected the right FEF and right IPS from the four clusters identified in Experiment 1 as the target regions for cTBS (Gallotto et al., 2022).”

(2) Theoretically, how the eye-related functions in this area could be achieved, and how it interacts with the ocular representation in V1 warrant further clarification.

Thanks for the reviewer’s comment! In the revised manuscript, we have discussed how FEF regulates attention-induced ocular dominance shift (see page 21 second paragraph to page 23 first paragraph or the quoted paragraphs under this reviewer’s first Public comment).

Reviewer #2 (Public Review):

Summary

Song et al investigate the role of the frontal eye field (FEF) and the intraparietal sulcus (IPS) in mediating the shift in ocular dominance (OD) observed after a period of dichoptic stimulation during which attention is selectively directed to one eye. This manipulation has been previously found to transiently shift OD in favor of the unattended eye, similar to the effect of short-term monocular deprivation. To this aim, the authors combine psychophysics, fMRI, and transcranial magnetic stimulation (TMS). In the first experiment, the authors determine the regions of interest (ROIs) based on the responses recorded by fMRI during either dichoptic or binocular stimulation, showing selective recruitment of the right FEF and IPS during the dichoptic condition, in line with the involvement of eye-based attention. In a second experiment, the authors investigate the causal role of these two ROIs in mediating the OD shift observed after a period of dichoptic stimulation by selectively inhibiting with TMS (using continuous theta burst stimulation, cTBS), before the adaptation period (50 min exposure to dichoptic stimulation). They show that, when cTBS is delivered on the FEF, but not the IPS or the vertex, the shift in OD induced by dichoptic stimulation is reduced, indicating a causal involvement of the FEF in mediating this form of short-term plasticity. A third control experiment rules out the possibility that TMS interferes with the OD task (binocular rivalry), rather than with the plasticity mechanisms. From this evidence, the authors conclude that the FEF is one of the areas mediating the OD shift induced by eye-selective attention.

Strengths

(1) The experimental paradigm is sound and the authors have thoroughly investigated the neural correlates of an interesting form of short-term visual plasticity combining different techniques in an intelligent way.

(2) The results are solid and the appropriate controls have been performed to exclude potential confounds.

(3) The results are very interesting, providing new evidence both about the neural correlates of eye-based attention and the involvement of extra-striate areas in mediating short-term OD plasticity in humans, with potential relevance for clinical applications (especially in the field of amblyopia).

Weaknesses

(1) Ethics: more details about the ethics need to be included in the manuscript. It is only mentioned for experiment 1 that participants "provided informed consent in accordance with the Declaration of Helsinki. This study was approved by the Institutional Review Board of the Institute of Psychology, Chinese Academy of Sciences". (Which version of the Declaration of Helsinki? The latest version requires the pre-registration of the study. The code of the approved protocol together with the code and date of the approval should be provided.) There is no mention of informed consent procedures or ethics approval for the TMS experiments. This is a huge concern, especially for brain stimulation experiments!

Response: Thanks for the reviewer’s comment! In the revised manuscript, we have provided the code of the approved protocol and date of the approval (see page 25 second paragraph or below):

“This study was approved (H21058, 11/01/2021) by the Institutional Review Board of the Institute of Psychology, Chinese Academy of Sciences.”

Indeed, ethics approval and informed consent were obtained for each experiment. To avoid duplication in the text, we only presented the ethics instructions in the Methods section of Experiment 1. We have now clarified in that section that all the experiments in this study were approved by the IRB in our Institute.

(2) Statistics: the methods section should include a sub-section describing in detail all the statistical analyses performed for the study. Moreover, in the results section, statistical details should be added to support the fMRI results. In the current version of the manuscript, the claims are not supported by statistical evidence.

Response: Thanks for the reviewer’s suggestion! In the Methods section of revised manuscript, we have added a section to describe the detailed statistical analyses for each experiment (see page 37 last paragraph for Experiment 2 and page 38 last paragraph for Experiment 3 or below):

“Statistical analyses were performed using MATLAB. A 3 (stimulation site: Vertex, FEF, IPS) × 2 (test phase: pre-test and post-test) repeated measures ANOVA was used to investigate the effect of cTBS delivery on ocular dominance shift. Moreover, for the blob detection test, the target detection rate of each experimental condition was calculated by dividing the summed number of detected blob targets by the total number of blob targets. Then, a 2 (eye: attended eye, unattended eye) × 3 (stimulation site: Vertex, FEF, IPS) repeated measures ANOVA on the detection performance was performed. Post-hoc tests were conducted using paired t-tests (2-tailed significance level at α = 0.05), and the resulting p-values were corrected for multiple comparisons using the false discovery rate (FDR) method (Benjamini & Hochberg, 1995).”

“In addition to the data analysis in Experiment 2, we complemented the standard inferential approach with the Bayes factor (van den Bergh et al., 2023; van Doorn et al., 2021; Wagenmakers et al., 2018), which allows quantifying the relative evidence that the data provide for the alternative (H1) or null hypothesis (H0). We conducted the Bayesian repeated measures ANOVA using JASP with default priors and computed inclusion Bayes factors (BFincl) which suggest the evidence for the inclusion of a particular effect calculated across matched models. A BF greater than 1 provides support for the alternative hypothesis. Specifically, a BF between 1 and 3 indicates weak evidence, a BF between 3 and 10 indicates moderate evidence, and a BF greater than 10 indicates strong evidence (van Doorn et al., 2021). In contrast, a BF below 1 provides evidence in favor of the null hypothesis.”

Furthermore, in the Results section of revised manuscript, we have added the statistical details to support the fMRI results (see page 9 last paragraph or below):

“To seek these brain regions, we used the AFNI program “3dttest++” to access the difference of ‘dichoptic-binocular’ contrast between the experimental and control runs. The AFNI program “ClustSim” was then applied for multiple comparison correction, yielding a minimum significant cluster size of 21 voxels (voxel wise p = .001; cluster threshold α = 0.05). We found 4 clusters showing stronger responses to the dichoptic movies than to the binocular movies especially in the experimental runs.”

(3) Interpretation of the results: the TMS results are very interesting and convincing regarding the involvement of the FEF in the build-up of the OD shift induced by dichoptic stimulation, however, I am not sure that the authors can claim that this effect is related to eye-based attention, as cTBS has no effect on the blob detection task during dichoptic stimulation. If the FEF were causally involved in eye-based attention, one would expect a change in performance in this task during dichoptic stimulation, perhaps a similar performance for the unattended and attended eye. The authors speculate that the sound could have an additional role in driving eye-based attention, which might explain the lack of effect for the blob discrimination task, however, this hypothesis has not been tested.

Response: Thanks for the reviewer’s comment! Following this reviewer’s insightful suggestion, we have conducted a new experiment to examine the effect of sound on blob detection task (see Experiment 4 in the revised manuscript). The procedure was similar to that of Experiment 2 except that the sound was no longer presented during the dichoptic-backward-movie adaptation. The results showed that the interocular difference of blob detection rate after sound elimination remained unaffected by the cTBS, which disagreed with our explanation in the previous version of manuscript. Based on the new data, we now question the validity to use the blob detection rate to precisely quantify eye-based attention, and have tried to explain why the blob detection results do not contradict with our account for the function role of FEF in modulating the aftereffect in the Discussion of the revised manuscript (see page 23 second paragraph to page 24 first paragraph or below):

“An unresolved issue is why inhibiting the cortical function of FEF did not impair the performance of blob detection task. One potential explanation is that the synchronized audio in Experiment 2 might help increase the length of time that the regular movie dominated awareness. However, the results of Experiment 4 did not support this explanation, in which the performance of blob detection survived from the inhibition of FEF even when silent movies were presented. Although this issue remains to be explored in future work, it does not contradict with our notion of FEF modulating AE-UAE opponency neurons. It should be noted that our notion merely states that FEF is the core area for attentional modulations on activities of AE-UAE opponency neurons. No other role of FEF during the adaptation is assumed here (e.g. boosting monocular responses or increasing conscious level of stimuli in the attended eye). In contrast, according to the most original definition, the blob detection performance serves as an estimation of visibility (or consciousness level) of the stimuli input from each eye, despite the initial goal of adopting this task is to precisely quantify eye-based attention (which might be impractical). Thus, according to our notion, inhibition of FEF does not necessarily lead to deteriorate performance of blob detection. Furthermore, our findings consistently indicated that the visibility of stimuli in the attended eye was markedly superior to that of stimuli in the unattended eye, yet the discrepancy in the SSVEP monocular responses between the two eyes was minimal though it had reached statistical significance (Song et al., 2023). Therefore, blob detection performance in our work may only faithfully reflect the conscious level in each monocular pathway, but it is probably not an appropriate index tightly associated with the attentional modulations on monocular responses in early visual areas. Indeed, previous work has argued that attention but not awareness modulates neural activities in V1 during interocular competition (Watanabe et al., 2011), but see (Yuval-Greenberg & Heeger, 2013). We have noticed and discussed the counterintuitive results of blob detection performance in our previous work (Song et al., 2023). Here, with the new counterintuitive finding that inhibition of FEF did not impair the performance of blob detection, we suspect that blob detection performance in the “dichoptic-backward-movie” adaptation paradigm may not be an ideal index that can be used to accurately quantify eye-based attention.

(4) Writing: in general, the manuscript is well written, but clarity should be improved in certain sections.

(a) fMRI results: the first sentence is difficult to understand at first read, but it is crucial to understand the results, please reformulate and clarify.

Response: Thanks for the reviewer’s suggestion! In the revised manuscript, we have reformulated this sentence (see page 9 last paragraph or below):

“It was only in the dichoptic condition of experimental runs that participants had to selectively pay more attention to one eye (i.e., eye-based attention). Therefore, we speculate that if certain brain regions exhibit greater activities in the dichoptic condition as compared to the binocular condition in the experimental runs but not in the control runs, the activation of these brain regions could be attributable to eye-based attention.”

(b) Experiment 3: the rationale for experiment one should be straightforward, without a long premise explaining why it would not be necessary.

Response: Thanks for the reviewer’s suggestion! In the revised manuscript, we have streamlined the lengthy premise explaining to make the rationale of Experiment 3 more straightforward (see page 15 last two paragraphs or below):

“The results of Experiment 2 support the notion that eye-based attention was the cause for attention-induced ocular dominance plasticity. However, an alternative account is that the significant two-way interaction between test phase and stimulation site did not stem from any persistent malfunction of FEF in modulating ocular dominance, but rather it was due to some abnormality of binocular rivalry measures in the post-test that occurred after stimulation at the FEF only (and not at the other two brain sites). For instance, stimulation at the FEF might simply reduce the ODI measured in the binocular rivalry post-test.

Therefore, we conducted Experiment 3 to examine how suppression of the three target sites would impact binocular rivalry performance, in case that any unknown confounding factors, which were unrelated to adaptation but related to binocular rivalry measures, contributed to the results.”

(c) Discussion: the language is a bit familiar here and there, a more straightforward style should be preferred (one example: p.19 second paragraph).

Response: Thanks for the reviewer’s suggestion! We have carefully revised the language in the discussion. The discussion following the example paragraph has been largely rewritten.

(5) Minor: the authors might consider using the term "participant" or "observer" instead of "subject" when referring to the volunteers who participated in the study.

Response: Thanks for the reviewer’s suggestion! In the revised manuscript, we have replaced the term “subject” with “participant”.

Reviewer #3 (Public Review):

Summary:

This study studied the neural mechanisms underlying the shift of ocular dominance induced by "dichoptic-backward-movie" adaptation. The study is self-consistent.

Strengths:

The experimental design is solid and progressive (relationship among three studies), and all of the raised research questions were well answered.

The logic behind the neural mechanisms is solid.

The findings regarding the cTMS (especially the position/site can be useful for future medical implications).

Weaknesses:

Why does the "dichoptic-backward-movie" adaptation matter? This part is severely missing. This kind of adaptation is neither intuitive like the classical (Gbison) visual adaptation, nor practical as adaptation as a research paradigm as well as the fundamental neural mechanism. If this part is not clearly stated and discussed, this study is just self-consistent in terms of its own research question. There are tons of "cool" phenomena in which the neural mechanisms are apparent as "FEF controls vision-attention" but never tested using TMS & fMRI, but we all know that this kind of research is just of incremental implications.

Response: Thanks for the reviewer’s comment! We designed the "dichoptic-backward-movie" adaptation to study the perceptual consequence and mechanisms of sustained attention to a monocular pathway. Since the overall visual input to both eyes during adaptation were identical, any effect (i.e. the change of ocular dominance in our study) after adaptation can be easily ascribed to unbalanced eye-based attention between the two eyes rather than unbalanced input energy across the eyes. In typical short-term monocular deprivation, input signal from one eye is blocked. Accordingly, attention is undoubtedly distributed to the non-deprived eye. The fact that in a short-term monocular deprivation paradigm the deprived eye is also the unattended eye prevents researchers from ascertaining whether unbalanced eye-based attentional allocation contributes to the shift of ocular dominance just like unbalanced visual input across the two eyes. That is why the “dichoptic-backward-movie” adaptation was adopted in the present study. This new paradigm balances the input energy across the eyes but leaves attention unbalanced across the eyes. In the revised manuscript, we have added the description of the “dichoptic-backward-movie” adaptation (see page 3 last paragraph and page 4 first paragraph or below). Hope this complementary information improves the clarity.

“In Song et al. (2023)’s “dichoptic-backward-movie” adaptation paradigm (see Figure 1B), participants are presented with regular movie images in one eye (i.e., attended eye) while the other eye (i.e., unattended eye) received the backward movie images of the same episode. They were also instructed to try their best to follow the logic of the regular movie and ignore the superimposed backward movie. Therefore, the goal-directed eye-based attention was predominantly focused on the attended eye. Song et al. (2023) found that the predominance of the unattended eye in binocular rivalry increased after one hour of adaptation to the “dichoptic-backward-movie”, indicating a shift of perceptual ocular dominance towards the unattended eye. Since the overall energy of visual input from the two eyes was balanced throughout the adaptation period, the change of ocular dominance after adaptation is thought to result from unbalanced eye-based attention rather than unbalanced input energy as in typical short-term monocular deprivation (Bai et al., 2017; Lunghi et al., 2011; Zhou et al., 2014).” In short-term monocular deprivation, input signal from one eye is blocked. Accordingly, attention is biased towards the non-deprived eye. However, it is difficult to tease apart the potential contribution of unbalanced eye-based attention from the consequence of the unbalanced input energy, as the deprived eye is also the unattended eye. Therefore, the advantage of the “dichoptic-backward-movie” adaptation paradigm is to balance the input energy across the eyes but leave attention unbalanced across the eyes.

Our previous work (Song et al., 2023) has shown that eye-based attention plays a role in the formation of ocular dominance shift following adaptation to dichoptic backward movie. However, because the “dichoptic-backward-movie” adaptation paradigm is new, to our knowledge, no literature has ever discovered the brain areas that are responsible for eye-based attention. Our fMRI experiment for the first time resolves this issue, which, we believe, is one of the novelties of the present study. Attention is a pretty general definition of our ability to select limited information for preferential or privileged processing, yet it includes numerous aspects (e.g. spatial attention for spatial locations, feature-based attention for visual features, object-based attention for objects, social attention for social cues, and eye-based attention for monocular pathways etc). Are we 100% sure that the same brain network always underlies every aspect of attention including eye-based attention? No test, no answer. Maybe the answer is Yes, but we are not aware of any evidence for that from literature. It is not unlikely that attention is like an elephant while researchers are like blind people touching the elephant from different angles. Even if all previous researchers have touched the side of the elephant and state that an elephant is no different from a wall, as long as one researcher grabs the elephant’s tail, the “wall” knowledge will be falsified. From this perspective of the essence of science (falsifiable), we have the confidence to say that our fMRI experiment on eye-based attention is novel, because to our knowledge our experiment is the first one to explore the issue. On the basis of the fMRI experiment (otherwise we would have no idea on which precise brain site to apply the cTBS), we could successfully complete the subsequent TMS experiments.

Of course, if the reviewer can kindly point out any previous neuroimaging work we missed that has already disclosed the neural mechanisms underlying human’s eye-based attention, we would truly appreciate the reviewer very much. But even so, we would like to emphasize that the purpose of the current study was actually not to use TMS & fMRI to confirm that “FEF controls visual attention”. As we mentioned in the Abstract and expanded the introduction in the last two paragraphs of Introduction, the goal of the TMS experiments is to examine the causal role of eye-based attention in producing the aftereffect of “dichoptic-backward-movie” adaptation. This research question is also new, thus we do not think the TMS experiments are incremental, either. Our findings provided direct causal evidence for the effect of FEF on modulating ocular dominance through eye-based attention. Please see the last two sentences in the first paragraph on page 20 in the revised manuscript or below,

“Interestingly, in our Experiment 2 this aftereffect was significantly attenuated after we temporarily inhibited the cortical function of FEF via cTBS. This finding indicates the crucial role of FEF in the formation of attention-induced ocular dominance shift.”

as well as the last sentence of the Abstract,

“…and in this network, FEF plays a crucial causal role in generating the attention-induced ocular dominance shift.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) The hemispheric asymmetry in the eye-based attention-related cortex should be further examined and discussed. For example, IPS in both hemispheres was identified in the fMRI experiment. It is not clear why only the right IPS was stimulated in the TMS experiment.

Response: Thanks for the comment. We have elucidated the reasons for the experimental design with hemispheric asymmetry in FEF and IPS. Please see our response to the Weakness #1 raised by Reviewer #1 in the Public Review section.

(2) It is known that the frontoparietal cortex plays a role in the contralateral shift of attentional allocation. Meanwhile, the latest stage of ocular-specific representation is V1. The authors should discuss how the eye-related function can be achieved in FEF.

Response: Thanks for the comment. we have discussed how FEF regulates attention-induced ocular dominance shift (see page 21 second paragraph to page 23 first paragraph in the revised manuscript, and our response to the Weakness #2 raised by Reviewer #1 in the Public Review section).

(3) To further validate the role of FEF in eye-related attention shifts, the authors may consider using the traditional monocular deprivation paradigm with fMRI and TMS. It would be valuable to compare the neural mechanisms related to the classical monocular deprivation paradigm with the current findings.

Response: Thanks for the reviewer’s suggestion! That is indeed an interesting research topic that we are currently exploring. The current study investigated the attention-induced ocular dominance shift with the “dichoptic-backward-movie-adaptation” paradigm. This paradigm is substantially different from traditional short-term monocular deprivation. In our Neuroscience Bulletin paper (Song et al. 2023), we discuss the reason as follows.

“An alternative account of our results is the homeostatic plasticity mechanism. The function of this mechanism is to stabilize neuronal activity and prevent the neuronal system from becoming hyperactive or hypoactive. For this goal, the mechanism moves the neuronal system back toward its baseline after a perturbation [51, 52]. In our case, the aftereffect can be explained such that the visual system boosts the signals from the unattended eye to maintain the balance of the network’s excitability. However, this account cannot easily explain why the change of neural ocular dominance led by prolonged eye-based attention was observed here using the binocular rivalry testing stimuli, but absent in the previous research using the binocularly fused stimuli [11]. In contrast, a recent SSVEP study also using the binocularly fused stimuli has successfully revealed a shift of neural ocular dominance after two hours of monocular deprivation [31], which is in line with the homeostatic plasticity account. Therefore, the mechanisms underlying the “dichoptic-backward-movie” adaptation and monocular deprivation are probably not fully overlapped with each other; and the binocular rivalry mechanism described in the ocular-opponency-neuron model seems to be more preferable than the homeostatic plasticity mechanism in accounting for the present findings.”

Therefore, before asking whether FEF plays a role in the attention-induced ocular dominance shift in a traditional monocular deprivation paradigm, one should probably first examine whether attention also plays a role in traditional monocular deprivation, and whether the ocular-opponency-neuron adaptation account can also be used to explain the traditional monocular deprivation effect. Our newly accepted paper “Negligible contribution of adaptation of ocular opponency neurons to the effect of short-term monocular deprivation” (https://www.frontiersin.org/articles/10.3389/fpsyg.2023.1282113/full) gives a generally negative answer to the second question. And as to the first question, we have one manuscript under review and another ongoing study. In other words, to get a satisfactory answer to this particular comment of this reviewer, we need to first obtain clear answers to the two above questions. We think this is far beyond the scope of one single manuscript.

(4) The authors only presented regular movies to the dominant eye to maximize the ocular dominance shift. This critical information of design should be clarified, not only in the method section.

Response: Thanks for the reviewer’s suggestion! In the Results section of Experiment 2, we have added a description of this critical information of design (see page 11 last paragraph to page 12 first paragraph or below):

“Then, participants adapted to the “dichoptic-backward-movie” in which regular movie images were presented to the dominant eye to maximize the effect of eye dominance shift (Song et al., 2023). Meanwhile they were asked to detect some infrequent blob targets presented on the movie images in one eye at the same time.”

(5) The frame rate of the movie is 30 fps, which is much lower than a typical 60 fps visual presentation, does this have an effect on the adaptation outcome?

Response: To our best of knowledge, there is no evidence that the frame rate of the movie influences the aftereffect of attention-induced ocular dominance shift. In our previous research, the frame rate of the movie during adaptation was 25 fps, which still produced a stable adaptation aftereffect (Song et al., 2023). And the frame rate of the movie was 30 fps in our monocular deprivation work (Lyu et al., 2020), which showed a similar monocular deprivation effect we previously observed in an altered reality study (Bai et al., 2017). The frame rate of the altered-reality video in Bai et al.’s (2017) work was 60 fps. All these clues suggest that the frame rate does not have an effect on the adaptation outcome.

(6) Figure 5: The ODSE derived from ODI in Experiment 3 should also be illustrated, for a better comparison with results from Experiment 2.

Response: Thanks for the reviewer’s suggestion! In the revised manuscript, we have added the results of ODSE in Experiment 3 to Figure 5 (see page 15 or below):

Author response image 1.

Figure 5. The results of (A) the ocular dominance index (ODI), (B) the ocular dominance shift effects (ODSE) in Experiment 2, (C) the ODI and (D) the ODSE in Experiment 3. The bars show the grand average data for each condition. The individual data are plotted with gray lines or dots. The dashed gray line represents the absolute balance point for the two eyes (ODI = 0.5). Error bars indicate standard errors of means. * p < .05; ** p < .01; n.s. p > .05.

(7) Spelling issues: "i.e." → "i.e.,"

Response: Thanks for the reviewer’s suggestion! In the revised manuscript, we have changed “i.e.” to “i.e.,”.

Reviewer #2 (Recommendations For The Authors):

Linked to weakness 3: Ideally, a control experiment with cTBS and dichoptic stimulation without sound but with the blob discrimination task should be performed to be able to make important claims about the neural mechanisms involved in eye-based attention.

Response: Thanks for the comment. We have performed a new experiment as the reviewer suggested. Please see our response to the Weakness #3 raised by Reviewer #2 in the Public Review section.

Reviewer #3 (Recommendations For The Authors):

(1) The neural mechanisms are so apparent. We all know the FEF\IPS\SC matter in vision and attention and gaze. This is not groundbreaking.

Response: As we addressed in our response to Reviewer #3’s public comment, the current study aimed at investigating the causal mechanism for eye-based attentional modulation of ocular dominance plasticity rather than simply the role of FEF\IPS\SC in visual attention. Moreover, eye-based attention is a less investigated aspect of visual attention. The neural mechanism underlying eye-based attention is still largely unknown, and seeking the brain areas for controlling eye-based attention is the necessary preparation work for applying the cTBS. We have responded in detail to Reviewer #3’s public comment why we think both the fMRI and TMS experiments are novel to the field, which we will not reiterate it here to avoid redundancy.

(2) Why does the "dichoptic-backward-movie" adaptation matter? Is playing a backward movie to one eye realistic? Does that follow the efficient coding? Is that a mere consequence of information theory?

Response: Thanks for the comments. We have added the description of the “dichoptic-backward-movie” adaptation paradigm in the revised manuscript (see page 3 last paragraph and page 4 first paragraph or our response to this reviewer’s Public comment).

Is it realistic to play backward movie to one eye? We feel this question is somehow ambiguous to us. If the reviewer means the technical operability for such stimulus presentation, we can assure it since we have used this paradigm in both the current and previously published studies. To be more specific, we made the video stimuli in advance. The left half of the video was the regular movie and the right half was the backward version of the same movie (or vice versa). When viewing such video stimuli through stereoscopes, participants could only see the left half of the video with the left eye and the right half of the video with the right eye. In other words, the regular movie and backward movie were viewed dichoptically. Alternatively, if the reviewer means that such dichoptic presentation rarely happens in real world thus not realistic, we agree with the reviewer on one hand. On the other hand, we have explained on page 3 last paragraph and page 4 first paragraph why it is a particular useful paradigm for the main purpose of the present study. Let us make a similar example. The phenomenon of binocular rivalry rarely happens in everyday life. So people may say binocular rivalry is not realistic. However, our visual system does have the ability to deal with such conflicting visual inputs across the eyes, even binocular rivalry is unrealistic! Sometimes it is fun to investigate those seemingly unrealistic functions of our brains since those may also reveal the mystery of our neural system. As we know, despite binocular rivalry is uncommon in daily life, it is frequently used to investigate awareness. And in our work, we use binocular rivalry to measure perceptual ocular dominance.

Finally, the reviewer queried about if the "dichoptic-backward-movie" adaptation paradigm follow efficient coding and information theory. The information theory and efficient coding assume that messages with low expectedness or of rare occurrence would attract more attention and induce larger neural responses than those with high expectedness. In the "dichoptic-backward-movie" adaptation paradigm, the backward movie should be less expected since the actions of the characters in the backward movie appeared illogical. Thus, according to the information theory and efficient coding, it would be expected that more attention was paid to the backward movie and thus the backward movie might dominate the awareness for a longer period during adaptation (Zhang et al., 2012). However, we instructed participants to follow the regular movie during adaptation. The results of blob detection task also showed a better task performance when the targets appeared in the eye presented with the regular movie, which contradicted with the prediction of the information theory and efficient coding. Thus, it seems not very likely that the "dichoptic-backward-movie" adaptation followed efficient coding and information theory.

References

Bai, J., Dong, X., He, S., & Bao, M. (2017). Monocular deprivation of Fourier phase information boosts the deprived eye’s dominance during interocular competition but not interocular phase combination. Neuroscience, 352, 122-130. https://doi.org/10.1016/j.neuroscience.2017.03.053

Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal statistical society: series B (Methodological), 57(1), 289-300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x

Choe, E., & Kim, M.-S. (2022). Eye-specific attentional bias driven by selection history. Psychonomic Bulletin & Review, 29(6), 2155-2166. https://doi.org/10.3758/s13423-022-02121-0

Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. Nature reviews neuroscience, 3(3), 201-215. https://doi.org/10.1038/nrn755

Dong, X., Gao, Y., Lv, L., & Bao, M. (2016). Habituation of visual adaptation. Sci Rep, 6, 19152. https://doi.org/10.1038/srep19152

Duecker, F., Formisano, E., & Sack, A. T. (2013). Hemispheric differences in the voluntary control of spatial attention: direct evidence for a right-hemispheric dominance within frontal cortex. Journal of Cognitive Neuroscience, 25(8), 1332-1342. https://doi.org/10.1162/jocn_a_00402

Esterman, M., Liu, G., Okabe, H., Reagan, A., Thai, M., & DeGutis, J. (2015). Frontal eye field involvement in sustaining visual attention: evidence from transcranial magnetic stimulation. Neuroimage, 111, 542-548. https://doi.org/10.1016/j.neuroimage.2015.01.044

Gallotto, S., Schuhmann, T., Duecker, F., Middag-van Spanje, M., de Graaf, T. A., & Sack, A. T. (2022). Concurrent frontal and parietal network TMS for modulating attention. iScience, 25(3), 103962. https://doi.org/10.1016/j.isci.2022.103962

Lega, C., Ferrante, O., Marini, F., Santandrea, E., Cattaneo, L., & Chelazzi, L. (2019). Probing the neural mechanisms for distractor filtering and their history-contingent modulation by means of TMS. Journal of Neuroscience, 39(38), 7591-7603. https://doi.org/10.1523/JNEUROSCI.2740-18.2019

Lunghi, C., Burr, D. C., & Morrone, C. (2011). Brief periods of monocular deprivation disrupt ocular balance in human adult visual cortex. Curr Biol, 21(14), R538-539. https://doi.org/10.1016/j.cub.2011.06.004

Lyu, L., He, S., Jiang, Y., Engel, S. A., & Bao, M. (2020). Natural-scene-based Steady-state Visual Evoked Potentials Reveal Effects of Short-term Monocular Deprivation. Neuroscience, 435, 10-21. https://doi.org/10.1016/j.neuroscience.2020.03.039

Mayrhofer, H. C., Duecker, F., van de Ven, V., Jacobs, H. I., & Sack, A. T. (2019). Hemifield-specific correlations between cue-related blood oxygen level dependent activity in bilateral nodes of the dorsal attention network and attentional benefits in a spatial orienting paradigm. Journal of Cognitive Neuroscience, 31(5), 625-638. https://doi.org/10.1162/jocn_a_01338

Rezec, A., Krekelberg, B., & Dobkins, K. R. (2004). Attention enhances adaptability: evidence from motion adaptation experiments. Vision Res, 44(26), 3035-3044. https://doi.org/10.1016/j.visres.2004.07.020

Sack, A. T. (2010). Using non-invasive brain interference as a tool for mimicking spatial neglect in healthy volunteers. Restorative neurology and neuroscience, 28(4), 485-497. https://doi.org/10.3233/RNN-2010-0568

Said, C. P., & Heeger, D. J. (2013). A model of binocular rivalry and cross-orientation suppression. PLoS computational biology, 9(3), e1002991. https://doi.org/10.1371/journal.pcbi.1002991

Song, F., Lyu, L., Zhao, J., & Bao, M. (2023). The role of eye-specific attention in ocular dominance plasticity. Cerebral Cortex, 33(4), 983-996. https://doi.org/10.1093/cercor/bhac116

van den Bergh, D., Wagenmakers, E.-J., & Aust, F. (2023). Bayesian Repeated-Measures Analysis of Variance: An Updated Methodology Implemented in JASP. Advances in Methods and Practices in Psychological Science, 6(2), 25152459231168024. https://doi.org/10.1177/25152459231168024

van Doorn, J., van den Bergh, D., Böhm, U., Dablander, F., Derks, K., Draws, T., Etz, A., Evans, N. J., Gronau, Q. F., Haaf, J. M., Hinne, M., Kucharský, Š., Ly, A., Marsman, M., Matzke, D., Gupta, A., Sarafoglou, A., Stefan, A., Voelkel, J. G., & Wagenmakers, E. J. (2021). The JASP guidelines for conducting and reporting a Bayesian analysis. Psychonomic Bulletin & Review, 28(3), 813–826. https://doi.org/10.3758/s13423-020-01798-5

Wagenmakers, E. J., Love, J., Marsman, M., Jamil, T., Ly, A., Verhagen, J., Selker, R., Gronau, Q. F., Dropmann, D., Boutin, B., Meerhoff, F., Knight, P., Raj, A., van Kesteren, E. J., van Doorn, J., Šmíra, M., Epskamp, S., Etz, A., Matzke, D., de Jong, T., van den Bergh, D., Sarafoglou, A., Steingroever, H., Derks, K., Rouder, J. N., & Morey, R. D. (2018). Bayesian inference for psychology. Part II: Example applications with JASP. Psychonomic Bulletin & Review, 25(1), 58–76. https://doi.org/10.3758/s13423-017-1323-7

Watanabe, M., Cheng, K., Murayama, Y., Ueno, K., Asamizuya, T., Tanaka, K., & Logothetis, N. (2011). Attention but not awareness modulates the BOLD signal in the human V1 during binocular suppression. Science, 334(6057), 829-831. https://doi.org/10.1126/science.1203161

Wong, S. P., Baldwin, A. S., Hess, R. F., & Mullen, K. T. (2021). Shifting eye balance using monocularly directed attention in normal vision. J Vis, 21(5), 4. https://doi.org/10.1167/jov.21.5.4

Yuval-Greenberg, S., & Heeger, D. J. (2013). Continuous flash suppression modulates cortical activity in early visual cortex. J Neurosci, 33(23), 9635-9643. https://doi.org/10.1523/jneurosci.4612-12.2013

Zhang, P., Jiang, Y., & He, S. (2012). Voluntary attention modulates processing of eye-specific visual information. Psychol Sci, 23(3), 254-260. https://doi.org/10.1177/0956797611424289

Zhou, J., Reynaud, A., & Hess, R. F. (2014). Real-time modulation of perceptual eye dominance in humans. Proc Biol Sci, 281(1795). https://doi.org/10.1098/rspb.2014.1717

Reviewer #1 (Public Review):

Summary:

Based on a "dichoptic-background-movie" paradigm that modulates ocular dominance, the present study combines fMRI and TMS to examine the role of the frontoparietal attentional network in ocular dominance shifts. The authors claimed a causal role of FEF in generating the attention-induced ocular dominance shift.

Strengths:

A combination of fMRI, TMS, and "dichoptic-background-movie" paradigm techniques is used to reveal the causal role of the frontoparietal attentional network in ocular dominance shifts. The conclusions of this paper are well supported by data.

Weaknesses:

My previous concerns have been addressed.

eLife assessment

This important study combines psychophysics, fMRI, and TMS to reveal a causal role of FEF in generating an attention-induced ocular dominance shift, with potential relevance for clinical applications. The evidence supporting the claims of the authors is convincing. The work will be of broad interest to perceptual and cognitive neuroscience.

Reviewer #2 (Public Review):

Summary

Song et al investigate the role of the frontal eye field (FEF) and the intraparietal sulcus (IPS) in mediating the shift in ocular dominance (OD) observed after a period of dichoptic stimulation during which attention is selectively directed to one eye. This manipulation has been previously found to transiently shift OD in favor of the unattended eye, similar to the effect of short-term monocular deprivation. To this aim, the authors combine psychophysics, fMRI, and transcranial magnetic stimulation (TMS). In the first experiment, the authors determine the regions of interest (ROIs) based on the responses recorded by fMRI during either dichoptic or binocular stimulation, showing selective recruitment of the right FEF and IPS during the dichoptic condition, in line with the involvement of eye-based attention. In a second experiment, the authors investigate the causal role of these two ROIs in mediating the OD shift observed after a period of dichoptic stimulation by selectively inhibiting with TMS (using continuous theta burst stimulation, cTBS), before the adaptation period (50 min exposure to dichoptic stimulation). They show that, when cTBS is delivered on the FEF, but not the IPS or the vertex, the shift in OD induced by dichoptic stimulation is reduced, indicating a causal involvement of the FEF in mediating this form of short-term plasticity. A third control experiment rules out the possibility that TMS interferes with the OD task (binocular rivalry), rather than with the plasticity mechanisms. From this evidence, the authors conclude that the FEF is one of the areas mediating the OD shift induced by eye-selective attention.

The authors have addressed the issues that I raised during the first round of review.<br /> While the results of the new experiment (Experiment 4), leave some unresolved isssues (addressed in the discussion section), they provide a very important replication of the main result, showing that even if the observed effect is small, it is robust.

Reviewer #3 (Public Review):

Summary:

This study studied the neural mechanisms underlying the shift of ocular dominance induced by "dichoptic-backward-movie" adaptation. The study is self-consistent.

Strengths:

The experimental design is solid and progressive (relationship among three studies), and all of the raised research questions were well answered.<br /> The logic behind the neural mechanisms is solid.<br /> The findings regarding the cTMS (especially the position/site can be useful for future medical implications).<br /> The updated Exp4 eliminates some concerns and thus makes the results even more solid.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

In chicken embryos, the counter-rotating migration of epiblast cells on both sides of the forming primitive streak (PS), a process referred to as polonaise movements, has attracted longstanding interest as a paradigm of morphogenetic cell movements. However, the association between these cell movements and PS development is still controversial. This study investigated PS development and polonaise movements separately at their initial stage, showing that both could be uncoupled (at least at the initial phase), being activated via Vg1 signaling.

Strengths of this study

Polonaise movements, i.e., the circular cell migration of epiblast cells on both sides of the forming PS in avian embryos, have been the subject of research through live imaging and promoted the development of new tools to analyze quantitatively such movements. However, conclusions from previous studies remain controversial, at least partly due to the nature of perturbations to PS development and polonaise movements.

This study performed the challenging technique of electroporation to successfully mark and manipulate Wnt/PCP pathways in unincubated chicken embryo cells at the initiation phase of these two processes. In addition, the authors separately altered PS development and polonaise movements: PS development was perturbed by inhibiting either the Wnt/PCP pathway or DNA synthesis using aphidicolin, while polonaise movements were modified by the development of a second PS after engrafting Vg1-expressing COS cells located at the opposite end of the blastoderm. The study concluded that Vg1 elicits both PS development and polonaise movements, which occur in a parallel and are not inter-dependent.

To support these conclusions, particle image velocimetry (PIV) of cell trajectories captured by live imaging was performed. These tools delineated visually appealing cell movements and gave rise to vorticity profiles, adding more value to this study.

Weaknesses of this study

Engrafted Vg1-expressing COS cells located at the anterior end of the blastoderm elicited both the development of a second PS and marked bilateral polonaise movements while perturbing these movements along the original PS. How do polonaise movements along the second PS dominate over those along the normal PS? The authors suggested a model in which Vg1 acts in a graded or dose-dependent manner since engrafted COS cells over-expressed Vg1. This model can be tested by reducing the mass of engrafted COS cells. Although the authors propose performing this analysis in further investigations, it would be preferable to incorporate into this study for better consistency.

We would like to express our gratitude to the editors and the reviewers for finding the valuable significances of our study and for giving thoughtful suggestions. We agree that it would be a logical next step to identify the driving mechanism(s) of the polonaise movements, although this is beyond the scope of the current study. Rather, it is the focus of ongoing studies, in which we are investigating how Vg1 works in this concentration context and resulting dose-dependent effect on downstream gene expression, in order to provide a comprehensive understanding of this interesting dual role of Vg1. The relationship between the intensity of Vg1 signaling and the polonaise movements can be tested by modifying the size of the Vg1/COS, as the reviewer pointed out.

The authors claim that chicken embryo development is representative of "amniotes," but it does not hold for all groups. Avian and mammal species are exceptional among amniotes in the sense they develop a PS (e.g., Coolen et al. 2008). Moreover, in certain mammalian embryos like mouse embryos, cells laterally to the PS do not move much (Williams et al. 2012). The authors should avoid the generalization that chicken embryos unequivocally represent amniotes as opposed to the observed in non-amniote embryos. The observations in chicken embryos as they stand are significant enough.

References:

Coolen M, et al. (2008). Molecular characterization of the gastrula in the turtle Emys orbicularis: an evolutionary perspective on gastrulation. PLoS One. 3(7):e2676. doi: 10.1371/journal.pone.0002676

Williams M, et al. (2012). Mouse primitive streak forms in situ by initiation of epithelial to mesenchymal transition without migration of a cell population. Dev Dyn. 241(2):270-283. doi: 10.1002/dvdy.23711

We modified the following sentences to the summary and introduction of the revised version as below:

In Summary:

(p.1, Lines 9-11.) “Large-scale cell flow characterizes gastrulation in animal development. In amniote gastrulation, particularly in avian gastrula, a bilateral vortex-like counter-rotating cell flow, called ‘polonaise movements’, appears along the midline.”

In Introduction:

(p.2, Lines 43-46.) “In amniotes, particularly in avian gastrula (i.e. embryonic disc), a bilateral vortex-like counter-rotating cell flow, termed ‘polonaise movements’, occurs within the epiblast along the midline axis, prior to and during primitive streak (PS) formation.”

Reviewer #2 (Public Review):

Summary:

The authors are interested in large-scale cell flow during gastrulation and in particular in the polonaise movement. This movement corresponds to a bilateral vortex-like counter-rotating cell flow and transport the mesendodermal cells allowing ingression of cells through the primitive streak and ultimately the formation of the mesoderm and endoderm. The authors specifically wanted to investigate the coupling of the polonaise movement and primitive streak to understand whether the polonaise movement is a consequence of the formation of the primitive streak or the other way around. They propose a model where the primitive streak elongation is not required for the cell flow but rather for its maintenance and that robust cell flow is not required for primitive streak extension.

Strengths:

Overall, the manuscript is well written with clear experimental designs. The authors have used live imaging and cell flow analysis in different conditions, where either the formation of the primitive streak or the cell flow was perturbed.

Their live imaging and PIV-based analyses convincingly support their conclusions that primitive streak deformation or mitotic arrest do not impact the initiation of the polonaise movement but rather the location or maintenance of these rotations. They additionally showed that disruption of the polonaise movement in the authentic primitive streak by elegant addition of an ectopic primitive streak does not impact the original primitive streak elongation.

Weaknesses:

• When using the delta-DEP-GFP construct, the authors showed that they can manipulate the shape of the primitive streak without affecting the identity and number of primitive streak cells. It is not clear however how this can affect the shape, volume or adhesion of the cells. Some mechanistic insights would strengthen the paper.

We appreciate the reviewer’s invaluable feedback. We agree that it would be informative to know how the ΔDEP-GFP construct led to PS deformation. This approach has been previously introduced by Voiculescu et al., (2007) to demonstrate an involvement of the Dsh(DEP) in PS shape regulation as described in text (please see pp4-5, lines 91-94 in Results and p13, lines 279-281 in Discussion). The previous study suggested that the Wnt/PCP pathway through Dsh(DEP) is a major regulator of cell intercalation, which plays an important role in PS morphogenesis (Voiculescu et al., 2007).

• Overall, frequencies of observation are missing for a better view of the phenomenon. For example, do Vg1/Cos cells always disrupt the flow at the authentic primitive streak? Can replicate vector fields be integrated to reflect quantification?

We agree and have added the numbers of embryos examined. In our experimental system, the Vg1/COS-implanted embryos always exhibited that the original polonaise movements along the authentic PS were always disrupted by the induced polonaise movements (n=4/4 embryos). The replicated vector fields were integrated to the Streamline and Vorticity plots (please see Fig. 1-4, Fig. S1, S4-7).

• Since myosin cables have been shown to be instrumental for the polonaise movement, it would be interesting to better investigate how the manipulations by the delta-DEP-GFP construct, or Vg1/Cos affect the myosin cables (as shown in preliminary form for the aphidicolin-treated embryos).

We agree that investigations of cytoskeletons and motor proteins would provide deeper understandings as to how the ΔDEP-GFP construct and perhaps Wnt/PCP components work in PS formation and morphogenesis. We plan to examine, as a future study, the patterns of the myosin cables in the ΔDEP-GFP-misexpressing or Vg1/COS-implanted embryos to get better understanding the mechanism(s) of the polonaise movements as the reviewer pointed out.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

• The authors named the dominant-negative Dsh lacking DEP [dnDsh(deltaDEP)]-fused GFP as deltaDEP-GFP, presumably to distinguish it from the construct dnDsh-deltaPDZ previously reported. However, the prefix "dnDsh" conveys the critical function in the present study. The reviewer recommends spelling out dnDsh(deltaDEP)-GFP to clarify to readers which signal was manipulated.

We agree that it is necessary to distinguish our construct used in this study from the dnDsh-deltaPDZ construct. We have, therefore, clarified the abbreviation in the main text as follows (please see pp 4-5, lines 91-97): ‘The DEP domain of Dishevelled (Dsh; a transducer protein of Wnt signaling) is responsible for the non-canonical Wnt/PCP pathway (43, 44), and misexpression of dominant-negative Dsh lacking DEP [dnDsh(ΔDEP)] leads to deformation of the midline structures, including the PS (21). Further, the Wnt/PCP pathway is involved in cellular polarity and migration, while the canonical Wnt pathway regulates cell proliferation (45). We refer the dnDsh(ΔDEP)-GFP construct that we generated, as ΔDEP-GFP, and tested its ability to alter cellular polarity, resulting in PS deformation’.

• The authors described the "Vg1 plasmid DNA" as a gift from Claudio D. Stern and Jane Dodd. However, they should indicate the vector backbone, especially whether the vector carries the SV40 ori sequence. Ori-containing plasmids multiply after transfection as COS cells express the SV40T antigen, leading to protein overexpression.

We added the name of the plasmid ‘pMT23-Vg1-myc-GDF1’ to the ‘Material and methods’ section (please see p25, line 574). pMT23 expression vector is a derivative of pMT21 (Hume and Dodd, 1993) and contains SV40 ori (Wong et al., 1985).

Reviewer #2 (Recommendations For The Authors):

• Most of the comments are indicated in the public review.

• There are additionally minor modifications that would help readers interpret the figures. In Figure S1B and D, it is not clear to the reader what the asterisks indicate.

We added the sentence ‘The white asterisks indicate GFP-expressing cells.’ to the figure legend of the Fig. S1 B and D (please see p34, line 874).

Reviewer #2 (Public Review):

Summary:

The authors are interested in large-scale cell flow during gastrulation and in particular in the polonaise movement. This movement corresponds to a bilateral vortex-like counter-rotating cell flow and transport the mesendodermal cells allowing ingression of cells through the primitive streak and ultimately the formation of the mesoderm and endoderm. The authors specifically wanted to investigate the coupling of the polonaise movement and primitive streak to understand whether the polonaise movement is a consequence of the formation of the primitive streak or the other way around. They propose a model where the primitive streak elongation is not required for the cell flow but rather for its maintenance and that robust cell flow is not required for primitive streak extension.

Strengths:

Overall, the manuscript is well written with clear experimental designs. The authors have used live imaging and cell flow analysis in different conditions, where either the formation of the primitive streak or the cell flow was perturbed.<br /> Their live imaging and PIV-based analyses convincingly support their conclusions that primitive streak deformation or mitotic arrest do not impact the initiation of the polonaise movement but rather the location or maintenance of these rotations. They additionally showed that disruption of the polonaise movement in the authentic primitive streak by elegant addition of an ectopic primitive streak does not impact the original primitive streak elongation.

Weaknesses:

- Since myosin cables have been shown to be instrumental for the polonaise movement, it would be interesting to better investigate how the manipulations by the delta-DEP-GFP construct, or Vg1/Cos affect the myosin cables (as shown in preliminary form for the aphidicolin-treated embryos).

Thank you for indicating that this will be a focus of future studies.

eLife assessment

Large scale cell movements occur during gastrulation in vertebrate embryos but their role in this major morphogenetic transition in formation of the body plan is poorly understood. Using the chick embryo model system, this study makes important advances using elegant methods to show that extension of the primitive streak during gastrulation, occurring through cell proliferation, polarisation and intercalation, and large-scale polonaise cell movements, can be uncoupled. Although the driving mechanism and precise role of these movements remains a mystery, the study provides convincing evidence for the uncoupling through independent approaches, the most creative of which are the effects shown after induction of a supernumerary primitive streak.

Reviewer #1 (Public Review):

In chicken embryos, the counter-rotating migration of epiblast cells on both sides of the forming primitive streak (PS), a process referred to as polonaise movements, has attracted longstanding interest as a paradigm of morphogenetic cell movements. However, the association between these cell movements and PS development is still controversial. This study investigated PS development and polonaise movements separately at their initial stage, showing that both could be uncoupled (at least at the initial phase), being activated via Vg1 signaling.

Strengths of this study

Polonaise movements, i.e., the circular cell migration of epiblast cells on both sides of the forming PS in avian embryos, have been the subject of research through live imaging and promoted the development of new tools to analyze quantitatively such movements. However, conclusions from previous studies remain controversial, at least partly due to the nature of perturbations to PS development and polonaise movements.

This study performed the challenging technique of electroporation to successfully mark and manipulate Wnt/PCP pathways in unincubated chicken embryo cells at the initiation phase of these two processes. In addition, the authors separately altered PS development and polonaise movements: PS development was perturbed by inhibiting either the Wnt/PCP pathway or DNA synthesis using aphidicolin, while polonaise movements were modified by the development of a second PS after engrafting Vg1-expressing COS cells located at the opposite end of the blastoderm. The study concluded that Vg1 elicits both PS development and polonaise movements, which occur in a parallel and are not inter-dependent.

To support these conclusions, particle image velocimetry (PIV) of cell trajectories captured by live imaging was performed. These tools delineated visually appealing cell movements and gave rise to vorticity profiles, adding more value to this study.

Weaknesses of this study

Engrafted Vg1-expressing COS cells located at the anterior end of the blastoderm elicited both the development of a second PS and marked bilateral polonaise movements while perturbing these movements along the original PS. How do polonaise movements along the second PS dominate over those along the normal PS? The authors suggested a model in which Vg1 acts in a graded or dose-dependent manner since engrafted COS cells over-expressed Vg1. This model can be tested by reducing the mass of engrafted COS cells. Although the authors propose performing this analysis in further investigations, it would be preferable to incorporate into this study for better consistency.

Thank you for indicating that this will be a focus of future studies.

Author Response

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

Joint Public Review

This study is concerned with the general question as to how pools of synaptic vesicles are organized in presynaptic terminals to support different types of transmitter release, such as fast synchronous and asynchronous release. To address this issue, the authors employed the classical method of load- ing synaptic vesicle membranes with FM-styryl dyes and assessing dye destaining during repetitive synapse stimulation by live imaging as a readout of the mobilization of vesicles for fusion. Among other 1ndings, the authors provide evidence indicating that there are multiple reserve vesicle pools, that quickly and slowly mobilized reserves do not mix, and that vesicle fusion does not follow a mono-exponential time course, leading to the notion that two separate reserve pools of vesicles - slowly vs. rapidly mobilizing - feed two distinct releasable pools - reluctantly vs. rapidly releasing. These 1ndings are valuable to the 1eld of synapse biology, where the organization of synaptic vesicle pools that support synaptic transmission in different temporal and stimulation regimes has been a focus of intense experimentation and discussion for more than two decades.

On the other hand, the present study has limitations, so that the authors’ key conclusions remain incompletely supported by the data, and alternative interpretations of the data remain possible. The approach of using bulk FM-styryl dye destaining as a readout of precise vesicle arrangements and pools in a population of functionally very diverse synapses bears problems. In essence, the approach is ’blind’ to many additional processes and confounding factors that operate in the back- ground, from other forms of release to inter-synaptic vesicle exchange. Further, averaging signals over many - functionally very diverse - synapses makes it diicult to distinguish the dynamics of separate vesicle pools within single synapses from a scenario where different kinetics of release originate from different types of synapses with different release probabilities.

We thank the editors and reviewers for their time and patience, and are happy that they found our results valuable.

We do not have a clear understanding of what the alternative interpretations might be - beyond those already addressed - but would like to. At present, we believe that the evidence for parallel processing of slowly and quickly mobilized reserve vesicles is solid and hope that people who are open to the possibility will evaluate the reasoning described within our report. The hypothesis that reserves are kept separate because they feed distinct subdivisions of the readily releasable pool remains to be tested.

Beyond that, we have used FM-dye de-staining as a bulk measurement of sub-synaptic events in the sense that we have made no attempt to measure mobilization of isolated individual vesicles. We do not see how this necessarily leaves viable alternative interpretations, but this is diZcult to evaluate without knowing what the alternatives might be. On the other hand, the FM-dye technique has had good resolution at the level of distinguishing between individual synapses since at least Murthy et al. (2001). For our part, we are con1dent that our analysis in Figure 3 combined with the results in Figures 4-11 shows that the multiple reserve pools co-occur in many individual presynaptic terminals. We did not use electron microscopy to con1rm that all of the punctae analyzed in Figure 3 were indeed single synapses, but the reviewers did not recommend this, and we believe there is already enough published about the spatial distribution of synapses in cell culture to be con1dent that many of the punctae that are smaller than 1.5 µm were individuals.

Overall, we have attempted to address all of the individual concerns raised by reviewers, and our understanding is that these concerns and our responses will be available on the eLife website. The reviewers were not convinced on every point, but these are cases where the nature of the concern was not clear to us. We hope that people who share these concerns will check out our responses and contact us with any further questions or alternative interpretations.

(1) The authors sincerely addressed many of the previous concerns, mainly by clari1cation. The data are consistent with the authors’ hypothesis. The pool concept is somewhat similar to that of Richards et al (2000) and Rey et al (2015). The authors further propose that two reserve pools feed vesicles to two readily-releasable pools independently.

To clarify further: The possibility that distinct reserve pools feed distinct readily releasable pools is predicted by our working model, and is something that we would like to test in the future, but is not a conclusion of the present study. Instead, in the present study, we tested the prediction that quickly and slowly mobilized reserve vesicles are processed in parallel without making assumptions about the the underlying mechanism.

Unfortunately, the heterogeneity among individual synapses remains a concern as shown in (some of) the raw data (Fig. 3 and supplements).

We emphasize that we have not attempted to minimize the extensive heterogeneity among synapses, but actually highlight this. In fact, we chose the image in Figure 3 for an example in part because of the lower left region replicated in Figure 3 supplement 2 demonstrating extensive heterogeneity along what appears to be a single axon. We are not the 1rst to notice the heterogeneity (see Waters and Smith, 2002), but we do provide a new possible explanation which, if correct, might be impor- tant for understanding biological computation (see our Discussion). At the same time, we believe that our evidence for multiple reserve pools within individual synapses with heterogenous properties is compelling. We see no contradiction, and indeed, our conclusion that the ratio of slowly to quickly mobilized varies extensively between synapses can only be correct if individual synapses contain mul- tiple types. We hope that people who are interested in our conclusions will evaluate the evidence and reasoning presented in our report.

Bulk imaging of FM de-staining does not really measure the fraction of non-stained vesicles, which changes dynamically during stimulation, so that the situation calls for an independent readout of stained and non-stained vesicles. Moreover, direct correspondence between two speci1c stimulation frequencies (with long stimulation) and vesicle pools is not straightforward. These issues make the experimentally measured pools not well-de1ned.

We think that the reviewer is suggesting an alternative scenario where decreases in the fractional rate of FM-dye de-staining seen during 1 Hz stimulation might be caused by a large (4-fold) increase in the total size of the reserve pool that dilutes the stained vesicles by mixing. This scenario is consis- tent with the results in Figures 2 and 4-7, and initially seems plausible because previous studies have shown that many vesicles are not mobilized, and therefore are not stained, during our standard load- ing protocol of 100 s at 20 Hz (Harata et al., 2001). However, liberation of this "deep reserve" as an explanation for the decrease in fractional destaining is not compatible with the results in Figures 10-11 that rule out mixing. For example, liberation of the deep reserve would cause fractional destaining to appear equally depressed during subsequent 20 Hz stimulation, and Figure 10 shows that this is not the case. The scenario cannot be rescued by postulating that the subsequent 20 Hz stimulation caused the deep reserve to quickly recapture the liberated vesicles because Figure 11D-E shows that fractional de-staining continues to be depressed at the very beginning of a second 1 Hz train that follows the 20 Hz stimulation.

(2) The authors’ latest round of responses did not alleviate most of my major previous concerns. The additional data now shown in Fig 3 rely on conceptually the same type of bulk measurements and thus suffer from the same limitations as outlined in the earlier review.

We believe that the new evidence in Figure 3 for multiple reserve pools at individual synapses is strong when evaluated in combination with the results in Figures 4-11. We do not, at present, see how the fact that FM-dye destaining is used as a bulk measurement at the sub-synaptic level could undercut our logic.

Moreover, the image of neuronal cultures shown in Fig. 3 might be problematic. It shows very bright staining with large round lumps, which may be indicative of unhealthy cultures.

Unhealthy cultures are not a concern because we used strict quantitative criteria to assess health that are better than we have seen elsewhere (details below). We think the reviewer might be reacting to the way we rendered the image; i.e., as “overexposed”. We did this to highlight the dimmest punctae, which is a key element of the analysis. The same image rendered with less contrast is now displayed in Author response image 1 (3rd panel from left).

Author response image 1.

Image to left is a reproduction of the example image in Figure 3, which was the average of 120 time lapse raw data images; scale bar is 20 µm. The second image is a replicate except all 69 punctae that were included in the study are occluded by 1.5 µm × 1.5 µm yellow squares. The third image is another replicate except with a different brightness setting. The rightmost image is one of the raw data images with brightness matched to the third image.

More details (relevance to in vivo is in point 4):

(1) Identifying unhealthy cultures is straightforward with our technique because synapses in un- healthy cultures destain spontaneously. Our criteria for accepting experiments for further analy- sis was less than 1.5 % spontaneous rundown/minute. This is a better way to judge health than we have seen elsewhere because it eliminates subjective decisions, and would be equally appli- cable for microscopes and imaging software of any quality. For our part, we used a 25X objective with a low numerical aperture and low intensity illumination that allowed us to completely avoid photobleaching. The images will look worse to some compared to when acquired with a higher quality microscope, but the absence of photobleaching is an important bene1t because it allowed us to avoid complicated corrections.

(2) Stained areas larger than 1.5 µm across - such as the ones noted by the reviewer - were expressly excluded from our study because they could have been clusters of multiple synapses. The size criteria are detailed in the Legend of Figure 3. Punctae and larger areas that were excluded are the ones that are not occluded by yellow squares in the 2nd image from the left, above; at least two of the largest were likely clusters of synapses that were out of focus. Nevertheless, despite being excluded, it is unlikely that the stained areas larger than 1.5 µm in the image in Figure 3 were characteristic of unhealthy cultures because these areas did not de-stain spontaneously, but instead de-stained in response to 1 and 20 Hz electrical stimulation much like the small punctae that were included in the analysis.

(3) Electron microscopy results have shown that individual synapses vary >10-fold in size, so a large range of brightness is expected (Murthy et al., 2001). The large range would either make the brighter punctae and clusters appear to be overexposed in a printed image, or render the dimmer punctae invisible. We have opted to present an image with overall brightness adjusted so that the dimmest punctae are visible. This is appropriate because one of the concerns was that analyzing the dimmest punctae would reveal underlying populations where the rate of fractional destaining was constant. In the end, no evidence for underlying populations emerged, which supports the conclusion that the decreases in fractional destaining occur at individual synapses. Note that adjusting brightness for example images was unavoidable; we used the camera in a range that was far below saturation and, because of this, images presented without adjusting brightness would appear to be completely black.

(4) Primary cell cultures are non-physiological by de1nition, so the concept of health is intrinsically arbitrary, and relevance to synapses in brains is questioned routinely. However, the new 1ndings in the present report are that: (1) individual hippocampal synapses contain multiple reserve pools; (2) the reserves remain separate but are not distinguishable by the timing of mobilization when the frequency of stimulation is high; and (3) the reserves are nevertheless processed in parallel even when the frequency of stimulation is high. Of these, 1nding (1) has been reported previously for other synapse types, but 1ndings (2) and (3) were both unexpected, and 1nding (3) was not compatible with current concepts. Nevertheless, all three 1ndings were predicted by a model that was developed to explain orthogonal results from studies of intact synapses in ex vivo slices that did not 1t with current concepts either, as referenced in the Introduction. Because of this, we think that the parallel processing of quickly and slowly mobilized reserve vesicles likely occurs in individual Schaffer collateral synapses in vivo, and is not a cell culture artifact; the alternative would be too much of an unlikely coincidence.

References

Harata N, Pyle JL, Aravanis AM, Mozhayeva M, Kavalali ET & Tsien RW (2001). Limited numbers of recycling vesicles in small CNS nerve terminals: implications for neural signaling and vesicular cycling. Trends in Neurosciences 24, 637–43.

Murthy VN, Schikorski T, Stevens CF & Zhu Y (2001). Inactivity produces increases in neurotransmitter release and synapse size. Neuron 32, 673–82.

Waters J & Smith SJ (2002). Vesicle pool partitioning in2uences presynaptic diversity and weighting in rat hippocampal synapses. Journal of Physiology 541, 811–23.

---

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

Reviewer 1

Mahfooz et al. investigated the time course of synaptic vesicle fusion of cultured mouse hippocampal synapses using FM-styryl dyes. The major finding is that the FM destaining time course deviates from a mono-exponential function during 1 Hz, but not 20 Hz stimulation. The deviation from a mono-exponential function was also seen during a second stimulus train applied after recovery periods of several minutes, or after depletion of the readily-releasable vesicle pool. Furthermore, this "decreased fractional destaining" was unlikely due to long-term synaptic depression, or incomplete dye clearance. Fractional destaining was enhanced when the dye was loaded with 1 Hz compared with 20 Hz stimulation, suggesting that vesicles recycled during 1 Hz stimulation are predominantly sorted into a rapidly mobilized pool. Finally, they show that 20 Hz stimulation does not affect the decrease in fractional destaining induced and recorded during 1 Hz stimulation. Based on these observations, they put forward a model in which slowly and quickly resupplied synaptic vesicles are mobilized in parallel.

The demonstration that FM destaining time courses deviate from single exponentials during 1 Hz stimulation (Figs 2-3) is a starting point used to rule out simple models where vesicles intermix freely and to introduce a mathematical technique for quantifying the extent of the deviations that is essential for the analysis of later experiments, where curve fitting could not be used. We then:

1) Show that the deviation from simple models is not caused by depletion of the readily releasable pool, as noted by the reviewer;

2) rule out a number of explanations for the deviation that do not involve reserve pools at all, again as noted;

3) provide affirmative evidence for the presence of multiple reserve pools by labeling them with distinct colors;

4) show that the vesicles within the distinct reserve pools do not intermix even when activity is intense enough to drive destaining with single exponential kinetics.

We believe that the 4th point - documented in Figs 10-11 - is a key element.

Beyond that, we note that our working model arose from previous studies, as referenced in the Introduction, not from the present results. The model did predict the parallel processing of quickly and slowly mobilized reserves, and the present study was designed to test this prediction. In that sense, the evidence in the current study supports our working model, not the other way around.

In any case, most readers in the near term will be more interested in the serial versus parallel question, and less in precisely what the present results mean for evaluating our working model. Because of this, we emphasize that evidence for parallel processing of separate reserve pools depends solely on experimental results within the study, and not on modeling. As a consequence, the evidence will continue to be equally strong even if problems with our working model arise later on (lines 382-386).

We do have additional unpublished evidence for the working model that does not bear directly on the parallel versus serial question. Some of this was removed from an earlier version of the manuscript and some has been newly gathered since the original submission. We will publish the additional evidence at a later point. We decided not to include it in the present manuscript expressly to avoid confusion about the relationship between modeling and the evidence for parallel processing in general.

The paper addresses an interesting question - the relationship between the resupply and release of synaptic vesicles. The study is based on a lot of data of high quality. Most data are solid. However, some of the major conclusions are not well supported by the data. Moreover, it remains unclear how speci1c the findings are to the experimental design.

The following points should be addressed:

1) Most traces display a decrease in fluorescence intensity before stimulation. Data with a decrease in baseline fluorescence intensity of up to 1.5 % were considered for the analysis (Fig 2-supplement 2). I may have missed it, but were the data corrected for the observed decrease in baseline fluorescence intensity? (In the model shown in Appendix 1 Figure 1, they correct for "rundown"). For instance, are the residuals shown in Fig 2D, E based on corrected data? In case the data would not be corrected for a decrease in baseline fluorescence, would the decay kinetics also deviate from a single exponential after correction?

We did not correct for rundown - as now noted on lines 96-97 - except in the figure in the Appendix, noted by the reviewer, where the uncorrected and corrected time courses are plotted side by side for easy comparison. However, our study includes an analysis showing that correcting for rundown during 1 Hz stimulation would increase - not decrease - the deviation from a single exponential (2 bars in rightmost panel in Fig 2C, and lines 113-116 of Results), so the absence of a correction does not weaken our conclusions.

2) The analysis of "fractional destaining" is not clear to me. How many intervals of which length were chosen and why? For instance, the intervals often differ in length, number and do not cover the complete decay (e.g., Fig 2B).

We calculated fractional destaining from longer intervals at later times because the overall amount of stain was less, meaning signal/noise was less, and scatter was more. We did this because increased scatter at later times could be counteracted by estimating the slope of destaining from longer intervals. An additional bene1t is that elongating the later intervals allowed us to plot only 6 bars for 25 min of 1 Hz destaining, which works better visually than 17.

Increasing the interval length for later times is mathematically sound because the key factor causing distortions related to deviations from linearity is not the length of the interval per se but, instead, the fractional destaining over the interval. The fractional destaining is greater at the start of 1Hz stimulation, thus requiring shorter intervals.

It would be possible to choose inappropriately long intervals that would distort estimates of the change in fractional destaining. However, we now include Fig 2-supplement 6 – which includes all 17 1.5 min intervals - to con1rm that any distortions after the first interval were minimal. The Appendix predicts a biologically important distortion for the first interval which we are following up, but this would underestimate the true deviation from quickly mixing pools, so would not be problematic for the present conclusions.

Sometimes, only the interval right after stimulation onset was considered (e.g., Fig 7, 8).

Figs 7, 8 in the previous version are now Figs 8, 9.

This is appropriate because the goal was to estimate the fractional destaining at the very start, before the quickly mobilized fraction has destained.

How quickly fractional destaining is expected to revert to the lowest value seen after 15 min of 1Hz stimulation in Fig 2 (and elsewhere) depends very much on assumptions - such as the number of reserve pools, etc. We sought to avoid this kind of additional analysis because we are keen to avoid the impression that our main conclusions depend on the speci1cs of modeling.

How sensitive are the changes in fractional destaining to the choice of the intervals?

Minimally. This can be seen by eye because the magenta lines in Fig 2B 1t the data well, but see Fig 2-supplement 6 for a quantitative comparison.

For instance, would fractional destaining be increased if later intervals would have been chosen for the second 20 Hz stimulus in the experiment shown in Fig 9B?

Previous Fig 9B is now Fig 10B.

We cannot be certain, but think it probably would not be different. Neither an increase nor a decrease would be problematic for our conclusions.

More detail: There is not enough data to evaluate this specifically for Fig 10B because the total amount of stain remaining at later intervals is little, meaning signal/noise is low, which causes extensive experimental scatter. However, synapses were even more extensively destained prior to time course c of Figure2-supplement 2C, which nevertheless matches time courses a, b, and d.

I propose fitting all baseline-corrected data with a single and a double-exponential function (as well as single exponential plus line?) and reporting the corresponding time constants (slopes) and amplitudes.

As noted above, we purposefully do not baseline correct data in a way that would make this possible. However, we do include exponential fits when appropriate, in Fig 2D-E, Fig 2- supplement 1, Fig 2-supplement-7, Fig 2-supplement-8, and Fig 12B.

Indeed, the absence of any change in the weighting parameter despite substantial changes for both time constants seen after raising the temperature to 35C (Fig 2-supplement-8 vs Fig12B) is notable because it suggests that the contents of the reserve pools are not altered by changing temperature, even though vesicle trafficking is accelerated. Fig 2-supplement-8 is a supplementary figure because the result is outside the scope of the main point, not because the quality is lower than for other figures.

Beyond that, exponential fits would not be adequate for most of the study because many experiments - including the core experiments in Figs 10-11 - require discontinuous stimulation, such as when we stop stimulating at 1 Hz, rest for minutes, and then start up again at 1 or 20 Hz. And, although widely used, exponentials are non-linear equations after all. Even when they can be used to quantify time courses, the fractional destaining measurement is almost always more informative, in the technical sense, because it avoids complications when estimating the importance of deviations occurring at the two extremes versus deviations in the middle of the time course.

3) Along the same lines, is the average slow time constant indeed around 40 min? (Are the data shown in Fig 2 S7 based on an average?) If this would be the case, I suggest conducting a control experiment with a recording time > 40 min. Would fitting an exponential or a line to baseline data (without stimulation) also give a similar slow component?

Fig 2-supplement 7 in the previous version is now Fig 2-supplement 8.

First, yes, the time course shown in Fig 2-supplement 8 is the mean across preparations. The time courses of the individual preparations were quanti1ed as the median value of the individual ROIs before averaging.

Second, no, fitting baseline data would give an approximately 3-fold greater time constant (i.e., 120 min) because fractional destaining decreases by about 3-fold when we stop stimulating after 25 min of 1 Hz stimulation (i.e., Fig 2C, 3B, and many others).

The key point is that fractional destaining decreases greatly over long trains of 1 Hz stimulation.

For Fig 2, we saw a 2.7+/-0.1-fold decrease before accounting for baseline destaining (lines 106-110), which increased to a 4.4-fold decrease when we did account for baseline destaining (lines 113-116). Overall, the 2.7-fold value is simultaneously a safe minimum boundary, and much greater than the value of 1.0 expected from models where vesicles mix freely.

Note that future studies will show that even the 4.4-fold value is probably an underestimate because 1 Hz stimulation misses a fast component at the very beginning of the time courses, as predicted in the Appendix.

4) How speci1c are the findings to 1 Hz (and 20 Hz) stimulation? From which frequency onward can a decrease in fractional destaining be no longer observed?

Our logic depends only on the premise that we are able to find some frequency where fractional destaining no longer decreases. We knew that 20 Hz was a good place to start because of previous electrophysiological experiments - frequency jumps (Fig 1 of Wesseling and Lo, 2002 and Fig 2C of Garcia-Perez and Wesseling, 2008), and trains of action potentials followed by osmotic shocks (Fig 2A of Garcia-Perez et al., 2008) - showing that 20 Hz stimulation is enough to nearly completely exhaust the readily releasable pool. This is noted in lines 202-203, and Box 2.

would previous stimulation with frequencies <20 Hz interfere with fractional destaining? These control experiments would help assessing how general/speci1c the findings are.

Yes (Figs 4 and 11A at 1 Hz). Also, we have done experiments at 0.1 Hz, which will be published later; some of these were actually removed from an earlier version of the manuscript because the results are primarily relevant to deciding between particular parallel models, and are not relevant to the conclusion of the present study that quickly and slowly mobilized reserves are processed in parallel.

Similarly, a major conclusion of the paper - the parallel mobilization of two vesicle pools - is largely based on these two stimulation frequencies. Can they exclude that mixing between the two pools occurs at other frequencies?

We cannot exclude the possibility of breakdown at a higher frequency, but this would not undercut our conclusions. We do not have plans to try this experiment because: (1) a positive result would be open to concerns about non-physiologically heavy stimulation; and (2) a negative result would be difficult to interpret because of the possibility that the axons cannot follow at higher frequencies.

6) Some information in the methods section is lacking. For instance, which species is the cell culture based on?

Mice from both sexes were used. This is now speci1ed in the Methods.

Reviewer 2

By using optical monitoring of synaptic vesicles with FM1-43 at hippocampal synapses, the authors try to show the evidence for two parallel reserve pools of synaptic vesicles, which feed the vesicles to the readily releasable pool. The major strength of the study is the use of a quantitative model, which can be readily testable by experiments: in the course of the study, the authors propose the best vesicle pool model, which fits the experimental data "averaged over synapses" nicely. On the other hand, the weak point of the study comes from the optical method and the data: bulk imaging of vesicle dynamics monitored at each synapse is noisy and the signals vary considerably among synapses. Therefore, the average signals over many synapses may not reflect the vesicle dynamics of two reserve pools within a synapse, but something else, such as the different kinetics of release from multiple synapses with different release probability. Nevertheless, a new framework of two reserve pools offers a testable hypothesis of vesicle dynamics, and the use of single vesicle tracking and EM may allow one to give a de1nitive answer in the future studies Therefore, the study may be of interest to the community of synaptic neurobiology.

1) The current version includes a new figure (Fig 3) showing that the deviations from single pool models seen in populations are caused by deviations occurring at the level of single synapses. The heterogeneity between synapses actually causes population statistics to underestimate - not overestimate - the mean and median size of the deviations at individuals.

We think the new evidence in Fig 3 and supplements is conclusive without follow-on EM of the same punctae given the substantial body of already published EM on similar cultures. Essentially, the only way to explain the results without invoking multiple reserve pools in individual synapses would be to say that individual synapses ALWAYS come in clumps containing multiple types and are NEVER separated from neighbors by more than 1.5 microns - even when the clumps are separated from each other by 5 microns. There is already clear evidence against this.

2) No new model is proposed here, see the first response to the first reviewer.

3) We are not aware of alternative hypotheses that could account for our results, so cannot evaluate if single vesicle tracking and EM could add meaningful additional support.

1) The existence of non-stained vesicles complicates the interpretation of the data. Because the release by 20 Hz and 1 Hz stimulation do not entirely reflect the release from fast and slow vesicle pools. the estimation of non-stained vesicles using synaptopHluorin (+ba1lomycin) and EPSCs would be helpful to examine fraction of non-stained / stained vesicles over time (with stimulation, the ratio may change dynamically, which may bring complications).

Non-stained vesicles are not a complication, but instead a key element of our logic which is included in the diagrams in Boxes 1 and 2 and Figure 9. That is, quickly and slowly mobilized reserves can be distinguished at 1 Hz precisely because 1 Hz is not intense enough to exhaust the readily releasable pool (Box 2). The corollary is that stained vesicles must be replaced by non-stained vesicles, because otherwise 1 Hz stimulation would exhaust the readily releasable pool. And this is why FM-dyes (plus a beta-cyclodextrin during washing) are ideal for the current questions whereas other techniques, such as electrophysiology or synaptopHluorin imaging are obviously indispensable for other questions, but could not replace the FM-dyes in the current study. This is now noted on lines 86-89.

We are aware that synaptopHluorin + ba1lomycin could, in principle, accomplish some of the same goals. However, ba1lomycin ended up being toxic when applied for tens of minutes, as it would have to be in our experiments. And, we do not see what critical question is not already answered with strong evidence using FM dyes.

2) Individual synapses show marked differences in the time course of de-staining, suggesting differences in release probability. The averaging of the whole data may reflect "average" behavior of synapses, but for example, bi-exponential time course may reflect high Pr and low Pr synapses, rather than vesicle recruitment.

The authors may comment on this issue.

See newly added Fig 3, and responses above.

3) Some differences are very small (Fig 10, the same amplitude as bleaching time course), and I am not certain if the observed differences are meaningful, given low signal to noise ratio in each synapse.

Fig 10 in the previous version is Fig 11 in the current version.

Even if correct, this would not be problematic because 20 Hz stimulation clearly did not cause fractional destaining to return to the initial value when stimulation was resumed at 1 Hz (compare d and f in Fig 11E). In any case, Figs 2C, 3B, 5B, 7B, and Fig 10-supplement 2A all show that the minimum fractional destaining value during 1 Hz stimulation is about 3-fold greater than during subsequent rest intervals, which is not a small difference. Also, note that Fig 2-supplement 3 shows that photobleaching likely did not play a role.

Reviewer 3

Reviewer #3 (Recommendations For The Authors):

This study attempts to conceptualize the long-standing question of vesicle pool organization in presynaptic terminals. Authors used classical FM dye release experiments to support a hypothesis that rapidly and slowly releasing vesicles are mobilized in parallel without intermixing. This modular model is also supported indirectly by the authors’ recent findings of molecular links that connect a subset of vesicles in linear chains (published elsewhere).

Our study should be seen as a test of the hypothesis that quickly and slowly mobilized reserves are processed in parallel. The evidence is independent of any modeling, and would continue to be equally strong if our working model turns out to be incorrect (lines 382-386).

The scope of the original model was limited by a number of caveats. The main concerns included a limited data set measured in bulk from a highly heterogeneous synapse population, and a complex interrelationship between vesicle mobilization and the bulk FM dye de-staining kinetics. The second major limitation was measurements being performed at room temperature, which inhibits or alters a number of critical synaptic processes that are being modeled. This includes the efficiency of exo/endocytosis coupling, vesicle mobility and release site refractory period, which are stimulus- and temperature-dependent, but were not accounted for in the original model.

The present study contains experiments at body temperature (Fig 12 and Fig 12-supplement 1 in the current version) and analyses of individual synapses (especially Fig 3 in the current version). To our knowledge all results are consistent with everything that is known about the efficiency of exo/endocytosis coupling, vesicle mobility and release site refractory periods.

The authors made strong efforts to address previous concerns. However, the main conceptual point, i.e. linking the bulk FM dye de-staining kinetics with precise arrangement of vesicle pools, is not well supported and is generally highly problematic because it ignores many additional processes and confounding factors.

For example, vesicle exchange between neighboring synapses constitutes from 15% to over 50% of total recycling vesicle population, and therefore is a major contributing factor to FM dye loss/redistribution, but is not considered in this study. Additionally, this vesicle exchange process undergoes calcium/activity-dependent changes, contributing to difficulty in interpreting the current experiments comparing FM de-staining at different stimulation frequencies.

We do not see how exchange of vesicles between synapses could be a problem for our logic, so cannot evaluate this without a more detailed description of the concern. Instead, our results rule out random inter-synaptic exchange between quickly and slowly mobilized reserve pools because this would show up in our assays as mixing, which does not occur. We think there are three remaining possibilities:

1) vesicles are exchanged primarily between quickly mobilized reserve pools

2) vesicles are exchanged primarily between slowly mobilized reserve pools

3) vesicles in quickly mobilized reserve pools are targeted to quickly mobilized reserve pools in other synapses and vesicles in slowly mobilized reserve pools are targeted to slowly mobilized reserve pools in other synapses.

It would be interesting to know which of these is correct, but this is outside the scope of the current study.

Moreover, other forms of release, such as asynchronous release, contribute a large fraction of released vesicles, but are not factored in. Asynchronous release varies widely in synapse population from 0.1 to >0.4 of synchronous release, but is entirely ignored. Spontaneous release may also contribute to FM dye loss over extended 25min recordings used.

Spontaneous release and asynchronous release are not caveats.

First, spontaneous: We suspect that spontaneous release contributes to the background destaining rate, but this is 3-fold slower than the minimum during 1 Hz stimulation on average (Figs 2C, 3C, 5B etc), so we know that the slowly mobilized reserve is mobilized by low frequency trains of action potentials (lines 410-412). Note that a different outcome - where the rate of destaining decreased to a very low level during long trains of 1 Hz stimulation - would not have been consistent with the idea that slowly mobilized vesicles are only released spontaneously because the remaining fluorescence can always be destained rapidly by increasing the stimulation intensity to 20 Hz (e.g., see examples in Fig 3).

Second, asynchronous: We know that slowly mobilized reserves must be released synchronously at 35C because the asynchronous component is eliminated at this temperature (Huson et al., 2019), without altering the quantity of slowly mobilized reserves that are mobilized by 1 Hz stimulation (lines 350-360 of Results, and 445-452 of Discussion; we can con1rm from our own unpublished experiments that the disappearance of asynchronous release at 35C is a robust phenomenon in these cell cultures). Asynchronous release of slowly mobilized vesicles might occur at room temperature, but this would not argue against the conclusion that slowly mobilized vesicles are processed in parallel with quickly mobilized.

Speci1c comments:

Points 1-4 are already addressed above.

5) The notion of the chained vesicles is somewhat confusing: how does the "first" vesicle located at the plasma membrane/release site get released if it is attached to the chain? Wouldn’t this "first" vesicle be non-immediately releasable since it must first be liberated? Since all vesicles shown in the Figure 1 have chains attached to them, what vesicle population then give rise to sub-millisecond release?

This is not a concern relevant to the present study because none of the conclusions rely on the model in any way (see Introduction, and lines 382-386 of the Discussion). Beyond that: We previously published clear evidence that docked vesicles are tethered to non-docked vesicles (Figure 8 of Wesseling et al., 2019). We see no reason to suspect that a tether to an internal vesicle would prevent the docked vesicle from priming for release.

7) Model: For fitting de-staining during 20 Hz stimulation, authors state that it was necessary to allow >5-fold Facilitation. This seems to be non-physiologically relevant, since previous studies found only very mild facilitation at room temperature (typically below a factor of 1.5-2.0) and the authors themselves state that, at most, a 1.3 fold facilitation was found.

If the 1.3-fold facilitation estimate comes from us, it must have been in a different context.

Most estimates of facilitation that are published are heavily convolved with simultaneous depression, and there is additionally a saturation mechanism for readily releasable vesicles with high release probability that is not widely known (Garcia-Perez and Wesseling, 2008). The standard method for eliminating the depression is to lower the probability of release by lowering extracellular [Ca2+], which additionally relieves occlusion by the saturation mechanism. And, lowering [Ca2+] uncovers an enormous amount facilitation at synapses in hippocampal cell culture. For example, see Figure 2B of Stevens and Wesseling (1999), which shows a 7-fold enhancement during 9 Hz stimulation, and Figure 3 of the same study, which shows a linear relationship with frequency. Taken together these two results suggest 15-fold enhancement during 20 Hz stimulation, which far exceeds the 5-fold value needed at inefficient release sites to make our working model 1t the FM-dye destaining results.

References

Garcia-Perez E, Lo DC & Wesseling JF (2008). Kinetic isolation of a slowly recovering component of short-term depression during exhaustive use at excitatory hippocampal synapses. Journal of Neurophysiology 100, 781–95.

Garcia-Perez E & Wesseling JF (2008). Augmentation controls the fast rebound from depression at excitatory hippocampal synapses. Journal of Neurophysiology 99, 1770–86.

Huson V, van Boven MA, Stuefer A, Verhage M & Cornelisse LN (2019). Synaptotagmin-1 enables frequency coding by suppressing asynchronous release in a temperature dependent manner. Scienti1c reports 9, 11341.

Stevens CF & Wesseling JF (1999). Augmentation is a potentiation of the exocytotic process. Neuron 22, 139–46.

Wesseling JF & Lo DC (2002). Limit on the role of activity in controlling the release-ready supply of synaptic vesicles. Journal of Neuroscience 22, 9708–20.

Wesseling JF, Phan S, Bushong EA, Siksou L, Marty S, Pérez-Otaño I & Ellisman M (2019). Sparse force-bearing bridges between neighboring synaptic vesicles. Brain Structure and Function 224, 3263–3276.

Joint Public Review:

This study is concerned with the general question as to how pools of synaptic vesicles are organized in presynaptic terminals to support different types of transmitter release, such as fast synchronous and asynchronous release. To address this issue, the authors employed the classical method of loading synaptic vesicle membranes with FM-styryl dyes and assessing dye destaining during repetitive synapse stimulation by live imaging as a readout of the mobilization of vesicles for fusion. Among other findings, the authors provide evidence indicating that there are multiple reserve vesicle pools, that quickly and slowly mobilized reserves do not mix, and that vesicle fusion does not follow a mono-exponential time course, leading to the notion that two separate reserve pools of vesicles - slowly vs. rapidly mobilizing - feed two distinct releasable pools - reluctantly vs. rapidly releasing. These findings are valuable to the field of synapse biology, where the organization of synaptic vesicle pools that support synaptic transmission in different temporal and stimulation regimes has been a focus of intense experimentation and discussion for more than two decades.

On the other hand, the present study has limitations, so that the authors' key conclusions remain incompletely supported by the data, and alternative interpretations of the data remain possible. The approach of using bulk FM-styryl dye destaining as a readout of precise vesicle arrangements and pools in a population of functionally very diverse synapses bears problems. In essence, the approach is 'blind' to many additional processes and confounding factors that operate in the background, from other forms of release to inter-synaptic vesicle exchange. Further, averaging signals over many - functionally very diverse - synapses makes it difficult to distinguish the dynamics of separate vesicle pools within single synapses from a scenario where different kinetics of release originate from different types of synapses with different release probabilities.

The reviewers commented on the revised version of your paper, in essence reiterating the limitations of the approach of bulk imaging of FM de-staining:

(1) The authors sincerely addressed many of the previous concerns, mainly by clarification. The data are consistent with the authors' hypothesis. The pool concept is somewhat similar to that of Richards et al (2000) and Rey et al (2015). The authors further propose that two reserve pools feed vesicles to two readily-releasable pools independently. Unfortunately, the heterogeneity among individual synapses remains a concern as shown in (some of) the raw data (Fig. 3 and supplements). Bulk imaging of FM de-staining does not really measure the fraction of non-stained vesicles, which changes dynamically during stimulation, so that the situation calls for an independent readout of stained and non-stained vesicles. Moreover, direct correspondence between two specific stimulation frequencies (with long stimulation) and vesicle pools is not straightforward. These issues make the experimentally measured pools not well-defined.

(2) The authors' latest round of responses did not alleviate most of my major previous concerns. The additional data now shown in Fig 3 rely on conceptually the same type of bulk measurements and thus suffer from the same limitations as outlined in the earlier review. Moreover, the image of neuronal cultures shown in Fig. 3 might be problematic. It shows very bright staining with large round lumps, which may be indicative of unhealthy cultures.

Reviewer #3 (Public Review):

Summary:<br /> In the manuscript by Valenzisi et al., the authors report on the role of WRNIP1 to prevent R-loop and TRC-associated DNA damage. The authors claim WRNIP1 localizes to TRCs in response to replication stress and prevents R-loop accumulation, TRC formation, replication fork stalling, and subsequent DNA damage. While the findings are of potential significance to the field, the strength of evidence in support of the conclusions is lacking.

In the revised submission by Velenzisi et al., the manuscript is still missing the controls that were requested in the original review. One cannot conclude the D37A mutant is unable to rescue DNA damage unless it is shown in the same experiment that the WT is able to rescue it. This is also true for the fork speed, stalled forks, and restarting forks experiments. Below is a list of Figures missing key controls.

Figure 1B -missing the shWRNIP1WT control<br /> Figure 1C - missing the MRC5SV control<br /> Figure 1D - missing the shWRNIP1WT control<br /> Figure 3C - missing the shWRNIP1WT control<br /> Figure 5A - missing the shWRNIP1WT control<br /> Figure 5B - missing the shWRNIP1WT control<br /> Figure 5C - missing the shWRNIP1WT control<br /> Figure 5D - missing the shWRNIP1WT control<br /> Figure 6C - missing the shWRNIP1WT control

Also, the authors did not explain the result showing shWRNIP1 decreases DNA damage compared to MRC5SV in Figure 1D (compare lanes 4 and 8). Again, this suggests WRNIP1 actually increases DNA damage in response to Aph and DRB. This concern was raised in the original peer review, and it remains unaddressed in the revised manuscript.

The use of RNaseHIII increases the specificity of the S9.6 antibody and improves confidence in the DNA-RNA hybrid data throughout the manuscript.

eLife assessment

This valuable paper examines the role of the WRNIP1 AAA+ ATPase in regulating R-loop formation, which induces a conflict with active replication forks and transcription. The authors provide convincing evidence to support a role of the ubiquitin-binding UBZ domain of WRNIP1 in R-loop suppression generated by this conflict. The work is of interest to researchers who work on genome stability/instability.

Reviewer #1 (Public Review):

This paper describes the role of WRNIP1 AAA+ ATPase, particularly its UBZ domain for ubiquitin-binding, but not ATPase, to prevent the formation of the R-loop when DNA replication is mildly perturbated. By combining cytological analysis for DNA damage, R-loop and chromosome aberration with the proximity ligation assay for colocalization of various proteins involved in DNA replication and transcription, the authors provide solid evidence to support the claim. The authors also revealed a distinct role of WRNIP1 in the prevention of R-loop-induced DNA damage from FANCD2, which is inconsistent with the known relationship between WRNIP1 and FANCD2 in the repair of crosslinks.

Reviewer #2 (Public Review):

This paper aims at establishing the role of WRN-interacting protein 1 (WRNIP1) and its UBZ domain (an N-terminal ubiquitin-binding zinc finger domain) on genome instability caused by mild inhibition of DNA synthesis by aphidicolin. The authors used human MRC5 fibroblasts investigated with standard methods in the field. The results clearly showed that WRNIP1 silencing and UBZ-mutation (D37A) increased DNA damage, chromosome aberrations, and transcription-replication conflicts caused by aphidicolin.

The conclusions of the paper are overall well supported by results, however, aspects of some data analyses would need to be clarified and/or extended.

(1) The methods (immunofluorescence microscopy and dot-blots) to determine R-loop levels can lack sensitivity and specificity. In particular, since the S9.6 antibody can bind to other structures besides heteroduplex, dot-blot analyses only grossly assess R-loop levels in cellular samples of purified nucleic acids, which are constituted by many different types of DNA/RNA structures.

(2) Experimental plan has analyzed the impact of WRNIP1 lack or mutations at steady-state conditions. Thus, the possible role of WRNIP1 at an early step of the mechanism would require some sort of kinetics analysis of the molecular process, therefore not at steady-state conditions. The findings of a co-localization of R-loops and WRNIP1 have been obtained with the S9.6 antibody, which recognizes DNA-RNA heteroduplexes. Since WRNIP1 is known to be recruited at stalled forks and DNA cleavage sites, it is not surprising that WRNIP1 is very close to heteroduplexes, abundant structures at replication forks and cleavage sites. Similar interpretations may also be valid for Rad51/S9.6 co-localization findings.

(3) Determination of DNA damage, chromosome aberration, and co-localization data are reported as means of measurements with appropriate statistics. However, the fold-change values relative to corresponding untreated samples are not reported. In some instances, it seems that WRNIP1 silencing or mutations actually reduce or do not affect aphidicolin effects. That leaves open the interpretation of specific results.

Author Response

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

Reviewer #1:

Summary:

In this manuscript, the authors used machine learning algorithm to analyze published exosome datasets to find biomarkers to differentiate exosomes of different origin.

Strengths:

The performance of the algorithm are generally of good quality.

Weaknesses:

The source datasets are heterogeneous as described in Figure 1 and Figure 2, or Line 72-75; and therefore questionable.

Response: We thank the reviewer for this assessment. The commonly used biomarkers of exosomes exhibit heterogeneous presence and abundance within the exosomes derived from different cell lines, tissue, and biological fluids. The primary goal of this study was to identify universal exosomal biomarkers that remain consistent across different sources of exosomes, unaffected by potential isolation and quantification bias. This objective was achieved through an integration of datasets from different sources, which allowed for the subsequent identification of common proteins associated with exosomes. Among the 18 protein markers identified, it is noteworthy that they are universally abundant in all cell lines and their exosomes. We believe that despite the heterogeneity of the datasets used here, the identification of 18 universal protein markers in exosomes from diverse sources is a strength of this analysis.

(1) Nomenclature: Extracellular vesicles (EVs) are small vesicles released by cells into the extracellular space, exhibiting high heterogeneity in origin across species. Exosomes are typically defined as being of multivesicular body origin. However, the absence of several crucial common exosomal markers, including CD63, suggests that the proteomics analysis may include various other vesicular and non-vesicular materials.

Response: As we reported previously (Kugeratski et al., Nature Cell Biology, 2021), the commonly used exosomal markers, such as CD9, CD63 and CD81 exhibit heterogeneity with respect to presence and abundance in the exosomes derived from different cell types. For example, CD63 demonstrated remarkably lower abundance in the exosomes derived from Raji cell lines. In our study, the detection rate of CD63 (< 50%) is quite low in the tissue-derived exosomes, which is consistent with the observations made in another proteomics based study (Hoshino et al., Cell, 2020). Therefore, relying solely on these markers is inadequate for the comprehensive characterization of EVs as exosomes. Therefore, we conducted this study to identify universal protein markers of exosomes by integrating data from multiple sources, thereby circumventing potential confounding effects due to their diverse origins and other technical differences.

(2) Line 90: IPA is not prior in the manuscript.

Response: We provided the full definition of IPA (Ingenuity Pathway Analysis) in the revised manuscript.

(3) Figure 2B: Considering the large number of variables, it is unsurprising that the 2D PCA (Principal Component Analysis) falls short in the classification task. Including a few additional dimensions (principal components) might have the potential to better distinguish the cancer groups from the control group.

Response: Thank you for this insightful query. The purpose of utilizing PCA here is to appreciate the heterogeneity associated with exosomes from different studies. While we acknowledge that additional dimensions may be more useful in distinguishing between cancer and control exosomes, we believe that derived performance will remain inferior to the machine learning approach we developed here.

(4) Figure 2D: Exosomes primarily derive from multivesicular bodies, rather than the plasma membrane. It remains unclear why the authors focus specifically on proteins in the plasma membrane. Is it intended to encompass all membrane proteins? Clarification is needed on this point.

Response: A good point. This study attempted to identify protein biomarkers of exosomes originating from different sources. Our approach involved considering proteins present on the plasma membrane as potential biomarkers also because many of them have been detected on the surface of exosomes.

(5) Figure 2F: The 18 identified proteins are also abundantly present in control cells, not solely in cancer-derived "exosomes." The statement in line 104 is misleading in this regard.

Response: We apologize for the misleading sentence. We have revised the statement to state that “In total, we identified a set of 18 exosome protein markers that are present at a higher abundance in all exosomes examined”.

(6) Figure 3B: Considering the definition of exosomes, CD63 and TSG101 should be present in all samples, and their absence raises concerns.

Response: We understand the concern of the reviewer. In this Figure, we analyzed CD63 and TSG101 in tissue-derived exosomes. Our results are consistent with the previous study also shows the paucity of these makers in the tissue-derived exosomes (Hoshino et al., Cell, 2020). Our study highlights that CD63 and TSG101 cannot always identify exosomes from diverse cell lines and tissues. Such initial observations motivated us to conduct this study to identify the universal biomarkers of exosomes across different sources.

(7) Figure 6G&H: Achieving an accuracy of 80% cannot be deemed "excellent."

Response: We employed the word “excellent” in line 225 to describe the sensitivity and specificity associated with AUROC.

(8) Other comments on methods: The manuscript lacks an explanation of the neural network structure and why it outperforms other methods. Additionally, details about the calculation of MI (mutual information), IPA, and other methods should be provided.

Response: This is a good suggestion but in this work we did not employ the neural networks for the analysis. We provided additional details and explanations regarding the methodology for mutual information score calculation, as well as insights into the improved use of IPA and other relevant methods in the revised manuscript.

Reviewer #2:

Summary:

This is a fine work on the development of computational approaches to detect cancer through exosomes. Exosomes are an emerging biomarker resource and have attracted considerable interests in the biomedical field. Kalluri and co-workers collected a large sample pool and used random forest to identify a group of protein markers that are universal to exosomes and to cancer exosomes. The results are very exciting and not only added new knowledge in cancer research but also a new and advanced method to detect cancer. Data was presented very nicely and the manuscript was well written.

Strengths:

Identified new biomarkers for cancer diagnosis via exosomes.

Developed a new method to detect cancer non-invasively.

Results were presented nicely and manuscript were well written.

Weaknesses:

N/A.

Response: We appreciate the the enthusiastic assessment of our study by the reviewer.

Reviewer #3:

In the current study, Li et al. address the difficulty in early non-invasive cancer diagnosis due to the limitations of current diagnostic methods in terms of sensitivity and specificity. The study brings attention to exosomes - membrane-bound nanovesicles secreted by cells, containing DNA, RNA, and proteins reflective of their originating cells. Given the prevalence of exosomes in various biological fluids, they offer potential as reliable biomarkers. Notably, the manuscript introduces a new computational approach, rooted in machine learning, to differentiate cancers by analyzing a set of proteins associated with exosomes. Utilizing exosome protein datasets from diverse sources, including cell lines, tissues, and various biological fluids, the study spotlights five proteins as predominant universal exosome biomarkers. Furthermore, it delineates three distinct panels of proteins that can discern cancer exosomes from non-cancerous ones and assist in cancer subtype classification using random forest models. Impressively, the models based on proteins from plasma, serum, or urine exosomes achieve AUROC scores above 0.91, outperforming other algorithms such as Support Vector Machine, K Nearest Neighbor Classifier, and Gaussian Naive Bayes. Overall, the study presents a promising protein biomarker signature tied to cancer exosomes and proposes a machine learning-driven diagnostic method that could potentially revolutionize non-invasive cancer diagnosis.

Response: We appreciate this positive assessment of our work.

(1) The authors should clarify why they focused solely on protein markers. Why weren't RNA transcripts also considered? Do the authors see value in incorporating RNA/micro RNA transcripts to enhance diagnostic capabilities?"

Response: This is a very important point for further consideration. The current datasets for exosomal proteins are extensive and generally proteins might offer distinct advantages in cancer diagnostics compared to nucleic acids due to their stability in exosomes and extended half-life (Schey et al., Methods, 2015). We do agree that the power of analysis can only get better if also add DNA, RNAs and other constituents and we hope to pursue such analysis in the future.

(2) Can the identified exosomal markers also be evaluated as prognostic indicators?

Response: We appreciate this intriguing question. Indeed, proteins such as apolipoprotein E (APOE) may serve as a potential prognostic marker in various cancers (Ren et al., Cancer Medicine, 2019). APOE is being extensively studied as a prognostic and diagnostic marker for multiple cancer types, including colorectal cancer (Martin et al., BMC Cancer, 2014), gastric cancer (Sakashita et al., Oncology Reports, 2008), pancreatic cancer (Chen et al., Medical Oncology, 2013; Xu et al., Tumor Biology, 2016), and human hepatocellular carcinoma (Yokoyama et al., International Journal of Oncology, 2006). In these studies, APOE levels were found to be elevated in the serum of cancer patients and correlated with survival outcomes.

(3) The discussion should emphasize if the identified protein markers are tumor-specific or if they indicate, for instance, the patient's immune reaction to the tumor.

Response: A good point. We believe that the identified biomarkers are tumor-specific and a significant number of these proteins have been previously associated with tumor initiation and progression. Further studies will likely identify immune response-related biomarkers when more in-depth tumor-level analyses are performed.

References:

Chen, J., Chen, L. J., Yang, R. B., Xia, Y. L., Zhou, H. C., Wu, W., Lu, Y., Hu, L. W., & Zhao, Y. (2013). Expression and clinical significance of apolipoprotein E in pancreatic ductal adenocarcinoma. Med Oncol, 30(2), 583. https://doi.org/10.1007/s12032-013-0583-y

Hoshino, A., Kim, H. S., Bojmar, L., Gyan, K. E., Cioffi, M., Hernandez, J., Zambirinis, C. P., Rodrigues, G., Molina, H., Heissel, S., Mark, M. T., Steiner, L., Benito-Martin, A., Lucotti, S., Di Giannatale, A., Offer, K., Nakajima, M., Williams, C., Nogues, L., . . . Lyden, D. (2020). Extracellular Vesicle and Particle Biomarkers Define Multiple Human Cancers. Cell, 182(4), 1044-1061 e1018. https://doi.org/10.1016/j.cell.2020.07.009

Kugeratski, F. G., Hodge, K., Lilla, S., McAndrews, K. M., Zhou, X., Hwang, R. F., Zanivan, S., & Kalluri, R. (2021). Quantitative proteomics identifies the core proteome of exosomes with syntenin-1 as the highest abundant protein and a putative universal biomarker. Nat Cell Biol, 23(6), 631-641. https://doi.org/10.1038/s41556-021-00693-y

Martin, P., Noonan, S., Mullen, M. P., Scaife, C., Tosetto, M., Nolan, B., Wynne, K., Hyland, J., Sheahan, K., Elia, G., O'Donoghue, D., Fennelly, D., & O'Sullivan, J. (2014). Predicting response to vascular endothelial growth factor inhibitor and chemotherapy in metastatic colorectal cancer. BMC Cancer, 14, 887. https://doi.org/10.1186/1471-2407-14-887

Ren, L., Yi, J., Li, W., Zheng, X., Liu, J., Wang, J., & Du, G. (2019). Apolipoproteins and cancer. Cancer Med, 8(16), 7032-7043. https://doi.org/10.1002/cam4.2587

Sakashita, K., Tanaka, F., Zhang, X., Mimori, K., Kamohara, Y., Inoue, H., Sawada, T., Hirakawa, K., & Mori, M. (2008). Clinical significance of ApoE expression in human gastric cancer. Oncol Rep, 20(6), 1313-1319. https://www.ncbi.nlm.nih.gov/pubmed/19020708

Schey, K. L., Luther, J. M., & Rose, K. L. (2015). Proteomics characterization of exosome cargo. Methods, 87, 75-82. https://doi.org/10.1016/j.ymeth.2015.03.018

Xu, X., Wan, J., Yuan, L., Ba, J., Feng, P., Long, W., Huang, H., Liu, P., Cai, Y., Liu, M., Luo, J., & Li, L. (2016). Serum levels of apolipoprotein E correlates with disease progression and poor prognosis in breast cancer. Tumour Biol. https://doi.org/10.1007/s13277-016-5453-8

Yokoyama, Y., Kuramitsu, Y., Takashima, M., Iizuka, N., Terai, S., Oka, M., Nakamura, K., Okita, K., & Sakaida, I. (2006). Protein level of apolipoprotein E increased in human hepatocellular carcinoma. Int J Oncol, 28(3), 625-631. https://www.ncbi.nlm.nih.gov/pubmed/16465366

eLife assessment

This important study introduces a novel AI method for the analysis of published data, with practical implications for early cancer diagnosis. The results are supported by compelling evidence.

Reviewer #1 (Public Review):

Summary:

In this manuscript, the authors used machine learning algorithm to analyze published exosome datasets to find biomarkers to differentiate exosomes of different origin. By applying the method to "exosomes" sample, the author discovered common exosome markers and cancer-type specific markers.

Strengths:

The performance of the algorithm are generally of good quality.

Reviewer #2 (Public Review):

Summary:

This is a fine work on the development of computational approaches to detect cancer through exosomes. Exosomes are an emerging biomarker resource and have attracted considerable interests in the biomedical field. Kalluri and co-workers collected a large sample pool and used random forest to identify a group of protein markers that are universal to exosomes and to cancer exosomes. The results are very exciting and not only added new knowledge in cancer research but also a new and advanced method to detect cancer. Data was presented very nicely and the manuscript was well written.

Strengths:

Identified new biomarkers for cancer diagnosis via exosomes.<br /> Developed a new method to detect cancer noninvasively.<br /> Results were presented nicely and manuscript were well written.

Author Response

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

eLife assessment

This valuable study reports comprehensive multi-omic data on the changes induced in young and aged male mouse tail fibroblasts after treatment with chemical reprogramming factors. The authors claim that chemical reprogramming factors induce changes consistent with a reduction of cellular 'biological' age (e.g., correlations with established aging markers in whole tissues). However, the study relies on previously identified aging markers (instead of aging in the tail fibroblast system itself), and thus, at this stage, the evidence in support of the observed molecular changes truly reflecting changes in biological age in the study system is still incomplete.

Essential revisions

After discussion with reviewers, we believe that the conclusions of the manuscript would be significantly strengthened with the following revisions:

(1) Rather than basing the analysis of age-related markers on public tissue data, it is recommended that authors use their own data on pre-reprogramming fibroblasts to define molecular aging-related markers/signatures specifically for male tail fibroblasts at 4 vs 20 months. This should also always be included in figures as reference points.

We appreciate these helpful comments. Please refer to our responses to Reviewers #1 and #2 concerning these suggestions and the corresponding changes we have made in the revised manuscript.

(2) In general, the methods as written lack the details necessary to fully understand the study/reproduce it independently, notably in terms of data analysis choices (e.g. use of FWER/FDR type correction for multiple testing, use of raw vs normalized RNA counts for PCA, etc).

Thank you for this feedback. We have modified our text to address this issue. Please refer to our responses to Reviewer #1 for the specific changes we have made.

(3) More generally, the authors should better outline the limitations/caveats of their experimental design in the discussion and/or abstract, including the specific cell type and the choice of using only male data (since aging itself is very sex-dimorphic, and the impact of partial reprogramming on aging phenotypes may also be sex-dimorphic).

Thank you for this important feedback. We have now added a section to our Discussion in which we directly address potential limitations of our study concerning sex-specific differences and the cell type used.

Public Reviews:

Reviewer #1:

Summary:

The investigators employed multi-omics approach to show the functional impact of partial chemical reprogramming in fibroblasts from young and aged mice.

Strengths:

Multi-omics data was collected, including epigenome, transcriptome, proteome, phosphoproteome, and metabolome. Different analyses were conducted accordingly, including differential expression analysis, gene set enrichment analysis, transcriptomic and epigenetic clock-based analyses. The impact of partial chemical reprogramming on aging was supported by these multi-source results.

We appreciate the reviewer noting the strength and comprehensiveness of our approach.

Weaknesses:

More experimental data may be needed to further validate current findings.

We thank the reviewer for this suggestion. To further validate our findings, we have proceeded as follows: (1) First, we have investigated the role of Prkaca activation during partial chemical reprogramming with 7c (see updated Fig. 5C, Fig. 5 – figure supplement 1B). By confocal microscopy, we show that partial chemical reprogramming with 7c does not cause Prkaca to localize to mitochondria; rather, its cellular distribution is altered to favor nuclear localization. We also use RNAi to knockdown Prkaca and find that Prkaca is not necessary for mediating the increase in mitochondrial membrane potential upon partial chemical reprogramming with 7c.

(2) We have determined the effect of partial chemical reprogramming with 7c on apoptosis using Annexin V assay (see updated Fig. 5 – figure supplement 1C). We show that during the course of partial chemical reprogramming, the proportion of apoptotic cells steadily increases to about 20 percent.

(3) We have re-analyzed our multi-omics data to determine the molecular differences (e.g. at the epigenome, transcriptome, proteome, and metabolome levels) between fibroblasts isolated from young and old mice (see updated Fig. 2 – figure supplement 1, Fig. 6 – figure supplement 1, and Fig. 7 – figure supplement 2). Additionally, we have updated Fig. 7A to include statistical comparisons of transcriptomic age of 4-month-old and 20-month-old fibroblasts. Finally, we have updated Fig. 3D to include functional enrichment of gene and protein expression levels of aged fibroblasts.

(4) We have more thoroughly characterized the effects of partial chemical reprogramming on the epigenome (see Fig. 7 – figure supplement 3).

(5) Julie Y. Chen was added on as an additional co-author for producing the analyses shown in Fig. 7 – figure supplement 2, and Fig. 7 – figure supplement 3.

Reviewer #2:

The short-term administration of reprogramming factors to partially reprogram cells has gained traction in recent years as a potential strategy to reverse aging in cells and organisms. Early studies used Yamanaka factors in transgenic mice to reverse aging phenotypes, but chemical cocktails could present a more feasible approach for in vivo delivery. In this study, Mitchell et al sought to determine the effects that short-term administration of chemical reprogramming cocktails have on biological age and function. To address this question, they treated young and old mouse fibroblasts with chemical reprogramming cocktails and performed transcriptome, proteome, metabolome, and DNA methylation profiling pre- and post-treatment. For each of these datasets, they identified changes associated with treatment, showing downregulation of some previously identified molecular signatures of aging in both young and old cells. From these data, the authors conclude that partial chemical reprogramming can rejuvenate both young and old fibroblasts.

The main strength of this study is the comprehensive profiling of cells pre- and post-treatment with the reprogramming cocktails, which will be a valuable resource for better understanding the molecular changes induced by chemical reprogramming. The authors highlighted consistent changes across the different datasets that are thought to be associated with aging phenotypes, showing reduction of age-associated signatures previously identified in various tissues. However, from the findings, it remains unclear which changes are functionally relevant in the specific fibroblast system being used. Specifically:

(1) The 4 month and 20 month mouse fibroblasts are designated "young" vs "old" in this study. An important analysis that was not shown for each of the profiled modalities was a comparison of untreated young vs old fibroblasts to determine age-associated molecular changes in this specific model of aging. Then, rather than using aging signatures defined in other tissues, it would be more appropriate to determine whether the chemical cocktails reverted old fibroblasts to a younger state based on the age-associated changes identified in this comparison.

In our study, we have used 4 biological samples per group for young and old untreated fibroblasts, and these samples have been used to calculate the effect of 7c and 2c cocktails on gene expression in each age group. Therefore, the correlation between logFC induced by 7c/2c treatment and logFC between young and old fibroblasts would be biased, since the same untreated samples would be used in both calculations: estimates B-A and C-B will be, on average, negatively correlated even if A, B and C are independent random variables. For this reason, to investigate the effect of cocktails on biological age, we utilized gene expression signatures of aging, estimated based on more than 2,600 samples of different ages from 25 data sources (PMID: 37269831). Notably, our multi-tissue signatures of aging were identified based on data from 17 tissues, including skin. Therefore, these biomarkers seem to represent more reliable and universal molecular mechanisms of aging. Since they have been identified using independent data, the signatures also don’t introduce the statistical bias described above. For these reasons, we think that they are more applicable for the current analysis. To demonstrate that the utilized aging signatures are overall consistent with the changes observed in studied fibroblasts, we performed GSEA-based analysis, testing association between logFC in aged fibroblasts and various signatures of aging and reprogramming (similar to our analysis in Fig. 2E). We found that the changes in aged fibroblasts from the current study demonstrated positive association with the majority of aging signatures (kidney, liver and multi-tissue signatures in mouse and rat) (Fig. 2 – figure supplement 1A) and were negatively associated with signatures of reprogramming. In addition, we characterized functional changes perturbed in untreated aged fibroblasts at the level of gene expression and protein concentrations and observed multiple changes consistent with the aging signatures, such as upregulation of genes and proteins involved in inflammatory response and interferon signaling (Fig. 3D, Fig. 2 – figure supplement 1C). Therefore, changes observed in untreated aged fibroblasts seem to agree with age-related molecular changes identified across mammalian tissues in our previous studies.

We would also like to mention that the epigenetic clocks used in this study consistently show that the fibroblasts from 20-month-old fibroblasts are significantly older than the fibroblasts from 4-month-old mice (Fig. 7B). Moreover, we have revised the manuscript to show that these epigenetic differences between young and old untreated fibroblasts are not due to overall changes in mean DNA methylation (Fig. 7 – figure supplement 2). In contrast, in the revised manuscript, we observe that 7c treatment is reducing the epigenetic age of cells by decreasing mean DNA methylation levels (Fig. 7 – figure supplement 3).

(2) Across all datasets, it appears that the global profiles of young vs old mouse fibroblasts are fairly similar compared to treated fibroblasts, suggesting that the chemical cocktails are not reverting the fibroblasts to a younger state but instead driving them to a different cell state. Similarly, in most cases where specific age-related processes/genes are being compared across untreated and treated samples, no significant differences are observed between young and old fibroblasts.

We agree that our data shows that partial chemical reprogramming seems to induce a similar effect on young and old fibroblasts. In Fig. 2 – figure supplement 1B, the Spearman correlation coefficients for the effects on gene expression in young and old fibroblasts are 0.80 and 0.85 for 2c and 7c, respectively. It is important to note that the effect of partial chemical reprogramming is a magnitude higher (say in terms of number of differentially expressed genes) than the effect of aging in the untreated fibroblasts. Partial chemical reprogramming with 7c, we believe, is pushing the cells to a younger state as a byproduct of producing a different cellular metabolic state with a strong increase in OXPHOS capacity.

(3) Functional validation experiments to confirm that specific changes observed after partial reprogramming are indeed reducing biological age is limited.

Functional validation of rejuvenating interventions is limited in vitro, as cells do not completely maintain their “aged” phenotype once isolated and cultured, and pursuing partial chemical reprogramming in vivo in naturally-aged mice was beyond the scope of the study. One of the best reporters of biological age that are preserved in primary cells in vitro are epigenetic and transcriptomic clocks, which were both utilized in this manuscript to show that 7c treatment, but not 2c, reduces biological age. We show that splicing-related damage is marginally elevated in old fibroblasts compared to young, and that 7c reduces splicing damage by reducing intron retention. Moreover, the epigenetic clocks used in this study show that the 20-month-old fibroblasts are significantly older than the 4-month-old fibroblasts, indicating that the “aged” phenotype is at least partially preserved. Furthermore, according to previous studies (PMIDs: 37269831, 31353263), one of the strongest functional biomarkers of aging is downregulation of mitochondrial function and energy metabolism, including oxidative phosphorylation, while upregulation of these functions is usually associated with extended lifespan in mice. For this reason, we have focused on these pathways in our study and assessed them with functional assays.

(4) Partial reprogramming appears to substantially reduce biological age of the young (4 month) fibroblasts based on the aging signatures used. It is unclear how this result should be interpreted.

This is a caveat of all reprogramming strategies/”anti-aging” interventions developed and tested to date. Currently, there are no genetic or pharmacological methods that target only the “aged” state and not the “young” state as well (i.e. an intervention that would only cause a change in old cells and revert them to a younger state). However, “young” cells in our study and many other studies are still the cells of an intermediate age, as aging appears to begin early during development. Therefore, perhaps unsurprisingly, partial chemical reprogramming seemed to have similar effects on fibroblasts isolated from young and old mice, which is in line with OSK/OSKM reprogramming. These results should be interpreted as follows: partial chemical reprogramming does not depend on the epigenetic state (biological age) of adult cells to induce rejuvenation. We have updated the discussion section of our manuscript accordingly.

Recommendations for the authors:

Reviewer #1:

(1) How was the PCA conducted for RNA-seq data? Were the raw or normalized counts used for PCA?

Normalized counts were used for PCA of the RNA-seq data.

(2) Supplementary Fig 3c, why was the correlation between the red rows and red columns low? Was the color of group messed up? Why was the Pearson correlation used instead of Spearman correlation? Most of the correlation analyses in the manuscript used Spearman correlation.

We thank the reviewer for noticing this mistake. The colors of the groups have now been corrected. Furthermore, to be consistent with the rest of the manuscript, we have performed a Spearman correlation analysis on the normalized proteomics data to evaluate sample-to-sample similarities and updated Fig. 3 – figure supplement 1 accordingly. Overall, the results are similar to those obtained by Pearson correlation.

(3) Were the significant metabolites tested by one-way ANOVA adjusted for family-wise type I error rate? It is surprising that over 50% metabolites were significant.

Yes, the significant metabolites were adjusted for family-wise type I error rate (with a 5% significance threshold) in Fig. 6B.

(4) Missing full names of several abbreviations, such as NIA, RLE, PSI, etc.

Thank you for noticing the missing abbreviations. We have corrected this by writing out the full term in the first instance in which each abbreviation appears.

(5) Methods section may be too long. Some paragraphs could be moved to supplementary text.

eLife does not have a limit to the number of figures or amount of text. Therefore, we have kept the methods section largely unaltered as we feel that they would be helpful to the scientific community.

Reviewer #2:

(1) As discussed in the public review, I would recommend first establishing what differences exist between 4 month and 20 month fibroblasts to identify potential age-related changes in these fibroblasts.

We thank the reviewer for this suggestion. We have now thoroughly characterized the molecular differences between fibroblasts taken from young and old mice at the epigenome, transcriptome, proteome, and metabolome levels. Please refer to previous responses for more specific details.

We have also attempted to establish aging-related differences at the phosphoproteome level, particularly in regards to mitochondrial processes (see figure below), but only GOcc: mitochondrion and GObp: mitochondrial transport come close to being statistically significant (raw p-values of 0.05 and 0.08, respectively) in the control comparison.

Author response image 1.

(2) While the global changes currently highlighted in the study are informative and should remain in the revised manuscript, additional analyses to show which age-related changes identified in point 1 are reverted upon 2c or 7c treatment would better address the question of whether these cocktails revert age-related changes seen in fibroblasts. These analyses should be performed for each dataset (i.e transcriptomic, proteomic, epigenomic, metabolomic) generated.

Thank you for this comment. We have now evaluated the effects of partial chemical reprogramming on the specific molecular differences between fibroblasts isolated from young and old mice (see updated Fig. 2 – figure supplement 1, Fig. 6 – figure supplement 1, Fig. 7 – figure supplement 2, and Fig. 7 – figure supplement 3). For functional enrichment of aged fibroblasts at the gene and protein level, please refer to updated Fig. 3D.

(3) Comparisons between partial reprogramming and OSKM reprogramming signatures are repeatedly made in the paper, but it is not clear from the text whether similarity to OSKM reprogramming signatures is a desired or undesired feature. Since there are likely both rejuvenating and oncogenic aspects of the OSKM signatures, it is unclear what conclusions can be made from these comparisons.

Two central questions of this study were (1) if partial chemical reprogramming could induce cellular rejuvenation, and (2) if so, would it do so by merely chemically activating expression of Yamanaka factors. In this study, we find that 7c, the cocktail that demonstrated the most profound effect on biological age, only minorly upregulates Klf4, downregulates c-Myc, and has no effect on Sox2 or Oct4 expression. Thus, partial chemical reprogramming seems to operate through a mechanism independent of upregulating OSK/OSKM gene expression. This is crucial as it suggests that there are other transcription factors outside of OSKM that can be targeted to induce cellular rejuvenation and reversal of biological age. However, the direct transcriptional targets of partial chemical reprogramming are currently unknown and require further investigation.

Partial reprogramming with OSK/OSKM has several limitations, including low efficiency, oncogenic risk, and differences in the speed of reprogramming according to cell/tissue type. These risks could be inherently tied to the transcription factors OSKM themselves; thus, partial chemical reprogramming, by avoiding strong activation of these genes, could potentially avoid these risks and provide a safer means for reversing biological age in vivo. However, extensive follow-up studies beyond the scope of this manuscript are certainly required to determine this.

We have addressed this comment by modifying the discussion to include these points.

(4) When analyzing the phospho-proteomics data, results are discussed as general changes in phosphorylation of proteins involved in different cellular processes. However, phosphorylation can either activate or inhibit a specific protein, and can depend on the specific residue in a protein that is modified. Different proteins in a cellular process can also respond in opposite directions to phosphorylation. Treating activating and inactivating phosphorylation events separately in describing these results would be more informative.

We agree that an analysis that considers for each specific phosphosite whether it activates or inactivates a particular pathway would in principle be preferable over our current enrichment analysis that only accounts for the increase or decrease in phosphorylation of each site without knowing its biological meaning. However, unfortunately, we think it is currently practically not possible to conduct such an analysis. The proposed analysis would require a database with information on which residues are (de-)phosphorylated when a certain pathway is activated. However, as far as we know, there are currently no databases that link activation or inactivation of specific phosphosites to pathways in repositories like KEGG, HALLMARK, GObp, GOcc, GOmf, Reactome, etc.

Some databases link phosphosites to drugs, diseases and kinases (e.g. PTMsigDB (PMID: 30563849)). However, these authors explicitly state: “We note that we do not capture functional annotations of PTM sites in PTMsigDB, such as activating or inactivating effect on the modified protein.” Furthermore, even in these databases, for the vast majority of the registered phosphosites, the responsible kinases are unknown, especially in mice. In our work, we made use of PhosphoSitePlus for kinase substrate enrichment analysis (see Fig. 5B). Such analyses, where kinase activity is inferred based on activated phosphosites are indeed commonly performed (see PMIDs: 34663829, 37269289, 37585503).

In the absence of a repository that assigns activity to phosphosites, if enrichment analysis is being done for biological pathways, it is standard practice to so without accounting for whether phosphosites are activating or inactivating (see PMID: 34663829), as we have done in our manuscript (Fig. 5A).

Despite the drawbacks, we believe our analysis is relevant, as it demonstrates important biological activity in these pathways uopn 2c/7c treatments as compared to controls. For example, the observed increase in abundance in mitochondrial OXPHOS complexes (Fig. 3E) combined with an increase in general phosphorylation of mitochondrial proteins (Fig. 5A) likely points to an increase mitochondrial activity, although one cannot exclude that some individual phosphorylation events might have inhibitory effects on certain mitochondrial proteins, while others might indicate increases in activity.

(5) For the transcriptomic and epigenetic aging clocks used in Fig 7, significance tests need to be included for untreated 4 month vs 20 month fibroblasts. Particularly for the transcriptional clock, the differences are small and suggest that it may not be a strong aging signature.

We have updated our clock analysis with the most recent versions of the clocks and added statistical significance between 4-month-old and 20-month-old untreated fibroblasts there (Fig. 7A). The difference is statistically significant for the chronological clock. However, when the lifespan-adjusted clock was applied, no statistical significance was observed, suggesting that 20-month-old fibroblasts do not exhibit substantial changes in gene expression associated with decreased healthspan and increased mortality.

(6) For heatmaps shown in Figure 3D and Figure 4, please include untreated 4 month and 20 month fibroblasts as well to determine if pathways being compared are different between young and old fibroblasts.

We have updated Figure 3D with functional enrichment results for aged fibroblasts at gene and protein expression levels, as requested. As for Fig. 4, we explained in our reply to point 1 of Reviewer #2 in the public review why addition of aged fibroblasts there would be biased there. Instead, we have performed GSEA-based association analysis for changes observed in aged fibroblasts and signatures of aging (Fig. 2 – figure supplement 1), confirming that our signatures are overall consistent with patterns of 20-month-old fibroblasts from the current study.

eLife assessment

This important study reports comprehensive multi-omics data on the changes induced in young and aged male mouse tail fibroblasts after treatment with chemical reprogramming factors. The authors provide solid evidence to support their claim that chemical reprogramming factors induce changes consistent with a reduction of cellular 'biological' age (e.g., correlations with established aging markers in whole tissues).

Reviewer #1 (Public Review):

Summary:

The investigators employed multi-omics approach to show the functional impact of partial chemical reprogramming in fibroblasts from young and aged mice.

Strengths:

Multi-omics data was collected, including epigenome, transcriptome, proteome, phosphoproteome, and metabolome. Different analyses were conducted accordingly, including differential expression analysis, gene set enrichment analysis, transcriptomic and epigenetic clock-based analyses. The impact of partial chemical reprogramming on aging was supported by these multi-source results.

Reviewer #2 (Public Review):

The short-term administration of reprogramming factors to partially reprogram cells has gained traction in recent years as a potential strategy to reverse aging in cells and organisms. Early studies used Yamanaka factors in transgenic mice to reverse aging phenotypes, but chemical cocktails could present a more feasible approach for in vivo delivery. In this study, Mitchell et al sought to determine the effects that short-term administration of chemical reprogramming cocktails have on biological age and function. To address this question, they treated young and old mouse fibroblasts with chemical reprogramming cocktails and performed transcriptome, proteome, metabolome, and DNA methylation profiling pre- and post-treatment. For each of these datasets, they identified changes associated with treatment, showing downregulation of some previously identified molecular signatures of aging in both young and old cells. From these data, the authors conclude that partial chemical reprogramming can rejuvenate both young and old fibroblasts.

The main strength of this study is the comprehensive profiling of cells pre- and post-treatment with the reprogramming cocktails, which will be a valuable resource for better understanding the molecular changes induced by chemical reprogramming. The authors highlighted consistent changes across the different datasets that are thought to be associated with aging phenotypes, showing reduction of age-associated signatures previously identified in various tissues.

Author Response

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

Reviewer #1 (Public Review):

Summary

The authors investigated the antigenic diversity of recent (2009- 2017) A/H3N2 influenza neuraminidases (NAs), the second major antigenic protein after haemagglutinin. They used 27 viruses and 43 ferret sera and performed NA inhibition. This work was supported by a subset of mouse sera. Clustering analysis determined 4 antigenic clusters, mostly in concordance with the genetic groupings. Association analysis was used to estimate important amino acid positions, which were shown to be more likely close to the catalytic site. Antigenic distances were calculated and a random forest model was used to determine potential important sites.

This has the potential to be a very interesting piece of work. At present, there are inconsistencies in the methods, results and presentation that limit its impact. In particular, there are weaknesses in some of the computational work.

Strengths

(1) The data cover recent NA evolution and a substantial number (43) of ferret (and mouse) sera were generated and titrated against 27 viruses. This is laborious experimental work and is the largest publicly available neuraminidase inhibition dataset that I am aware of. As such, it will prove a useful resource for the influenza community.

(2) A variety of computational methods were used to analyse the data, which give a rounded picture of the antigenic and genetic relationships and link between sequence, structure and phenotype.

Weaknesses

(1) Inconsistency in experimental methods

Two ferret sera were boosted with H1N2, while recombinant NA protein for the others. This, and the underlying reason, are clearly explained in the manuscript. The authors note that boosting with live virus did not increase titres. Nevertheless, these results are included in the analysis when it would be better to exclude them (Figure 2 shows much lower titres to their own group than other sera).

As an exercise, we have excluded the H1N2 boosted ferrets sera and no major impact was observed in the antigenic grouping (see Author response image 1a). Another way to control for differences in immunogenicity is to normalize the NAI values with the homologous ELISA titers for each antigen. Clustering based on these ELISA normalized NAI titers reveals the same 4 distinct antigenic groups but with one change: Kan17 is shifted from group 1 to group 2 (Author response image 1b). Note that a homologous ELISA titer is not available for A/West-Virginia/17/2012 and thus this serum sample is not included in Author response image 1b.

Author response image 1.

Antigenic and phylogenetic relatedness of N2 NAs. Phylogenetic tree based on the N2 NA head domain amino acid sequences and heat-map representing the average of normalized neuraminidase inhibition titer per H6N2 [log2 (max NAI/NAI)] determined in ferret sera after the boost (listed vertically). The red-to-blue scale indicates high-to-low NAI observed in ELLA against the H6N2 reassortants (listed at the bottom). UPGMA clustering of H6N2s inhibition profiles are shown on top of the heat map and colored according to the phylogenetic groups.(a) Based on the ferret sera with exclusion of the sera that were obtained following prime-boost by infection with H1N2 (A/Estonia/91625/2015 and A/Stockholm/15/2014). (b) Based on serum NAI titers that were normalized by the homologous ELISA titer.

(2) Inconsistency in experimental results

Clustering of the NA inhibition results identifies three viruses which do not cluster with their phylogenetic group. Again, this is clearly pointed out in the paper. Further investigation of this inconsistency is required to determine whether this has a genetic basis or is an experimental issue. It is difficult to trust the remaining data while this issue is unresolved.

We understand the concern of the reviewer. It is important to keep in mind that discrete grouping of antigens allows to visualize major antigenic drifts. However, within closely related groups the cross reactivity of antisera is more likely distributed in a spectrum. When we constructed an antigenic map based on the antigenic cartography algorithm (as described by Smith D. et al, 2004), Kansas17, Wis15, and Ala15 are positioned more closely to antigenic group 1 than the majority of other antigens that were classified as group 2 (Author response image 2a). Similar results were obtained when individual ferret sera from the biological duplicates were used (Author response image 2b). This antigenic cartography map is now added as Figure 2. Figure supplement 3 to the revised manuscript.

Author response image 2.

The antigenic cartography was constructed using averaged data from pairs of ferrets (a). Similar analysis was performed on individual ferrets sera (b).

(3) Inconsistency in group labelling

A/Hatay/4990/2016 & A/New Caledonia/23/2016 are in phylogenetic group 1 in Figure 2 and phylogenetic group 1 in Figure 5 - figure supplement 1 panel a.

Our apologies: there was indeed a mistake in labeling of Figure 5. A new antigenic cartography was constructed and included in the revised manuscript. As a result Figure 5 - figure supplement has now become redundant and was removed from the manuscript.

A/Kansas/14/2017 is selected as a representative of antigenic group 2, when in Figure 2 it is labelled as AC1 (although Figure 2 - supplement 4 which the text is referring to shows data for A/Singapore/Infimh-16-0019/2016 as the representative of AC2). A/Kansas/14/2017 is coloured and labelled as AC2 in Figure 2 - supplement 5.

Thank you for pointing out this inconsistency. Kan17 clustered antigenically in group 1 based on the NAI values that were normalized relative to the serum with the maximal NAI value against the H6N2 virus that was tested. When using NAI titers that are normalization with the homologous ELISA titer, Kan17 is positioned in group 2. Likewise, antigenic cartography mapping positions Kan17 in group 2. Therefore, we conclude that A/Kansas/14/2017 NA is a representative of group 2.

The colouring is changed for Figure 3a at the bottom. A/Heilongjiang-Xiangyang/1134/2011 is coloured the same as AC4 viruses when it is AC1 in Figure 2. This lack of consistency makes the figures misleading.

We apologize for this mistake. The coloring in Figure 3a has been corrected.

(4) Data not presented, without explanation

The paper states that 44 sera and 27 H6N2 viruses were used (line 158). However, the results for the Kansas/14/2017 sera do not appear to be presented in any of the figures (e.g. Figure 2 phylogenetic tree, Figure 5 - figure supplement 1). It is not obvious why these data were not presented. The exclusion of this serum could affect the results as often the homologous titre is the highest and several heatmaps show the fold down from the highest titre.

Serum against A/Kansas/14/2017 was not prepared. For that reason, it is not included in the analysis. We agree that such homologous serum ideally should have been included and in the NAI assay would have resulted in a high if not the highest titer. However, we noticed that homologous sera did not always have the highest titers, especially in panels like ours were some antigens are closely related. The highest titer obtained against Kan17 H6N2 was from A/Bris/16 sera: 1/104, a titer that is in the range of other, homologous titers observed in the panel (Table S3). The Bris16 and Kan17 NAs have five amino acid differences. In summary, inclusion of Kan17 homologous sera would likely not impact the analysis and interpretation of the results because there are multiple highly cross-inhibiting heterologous serum samples against Kan17.

(5) The cMDS plot does not have sufficient quality assurance A cMDS plot is shown in Figure 5 - figure supplement 1, generated using classical MDS. The following support for the appropriateness of this visualisation is not given. a. Goodness of fit of the cMDS projection, including per point and per titre. b. Testing of the appropriate number of dimensions (the two sera from phylogenetic group 3 are clustered with phylogenetic group 2; additional dimensions might separate these groups). c. A measure of uncertainty in positioning, e.g. bootstrapping. d. A sensitivity analysis of the assumption about titres below the level of detection (i.e. that <20 = 10). Without this information, it is difficult to judge if the projection is reliable.

We agree with these comments. We have removed Figure 5 – figure supplement 1, and added new figure 2 – figure supplement 3 (antigenic cartography) instead.

(6) Choice of antigenic distance measure

The measure of antigenic distance used here is the average difference between titres for two sera. This is dependent on which viruses have been included in the analysis and will be biased by the unbalanced number of viruses in the different clusters (12, 8, 2, 5).

To verify the impact of the number of antigens on our analysis, the matrix of differences was generated with only 4 H6N2s representing at least one phylogenetic group (Per09, Sin16, Hel823 and Ind11) (Author response image 3a). This matrix is very similar to the one calculated based on all 27 antigens (Author response image 3b). The obtained matrix (Author response image 3a) was used in random forest to model antigenic distances and the result of prediction was plotted against real differences calculated based on the full data. The correlation coefficient (R2) of predicted vs observed values dropped from 0.81 to 0.71, suggesting that the number of antigens tested does not drastically affect the antigenic differences calculated based on serum values (Author response image 3e). Importantly, amino acid substitutions potentially associated with increased antigenic distances are similarly identified (Author response image 3c, d and f).

Author response image 3.

Matrix of differences was calculated using only 4 H6N2 antigens (a) or the full panel (b). The matrixes from (c) 4 or (d) 27 antigens were used in random forest modeling to estimate the impact of amino acid changes, respectively. The rf modeling data generated from 4 H6N2 only was plotted and correlated with values calculated from the full panel of 27 H6N2s (e). The multi-way importance plot indicates in red that 7 out of the 10 most important substitutions were identified by the analysis using only 4 H6N2s (f).

Interestingly, when matrix of differences is calculated using only 4 H6N2s data but not including at least one representative of antigenic group 1 and 2, the correlation coefficient between the predicted values and values obtained from the full panel is dramatically impacted (R2 values drops from 0.81 to 0.5 and 0.57. It is important to note that most of the sera also belong to phylogenetic antigens from groups 1 and 2. As a consequence, poorer prediction of those antigens would more drastically impact the correlation. No drastic drop was observed when representative H6N2s from group 3 or 4 were excluded from the data (from 0.81 to 0.75 and 0.73, Author response image 4 c and d).

Author response image 4.

Random forest analysis was repeated using only 4 antigens, but excluding representatives of one of the phylogenetic groups (a) no group 1, (b) no group 2, (c) no group 3, and (d) no group 4.

We also used Euclidean distances as a measure of differences (Author response image 5). The predictive values obtained in rf have a slightly reduced R2 compared to the values obtained using average of differences.

In conclusion the unbalanced number of antigens used per group and metric of distance does not seem to impact per se our analysis.

Author response image 5.

Antigenic distances were calculated using Euclidian distances of sera to sera. Those antigenic distances were used in rf for estimation of antigenic distance and importance of each amino acid substitution.

(7) Association analysis does not account for correlations

For each H6N2 virus and position, significance was calculated by comparing the titres between sera that did or did not have a change at that position. This does not take into account the correlations between positions. For haemagglutinin, it can be impossible to determine the true antigenic effects of such correlated substitutions with mutagenesis studies.

Most of the potential correlated effects cannot be addressed with the panel of N2s, except for combinations of substitution that are included in the panel, such as 245/247 with or without 468. Only mutagenesis studies would shed light on the epistatic effects. However, it is important to keep in mind that those individual substitutions in such kind of study likely do not reflect natural evolution of N2 (cfr. the importance of the NA charge balance (Wang et al., 2021: 10.7554/eLife.72516).

(8) Random forest method

25 features are used to classify 43 sera, which seems high (p/3 is typical for classification). By only considering mismatches, rather than the specific amino acid changes, some signals may be lost (for example, at a given position, one amino acid change might be neutral while another has a large antigenic effect). Features may be highly, or perfectly correlated, which will give them a lower reported importance and skew the results.

The number of features were optimized in the range from 5 to 80, with 25 being optimal (best R-value in predicted vs observed antigenic distances). Those features refer to the number of amino acid substitutions used in each tree. The number of trees was also optimized in the range of 100 to 2000.

In random forest the matrix of differences is made considering only position based and not the type of substitution in pairs of NA. Indeed, substitutions with distinct effects may skew results by indicating lower reported importance.

We have highlighted such potential bias in our discussion:

“Also, our modelling does not consider that substitution by other amino acids can have a distinct impact on the antigenic distance. As a consequence, predictions based on the model could underestimate or overestimate the importance of a particular amino acid residue substitution in some cases.”

Reviewer #2 (Public Review):

Summary:

The authors characterized the antigenicity of N2 protein of 44 selected A(H3N2) influenza A viruses isolated from 2009-2017 using ferret and mice immune sera. Four antigenic groups were identified, which correlated with their respective phylogenic/ genetic groups. Among 102 amino acids differed by the 44 selected N2 proteins, the authors identified residues that differentiate the antigenicity of the four groups and constructed a machine-learning model that provides antigenic distance estimation. Three recent A(H3N2) vaccine strains were tested in the model but there was no experimental data to confirm the model prediction results.

Strengths:

This study used N2 protein of 44 selected A(H3N2) influenza A viruses isolated from 2009-2017 and generated corresponding panels of ferret and mouse sera to react with the selected strains. The amount of experimental data for N2 antigenicity characterization is large enough for model building.

Weaknesses:

The main weakness is that the strategy of selecting 44 A(H3N2) viruses from 2009-2017 was not explained. It is not clear if they represent the overall genetic diversity of human A(H3N2) viruses circulating during this time. A comprehensive N2 phylogenetic tree of human A(H3N2) viruses from 2009-2017, with the selected 44 strains labeled in the tree, would be helpful to assess the representativeness of the strains included in the study.

The selection of antigens was performed using the method described by Bien and Tibshirani 2011 (doi: 10.1198/jasa.2011.tm10183). This method calculates MinMax distances to identify a central representative among distinct clusters.

To facilitate visualization of in a phylogenetic tree, only 180 representative N2 proteins from 2009-2017 were randomly selected (20 strains per year, unlabelled). Those 180 representatives and 44 readout panel strains (labelled) are shown in the phylogenetic tree below. Readout strains cover the major branches of the tree. The tree has been built using PhyML 3.0 using JTT substitution model and default parameters (Guindon S. et al, Systematic Biology 59(3):307-21, 2010) and visualized using ETE3 (Huerta-Cepas J. et al, Mol. Biol. Evol 33(6):1635-38, 2016).

Author response image 6.

The second weakness is the use of double-immune ferret sera (post-infection plus immunization with recombinant NA protein) or mouse sera (immunized twice with recombinant NA protein) to characterize the antigenicity of the selected A(H3N2) viruses. Conventionally, NA antigenicity is characterized using ferret sera after a single infection. Repeated influenza exposure in ferrets has been shown to enhance antibody binding affinity and may affect the cross-reactivity to heterologous strains (PMID: 29672713). The increased cross-reactivity is supported by the NAI titers shown in Table S3, as many of the double immune ferret sera showed the highest reactivity not against its own homologous virus but to heterologous strains. Although the authors used the post-infection ferret sera to characterize 5 viruses (Figure 2, Figure Supplement 4), the patterns did not correlate well. If the authors repeat the NA antigenic analysis using the post-infection ferret sera with lower cross-reactivity, will the authors be able to identify more antigenic groups instead of 4 groups?

This is a very valuable remark. In their paper, Kosikova et al. (CID 2018) report that repeated infection of ferrets with antigenically slightly different H3N2 viruses results in a broader anti-HA response, compared to a prime infection of an influenza naïve ferret, which results in a narrower anti-HA response. In our ferret immunizations the boost was performed with recombinant, enzymatically active NA that was homologous to the NA of the H1N2 virus that was used for the priming by infection. We determined the NAI responses in sera from ferrets after H1N2 infection against 5 different H6N2 viruses (Figure 2 – figure supplement 5). Compared to NAI responses in sera from H1N2 infected and subsequently NA protein boosted ferrets, the NAI titers obtained after a single infection were considerably lower. Although the normalized NAI titers of day 14 and day 42 sera correlated well, we cannot exclude a degree of broadening of the NAI response in the NA protein boost sera (Author response image 7). On the other hand, repeated influenza antigen exposure is the reality for the majority of people.

Author response image 7.

Correlation obtained on NAI data from ferrets at day 14 after infection vs data from day 42 after boost.

Another weakness is that the authors used the newly constructed model to predict the antigenic distance of three recent A(H3N2) viruses but there is no experimental data to validate their prediction (eg. if these viruses are indeed antigenically deviating from group 2 strains as concluded by the authors).

Indeed, there is no experimental data from A/Hong_Kong/45/2018, A/Tasmania/503/2020, or A/Darwin/9/2021. The generation of data to determine experimental values for A/Hong_Kong/45/2018, A/Tasmania/503/2020, or A/Darwin/9/2021 would require the generation of new reassortant viruses (H1N2s), recombinant protein and immunization of new ferrets. The ferrets sera would have to be analyzed against all 27 H6N2s, including duplicated control sera for normalization. The major point of the modeling was to evaluate if it is possible to predict the antigenic behavior based on amino acid substitutions.

As an exercise we have run the model again but this time excluding the Swe17 and HK17 antigens from the data set. Sequences of Sw17 or HK17 were then used to predict antigenic distances. The modeled versus experimental data are plotted in Author response image 8 and show a robust predictive outcome with R2 values of 0.94 and 0.91 for Sw17 and HK17, respectively.

Author response image 8.

Antigenic distances from Swe17 and HK17 calculated using the random forest algorithm that was constructed without experimental data from Swe17 and HK17. The predicted distances were plotted side by side to the experimental distances in (a) and correlations are shown in (b).

Reviewer #3 (Public Review):

Summary:

This paper by Portela Catani et al examines the antigenic relationships (measured using monotypic ferret and mouse sera) across a panel of N2 genes from the past 14 years, along with the underlying sequence differences and phylogenetic relationships. This is a highly significant topic given the recent increased appreciation of the importance of NA as a vaccine target, and the relative lack of information about NA antigenic evolution compared with what is known about HA. Thus, these data will be of interest to those studying the antigenic evolution of influenza viruses. The methods used are generally quite sound, though there are a few addressable concerns that limit the confidence with which conclusions can be drawn from the data/analyses.

Strengths:

• The significance of the work, and the (general) soundness of the methods.

• Explicit comparison of results obtained with mouse and ferret sera.

Weaknesses:

• Approach for assessing the influence of individual polymorphisms on antigenicity does not account for the potential effects of epistasis.

Indeed, possible epistatic effects or individual polymorphisms were not assessed, which is limited by the nature of the panel of N2s selected in the study. We now emphasize this in the discussion as follows:

“Also, our modelling does not consider that substitution by different amino acids can have distinct impact on antigenic distance. As a consequence, predictions based on the model could underestimate the importance of a particular amino acid residue substitution in some cases.”

• Machine learning analyses were neither experimentally validated nor shown to be better than simple, phylogenetic-based inference.

This is a valid remark and indeed we have found a clear correlation between NAI cross reactivity and phylogenetic relatedness. However, besides achieving good prediction of the experimental data (as shown in Figure 5 and in FigureR7), machine Learning analysis has the potential to rank or indicate major antigenic divergences based on available sequences before it has consolidated as new clade. ML can also support the selection and design of broader reactive antigens.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Major corrections

No major corrections, beyond the issues I touched on in the public review, for which I give a little more detail below:

Point 2. If there's not a putative genetic basis for the unexpected clustering seen in the NAI, then reiterating a small subset of the data would show the reliability of the experimental methods and substantiate this unexpected finding.

We thank the reviewer for this pertinent point and suggestion. We have modified our analysis by reiterating individual ferret data normalized with the homologous ELISA titers. This reiteration is shown in figure R1b. In this case both Kan17 and Wis15 are switched to antigenic group 2. The profile of sera inhibition against those 2 strains that shift from antigenic cluster 1 to 2, is clearly an intermediate between profiles observed in those 2 groups. Considering that antigenic evolution occurs gradually, it is not unexpected that those intermediate profiles would swing from one side to another when pushed to forced discrimination. Antigenic cartography mapping, as in Smith et al. (2004), also indicated that those H6N2s are located closer to G1 than overall antigens from G2. Raw data distribution (max and min EC50) also do not indicate potential bias in analysis.

Point 5. If you want to use antigenic cartography (Smith et al 2004), there is the R CRAN package (https://CRAN.R-project.org/package=Racmacs) which can handle threshold titres (like <20) and has functions for the diagnostic tools I describe, in order to quality assure the resulting plot. It does use a different antigenic distance metric than the paper currently uses, so you might not want to take that route.

Thank you for this suggestion. We have performed antigenic cartography using the methodology described by Smith et al made accessible by Sam Wilks. The outcome of this analysis has been added to the manuscript as Figure 2 – Figure supplement 3.

Point 6. More robust measures of antigenic distance take into account the homologous titre, homologous and heterologous titres (Archetti & Horsfall, 1950) or use the highest observed titre for a serum (Smith et al 2004). A limitation of the first two is that the antigenic distance can only be calculated when you have the homologous titre, which will limit you as you only have this for 26/43 sera. They may give similar results to your average antigenic distance, in which case your analysis still stands. Calculating antigenic distance using the homologous or maximum titre only gives the antigenic distance between the antigen and the serum. If you want the distance between all the sera, then further analysis is required (making an antigenic map and outputting the serum-serum distances, see the point above).

We thank the reviewer for these suggestions. A complete set of 43 H6N2 viruses that matches all 43 sera would have been ideal. This would require the generation of 17 additional H6N2 viruses and their testing in ELLA, a significant amount of work in terms of time and resources. Instead, we have generated an antigenic map of the 27 antigens and homologous sera (cfr. our response to point 5 above). Despite different methods the outcome showing 4 major antigenic groups is consistent.

Minor corrections

Table S1

A/New_Castle/67/2016 should be A/Newcastle/67/2016

A/Gambia/2012 is not the full virus name

Corrected.

Table S3 has multiple values of exactly 10.0. I think these should be <20 as they are below the threshold of detection for the assay.

All the values lower than 20 in Table S3 were replaced by “< 20”.

Line 376: A/Sidney/5/1997 should be A/Sydney/5/1997

Corrected.

Line 338: "25 randomly sampled data" is a bit vague, "25 randomly sampled features" would be better

Corrected.

Include RMSE of the random forest model.

RMSE=19.6 RMSE/mean = 0.207 is now mentioned in the manuscript.

Figure 5 - supplement 1: These plots are difficult to interpret as the aspect ratio is not 1:1, and panels a & b are difficult to compare as they have not been aligned (using a Procrustes analysis). It would be neater if they were labelled with short names.

We have generated an antigenic cartography map instead. As a consequence, the MDS has become redundant and Figure 5 – supplement 1 was removed.

Line 562: 98 variable residues, where it is 102 elsewhere in the text.

There are 4 mutations near the end of the NA stalk domain, which are not resolved in the N2 structure. Therefore, amino acid distances to these residues cannot be calculated.

No data availability statement. Some of the raw data is available in Table S3 and there is no link to the code.

The data and code used for generation of rf modelling was uploaded to Github and made available. The following statement has been added to the manuscript: “The data and code used for the generation of the rf model is available at https://github.com/SaelensLAB/RF..”

Reviewer #2 (Recommendations For The Authors):

(1) More than 42,000 NA sequences are available for the mentioned period on GISAID, it is therefore important to understand the selection criteria for the 44 strains and if these strains represent the overall genetic diversity of N2 of human A(H3N2) viruses. To demonstrate the representativeness of the 44 selected strains, please construct a representative N2 phylogenetic tree for human A(H3N2) viruses circulated in 2009-2017 and label the 44 selected strains on the tree.

The selection of antigens was performed using the method described by Bien and Tibshirani 2011 (doi: 10.1198/jasa.2011.tm10183). This method uses MinMax distances to identify a central representative among distinct clusters.

To facilitate visualization tree only of 180 representative N2 proteins from 2009-2017 were randomly selected (20 strains per year, unlabelled). Those 180 representatives and 44 readout panel strains (labelled) are shown in the phylogenetic tree below. Readout strains cover the major branches of the tree. The tree has been built using PhyML 3.0 using JTT substitution model and default parameters (Guindon S. et al, Systematic Biology 59(3):307-21, 2010) and visualized using ETE3 (Huerta-Cepas J. et al, Mol. Biol. Evol 33(6):1635-38, 2016).

Author response image 9.

(2) Double immune ferret sera may increase antibody binding affinity and cross-reactivity against heterologous strains. Using single-infection ferret sera may yield different antigenic grouping results (eg. may identify more antigenic groups). Can the authors repeat the NA antigenic grouping using single-infection ferret sera? Although data from a subset of 5 strains was presented (Figure 2, Figure Supplement 4), the information was not sufficient to support if the use of single-infection or double immune ferret sera will yield similar antigenic grouping results.

In our ferret immunizations the boost was performed with recombinant, enzymatically active NA that was homologous to the NA of the H1N2 virus that was used for the priming by infection. We determined the NAI responses in sera from ferrets after H1N2 infection against 5 different H6N2 viruses (Figure 2 – figure supplement 5). Compared to NAI responses in sera from H1N2 infected and subsequently NA protein boosted ferrets, the NAI titers obtained after a single infection were considerably lower. Although the normalized NAI titers of day 14 and day 42 sera correlated well, we cannot exclude a degree of broadening of the NAI response in the NA protein boost sera (Figure R6). On the other hand, repeated influenza antigen exposure is the reality for the majority of people.

(3) NA antigenicity data is presented in heat maps and the authors would often describe the heat map patterns matches without further explanations. Line 234-235, the heat map of mouse sera (Figure 2. Figure supplement 5) was described to match the results of ferret sera (Figure 2), but this tends to be subjective. A correlation analysis of 7 selected antigens showed a positive correlation, what about the other 37 antigens?

The interpretation of heatmaps is indeed very subjective, for this reason the correlation of the 7 selected antigens was also provided. The other 37 antigens were not tested. Considering the results using post boost sera, a simulation of using random forest modeling indicate that the data from one antigen of each antigenic group is sufficient to achieve a reliable predictive output (R2=0.71) (Figure R3 of this rebuttal).

(4) Can the authors explain in more detail how data in Figure 4a was generated? According to the authors, residues close to the catalytic pocket are more likely to impact NAI. Can the authors explain how they define if a residue is close to the catalytic pocket?

The correlation of distances of amino acid residues with significance values is explained as follows. Consider 7 distinct elements that are distributed horizontally as shown by the squares in the figure below (Author response image 10a). The elements highlighted in yellow have a numerical propriety (in case of N2 neuraminidase this was the significance values obtained in the association study). Taking P1 as reference we can calculate the distance (red arrows) between P1 and P2, P4 and P7, those distances can them be correlated to intrinsic values of P2, P4 and P7, which enables the calculation of the correlation coefficient Tau. This same process is repeated for each position (or each amino acid), as a consequence every position will have a correlation coefficient calculated (Author response image 8b). This correlation coefficient can be represented as a heat map at the surface of N2.

Author response image 10.

The 2D scheme represents the strategy used to calculate the correlation (i.e. the Tau values) between distances and p-values. Tau values can then be presented in a heat map.

(5) Can the authors provide experimental data using the three recent A(H3N2) viruses as antigens and perform NAI assay to confirm if they are antigenic all deviating from group 2 viruses?

The generation of data to determine experimental values for A/Hong_Kong/45/2018, A/Tasmania/503/2020, or A/Darwin/9/2021 would require the generation of new reassortant viruses (H1N2s), recombinant protein and immunization of new ferrets. The ferrets sera would have to be analyzed against all 27 H6N2s, including duplicated control sera for normalization. The major point of the modeling was to evaluate if it is possible to predict the antigenic behavior based on amino acid substitutions.

As an exercise we have run the model again but this time excluding the Swe17 and HK17 antigens from the data set. Sequences of Sw17 or HK17 were then used to predict antigenic distances. The modeled versus experimental data are plotted in Author response image 7 and show a robust predictive outcome with R2 values of 0.94 and 0.91 for Sw17 and HK17, respectively.

(6) According to Ge et al. 2022 (PMID: 35387078), N2 NA's before 2014 (2007-2013) showed a 329-N-glycosylation and E344, and they were subsequently replaced by H3N2 viruses with E344K and 329 non-glycosylation changing the NI reactivity in ferret antisera towards later strains. Were these residues also predicted to be important to N2 antigenicity from your machine-learning method?

Three of the N2 NAs used in our panel, A/Victoria/361/2011, A/Hong_Kong/3089/2017, and A/Tennessee/18/2017, lack this N-glycosylation motif. The E344K substitution is present in another 3 NAs, derived from A/Nagano/2153/2017, A/Minnesota/11/2010, and A/Indiana/08/2011. The importance of those mutations is among the lowest ones predicted in our modeling. However, the differences in NAI reported by Ge et al. are low (not even twofold). The experimental variability in our study potentially limits the identification of substitutions with a subtle impact NAI. We have added the following to the discussion in our revised manuscript:

“It has been reported that an N-glycosylation site at position 329 combined with E344 in NA from human H3N2 viruses from 2007 to 2013 was gradually lost in later H3N2 viruses (Ge et al., 2022). This loss of an N-glycosylation site at position 329 combined with an E344K substitution was associated with a change in NAI reactivity in ferret sera. Three N2 NAs in our panel, derived from A/Victoria/361/2011, A/Hong_Kong/3089/2017, and A/Tennessee/18/2017, lack this N-glycosylation motif. The E344K substitution is present in three other NAs, derived from A/Nagano/2153/2017, A/Minnesota/11/2010, and A/Indiana/08/2011. The importance of those mutations is among the lowest ones predicted by our modeling. However, the differences in NAI reported by Ge et al. are very modest (lower than twofold). The experimental variability in our study potentially limits the identification of substitutions with a subtle impact NAI.”

Reviewer #3 (Recommendations For The Authors):

Specific suggestions:

Line 132: Did the authors confirm the absence of compensatory mutations due to a heterologous H6 background that could potentially confound downstream NAI results?

All NAs genes of the rescued H6N2 viruses were fully sequenced and were found to be identical to the expected NA sequences, with the only exception being the A/Tasmania/1018/2015 were a mixed population of wt and M467I was found. This substitution is located at the surface and at the top of the NA head domain, and thus could potentially impact NA antigenicity. However, A/Tasmania/1018/2015 H6N2s had a similar inhibition profile as other H6N2s in phylogenetic and antigenic group 1. This indicates that, at least in this mixed population, antigenicity was not drastically affected by the M467I substitution.

Line 96: how do these data rule out variation in the fraction of properly folded protein across NAs? They certainly show that properly folded NA protein is present, but not whether amounts vary between the different NAs.

SEC-MALS (size exclusion chromatography-Multiangle light scattering) data and enzymatic activity were considered as a proxy for correctly folded NA. Although the specific activity of the recombinant N2 NAs is expressed per mass unit (microgram), we cannot exclude that the fraction of properly folded protein across the different recombinant NAs may vary.

Lines 262-269: this analysis approach (based on my reading) seems to consider each polymorphism in isolation and thus does not seem well suited for accounting for epistatic interactions within the NA. For example, the effect of a substitution on NAI may be contingent upon other alleles within NA that are not cleanly segregated between the two serum comparator groups. Can the authors address the potential of epistasis within NA to confound the results shown in Figure 3?

Unfortunately, epistatic interactions cannot be solved using the panel of N2 selected for the study. This limitation is mentioned in our discussion:

“It is important to highlight that co-occurring substitutions in our panel (the ones present in the main branches of the phylogenetic tree) cannot be individually assessed by association analysis or the random forest model. The individual weight of those mutation on NA drift thus remains to be experimentally demonstrated.”

Line 331: is there a way to visualize and/or quantify how these two plots (F5 supplement 1a/b) reflect each other or not? Without this, it is hard to ascertain how they relate to each other.

We have generated an antigenic cartography map instead. As a consequence, the MDS has become redundant and Figure 5 – supplement 1 was removed.

Figure 4B structural images are not well labelled.

The active site in 1 of the protomers is now indicated with an arrow in the top and side views of the NA tetramer.

Lines 339-359: the ML predictions are just predictions and kind of meaningless without experimental validation of the predicted antigenic differences between recent NAs. This section would also be strengthened by an assessment of whether the ML approach obtains more accurate results than simply using phylogeny to predict antigenic relationships.

Indeed, there is no experimental data from A/Hong_Kong/45/2018, A/Tasmania/503/2020, or A/Darwin/9/2021. The generation of data to determine experimental values for A/Hong_Kong/45/2018, A/Tasmania/503/2020, or A/Darwin/9/2021 would require the generation of new reassortant viruses (H1N2s), recombinant protein and immunization of new ferrets. The ferrets sera would have to be analyzed against all 27 H6N2s, including duplicated control sera for normalization. The major point of the modeling was to evaluate if it is possible to predict the antigenic behavior based on amino acid substitutions.

As an exercise we have run the model again but this time excluding the Swe17 and HK17 antigens from the data set. Sequences of Sw17 or HK17 were then used to predict antigenic distances. The modeled versus experimental data are plotted in figure R7 and show a robust predictive outcome with R2 values of 0.94 and 0.91 for Sw17 and HK17, respectively. A major advantage of antigenic modeling is the potential to rank or indicate major antigenic divergences based on available sequences before it has consolidated as new clade. The support in selecting or designing broader reactive antigens is another advantage of machine learning analysis.

Lines 416-421: appreciate the direct comparison of results obtained from ferrets versus mice.

We thank the reviewer for expressing this appreciation.

eLife assessment

This study presents valuable data on the antigenic properties of neuraminidase proteins of human A/H3N2 influenza viruses sampled between 2009 and 2017. The antigenic properties are found to be generally concordant with genetic groups. Additional analysis have strengthened the revised manuscript, and the evidence supporting the claims is solid.

Reviewer #1 (Public Review):

Summary

The authors investigated the antigenic diversity of recent (2009-2017) A/H3N2 influenza neuraminidases (NAs), the second major antigenic protein after haemagglutinin. They used 27 viruses and 43 ferret sera and performed NA inhibition. This work was supported by a subset of mouse sera. Clustering analysis determined 4 antigenic clusters, mostly in concordance with the genetic groupings. Association analysis was used to estimate important amino acid positions, which were shown to be more likely close to the catalytic site. Antigenic distances were calculated and a random forest model used to determine potential important sites.

This revision has addressed many of my concerns of inconsistencies in the methods, results and presentation. There are still some remaining weaknesses in the computational work.

Strengths

(1) The data cover recent NA evolution and a substantial number (43) of ferret (and mouse) sera were generated and titrated against 27 viruses. This is laborious experimental work and is the largest publicly available neuraminidase inhibition dataset that I am aware of. As such, it will prove a useful resource for the influenza community.

(2) A variety of computational methods were used to analyse the data, which give a rounded picture of the antigenic and genetic relationships and link between sequence, structure and phenotype.

(3) Issues raised in the previous review have been thoroughly addressed.

Weaknesses

(1) Some inconsistencies and missing data in experimental methods<br /> Two ferret sera were boosted with H1N2, while recombinant NA protein for the others. This, and the underlying reason, are clearly explained in the manuscript. The authors note that boosting with live virus did not increase titres. Additionally, one homologous serum (A/Kansas/14/2017) was not generated, although this would not necessarily have impacted the results.

(2) Inconsistency in experimental results<br /> Clustering of the NA inhibition results identifies three viruses which do not cluster with their phylogenetic group. Again this is clearly pointed out in the paper and is consistent with the two replicate ferret sera. Additionally, A/Kansas/14/2017 is in a different cluster based on the antigenic cartography vs the clustering of the titres

(3) Antigenic cartography plot would benefit from documentation of the parameters and supporting analyses<br /> a. The number of optimisations used<br /> b. The final stress and the difference between the stress of the lowest few (e.g. 5) optimisations, or alternatively a graph of the stress of all the optimisations. Information on the stress per titre and per point, and whether any of these were outliers<br /> c. A measure of uncertainty in position (e.g. from bootstrapping)

(4) Random forest<br /> The full dataset was used for the random forest model, including tuning the hyperparameters. It is more robust to have a training and test set to be able to evaluate overfitting (there are 25 features to classify 43 sera).

Reviewer #2 (Public Review):

Summary:<br /> The authors characterized the antigenicity of N2 protein of 43 selected A(H3N2) influenza A viruses isolated from 2009-2017 using ferret and mice immune sera. Four antigenic groups were identified, which the authors claimed to be correlated with their respective phylogenic/ genetic groups. Among 102 amino acids differed by the 44 selected N2 proteins, the authors identified residues that differentiate the antigenicity of the four groups and constructed a machine-learning model that provides antigenic distance estimation. Three recent A(H3N2) vaccine strains were tested in the model but there was no experimental data to confirm the model prediction results.

Strengths:<br /> This study used N2 protein of 44 selected A(H3N2) influenza A viruses isolated from 2009-2017 and generated corresponding panels of ferret and mouse sera to react with the selected strains. The amount of experimental data for N2 antigenicity characterization is large enough for model building.

Weaknesses:<br /> The main weakness is that the strategy of selecting 43 A(H3N2) viruses from 2009-2017 was not explained. It is not clear if they represent the overall genetic diversity of human A(H3N2) viruses circulating during this time. In response to the reviewer's comment, the authors have provided a N2 phylogenetic tree using180 randomly selected N2 sequences from human A(H3N2) viruses from 2009-2017. While the 43 strains seems to scatter across the N2 tree, the four antigenic groups described by the author did not correlated with their respective phylogenic/ genetic groups as shown in Fig. 2. The authors should show the N2 phylogenic tree together with Fig. 2 and discuss the discrepancy observed.

The second weakness is the use of double-immune ferret sera (post-infection plus immunization with recombinant NA protein) or mouse sera (immunized twice with recombinant NA protein) to characterize the antigenicity of the selected A(H3N2) viruses. Conventionally, NA antigenicity is characterized using ferret sera after a single infection. Repeated influenza exposure in ferrets has been shown to enhance antibody binding affinity and may affect the cross-reactivity to heterologous strains (PMID: 29672713). The increased cross-reactivity is supported by the NAI titers shown in Table S3, as many of the double immune ferret sera showed the highest reactivity not against its own homologous virus but to heterologous strains. In response to the reviewer's comment, the authors agreed the use of double-immune ferret sera may be a limitation of the study. It would be helpful if the authors can discuss the potential effect on the use of double-immune ferret sera in antigenicity characterization in the manuscript.

Another weakness is that the authors used the newly constructed a model to predict antigenic distance of three recent A(H3N2) viruses but there is no experimental data to validate their prediction (eg. if these viruses are indeed antigenically deviating from group 2 strains as concluded by the authors). In response to the comment, the authors have taken two strains out of the dataset and use them for validation. The results is shown as Fig. R7. However, it may be useful to include this in the main manuscript to support the validity of the model.

Reviewer #3 (Public Review):

Summary:<br /> This paper by Portela Catani et al examines the antigenic relationships (measured using monotypic ferret and mouse sera) across a panel of N2 genes from the past 14 years, along with the underlying sequence differences and phylogenetic relationships. This is a highly significant topic given the recent increased appreciation of the importance of NA as a vaccine target, and the relative lack of information about NA antigenic evolution compared with what is known about HA. Thus, these data will be of interest to those studying the antigenic evolution of influenza viruses. The methods used are generally quite sound, though there are a few addressable concerns that limit the confidence with which conclusions can be drawn from the data/analyses.

Strengths:<br /> -The significance of the work, and the (general) soundness of the methods.<br /> -Explicit comparison of results obtained with mouse and ferret sera

Weaknesses:<br /> - Approach for assessing influence of individual polymorphisms on antigenicity does not account for potential effects of epistasis (this point is acknowledged by the authors).<br /> - Machine learning analyses neither experimentally validated nor shown to be better than simple, phylogenetic-based inference.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

This study presents fundamental new insights into vesicular monoamine transport and the binding pose of the clinical drug tetrabenazine (TBZ) to the mammalian VMAT2 transporter. Specifically, this study reports the first structure for the mammalian VMAT (SLC18) family of vesicular monoamine transporters. It provides insights into the mechanism by which this inhibitor traps VMAT2 into a 'dead-end' conformation. The structure also provides some evidence for a novel gating mechanism within VMAT2, which may have wider implications for understanding the mechanism of transport in the wider SLC18 family.

Strengths:

The structure is high quality, and the method used to determine the structure via fusing mVenus and the anti-GFP nanobody to the amino and carboxyl termini is novel. The binding and transport data are convincing and provide new insights into the role of conserved side chains within the SLC18 members. The binding position of TBZ is of high value, given its role in treating Huntington's chorea and for being a 'dead-end' inhibitor for VMAT2.

We thank reviewer #1 for their constructive comments and input which we feel has greatly improved the manuscript.

Reviewer #2 (Public Review):

This public review is the same review that was posted earlier and has not been updated in response to our comments or to the revised manuscript. Please see our earlier response to these comments. We thank reviewer #2 for their input and we have incorporated many of these suggestions into our revised manuscript. With regard to the question of ‘how TBZ got there’, we have revised this sentence in the discussion to be more speculative. As pointed out earlier, our interpretation of the structure is based on a wealth of experimental and structural data which support our interpretations. Thus, our conclusions have not been overstated. This has been explained in our earlier public response and these key studies have been cited throughout the manuscript. We also note that reviewer #3 found the AlphaFold comparisons to be quite meaningful.

Overview:

As a report of the first structure of VMAT2, indeed the first structure of any vesicular monoamine transporter, this manuscript represents an important milestone in the field of neurotransmitter transport. VMAT2 belongs to a large family (the major facilitator superfamily, MFS) containing transporters from all living species. There is a wealth of information relating to the way that MFS transporters bind substrates, undergo conformational changes to transport them across the membrane and couple these events to the transmembrane movement of ions. VMAT2 couples the movement of protons out of synaptic vesicles to the vesicular uptake of biogenic amines (serotonin, dopamine and norepinephrine) from the cytoplasm. The new structure presented in this manuscript can be expected to contribute to an understanding of this proton/amine antiport process.

The structure contains a molecule of the inhibitor TBZ bound in a central cavity, with no access to either luminal or cytoplasmic compartments. The authors carefully analyze which residues interact with bound TBZ and measure TBZ binding to VMAT2 mutated at some of those residues. These measurements allow well-reasoned conclusions about the differences in inhibitor selectivity between VMAT1 and VMAT2 and differences in affinity between TBZ derivatives.

The structure also reveals polar networks within the protein and hydrophobic residues in positions that may allow them to open and close pathways between the central binding site and the cytoplasm or the vesicle lumen. The authors propose involvement of these networks and hydrophobic residues in coupling of transport to proton translocation and conformational changes. However, these proposals are quite speculative in the absence of supporting structures and experimentation that would test specific mechanistic details.

Critique:

Although the structure presented in this MS is clearly important, I feel that the authors have overstated several of the conclusions that can be drawn from it. I don't agree that the structure clearly indicates why TBZ is a non-competitive inhibitor; the proposal that specific hydrophobic residues function as gates will depend on lumen- and cytoplasm-facing structures for verification; the polar networks could have any number of functions - indeed it would be surprising if they were all involved in proton transport. Several of these issues could be resolved by a clearer illustration of the data, but I believe that a more rigorous description of the conclusions and where they fall between firm findings and speculation would help the reader put the results in perspective.

Non-competitive inhibition occurs when the action of an inhibitor can't be overcome by increasing substrate concentration. The structure shows TBZ sequestered in the central cavity with no access to either cytoplasm or lumen. The explanation of competitive vs non-competitive inhibition depends entirely on how TBZ got there. If it bound from the cytoplasm, cytoplasmic substrate should have been able to compete with TBZ and overcome the inhibition. If it bound from the lumen, or from within the bilayer, cytoplasmic substrate would not be able to compete, and inhibition would be non-competitive. The structure does not tell us how TBZ got there, only that it was eventually occluded from both aqueous compartments and the bilayer.

The issue of how VMAT2 opens access to the central binding site from luminal and cytoplasmic sides is an important and interesting one, and comparison with other MFS structures in cytoplasmic-open or extracellular/luminal-open is a very reasonable approach. However, any conclusions for VMAT2 should be clearly indicated as speculative in the absence of comparable open structures of VMAT2. As a matter of presentation, I found the illustrations in ED Fig. 6 to be less helpful than they could have been. Specifically, illustrations that focus on the proposed gates, comparing that region of the new structure with the corresponding region of either VGLUT or GLUT4 would better help the reader to compare the position of the proposed gate residues with the corresponding region of the open structure. I realize that is the intended purpose of ED Fig. 6b and 6c, but currently, those show the entire protein and a focus on the gate regions might make the proposed gate movements clearer. I also appreciate the difference between the Alphafold prediction and the new structure, but I'm not convinced that ED Fig. 6a adds anything helpful.

The polar networks described in the manuscript provide interesting possibilities for interactions with substrates and protons whose binding to VMAT2 must control conformational change. Aside from the description of these networks, there is little evidence presented to assess the role of these networks in transport. Are the networks conserved in other closely related transporters? How could the interaction of the networks with substrate or protons affect conformational change? Of course, any potential role proposed for the networks would be highly speculative at this point, and any discussion of their role should point out their speculative nature and the need for experimental verification. Some speculation, however, can be useful for focusing the field's attention on future directions. However, statements in the abstract (three distinct polar networks... play a role in proton transduction.) and the discussion (...are likely also involved in mediating proton transduction.) should be clearly presented as speculation until they are validated experimentally.

The strongest aspect of this work (aside from the structure itself) is the analysis of TBZ binding. I will comment on some minor points below, but there is one problematic aspect to this analysis. The discussion on how TBZ stabilizes the occluded conformation of VMAT2 is premature without structures of apo-VMAT2 and possibly structures with other ligands bound. We don't really know at this point whether VMAT2 might be in the same occluded conformation in the absence of TBZ. Any statements regarding the effect of interactions between VMAT2 and TBZ depend on demonstrating that TBZ has a conformational effect. The same applies to the discussion of the role of W318 on conformation and to the loops proposed to "occlude the luminal side of the transporter" (line 131).

The description of VMAT2 mechanism makes many assumptions that are based on studies with other MFS transporters. Rather than stating these assumptions as fact (VMAT2 functions by alternating access...), it would be preferable to explain why a reader should believe these assumptions. In general, this discussion presents conclusions as established facts rather than proposals that need to be tested experimentally.

The MD simulations are not described well enough for a general reader. What is the significance of the different runs? ED Fig. 4d is not high enough resolution to see the details.

Reviewer #3 (Public Review):

Summary:

The vesicular monoamine transporter is a key component in neuronal signaling and is implicated in diseases such as Parkinson's. Understanding of monoamine processing and our ability to target that process therapeutically has been to date provided by structural modeling and extensive biochemical studies. However, structural data is required to establish these findings more firmly.

Strengths:

Dalton et al resolved a structure of VMAT2 in the presence of an important inhibitor, tetrabenazine, with the protein in detergent micelles, using cryo-EM and with the aid of protein domains fused to its N- and C-terminal ends, including one fluorescent protein that facilitated protein screening and purification. The resolution of the maps allows clear assignment of the amino acids in the core of the protein. The structure is in good agreement with a wealth of experimental and structural prediction data, and provides important insights into the binding site for tetrabenazine and selectivity relative to analogous compounds. The authors provide additional biochemical analyses that further support their findings. The comparison with AlphaFold models is enlightening.

We appreciate this summary and thank reviewer #3 for their helpful suggestions to improve the manuscript.

Weaknesses:

The authors follow up their structures with molecular dynamics simulations of the tetrabenazine-bound state, and test several protonation states of acidic residues in the binding pocket, but not all possible combinations; thus, it is not clear the extent to which tetrabenazine rearrangements observed in these simulations are meaningful. Additional simulations of the substrate dopamine docked into this structure were also carried out, although it is unclear whether this "dead-end" occluded state is a relevant state for dopamine binding. The authors report release of dopamine during these simulations, but it is notable that this only occurs when all four acidic binding site residues were protonated and when an enhanced sampling approach was applied.

As an occluded neurotransmitter bound structure has yet to be solved experimentally, it is not possible to address whether this state resembles the docked dopamine structure. However, it is reasonable to hypothesize that this is a relevant state for dopamine binding and if so, these simulations would be of great interest. The MD simulations which were performed are logical, based on the calculated pKa of the residues and the known pH of the vesicle lumen (5.5). Note that we have carried out a total of more than 2 microseconds of simulations, which required a significant computing time/memory allocation for the current runs in explicit water and membrane. To investigate all possible combinations, it would require at least 16 independent simulations, to be performed in duplicates, to vary protonation status of the four highlighted acidic residues alone, not including proper experimental replicates. We do not believe this to be a feasible suggestion, nor necessary given that the selected combinations were based on rational evaluation of on-path amino acids that were assessed to be potentially protonated.

eLife assessment

The report presents the cryo-EM structure of human vesicular monoamine transporter 2 (VMAT2) bound to tetrabenazine, a clinical drug. VMAT2 is critical for neurotransmission, and the study constitutes an important milestone in neurotransmitter transport research. The evidence presented in the report is convincing and provides new opportunities for developing improved therapeutic interventions and furthering our understanding of this vital protein's function.

Reviewer #1 (Public Review):

Summary:

This study presents fundamental new insights into vesicular monoamine transport and the binding pose of the clinical drug tetrabenazine (TBZ) to the mammalian VMAT2 transporter. Specifically, this study reports the first structure for the mammalian VMAT (SLC18) family of vesicular monoamine transporters. It provides insights into the mechanism by which this inhibitor traps VMAT2 into a 'dead-end' conformation. The structure also provides some evidence for a novel gating mechanism within VMAT2, which may have wider implications for understanding the mechanism of transport in the wider SLC18 family.

Strengths:

The structure is high quality, and the method used to determine the structure via fusing mVenus and the anti-GFP nanobody to the amino and carboxyl termini is novel. The binding and transport data are convincing and provide new insights into the role of conserved side chains within the SLC18 members. The binding position of TBZ is of high value, given its role in treating Huntington's chorea and for being a 'dead-end' inhibitor for VMAT2.

Reviewer #2 (Public Review):

As a report of the first structure of VMAT2, indeed the first structure of any vesicular monoamine transporter, this manuscript represents an important milestone in the field of neurotransmitter transport. VMAT2 belongs to a large family (the major facilitator superfamily, MFS) containing transporters from all living species. There is a wealth of information relating to the way that MFS transporters bind substrates, undergo conformational changes to transport them across the membrane and couple these events to the transmembrane movement of ions. VMAT2 couples the movement of protons out of synaptic vesicles to the vesicular uptake of biogenic amines (serotonin, dopamine and norepinephrine) from the cytoplasm. The new structure presented in this manuscript can be expected to contribute to an understanding of this proton/amine antiport process.

The structure contains a molecule of the inhibitor TBZ bound in a central cavity, with no access to either luminal or cytoplasmic compartments. The authors carefully analyze which residues interact with bound TBZ and measure TBZ binding to VMAT2 mutated at some of those residues. These measurements allow well-reasoned conclusions about the differences in inhibitor selectivity between VMAT1 and VMAT2 and differences in affinity between TBZ derivatives.

The structure also reveals polar networks within the protein and hydrophobic residues in positions that may allow them to open and close pathways between the central binding site and the cytoplasm or the vesicle lumen. The authors propose involvement of these networks and hydrophobic residues in coupling of transport to proton translocation and conformational changes.

Reviewer #3 (Public Review):

Summary:

The vesicular monoamine transporter is a key component in neuronal signaling and is implicated in diseases such as Parkinson's. Understanding of monoamine processing and our ability to target that process therapeutically has been to date provided by structural modeling and extensive biochemical studies. However, structural data is required to establish these findings more firmly.

Strengths:

Dalton et al resolved a structure of VMAT2 in the presence of an important inhibitor, tetrabenazine, with the protein in detergent micelles, using cryo-EM and with the aid of protein domains fused to its N- and C-terminal ends, including one fluorescent protein that facilitated protein screening and purification. The resolution of the maps allows clear assignment of the amino acids in the core of the protein. The structure is in good agreement with a wealth of experimental and structural prediction data, and provides important insights into the binding site for tetrabenazine and selectivity relative to analogous compounds. The authors provide additional biochemical analyses that further support their findings. The comparison with AlphaFold models is enlightening.

Author Response

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

We thank the editor for organizing the review of our manuscript. We have carefully read and analyzed the reviewers’ comments, addressed each criticism point-by-point as outlined below, and modified the manuscript and figures accordingly. In this regard, we would also like to take the opportunity to thank both reviewers for their thoughtful suggestions for improvement of our manuscript. We believe that our manuscript has improved as a result, and hope that it is now suitable for publication.

Public Reviews:

Reviewer #1 (Public Review):

Aiming at the problem that Staphylococcus aureus can cause apoptosis of macrophages, the author found and verified that drug (R)-DI-87 can inhibit mammalian deoxycytidine kinase (dCK), weaken the killing effect of staphylococcus aureus on macrophages, and reduce the apoptosis of macrophages. And increase the infiltration of macrophages to the abscess, thus weakening the damage of Staphylococcus aureus to the host. This work provides new insights and ideas for understanding the effects of Staphylococcus aureus infection on host immunity and discovering corresponding therapeutic interventions.

The logic of the study is commendable, and the design is reasonable.

Some data related to the conclusion of the paper need to be supplemented, and some experimental details need to be described.

Response: We thank the reviewer for the positive feedback along with the detailed and knowledgeable analysis of this paper. Specific details and comments on all raised concerns can be found below.

Reviewer #2 (Public Review):

Summary:

In this study, Winstel and colleagues test if the deoxycytidine kinase inhibitor, (R)-DI-87 provides therapeutic benefit during infection with Staphylococcus aureus. The premise behind the current work is a series of prior studies that found that S. aureus can disable functional immune clearance by generating NET-derived deoxyribonucleosides to induce macrophage apoptosis via purine salvage. Here, the authors use in vitro and in vivo experiments with (R)-DI-87 to demonstrate that inhibition of deoxycytidine kinase prevents S. aureus-induced deoxyribonucleoside-mediated macrophage cell death, to bolster immune cell function and promote more effective clearance during infection. The authors conclude that (R)-DI-87 represents and potentially important Host-Directed Therapy (HDT) with good potential to promote natural clearance of infection without targeting the bacterium. Overall, the study represents an important next step in the exploration of purine salvage and deoxyribonucleoside toxicity as a targetable pathway to bolster infection clearance and provides early-stage evidence of the therapeutic potential of (R)-DI-87 during S. aureus infection.

Response: We thank the reviewer for the thoughtful suggestions for improvement of our manuscript. Specific details and comments on all raised concerns can be found below.

Strengths:

The study has several strengths that support its conclusions:

(1) Well-controlled in vitro studies that firmly establish (R)-DI-87 is capable of blocking deoxyribonucleoside-mediated apoptosis of immune cell lines and primary cells.

(2) Solid evidence to support that administration of (R)-DI-87 can have therapeutic benefits during infection (reduced number of abscesses and reduced CFU).

(3) Controls included to ascertain the degree to which (R)-DI-87 might have secondary effects on immune cell distribution.

(4) Controls included to ascertain whether or not (R)-DI-87 has intrinsic antibacterial properties.

Weaknesses:

However, there are several important weaknesses related to the rigor of the research and the conclusions drawn. The most relevant weaknesses noted by this reviewer are:

(1) Drawing firm conclusions about the therapeutic potential of (R)-DI-87 using only S. aureus strain Newman, a methicillin-susceptible S. aureus, that while a clinical isolate is not clearly representative of the strains of S. aureus causing infection in hospitals and communities. Newman also harbors an unusual mutation in a regulator that dramatically changes virulence factor gene expression. While the data with Newman remains valuable, the absence of consideration of other strains, including MRSA, makes it more difficult to support the relatively broad conclusions about therapeutic potential made by the authors.

Response: We assume that this is a misunderstanding. S. aureus Newman is a patient-derived isolate and not a regulator mutant and/or laboratory strain (Duthie and Lorenz LL 1952, J Gen Microbiol 6(1-2), 95107). Its genome is fully sequenced (Baba et al. 2008, J Bacteriol 190(1):300-10) and it is highly virulent in mouse or human ex vivo models (e.g. Alonzo 3rd et al. 2013, Nature 493(7430):51-5.; DuMont et al. 2011, Mol Microbiol 79(3):814-25; Skaar et al. 2004, Science 305(5690):1626-8). Moreover, S. aureus Newman has served as a gold standard to study abscess formation in the past (e.g. Thammavongsa et al. 2013, Science 342(6160):863-6; Cheng et al. 2009, FASEB J 23(10):3393-404; Corbin et al. 2008, Science 319(5865):962-5) and has further also been used multiple times to test the therapeutic efficacy of antimicrobial or anti-infective agents in various animal models of infectious disease (e.g. Buckley et al. 2023, Cell Host Microbe 31(5):751-765.e11; Zhang et al. 2014, PNAS 111(37):13517-22; Richter et al. 2013, PNAS 110(9):3531-6). Apart from this, it is crucial to note that methicillin-sensitive isolates such as S. aureus Newman are typically more frequently isolated in hospitals as compared to MRSA. Specifically, public health system- and population-based surveillance studies clearly indicate that annual incidence rates for MSSA infections are dominant over those associated with MRSA infections (e.g. Gagliotti et al. 2021, Euro Surveill 26(46):2002094; Jackson et al. 2020, Clin Infect Dis 70(6):1021-1028; Laupland et al. 2013, Clin Microbiol Infect 19(5):465-71), even in groups at elevated risk (e.g. McMullan et al. 2016, JAMA Pediatr et al., 170(10):979-986; Ericson et al. 2015, JAMA Pediatr 169(12):1105-11). Although we understand and agree with the reviewer that certain MRSA clones can be a dominant cause of staphylococcal disease in specific geographic areas, we believe that S. aureus Newman adequately reflects staphylococcal isolates that cause the majority of infections in humans. In this regard, we would also like to highlight once more that (R)-DI-87 targets host dCK and not the bacterium. Accordingly, the antibiotic resistance status of S. aureus is not expected to impact our main findings and conclusions as (R)-DI-87 exclusively inhibits dCK, a key element of the mammalian purine salvage pathway.

(2) In vitro (R)-DI-87 efficacy studies with dAdo and dGuo are strong, however, the authors do not test the in vitro efficacy of (R)-DI-87 using S. aureus. They have done this type of work in prior studies (See doi: 10.1073/pnas.1805622115 - Figure 5). If included it would greatly strengthen their argument that (R)-DI87 is directly affecting the S. aureus --> Nuclease --> AdsA macrophage-killing pathway. Without it, the evidence provided remains indirect, and several conclusions may be overstated.

Response: We highly appreciate this comment and agree with the reviewer that such an experiment would support our main findings. Thus, we have performed additional experiments and took advantage of a previously described approach (Tantawy et al. 2022, Front Immunol 13:847171) to demonstrate that (R)DI-87-mediated inhibition of host dCK enhances macrophage survival upon treatment with culture media that had been conditioned by incubation with adsA-proficient or adsA-deficient staphylococci in the presence or absence of purine deoxyribonucleoside monophosphates. Our findings are described in the main text and in a new figure (Fig. 2K-L). Based on these new findings and together with our rAdsA-based approach (Fig. 2I-J), we are confident that (R)-DI-87 represents a suitable small molecule inhibitor of host dCK which can prevent host immune cell death induced by toxigenic products associated with the S. aureus Nuc/AdsA pathway.

(3) Caspase-3 immunoblot experiments seem to suggest an alternative conclusion to what was made by the authors. They point out that Caspase-3 cleavage does not occur upon treatment with (R)-DI-87. However, the data seem to argue that there is almost no caspase-3 present in (R)-DI-87 treated cells (cleaved or uncleaved). Might this suggest that caspase-3 is not even produced when cells are not experiencing deoxyribonucleoside toxicity? Perhaps the authors could reconsider the interpretation of this data.

Response: We believe that this is a misunderstanding. Our immunoblots (Fig. 3E-F) show only the processed forms of caspase-3. The antibody we have used can recognize full-length caspase-3 along with the p17 and p19 subunits that can result from cleavage. To clarify this point, we have slightly modified our main figure and provide the full immunoblots (Source data file) which clearly demonstrate that unprocessed caspase-3 (pro-caspase-3) is present in all samples. In this regard, we further note that caspase-3 can also form heterocomplexes with other proteins, presumably explaining some of the unknown bands in samples obtained from cells that have been exposed to death-effector deoxyribonucleosides. Additional bands are probably a result of cross-reactivity of the antibody and/or unspecific degradation of pro-caspase in cellular lysates.

(4) There are some concerns over experimental rigor and clarity of the experimental design in the methods. The most important points noted by this reviewer are included here. (a.) There is no description of the number of replicates or representation of the Western blots and no uncropped blots are provided. (b.) the methods describing the treatment conditions for in vivo studies are not sufficiently clear. For example, it is hard to tell when (R)-DI-87 is first administered to mice. Is it immediately before the infection, immediately after, or at the same time? This has important implications for interpreting the results in terms of therapeutic potential. (c.) There are several statements made that (R)-DI-87 does not have a negative impact on the mice however, it is not sufficiently clear that the studies conducted are sufficient to make this broader claim that (R)-DI-87 has no impact on the animal, except as it relates to the distribution of immune cells, which is directly tested. (d.) there are no quantitative measures of apoptosis or macrophage infiltration, which impacts the rigor of these imaging experiments. (e.) only female mice are used in the in vivo studies. There is no justification provided for this choice; however, the rigor of the study design and the ability to draw conclusions about therapeutic potential is impacted in the absence of consideration of both sexes.

Response: Thank you for raising these points here. (a) We have modified our figure legend and provide the full immunoblots (Source data file) in order to clarify this point. (b) Moreover, we now provide more experimental details on the treatment conditions that were used to administer (R)-DI-87 to mice (methods section). (c) Furthermore, we have conducted new experiments in order to demonstrate that administration of (R)-DI-87 has no impact on laboratory animals. Specifically, we provide new data along with additional text on organ cellularity following long-term exposure of mice to (R)-DI-87. In this regard, we have also applied our immuno-phenotyping approach to spleen tissues samples derived from mice that received (R)-DI-87 or vehicle. As outlined in our new results, neither developmental errors nor differences in lymphocyte development have been observed (new Fig. 4B-C; new supplementary Fig. 3). Together with our data on mouse body weight along with our immuno-phenotyping approach of blood cells (Fig. 4A and 4D) and the fact that (R)-DI-87 is extremely well tolerated in humans (personal communication; Kenneth A. Schultz, Trethera Corporation, Los Angeles, CA, USA), we are very confident that application of (R)-DI87 is safe and has no detrimental impact on the host. (d) Lastly, we would like to point out that due to the densely packed and extremely sticky cuff of immune cells within staphylococcal abscesses, it is technically not possible to extract enough abscess material required for a reliable quantification of apoptotic macrophages within infectious foci. Such an analysis would also not allow us to differentiate between lesion-infiltrating macrophages and macrophages that may reside at the periphery of the abscess. For these reasons, we have established a fluorescence microscopy-based approach to demonstrate increased macrophage infiltration rates into abscesses formed in organs of mice that have been treated with the dCK-specific inhibitor (R)-DI-87 (Fig. 5A-P). Nonetheless, we have slightly modified our figure and its legend in order to help the readership to localize S. aureus-derived tissue lesions and the periphery of abscesses in these images. (e) Finally, publicly available databases indicate that dCK is equally well expressed in various tissues in both sexes. Moreover, dCK is not encoded on a sex chromosome, neither in mice nor in humans. Thus, we believe that it is justified to test the in vivo efficacy of (R)-DI-87 in female mice. Nonetheless, we have conducted additional in vitro experiments to test whether (R)-DI-87 can protect male animal-derived BMDMs from death-effector deoxyribonucleosides in a manner similar to cells derived from female mice. As expected, we did not observe a sex-specific effect (new supplementary Fig. 5), and hope that this adequately addresses this point.

(5) Animal studies show significant disease burden (CFU) even after administration of (R)-DI-87. Given the absence of robust clearance of infection, the author's claims read as an overstatement of the data. The authors may wish to reframe their conclusions to better highlight the potential benefit of this therapy at reducing severe disease but also to point out relevant limitations, especially considering that it does not lead to clearance in this model. In general, the consideration of the limitations of the proposed therapeutic approach, as uncovered by the data, is not present. A more nuanced consideration of the data and its interpretations, including both strengths and limitations, would greatly help to frame the study.

Response: Thank you for raising this point here. To highlighting the limitations of our approach, we have modified several passages in the main text. Moreover, we have adjusted our discussion section accordingly.

Reviewer #1 (Recommendations For The Authors):

(1) In vivo experiments, the dose given to mice was 75mg/kg. How did the author determine the dose of this drug?

Response: We thank the reviewer for this question, which gives us the chance to clarify this point. The experimental condition used to block host dCK in mice has been adopted from a previous publication (Chen et al. 2023, Immunology 168(1):152-169). To improve the overall quality of our current manuscript, we now included more background information addressing this point. Specifically, we have added additional in vivo and biochemical data along with more conclusive text to our results section to better explain the reason for the dose given to mice (new Fig. 4E).

(2) The author established a mouse model of Staphylococcus aureus blood infection in vivo and divided four groups for related experiments. It is suggested that the authors should supplement the survival rate of mice in each group so that readers can understand the effect of the drug on the survival of mice with bloodstream infection.

Response: While this is an interesting suggestion by the reviewer, we believe that this is beyond the scope of our study. In particular, the current study focused on analyzing the capacity of the dCK-specific inhibitor (R)-DI-87 to improve macrophage survival during staphylococcal abscess formation in an effort to lower bacterial loads in infected organ tissues. However, we agree with the reviewer that (R)-DI-87 might also help to improve further clinical syndromes of staphylococcal infections, including lethal bloodstream infection. We therefore modified parts of our discussion to address this point.

(3) In the in vivo experiment, the author administered the drug by intragastric administration, but the treatment was for the bloodstream infection of Staphylococcus aureus, so the author needed to determine the actual effective concentration of the drug in the blood of mice.

Response: We thank the reviewer for this comment and agree that inclusion of more background information and data would be a valuable addition to our manuscript. As outlined above, we have designed our in vivo experiments based on the methodology of a previous publication (Chen et al. 2023, Immunology 168(1):152-169). Similar to Chen and colleagues, we have also used a dose of 75 mg/kg of (R)-DI-87 that allows complete inhibition of host dCK in vivo. In this regard, we have now performed additional in vivo experiments to address this point. More precisely, we took advantage of a highly sensitive and LC-MS/MSbased method to measure accumulation of deoxycytidine, the natural substrate of host dCK, in mouse plasma upon administration of the dCK-specific inhibitor. As shown in our new Fig. 4E, administration of (R)-DI-87 at a dose of 75 mg/kg strongly increased deoxycytidine levels in mouse plasma thereby indicating that host dCK activity is completely blocked under these experimental conditions.

(5) This work is to reduce the apoptosis of macrophages through drug inhibition of dck, but not directly inhibit the related virulence of Staphylococcus aureus. Therefore, it is suggested that the author modify the title to summarize the whole paper more accurately.

Response: We agree with the reviewer that our manuscript’s title might be a bit misleading as (R)-DI-87 does not directly target the bacterium or staphylococcal virulence factors. Thus, we have modified the title of our revised manuscript to: “Targeting host deoxycytidine kinase mitigates Staphylococcus aureus abscess formation”.

Reviewer #1 (Public Review):

Aiming at the problem that Staphylococcus aureus can cause apoptosis of macrophages, the authors found and verified that drug (R)-DI-87 can inhibit mammalian deoxycytidine kinase (dCK), weaken the killing effect of staphylococcus aureus on macrophages, and reduce the apoptosis of macrophages. And increase the infiltration of macrophages to the abscess, thus weakening the damage of Staphylococcus aureus to the host. This work provides new insights and ideas for understanding the effects of Staphylococcus aureus infection on host immunity and discovering corresponding therapeutic interventions. This work is important and groundbreaking.

Comments on revised version:

The changes made by the authors addressed my previous concerns about the manuscript and greatly improved the quality of the article.

Author Response

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

Reviewer #1 (Public Review):

Summary:

The manuscript by Adelus and colleagues investigates the snRNA sequencing of endothelial cells isolated from deceased heart donor aortic trimmings. From n=6 donors, the authors have identified 5 distinct endothelial cell (EC) populations. The expression levels of a set of genes are different among the different donors and different EC clusters. Furthermore, treatment with IL1B, TGFB, or ERGsi decreased the proportion of some of these clusters and increased others, with some migratory and ECM-producing capacity. Another interesting observation in this study is that IL-1B alone induces a shift in the clusters and that is different from the TGFB-induced cells. However, ex vivo analyses showed most of the TGFB-induced population matched the in vitro observations. Another interesting finding of the work is that the authors detected SNPs linked to chromatin accessibility to the set of genes identified within these EC populations.

Strengths:

Overall, the work is intriguing and has some novel aspects to it, especially the link between ECderived EndMT in culture and comparing that with ex vivo atherosclerotic samples.

In summary, we thank we thank Reviewer #1 for raising questions that prompted new speculations and clarifications of our data. We hope this Reviewer will now find our revised manuscript suitable for publication.

Weaknesses:

The experiments are lacking in controls, the purity of the isolation, and the use of multiple donors (deceased hearts) to draw conclusions. The lack of validation of the work is a concern.

We thank Reviewer #1 for raising these concerns. Controls were not available in the public in vivo data, likely due to the systemic nature of coronary artery disease (CAD) and the logistical difficulty in obtaining arterial samples from healthy participants. With respect to our in vitro data, controls were included in the design. We agree that it is critical to validate functions of endothelial cell (EC) populations with functional studies, and this is the subject of ongoing and future work. Regarding asymmetry of donors, we aimed to have at least three replicate donors per condition. In the study design, we had to load genetically different donors per 10x lane, which is why we utilized different donors for each condition. We address the purity of isolation in our response to Reviewer #2 below.

Reviewer #2 (Public Review):

This study by Adelus et al. profiled the transcriptome and chromatin accessibility in cultured human aortic endothelial cells (ECs) at single-cell resolution. They also stimulated these cells with EC-activating agents, such as IL1b, TGFB2, or si-EGR, to knock down this master transcription factor in ECs. The results show a subpopulation, EC3, with the highest plasticity and sensitivity to perturbations. The authors also reviewed and meta-analyzed three independent publicly available scRNA-seq datasets, identifying two distinct EC subpopulations. Additionally, they aligned CAD-related SNPs with open chromatin regions in EC subpopulations. This study provides fundamental evidence to enrich our understanding of vascular ECs and highlights potential subpopulations that may contribute to health and diseases. The work exhibits the potential impact in the field. While the manuscript is comprehensive, there are some concerns that should be addressed.

(1) My major concern is whether EC4 is derived from ECs. It seems that EC4 showed a lesser reaction to those perturbations and had lower expression levels of EC marker genes. Did the authors evaluate the purity of their isolated HAECs? Please discuss the potential cell lineage mapping of EC4.

We thank Reviewer #2 for raising the question on the purity of isolation. We have now included this in the Discussion:

“A major question raised by this work is the origin of cells in the mesenchymal cluster EC4. We originally hypothesized this cluster was the result of EndMT, which led to our investigations as to whether we could leverage EndMT-promoting exposures (IL1B, TGFB2, siERG) in vitro observe an expansion of treated cells in the EC4 population. To our surprise, the EC4 population did not expand. If anything, these exposures reduced the proportion of cells in ECs (Figure 4). Nonetheless, it remains a possibility that EC4 represents cells that had undergone EndMT in vivo prior to culture and that the exposures we presented in vitro were not sufficient to elicit a complete EndMT transition. Another viable hypothesis is that cells in EC4 are of SMC origin and have persisted in culture alongside their EC counterparts. Cells used in this study were isolated by luminal collagenase digestion of explanted aortic segments and were tested at early passage for EC phenotypic markers including VWF expression, cobblestone morphology, and uptake of acetylated LDL. Notably, these rigorous metrics to ensure pure EC isolation occurred prior to our group’s studies. In addition, if some of the isolated cells had undergone EndMT in vivo prior to isolation, it would be nearly impossible to distinguish their cell of origin after isolation since their collective molecular phenotypes would appear as an SMC. Without lineage tracing, which is currently not possible in human tissue explants, it would not be possible to distinguish cell origin. Nonetheless, this remains an important issue that is the subject of ongoing investigations. What we can confidently discern from these data is that these distinct cell subpopulations respond differently to the disease-relevant exposures of IL1B, TGFB2, and ERG depletion.”

(2) Although all the donors are de-identified, is there any information about the severity of their vascular impairment, particularly in the case of patient 5, who exhibits the unique EC5?

All donors are de-identified, and we only have access to their genotypes. We have now clarified this in Methods, “Tissue Procurement and Cell Culture”:” Primary HAECs were isolated from eight de-identified deceased heart donor aortic trimmings (belonging to three females and five males of Admixed Americans, European, and East Asian ancestries) at the University of California Los Angeles Hospital as described previously (42) (Table S7 in the Data Supplement). The only clinically relevant information collected for each donor was their genotype (Methods, “Genotyping and Multiplexing Cell Barcodes for Donor Identification”).”

(3) The meta-analysis of the published datasets is comprehensive. The identified EC heterogeneity corresponds to their in vitro data. I am wondering, in terms of transcriptome, is there any similarity between endo1 and EC1/EC2, and also endo2 and EC3/EC4?

This was addressed in Results, “Ex Vivo-derived Module Score Analysis Reveals Differences among In Vitro EC Subtypes and EndMT Stimuli”: “Cells scoring high for Endo1 are concentrated in the in vitro EC1 cluster, while cells scoring high in Endo2 are concentrated to the in vitro EC3 locale (Figure S7B-E in the Data Supplement).”

(4) The in vitro data indicates that EC3 shows the highest plasticity and sensitivity to perturbations, which may act as the major subtype of ECs responding to risk factors. It's very interesting that CAD-related SNPs do not seem to be enriched in EC3. Please discuss this discrepancy.

We thank Reviewer #2 for bringing up this interesting point, which we have now included in our Discussion: “While EC3 was found to be more sensitive to perturbations in our in vitro experiments, we did not expect to see CAD-related SNPs enriched in EC3 because plasticity does not necessarily imply a pathological process. Moreover, while EC3 and EC4 both have mesenchymal phenotypes, EC3 may represent a reversible state that is lacking in EC4. This hypothesis would explain the enrichment of EC4, but not EC3, in CAD-related SNPs.”

(5) The last sentence in the legend of Figure 1 seems incomplete: 'Module scores are generated for each cell barcode with Seurat function AddModuleScore().'

We have made changes to this sentence so that it now reads: “Module scores are generated for each cell barcode with the Seurat function AddModuleScore().”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

The manuscript by Adelus and colleagues investigates the snRNA sequencing of endothelial cells isolated from deceased heart donor aortic trimmings. From n=6 donors, the authors have identified 5 distinct endothelial cell (EC) populations. The expression levels of a set of genes are different among the different donors and different EC clusters. Furthermore, treatment with IL1B, TGFB, or ERGsi decreased the proportion of some of these clusters and increased others, with some migratory and ECM-producing capacity. Another interesting observation in this study is that IL-1B alone induces a shift in the clusters and that is different from the TGFB-induced cells. However ex vivo analyses showed most of the TGFB-induced population are the ones that matched the in vitro observations. Another interesting finding of the work is that the authors detected SNPs linked to chromatin accessibility to the set of genes identified within these EC populations. Overall, the work is intriguing and has some novel aspects to it, especially the link between EC-derived EndMT in culture and comparing that with ex vivo atherosclerotic samples. However, the experiments are lacking in controls, the purity of the isolation, and the use of multiple donors (deceased hearts) to draw conclusions. The lack of validations for the work is a huge concern. Additional major and minor concerns are:

Major concerns:

(1) Abstract: line 15: ECs are a major cell type in atherosclerosis progression - That is a bold statement: What about macrophages and VSMCs?

We have made changes to this sentence so that it now reads: “Endothelial cells (ECs), macrophages, and vascular smooth muscle cells (VSMCs) are major cell types in atherosclerosis progression, and heterogeneity in EC sub-phenotypes are becoming increasingly appreciated.”

(2) Methods: The cells were isolated from the deceased heart by a device? What kind of device? Is it a standard method, showing a figure or data suggesting the purity of the isolates. Also, the authors mentioned that they assessed EC function, but no single figure suggests that. Why were the cells treated with fibronectin?

We thank Reviewer #1 for bringing this to our attention. We did not isolate and identify the cells ourselves. This was done in a prior study as described in reference 41. The only function of the device was to hold the aortic explanted tissue in place so the luminal surface of the ECs could be digested with collagenase. We have made edits to clarify these points in Methods, “Tissue Procurement and Cell Culture”: “HAECs were isolated from the luminal surface of the aortic trimmings using collagenase, and identified by Navab et al. using their typical cobblestone morphology, presence of Factor VIII-related antigen, and uptake of acetylated LDL labeled with 1,1’-dioctadecyl-1-3,3,3’,3’-tetramethyl-indo-carbocyan-ine perchlorate (Di-acyetl-LD) (42).”

(3) Why did the authors elect to treat each donor cell with different treatment types and different concentrations, also why 1ng/ml of IL-1B?

We have addressed the study design asymmetry above. We chose the treatments because we questioned whether HAECs responded heterogeneously to these stimuli. We were interested in using these stimuli, because they have previously been used in vitro to induce EndMT and/or inflammation, two major pathophysiological processes in CAD. This is outlined in the Introduction: “We also quantified single cell responses to three perturbations known to be important in EC biology and atherosclerosis. The first was activation of transforming growth factor beta (TGFB) signaling, which is a hallmark of phenotypic transition and a regulator of EC heterogeneity (20, 30). The second was stimulation with the pro-inflammatory cytokine interleukin-1 beta (IL1B), which has been shown to model inflammation and EndMT in vitro (31-35), and whose inhibition reduced adverse cardiovascular events in a large clinical trial (36). The third perturbation utilized in our study was knock-down of the ETS related gene (ERG), which encodes a transcription factor of critical importance for EC fate specification and homeostasis (37-41).”

(4) The justification for comparing the EC population in ERGsi is unclear? This was detected as the highest in EC2 but EC2 is not the main cell type across the donors.

We include a justification for comparing the EC populations with siERG in the Introduction:

“There are notable benefits and limitations for studying heterogeneity using in vitro and in vivo approaches in atherosclerosis research. In vitro approaches provide unique opportunities for interrogating consequences of genetic and chemical perturbations in highly controlled environments and are adept at identifying mechanistic relationships on accelerated timelines.”

…and…

“We… quantified single cell responses to three perturbations known to be important in EC biology and atherosclerosis…The third perturbation utilized in our study was knock-down of the ETS related gene (ERG), which encodes a transcription factor of critical importance for EC fate specification and homeostasis (37-41).”

Notably, we found the highest proportion of cells in EC3 with siERG, not EC2:

The one cluster exhibiting increased proportions of cells upon EndMT perturbations was EC3, with 3 of 4 EC IL1B-exposed donors having increased proportions in EC3 (p = 0.08 by 2-sided paired t-test; Figure 3A), 4 of 5 TGFB2-exposed donors having increased proportions (p = 0.04 by 2-sided paired t-test; Figure 3A), and 3 of 3 donors having increased EC3 proportions upon ERG knock-down (Figure 3B).

(5) The different proportions of clusters per donor and their responses are different. These donors are from deceased hearts, could the postmortem induce changes in the ECs? The presence of SMC pathways in their analysis may indicate SMC contamination within the isolation rather than EndMT?

We have now included the possibility of postmortem effects in the Discussion:

“We cannot exclude the possibility that EC3 is an EndMT cluster, although we would have expected more significant deviation from clusters EC1 and EC2. It is also possible that the postmortem could induce changes in the ECs, or that the duration and doses of perturbations chosen were not sufficient to elicit complete EndMT.”

As aforementioned, we addressed the purity of isolation within the Discussion.

(6) Figure 4A is confusing, what do the dots indicate and the intersection size mean? What is the difference between Figure 4 C and 4 E?

We have added a description of rows and columns to the legend for Figure 4A:

“(A), Upset plots of up- and down-regulated DEGs across EC subtypes with siERG (grey), IL1B (pink), and TGFB2 (blue). Upset plots visualize intersections between sets in a matrix, where the columns of the matrix correspond to the sets, and the rows correspond to the intersections. Intersection size represents the number of genes at each intersection.”

Figure 4E depicts up- and down-regulated DEGs that are mutually exclusive and shared between IL1B and siERG in EC3, whereas Figure 4C depecits up- and down-regulated DEGs with IL1B alone compared to siSCR in EC2, EC3, and EC4. This is described within the legend for Figure 4C and Figure 4E:

“C), PEA for EC2-4 up- and down-regulated DEGs with IL1B compared to control media… (E), PEA comparing up- and down-regulated DEGs that are mutually exclusive and shared between IL1B and siERG in EC3.”

(7) VSMCS 5 in Figure 5 is interesting, but it could be contaminated with SMCs in your EC population and they are SMCs indeed with some mesenchymal transdifferentiation?

As abovementioned, we addressed the purity of isolation within the Discussion.

Minor concerns:

(1) All growth supplements, kits, and reagents should be provided with their sources and catalogue numbers.

Sources and catalogue numbers have now been added to the following Methods sections:

“Tissue Procurement and Cell Culture”: “Cells were grown in culture in M-199 (ThermoFisher Scientific, Waltham, MA, MT-10-060-CV) supplemented with 1.2% sodium pyruvate (ThermoFisher Scientific, cat. no. 11360070), 1% 100X Pen Strep Glutamine (ThermoFisher Scientific, cat. no. 10378016), 20% fetal bovine serum (FBS, GE Healthcare, Hyclone, Pittsburgh, PA), 1.6% Endothelial Cell Growth Serum (Corning, Corning, NY, cat. no. 356006), 1.6% heparin, and 10µL/50 mL Amphotericin B (ThermoFisher Scientific, cat. no. 15290018). HAECs at low passage (passage 3-6) were treated prior to harvest every 2 days for 7 days with either 10 ng/mL TGFB2 (ThermoFisher Scientific, cat. no. 302B2002CF), IL1B (ThermoFisher Scientific, cat. no. 201LB005CF), or no additional protein, or two doses of small interfering RNA for ERG locus (siERG; Table S18 in the Data Supplement), or randomized siRNA (siSCR; Table S18 in the Data Supplement).”

…and…

“siRNA Knock-down, qPCR, and Western Blotting”: “Knockdown of ERG was performed as previously described (40) using 1 nM siRNA oligonucleotides in OptiMEM (ThermoFisher Scientific, cat. no. 11058021) with Lipofectamine 2000 (ThermoFisher Scientific, cat. no. 11668030).”

(2) The quantification of western blot how?

Methods, “siRNA Knock-down, qPCR, and Western Blotting” now reads: “Western blots were quantified using ImageJ (76).”

(3) All the supplemental figures are listed incorrectly in the manuscript. For example, the authors refer to Figure S11B which should be S10. Please review the manuscript throughout to refer to the correct figures.

We thank Reviewer #1 for bringing this to our attention. Figure S4 was missing, leading to incorrectly listed supplemental figures for Figures S4-S12. Figure S4 has now been included, and Figures S4-S12 are now listed correctly within the manuscript text.

(4) Please refer to IL-1B as IL-1beta, same with TGFB.

We have left the terms as is, since it is also routine to refer to IL-1beta as IL1B, and TGFbeta as TGFB.

(5) here are typos throughout the manuscript, such as 4C, VW Fexpression, VWFand VCAM-1.

We could not locate typos “VW Fexpression” or “VWFand VCAM-1”. We do not consider “4C” a typo, as it refers to the temperature at which the centrifuge was set to in Methods, “Nuclear Dissociation and Library Preparation”: “Samples were centrifuged at 500 rcf for 5 minutes at 4C…”

(6) Please define the abbreviations: line 69 and also cite the source of the use of aSMA/PECAM1 as EndMT?

We have now included abbreviation definitions and the cited source for ECs that co-express aSMA/PECAM-1 in atherosclerotic lesions within the Introduction: “These studies have described an unexpectedly large number of cells co-expressing pairs of endothelial and mesenchymal proteins, including fibroblast activating protein/von Willebrand factor (FAP/VWF), fibroblastspecific protein-1/VWF (FSP-1/VWF), FAP/platelet-endothelial cell adhesion molecule-1 (CD31 or PECAM-1), FSP-1/CD31 (20), phosphorylation of TGFB signaling intermediary SMAD2/FGF receptor 1 (p-SMAD2/FGFR1) (22), and α-smooth muscle actin (αSMA)/PECAM-1 (23).”

(7) The changes in % cells in cluster per donor per condition in Figure 3 are interesting, have the authors observed a change of one cluster at the expense of another i.e. do they transdifferentiate into another with different treatments?

Figure 3 shows that as percent of cells in EC3 go up with TGFB or IL1B, they go down in EC4 with these treatments. This has been added to the Discussion: “Moreover, as the percent of cells in EC3 go up with TGFB or IL1B, they go down in EC4, suggesting trans-differentiation from EC4 into EC3 with these perturbations.”

(8) Functional analysis of these clusters with and without treatment is required to confirm the EndMT.

We do not claim that the cells underwent EndMT. Rather, we use pro-EndMT perturbations previously described in the literature to test whether ECs respond heterogeneously to stimuli which are relevant to CAD. We found that EC subtype was a greater determinant of cell state than treatment.

(9) No blank line at 266. The break is in the middle of the sentence, also cytoplasmic cytoplasmic ribosomal proteins (typo?).

We have revised these sentences to read: “Shared IL1B- and siERG-upregulated genes were enriched in COVID-19 adverse outcome pathway (WP4891; p-value 1.9x10-9) (52). Shared IL1B- and siERG-attenuated genes are enriched in several processes involving ribosomal proteins, including ribosome, cytoplasmic (CORUM:306; p-value 3.3x10-7), cytoplasmic ribosomal proteins (WP477; p-value 5.3x10-7), and peptide chain elongation (R-HSA-156902; pvalue 5.9x10-7) (Figure 4E).”

(10) The sentence in line 321 "These observations support ....of human, seems incomplete.

We revised these sentences to read: “Expected pathway enrichments are observed for annotated cell types, including NABA CORE MATRISOME (M5884; p-value 4.8x10-41) for fibroblasts, blood vessel development (GO:0001568; p-value 5.6x10-33) for ECs, and actin cytoskeleton organization (GO:0030036; p-value 1.3x10-15) for VSMCs (Figure S5D-G in the Data Supplement). These observations support the diverse composition of human atherosclerotic lesions.”

(11) What do the authors mean by (at least partially) line 444?

We revised this sentence to read: “In fact, the limited correlation with ex vivo data supports this interpretation.”

(12) Some unrelated data in the paper, like supplemental figure 10B and supplemental figure 11?

These data are relevant to methods, and have been kept.

Reviewer #2 (Recommendations For The Authors):

We need this work to expand our knowledge of endothelial biology. Please address my concerns to further strengthen this work.

eLife assessment

This is a fundamental resource of snRNA-seq and and chromatin accessibility data from human aortic endothelial cells (ECs), treated with relevant perturbations such as IL1b, TGFB2, or si-EGR. The authors show that ECs can be categorized by distinct subpopulations of differing plasticity. The support for the existence of these subpopulations is compelling, supported also by three publicly available scRNA-seq datasets, and differential enrichment of coronary artery disease associated SNPs in open chromatin in these subpopulations.

Reviewer #2 (Public Review):

This study by Adelus et al. profiled the transcriptome and chromatin accessibility in cultured human aortic endothelial cells (ECs) at single-cell resolution. They also stimulated these cells with EC-activating agents, such as IL1b, TGFB2, or si-EGR, to knock down this master transcription factor in ECs. The results show a subpopulation, EC3, with the highest plasticity and sensitivity to perturbations. The authors also reviewed and meta-analyzed three independent publicly available scRNA-seq datasets, identifying two distinct EC subpopulations. Additionally, they aligned CAD-related SNPs with open chromatin regions in EC subpopulations. This study provides fundamental evidence to enrich our understanding of vascular ECs and highlights potential subpopulations that may contribute to health and diseases. The work exhibits the potential impact in the field.

Comments on revised version:

I appreciate their revision, which addressed all my concerns. I understand the current technique's limitation in distinguishing bona fide cell lineages from human tissue explants, but it merits further investigation. This is because EC4 may also be involved in critical pathological processes. Again, this work established a solid foundation for exploring endothelial cell plasticity.

Author Response

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

Reviewer #1 (Recommendations For The Authors):

Revised manuscript

The authors have addressed most of my points, but I still have one outstanding concern about the statistics:

My Original Question:

I have a few concerns and questions that I would like to see addressed: 1) Figure 1L - the statistics are a little unusual here as the errors are across visual areas rather than across mice or hemispheres. This isn't ideal as ideally, we want to generalize the results across animals, not areas, and the results seem to be driven mostly by V1/RSC. I would like to see comparisons using mice as the statistical unit either in an ANOVA with areas as factors or post-hoc comparisons per area.

Author Reply:

Based on the assumption that visual cortex should respond to visual stimuli, we would have expected to find a difference between closed and open loop locomotion onset responses in all cell types in visual areas of cortex (a closed loop locomotion onset being the combination of locomotion and visual flow onset, while an open loop locomotion onset lacks the visual flow component). Thus, the first surprise was that in most cell types we found very little difference between these two locomotion onset types. Conversely, in Tlx3-positive L5 IT neurons the difference was apparent well outside of the visual areas of cortex (even though the difference was indeed strongest in V1/RSC). To quantify the extent to which closed and open loop locomotion onsets result in different activity patterns across dorsal cortex we performed the analyses shown in Figures 1L and 2. To make the point that the effect was observable on average across cortical areas, we used cortical area as a unit in Figure 1L. We have added the analysis shown in Figure 1L with mice as the statistical unit as Figure S4J and have added the ANOVA information to Table S1, as suggested.

My revised question:

The authors have only partially addressed my concerns here. I disagree with the authors that they were making a point about the effect being observable across visual areas. The primary statistical statement they are trying to make is that the similarity between open and closed-loop stimulation is different for Tlx mice, e.g. Line 122: "However, comparing locomotion onsets in mice that expressed GCaMP6 only in Tlx3 positive L5 IT neurons, we found that the activation pattern was strikingly different between closed and open loop conditions" and Line 172-3: "Thus, excitatory neurons of deep cortical layers exhibited the strongest differences between closed and open loop locomotion related activation". These statements are not correctly supported by the statistical analysis as presented in Figure 1L as it is the variability across mice that is relevant to draw this conclusion.

In the example "However, comparing locomotion onsets in mice that expressed GCaMP6 only in Tlx3 positive L5 IT neurons, we found that the activation pattern was strikingly different between closed and open loop conditions (Figure 1D)" we talk about the example mouse shown. We have not changed phrasing here.

We have, however, changed the way we talk about Figure 1L and S4J (the second example given by the reviewer), and have rephrased much of this paragraph. Please note, we have also changed Figure S4J to quantify the difference only for V1.

This is partially addressed by Figure S4J where the authors show standard-errors across mice and report statistics across mice. In Table S1 the statistical test is reported to be a bootstrap test with mice as the statistical unit, however, according to line 985 this was a non-hierarchical bootstrap test. Does this mean that the authors resampled onsets without regard to which mouse they came from to regenerate the response-curves and recalculate the correlation coefficient? Or did they directly resample from the distribution of correlation coefficient values? I suspect the latter, but for some comparisons (e.g. Tlx3 vs PV) there are only two mice in one group, yielding two correlation coefficients, and resampling 2 values 10,000 times would lead to very biased statistics. Either way the approach is far from ideal. There is also no protection against multiple-comparisons in these tests.

We have adapted Figure S4J to include only V1, where we find the largest effect (the text is adapted to reflect this) and have added individual data points as suggested in the following comment. The reviewer is correct that we created a bootstrap distribution by resampling correlation values. This means we are resampling 2, 3, 4, 6, 7, or 14 values depending on the comparison. This should now be clearer in the text. We agree that this is not ideal, but when using mice as a statistical unit, analysis is almost always underpowered. To the best of our knowledge, bootstrap resampling is the best approach to alleviate this problem. Regarding the concern for multiple comparisons: We have now adjusted the significance threshold in Figures 1L and S4J by dividing through the number of groups (here: 9 genotypes).

The ANOVA reported in Table S1 for Figure S4J isn't described in the methods so I can't say what they did and it doesn't seem to be referred to in the text and is non-significant in any case. Figure S4J also only shows summary statistics whereas individual mice should be plotted. The correct statistical test is either a one-way ANOVA with one factor (genotype) with post-hoc tests between the Tlx3 genotype and the others with suitable multiple-comparisons corrections (this may be the non-significant test in table S1). Alternatively, a linear mixed effects model with Genotype as a fixed effect and Mouse as a random intercept term. This approach is more powerful as it would allow them to use data from all locomotion onsets, but it may struggle to fit datasets with only 2 members for certain genotypes. If they wish to make the more extended point that the pattern across visual areas differs between Tlx3 and other mice they could include 'Area' as another (fixed) factor in the design and look for an interaction with Genotype.

The ANOVA was indeed a one-way ANOVA with one factor. We have added this information to the methods. As suggested, we have added individual data points to Figure S4J.

I also agree with the other reviewers that the presentation of standard-errors in Figures 1F-K and elsewhere is somewhat misleading as these are s.e.m. across onsets without taking into account the hierarchical nature of the data. Across mice s.e.m. would give a more accurate view of the variability in the data across the population. I also understand that first averaging across onsets within mice before taking a grand-average throws away a lot of data and s.e.m.s will be considerably larger. The authors should consider linear mixed effects models as an optimal solution for estimating s.e.m. If this is not feasible then the authors could consider showing data from individual mice in a supplementary figure or at least reporting the number of onsets that came from each mouse.

We have now changed all plots in which we show time course data of widefield calcium imaging to show a hierarchical bootstrap estimate of mean and 90% confidence interval of the mean estimate.

Reviewer #2 (Recommendations For The Authors):

Congratulations to the authors on the revision! The revised article has substantially improved, and I have no further comments. I am particularly reassured by the new hierarchical bootstrap analyses as well as by the new analysis with mouse as a statistical unit that reproduces the key finding from the analyses with region as a statistical unit. Moreover, the authors added a vehicle control condition which does not yield any results. Therefore, I have no further methodological concerns and removed my mention of this previous weakness from my public review. Also, the readability of the manuscript has much improved in the revised version. Congratulations again on this important work!

We thank the reviewer for the help in improving the manuscript.

Reviewer #3 (Recommendations For The Authors):

Comments on rebuttal:

(1) It is greatly appreciated that the authors have improved aspects of their statistics, I have revised my comments accordingly.

We are happy to hear.

(2) However, I should clarify my comments regarding statistical concerns were not merely pertaining to a given Figure (e.g. Figure 1) I was only using it as an example. The authors have redone aspects of their analysis using N = number of mice (for statistics/trace figures), but is there a reason they cannot do this for other problematic figures/traces in the manuscript?

Prompted also by reviewer 1, we have changed all time course plots in the manuscript to show a hierarchical bootstrap estimate of mean and 90% confidence interval of mean.

Using mice as a statistical unit throughout the manuscript unfortunately is not viable in most cases, as we simply do not have enough mice in our dataset and statistical tests based on mice would be underpowered. The manuscript currently contains data from 77 mice, and we would likely need multiples of that to do statistics over mice.

For Figure 1 - I do take the point why regions are being used as the independent N (though the authors justification should be made more clearly in the manuscript) making an N of 12 (though I am less clear why the same region across 2 hemispheres is counted as 2 Ns instead of 1; are they really independent?). However, I am less clear as to the choice in N in other figures. Could the authors clarify this more explicitly in the manuscript.

We use regions as a statistical unit in Figures 6 and 7, S6-S8. Regarding the independence of hemispheres, this depends on cell type and region. E.g. activity in left V1 exhibits a higher correlation with activity left V2am than with right V2 (see Figure 5). On average callosal pairs exhibit correlation levels comparable to near cortical neighbors. See also, other work on the topic, for example (Calhoun et al., 2023).

Regarding choice of N in other figures, this is either “recording session” or “pairs of regions”. We have made this clearer in the figure legends. In the case of testing using recording sessions, the idea is that each recording session constitutes a measurement. Measurements in the same mouse are not independent, and hence we use hierarchical bootstrap for all testing on recording sessions. The choice of “pairs of regions” for the correlation analysis follows from the use of regions as a statistical unit.

(3) Regarding using N = locomotion onsets (or other definitions other than N = mice) when deriving trace averages/SEMs (for example, as in Figure 1) is visually misleading for the reader as it masks the true variability of the data, and even more misleading given that the authors do necessarily use that definition of N in their statistical tests associated with the data (as the authors commented). Whilst the authors have shown some traces with N=mice for some data, is there a reason they cannot do this for all figures in the manuscript? At the very least the practice of using other definitions of N for the purpose of showing trace averages/SEMs should be justified in the MS.

We have replaced all time course plots that used SEM over events (for example locomotion onsets or visual stimuli) with a hierarchical bootstrap estimate of mean and 90% confidence interval of the mean throughout the manuscript. See also response to comment 2 above, and to reviewer 1, comment 4.

References

Calhoun, G., Chen, C.-T., Kanold, P.O., 2023. Bilateral widefield calcium imaging reveals circuit asymmetries and lateralized functional activation of the mouse auditory cortex. Proc. Natl. Acad. Sci. U. S. A. 120, e2219340120. https://doi.org/10.1073/pnas.2219340120

eLife assessment

This important study uses calcium imaging in mice to advance our understanding of the effect of antipsychotic drugs on neural functioning. The evidence supporting the conclusions is convincing, and this work will be of interest to neuroscientists working on visual processing and psychosis researchers.

Reviewer #1 (Public Review):

The authors present a study of visuo-motor coupling primarily using wide-field calcium imaging to measure activity across the dorsal visual cortex. They used different mouse lines or systemically injected viral vectors to allow imaging of calcium activity from specific cell-types with a particular focus on a mouse-line that expresses GCaMP in layer 5 IT (intratelencephalic) neurons. They examined the question of how the neural response to predictable visual input, as a consequence of self-motion, differed from responses to unpredictable input. They identify layer 5 IT cells as having a different response pattern to other cell-types/layers in that they show differences in their response to closed-loop (i.e. predictable) vs open-loop (i.e. unpredictable) stimulation whereas other cell-types showed similar activity patterns between these two conditions. They also analyzed the responses to visuomotor prediction errors obtained by briefly pausing the display while the mouse is running, causing a negative prediction error, or by presenting an unpredicted visual input causing a positive prediction error. Surprisingly, they find that presentation of a visual grating actually decreases the responses of L5 IT cells in V1. They interpret their results within a predictive coding framework that the last author has previously proposed. The response pattern of the L5 IT cells leads them to propose that these cells may act as 'internal representation' neurons that carry a representation of the brain's model of its environment. Though this is rather speculative. They subsequently examine the responses of these cells to anti-psychotic drugs (e.g. clozapine) with the reasoning that a leading theory of schizophrenia is a disturbance of the brain's internal model and/or a failure to correctly predict the sensory consequences of self-movement. They find that anti-psychotic drugs strongly enhance responses of L5 IT cells to locomotion while having little effect on other cell-types. Finally, they suggest that anti-psychotics reduce long-range correlations between (predominantly) L5 cells and reduce the propagation of prediction errors to higher visual areas and suggest this may be a mechanism by which these drugs reduce hallucinations/psychosis.

This is a large study containing a screening of many mouse-lines/expression profiles using wide-field calcium imaging. Wide-field imaging has its caveats, including a broad point-spread function of the signal and susceptibility to hemodynamic artifacts, which can make the interpretation of results difficult. The authors acknowledge these problems and directly address the hemodynamic occlusion problem. It was reassuring to see supplementary 2-photon imaging of soma to complement this data-set, even though this is rather briefly described in the paper. Some comparisons in the paper are underpowered as a result of including only a small number of mice (e.g. the PV, Ntsr1 and Cux2 mice) and results involving these mice should be cautiously interpreted, but in general the results are robust. Overall the paper's strengths are its identification of a very different response profile in the L5 IT cells compared to other layers/cell-types which suggests an important role for these cells in handling integration of self-motion generated sensory predictions with sensory input. The interpretation of the responses to anti-psychotic drugs is more speculative but the result appears robust and provides an interesting basis for further studies of this effect with more specific recording techniques and possibly behavioral measures.

Author Response

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

Reviewer #1 (Public Review):

The modeling approaches are very sophisticated, and clearly demonstrate the selective nature of acute ketamine to reduce the impact of trial losses on subsequent performance, relative to neutral or gain outcomes. The authors then, not unreasonably, suggest that this effect is important in the context of the negative bias in interpreting events that is prominent in depression, in that if ketamine reduces the ability of negative outcomes to alter behavior, this may be a mechanism for its rapid acting antidepressant effects.

However, there is a very strong assumption in this regard, as shown by the first sentence of the discussion which implies this is a systematic study of ketamine's acute antidepressant effects. In actuality, this is a study of the acute effects of ketamine on reinforcement learning (RL) modeled parameters. A primary concern here is that an effect presented as a "robust antidepressant-like behavioral effect" should be more enduring than just an alteration during the acute administration. As it is, the link to an "anti-depressant effect" is based solely on the selective effects on losses. This is not to say this is not an interesting observation, worthy of exploration. It is noted that a similar lack of enduring effects on outcome evaluation is observed in humans, as shown in supplemental fig. S4, but there is not accompanying citation for the human work.

We agree with the reviewer that the way we linked the study results to ketamine’s antidepressant action can be misleading and based on a rather strong assumption which was not systematically tested in the study. We made the following changes to the manuscript:

(1) These results constitute a rare report of a robust antidepressant-like behavioral effect produced by therapeutic doses of ketamine during acute phase (<1 hour) after injection (Introduction, 3rd paragraph, line 8-9 in the original manuscript).

Changed to: These results constitute a rare report of an acute effect of therapeutic dose of ketamine on the processing of affectively negative events during dynamic decision-making.

(2) We clarified in the Discussion that our study is to gain insights into, but not a systematic investigation of ketamine’s antidepressant action as follows:

(2.1) A sentence was added (1st paragraph of Discussion): Using a token-based decision task and extensive computational modeling, we examined the behavioral modulation induced by therapeutic doses of ketamine to gain insights into possible early signs of ketamine’s antidepressant activity.

(2.2) Consistent with the findings from humans, ketamine’s effect on outcome evaluation was acute and did not last over subsequent days (Supplemental Figure S4) (Discussion, 2nd paragraph, line 6-7 in the original manuscript).

Changed to: While ketamine’s antidepressant effect is reported to be sustained over a week of period (5), ketamine’s effect on outcome evaluation was acute and did not last over subsequent days (Supplemental Figure S4). This discrepancy might be attributable to the possible differences in the state of brain network between healthy subjects and those with depression as well as the type of measures taken to assess ketamine’s effect.

(2.3) A sentence was added (Discussion, last sentence of the 2nd paragraph) : Nevertheless, systematic studies are required to understand whether the reduced aversiveness to loss in our task might share the same mechanisms that underlie ketamine’s antidepressant action.

One question that comes to mind in terms of the selectivity observed is whether similar work has been done to examine the acute effects of any other drugs. If ketamine is unique in this regard, that would be quite interesting.

We think this is an interesting idea. However, comparing ketamine’s effect to that of other drugs is not the scope of the current study. We hope that we will be able to answer this question with future studies.

Reviewer #2 (Public Review):

Oemisch and Seo set out to examine the effects of low-dose ketamine on reinforcement learning, with the idea that alterations in reinforcement learning and/or motivation might inform our understanding of what alterations co-occur with potential antidepressant effects. Macaques performed a reinforced/punished matching pennies task while under effects of saline or ketamine administration and the data were fit to a series of reinforcement learning models to determine which model described behavior under saline most closely and then what parameters of this best-fitting model were altered by ketamine. They found a mixed effect, with two out of three macaques primarily exhibiting an effect of ketamine on processing of losses and one out of three macaques exhibiting an effect of ketamine on processing of losses and perseveration. They found that these effects of ketamine appeared to be dissociable from the nystagmus effects of the ketamine.

The findings are novel and the data suggesting that ketamine is primarily having its effects on processing of losses (under the procedures used) are solid. However, it is unclear whether the connection between processing of losses and the antidepressant effects of ketamine is justified and the current findings may be more useful for those studying reinforcement learning than those studying depression and antidepressant effects. In addition, the co-occurrence of different behavioral procedures with different patterns of ketamine effects, with one macaque tested with different parameters than the other two exhibiting effects of ketamine that were best fit with a different model than the other two macaques, suggests that there may be difficulty in generalizing these findings to reinforcement learning more generally.

(1) First, the authors should be more explicit and careful in the connection they are trying to make about the link between loss processing and depression. The authors call their effect a "robust antidepressant-like behavioral effect" but there are no references to support this or discussion of how the altered loss processing would relate directly to the antidepressant effects.

We agree with the reviewer’s point on the way we made the connection between the study results and ketamine’s antidepressant action. This concern overlaps with the reviewer #1’s concern. Please refer to our response 2, 2-1, 2-2 and 2-3.

(2) It appears that the monkey P was given smaller rewards and punishers than the other two monkeys and this monkey had an effect of ketamine on perseveration that was not observed in the other two monkeys. Is this believed to be due to the different task, or was this animal given a different task because of some behavioral differences that preceded the experiment? The authors should also discuss what these differences may mean for the generality of their findings. For example, might there be some set of parameters where ketamine would only alter perseveration and not processing of losses?

Although the best-fitting ketamine model for monkey P includes an additional element – perseveration, we believe that monkey P’s baseline behavior and ketamine’s effect are not significantly different from the other two monkeys for the following reasons.

First, monkey P was the first animal that we tested ketamine’s effect, and therefore we aimed to match the other two monkeys’ baseline behavior similar to monkey P’s behavior in order to reduce variability in ketamine’s effect potentially attributable to the difference in baseline behavior before pharmacological manipulation. We had to adjust the payoff matrix for the subsequent animals (Y and B) because these monkeys were more sensitive to loss, and seldom chose “risky” target (yielding loss). In order to make the other two monkeys’ behavior similar to that of monkey P, we adjusted the asymmetry between the risky and the safe target in the way that loss (neutral) outcome occurred from the safe (risky) target as well. Eventually, this adjustment made the baseline behavior similar across all three monkeys. The goal of the study was to reliably measure the ketamine’s effect, and not to study individual differences that can naturally occur with the same task parameters. Therefore, we believe that the adjustment of payoff matrix helped to reliably detect ketamine’s effect starting from the common baseline behavior.

Second, the best-fitting model for monkey P (K-model 7) and that for the other two monkeys (K-model 4) make very similar predictions both qualitatively and quantitatively as are seen in the revised Figure 4. The parameters for outcome values estimated from these two models in monkey P are very similar as is seen in the revised Table 3. In addition, the difference in BIC between the model which includes only perseveration modulation (K-model 6) and the model incorporating outcome value modulation as well (K-model 7) is 441, whereas the difference in BIC between K-model 7 and the model that includes only outcome value modulation (K-model 4) is as small as 4. These BIC results indicate that the variability explained by ketamine’s modulation of outcome evaluation is remarkably larger that that explained by its modulation of perseveration in monkey P.

Therefore, we conclude that ketamine’s effect was not significantly different between monkey P and the other two monkeys. We clarified this in the revised manuscript by adding the following paragraph in the Result section:

“Unlike monkey Y and B, the best-fitting model for monkey P indicated that ketamine increased overall tendency to switch choice in addition to outcome-dependent modulation of outcome evaluation. However, BIC differed only slightly (dBIC = 3.99) between the best-fitting (K-model 7) and the second-best model (K-model 4) and the model predictions for choice behavior were very similar both qualitatively and quantitatively (Table 3, Figure 4). We conclude that the behavioral effects of ketamine were consistent across all three monkeys.”

(3) The authors should discuss whether the plasma ketamine levels they observed are similar to those seen with rapid antidepressant ketamine or are higher or lower.

We added a sentence in the first paragraph of the Result section as follows with a reference.

“Plasma concentration and its time course over 60 minutes were also comparable to those measured after 0.5mg/kg in human subjects (35).”

(35) Zarate CA, Brutsche N, Laje G, Luckenbaugh DA, Venkata SLV, Ramamoorthy A, et al (2012): Relationship of ketamine’s plasma metabolites with response, diagnosis, and side effects in major depression. Biol Psychiatry, 72: 331-338.

(4) For Figure 4 or S3, the authors should show the data fitted to model 7, which was the best for one of the animals.

We added the parameters and model predictions from both K-model 7 and K-model 4 for monkey P to help comparison between two models in Table 3, and Figure 4. Revised Table 3 and Figure 4 are as follows:

Author response table 1.

Maximum likelihood parameter estimates of the best models for saline and ketamine sessions.

In all three animals, the model incorporating valence-dependent change in outcome evaluation best fit the choice data from ketamine sessions with (K-model 7 in the parenthesis, P) or without (K-model 4, P and Y/B) additional change in the tendency of choice perseveration (Figure 3, Table 3).

Author response image 1.

ketamine-induced behavioral modulation simulated with differential forgetting model (for saline session) and best-fitting K-model (for ketamine session).

eLife assessment

The authors use reinforcement learning modeling to study the alterations following acute ketamine in macaques. The evidence supporting the conclusion that ketamine reduces the impact of losses vs. neutral/gains is solid. In this version of this valuable study, the authors make more measured interpretations about the relationship between the processing of losses and ketamine's antidepressant effects.

Reviewer #1 (Public Review):

Oemisch and Seo use sophisticated reinforcement learning (RL) modeling to show that acute ketamine reduces the strength impact of losses vs neutral/gains on the subsequent trial performance of a token-based biased matching-pennies task. In this version, the authors make more measured interpretations about the potential relevance of their results to ketamine's antidepressant effects for the most part.

My prior review emphasized what I considered to be an over-interpretation of the relevance of their data (that I find interesting and of value) to mechanisms of action of ketamine's antidepressant effects. The authors have corrected those excesses exception for the last sentence of the introduction, which continues to suggest they are studying both mechanisms of antidepressant actions as well as the pathophysiology of depression.

Reviewer #2 (Public Review):

Oemisch and Seo set out to examine the effects of low-dose ketamine on reinforcement learning, with the idea that alterations in reinforcement learning and/or motivation might inform our understanding of what alterations co-occur with potential antidepressant effects. Macaques performed a reinforced/punished matching pennies task while under effects of saline or ketamine administration and the data were fit to a series of reinforcement learning models to determine which model described behavior under saline most closely and then what parameters of this best-fitting model were altered by ketamine. They found a mixed effect, with two out of three macaques primarily exhibiting an effect of ketamine on the processing of losses and one out of three macaques exhibiting an effect of ketamine on processing losses and perseveration. They found that these effects of ketamine appeared to be dissociable from the nystagmus effects of the ketamine.

The findings are novel, and the data suggesting that ketamine primarily affects on the processing of losses (under the procedures used) are solid. However, it is unclear whether the connection between the processing of losses and the antidepressant effects of ketamine is justified, and the current findings may be more useful for those studying reinforcement learning than those studying depression and antidepressant effects. In addition, the co-occurrence of different behavioral procedures with different patterns of ketamine effects, with one macaque tested with different parameters than the other two exhibiting effects of ketamine that were best fit with a different model than the other two macaques, suggests that there may be difficulty in generalizing these findings to reinforcement learning more generally.

(1) First, the authors should be more explicit and careful in the connection they are trying to make about the link between loss processing and depression. The authors call their effect a "robust antidepressant-like behavioral effect." However, there are no references to support this or discussion of how the altered loss processing would relate directly to the antidepressant effects. A few statements about a link to antidepressant effects have been removed or moderated, but many remain, including those in the abstract. The authors provide little to no support for this link, so the current version represents solid evidence for an effect on loss processing and incomplete or weak evidence for an antidepressant effect.

(2) It appears that the monkey P was given smaller rewards and punishers than the other two monkeys, and this monkey had an effect of ketamine on perseveration that was not observed in the other two monkeys. This may be due to this monkey being trained and tested before the other animals, but it does raise the issue of the generality of the authors' findings. It seems possible that the procedures used for the other two monkeys (with no deviation at all) might support the best-fit model that the authors favor. However, if changes in the size of the rewards and punishments suddenly make ketamine affect perseveration, then it suggests that ketamine's effect is highly parameter-specific. For example, might there be some parameters where ketamine would only alter perseveration and not loss processing?

Author Response

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

To Reviewer #1

We sincerely appreciate the constructive and insightful comments provided by the reviewer. Their valuable suggestions have been meticulously considered, leading to comprehensive modifications within the article.

In addition, we want to stress that we have implemented a significant additional modification by introducing a new figure (Fig. 6). This figure highlights the collaborative impact of FMRP and Map1B on the microtubular structure of migrating neurons. We firmly believe that this molecular elucidation of the migration phenotype constitutes a noteworthy addition to our work.

Public Review

(1) We have taken the necessary steps to enhance the material and methods section of our neuronal migration analysis. We apologize for any initial lack of detail, including the omission of information on sinuosity index and directionality radar. Regarding the query about speed, we want to clarify that it indeed encompasses the percentage of pausing time. The speed is calculated by dividing the total distance traveled by the cell by the total time it migrated.

(2) We would like to provide a clarification regarding the statistical analysis in our figures. The figures now represent the median, and the legend indicates the median along with the interquartile range. This approach is in line with the use of non-parametric analysis for variables that do not adhere to a normal distribution. Regrettably, in the previous version, there was an oversight in the figure legends where the mean, along with the standard error of the mean, was incorrectly stated instead of the intended representation of the median. We sincerely apologize for any confusion this may have caused. Moving forward, the corrected legend now accurately reflects the statistical measures used in the analysis.

The global Kruskal Wallis analysis, followed by Dunn’s post hoc analysis, does indeed indicate that Fmr1 KD globally replicates the Fmr1-null phenotype. However, we concur with the reviewer's point regarding directionality, and we apologize for any lack of precision in the initial version. Upon further analysis, we have identified a significant difference in directionality (Fisher test p < 0.001). This more pronounced directionality defect in the KD could potentially be indicative of a lack of compensation, a factor that may not be at play in the Fmr1 null context. We appreciate the opportunity to address this issue and our revised version includes the necessary details to accurately convey these findings.

(3) We appreciate the referee's agreement with our perspective.

(4) In response to the recommendations from all referees, we have expanded both the introduction and discussion sections of our manuscript. The initial brevity of these sections was due to the short format we had initially chosen. We believe that these expansions contribute to a more comprehensive and nuanced presentation of our work, addressing the concerns raised by the referees.

Recommendations for the authors

The time stamp and scale bars were added.

The median versus mean issue is addressed above.

Figure numbering has been corrected (sorry for the mistake). The efficiency of CK is defined in the Mat and Met section.

To Reviewer #2

Public review

We express our gratitude to the referee for their positive appreciation of our work. We have carefully considered their suggestions and have modified the article accordingly.

In addition, as said to Referee #1, we want to stress that we have implemented a significant additional modification by introducing a new figure (Fig. 6). This figure highlights the collaborative impact of FMRP and Map1B on the microtubular structure of migrating neurons. We firmly believe that this molecular elucidation of the migration phenotype constitutes a noteworthy addition to our work.

Recommendations for the authors

(1) In light of the referee's recommendation, we conducted more resolutive staining of FMRP in SVZ neurons cultured in Matrigel, providing a more precise depiction of its subcellular localization (see Figure 1). Additionally, we have removed the sentence referring to growth cone staining, as it was not visibly present in cultured neurons. We appreciate the guidance from the referee in refining our study.

(2) We have also added a new figure 4 with better staining of MAP1B in the RMS as well as a more resolutive MAP1B staining in cultured neurons.

With all due respect, we maintain that the western blot experiments, conducted in three independent experiments, unequivocally support the conclusion of a 1.6X increase in MAP1B in the RMS of Fmr1null mutants, a trend observed in other systems.

In accordance with the referee's suggestion, we endeavored to quantify RMS immunostainings. Regrettably, the results proved inconclusive. This outcome is not entirely unexpected, as immunostainings are recognized for their inherent challenges in quantification. The additional complexity introduced by neonate perfusion further contributes to the notable interindividual variability observed.

(3) The efficiency of the two interfering RNAs is now documented in the text. Regarding the directionality radar, as highlighted for Ref 1 (public review, point #2), we acknowledge that, while Fmr1KD generally recapitulates the migratory phenotype of the Fmr1 mutants, more precise statistical analysis reveals differences in directionality, which is now documented. We apologize for the previous lack of precision.

(4) The suggested experiment of overexpression is interesting but we faced challenges in its execution. Attempts to overexpress MAP1B through intraventricular electroporation of a CMV-MAP1B plasmid resulted in the immobilization of transfected cells in the SVZ, hindering further analysis of migration. We hypothesize that this outcome may be attributed to a discrepancy in the actual dosage of MAP1B in the mutants.

(5) Concerning this point, and as mentioned above, we have incorporated a crucial piece of information into the manuscript, presented in Figure 6. The data reveal a severe disruption in the microtubular cage surrounding the nucleus of migrating neurons in Fmr1 mutants, a phenomenon rescued by MAP1B knock-down. Based on these findings, we believe we can confidently conclude that the microtubule-dependent functions of MAP1B play a role in the migratory phenotype of Fmr1 mutants. We consider this experiment to be a highly valuable addition to our work, shedding light on the underlying molecular mechanisms.

To Reviewer #3

We thank the referee for their insightful comments and have taken their consideration with great considerations.

In addition and as said above, we want to stress that we have implemented a significant additional modification by introducing a new figure (Fig. 6). This figure highlights the collaborative impact of FMRP and Map1B on the microtubular structure of migrating neurons. We firmly believe that this molecular elucidation of the migration phenotype constitutes a noteworthy addition to our work.

Public review

With regard to the perceived 'incompleteness' of our work, we believe that the addition of Figure 6, illustrating the molecular underpinnings of the Fmr1 mutation on the microtubular cytoskeleton and its rescue in the MAP1B KD, significantly enhances the completeness of our study.

In response to the comment on the introduction and discussion sections, we acknowledge that their brevity was due to the Short Format initially chosen. We have since expanded these sections, incorporating additional information about FMRP and MAP1B and their influences on migration.

Regarding the La Fata article, as highlighted in our discussion, it's important to note that while the study did not strongly indicate an impact on radial locomotion per se, drawing conclusive results is challenging due to the relatively low number of analyzed neurons. Consequently, we do not believe that it poses a challenge to our findings.

With respect to MAP1B overexpression, as previously mentioned in response to Ref #2, point 4, our attempts resulted in the inhibition of migration, potentially due to an overdosage of the protein.

In terms of anatomical consequences, as highlighted in our discussion, while our neurons experience a delay in migration, they eventually reach their destination. Although a delay in migration may not directly result in significant anatomical anomalies, we acknowledge that the timing of differentiation can be crucial. As noted by Bocchi et al. (2017), a delay in the timing of differentiation for neurons reaching their target could lead to notable functional consequences. In any case, we have tOned down any references to the implication for the pathology.

Recommendation for the authors

• The size of the figures has been modified

• The pausing time and sinuosity are now defined

• The centrin-RFP labeling was indeed too weak in the previous version, which we corrected. We apologize for this.

• Fig S3 has been revised to address concerns. Notably, the decision to present the two bands for Vinculin and MAP1B separately is intentional. The blot is cut to allow independent development due to the substantial difference in their development times. We believe this approach provides a more accurate representation of the data.

• The numbering of the figures has been corrected. Sorry for the initial mistake.

• The Mat and Meth section has been corrected. Please note that we did not use any culture insert in this study.

• The tittle has been modified

• Comments about the Map1B overexpression experiment are expressed above and in replies to ref #2.

Reviewer #3 (Public Review):

Neuronal migration is one of the key processes for appropriate neuronal development. Defects in neuronal migration are associated with different brain disorders often accompanied by intellectual disabilities. Therefore, the study of the mechanisms involved in neuronal migration helps to understand the pathogenesis of some brain malformations and psychiatric disorders.

FMRP is an RNA-binding protein implicated in RNA metabolism regulation and mRNA local translation. FMRP loss of function causes fragile X syndrome (FXS), the most common form of inherited intellectual disability. Previous studies have shown the role of FMRP in the multipolar to bipolar transition during the radial migration in the cortex and its possible relation with periventricular heterotopia and altered synaptic communication in humans with FXS. However, the role of FMRP in neuronal tangential migration is largely unknown. In this manuscript, the authors aim to decipher the role of FMRP in the tangential migration of neuroblasts along the rostral migratory stream (RMS) in the postnatal brain. By extensive live-imaging analysis of migrating neuroblasts along the RMS, they demonstrate the requirement of FMRP for neuroblast migration and centrosomal movement. These migratory defects are cell-autonomous and mediated by the microtubule-associated protein Map1b.

Overall, the manuscript highlights the importance of FMRP in neuronal tangential migration. They performed an analysis of different aspects of migration such as nucleokinesis and cytokinesis in migrating neuroblasts from live-imaging videos. The authors have reinforced the results that associate defects in microtubule organization in Fmrp1 KO neurons and this rescue with the microtubule-associated protein Map1b. Overall, results concerning the role of Fmr1 in the tangential migration of neuroblasts are solid and convincing.

However, the work is still quite incomplete. My main concern is still what are the functional consequences of delay in neuroblast migration in the integration and function of OB interneurons and this relation with FXS pathophysiology. An anatomical examination of the RMS in the Fmr1KO mice is still missing.

eLife assessment

This study addresses the role of FMRP in the migration of newborn neuroblasts in the postnatal brain. Through extensive and convincing analysis of living imaging videos, the authors showed that neurons with FMRP deletion migrate aberrantly and exhibit defects in nucleokinesis and centrokinesis. The study presents a valuable finding on the mechanism of neuroblast migration in the postnatal brain.

Reviewer #1 (Public Review):

This study investigated Fragile X Messenger Ribonucleoprotein (FMRP) protein impact on neuroblast tangential migration in the postnatal rostral migratory stream (RMS). Authors conducted a series of live-imaging on organotypic brain slices from Fmr1-null mice. They continued their analysis silencing Fmr1 exclusively from migrating neuroblasts using electroporation-mediated RNA interference method (MiRFmr1 KD). These impressive approaches show that neuroblasts tangential migration is impaired in Fmr1-null mice RMS and these defects are mostly recapitulated in the MiRFmr1neuroblasts. This nicely supports the idea that FMRP have a cell autonomous function in tangentially migrating neuroblasts. Authors also confirm that FMRP mRNA target Microtubule Associated Protein 1B (MAP1B) is overexpressed in the Fmr1-null mice RMS. They successfully use electroporation-mediated RNA interference method to silence Map1b in the Fmr1-null mice neuroblasts. This discreet and elaborate experiment rescues most of the migratory defects observed both in Fmr1-null and MiRFmr1neuroblasts. Altogether, these results strongly suggest that FMRP-MAP1B axis has an important role in regulation of the neuroblasts tangential migration in the RMS. Neurons move forward in cyclic saltatory manner which includes repeated steps of leading process extension, migration of the cell organelles and nuclear translocation. Authors reveal by analyzing the live-imaging data that FMRP-MAP1B axis is affecting movement of centrosome and nucleus during saltatory migration. An important part of the centrosome and nucleus movement is forces mediated by microtubule dynamics. Authors propose that FMRP regulate tangential migration via microtubule dynamics regulator MAP1B. This work provides valuable new information on regulation of the neuroblasts tangential saltatory migration. These findings also increase and improve our understanding of the issues involved in Fragile X Syndrome (FXS) disorders. The conclusions of this work are supported by the presented data.

The current version of the study has improved substantially. Authors have enhanced the material and methods section including a more detailed section on the neuronal migration analysis. This amendment is a very valuable addition and strengthens the interpretation of the results, analysis and conclusions. Authors also have strengthened and clarified their results providing a more profound analysis of the migration directionality between controls, Fmr1-null, MiRFmr1 KD and MiRMap1b KD neuroblasts. They have incorporated new results in the study which elaborate FMRP and MAP1B participation in microtubule organization during tangential migration. Authors show that FMRP-MAP1B axis act on microtubule cage surrounding the nucleus. Microtubule cage participate on proper nuclear movement during neuron migration. These results emphasize more the interplay between FMRP, MAP1B, and the microtubule cytoskeleton. The authors have successfully expanded both the introduction and discussion sections of the manuscript.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript, Lee et al. compared encoding of odor identity and value by calcium signaling from neurons in the ventral pallidum (VP) in comparison to D1 and D2 neurons in the olfactory tubercle (OT).

Strengths:

They utilize a strong comparative approach, which allows the comparison of signals in two directly connected regions. First, they demonstrate that both D1 and D2 OT neurons project strongly to the VP, but not the VTA or other examined regions, in contrast to accumbal D1 neurons which project strongly to the VTA as well as the VP. They examine single unit calcium activity in a robust olfactory cue conditioning paradigm that allows them to differentiate encoding of olfactory identity versus value, by incorporating two different sucrose, neutral and air puff cues with different chemical characteristics. They then use multiple analytical approaches to demonstrate strong, low-dimensional encoding of cue value in the VP, and more robust, high-dimensional encoding of odor identity by both D1 and D2 OT neurons, though D1 OT neurons are still somewhat modulated by reward contingency/value. Finally, they utilize a modified conditioning paradigm that dissociates reward probability and lick vigor to demonstrate that VP encoding of cue value is not dependent on encoding of lick vigor during sucrose cues, and that separable populations of VP neurons encode cue value/sucrose probability and lick vigor.

Weaknesses:

The conclusions of the data are mostly well supported by the analyses, but the statistical analysis is somewhat limited and needs to be clarified and extended.

(1) The manuscript includes limited direct statistical comparison of the neural populations, and many of the comparisons between the subregions are descriptive, including descriptions of the percentage of neurons having specific response types, or differences in effect sizes or differing "levels" of significance. An additional direct comparison of data from each subpopulation would help to confirm whether the differences reported are statistically meaningful.

Response: We thank the reviewer for their helpful suggestions. As the reviewer noted, the first version of our manuscript had limited direct comparisons of single-neuron metrics across subpopulations. These analyses were also limited to the supplementary figures: 1) {SK vs. XK} and {SK vs. ST} decoder auROC (S10F), 2) Valence scores (S10G), and 3) S-cue confusion after MNR classification (S11D). We have now included the following statistical comparisons of single-neuron metrics across subpopulation: 1) % of neurons that respond to both S cues (Tables S10, S11), 2) % of neurons that have auROC >0.75 for {SK vs. XK}, {SK vs. PK}, and {SK vs. ST} (Tables S12-S17), 3) response magnitudes to S cues (Table S38), and 4) valence scores (Tables S44-46).

(2) When hypothesis tests are conducted between the neural populations, it is not clear whether the authors have accounted for the random effect of the subject, or whether individual units were treated as fully independent. For instance, pairwise differences are reported in Figures 4I, 5G/I/L, and others, but the statistical methods are unclear. Assessment of the statistics is further limited by the lack of reporting of degrees of freedom. If the individual neurons are treated as independent in these analyses, it could increase the likelihood of

Response: We have clarified when statistical analyses are comparing individual neurons vs. simultaneously recorded populations. Per the reviewer’s recommendation, we have also incorporated linear mixed-effects models when statistically analyzing individual neurons. Lastly, to further clarify the statistical analyses used, we have added multiple supplementary tables that better describe the statistical tests used and the relevant outputs.

Reviewer #2 (Public Review):

Summary:

This work is interesting since the authors provide an in vivo analysis into how odor-associations may change as represented at the level of olfactory tubercle (presynaptic) and next at the level of the ventral pallidum (postsynaptic). First the authors start-off with a seemingly careful characterization of the anterograde and retrograde connectivity of dopamine 1 receptor (D1) and dopamine 2 receptor (D2) expressing medium spiny neurons in the olfactory tubercle and neurons in the ventral pallidum. From this work they claim that regardless of D1 or D2 expression, tubercle neurons mainly project to the lateral portion of the ventral pallidum. Next, to compare how odor-associated neuronal activity in the ventral pallidum and the olfactory tubercle (D1 vs D2 MSNs) transforms across association learning, the authors performed 2photon calcium imaging while mice engaged in a lick / no-lick task wherein two odors are associated with reward, two odors are associated with no outcome, and two odors are associated with an air puff.

This manuscript builds off of prior work by several groups indicating that the olfactory tubercle neurons form flexible learned associations to odors by looking at outputs into the pallidum (but without looking specifically at palladial neurons that truly get input from tubercle I should highlight) and with that, this work is novel. We appreciated the use of a straight-forward odoroutcome behavioral paradigm and the careful computational methods and analyses utilized to disentangle the contributions of single neurons vs population level responses to behavior. With one exception from the Murthy lab, 2P imaging in the tubercle is a new frontier and that is appreciated - as is the 2P imaging in the pallidum which was well-supported by the histology. The anatomical work is also well presented.

Overall the approach and methods are superb. The issues come when considering how the authors present the story and what conclusions are made from these data. Several key points before going into specifics about each are: 1) The authors can not conclude that their results are contradictory to prior results, 2) The authors over-interpret the results and do not discuss several key methodological issues. We were concerned with the ability to make strong claims regarding the circuitry presented, especially given how much the presented claims contradict prior work. There were also issues with the interpretability of neuronal encoding of value vs valence based on the present behavior (in which a distinction between the air puff and neutral trial types was not clear) and the imaging methodology (in which the neuronal populations analyzed were not clearly defined). In addition to toning down and rectifying some of the language and interpretations, we suggest including a study limitations section where these methodological and interpretation issues are discussed. Over-interpreting and playing up the significance of this work is unnecessary, especially given eLife's new review and publication policy. Readers should be given a sufficiently detailed and nuanced presentation of these thought-provoking results, and from there allowed to interpret the results as they want.

Strengths:

State-of-the-art approaches (as detailed above)

Possible conceptual innovation in terms of looking into output from the olfactory tubercle which has yet to be investigated in this avenue.

Weaknesses:

On the first point regarding the authors repeated and unsupported claims that their results are contradictory. There are papers by numerous groups, in respected journals including this one, all together which used 5 different methods (cfos, photometry, 2P, units, fMRI), in animals ranging from humans to mice, which support that tubercle neurons reflect the emotional association of an odor, whether spontaneous or learned. With that, it is on the authors to not claim that their results contradict as if the other papers are suspect, but instead, from our standpoint it is on the authors to explain how and why their results differ from these other papers versus just simply saying they found something different [which at present is framed in a way that is 'correct' due to primacy if nothing else].

Response: We acknowledge that the first version of the manuscript contained unnecessary disagreeing language. We do not think that our results are broadly in disagreement with the existing literature, but we do come to different conclusions about what the OT is representing. Namely, our comparison of valence encoding in OT to that in the VP strongly indicates that the anteromedial OT has a less robust representation of valence, and we argue that this reflects either an intermediate form of valence representation or potentially might not be important for valence representation at all. We have toned down our conclusions, made clear that we are only recording from one domain of the OT, limited our speculation to the discussion and added a “speculations” section.

Second, onto the points of interpretation of results, there are several specific areas where this should be rectified. As is, the authors overinterpret their results and draw too far-reaching conclusions. This needs to be corrected.

In particular, the claims that D1 and D2 neurons of the olfactory tubercle nearly exclusively send projections to the ventral pallidum must be interpreted with caution given that the authors injected an anterograde AAV into the anteromedial olfactory tubercle, and did not examine the projections from either the posterior or lateral portions of the olfactory tubercle. This is especially significant since the retrograde tracing performed from the ventral pallidum indicates that the lateral olfactory tubercle, not the medial olfactory tubercle, primarily projects to the ventral pallidum (Fig 1D-F), however this may be due to leakage into the nucleus accumbens, as seen in the supplementary figure, S1G.

Response: We thank the reviewer for the point of caution. We have now made it clear that our conclusions are limited to the anteromedial portion of the OT, and other areas may have other projections.

The same caution must be advised when interpreting the retrograde tracing performed in Fig 1G-I, since the neuronal tracer used and the laterality and rostral-caudal injection site within the VTA could result in different projection patterns and under- or over-labelling. Additionally, the metric used, %Fiber Density (Figure 1C), as in the percentage of 16-bit pixels within the region of interest with an intensity greater than 200, is semi-quantitative, and is more applicable for examining axonal fibers that pass through a region rather than the synaptic terminals (like with a synaptophysin fusion protein-based tracing paradigm) found within a region (puncta). The statements made in contrast to prior studies should therefore be softened, and these concerns should be addressed in the introduction, discussion, and the limitations section if added.

Response: We have added statements to address these limitations.

The other major concern is whether the behavioral data generated is indicative of the full spectrum of valence. The authors appropriately state that the mice "perceive" the air puff, yet based on their data the mice did not clearly experience the puff-associated odor as emotionally aversive (viz., negative valence). The way the authors describe these results, it seems they agree with this. With that, the authors can't say the puff is aversive without data to show such - that is an assumption which, while seemingly intuitive, is not supported by the data unfortunately. To elaborate more since this is important to the messaging of the paper: The authors utilized a simple behavioral design, wherein two molecular classes of odors were included in either a sucrose rewarded, neutral no outcome, or air puff punished trial type. The odor-outcome pairs were switched after three days, allowing the authors to compare neuronal responses on the basis of odor identity and the later associated outcome. While the mice showed clear learning of the rewarded trial types by an increase in anticipatory licking during the odor, they did not show any significant changes in behavior that indicated learning of the air puff trial type (change in running velocity or % maximal eye size), especially in contrast to the neutral trial type. This brings up the concern that either the odor-air puff aversive associations (to odors) were not learned, or that the neutral trial types, in which a reward was omitted, were just as aversive as the air puff to the rear, despite the lack of startle response - perhaps due to stimulus generalization between neutral and air puff odor. The possibility of lack of learning is addressed in the paragraph starting at line 578, but does not account for the possibility that the lack of reward is also sufficiently punishing. The authors also address the possibility that laterality in the VP contributed to the lack of neural responsivity observed, but should also include a statement regarding laterality in the olfactory tubercle, as described in https://doi.org/10.7554/eLife.25423 and https://doi.org/10.1523/JNEUROSCI.0073-15.2015, since the effects of modulating the lateral portion of the olfactory tubercle are not yet reported. Lastly, use of the term "reward processing" should be avoided/omitted since the authors did not specifically study the processing of reinforcers.

Response: As the reviewer points out, we tried to be cautious interpreting the “aversive” odor response, and focused mainly on the reward association. This was discussed in the discussion. We don’t see the need to further add a redundent statement to a “limitations section”. We have also added a note about the previously identified laterality of the OT, which might account for lack of aversive responsive neurons in the OT. The reviewer makes an interesting suggestion that behavioral responses to airpuff-associated odors are not significantly different from un-associated because the lack of reward in this context is already aversive. We note that the walking velocity between reward- and puff-associated odor is significantly different, but not that to unassociated. This is in agreement with the suggestion, and we have added a statement to reflect this.

Also, I would appreciate justification of the term "value". How specifically does the assay used assess value versus a more simplistic learned association which influences perceived hedonics or valence of the odors.

Response: We have removed the term “value” with the exception of areas where we cite the work of others. We acknowledge that the word value is complicated in the incentive learning field and appreciate the suggestion. Our experimental design was meant to investigate learned association for positive and negative stimuli, thus valence is more appropriate and we have used this term.

More information is needed regarding how neurons are identified day-to-day, both in textual additions to the Methods and also in terms of elaborating more in the results and/or figure legends about what neurons are included:

(a) The ROI maps for identifying/indicating cells in the FOVs are nice to see and at the same time raise some concerns about how cells are identified and/or borders for those specific ROIs drawn. For instance, Figure 4, A & D, ROI #13 (cell #13) between those two panels is VERY different in shape/size. Also see ROIs 15 and 4. Why was an ROI map not made on day 1 and then that same map applied and registered to frames from consecutive imaging days in that same mouse? As it is new ROIs are drawn, smaller for some "cells" and larger for others. And at least in ROI #13 above, one ROI is about twice as large as the other. This inconsistency in the work flow and definition of the ROIs is needing to be addressed in Methods. Also, the authors should address if and how this could possibly impact their results.

Response: We have added details and clarified the methods section to make this more clear. We note that we extracted calcium transients from the raw data with the the widely used Constrained Nonnegative Matrix Factorization (CNMF) algorithm. This processing algorithm simultaneously identifies spatial and temporal components using modeled kinetics of calcium transients and pre-trained CNN classifiers. Using 2-photon microscopy the optical resolution in the z plane is narrow and we may not always capture components of a neuron that look like “neurons”, but all ROIs were confirmed manually to ensure they were not artifacts.

(b) Also, more details are needed in results and/or figure legends regarding the changes in cell numbers over days that are directly compared in the results. Some days there are 10% or more or less cells. Why? It is not the same population being compared in this case and so some Discussion of this is needed.

Response: The shapes of the spatial components can vary across days due to nonrigid motion in the brain and/or miniscule differences in the imaging angle across days. Although we visually verified that we are imaging approximately the same z plane across days, we cannot (and do not) claim to image identical populations of neurons across days.

Reviewer #3 (Public Review):

Summary:

This manuscript describes a study of the olfactory tubercle in the context of reward representation in the brain. The authors do so by studying the responses of OT neurons to odors with various reward contingencies and compare systematically to the ventral pallidum. Through careful tracing, they present convincing anatomical evidence that the projection from the olfactory tubercle is restricted to the lateral portion of the ventral pallidum.

Using a clever behavioral paradigm, the authors then investigate how D1 receptor- vs. D2 receptor-expressing neurons of the OT respond to odors as mice learn different contingencies. The authors find that, while the D1-expressing OT neurons are modulated marginally more by the rewarded odor than the D2-expressing OT neurons as mice learn the contingencies, this modulation is significantly less than is observed for the ventral pallidum. In addition, neither of the OT neuron classes shows significant modulation by the reward itself. In contrast, the OT neurons contained information that could distinguish odor identities. These observations have led the authors to conclude that the primary feature represented in the OT is not reward.

Strengths:

The highly localized projection pattern from olfactory tubercle to ventral pallidum is a valuable finding and suggests that studying this connection may give unique insights into the transformation of odor by reward association.

Comparison of olfactory tubercle vs. ventral pallidum is a good strategy to further clarify the olfactory tubercle's position in value representation in the brain.

Weaknesses:

The authors' interpretation of the physiologic results - that a novel framework is needed to interpret the OT's role - requires more careful treatment.

Response: We thank the reviewer for their recommendation. We have toned down the conclusiveness of our language in the discussion. Additionally, we have removed several speculative sentences from the concluding paragraph.

Reviewer recommendations for Authors:

We thank the reviewers for this helpful list of recommended changes to the manuscript.<br /> Regrettably, a few of the recommendations were overlooked in the revision, as indicated below.<br /> We do agree with the suggestions and plan to add appropriate changes to the version of record.

Reviewer #1 (Recommendations For The Authors):

If the comparisons mentioned in point 2 in the public review do not account for the lack of independence of individual neurons, I suggest the authors do so by either running linear mixed effects models with a random effect for subject, or one-way ANOVAs with a random effect of subject, where appropriate. The authors could also run analyses on summarized individual subject data (averages, % of neurons, etc.), though the authors would lose substantial power when assessing whether average changes differ between subjects in each recording group.

We have clarified when statistical analyses are comparing individual neurons vs. simultaneously recorded populations. Per the reviewer’s recommendation, we have also incorporated linear mixed-effects models when statistically analyzing individual neurons. Lastly, to further clarify the statistical analyses used, we have added supplementary tables for every statistical test that better describe the parameters used and the relevant outputs.

Reviewer #2 (Recommendations For The Authors):

Of minor note, there are some symbols/special characters that did not translate in the figure caption for Figure 6C, repeated text between lines 700-705 and 707-712, and some other small grammatical errors. Additionally, the source of the anterograde tracing virus (AAV9-phSyn1FLEX-tdTomato-T2A-SypEGFP-WPRE) needs to be stated.

Thank you for pointing these out. We have added description to the figure legend, and deleted the repeated lines and fixed grammatical errors. During the revision, we Regrettably overlooked the request to provide the source for the AAV9-phSyn1-FLEX-tdTomato-T2A-SypEGFP-WPRE. We agree that this small detail is important and will add it before publication of the version of record. This viral vector was purchased from The Salk Institute GT3 Core.

Reviewer #3 (Recommendations For The Authors):

The authors' interpretation of the physiologic results - that a novel framework is needed to interpret the OT's role - requires more careful treatment. As the authors note, there is rewardcontingency modulation in OT, especially when D1 neurons are compared against D2, as shown in Fig. 3D,E, Fig. 4I, and Fig. F,J. Though small in effect size, presumably, these modulations cannot be explained by the odor identity. These observations, to this reviewer, suggest the D1 neurons of OT have a component of cue-reward representation. In other words, rather than developing an entirely new framework, an alternative possibility that D1 neurons of OT occupy an intermediate stage in associating cues with reward (i.e., under the same framework, but occupying a different position in the emergence of value representation) should be considered.

We thank the reviewer for this thoughtful comment. We have eliminated the statement that “novel framework is needed” and have been more conservative in our interpretations. We have also acknowledged that our results are not necessarily in conflict with existing literature, but we do draw different conclusions, namely that the anteromedial OT is not a robust valence encoding population in comparison to that in the VP. We appreciate the suggestion of the term “intermediate stage” in reward association and have now included this in the discussion. Lastly, we have limited broader speculation to a “speculation” section of the discussion.

Related to the above point, have the authors analyzed if the similarities in the chemical structures correspond to perceptual and neural similarities? In the data presented in Figure S4, there are greater similarities in the population patterns within the same rewarding condition than within chemical groups. A comparison of the reward vs. chemical group (a simpler version of Fig. 5B) may be beneficial and take full advantage of the experimental design.

This comparison already exists in 5B and lines 285-289 of results. In VP populations, the distribution was structured such that intervalence pairwise comparisons between sucrose-paired and not sucrose-paired odors (e.g. ||SK-PK|| and ||SK-XK||) were larger than intravalence pairwise comparisons (e.g. ||SK-ST||, or ||XK-XT||). OTD1 populations showed an intermediate trend where most intravalence pairwise distances were smaller than intervalence pairwise distances with the exception of ||SK-ST||.

Related to the point about chemical similarities - is the smaller effect size (amount of modulation associated with reward contingency) in this study, compared to the study by Martiros et al, explained by the similarities of odorants used?

This is an interesting point. Although the odorants we use are different from those in Martiros et al, we think it is unlikely to the basis of smaller effect size due to reward modulation. If OT represents odor in a population code, whereby identity is encoded in unique ensembles of activity, then variation in the expression of D1R between OT neurons could account for different effects in different ensembles. However, there is no evidence for such varied expression and it doesn’t seem like an ideal mechanism for the OT to broadly associate odor with reward. Moreover, we do not observe any differences in effect size of reward association between the different odorants used in our study. Rather, we think the difference between our findings is more likely to result from recording in different populations of neurons, which is addressed in lines 522-535.

Regarding the data presented in Fig. 3I - the rewarded odor responses (Sk) are compared against neutral ones (Xk responses), but an S vs. P comparison may be informative, too. Even though the authors mention that the effect of air puff is subtle, the behavioral data presented in Fig. 2F and G suggest that these serve as aversive stimuli. For example, on day 4, the first day after the reward contingency switch, the licking levels seem the lowest for the P odors.

We have added the S vs P comparison. Indeed, we had originally omitted this because the neural and behavioral response to puff cues was not robust. This is discussed in the discussion (lines 563-579), and our conclusions about aversive conditioning are cautious.

Regarding the data presented in Fig. 4G: it is difficult to interpret the data when the data for day 1 reward period and day 3 reward cue period are combined. Or do the authors mean day 1 S cue and day 3 S cue?

These data were based on an observation that some neurons in the VP only responded to sucrose (not odor) on day 1, but later became responsive to the associated odor on day 4. To quantify this, Fig. 4G shows the percentage of these neurons by reporting the percentage that were both responsive to sucrose (not odor) on day 1 and also rewarded odor on day 3. This is described in lines 260-274.

Figure 6 presentation would benefit from a revision. For example, it is unclear if the water port becomes available for the "N" odors with 100% or 50% chance of reward delivery, and if so, how that happens. There are some errors e.g., colormap used for panel G; odors listed may be wrong in line 752 etc. It was unfortunately not possible to understand what was presented.

We have added a schematic (Fig 6B) to better describe the movement of the port and details to the methods. The color scale was indeed inverted in panel G (now H), and it has been corrected. We have verified that the odors listed in the methods are correct. Although not included in the revision, in the version of record we will also add corresponding descriptors (e.g., LHi & Lx) to the odors in the methods for easier comparison.

Minor comments

For Figure 2H, an alternative description in the legend may be beneficial, as the phrasing is not intuitive. A suggested alternative is "licks in response to sugar-associated odors expressed as fraction of all odors".

We appreciate the suggestion and have changed this to “licks during either sucrose cue expressed as a fraction of all licks during any odor.”

Figure 2H: please explain the color code for crosses in the legend and the statistical comparison shown in the figure.

We have added a legend to explain the color code and included a statement about the statistics in the legend with a link to a supplemental table for statistical parameters.

Figure 3D: may contain mislabeling in the legend - the legend for 3D does not match the plot (legend refers to bar graph while plot shows line graphs)

Unclear what is meant. 3D legend says: “Percentage of total neurons that were significantly excited or inhibited by each odor (Bonferroni- adjusted FDR < 0.05) as a function of time relative to odor. Lines represent the mean across biological replicates and the shaded area reflects the mean ± SEM.” This is not a bar plot and is not referred to as one. 3E does show bar plots and is correctly described in the legend.

Figure 3M: uses letters to refer to cell populations that are identical to the roman numerals used in Fig 3 A-C as well as colours similar to the ones in Fig 3C. However, the cell groups are unrelated; splitting the figures or using a different nomenclature might help

We have adapted a different color code that we think makes this more distinct.

Figure 4I: statistical comparison shown in figure not explained (neither in main text nor legend)

We have added a statement about the statistical comparison and referenced a supplementary table.

Figure 5 D: color code appears to have a different range than the values shown (i.e. lower limit is 0.7 while the plot shows values below 0.7)

We confirm this is not a mistake but a stylistic choice. The displayed color scale does only show values to lower limit of 0.7, while the lower limit of values is 0.67. Although the color for 0.67 is not shown in the scale it is approximately the same as the lower limit. The values are reported for full transparency and accuracy.

Figure 5 G, I, & L: statistical comparison shown in figure not explained

The comparisons have been explained in supplemental tables (S22-29) and referenced in the legend.

Figure 5 I: meaning of symbols overlayed over bars not explained

“Markers represent the mean across biological replicates” has been added.

Figure 5 J&K: please state if error bars show SEM or SD; also please describe individual thinner lines in the legend

This has been added to describe 5I. The same format applies to J&K.

Figure 5L: please describe the individual crosses overlayed over bars in the legend

Described in 5I.

Figure S6A-C: please mention the odors used.

S6A-C shows kinetics for the odor a-terpinene, which is now indicated in the legend.

Line 129: mentions a 70 psi airpuff but methods say 75 psi - please clarify This has been corrected. 70 psi is the correct value.

Line 134 typo: SP should be PK

This has been corrected.

Line 428: typo; should be cluster 3, not 2

This has been corrected.

Line 474 (and figure 6O): please explain what "P" is

“P” is probability, used as P(S), as in probability of sucrose. This is defined in in line 466.

Line 692: please describe the staining protocol in the methods (rather than just listing the antibodies and concentrations)

We have added more details (lines 692-699).

Line 707-712: duplicate text (identical to Line 700-705)

This has been deleted.

Reviewer #3 (Public Review):

Summary:

This manuscript describes a study of the olfactory tubercle in the context of reward representation in the brain. The authors do so by studying the responses of OT neurons to odors with various reward contingencies and compare systematically to the ventral pallidum. Through careful tracing, they present convincing anatomical evidence that the projection from the olfactory tubercle is restricted to the lateral portion of the ventral pallidum.

Using a clever behavioral paradigm, the authors then investigate how D1 receptor- vs. D2 receptor-expressing neurons of the OT respond to odors as mice learn different contingencies. The authors find that, while the D1-expressing OT neurons are modulated marginally more by the rewarded odor than the D2-expressing OT neurons as mice learn the contingencies, this modulation is significantly less than is observed for the ventral pallidum. In addition, neither of the OT neuron classes shows conspicuous amount of modulation by the reward itself. In contrast, the OT neurons contained information that could distinguish odor identities. These observations have led the authors to conclude that the primary feature represented in the OT may not be reward.

Strengths:

The highly localized projection pattern from olfactory tubercle to ventral pallidum is a valuable finding and suggests that studying this connection may give unique insights into the transformation of odor by reward association.

Comparison of olfactory tubervle vs. ventral pallidum is a good strategy to further clarify the olfactory tubercle's position in value representation in the brain.

Weaknesses:

The study comes to a different conclusion about the olfactory tubercle regarding reward representations from several other prior works. Whether this stems from a difference in the experimental configurations such as behavioral paradigms used or indeed points to a conceptually different role for the olfactory tubercle remains to be seen.

eLife assessment

This important study by Lee and colleagues examined how neural representations are transformed between the olfactory tubercle (OT) and the ventral pallidum (VP) using single neuron calcium imaging in head-fixed animals trained in classical conditioning. They show that the dimensionality of neural responses is lower in the VP than in the OT and suggest that VP responses represent values in a more abstract form while OT contains more odor information, potentially enhancing odor contrast. The reviewers found the results overall convincing although the nature of OT responses needs to be investigated further.

Reviewer #1 (Public Review):

In this manuscript, Lee et al. compared encoding of odor identity and value by calcium signaling from neurons in the ventral pallidum (VP) in comparison to D1 and D2 neurons in the olfactory tubercle (OT).

Strengths:

They utilize a strong comparative approach, which allows the comparison of signals in two directly connected regions. First, they demonstrate that both D1 and D2 OT neurons project strongly to the VP, but not the VTA or other examined regions, in contrast to accumbal D1 neurons which project strongly to the VTA as well as the VP. They examine single unit calcium activity in a robust olfactory cue conditioning paradigm that allows them to differentiate encoding of olfactory identity versus value, by incorporating two different sucrose, neutral and air puff cues with different chemical characteristics. They then use multiple analytical approaches to demonstrate strong, low-dimensional encoding of cue value in the VP, and more robust, high-dimensional encoding of odor identity by both D1 and D2 OT neurons, though D1 OT neurons are still somewhat modulated by reward contingency/value. Finally, they utilize a modified conditioning paradigm that dissociates reward probability and lick vigor to demonstrate that VP encoding of cue value is not dependent on encoding of lick vigor during sucrose cues, and that separable populations of VP neuros encode cue value/sucrose probability and lick vigor. Direct comparisons of single unit responses between the two regions now utilize linear mixed effects models with random effects for subject,

Weaknesses:

The manuscript still includes mention of differences in effect size or differing "levels" of significance between VP and OT D1 neurons without reports of a direct comparisons between the two populations. This is somewhat mitigated by the comprehensive statistical reporting in the supplemental information, but interpretation of some of these results is clouded by the inclusion of OT D2 neurons in these analyses, and the limited description or contextualization in the main text.

Reviewer #2 (Public Review):

We appreciate the authors revision of this manuscript and toning down some of the statements regarding "contradictory" results. We still have some concerns about the major claims of this paper which lead us to suggest this paper undergo more revision as follows since, in its present form, we fear this paper is misleading for the field in two areas. here is a brief outline:

(1) Despite acknowledging that the injections only occurred in the anteromedial aspect of the tubercle, the authors still assert broad conclusions regarding where the tubercle projects and what the tubercle does. for instance, even the abstract states "both D1 and D2 neurons of the OT project primarily to the VP and minimally elsewhere" without mention that this is the "anteromedial OT". Every conclusion needs to specify this is stemming from evidence in just the anteromedial tubercle, as the authors do in some parts of the the discussion.

(2) The authors now frame the 2P imaging data that D1 neuron activity reflects "increased contrast of identity or an intermediate and multiplexed encoding of valence and identity". I struggle to understand what the authors are actually concluding here. Later in discussion, the authors state that they saw that OT D1 and D2 neurons "encode odor valence" (line 510). We appreciate the authors note that there is "poor standardization" when it comes to defining valence (line 521). We are ok with the authors speculating and think this revision is more forthcoming regarding the results and better caveats the conclusions. I suggest in abstract the authors adjust line 14/15 to conclude that, "While D1 OT neurons showed larger responses to rewarded odors, in line with prior work, we propose this might be interpreted as identity encoding with enhanced contrast." [eliminating "rather than valence encoding" since that is a speculation best reserved for discussion as the authors nicely do.

The above items stated, one issue comes to mind, and that is, why of all reasons would the authors find that the anteromedial aspect of the tubercle is not greatly reflecting valence. the anteromedial aspect of the tubercle, over all other aspects of the tubercle, is thought my many to more greatly partake in valence and other hedonic-driven behaviors given its dense reception of VTA DAergic fibers (as shown by Ikemoto, Kelsch, Zhang, and others). So this finding is paradoxical in contrast to if the authors would had studied the anterolateral tubercle or posterior lateral tubercle which gets less DA input.

eLife assessment

This study provides valuable observations indicating that human pyramidal neurons propagate information as fast as rat pyramidal neurons despite their larger size. Solid evidence demonstrates that this property is due to several biophysical properties of human neurons. This study will be of interest to neurophysiologists.

Reviewer #1 (Public Review):

The propagation of electrical signals within neuronal circuits is tightly regulated by the physical and molecular properties of neurons. Since neurons vary in size across species, the question arises whether propagation speed also varies to compensate for it. The present article compares numerous speed-related properties in human and rat neurons. They found that the larger size of human neurons seems to be compensated by a faster propagation within dendrites but not the axons of these neurons. The faster dendritic signal propagation was found to arise from wider dendritic diameters and greater conductance load in human neurons. In addition, the article provides a careful characterization of human dendrites and axons, as the field has only recently begun to characterize post-operative human cells. There are only a few studies reporting dendritic properties and these are not all consistent, hence there is the added value of reporting these findings, particularly given that the characterization is condensed in a compartmental model.

Strengths:<br /> The study was performed with great care using standard techniques in slice electrophysiology (pharmacological manipulation with somatic patch-clamp) as well as some challenging ones (axonal and dendritic patch-clamp). Modeling was used to parse out the role of different features in regulating dendritic propagation speed. The finding that propagation speed varies across species is novel as previous studies did not find a large change in membrane time constant or axonal diameters (a significant parameter affecting speed). A number of possible, yet less likely factors were carefully tested (Ih, membrane capacitance). The main features outlined here are well-known to regulate speed in neuronal processes. The modeling was also carefully done to verify that the magnitude of the effects is consistent with the difference in biophysical properties. Hence, the findings appear very solid to me.

Weaknesses:<br /> The role of diameter in regulating propagation speed is well-known in the axon literature.

Reviewer #2 (Public Review):

Summary:<br /> In this paper, Oláh and colleagues introduce new research data on the cellular and biophysical elements involved in transmission within the pyramidal circuits of the human neocortex. They gathered a comprehensive set of patch-clamp recordings from human and rat pyramidal neurons to compare how the temporal aspect of neuronal processing is maintained in the larger human neocortex. A broad range of experimental, theoretical, and computational methods are used, including two-photon guided dual whole-cell recordings, electron microscopy, and computational simulations of reconstructed neurons.

Recordings from synaptically connected pyramidal neurons revealed longer intercellular path lengths within the human neocortex. Further, by using dual whole-cell recordings from soma-dendrite and soma-axon locations, they found that short latencies from soma to soma can be partly attributed to an increased propagation speed for synaptic potentials, but not for the propagation of action potentials along the axon.

Next, in a series of extensive computational modeling studies focusing on the synaptic potentials, the authors observe that the short-latency within large human pyramidal neural circuits may have a passive origin. For a wide array of local synaptic input sites, the authors show that the conductance load of the dendrites, electrically coupled to a large diameter apical dendrite, affects the cable properties. The result is a relatively faster propagation of EPSPs in the human neuron.

The manuscript is well-written and the physiological experiments and biophysical arguments are very well explained. I appreciated the in-depth theoretical steps for the simulations. That passive cable properties of the dendrites are causing a higher velocity in human dendrites is interesting but there is a disconnect between the experimental findings and the model simulations. Based on the present data the contribution of active membrane properties cannot be dismissed and deserves further experiments.

Strengths:<br /> The authors present state-of-the-art 2P-guided dual whole-cell recordings in human neurons. In combination with detailed reconstructions, these approaches represent the next steps in unravelling the information processing in human circuits.

The computational modeling based on cable theory and experimentally constrained simulations provides an excellent integrated view of the passive membrane properties.

Weaknesses:<br /> There are smaller and larger issues with the statistical analyses of the experimental data which muddles the interim conclusions.

That the cable properties alone are the main explanation for speeding the electrical signaling in human pyramidal neurons appears inconsistent with the experimental data.

Some of the electrophysiological experiments require further control experiments to make robust conclusions.

Reviewer #3 (Public Review):

Summary:<br /> This study indicates that connections across human cortical pyramidal cells have identical latencies despite a larger mean dendritic and axonal length between somas in the human cortex. A precise demonstration combining detailed electrophysiology and modeling indicates that this property is due to faster propagation of signals in proximal human dendrites. This faster propagation is itself due to a slightly thicker dendrite, a larger capacitive load, and stronger hyperpolarizing currents. Hence, the biophysical properties of human pyramidal cells are adapted such that they do not compromise information transfer speed.

Strengths:<br /> The manuscript is clear and very detailed. The authors have experimentally verified a large number of aspects that could affect propagation speed and have pinpointed the most important one. This paper provides an excellent comparison of biophysical properties between rat and human pyramidal cells. Thanks to this approach a comprehensive description of the mechanisms underlying the acceleration of propagation in human dendrite is provided.

Weaknesses:<br /> Several aspects having an impact on propagation speed are highlighted (dendritic diameter, ionic channels, capacitive load) and there is no clear ranking of their impact on signal propagation speed. It seems that the capacitive load plays a major role, much more than dendritic diameter for which only a 10% increase is observed across species. Both aspects actually indicate that there is an increase in passive signal propagation speed with bigger cells at least close to the soma. This suggests that bigger cells are mechanically more rapid. An intuitive reason why capacitive load increases speed would also help the reader follow the demonstration.

Author Response

Comments on eLife Reviews

We thank the reviewers for their positive comments and constructive feedback following their thorough reading of the manuscript. In this provisional reply we will briefly address the reviewer’s comments and suggestions point by point. In the forthcoming revised manuscript, we will more thoroughly address the reviewer’s comments and provide additional supporting data.

(1) The expression 'randomly clustered networks' needs to be explained in more detail given that in its current form risks to indicate that the network might be randomly organized (i.e., not organized). In particular, a clustered network with future functionality based on its current clustering is not random but rather pre-configured into those clusters. What the authors likely meant to say, while using the said expression in the title and text, is that clustering is not induced by an experience in the environment, which will only be later mapped using those clusters. While this organization might indeed appear as randomly clustered when referenced to a future novel experience, it might be non-random when referenced to the prior (unaccounted) activity of the network. Related to this, network organization based on similar yet distinct experiences (e.g., on parallel linear tracks as in Liu, Sibille, Dragoi, Neuron 2021) could explain/configure, in part, the hippocampal CA1 network organization that would appear otherwise 'randomly clustered' when referenced to a future novel experience.

As suggested by the reviewer, we will revise the text to clarify that the random clustering is random with respect to any future, novel environment. The cause of clustering could be prior experiences (e.g. Bourjaily M & Miller P, Front. Comput. Neurosci. 5:37, 2011) or developmental programming (e.g. Perin R, Berger TK, & Markram H, Proc. Natl. Acad. Sci. USA 108:5419, 2011).

(2) The authors should elaborate more on how the said 'randomly clustered networks' generate beyond chance-level preplay. Specifically, why was there preplay stronger than the time-bin shuffle? There are at least two potential explanations:

(2.1) When the activation of clusters lasts for several decoding time bins, temporal shuffle breaks the continuity of one cluster's activation, thus leading to less sequential decoding results. In that case, the preplay might mainly outperform the shuffle when there are fewer clusters activating in a PBE. For example, activation of two clusters must be sequential (either A to B or B to A), while time bin shuffle could lead to non-sequential activations such as a-b-a-b-a-b where a and b are components of A and B;

(2.2) There is a preferred connection between clusters based on the size of overlap across clusters. For example, if pair A-B and B-C have stronger overlap than A-C, then cluster sequences A-B-C and C-B-A are more likely to occur than others (such as A-C-B) across brain states. In that case, authors should present the distribution of overlap across clusters, and whether the sequences during run and sleep match the magnitude of overlap. During run simulation in the model, as clusters randomly receive a weak location cue bias, the activation sequence might not exactly match the overlap of clusters due to the external drive. In that case, the strength of location cue bias (4% in the current setup) could change the balance between the internal drive and external drive of the representation. How does that parameter influence the preplay incidence or quality?

Based on our finding that preplay occurs only in networks that sustain cluster activity over multiple decoding time bins (Figure 5d-e), our understanding of the model’s function is consistent with the reviewers first explanation. We will provide additional analysis in the forthcoming revised manuscript in order to directly test the first explanation and will also test the intriguing possibility that the reviewer’s second suggestion contributes to above-chance preplay.

(3) The manuscript is focused on presenting that a randomly clustered network can generate preplay and place maps with properties similar to experimental observations. An equally interesting question is how preplay supports spatial coding. If preplay is an intrinsic dynamic feature of this network, then it would be good to study whether this network outperforms other networks (randomly connected or ring lattice) in terms of spatial coding (encoding speed, encoding capacity, tuning stability, tuning quality, etc.)

We agree that this is an interesting future direction, but we see it as outside the scope of the current work. There are two interesting avenues of future work: 1) Our current model does not include any plasticity mechanisms, but a future model could study the effects of synaptic plasticity during preplay on long-term network dynamics, and 2) Our current model does not include alternative approaches to constructing the recurrent network, but future studies could systematically compare the spatial coding properties of alternative types of recurrent networks.

(4) The manuscript mentions the small-world connectivity several times, but the concept still appears too abstract and how the small-world index (SWI) contributes to place fields or preplay is not sufficiently discussed.

For a more general audience in the field of neuroscience, it would be helpful to include example graphs with high and low SWI. For example, you can show a ring lattice graph and indicate that there are long paths between points at opposite sides of the ring; show randomly connected graphs indicating there are no local clustered structures, and show clustered graphs with several hubs establishing long-range connections to reduce pair-wise distance.

How this SWI contributes to preplay is also not clear. Figure 6 showed preplay is correlated with SWI, but maybe the correlation is caused by both of them being correlated with cluster participation. The balance between cluster overlap and cluster isolation is well discussed. In the Discussion, the authors mention "...Such a balance in cluster overlap produces networks with small-world characteristics (Watts and Strogatz, 1998) as quantified by a small-world index..." (Lines 560-561). I believe the statement is not entirely appropriate, a network similar to ring lattice can still have the balance of cluster isolation and cluster overlap, while it will have small SWI due to a long path across some node pairs. Both cluster structure and long-range connection could contribute to SWI. The authors only discuss the necessity of cluster structure, but why is the long-range connection important should also be discussed. I guess long-range connection could make the network more flexible (clusters are closer to each other) and thus increase the potential repertoire.

We agree that the manuscript would benefit from a more concrete explanation of the small-world index. We will revise the text and add illustrative figures.

We note that while our most successful clustered networks are indeed those with small-world characteristics, there are other ways of producing small-world networks which may not show good place fields or preplay. We will test another type of small-world network if time permits.

Our discussion of “cluster overlap” is specific to our type of small-world network in which there is no pre-determined spatial dimension (unlike the ring network of Watts and Strogatz). Therefore, because clusters map randomly to location once a particular spatial context is imposed, the random overlap between clusters produces long-range connections in that context (and any other context) so one can think of the amount of overlap between clusters as representing the number of long-range connections in a Watts-Strogatz model, except, we wish to iterate, such models involve a spatial topology within the network, which we do not include.

(5) What drives PBE during sleep? Seems like the main difference between sleep and run states is the magnitude of excitatory and inhibitory inputs controlled by scaling factors. If there are bursts (PBE) in sleep, do you also observe those during run? Does the network automatically generate PBE in a regime of strong excitation and weak inhibition (neural bifurcation)?

During sleep simulations, the PBEs are spontaneously generated by the recurrent connections in the network. The constant-rate Poisson inputs drive low-rate stochastic spiking in the recurrent network, which then randomly generates population events when there is sufficient internal activity to transiently drive additional spiking within the network.

During run simulations, the spatially-tuned inputs drive greater activity in a subset of the cells at a given point on the track, which in turn suppress the other excitatory cells through the feedback inhibition.

(6) Is the concept of 'cluster' similar to 'assemblies', as in Peyrache et al, 2010; Farooq et al, 2019? Does a classic assembly analysis during run reveal cluster structures?

Yes, we are highly confident that the clusters in our network would correspond to the functional assemblies that have been studied through assembly analysis and will present the relevant data in a revision.

(7) Can the capacity of the clustered network to express preplay for multiple distinct future experiences be estimated in relation to current network activity, as in Dragoi and Tonegawa, PNAS 2013?

We agree this is an interesting opportunity to compare the results of our model to what has been previously found experimentally and will test this if time permits.

Reviewer # 2

Weaknesses:

My main critiques of the paper relate to the form of the input to the network.

First, because the input is the same across trials (i.e. all traversals are the same duration/velocity), there is no ability to distinguish a representation of space from a representation of time elapsed since the beginning of the trial. The authors should test what happens e.g. with traversals in which the animal travels at different speeds, and in which the animal's speed is not constant across the entire track, and then confirm that the resulting tuning curves are a better representation of position or duration.

We agree that this is an important question, and we plan to run further simulations where we test the effects of varying the simulated speed. We will present results in the resubmission.

Second, it's unclear how much the results depend on the choice of a one-dimensional environment with ramping input. While this is an elegant idealization that allows the authors to explore the representation and replay properties of their model, it is a strong and highly non-physiological constraint. The authors should verify that their results do not depend on this idealization. Specifically, I would suggest the authors also test the spatial coding properties of their network in 2-dimensional environments, and with different kinds of input that have a range of degrees of spatial tuning and physiological plausibility. A method for systematically producing input with varying degrees of spatial tuning in both 1D and 2D environments has been previously used in (Fang et al 2023, eLife, see Figures 4 and 5), which could be readily adapted for the current study; and behaviorally plausible trajectories in 2D can be produced using the RatInABox package (George et al 2022, bioRxiv), which can also generate e.g. grid cell-like activity that could be used as physiologically plausible input to the network.

We agree that testing the robustness of our results to different models of feedforward input is important and we plan to do this in our revised manuscript for the linear track and W-track.

Testing the model in a 2D environment is an interesting future direction, but we see it as outside the scope of the current work. To our knowledge there are no experimental findings of preplay in 2D environments, but this presents an interesting opportunity for future modeling studies.

Finally, I was left wondering how the cells' spatial tuning relates to their cluster membership, and how the capacity of the network (number of different environments/locations that can be represented) relates to the number of clusters. It seems that if clusters of cells tend to code for nearby locations in the environment (as predicted by the results of Figure 5), then the number of encodable locations would be limited (by the number of clusters). Further, there should be a strong tendency for cells in the same cluster to encode overlapping locations in different environments, which is not seen in experimental data.

Thank you for making this important point and giving us the opportunity to clarify. We do find that subsets of cells with identical cluster membership have correlated place fields, but as we show in Figure 7b the network place map as a whole shows low remapping correlations across environments, which is consistent with experimental data (Hampson RE et al, Hippocampus 6:281, 1996; Pavlides C, et al, Neurobiol Learn Mem 161:122, 2019). Our model includes a relatively small number of cells and clusters compared to CA3, and with a more realistic number of clusters, the level of correlation across network place maps should reduce even further in our model network. The reason for a low level of correlation is because cluster membership is combinatorial, whereby cells that share membership in one cluster can also belong to separate/distinct other clusters, rendering their activity less correlated than might be anticipated. In our revised manuscript we will address this point more carefully and cite the relevant experimental support.

Reviewer # 3

Weaknesses:

To generate place cell-like activity during a simulated traversal of a linear environment, the authors drive the network with a combination of linearly increasing/decreasing synaptic inputs, mimicking border cell-like inputs. These inputs presumably stem from the entorhinal cortex (though this is not discussed). The authors do not explore how the model would behave when these inputs are replaced by or combined with grid cell inputs which would be more physiologically realistic.

We chose the linearly varying spatial inputs as the minimal model of providing spatial input to the network so that we could focus on the dynamics of the recurrent connections. We agree our results will be strengthened by testing alternative types of border-like input so will present such additional results in our revised version. However, given that a sub-goal of our model was to show that place fields could arise in locations at which no neurons receive a peak in external input, whereas combining input from multiple grid cells produces peaked place-field like input, adding grid cell input (and the many other types of potential hippocampal input) is beyond the scope of the paper.

Even though the authors claim that no spatially-tuned information is needed for the model to generate place cells, there is a small location-cue bias added to the cells, depending on the cluster(s) they belong to. Even though this input is relatively weak, it could potentially be driving the sequential activation of clusters and therefore the preplays and place cells. In that case, the claim for non-spatially tuned inputs seems weak. This detail is hidden in the Methods section and not discussed further. How does the model behave without this added bias input?

First, we apologize for a lack of clarity if we have caused confusion about the type of inputs (linear and cluster-dependent as we had attempted to portray prominently in Figure 1, where it is described in the caption, l. 156-157, and Results, l. 189-190 & l. 497-499, as well as in the Methods, l. 671-683) and if we implied an absence of spatially-tuned information in the network. In the revision we will clarify that for reliable place fields to appear, the network must receive spatial information and that one point of our paper is that the information need not arrive as peaks of external input already resembling place cells or grid cells. We chose linearly ramping boundary inputs as the minimally place-field like stimulus (that still contains spatial information) but in our revision we will include alternatives. We should note that during sleep, when “preplay” occurs, there is no such spatial bias (which is why preplay can equally correlate with place field sequences in any context). In the revision, we will update Figure 1 to show more clearly the cluster-dependent linearly ramping input received by some specific cells with both similar and different place fields.

Unlike excitation, inhibition is modeled in a very uniform way (uniform connection probability with all E cells, no I-I connections, no border-cell inputs). This goes against a long literature on the precise coordination of multiple inhibitory subnetworks, with different interneuron subtypes playing different roles (e.g. output-suppressing perisomatic inhibition vs input-gating dendritic inhibition). Even though no model is meant to capture every detail of a real neuronal circuit, expanding on the role of inhibition in this clustered architecture would greatly strengthen this work.

This is an interesting future direction, but we see it as outside the scope of our current work. While inhibitory microcircuits are certainly important physiologically, we focus here on a minimal model that produces the desired place cell activity and preplay, as measured in excitatory cells.

For the modeling insights to be physiologically plausible, it is important to show that CA3 connectivity (which the model mimics) shares the proposed small-world architecture. The authors discuss the existence of this architecture in various brain regions but not in CA3, which is traditionally thought of and modeled as a random or fully connected recurrent excitatory network. A thorough discussion of CA3 connectivity would strengthen this work.

We agree this is an important point that is missing, and we will revise the text to specifically address CA3 connectivity (Guzman et al., Science 353 (6304), 1117-1123 2016) and the small-world structure therein due to the presence of “assemblies”.

Reviewer #3 (Public Review):

Summary:<br /> This work offers a novel perspective on the question of how hippocampal networks can adaptively generate different spatial maps and replays/preplays of the corresponding place cells, without any such maps pre-existing in the network architecture or its inputs. Unlike previous modeling attempts, the authors do not pre-tune their model neurons to any particular place fields. Instead, they build a random, moderately-clustered network of excitatory (and some inhibitory) cells, similar to CA3 architecture. By simulating spatial exploration through border-cell-like synaptic inputs, the model generates place cells for different "environments" without the need to reconfigure its synaptic connectivity or introduce plasticity. By simulating sleep-like random synaptic inputs, the model generates sequential activations of cells, mimicking preplays. These "preplays" require small-world connectivity, so that weakly connected cell clusters are activated in sequence. Using a set of electrophysiological recordings from CA1, the authors confirm that the modeled place cells and replays share many features with real ones. In summary, the model demonstrates that spontaneous activity within a small-world structured network can generate place cells and replays without the need for pre-configured maps.

Strengths:<br /> This work addresses an important question in hippocampal dynamics. Namely, how can hippocampal networks quickly generate new place cells when a novel environment is introduced? And how can these place cells preplay their sequences even before the environment is experienced? Previous models required pre-existing spatial representations to be artificially introduced, limiting their adaptability to new environments. Other models depended on synaptic plasticity rules which made remapping slower than what is seen in recordings. This modeling work proposes that quickly-adaptive intrinsic spiking sequences (preplays) and spatially tuned spiking (place cells) can be generated in a network through randomly clustered recurrent connectivity and border-cell inputs, avoiding the need for pre-set spatial maps or plasticity rules. The proposal that small-world architecture is key for place cells and preplays to adapt to new spatial environments is novel and of potential interest to the computational and experimental community.

The authors do a good job of thoroughly examining some of the features of their model, with a strong focus on excitatory cell connectivity. Perhaps the most valuable conclusion is that replays require the successive activation of different cell clusters. Small-world architecture is the optimal regime for such a controlled succession of activated clusters.

The use of pre-existing electrophysiological data adds particular value to the model. The authors convincingly show that the simulated place cells and preplay events share many important features with those recorded in CA1 (though CA3 ones are similar).

Weaknesses:<br /> To generate place cell-like activity during a simulated traversal of a linear environment, the authors drive the network with a combination of linearly increasing/decreasing synaptic inputs, mimicking border cell-like inputs. These inputs presumably stem from the entorhinal cortex (though this is not discussed). The authors do not explore how the model would behave when these inputs are replaced by or combined with grid cell inputs which would be more physiologically realistic.

Even though the authors claim that no spatially-tuned information is needed for the model to generate place cells, there is a small location-cue bias added to the cells, depending on the cluster(s) they belong to. Even though this input is relatively weak, it could potentially be driving the sequential activation of clusters and therefore the preplays and place cells. In that case, the claim for non-spatially tuned inputs seems weak. This detail is hidden in the Methods section and not discussed further. How does the model behave without this added bias input?

Unlike excitation, inhibition is modeled in a very uniform way (uniform connection probability with all E cells, no I-I connections, no border-cell inputs). This goes against a long literature on the precise coordination of multiple inhibitory subnetworks, with different interneuron subtypes playing different roles (e.g. output-suppressing perisomatic inhibition vs input-gating dendritic inhibition). Even though no model is meant to capture every detail of a real neuronal circuit, expanding on the role of inhibition in this clustered architecture would greatly strengthen this work.

For the modeling insights to be physiologically plausible, it is important to show that CA3 connectivity (which the model mimics) shares the proposed small-world architecture. The authors discuss the existence of this architecture in various brain regions but not in CA3, which is traditionally thought of and modeled as a random or fully connected recurrent excitatory network. A thorough discussion of CA3 connectivity would strengthen this work.

eLife assessment

This study presents a valuable finding on the spontaneous emergence of structured activity in artificial neural networks endowed with specific connectivity profiles. The evidence supporting the claims of the authors is potentially solid but still incomplete at this stage, as it is important to demonstrate that similar properties are observed with more diverse inputs and in more complex environments. The work will be of interest to systems and computational neuroscientists.

Reviewer #1 (Public Review):

In this manuscript, the authors investigated the dynamics of a neural network model characterized by sparsely connected clusters of neuronal ensembles. They found that such a network could intrinsically generate sequence preplay and place maps, with properties like those observed in the real-world data. Strengths of the study include the computational model and data analysis supporting the hippocampal network mechanisms underlying sequence preplay of future experiences and place maps.

Previous models of replay or theta sequences focused on circuit plasticity and usually required a pre-existing place map input from the external environment via upstream structures. However, those models failed to explain how networks support rapid sequential coding of novel environments or simply transferred the question to the upstream structure. On the contrary, the current proposed model required minimal spatial inputs and was aimed at elucidating how a preconfigured structure gave rise to preplay, thereby facilitating the sequential encoding of future novel environments.

In this model, the fundamental units for spatial representation were clusters within the network. Sequential representation was achieved through the balance of cluster isolation and their partial overlap. Isolation resulted in a self-reinforced assembly representation, ensuring stable spatial coding. On the other hand, overlap-induced activation transitions across clusters, enabling sequential coding.

This study is important when considering that previous models mainly focused on plasticity and experience-related learning, while this model provided us with insights into how network architecture could support rapid sequential coding with large capacity, upon which learning could occur efficiently with modest modification via plasticity.

I found this research very inspiring and, below, I provide some comments aimed at improving the manuscript. Some of these comments may extend beyond the scope of the current study, but I believe they raise important questions that should be addressed in this line of research.

(1) The expression 'randomly clustered networks' needs to be explained in more detail given that in its current form risks to indicate that the network might be randomly organized (i.e., not organized). In particular, a clustered network with future functionality based on its current clustering is not random but rather pre-configured into those clusters. What the authors likely meant to say, while using the said expression in the title and text, is that clustering is not induced by an experience in the environment, which will only be later mapped using those clusters. While this organization might indeed appear as randomly clustered when referenced to a future novel experience, it might be non-random when referenced to the prior (unaccounted) activity of the network. Related to this, network organization based on similar yet distinct experiences (e.g., on parallel linear tracks as in Liu, Sibille, Dragoi, Neuron 2021) could explain/configure, in part, the hippocampal CA1 network organization that would appear otherwise 'randomly clustered' when referenced to a future novel experience.

(2) The authors should elaborate more on how the said 'randomly clustered networks' generate beyond chance-level preplay. Specifically, why was there preplay stronger than the time-bin shuffle? There are at least two potential explanations:

(1) - When the activation of clusters lasts for several decoding time bins, temporal shuffle breaks the continuity of one cluster's activation, thus leading to less sequential decoding results. In that case, the preplay might mainly outperform the shuffle when there are fewer clusters activating in a PBE. For example, activation of two clusters must be sequential (either A to B or B to A), while time bin shuffle could lead to non-sequential activations such as a-b-a-b-a-b where a and b are components of A and B;

(2) - There is a preferred connection between clusters based on the size of overlap across clusters. For example, if pair A-B and B-C have stronger overlap than A-C, then cluster sequences A-B-C and C-B-A are more likely to occur than others (such as A-C-B) across brain states. In that case, authors should present the distribution of overlap across clusters, and whether the sequences during run and sleep match the magnitude of overlap. During run simulation in the model, as clusters randomly receive a weak location cue bias, the activation sequence might not exactly match the overlap of clusters due to the external drive. In that case, the strength of location cue bias (4% in the current setup) could change the balance between the internal drive and external drive of the representation. How does that parameter influence the preplay incidence or quality?

(3). The manuscript is focused on presenting that a randomly clustered network can generate preplay and place maps with properties similar to experimental observations. An equally interesting question is how preplay supports spatial coding. If preplay is an intrinsic dynamic feature of this network, then it would be good to study whether this network outperforms other networks (randomly connected or ring lattice) in terms of spatial coding (encoding speed, encoding capacity, tuning stability, tuning quality, etc.)

(4) The manuscript mentions the small-world connectivity several times, but the concept still appears too abstract and how the small-world index (SWI) contributes to place fields or preplay is not sufficiently discussed.

For a more general audience in the field of neuroscience, it would be helpful to include example graphs with high and low SWI. For example, you can show a ring lattice graph and indicate that there are long paths between points at opposite sides of the ring; show randomly connected graphs indicating there are no local clustered structures, and show clustered graphs with several hubs establishing long-range connections to reduce pair-wise distance.

How this SWI contributes to preplay is also not clear. Figure 6 showed preplay is correlated with SWI, but maybe the correlation is caused by both of them being correlated with cluster participation. The balance between cluster overlap and cluster isolation is well discussed. In the Discussion, the authors mention "...Such a balance in cluster overlap produces networks with small-world characteristics (Watts and Strogatz, 1998) as quantified by a small-world index..." (Lines 560-561). I believe the statement is not entirely appropriate, a network similar to ring lattice can still have the balance of cluster isolation and cluster overlap, while it will have small SWI due to a long path across some node pairs. Both cluster structure and long-range connection could contribute to SWI. The authors only discuss the necessity of cluster structure, but why is the long-range connection important should also be discussed. I guess long-range connection could make the network more flexible (clusters are closer to each other) and thus increase the potential repertoire.

(5) What drives PBE during sleep? Seems like the main difference between sleep and run states is the magnitude of excitatory and inhibitory inputs controlled by scaling factors. If there are bursts (PBE) in sleep, do you also observe those during run? Does the network automatically generate PBE in a regime of strong excitation and weak inhibition (neural bifurcation)?

(6) Is the concept of 'cluster' similar to 'assemblies', as in Peyrache et al, 2010; Farooq et al, 2019? Does a classic assembly analysis during run reveal cluster structures?

(7) Can the capacity of the clustered network to express preplay for multiple distinct future experiences be estimated in relation to current network activity, as in Dragoi and Tonegawa, PNAS 2013?

Reviewer #2 (Public Review):

Summary:<br /> The authors show that a spiking network model with clustered neurons produces intrinsic spike sequences when driven with a ramping input, which are recapitulated in the absence of input. This behavior is only seen for some network parameters (neuron cluster participation and number of clusters in the network), which correspond to those that produce a small world network. By changing the strength of ramping input to each network cluster, the network can show different sequences.

Strengths:<br /> A strength of the paper is the direct comparison between the properties of the model and neural data.

Weaknesses:<br /> My main critiques of the paper relate to the form of the input to the network.

First, because the input is the same across trials (i.e. all traversals are the same duration/velocity), there is no ability to distinguish a representation of space from a representation of time elapsed since the beginning of the trial. The authors should test what happens e.g. with traversals in which the animal travels at different speeds, and in which the animal's speed is not constant across the entire track, and then confirm that the resulting tuning curves are a better representation of position or duration.

Second, it's unclear how much the results depend on the choice of a one-dimensional environment with ramping input. While this is an elegant idealization that allows the authors to explore the representation and replay properties of their model, it is a strong and highly non-physiological constraint. The authors should verify that their results do not depend on this idealization. Specifically, I would suggest the authors also test the spatial coding properties of their network in 2-dimensional environments, and with different kinds of input that have a range of degrees of spatial tuning and physiological plausibility. A method for systematically producing input with varying degrees of spatial tuning in both 1D and 2D environments has been previously used in (Fang et al 2023, eLife, see Figures 4 and 5), which could be readily adapted for the current study; and behaviorally plausible trajectories in 2D can be produced using the RatInABox package (George et al 2022, bioRxiv), which can also generate e.g. grid cell-like activity that could be used as physiologically plausible input to the network.

Finally, I was left wondering how the cells' spatial tuning relates to their cluster membership, and how the capacity of the network (number of different environments/locations that can be represented) relates to the number of clusters. It seems that if clusters of cells tend to code for nearby locations in the environment (as predicted by the results of Figure 5), then the number of encodable locations would be limited (by the number of clusters). Further, there should be a strong tendency for cells in the same cluster to encode overlapping locations in different environments, which is not seen in experimental data.

Reviewer #2 (Public Review):

Summary:

PKA is a major signaling protein that has been long studied and is vital for synaptic plasticity. Here, the authors examine the mechanism of PKA activity and specifically focus on addressing the question of PKA dissociation as a major mode of its activation in dendritic spines. This would potentially allow us to determine the precise mechanisms of PKA activation and address how it maintains spatial and temporal signaling specificity.

Strengths:

The results convincingly show that PKA activity is governed by the subcellular localization in dendrites and spines and is mediated via subunit dissociation. The authors make use of organotypic hippocampal slice cultures, where they use pharmacology, glutamate uncaging, and electrophysiological recordings.

Overall, the experiments and data presented are well executed. The experiments all show that at least in the case of synaptic activity, the distribution of PKA-C to dendritic spines is necessary and sufficient for PKA-mediated functional and structural plasticity.

The authors were able to persuasively support their claim that PKA subunit dissociation is necessary for its function and localization in dendritic spines. This conclusion is important to better understand the mechanisms of PKA activity and its role in synaptic plasticity.

Weaknesses:

While the experiments are indeed convincing and well executed, the data presented is similar to previously published work from the Zhong lab (Tillo et al., 2017, Zhong et al 2009). This reduces the novelty of the findings in terms of re-distribution of PKA subunits, which was already established. A few alternative approaches for addressing this question: targeting localization of endogenous PKA, addressing its synaptic distribution, or even impairing within intact neuronal circuits, would highly strengthen their findings. This would allow us to further substantiate the synaptic localization and re-distribution mechanism of PKA as a critical regulator of synaptic structure, function, and plasticity.

eLife assessment

In this important study, Xiong and colleagues studied PKA regulation in synaptic plasticity. They provide convincing evidence that dissociation of PKA catalytic subunits is essential for the proper function of the kinase. Experiments using a PKA regulatory-catalytic subunit fusion establish that dissociative activation is required for both structural long-term potentiation and basal priming of AMPA receptors.

Reviewer #1 (Public Review):

Summary:

This is a short self-contained study with a straightforward and interesting message. The paper focuses on settling whether PKA activation requires dissociation of the catalytic and regulatory subunits. This debate has been ongoing for ~ 30 years, with renewed interest in the question following a publication in Science, 2017 (Smith et al.). Here, Xiong et al demonstrate that fusing the R and C subunits together (in the same way as Smith et al) prevents the proper function of PKA in neurons. This provides further support for the dissociative activation model - it is imperative that researchers have clarity on this topic since it is so fundamental to building accurate models of localised cAMP signalling in all cell types. Furthermore, their experiments highlight that C subunit dissociation into spines is essential for structural LTP, which is an interesting finding in itself. They also show that preventing C subunit dissociation reduces basal AMPA receptor currents to the same extent as knocking down the C subunit. Overall, the paper will interest both cAMP researchers and scientists interested in fundamental mechanisms of synaptic regulation.

Strengths:

The experiments are technically challenging and well executed. Good use of control conditions e.g untransfected controls in Figure 4.

Weaknesses:

The novelty is lessened given the same team has shown dissociation of the C subunit into dendritic spines from RIIbeta subunits localised to dendritic shafts before (Tillo et al., 2017). Nevertheless, the experiments with RII-C fusion proteins are novel and an important addition.

Reviewer #3 (Public Review):

Summary:

Xiong et al. investigated the debated mechanism of PKA activation using hippocampal CA1 neurons under pharmacological and synaptic stimulations. Examining the two PKA major isoforms in these neurons, they found that a portion of PKA-C dissociates from PKA-R and translocates into dendritic spines following norepinephrine bath application. Additionally, their use of a non-dissociable form of PKC demonstrates its essential role in structural long-term potentiation (LTP) induced by two-photon glutamate uncaging, as well as in maintaining normal synaptic transmission, as verified by electrophysiology. This study presents a valuable finding on the activation-dependent re-distribution of PKA catalytic subunits in CA1 neurons, a process vital for synaptic functionality. The robust evidence provided by the authors makes this work particularly relevant for biologists seeking to understand PKA activation and its downstream effects essential for synaptic plasticity.

Strengths:

The study is methodologically robust, particularly in the application of two-photon imaging and electrophysiology. The experiments are well-designed with effective controls and a comprehensive analysis. The credibility of the data is further enhanced by the research team's previous works in related experiments. The conclusions of this paper are mostly well supported by data. The research fills a significant gap in our understanding of PKA activation mechanisms in synaptic functioning, presenting valuable insights backed by empirical evidence.

Weaknesses:

The physiological relevance of the findings regarding PKA dissociation is somewhat weakened by the use of norepinephrine (10 µM) in bath applications, which might not accurately reflect physiological conditions. Furthermore, the study does not address the impact of glutamate uncaging, a well-characterized physiologically relevant stimulation, on the redistribution of PKA catalytic subunits, leaving some questions unanswered.

Reviewer #1 (Public Review):

In this manuscript, Lebedeva et al. report the input/output wiring diagram of a population of previously identified giant excitatory neurons (abbreviated as ExNr) in the CA1 region of the rat hippocampus. Overall, Lebedeva et al. report that 1) ExNr are driven by Schaffer collaterals; 2) ExNr do not contact CA1Pyrs; 3) ExNr innervate PV interneurons; 4) ExNr received inhibition from bistratified cells, but not basket cells; and 5) ExNr -> PV synapse is strong enough to massively inhibit CA1Pyrs. Some of the findings reported here appear interesting. However, my appreciation of this manuscript was dampened by the limited scientific novelty, strong statements that are sometimes not supported by data, and vague, imprecise, and oversimplified narratives used throughout.

(1) The identity of ExNr reported here is unclear. It is unclear how ExNr are identified, and how robust the identification criteria are. A single anatomical reconstruction is provided together with depolarization-induced firing. However, whether all cells are consistent with the examples provided is unclear. Giant radiatum cells (previously known as RGCs, here abbreviated as ExNr) were previously identified by Maccaferri (1996) and Gulyas (1998). Based on anatomical criteria alone, it was suggested that these cells could take 4 different forms. The current manuscript mostly ignores this past finding. Given the topic of this paper, a careful anatomical and electrophysiological examination of ExNr is required.

(2) The identity of recorded interneurons is unclear. A major and potentially interesting finding reported here is the differential connectivity of ExNr to basket and bistratified neurons. However, it seems like basket and bistratified cells were mostly identified on the basis of electrophysiological criteria, and that 'only 5 neurons of each group were filed with biocytin, and the identity of interneurons was confirmed by axonal arborization pattern.' First, this significantly departs from the general current practices in the field where interneurons are identified based on combined anatomical and electrophysiological properties. This is because multiple examples in the literature support the extreme heterogeneity of interneurons, and that a combination of criteria is usually required for their appropriate identification. Second, the reconstruction of these neurons should be provided. Since the circuit wiring diagram proposed by the authors is based on these results, proper interneuron classification is necessary.

(3) Multiple conclusions are overstatements. For example, the interpretation that ExNr escapes perisomatic inhibition, as reported in the title, seems to ignore large families of cholecystokinin- or Sncg-expressing basket cells.

(4) Some of the more exciting findings appear preliminary, and the robustness of the findings is hard to evaluate. An example of that is found on Page 8, line 179: 'Thus, ExNR can operate as an amplification relay station for feed-forward inhibition of neurons in the CA area.' This conclusion appears only loosely supported by a few observations, (n = 3), as stated above. Similarly, the next section investigates the downstream effect of ExNr firing on CA1 pyramidal cells. The author reports that 'In 24% of the slices unitary APs in ExNr generated an fIPSP, delayed relative to the peak of the AP by 5.5 ms (n=6; Fig 3D-F).' In my opinion, 24% is a relatively low occurrence, even if we consider potentially cut axons (rightfully acknowledged by the authors) during the slicing procedure. Overall, this clearly doesn't fit the 'massive inhibition of downstream CA1Pyrs' proposed by the authors.

(5) The abstract and introduction are often too vague or oversimplified.

eLife assessment

This work addresses the connectivity of giant excitatory neurons in a part of CA1 of the hippocampus. Recordings in rat brain slices provide new evidence that these cells excite bistratified and basket inhibitory neurons, and have weak inhibitory input from basket cells, as well as other findings. This circuitry gives these cells unique potential, making the work valuable, however the strength of the evidence is currently incomplete.

Reviewer #2 (Public Review):

Summary:

This study addresses an intriguing and little-studied population of large excitatory cells that lie in the stratum radiatum, outside the classical cell body layers in the hippocampus. Interestingly, the authors show that these "giant excitatory neurons in stratum radiatum" strongly drive both bistratified and basket interneurons. Activating a single giant cell could induce action potential firing in postsynaptic interneurons, which in turn inhibit their postsynaptic pyramidal cell targets. They appear to receive excitatory input from CA3 but not the entorhinal cortex; at a local level, they are not strongly interconnected with CA1 pyramidal cells, and receive inhibitory input from bistratified but not basket cells.

The lack of perisomatic input from basket cells is unique in comparison with the vast majority of excitatory cells in the hippocampus. It is however not surprising, given the fact that the giant excitatory neurons studied in this paper are defined by their position in a particular hippocampal layer (stratum radiatum), and the axons of inhibitory basket cells are largely restricted to another layer (stratum pyramidale). Nonetheless, the fact that this study draws attention to this unique property, and also provides data to support it, is valuable. As the authors also point out, given the importance of such perisomatic input for rhythmogenesis in the hippocampus, the lack of such input may leave these cells free to operate outside of the dominant rhythm.

In combination with the strong drive onto interneurons, which strongly control the activity of pyramidal cells, the giant excitatory cells in the stratum radiatum appear to be in a unique position to influence the hippocampal circuit. Although clearly such an alternative pathway provides the potential for more diverse functions within the hippocampal circuit, and the connectivity shown in this study will likely be of interest to anyone interested in hippocampal function, the authors do not show a concrete function for this pathway.

Strengths:

Overall, the main value of this study is to demonstrate that this small population of oft-neglected cells could have an unexpectedly large impact on hippocampal function via a uniquely strong excitatory output onto two types of interneurons. Whereas activating a "classical" pyramidal cell produces only subthreshold activity in postsynaptic interneurons, meaning that several pyramidal cells have to be co-active to drive their postsynaptic targets to fire, here the authors show that a single giant excitatory neuron in the stratum radiatum can directly drive at least a subset of its postsynaptic targets to fire.

The authors also show the effect of this output both on the membrane potential of CA1 pyramidal cells and on the extracellular field potential as measured with silicon probes. The fact that the authors identified a relatively large number of these sparse giant excitatory cells in the stratum radiatum and performed paired recordings from them is itself a strength of this study.

Another strength is the fact that the authors also investigate the inputs to these giant excitatory cells. The method of paired patch-clamp recordings in rat brain slices enables in principle to record connectivity in both directions, by stimulating one and checking for a response in the other. Recording the interconnectivity of giant excitatory cells with bistratified, basket, and pyramidal cells, as well as the connectivity between pyramidal cells and the two types of interneurons, allows insightful comparisons between "classic" CA1 pyramidal cells and the displaced giant excitatory cells. Although the lack of connectivity between the latter two cell types that the authors report is not so surprising (given the generally very low connectivity between excitatory cells in CA1), it is nonetheless important data. To also check non-local inputs the authors used optogenetics, whereby a Camk2a promoter likely limited cells expressing channelrhodopsin to mostly excitatory cells.

Weaknesses:

The main weakness of this study is perhaps the lack of a clear function for the described circuitry. Although the authors do speculate on this, it remains to be demonstrated what the role of these cells and their connections with the identified interneuron types might be for hippocampal function.

For the first experimental result, it's not fully clear from the evidence the authors present, that indeed the injections were limited to CA3 (for Figure 1c) and to EC (for Figure 1d). This is important since in theory the CA3 injection could also include e.g. CA2 or CA1 itself, which is not that unlikely given the relatively large injected volume of 1ul per side (bilateral). Similarly for the EC injection, it appears the injection may be 2 ul per side (the methods are a bit ambiguous, unfortunately), and this could lead to infection in e.g. Subiculum. Given that these potential mistargeted areas may also project to CA1, this could obviously change the conclusions one can draw from the optogenetic stimulation results the authors present. Furthermore, for the EC result, the authors assume the response they measure is not monosynaptic, which indeed is likely given the long delay, but to interpret this properly a few recordings with pharmacology would be helpful to really show monosynaptic connections (also for the CA3 inputs). One could also cut the inputs to the DG to show that the delayed EC inputs are abolished then (or instead they may be relayed via local CA1 pyramids receiving EC input). Either way, some additional line of evidence beyond simply the delay would be reassuring. A further worry for the EC result relates to the angle of slicing: can the authors give the reader some reassurance that the lack of monosynaptic inputs is not simply a result of cut connections in the slices they used? Especially since only 5 neurons were recorded with stimulation of presumed EC fibers, it is hard to rule out EC input based on the presented evidence. Related to this, one wonders why in Figures 1D and 1E there are no reported connections from EC to CA1 pyramidal cells (while the authors do include CA1 pyramidal cell recordings for the CA3 stimulation experiments); again this might suggest the connections are simply cut in the slice preparation.

For the connectivity results, the data seem to support the claims, but the conclusions would be improved if the terminology of "privileged" and "escaped" could be avoided. More importantly, the exact criteria for distinguishing between bistratified and basket cells are not fully clear; it seems that the amount of current needed to induce AP firing was the main criterion but there is no figure showing this data (only an example in S1A). The input resistance distributions are overlapping, so this was clearly not used as the main criterion. Showing some pictures of the filled cells as supplemental material would also be helpful to give the reader a bit more confidence that the classification is reliable. In the methods, it is mentioned that 10 cells were filled with biocytin, but the authors don't explicitly state (or show) that the identity was confirmed for all 10 filled cells, and what this was based on. Overall, a bit more info on the giant excitatory cells in the stratum radiatum would be helpful (e.g. soma locations, extent of dendrites relative to layers, density/nr of cells); a brief mention of this in the introduction or discussion would help the reader to place the work in context.

The number of tested pairs or cells is also a bit low (or unclear) in some cases. For instance, the relatively low number of recordings (n=30) between CA1 pyramidal cells and giant excitatory cells in the radiatum means a low connectivity rate on the order of a few percent cannot be ruled out; it has been shown that even in CA3, which is classically considered a "reciprocally connected" area, such low connectivity rates can still be functionally important (Guzman et al).

For the feedforward inhibition result, the concept of "amplification relay station" that was introduced is not so clear. It is not unexpected that when you strongly innervate BC cells and bring them to spike, as the giant cells in this study do, this activity will in turn inhibit pyramidal cells (and actually quite a lot of them, so that it is not surprising that you can measure IPSCs). Furthermore, the rationale for doing the silicon probe recordings is not well explained, and it would be helpful if the authors could discuss the significance of performing such LFP recordings in slices.

Conceptually, the presentation of perisomatic inhibition as simply silencing pyramidal and granule cells, forming a "burden" that needs to be "overcome" or "bypassed" via an alternative pathway (as in the example the authors give of having an axon coming from a dendrite instead of the presumably "blocked" soma), is not so convincing. Perisomatic inhibition is much more than that, particularly if one takes timing into account (indeed the authors point to its role in rhythmogenesis). This does not detract from the fact that the lack of perisomatic inhibition (at least from fast-spiking basket cells) is likely to have large functional implications, which the authors rightly emphasize.

Reviewer #3 (Public Review):

Summary:

This paper reexamined large excitatory neurons in the stratum radiatum with optogenetics. The findings are valuable because prior studies of the circuitry were confounded by the use of stimulating electrodes placed in different layers where multiple inputs were stimulated at one time. The strength of the evidence for the conclusions is incomplete because of several concerns with the manuscript.

Strengths:

The strength of the study is the very nice presentation of data. Also, there is a nice combination of patching, LFPs, paired recordings, and microelectrode arrays.

Weaknesses:

The limitations are in the conclusions which don't seem fully justified.

Author Response

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

eLife assessment

This study investigated transcriptional profiles of midbrain dopamine neurons using single nucleus RNA (snRNA) sequencing. The authors found more nuanced subgroups of dopamine neurons than previous studies, and idenfied some genes that are preferenally expressed in subpopulaons that are more vulnerable to neurochemical lesions using 6-hydroxydopamine (6OHDA). The reviewers found the results are solid, and the study is overall valuable, providing crical informaon on the heterogeneity and vulnerability of dopamine neurons although the scope is somewhat limited because the result with snRNA is similar to previous results and cell deaths were induced by 6OHDA injecons.

Public Reviews:

Reviewer #1 (Public Review):

In this study by Yaghmaeian Salmani et al., the authors performed single-nuclei RNA sequencing of a large number of cells (>70,000) in the ventral midbrain. The authors focused on cells in the ventral tegmental area (VTA) and substana nigra (SN), which contain heterogeneous cell populaons comprising dopaminergic, GABAergic, and glutamatergic neurons. Dopamine neurons are known to consist of heterogeneous subtypes, and these cells have been implicated in various neuropsychiatric diseases. Thus, idenfying specific marker genes across different dopamine subpopulaons may allow researchers in future studies to develop dopamine subtype-specific targeng strategies that could have substanal translaonal implicaons for developing more specific therapies for neuropsychiatric diseases.

A strength of the authors' approach compared to previous work is that a large number of cells were sequenced, which was achieved using snRNA-seq, which the authors found to be superior compared to scRNA-seq for reducing sampling bias. A weakness of the study is that relavely litle new informaon is provided as the results are largely consistent with previous studies (e.g., Poulin et al., 2014). Nevertheless, it should be noted that the authors found some more nuanced subdivisions in several genecally idenfied DA subtypes.

On this point we respectfully disagree with the reviewer. In this study, over 30,000 mDA neurons have been analyzed at the genome-wide gene expression level, idenfying mDA territories and neighborhoods (that some may call “subtypes”), a descripon of the mDA neuron diversity that goes far beyond what has been published previously.

Although several single-cell RNA sequencing studies of mDA neurons have added to our understanding of mDA diversity, they have been limited by the low numbers of sequenced mDA neurons. As the reviewer specifically referred to the study by Poulin et al., 2014, it should be noted that in this report, 159 mDA neurons were analyzed by qPCR – not by RNAseq – of 96 previously identified marker genes. Despite those limitaons, this was indeed a highly impressive study, suggesng five different mDA neuron subtypes (as compared to the 16 neighborhoods described here), published before the era of single-cell genome-wide gene expression methods and advanced bioinformac tools were available. On average, the following scRNAseq studies typically captured a few hundred mDA neurons - compared to over 30,000 in this study. None of the studies menoned in our manuscript were close to capturing the full diversity, and the informaon on mDA neuron diversity is, for this reason, somewhat fragmented in the scienfic literature. Indeed, the seven mDA “subtypes” described in the excellent reviews by Poulin et al., 2020 in Trends in Neurosciences and Garritsen et al., 2023 in Nature Neuroscience are integrated interpretaons of the results from numerous independent studies, each methodologically unique. Several previously idenfied groups, especially Vglut2+ populaons in VTA and SNpc, have been considered poorly defined. As menoned above, our findings in this study could reliably idenfy, by computaonal analyses and combinatorial marker expression in situ, 16 different neighborhoods within the mDA populaon and localize them in the ssue (Figure 4, Supplementary figures 4-1 to 4-3, described further in Supplementary Results). To menon three examples: Within Sox6+ SNpc, we idenfied four different variants (neighborhoods) with partly unique anatomical localizaon. In addion, the large group of mDA neurons referred to as the Pcsk6 territory has not been clearly defined in earlier studies. We also idenfied a novel mDA neuron group that is related to the previously well described Vip-expressing mDA neurons. These and other novel features are menoned in the manuscript and in Supplementary Figure 4-1 to 4-3.

Although we have, for the consideraon of the space and intelligibility, characterized the 16 neighborhoods with only a few selected key marker genes, we have idenfied numerous addional novel markers, some of which are shown in dot plots in Figure 3 and Supplementary Figure 3, which can be used to characterize these groups further. We also provide all our sequencing data and our Padlock probe ISS data for anyone to download and analyze further, and we have made a web-based tool, CELLxGENE, available on our group’s website to facilitate exploraon of the different aspects of our dataset.

Lastly, the authors performed molecular analysis of ventral midbrain cells in response to 6-OHDA exposure, which leads to the degeneraon of SN dopamine neurons, whereas VTA dopamine neurons are mainly unaffected. Based on this analysis, the authors idenfied several candidate genes that may be linked to neuronal vulnerability or resilience.

Overall, the authors present a comprehensive mouse brain atlas detailing gene expression profiles of ventral midbrain cell populaons, which will be important to guide future studies that focus on understanding dopamine heterogeneity in health and disease.<br /> We thank the reviewer for poinng this out.

Reviewer #2 (Public Review):

In the manuscript by Salmani et al., the authors explore the transcriptomic characterizaon of dopamine neurons in order to explore which neurons are parcularly vulnerable to 6-OHDA-induced toxicity. To do this they perform single nucleus RNA sequencing of a large number of cells in the mouse midbrain in control animals and those exposed to 6-OHDA. This manuscript provides a detailed atlas of the transcriptome of various types of ventral midbrain cells - though the focus here is on dopaminergic cells, the data can be mined by other groups interested in other cell types as well.

The results in terms of cell type classificaon are largely consistent with previous studies, though a more nuanced picture of cellular subtypes is portrayed here, a unique advantage of the large dataset obtained. The major advance here is exploring the transcriponal profile in the ventral midbrain of animals treated with 6-OHDA, highlighng potenal candidate genes that may influence vulnerability. This approach could be generalizable to invesgate how various experiences and insults alter unique cell subtypes in the midbrain, providing valuable informaon about how these smuli impact DA cell biology and which cells may be the most strongly affected.

We appreciate these comments. We want to state that the study not only gives a more nuanced picture but goes far beyond previously published studies and provides a highly resolved and detailed atlas of mDA neurons. Thus, it clarifies poorly described diversity and idenfies enrely novel groups of diverse mDA neurons at the genome-wide gene expression level.

Overall, the manuscript is relavely heavy on characterizaon and comparavely light on funconal interpretaon of findings. This limits the impact of the proposed work. It also isn't clear what the vulnerability factors may be in the neurons that die. Beyond the characterizaon of which neurons die - what is the reason that these neurons are suscepble to lesion? Also, the interpretaon of these findings is going to be limited by the fact that 6-OHDA is an injectable, and the effects depend on the accuracy of injecon targeng and the equal access of the toxin to access all cell populaons. Though the site of injecon (MFB) should hit most/all of the forebrain-projecng DA cells, the injecon sites for each animal were not characterized (and since the cells from animals were pooled, the effects of injecon targeng on the group data would be hard to determine in any case).

We agree that the results are presented to provide a comprehensive and valuable resource rather than explaining molecular mechanisms. The reviewer points out that “what the vulnerability factors may be in the neurons that die” is unclear. However, our study was designed to answer the queson: What genes are enriched in clusters of mDA neurons that are parcularly likely to die aer toxic stress? Using single-cell analysis, we believe this queson had higher priority than atempng to idenfy gene expression changes occurring during the cell death process. We agree that we cannot answer why neurons are suscepble to lesions, only idenfy genes that correlate with either high or low sensivity. Thus, the genes we refer to as “vulnerability genes” and “resilience genes” are candidates for influencing differenal vulnerability. Hard evidence for such influence will require addional and extensive funconal analysis. As for the variability of injecon and the characterizaon of individual animals, we wish to menon the online interacve explorer available at htps://perlmannlab.org/resources/. It allows visualizaon of nuclei distribuon per territory and neighborhood for each mouse, making it easy to determine the cell loss rao and cell distribuon per animal. There is indeed variance in the proporons of intact/lesioned total nuclei per animal. This is also evident from the DAT autoradiographs shown for each lesioned animal and presented in Figure Supplement 5-1 A. Importantly, the relave UMAP distribuon of nuclei is quite similar between individual animals. To further invesgate this, we used Pearson’s Chi square test of independence with a conngency table for animals, each with two categorical variables as the proporon of nuclei from intact vs lesioned parts of the vMB (see added Supplementary figure 5-1 C ). This shows that – while there is a difference in the number of nuclei remaining aer lesioning – the relave distribuon among clusters and neighborhoods is similar between animals. We have clarified this point in the manuscript (see page 12 ).

I am also not clear why the authors don't explore more about what the genes/pathways are that differenate these condions and why some cells are parcularly vulnerable or resilient. For example, one could run GO analyses, weighted gene co-expression network analysis, or any one of a number of analysis packages to highlight which genes/pathways may give rise to vulnerability or resilience. Since the manuscript is focused on idenfying cells and gene expression profiles that define vulnerability and resilience, there is much more that could have been done with this based on the data that the authors collected.

We performed GO analysis for the genes upregulated and downregulated in the ML clusters (specific to the lesion condion) in the original manuscript (Please see figure supplement 7-1 C-E, and the newly added Supplementary file 10), but we agree with the reviewer that we could also have analyzed funconal categories of genes correlang with differenal vulnerability. Thus, we have used tools recently developed by Morabito et al., Cell Reports Methods (2023), and their hdWGCNA package to address this queson. This method is parcularly suitable for analyzing high-dimensional transcriptomics data such as single-cell RNA-seq or spaal transcriptomics. We calculated the coexpression network based on the lesioned nuclei of the mDA territories. Of the 9 co-expression modules calculated, one has the highest expression in Sox6 territory and has genes in common with the vulnerability module. Another co-expression module has genes in common with the resilience module and is most highly expressed in Otx2 and Ebf1 territories. We also did GO analysis for these co-expression modules and added addional GO analysis of the ML-enriched genes (see Supplementary Figure 7-1 D,E, the newly added Supplementary Figure 6-3, and the newly added Supplementary file 9). Text describing these addional analyses are menoned on page 15 and 17.

In addition, we wish to emphasize our idenficaon of the genes we refer to as vulnerability and resilience modules in the previous version of the manuscript. Several of the genes were discussed in the previous version of the manuscript but we have now included more informaon on these genes, based on previously published studies and discuss their potenal funconal roles (see pages 22 & 23 in the Discussion).

Another limitation of this study as presented is the missed opportunity to integrate it with the rich literature on midbrain dopamine (and non-dopamine) neuron subtypes. Many subtypes have been explored, with divergent funcons, and can usually be disnguished by either their projecon site, neurotransmiter identy, or both. Unfortunately, the projecon site does not seem to track parcularly well with transcriptomic idenes, aside from a few genes such as DAT or the DRD2 receptor. However, this could have been more thoroughly explored in this manuscript, either by introducing AAVretro barcodes through injecon into downstream brain sites, or through exisng evidence within their sequencing dataset. There are likely clear interpretaons from some of that literature, some of which may be more excing than others. For example, the authors note that vGluT2-expressing cells were part of the resilient territory. This might be because this is expressed in medially-located DA cells and not laterally-located ones, which tends to track which cells die and which don't.

The manuscript consists of a comprehensive descripon of transcriponal diversity. Although of clear value, we believe that addional, comprehensive analysis that combines snRNAseq with, e.g., AAVretro barcodes must be done in a separate study. It should also be noted that we describe each territory and neighborhoods in the further detail in the Supplementary Results, which contains references to the relevant literature. In line with the comments, this secon has now been expanded with further references to relevant studies (see Supplementary Results related to Figure 4-figure supplements 1-3).

It is not immediately clear why the authors used a relaxed gate for mCherry fluorescence in Figure 1. This makes it difficult to definively isolate dopaminergic neurons - or at least, neurons with a DATCre expression history. While the expression of TH/DAT should be able to give a fairly reliable idenficaon of these cells, the reason for this decision is not made clear in the text.

We used a relaxed gang to ensure that we could capture nuclei expressing low levels of RFP, which we believe could be especially relevant for the lesioned dataset (see page 5). We did not find that it would be advantageous to use a more stringent gang that would risk losing all cells expressing no (or very low levels) RFP. Idenfying mDA neurons based on their typical markers is straighorward, as their transcriponal relaonship is evident from the expression profile of several markers, including transcripon factors such as Nr4a2, Pitx3, and En1. In addion, as pointed out in response to Reviewer #1, point 5, atypical DA neurons expressing Th and other mDA markers with no or low levels of Slc6a3 (DAT) were isolated. We believe the study is more complete by the inclusion of these cells. Moreover, we included a sufficiently large number of cells, which ensured a comprehensive analysis of mDA neurons in relaon to other cell types dissected from the ventral midbrain.

Reviewer #1 (Recommendations For The Authors):

(1) The authors state that a major advantage of their approach is that it prevents biased datasets when compared to methods that rely on capturing certain cell types. I was wondering if the authors could follow up on this topic with a more detailed descripon of their methodological advantages regarding potenal sampling bias. This is somewhat unclear to me, given that the results of the present study are largely consistent with previous work on this topic.

As expanded on above (see response to the inial comment in the public review), we strongly disagree that there is litle novelty in our study. None of the previous studies come close to describing the mDA neuron populaon with a similar resoluon, which is unsurprising given the differences in the number of analyzed mDA neurons in this versus previous reports. We agree with the reviewer that our data is consistent with previous studies, when they are all combined. Thus, we idenfied mDA neuron groups that correspond (or roughly correspond) to major DA neuron groups idenfied in previous studies (see pages 8-14 in the Supplementary Results). However, the atlas presented here goes well beyond anything published in scope and resoluon. The diversity we define is comparable to findings that, with careful cross-paper analyses, can be stched together from previous single-cell studies. However, even such a combined analysis does not unravel the resoluon and diverse categorizaon of what we have demonstrated herein (16 neighborhoods in midbrain dopaminergic territories). Considering the well-established problems of dissociang and isolang whole neurons from adult brain ssue, this is likely due to sampling bias, resulng in an almost complete exclusion of some sub-populaons of neurons. We have added text on page 20 to clarify this point.

(2) In the abstract, the authors state that their "results showed that differences between mDA neuron group could best be understood as a connuum without sharp differences between subtypes". However, I am not sure whether this is the most appropriate descripon of the authors' results, parcularly when looking at the schemac overview shown in Fig. 4F. To me, it seems more likely that genecally-defined DA subtypes overlap with discrete ventral midbrain subnuclei - parcularly in the case of Sox6-expressing cells, which are almost exclusively located in the SNc. In the case of genes that are specific for the VTA, there also seems to be a strong bias toward certain VTA subnuclei, although I agree that arguments can be made that there is some topographic organizaon along a dorso-ventral and medio-lateral gradient, which seems to be largely consistent with the anatomical locaon of projecon-defined dopamine neurons as described previously by Poulin et al., 2018 (Nature Neuroscience).

What was meant by connuum must be interpreted in the context of the transcriponal landscape of mDA neurons and not their anatomical localizaon. As stated in the paper, the dendrogram depicon of mDA neurons’ transcriptome can be misinterpreted as an indicaon of sharp boundaries and discrete groups in transcriponal profiles. In contrast, we assert that differences between developmentally related mDA neurons are beter described as a connuum with areas in the gene expression landscape defined by the expression of shared genes but without sharp borders between them. We decided to name different areas within this connuum as “territories” at the higher hierarchical level and “neighborhoods” at the more highly resolved level. Hypothecally, such categorizaon can be even more fine-grained, but we find it unlikely that a resoluon beyond the neighborhood level is biologically relevant. As pointed out, the Sox6 territory is the territory that best qualifies as a disncve subtype, while mDA neurons in, e.g., the VTA consist of much higher and nuanced diversity. Importantly, all mDA neurons are much more related to each other than cell types lacking a common developmental origin, including hypothalamic DA neurons. Thus, our effort to define differences in such a gene expression connuum is, in our opinion, more accurate than conveying the message that the diversity consists of subtypes comparable in difference to other cell types that lack a close developmental relaonship with the mDA neuron populaon. Such disnct neuron types, despite using the same neurotransmiter as hypothalamic DA neurons, appear as disnct islands in the UMAP snRNA-seq landscape and typically harbor hundreds of differenally expressed genes. As pointed out in the Discussion, several other studies have noted similar difficules in defining different subtypes among related neurons in e.g. the cortex, striatum, and hippocampus (Kozareva et al., 2021; Saunders et al., 2018; Tasic et al., 2018; Yao et al., 2021). For example, Yao et al., 2021, used a similar hierarchical definion to avoid the implicaon that different groups (“neighborhoods” in this study) should be defined as disnct subtypes of neurons with obvious disncve funcons.

(3) I recommend that the authors revise the introducon to include more current literature on this topic. The review by Bjoerklund and Dunnet, 2006, is very informave and important, but there is more current literature available that discusses anatomical, molecular, and funconal heterogeneity in the ventral midbrain. For example, it would be nice to incorporate recent work from the Awatramani lab on the mapping of the projecon of molecularly defined dopamine neurons (Poulin et al., 2018; Nature Neuroscience).

We deliberately avoided including primary references to previously described diversity in the Introducon since numerous papers are relevant to cite. Instead, we refer to three essenal reviews, including the recent arcles from Awatramani and Pasterkamp. In the Supplementary Results related to Figure 4 (pages 8-14 in the Supplementary Results), we include many references and the Poulin 2018 paper. We believe that this is the appropriate place for a comprehensive discussion on anatomical, molecular, and funconal heterogeneity. In the revised manuscript's main body, we now emphasize that previous literature is discussed in the Supplementary Results (see page 11).

(4) In Fig. 1C, the authors show a sample image demonstrang overlap between TH and mCherry, but this has not been quanfied. Similarly, there seem to be no sample images and quanficaon for the contralateral side that was exposed to 6-OHDA.

The mouse lines used here (Dat-Cre and Rpl10a-mCherry) have been characterized before (Toskas et al., Science Advances 2022). The labelling colocalizes nearly fully with TH, with some excepons (see response below to point #5). We have now complemented with addional data showing an IHC image of one of the midbrain of a unilaterally lesioned mouse in Figure Supplement 5-1E.

(5) The authors state that they focused their analysis on 33,052 nuclei expressing above-threshold levels of either Th OR Slc6a3. However, there seem to be cell populaons in the ventral midbrain of mice that express TH mRNA but not TH protein, and these cells do not seem to be bona fide dopamine neurons (see work from the Morales lab). Similarly, not all dopamine neurons may express DAT mRNA. I was wondering how these discrepancies may influence the authors' analysis and interpretaon.

Indeed, the presence of cells lacking TH protein despite Th mRNA being expressed has been previously described. We also detected these cells across SNpc and VTA and now show these data as a newly added supplementary figure 2-1. In our dataset, the Gad2 territory, located in the ventromedial VTA, contains cells that express many typical mDA markers, such as Pitx3, but very low levels of TH protein. We have idenfied these based on Pitx3-EGFP and Gad2 mRNA co-expression (figure supplement 4-3). In other parts of VTA and SNpc, most cells seem to co-express Th mRNA and protein and are labeled with Dat-Cre. Also scatered in these areas, we could detect some rare mDA cells that lack TH protein. It should be noted that in our mDA territories other typical mDA neuron genes were expressed, such as Slc18a2, Ddc, Nr4a2 and Pitx3, and thus, they were not solely defined by the presence of Th and/or Slc6a3. Cells that do not have a history of DAT-expression, and therefore were not mCherry labelled, were also included in the analysis due to the relaxed gang used during FANS isolaon.

(6) The sex and age of the mice that are used for the experiments are not stated in the Materials and Methods secon under "Mouse lines and genotyping".

Thank you for pointing this out. This informaon has been added to the updated manuscript in the methods secon.

Reviewer #2 (Recommendations For The Authors):

I think that the manuscript can be significantly improved just by providing deeper analyses of the exisng data and linking them to the current state of the art in terms of defining midbrain dopamine neurons (e.g., by projecon). The dataset is likely richer than was explored in the manuscript and more valuable insights could be gleaned with a deeper analysis.

Please see our response to Reviewer #2 (Public Review), regarding WGCNA analysis, and the comments on ML-based GO analysis, as well as the comments on the added secons in the supplementary results file.

eLife assessment

This important study investigated transcriptional profiles of midbrain dopamine neurons using single nucleus RNA (snRNA) sequencing. The authors found more nuanced subgroups of dopamine neurons than previous studies, and identified some genes that are preferentially expressed in subpopulations that are more vulnerable to neurochemical lesions using 6-hydroxydopamine (6OHDA). The results are convincing and provide critical information on the heterogeneity and vulnerability of dopamine neurons which will be a foundation for future studies.

Reviewer #1 (Public Review):

In this study by Yaghmaeian Salmani et al., the authors performed single-nuclei RNA sequencing of a large number of cells (>70,000) in the ventral midbrain. The authors focused on cells in the ventral tegmental area (VTA) and substantia nigra (SN), which contain heterogeneous cell populations comprising dopaminergic, GABAergic, and glutamatergic neurons. Dopamine neurons are known to consist of heterogeneous subtypes, and these cells have been implicated in various neuropsychiatric diseases. Thus, identifying specific marker genes across different dopamine subpopulations may allow researchers in future studies to develop dopamine subtype-specific targeting strategies that could have substantial translational implications for developing more specific therapies for neuropsychiatric diseases.

A strength of the authors' approach compared to previous work is that a large number of cells were sequenced, which was achieved using snRNA-seq, which the authors found to be superior compared to scRNA-seq for reducing sampling bias. A weakness of the study is that relatively little new information is provided as the results are largely consistent with previous studies (e.g., Poulin et al., 2014). Nevertheless, it should be noted that the authors found some more nuanced subdivisions in several genetically identified DA subtypes.

Lastly, the authors performed molecular analysis of ventral midbrain cells in response to 6-OHDA exposure, which leads to the degeneration of SN dopamine neurons, whereas VTA dopamine neurons are largely unaffected. Based on this analysis, the authors identified several candidate genes that may be linked to neuronal vulnerability or resilience.

Overall, the authors present a comprehensive mouse brain atlas detailing gene expression profiles of ventral midbrain cell populations, which will be important to guide future studies that focus on understanding dopamine heterogeneity in health and disease.

Comments on the revised version

The authors have addressed all of my concerns.

Reviewer #2 (Public Review):

In the manuscript by Salmani et al., the authors explore the transcriptomic characterization of dopamine neurons in order to explore which neurons are particularly vulnerable to 6-OHDA-induced toxicity. To do this they perform single nucleus RNA sequencing of a large number of cells in the mouse midbrain in control animals and those exposed to 6-OHDA. This manuscript provides a detailed atlas of the transcriptome of various types of ventral midbrain cells - though the focus here is on dopaminergic cells, the data can be mined by other groups interested in other cell types as well. The results in terms of cell type classification are largely consistent with previous studies, though a more nuanced picture of cellular subtypes is portrayed here, a unique advantage of the large dataset obtained. The major advance here is exploring the transcriptional profile in the ventral midbrain of animals treated with 6-OHDA, highlighting potential candidate genes that may influence vulnerability. This approach could be generalizable to investigate how various experiences and insults alter unique cell subtypes in the midbrain, providing valuable information about how these stimuli impact DA cell biology and which cells may be the most strongly affected.

Comments on the revised version

The authors addressed most of my concerns about the depth of analysis and implemented further analyses of the data. However I still think that the manuscript would be strengthened with an acknowledgement and deeper integration with the concepts from recent papers in the field, as mentioned by Reviewer 1. There is a rich amount of biology that can be gleaned from understanding the anatomical topology of the VTA and how that relates to gene expression patterns, both at a basal state and following 6-OHDA injection. For example, I made the point about medially-located DA cells in the VTA being the DA that co-express vGluT2. The work would provide more value to the field if more effort was made in the introduction and discussion to briefly mention the recent key papers in the field and how their work relates to our knowledge of the VTA and adjacent SNc in terms of cell-type identity, spatial location, and co-expression of various genes e.g., DAT and vGluT2.

Reviewer #1 (Public Review):

Summary:<br /> Fiber photometry has become a very popular tool in recording neuronal activity in freely behaving animals. Despite the number of papers published with the method, as the authors rightly note, there are currently no standardized ways to analyze the data produced. Moreover, most of the data analyses confine to simple measurements of averaged activity and by doing so, erase valuable information encoded in the data. The authors offer an approach based on functional linear mixed modeling, where beyond changes in overall activity various functions of the data can also be analyzed. More in-depth analysis, more variables taken into account, and better statistical power all lead to higher quality science.

Strengths:<br /> The framework the authors present is solid and well-explained. By reanalyzing formerly published data, the authors also further increase the significance of the proposed tool opening new avenues for reinterpreting already collected data.

Weaknesses:<br /> However, this also leads to several questions. The normalization method employed for raw fiber photometry data is different from lab to lab. This imposes a significant challenge to applying a single tool of analysis. Does the method that the authors propose work similarly efficiently whether the data are normalized in a running average dF/F as it is described in the cited papers? For example, trace smoothing using running averages (Jeong et al. 2022) in itself may lead to pattern dilution. The same question applies if the z-score is calculated based on various responses or even baselines. How reliable the method is if the data are non-stationery and the baselines undergo major changes between separate trials?

Finally, what is the rationale for not using non-linear analysis methods? Following the paper's logic, non-linear analysis can capture more information that is diluted by linear methods.

eLife assessment

This important work presents a new methodology for the statistical analysis of fiber photometry data, improving statistical power while avoiding the bias inherent in the choices that are necessarily made when summarizing photometry data. The reanalysis of two recent photometry data sets, the simulations, and the mathematical detail provide convincing evidence for the utility of the method and the main conclusions, however, the discussion of the re-analyzed data is incomplete and would be improved by a deeper consideration of the limitations of the original data. In addition, consideration of other data sets and photometry methodologies including non-linear analysis tools, as well as a discussion of the importance of the data normalization are needed.

Reviewer #2 (Public Review):

Summary:<br /> This work describes a statistical framework that combines functional linear mixed modeling with joint 95% confidence intervals, which improves statistical power and provides less conservative statistical inferences than in previous studies. As recently reviewed by Simpson et al. (2023), linear regression analysis has been used extensively to analyze time series signals from a wide range of neuroscience recording techniques, with recent studies applying them to photometry data. The novelty of this study lies in 1) the introduction of joint 95% confidence intervals for statistical testing of functional mixed models with nested random-effects, and 2) providing an open-source R package implementing this framework. This study also highlights how summary statistics as opposed to trial-by-trial analysis can obscure or even change the direction of statistical results by reanalyzing two other studies.

Strengths:<br /> The open-source package in R using a similar syntax as the lme4 package for the implementation of this framework on photometry data enhances the accessibility, and usage by other researchers. Moreover, the decreased fitting time of the model in comparison with a similar package on simulated data, has the potential to be more easily adopted.

The reanalysis of two studies using summary statistics on photometry data (Jeong et al., 2022; Coddington et al., 2023) highlights how trial-by-trial analysis at each time-point on the trial can reveal information obscured by averaging across trials. Furthermore, this work also exemplifies how session and subject variability can lead to opposite conclusions when not considered.

Weaknesses:<br /> Although this work has reanalyzed previous work that used summary statistics, it does not compare with other studies that use trial-by-trial photometry data across time-points in a trial.

As described by the authors, fitting pointwise linear mixed models and performing t-test and Benjamini-Hochberg correction as performed in Lee et al. (2019) has some caveats. Using joint confidence intervals has the potential to improve statistical robustness, however, this is not directly shown with temporal data in this work. Furthermore, it is unclear how FLMM differs from the pointwise linear mixed modeling used in this work.

In this work, FLMM usages included only one or two covariates. However, in complex behavioral experiments, where variables are correlated, more than two may be needed (see Simpson et al. (2023), Engelhard et al. (2019); Blanco-Pozo et al. (2024)). It is not clear from this work, how feasible computationally would be to fit such complex models, which would also include more complex random effects.

Reviewer #3 (Public Review):

Summary:<br /> Loewinger et al., extend a previously described framework (Cui et al., 2021) to provide new methods for statistical analysis of fiber photometry data. The methodology combines functional regression with linear mixed models, allowing inference on complex study designs that are common in photometry studies. To demonstrate its utility, they reanalyze datasets from two recent fiber photometry studies into mesolimbic dopamine. Then, through simulation, they demonstrate the superiority of their approach compared to other common methods.

Strengths:<br /> The statistical framework described provides a powerful way to analyze photometry data and potentially other similar signals. The provided package makes this methodology easy to implement and the extensively worked examples of reanalysis provide a useful guide to others on how to correctly specify models.

Modeling the entire trial (function regression) removes the need to choose appropriate summary statistics, removing the opportunity to introduce bias, for example in searching for optimal windows in which to calculate the AUC. This is demonstrated in the re-analysis of Jeong et al., 2022, in which the AUC measures presented masked important details about how the photometry signal was changing.

Meanwhile, using linear mixed methods allows for the estimation of random effects, which are an important consideration given the repeated-measures design of most photometry studies.

Weaknesses:<br /> While the availability of the software package (fastFMM), the provided code, and worked examples used in the paper are undoubtedly helpful to those wanting to use these methods, some concepts could be explained more thoroughly for a general neuroscience audience.

While the methodology is sound and the discussion of its benefits is good, the interpretation and discussion of the re-analyzed results are poor:

In section 2.3, the authors use FLMM to identify an instance of Simpson's Paradox in the analysis of Jeong et al. (2022). While this phenomenon is evident in the original authors' metrics (replotted in Figure 5A), FLMM provides a convenient method to identify these effects while illustrating the deficiencies of the original authors' approach of concatenating a different number of sessions for each animal and ignoring potential within-session effects. The discussion of this result is muddled. Having identified the paradox, there is some appropriate speculation as to what is causing these opposing effects, particularly the decrease in sessions. In the discussion and appendices, the authors identify (1) changes in satiation/habitation/motivation, (2) the predictability of the rewards (presumably by the click of a solenoid valve) and (3) photobleaching as potential explanations of the decrease within days. Having identified these effects, but without strong evidence to rule all three out, the discussion of whether RPE or ANCCR matches these results is probably moot. In particular, the hypotheses developed by Jeong et al., were for a random (unpredictable) rewards experiment, whereas the evidence points to the rewards being sometimes predictable. The learning of that predictability (e.g. over sessions) and variation in predictability (e.g. by attention level to sounds of each mouse) significantly complicate the analysis. The FLMM analysis reveals the complexity of analyzing what is apparently a straightforward task design. If this paper is not trying to arbitrate between RPE and ANCCR, as stated in the text, the post hoc reasoning of the authors of Jeong et al 2022 provided in the discussion is not germane. Arbitrating between the models likely requires new experimental designs (removing the sound of the solenoid, satiety controls) or more complex models (e.g. with session effects, measures of predictability) that address the identified issues.

Of the three potential causes of within-session decreases, the photobleaching arguments advanced in the discussion and expanded greatly in the appendices are not convincing. The data being modeled is a processed signal (ΔF/F) with smoothing and baseline correction and this does not seem to have been considered in the argument. Furthermore, the photometry readout is also a convolution of the actual concentration changes over time, influenced by the on-off kinetics of the sensor, which makes the interpretation of timing effects of photobleaching less obvious than presented here and more complex than the dyes considered in the cited reference used as a foundation for this line of reasoning.

Within this discussion of photobleaching, the characterization of the background reward experiments used in part to consider photobleaching (appendix 7.3.2) is incorrect. In this experiment (Jeong et al., 2022), background rewards were only delivered in the inter-trial-interval (i.e. not between the CS+ and predicted reward as stated in the text). Both in the authors' description and in the data, there is a 6s before cue onset where rewards are not delivered and while not described in the text, the data suggests there is a period after a predicted reward when background rewards are not delivered. This complicates the comparison of this data to the random reward experiment.

The discussion of the lack of evidence for backpropagation, taken as evidence for ANCCR over RPE, is also weak. A more useful exercise than comparing FLMM to the methods and data of Jeong et al., 2022, would be to compare against the approach of Amo et al., 2022, which identifies backpropagation (data publicly available: DOI: 10.5061/dryad.hhmgqnkjw). The replication of a positive result would be more convincing of the sensitivity of the methodology than the replication of a negative result, which could be a result of many factors in the experimental design. Given that the Amo et al. analysis relies on identifying systematic changes in the timing of a signal over time, this would be particularly useful in understanding if the smoothing steps in FLMM obscure such changes.

Author Response

eLife assessment

This study, which seeks to identify factors from the glial niche that support and maintain neural stem cells, unveils a novel role for ferritin in this process. Furthermore, the work shows that defects in larval brain development resulting from ferritin knockdown can be attributed to impaired Fe-S cluster activity and ATP production. These findings will be valuable to both oncologists and neurobiologists, though the supporting evidence is currently incomplete.

Public Reviews

Reviewer #1 (Public Review):

Summary:

This study unveils a novel role for ferritin in Drosophila larval brain development. Furthermore, it pinpoints that the observed defects in larval brain development resulting from ferritin knockdown are attributed to impaired Fe-S cluster activity and ATP production. In addition, knocking down ferritin genes suppressed the formation of brain tumors induced by brat or numb RNAi in Drosophila larval brains. Similarly, iron deficiency suppressed glioma in the mice model. Overall, this is a well-conducted and novel study.

Strengths:

Thorough analyses with the elucidation of molecular mechanisms.

Weaknesses:

Some of the conclusions are not well supported by the results presented.

We really appreciate your review and positive feedback. As for weaknesses, we will try our best to solidate the related conclusions.

Reviewer #2 (Public Review):

Summary:

Zhixin and collaborators have investigated if the molecular pathways present in glia play a role in the proliferation, maintenance, and differentiation of Neural Stem Cells. In this case, Drosophila Neuroblasts are used as models. The authors find that neuronal iron metabolism modulated by glial ferritin is an essential element for Neuroblast proliferation and differentiation. They show that loss of glial ferritin is sufficient to impact on the number of neuroblasts. Remarkably, the authors have identified that ferritin produced in the glia is secreted to be used as an iron source by the neurons. Therefore iron defects in glia have serious consequences in neuroblasts and likely vice versa. Interestingly, preventing iron absorption in the intestine is sufficient to reduce NB number. Furthermore, they have identified Zip13 as another regulator of the process. The evidence presented strongly indicates that loss of neuroblasts is due to premature differentiation rather than cell death.

Strengths:

• Comprenhensive analysis of the impact of glial iron metabolism in neuroblast behaviour by genetic and drug-based approaches as well as using a second model (mouse) for some validations.

• Using cutting-edge methods such as RNAseq as well as very elegant and clean approaches such as RNAi-resistant lines or temperature-sensitive tools

• Goes beyond the state of the art highlighting iron as a key element in neuroblast formation as well as as a target in tumor treatments.

Weaknesses:

Although the manuscripts have clear strengths, there are also some strong weaknesses that need to be addressed.

• Some literature is missing

Thanks for your reminder and we will add the missing literatures.

• In general, the authors succeeded but in some cases, the authors´ claims are not fully supported by the evidence presented and additional experiments are critical to discriminate among different hypotheses.

We are greatly grateful to the reviewer for recognizing our work, and we will support our conclusions with further evidence.

• Moreover, some potential flaws might be present in the analysis of cell death and mitochondrial iron.

We used Caspase-3 or TUNEL to indicate the apoptosis signal. Further, we overexpressed the anti-apoptosis gene p35 to inhibit apoptosis and found no rescue effect on neuroblast number. The results of these experiments are consistent.

It is difficult to determine the mitochondrial iron of neuroblast, so we used indirect methods to test ferroptosis, such as TEM and iron (or iron chelator) supplement. We will perform more experiments according to recommendations to determine that.

Reviewer #3 (Public Review):

In this manuscript, Ma et al seek to identify stem cell niche factors. They perform an RNAi screen in glial cells and screen for candidates that support and maintain neuroblasts (NBs) in the developing fly brain. Through this, they identify two subunits of ferritin, which is a conserved protein that can store iron in cells in a non-toxic form and release it in a controlled manner when and where required. They present data to support the conclusion that ferritin produced in glia is released and taken up by NBs where it is utilised by enzymes in the Krebs cycle as well as in the electron transport chain. In its absence from glia, NBs are unable to generate sufficient energy for division and therefore prematurely differentiate via nuclear prospero resulting in small brains. The work will be of interest to those interested in neural stem cells and their non-cell autonomous control by niches.

The past decade has seen a growing appreciation of how glial cells support and maintain NBs during development.

The authors' discovery of glial-derived ferritin providing essential iron atoms for energy production is interesting and important. They have employed a variety of genetic tools and assays to uncover how ferritin in glia might support NBs. This is particularly challenging because there are no direct ways of assaying for iron or energy consumption in a cell-specific manner.

There are however instances where conclusions are drawn to support the story being developed without considering the equally plausible alternative explanations that should ideally be addressed.

For example, the data supporting the transfer of ferritin from glia to NBs was weak given the misexpression system used; the Shi[ts] experiment was also not convincing (perhaps they have more representative images?).

Thanks for your comment. We have the negative control, which excludes the misexpression. As for Shits experiment, we will substitute for more representative images.

The iron manipulation experiments are in the whole animal and it is likely that this affects general feeding behaviour, which is known to affect NB exit from quiescence and proliferative capacity. The loss of ferritin in the gut and iron chelators enhancing the NB phenotype are used as evidence that glia provide iron to NB to support their number and proliferation. Since the loss of NB is a phenotype that could result from many possible underlying causes (including low nutrition), this specific conclusion is one of many possibilities.

Iron chelator (or iron salt) feeding is a common method for investigating metal metabolism in Drosophila[1-3]. And other metal chelators (such as copper and zinc chelator) do not have similar phenotype (data not shown), which can partially exclude this possibility. Further, iron absorption was blocked by knockdown of ferritin only in the iron cell region[1], a small part of midgut, which phenocopied iron chelator feeding, implying iron deficiency is probably the main cause of the phenotype. More importantly, iron chelator only enhances the NB phenotype in the ferritin knockdown group, not the control group, suggesting iron deficiency results in the phenotype, which rules out other possibilities.

Similarly, knockdown of the FeS protein assembly components phenocopy glial ferritin knock down. Since iron is so important for the TCA and the ETC, this is not surprising, but the similarities in the two phenotypes seem insufficient to say that it's glial ferritin that's causing the lack of iron in the NB and therefore resulting in loss of NBs.

It is hard to get this conclusion just by FeS protein assembly components knockdown, so we just used “implied” to describe this result. However, we combine several results to address this issue, including iron chelator feeding, ferritin knockdown in the midgut, the enhancement of phenotype by iron chelators, aconitase activity, GO enrichment, KEGG enrichment, and Zip13. These results pointed to the interpretation that iron deficiency in NBs caused by glial ferritin defects leads to NB loss.

Pros RNAi will certainly result in an increase in NB numbers because the loss of pros results in an inability of NB progeny to differentiate. This (despite the slight increase in nuclear pros) is not sufficient to infer that glial ferritin knockdown results in premature differentiation of NBs via nuclear pros.

First, pros RNAi, brat RNAi, or numb RNAi can each result in an inability of NB progeny to differentiate, respectively[4-6]. If the rescue of NB number by pros RNAi mainly relies on the differentiation block of NB progeny, brat RNAi or numb RNAi is expected to similarly rescue the NB number. However, our results showed that only pros RNAi could rescue the NB number, while brat RNAi or numb RNAi could not.

Secondly, nuclear Pros represses genes required for self-renewal and is also required to activate genes for terminal differentiation[7]. Thus, Pros is kept in the cytoplasm and remains almost undetectable in the nuclei in normal NBs[8]. However, we observed the detectable Pros in the nuclei of some NBs after glial ferritin knockdown, and the NB number with detectable nuclear Pros was significantly increased when compared to control.

Altogether, we conclude that NBs tend to undergo premature differentiation after glial ferritin knockdown.

I recognise these are challenging to prove irrefutably, however, the frequency of such expansive interpretations of data is of concern.

(1) Tang X, Zhou B. Ferritin is the key to dietary iron absorption and tissue iron detoxification in Drosophila melanogaster. FASEB J, 2013,27(1):288-98

(2) Xiao G, Liu ZH, Zhao M, et al. Transferrin 1 Functions in Iron Trafficking and Genetically Interacts with Ferritin in Drosophila melanogaster. Cell Rep, 2019,26(3):748-58 e5

(3) Mukherjee C, Kling T, Russo B, et al. Oligodendrocytes Provide Antioxidant Defense Function for Neurons by Secreting Ferritin Heavy Chain. Cell Metab, 2020,32(2):259-72 e10

(4) Knoblich JA, Jan LY, Jan YN. Asymmetric Segregation of Numb and Prospero during Cell-Division. Nature, 1995,377(6550):624-7

(5) Zacharioudaki E, Magadi SS, Delidakis C. bHLH-O proteins are crucial for neuroblast self-renewal and mediate Notch-induced overproliferation. Development, 2012,139(7):1258-69

(6) Bello B, Reichert H, Hirth F. The brain tumor gene negatively regulates neural progenitor cell proliferation in the larval central brain of. Development, 2006,133(14):2639-48

(7) Choksi SP, Southall TD, Bossing T, et al. Prospero acts as a binary switch between self-renewal and differentiation in Drosophila neural stem cells. Developmental Cell, 2006,11(6):775-89

(8) Spana EP, Doe CQ. The Prospero Transcription Factor Is Asymmetrically Localized to the Cell Cortex during Neuroblast Mitosis in Drosophila. Development, 1995,121(10):3187-95

Reviewer #2 (Public Review):

Summary:<br /> Zhixin and collaborators have investigated if the molecular pathways present in glia play a role in the proliferation, maintenance, and differentiation of Neural Stem Cells. In this case, Drosophila Neuroblasts are used as models. The authors find that neuronal iron metabolism modulated by glial ferritin is an essential element for Neuroblast proliferation and differentiation. They show that loss of glial ferritin is sufficient to impact on the number of neuroblasts. Remarkably, the authors have identified that ferritin produced in the glia is secreted to be used as an iron source by the neurons. Therefore iron defects in glia have serious consequences in neuroblasts and likely vice versa. Interestingly, preventing iron absorption in the intestine is sufficient to reduce NB number. Furthermore, they have identified Zip13 as another regulator of the process. The evidence presented strongly indicates that loss of neuroblasts is due to premature differentiation rather than cell death.

Strengths:<br /> - Comprenhensive analysis of the impact of glial iron metabolism in neuroblast behaviour by genetic and drug-based approaches as well as using a second model (mouse) for some validations.<br /> - Using cutting-edge methods such as RNAseq as well as very elegant and clean approaches such as RNAi-resistant lines or temperature-sensitive tools<br /> - Goes beyond the state of the art highlighting iron as a key element in neuroblast formation as well as as a target in tumor treatments.

Weaknesses:<br /> Although the manuscripts have clear strengths, there are also some strong weaknesses that need to be addressed.<br /> - Some literature is missing<br /> - In general, the authors succeeded but in some cases, the authors´ claims are not fully supported by the evidence presented and additional experiments are critical to discriminate among different hypotheses.<br /> - Moreover, some potential flaws might be present in the analysis of cell death and mitochondrial iron.

eLife assessment

This study, which seeks to identify factors from the glial niche that support and maintain neural stem cells, unveils a novel role for ferritin in this process. Furthermore, the work shows that defects in larval brain development resulting from ferritin knockdown can be attributed to impaired Fe-S cluster activity and ATP production. These findings will be valuable to both oncologists and neurobiologists, though the supporting evidence is currently incomplete.

Reviewer #1 (Public Review):

Summary:<br /> This study unveils a novel role for ferritin in Drosophila larval brain development. Furthermore, it pinpoints that the observed defects in larval brain development resulting from ferritin knockdown are attributed to impaired Fe-S cluster activity and ATP production. In addition, knocking down ferritin genes suppressed the formation of brain tumors induced by brat or numb RNAi in Drosophila larval brains. Similarly, iron deficiency suppressed glioma in the mice model. Overall, this is a well-conducted and novel study.

Strengths:<br /> Thorough analyses with the elucidation of molecular mechanisms.

Weaknesses:<br /> Some of the conclusions are not well supported by the results presented.

Reviewer #3 (Public Review):

In this manuscript, Ma et al seek to identify stem cell niche factors. They perform an RNAi screen in glial cells and screen for candidates that support and maintain neuroblasts (NBs) in the developing fly brain. Through this, they identify two subunits of ferritin, which is a conserved protein that can store iron in cells in a non-toxic form and release it in a controlled manner when and where required. They present data to support the conclusion that ferritin produced in glia is released and taken up by NBs where it is utilised by enzymes in the Krebs cycle as well as in the electron transport chain. In its absence from glia, NBs are unable to generate sufficient energy for division and therefore prematurely differentiate via nuclear prospero resulting in small brains. The work will be of interest to those interested in neural stem cells and their non-cell autonomous control by niches.

The past decade has seen a growing appreciation of how glial cells support and maintain NBs during development. The authors' discovery of glial-derived ferritin providing essential iron atoms for energy production is interesting and important. They have employed a variety of genetic tools and assays to uncover how ferritin in glia might support NBs. This is particularly challenging because there are no direct ways of assaying for iron or energy consumption in a cell-specific manner.

There are however instances where conclusions are drawn to support the story being developed without considering the equally plausible alternative explanations that should ideally be addressed.

For example, the data supporting the transfer of ferritin from glia to NBs was weak given the misexpression system used; the Shi[ts] experiment was also not convincing (perhaps they have more representative images?).

The iron manipulation experiments are in the whole animal and it is likely that this affects general feeding behaviour, which is known to affect NB exit from quiescence and proliferative capacity. The loss of ferritin in the gut and iron chelators enhancing the NB phenotype are used as evidence that glia provide iron to NB to support their number and proliferation. Since the loss of NB is a phenotype that could result from many possible underlying causes (including low nutrition), this specific conclusion is one of many possibilities.

Similarly, knockdown of the FeS protein assembly components phenocopy glial ferritin knock down. Since iron is so important for the TCA and the ETC, this is not surprising, but the similarities in the two phenotypes seem insufficient to say that it's glial ferritin that's causing the lack of iron in the NB and therefore resulting in loss of NBs.

Pros RNAi will certainly result in an increase in NB numbers because the loss of pros results in an inability of NB progeny to differentiate. This (despite the slight increase in nuclear pros) is not sufficient to infer that glial ferritin knockdown results in premature differentiation of NBs via nuclear pros.

I recognise these are challenging to prove irrefutably, however, the frequency of such expansive interpretations of data is of concern.

Author Response

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

We are grateful to the reviewers for recognizing the importance of our work on transcription-independent early recovery of proteasome activity. We also thank them for their thoughtful criticisms and suggested improvements, which we addressed in the revised version as described below.

The reviewers and editors asked for data to support the model that early recovery of proteasome activity is due to accelerated proteasome assembly. This model is backed by published data that proteasome assembly intermediates increase dramatically in cells treated with proteasome inhibitors (Fig. 6 in Ref. 46 of the revised manuscript). We expanded the discussion of this paper in a paragraph that describes our model. Another key experiment to confirm this model would be to determine what fraction of nascent polypeptides is degraded within minutes after synthesis, which is not trivial, and Ibtisam ran out of time to conduct these experiments because she had to graduate in spring before the expiration of her visa. This type of experiment usually uses metabolic labeling by a heavy or radioactive amino acid that always includes a prior depletion of a non-labeled amino acid. However, the fundamental flaw of this approach, which is not recognized by the scientific community, is that depletion of an amino acid stresses cells and reduces the rate of protein synthesis, especially if this amino acid is methionine. Thus, this model is not easy to test, and should be considered a speculation. We therefore moved the description of this model, together with Fig. 4, into a separate "Ideas and Speculations" section and removed this model's description from the abstract.

Reviewer 1 raised the possibility that a background band detected on the western blot of DDI2 KO cells could be a highly homologous protease DDI1. This is highly unlikely because, according to Protein Atlas, DDI1 is selectively expressed in the testis and is not expressed in the cell lines we used. Reviewer 1 also suggested that we should base our conclusion on Nrf1 KD, which we de-facto did because we confirmed that DDI2 KD blocks Nrf1 activation (Fig. 1d).

In response to Reviewer 1 critiques regarding the presentation of proteasome subunits stability data in Fig. 4 (Ref. 45 of the revised manusript), we removed PSMB8 and replaced chaperons with the subunits of the 26S base. We changed color palettes, symbols, and axis scales to improve clarity.

We acknowledged in the discussion that our work did not exclude DDI2 role in the recovery of proteasome after repeated pulse treatments, as suggested by Reviewer 1.

We agree with Reviewer 2 that using “proteasome levels” is inaccurate when describing our activity measurement data. However, in the manuscript, we use "levels" only when discussing data in the literature. We believe measuring activity and not the total levels is more important because not all proteasomes are active, e.g., latent 20S proteasome core particles.

Reviewer 3 expressed concern that our conclusions were based on data in HAP1 cells, which are haploid, and appear not very sensitive to proteasome inhibitors. This is why we used DDI2 KD in MDA-MB-231 and SUM149 cells, which are highly sensitive to proteasome inhibitors (Weyburne et al., Ref. 11). In our experience, full extent of proteasome inhibitor cytotoxicity is not revealed until 48hr after treatments, and viability determined at 12hr and 24hr as on Fig. 1c should not be used to determine sensitivity (it was used for activity assay normalization). We added a new supplementary figure showing that HAP1 cells are as sensitive to proteasome inhibitors as MDA-MD-231 cells when cell viability is assayed 48hr after treatment (new Fig. S2). Another panel on this new figure demonstrates that the baseline proteasome activity is very similar in HAP1, MD-MB-231 and SUM149 cells. We also added data demonstrating that inactivation of DDI2 by mutation does not change the recovery of proteasome activity in HCT-116 cells (new Fig. 1g). Recovery in MDA-MB-231, SUM149, and HCT-116 cells was measured at 18hr, which is still within the 12 – 24hr window when other investigators observed partially DDI2-dependent recovery.

We have conducted an experiment in which we followed activity recovery for up to 72hr. We found that activity plateaued at 24hr and opted against the repeat because there were no changes. We feel that the manuscript should not include one biological replicate data. The fact that the recovery is incomplete and that cells seem to survive with lower levels of proteasome activity is interesting; however, investigating the molecular basis for this phenomenon is beyond the scope of the current project.

We were not disputing the conclusions of previous studies that DDI2/Nrf1 is responsible for enhanced expression of proteasomal mRNA in cells continuously treated with proteasome inhibitors. In fact, we confirmed that pulse-treatment causes similar increase (Fig. 2b). As for papers that measured activity recovery after pulse treatment, we objectively discuss our results in the context of these papers. In response to Reviewers' recommendations and minor points:

• We reviewed the revised version carefully to eliminate spelling and grammatical errors and typos.

• We no longer refer to DDI2 as a novel protease, as suggested by Reviewer 1.

• We agree with Reviewer 2 that our CHX results do not necessarily mean that recovery involves translation of proteasomal mRNAs, and we now conclude that proteasome recovery requires protein synthesis.

• We revised Fig. 1c, 3a and 4a to improve clarity.

• We have stated in the caption that data in Fig. 4a comes from Table S4 in Ref. 45.

• We accepted an excellent suggestion of Reviewer 3 to change "recovery" to "early recovery" in the title.

• Regarding Reviewer 3 request to assay activity recovery at additional time points before 12h, this was done in the cycloheximide experiment in Fig. 3A.

• Even if we assume that the differences in the observed recovery activity in MDA-MB-231 cells (Fig. 1f) are statistically significant, which may implicate DDI2 involvement in the activity recovery, the percentage is still small, suggesting that most activity recovery is DDI2-independent.

• We toned down the statement "the present findings suggest that DDI2 desensitizes cells to PI by a different mechanism," replacing "suggest" with "raise a possibility".

• We indicated that only Bortezomib is approved for mantle cell lymphoma.

• We changed the description of clinical dosing as suggested by Reviewer 3. We added a reference on PK of subcutaneous bortezomib (Ref. 9), even though the review we quoted (Ref. 7) discussed subcutaneous dosing.

Reviewer #2 (Public Review):

Summary:

In this work Ibtisam and Kisselev explore the role of DDI2 in the proteasome function recovery after a clinically relevant pulse dosing using different proteasome inhibitors and their corresponding PK properties. The authors report that despite lack of NRF1 activation by DDI2 there was no difference in recovery from pulsed proteasome inhibition observed in DDI2 KO cells as compared to WT controls suggesting DDI2 is not required for recovery in this system. They further show that transcription of the proteasome subunits is initiated only after partial recovery of proteasome activity is already observed suggesting that non-transcriptional mechanisms might be also involved. The authors further show that translation inhibition blocked the recovery from proteasome inhibitors.

Strengths:

Overall, it is very important and informative to use a pulse treatment type approach (mimicking the PK properties of the drugs) to explore the biology of PIs as used in this study. The authors also provide convincing data that DDI2 is not required for proteasome activity recovery post-PI pulse treatment in the systems they explored.

Weaknesses:

The authors show that the recovery of one specific catalytic activity of the proteasome post-PI treatment is transcription independent. However, in this work they do not explore the other catalytic activities of the proteasome, the protein levels of the individual subunits and most importantly the level of the different assembled proteasome complexes and how they change over time. Without this data the proposed mechanism is still speculative, in particular the conclusion on the role of translation, and ignores other findings in the field that suggest that alternative mechanisms (such as proteasome complex assembly regulation for instance) might be just as plausible.

Reviewer #1 (Public Review):

Summary:

There has been substantial prior work trying to understand the transcriptional control of proteasome expression as an adaptive response to proteasome inhibition. This field has been mired by fierce debates over the role of the protease Ddi2 in activating the transcription factor Nrf1/NFE2L1. As the authors of this manuscript point out, most of the previous research centers on the continuous treatment of cells with proteasome inhibitors rather than a brief pulse of inhibition that better models the situation when these drugs are used clinically. The authors find that the initial recovery of proteasome activity is independent of Ddi2 and involves a mechanism distinct from transcription. The authors intriguingly point to a model in which the assembly of proteasomes is regulated. If true, this would be a significant finding, but for now, this model remains more speculative.

Strengths:

The pulsed treatment of proteasome inhibitors is a strength of this lab that few others use. It better mimics the clinical use of these inhibitors and allows for a more detailed analysis of the initial response to inhibition. The authors have used multiple different clones of Ddi2 knockouts and siRNA against Ddi2 to rule out the necessity of Ddi2 in the early production of proteasomes when cells are inhibited with proteasomes. establishing a thorough knockout approach while also avoiding compensatory mutations. These experiments are well controlled, showing both the levels of Ddi2 upon knockout or knockdown and demonstrating that the cleavage of Nrf1, one of two known targets of Ddi2, is impaired. However, it should be noted that faint bands for Ddi2 mysteriously remain even in the knockout.

This article sensitively monitors the recovery of proteasome function with the β5 activity assay and for the production of new proteasome transcripts by qPCR. This precision and a detailed analysis of the timing are strengths that pointed to a more rapid recovery than transcription alone.

Weaknesses:

This paper's major weakness is the difficulty in establishing the authors' model that assembly is regulating this process. They do a convincing job demonstrating that activity recovers before transcription. The evidence that translation is unaffected depends entirely on the polysome RNA profiling from two replicates. Clearer and orthogonal data would help establish this finding. The stability of subunits is interesting and important in its own right.

In short, the authors establish that Ddi2 is unnecessary for the initial, non-transcriptional recovery of proteasome activity after a pulse of proteasome inhibition.

It is not clear what clinical impact this work will have. Although it models the pulse of proteasome inhibition more perfectly, it only looks at a single pulse rather than multiple treatments. Thus, ruling out Ddi2's importance for clinical benefit may be premature. More significantly, this work suggests that assembling proteasomes might be a regulated process worth substantial follow-up that will be interesting to follow.

eLife assessment

The study presents important findings regarding a transcription-independent component of the early recovery of proteasome activity from a short pulse of proteasome inhibitor treatment, which has not been appreciated before and which is independent of the DDI2-NRF2 axis. While the evidence is in principle solid, with recapitulation in several cell line models, the proposed alternative underlying mechanism, namely regulation at the level of proteasome assembly, lacks experimental support, and at this point remain speculative.

Author Response

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

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

The authors have addressed most of the points that were made. However, despite some things that may well be beyond the scope, I would like to insist on a few small points:

Point 1: If the authors have conducted a gross analysis of cardiac morphology by histology already, they should include this data in the manuscript and comment with 1-2 sentences that "cardiac healing"..."is unlikely influenced by developmental defects".

We agree with the reviewer that this analysis is important. Therefore, we are currently conducting an in-depth analysis of the cardiac phenotype of different mouse strains lacking distinct subpopulations of cardiac macrophages in development and non-stimulated (baseline) conditions, including functional, metabolic and even electrophysiological aspects. These analysis will also include FIRE mice. While a gross analysis in this mouse strain did not show pathologic aspects, we look forward to the very detailed tissue characterization before publishing any data from a first basic analysis.

Point 7: There is still no legend in Figure 6: what is read? What is blue?

We added the respective legend in the figure.

Point 8: Please add the information on the background of mice used for the different FIRE mice into the methods part of the paper

We added the information in the Methods Part (lines 344-347).

Reviewer #2 (Recommendations For The Authors):

The authors have responded to all questions. I have no further comments and congratulate the authors on their work.

We thank the reviewer for their important questions and the constructive feedback.

eLife assessment

Using state-of-the-art fate-mapping models and genetic and pharmacological targeting approaches, this study provides important findings on the distinct functions exerted by resident and recruited macrophages during cardiac healing after myocardial ischemia. Evidence supporting the conclusions are solid with the use of the FIRE mouse model in combination with fate-mapping to target fetal-derived macrophages. This study will be of interest for the macrophage biologists working in the heart but also in others tissues in the context of inflammation.

Reviewer #1 (Public Review):

Weinberger et al. use different fate-mapping models, the FIRE model and PLX-diet to follow and target different macrophage populations and combine them with single-cell data to understand their contribution to heart regeneration after I/R injury. This question has already been addressed by other groups in the field using different models. However, the major strength of this manuscript is the usage of the FIRE mouse model that, for the first time, allows specific targeting of only fetal-derived macrophages.

The data show that the absence of resident macrophages is not influencing infarct size but instead is altering the immune cell crosstalk in response to injury, which is in line with the current idea in the field that macrophages of different origins have distinct functions in tissues, especially after an injury.

To fully support the claims of the study, specific targeting of monocyte-derived macrophages or the inhibition of their influx at different stages after injury would be of high interest.

In summary, the study is well done and important for the field of cardiac injury. But it also provides a novel model (FIRE mice + RANK-Cre fate-mapping) for other tissues to study the function of fetal-derived macrophages while monocyte-derived macrophages remain intact.

Reviewer #2 (Public Review):

In this study Weinberger et al. investigated cardiac macrophage subsets after ischemia/reperfusion (I/R) injury in mice. The authors studied a ∆FIRE mouse model (deletion of a regulatory element in the Csf1r locus), in which only tissue resident macrophages might be ablated. The authors showed a reduction of resident macrophages in ∆FIRE mice and characterized its macrophages populations via scRNAseq at baseline conditions and after I/R injury. 2 days after I/R protocol ∆FIRE mice showed an enhanced pro inflammatory phenotype in the RNAseq data and differential effects on echocardiographic function 6 and 30 days after I/R injury. Via flow cytometry and histology the authors confirmed existing evidence of increased bone marrow-derived macrophage infiltration to the heart, specifically to the ischemic myocardium. Macrophage population in ∆FIRE mice after I/R injury were only changed in the remote zone. Further RNAseq data on resident or recruited macrophages showed transcriptional differences between both cell types in terms of homeostasis-related genes and inflammation. Depleting all macrophage using a Csf1r inhibitor resulted in a reduced cardiac function and increased fibrosis.

Strengths:

(1) The authors utilized robust methodology encompassing state of the art immunological methods, different genetic mouse models and transcriptomics.<br /> (2) The topic of this work is important given the emerging role of tissue resident macrophages in cardiac homeostasis and disease.

Comments on revised version:

The authors have responded to all questions. I have no further comments and congratulate the authors on their work.

Reviewer #2 (Public Review):

Harnessing macrophages to attack cancer is an immunotherapy strategy that has been steadily gaining interest. Whether macrophages alone can be powerful enough to permanently eliminate a tumor is a high-priority question. In addition, the factors making different tumors more vulnerable to macrophage attack have not been completely defined. In this paper, the authors find that MSP1 inhibition, most notable for causing chromosomal instability (CIN), in cancer cells improves the effect of macrophage targeted immunotherapies. They demonstrate that MSP1 inhibited tumors secrete factors that polarize macrophages to a more tumoricidal fate through several methods. The most compelling experiment is transferring conditioned media from MSP1 inhibited and control cancer cells, then using RNAseq to demonstrate that the MSP1-inhibited conditioned media causes a shift towards a more tumoricidal macrophage phenotype. In mice with MSP1 inhibited (CIN) B16 melanoma tumors, a combination of CD47 knockdown and anti-Tyrp1 IgG is sufficient for long term survival in nearly all mice. This combination is a striking improvement from conditions without CIN.

Like any interesting paper, this study leaves several unanswered questions. First, how do CIN tumors repolarize macrophages? The authors demonstrate that conditioned media is sufficient for this repolarization, implicating secreted factors, but the specific mechanism is unclear. The main caveat of the study is that chromosomal instability is driven by MSP1 inhibition in all the experiments, leaving open the possibility that some effects are due to MSP1 inhibition specifically rather than CIN more generally. To specifically connect CIN and macrophage repolarization, future studies will need to examine tumors with CIN unrelated to MSP1 inhibition to determine if these are also able to repolarize macrophages.

Overall, this is a thought-provoking study that will be of broad interest to many different fields including cancer biology, immunology and cell biology.

Author Response

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

Reviewer #1 (Public Review):

This is an interesting manuscript that extends prior work from this group identifying that a chemovar of Cannabis induces apoptosis of T-ALL cells by preventing NOTCH1 cleavage. Here the authors isolate specific components of the chemovar responsible for this effect to CBD and CBDV. They identify the mechanism of action of these agents as occurring via the integrated stress response. Overall the work is well performed but there are two lingering questions that would be helpful to address as follows:

• Exactly how CBD and CBDV result in the upregulation of the TRPV1/integrated stress response is unclear. What is the most proximal target of these agents that results in these changes?

The interaction of CBD and CBDV with TRPV1 has been thoroughly investigated by previous studies in the field. A few prominent examples are:

(1) De Petrocellis, Luciano, Alessia Ligresti, Aniello Schiano Moriello, Marco Allarà, Tiziana Bisogno, Stefania Petrosino, Colin G. Stott, and Vincenzo Di Marzo. "Effects of cannabinoids and cannabinoid‐enriched Cannabis extracts on TRP channels and endocannabinoid metabolic enzymes." British journal of pharmacology 163, no. 7 (2011): 1479-1494.

(2) Muller, Chanté, Paula Morales, and Patricia H. Reggio. "Cannabinoid ligands targeting TRP channels." Frontiers in molecular neuroscience 11 (2019): 487.

(3) Iannotti, Fabio Arturo, Charlotte L. Hill, Antonio Leo, Ahlam Alhusaini, Camille Soubrane, Enrico Mazzarella, Emilio Russo, Benjamin J. Whalley, Vincenzo Di Marzo, and Gary J. Stephens. "Nonpsychotropic plant cannabinoids, cannabidivarin (CBDV) and cannabidiol (CBD), activate and desensitize transient receptor potential vanilloid 1 (TRPV1) channels in vitro: potential for the treatment of neuronal hyperexcitability." ACS chemical neuroscience 5, no. 11 (2014): 1131-1141.

(4) Costa, Barbara, Gabriella Giagnoni, Chiara Franke, Anna Elisa Trovato, and Mariapia Colleoni. "Vanilloid TRPV1 receptor mediates the antihyperalgesic effect of the nonpsychoactive cannabinoid, cannabidiol, in a rat model of acute inflammation." British journal of pharmacology 143, no. 2 (2004): 247-250.

(5) de Almeida, Douglas L., and Lakshmi A. Devi. "Diversity of molecular targets and signaling pathways for CBD." Pharmacology research & perspectives 8, no. 6 (2020): e00682.

(6) Anand, Uma, Ben Jones, Yuri Korchev, Stephen R. Bloom, Barbara Pacchetti, Praveen Anand, and Mikael Hans Sodergren. "CBD effects on TRPV1 signaling pathways in cultured DRG neurons." Journal of Pain Research (2020): 22692278.

Similarly, other works have demonstrated the link between TRPV1 and the integrated stress response pathway (see below). These studies suggested increased reactive oxygen species (ROS) production, Cyclooxygenase-2 (COX-2) enzyme, as well as other stressors, lead to modulation of intracellular calcium levels by TRPV1.

(1) Ho, Karen W., Nicholas J. Ward, and David J. Calkins. "TRPV1: a stress response protein in the central nervous system." American journal of neurodegenerative disease 1, no. 1 (2012): 1.

(2) de la Harpe, Amy, Natasha Beukes, and Carminita L. Frost. "CBD activation of TRPV1 induces oxidative signaling and subsequent ER stress in breast cancer cell lines." Biotechnology and Applied Biochemistry 69, no. 2 (2022): 420-430.

(3) Soliman, Eman, and Rukiyah Van Dross. "Anandamide‐induced endoplasmic reticulum stress and apoptosis are mediated by oxidative stress in nonmelanoma skin cancer: Receptor‐independent endocannabinoid signaling." Molecular Carcinogenesis 55, no. 11 (2016): 1807-1821.

• Related to the above, all experiments to confirm the mechanism of action of CBD/CBDV rely on chemical agents, whose precise targets are not fully clear in some cases. Thus, some use of genetic means (such as by knockout of TRPV1, ATF4) to confirm the dependency of these pathways on drug response and NOTCH cleavage would be very helpful.

Knockdown experiments and inhibition with inhibitors are two different approaches to studying the function of a specific gene or protein. Each method has its advantages and limitations. We initially attempted to knock-down CHAC1, but only managed to silence ~50% (Incomplete knockdown). Following treatment of MOLT4 cells with the whole extract, we observed only a partial downregulation in the mRNA expression of the Notch intracellular domain (NICD) (left panel), which could account for the incomplete rescue from the extract-induced death (right panel). We therefore turned to confirm the signaling pathway by inhibition of different targets with chemical agents.

Author response image 1.

Partial knockdown of CHAC1 hinders extract-induced cell death. (A) MOLT-4 cells were treated with either an empty vector or shRNA for Chac1, 369 and 739 represent two different areas of Chac1, for 48 hrs. Then, the gene expression of CHAC1 was assessed via qRT-PCR (N=3). (B) MOLT-4 cells were treated as in A, then added vehicle control or whole Extract (3 µg/mL) for additional 24 hrs, and the viability of the cells was assessed with XTT.

Reviewer #2 (Public Review):

Summary:

The Meiri group previously showed that Notch1-activated human T-ALL cell lines are sensitive to a cannabis extract in vitro and in vivo (Ref. 32). In that article, the authors showed that Extract #12 reduced NICD expression and viability, which was partially rescued by restoring NICD expression. Here, the authors have identified three compounds of Extract #12 (CBD, 331-18A, and CBDV) that are responsible for the majority of anti-leukemic activity and NICD reduction. Using a pharmacological approach, the authors determined that Extract #12 exerted its anti-leukemic and NICD-reducing effects through the CB2 and TRPV1 receptors. To determine the mechanism, the authors performed RNA-seq and observed that Extract #12 induces ER calcium depletion and stress-associated signals -- ATF4, CHOP, and CHAC1. Since CHAC1 was previously shown to be a Notch inhibitor in neural cells, the authors assume that the cannabis compounds repress Notch S1 cleavage through CHAC1 induction. The induction of stress-associated signals, Notch repression, and anti-leukemic effects were reversed by the integrated stress response (ISR) inhibitor ISRIB. Interestingly, combining the 3 cannabinoids gave synergistic anti-leukemic effects in vitro and had growthinhibitory effects in vivo.

Strengths:

(1) The authors show novel mechanistic insights that cannabinoids induce ER calcium release and that the subsequent integrated stress response represses activated NOTCH1 expression and kills T-ALL cells.

(2) This report adds to the evidence that phytocannabinoids can show a so-called "entourage effect" in which minor cannabinoids enhance the effect of the major cannabinoid CBD.

(3) This report dissects the main cannabinoids in the previously described Extract #12 that contribute to T-ALL killing.

(4) The manuscript is clear and generally well-written.

(5) The data are generally high quality and with adequate statistical analyses.

(6) The data generally support the authors' conclusions. The exception is the experiments related to Notch.

(7) The authors' discovery of the role of the integrated stress response might explain previous observations that SERCA inhibitors block Notch S1 cleavage and activation in T-ALL (Roti Cancer Cell 2013). The previous explanation by Roti et al was that calcium depletion causes Notch misfolding, which leads to impaired trafficking and cleavage. Perhaps this explanation is not entirely sufficient.

Weaknesses:

(1) Given the authors' previous Cancer Communications paper on the anti-leukemic effects and mechanism of Extract #12, the significance of the current manuscript is reduced.

Our original manuscript consisted extensive multidisciplinary research, and we were asked to divide the research work into a paper that focuses on the cannabis plant and another paper that addresses finding the specific molecules and their underlying mechanism.

We understand that our publication of the initial observations with the whole extract dampened the overall novelty presented here, but the previous publication lacked the detailed and strong mechanistic work presented here that explains how the cannabis extract exerted its antitumoral effects.

In addition, the finding of the need for 3 phytocannabinoids and the synergy analysis supplies essential support to the ‘entourage effect’. This is a phenomenon in which the presence of minor proportions of cannabinoids and other plant components significantly modulate the effects of the main active components of cannabis and thereby produce more potent or more selective effects than the use of one major cannabinoid alone. It was well-demonstrated for endocannabinoids but was only demonstrated in very few studies for phytocannabinoids.

(2) It would be important to connect the authors' findings and a wealth of literature on the role of ER calcium/stress on Notch cleavage, folding, trafficking, and activation.

We mentioned three previous papers (ref. 34-36) that guided us in our investigation. Following this reviewer’s comment, we added to the discussion a few lines connecting our findings to previous works on ER stress and Notch activation with the appropriate references.

(3) There is an overreliance on the data on a single cell line -- MOLT4. MOLT4 is a good initial choice as it is Notch-mutated, Notch-dependent, and representative of the most common T-ALL subtype -- TAL1. However, there is no confirmatory data in other TAL1positive T-ALLs or interrogation of other T-ALL subtypes.

As mentioned by the reviewer, this study followed a previous publication in which 7 different cell lines were assessed (MOLT‐4, CCRF‐CEM, Jurkat, Loucy, HPB-ALL, DND-41and T-ALL1). MOLT-4 cells were used to investigate the mechanism, both MOLT-4 cells and CCRF-CEM cells were utilized to investigate the effect of the cannabinoid combination or the whole extract in-vivo.

(4) Fig. 6H. The effects of the cannabinoid combination might be statistically significant but seem biologically weak.

Survival rates are presented in Fig. 6H for the combination of the cannabinoids and in Supplementary Fig. S6C for the whole extract. While this mouse model provides valuable insights, the biological significance and the translation of findings to human patients require cautious interpretation.

(5) Fig. 3. Based on these data, the authors conclude that the cannabinoid combination induces CHAC1, which represses Notch S1 cleavage in T-ALL cells. The concern is that Notch signaling is highly context-dependent. CHAC1 might inhibit Notch in neural cells (Refs. 34-35), but it might not do this in a different context like T-ALL. It would be important to show evidence that CHAC1 represses S1 cleavage in the T-ALL context. More importantly, Fig. 3H clearly shows the cannabinoid combination inducing ATF4 and CHOP protein expression, but the effects on CHAC1 protein do not seem to be satisfactory as a mechanism for Notch inhibition. Perhaps something else is blocking Notch expression?

We understand the reviewer’s concern. Previous works had shown the upregulation of CHAC1 also in the context of Notch signaling in leukemia, particularly recently also for T-ALL:

(1) Meng, X., Matlawska-Wasowska, K., Girodon, F., Mazel, T., Willman, C.L., Atlas, S., Chen, I.M., Harvey, R.C., Hunger, S.P., Ness, S.A. and Winter, S.S., 2011. GSI-I (Z-LLNle-CHO) inhibits γ-secretase and the proteosome to trigger cell death in precursor-B acute lymphoblastic leukemia. Leukemia, 25(7), pp.11351146.

(2) Chang, Yoon Soo, Joell J. Gills, Shigeru Kawabata, Masahiro Onozawa, Giusy Della Gatta, Adolfo A. Ferrando, Peter D. Aplan, and Phillip A. Dennis. "Inhibition of the NOTCH and mTOR pathways by nelfinavir as a novel treatment for T cell acute lymphoblastic leukemia." International Journal of Oncology 63, no. 5 (2023): 1-12.

As for the second part of the reviewer’s comment, we tested both the mRNA transcript and protein expression of CHAC1. The increase is clearly shown at 60 min for the mRNA Fig. 3D and Fig. 4F and for the protein also in Supplementary Fig. S4G-I.

To show direct involvement of CHAC1 we utilized the means of knockdown. Though it was not completely effective and accounted for about ~50% reduction, it clearly shows the involvement of CHAC1 in the mechanism leading to the reduction in viability of these cancer cells.

Author response image 2.

Partial knockdown of CHAC1 hinders extract-induced cell death. (A) MOLT-4 cells were treated with either an empty vector or shRNA for Chac1, 369 and 739 represent two different areas of Chac1, for 48 hrs. Then, the gene expression of CHAC1 was assessed via qRT-PCR (N=3). (B) MOLT-4 cells were treated as in A, then added vehicle control or whole Extract (3 µg/mL) for additional 24 hrs, and the viability of the cells was assessed with XTT.

(6) Fig. 4B-C/S5D-E. These Western blots of NICD expression are consistent with the cannabinoid combination blocking Furin-mediated NOTCH1 cleavage, which is reversed by ISR inhibition. However, there are many mechanisms that regulate NICD expression. To support their conclusion that the effects are specifically Furin-medated, the authors should probe full-length (uncleaved) NOTCH1 in their Western blots.

We have probed for the full-length Notch1 in our previously published paper (Cancer Communications, Supplementary Fig. S1G-I). As we have shown here the three cannabinoids together mimic the effect of the whole extract, we did not repeat the experiments with full-length Notch1.

(7) Fig. S4A-B. While these pharmacologic data are suggestive that Extract #12 reduces NICD expression through the CB2 receptor and TRPV1 channel, the doses used are very high (50uM). To exclude off-target effects, these data should be paired with genetic data to support the authors' conclusions.

We performed a dose-response experiment before choosing the doses used for the inhibitors of CB2 and TRPV1 (see below). The dose of 50 µM was selected as it did not affect the viability of the cells.

Author response image 3.

Dose-response of CB2 and TRPV1 inhibitors in MOLT-4 cells. MOLT-4 cells were treated with increasing concentrations (µM) of (A) CB2 inhibitor AM630 or (B) TRPV1 inhibitor AMG9810; and 24 hrs later the viability of the cells was assessed with XTT.

Reviewer #2 (Recommendations For The Authors):

(1) In Fig. 6H, it is unclear why the authors are using CCRF-CEM cells, which are known to be resistant to Notch inhibitors, rather than popular cell lines that are Notch-dependent (e.g. CUTLL1, DND-41, HPB-ALL). Since cannabinoids seem to kill at least in part through Notch inhibition, the effects would be predicted to be greater in Notch-dependent T-ALL cell lines than Notch-independent cell lines like CCRF-CEM. To show stronger in vivo preclinical efficacy, another suggestion is to combine cannabinoids with tolerable dosing of gammasecretase inhibitors as published by the Michelle Kelliher group.

We have shown in our previous publication that both MOLT-4 and CCRF-CEM cells are dependent on Notch for their propagation, while other cell lines of T-ALL such as Loucy and Jurkat do not. Therefore, we treat CCRF-CEM as Notch-dependent. We discuss the possibility of using the cannabinoid combination with other treatments, specifically chemotherapy, to enhance effectiveness.

(2) To increase significance, this reviewer suggests strengthening the mechanism. However, this reviewer understands the challenge of identifying the correct mechanism. Thus, an alternative would be to increase clinical relevance. Some specific suggestions are described below.

(a) With regard to increasing mechanistic insights, the authors should be aware of some papers that might be helpful. Roti et al (Cancer Cell 2013) showed that SERCA inhibitors like thapsigardin reduce ER calcium levels and block Notch signaling by inhibiting NOTCH1 trafficking and inhibiting Furin-mediated (S1) cleavage of Notch1. Multiple EGF repeats and all three Lin12/Notch repeats in the extracellular domains of Notch receptors require calcium for proper folding (Aster Biochemistry 1999; Gordon Nat. Struct. Mol. Biol. 2007; Hambleton Structure 2004; Rand Protein Sci 1997). Thus, Roti et al concluded that ER calcium depletion blocks NOTCH1 S1 cleavage. This effect seems to be conserved in Drosophila as Periz and Fortiin (EMBO J, 1999) showed impaired Notch cleavage in Ca2+/ATPasemutated Drosophila cells. Besser et al should consider these papers when exploring the mechanism by which the ER calcium release by the cannabinoid combination blocks activated NOTCH1 expression. Similarities and differences should be discussed.

As mentioned above and stated also by the reviewer, many papers have shown the cleavage and/or activation of Notch following ER stress.

(b) With regard to increasing clinical relevance, the authors should consider testing the effects of the cannabinoid combination on primary samples, PDX models, and/or genetically engineered mouse models. Pan-Notch inhibitors like gamma-secretase inhibitors (GSIs) have been disappointing in clinical trials because of excessive on-target toxicity, in particular in the intestine. The authors should consider exploring whether the cannabinoids might be superior to GSIs with regard to intestinal toxicity and why that might be (e.g. receptor expression).

We thank the reviewer and agree that clinical relevance is of outmost importance. As obtaining primary tumor cells from patients is challenging, we assessed the whole cannabis extract in a PDX model. This extract is already being used by patients. We added this result as Supplementary fig. S7, and address it in the main text of the Results and in the Materials and Methods section.

(3) Since the authors have performed gene expression profiling, another test to confirm that Extract #12 acts through the Notch pathway is to perform enrichment analysis for known Notch target genes in T-ALL (e.g. Wang PNAS 2013).

We performed the analysis and this is how we pinpointed the involvement of ATF4, CHOP and CHAC1 of the integrated stress response pathway.

Minor concern:

Supplemental Table S4. According to the text (page 10, line 160) and table title, these data are RNA-seq. However, according to the GSE154287 annotation, these data are Affymetrix arrays There are no gene names in the GSE table. Are the IDs probesets rather than genes?

Indeed, the gene analysis data are Affymetrix arrays and the title was corrected.

eLife assessment

This important study follows up on previous work defining the anti-leukemic effects of a previously characterized cannabis extract on Notch-activated T cells and identifies several pathways that mediate its anti-cancer activity including the ER calcium and integrated stress response. The evidence is solid, but several concerns remain including the over reliance on a single cell line for the majority of the studies and lack of integration of the observations with existing literature

Reviewer #2 (Public Review):

Summary:

The Meiri group previously showed that Notch1-activated human T-ALL cell lines are sensitive to a cannabis extract in vitro and in vivo (Ref. 32). In that article, the authors showed that Extract #12 reduced NICD expression and viability, which was partially rescued by restoring NICD expression. Here, the authors have identified three compounds of Extract #12 (CBD, 331-18A, and CBDV) that are responsible for the majority of anti-leukemic activity and NICD reduction. Using a pharmacological approach, the authors determined that Extract #12 exerted its anti-leukemic and NICD-reducing affects through the CB2 and TRPV1 receptors. To determine mechanism, the authors performed RNA-seq and observed that Extract #12 induces ER calcium depletion and stress-associated signals -- ATF4, CHOP, and CHAC1. Since CHAC1 was previously shown to be a Notch inhibitor in neural cells, the authors assume that the cannabis compounds repress Notch S1 cleavage through CHAC1 induction. The induction of stress-associated signals, Notch repression, and anti-leukemic effects were reversed by the integrated stress response (ISR) inhibitor ISRIB. Interestingly, combining the 3 cannabinoids gave synergistic anti-leukemic effects in vitro and had growth inhibitory effects in vivo.

Strengths:

(1) The authors show novel mechanistic insights that cannabinoids induce ER calcium release and that the subsequent integrated stress response represses activated NOTCH1 expression and kills T-ALL cells.

(2) This report adds to the evidence that phytocannabinoids can show a so-called "entourage effect" in which minor cannabinoids enhance the effect of the major cannabinoid CBD.

(3) This report dissects out the main cannabinoids in the previously described Extract #12 that contribute to T-ALL killing.

(4) The manuscript is clear and generally well-written.

(5) The data are mostly high quality and with adequate statistical analyses.

(6) The data generally support the authors' conclusions. The main exception is the experiments related to Notch.

(7) The authors' discovery of the role of the integrated stress response might explain previous observations that SERCA inhibitors block Notch S1 cleavage and activation in T-ALL (Roti Cancer Cell 2013). The previous explanation by Roti et al was that calcium depletion causes Notch misfolding, which leads to impaired trafficking and cleavage. Perhaps this explanation is not entirely sufficient?

Weaknesses:

(1) Given the authors' previous Cancer Communications paper on the anti-leukemic effects and mechanism of Extract #12, the significance of the original manuscript was reduced. To increase significance, the authors provided a new Fig. S7 in the revision showing that Extract #12 inhibits PDX growth in vivo. This experiment is nicely supportive of the anti-leukemic effects of Extract #12, raising the significance of their previous Cancer Communication paper by using in vivo patient-derived cells. However, this reviewer had suggested testing the combination of 333-18A+CBVD+CBD since the combination is the focus of the current manuscript. For unclear reasons, the combination was not tested.

(2) It would be important to connect the authors' findings and a wealth of literature on the role of ER calcium/stress on Notch cleavage, folding, trafficking, and activation. The several references suggested by this reviewer were not included in the revised manuscript for unclear reasons. These references are important to show the current status of the field and help readers appreciate what this manuscript brings that is new to T-ALL. In particular, Roti et al (Cancer Cell 2013) showed that SERCA inhibitors like thapsigardin reduce ER calcium levels and block Notch signaling by inhibiting NOTCH1 trafficking and inhibiting Furin-mediated (S1) cleavage of Notch1 in T-ALL. Multiple EGF repeats and all three Lin12/Notch repeats in the extracellular domains of Notch receptors require calcium for proper folding (Aster Biochemistry 1999; Gordon Nat. Struct. Mol. Biol. 2007; Hambleton Structure 2004; Rand Protein Sci 1997). Thus, Roti et al concluded that ER calcium depletion blocks NOTCH1 S1 cleavage in T-ALL. This effect seems to be conserved in Drosophila as Periz and Fortiin (EMBO J, 1999) showed impaired Notch cleavage in Ca2+/ATPase-mutated Drosophila cells.

(3) There is an overreliance of the data on single cell line -- MOLT4. MOLT4 is a good initial choice as it is Notch-mutated, Notch-dependent, and representative of the most common T-ALL subtype -- TAL1. However, there is no confirmatory data in other TAL1-positive T-ALLs or interrogation of other T-ALL subtypes. While this reviewer appreciates that the authors showed that multiple T-ALL cell lines were killed in response to Extract #12 in a previous study, the current manuscript is a separate study that should stand on its own. T-ALLs can be killed by multiple mechanisms. It would be important to show a few pieces of key data illustrating that the mechanism of killing found in MOLT4 applies to other T-ALLs.

(4) Fig. 6H. The effects of the cannabinoid combination might be statistically significant but seems biologically weak.

(5) Fig. 3. Based on these data, the authors conclude that the cannabinoid combination induces CHAC1, which represses Notch S1 cleavage in T-ALL cells. The concern is that Notch signaling is highly context dependent. CHAC1 might inhibit Notch in neural cells (Refs. 34-35), but it might not do this in a different context like T-ALL. It would be important to show evidence that CHAC1 represses S1 cleavage in the T-ALL context. More importantly, Fig. 3H clearly shows the cannabinoid combination inducing ATF4 and CHOP protein expression, but the effects on CHAC1 protein do not seem to be satisfactory as a mechanism for Notch inhibition. Perhaps something else is blocking Notch expression?

In the rebuttal, the two references provided by the authors do not alleviate concern that CHAC1 might not be acting as a Notch cleavage inhibitor in the T-ALL context. The Meng et al paper studied B-ALL not T-ALL and did not evaluate CHAC1 as a possible Notch cleavage inhibitor. Likewise, the Chang et al paper did not evaluate CHAC1 as a possible Notch cleavage inhibitor. Therefore, whether CHAC1 is a Notch cleavage inhibitor in the T-ALL context remains an open question. While the authors are correct that Supplementary Fig. S4G-I show that Extract #12 clearly induces CHAC1 protein expression, Main Fig. 3H shows that the extract combination 333-18A+CBVD+CBD, which is the focus of this manuscript, has unclear effects. If the extract combination has no effect on CHAC1 but has the same effects on Notch1 expression as the full extract, then there is reduced support for the authors' conclusion that the full extract and the 333-18A+CBVD+CBD combination inhibit Notch through CHAC1 induction.

(6) The authors provide a new figure on page 5 of the rebuttal that was not requested. It is supposed to show that CHAC1 loss protects T-ALL cells from Extract #12-induced cell population decline. Unfortunately, this figure is not conclusive. The empty vector PLKO is not an appropriate negative control. The field uses non-targeting shRNA controls like pLKO-luciferase to control for induction of the RNA interference pathway. Further, the viability data in panel B is normalized such that the effect of shCHAC1 on viability is masked. Showing non-normalized data is important, because if shCHAC1 impairs viability compared to control shRNA, then CHAC1might have effects on non-Notch pathways, which would reinforce the above concern in Point #5 that CHAC1 might not act as a Notch inhibitor in the T-ALL context. Separately, if this experiment had tested whether CHAC1 knockdown increases Notch cleavage and Notch target gene expression like DTX1, HES1 and MYC, then such data would have helped address Point #5.

(7) Fig. 4B-C/S5D-E. These Western blots of NICD expression are consistent with the cannabinoid combination blocking Furin-mediated NOTCH1 cleavage, which is reversed by ISR inhibition. However, there are many mechanisms that regulate NICD expression. To support their conclusion that the effects are specifically Furin-medated, the authors should probe full length (uncleaved) NOTCH1 in their Western blots. While the authors showed that the full extract (#12) increased uncleaved NOTCH1 expression in their Cancer Communications paper, a major conclusion of the manuscript is that the cannabinoid combination 333-18A+CBVD+CBD reproduces the effect of the full extract (#12). To support this conclusion, the authors should probe key blots for full-length Notch to show that the cannabinoid combination increases uncleaved NOTCH1 just like Extract #12 did in the authors' Cancer Communications paper.

(8) Fig. S4A-B. While these pharmacologic data are suggestive that Extract #12 reduces NICD expression through the CB2 receptor and TRPV1 channel, the doses used are very high (50uM). To exclude off-target effects, these data should be paired with genetic data to support the authors' conclusions. In the rebuttal, the authors provide dose response cell viability curves of the CB2 and TRPV1 inhibitors. These curves do not exclude the possibility that 50uM has off-target effects. This reviewer notes that Reviewer #1 had similar concerns and that both reviewers requested genetic validation of the pharmacological data. These data were not provided in the revision.

(9) Since the authors have performed gene expression profiling, an orthogonal test to confirm that Extract #12 acts through the Notch pathway is to perform enrichment analysis using Notch target gene signatures in T-ALL (e.g. Wang PNAS 2013). In contrast to the authors' rebuttal, this reviewer does not see any enrichment analysis (e.g. GSEA plots) performed on the microarray data to show that Extract #12 inhibits the Notch pathway.

(10) The revised manuscript still retains references that microarray data are "RNA-seq" data, which is inaccurate (see page 10, line 160; Figure 3 legend; page 12, line 169; page 27, line 428; page 36, line 741)

Author Response

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

eLife assessment

The delineation of MBOAT function is important with theoretical and practical implications in MAFLD, alcohol-induced hepatic steatosis, and lysosomal diseases. The strength of evidence is convincing using methodology in line with current state-of-the-art, with good support for the claims.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors provide mechanistic insights into how the loss of function of MBOAT7 promotes alcoholassociated liver disease. They showed that hepatocyte-specific genetic deletion of Mboat7 enhances ethanol-induced hepatic steatosis and increased ALT levels in a murine model of ethanol-induced liver disease. Through lipidomic profiling, they showed that mice with Mboat7 deletion demonstrated augmented ethanol-induced endosomal and lysosomal lipids, together with impaired transcription factor EB (TFEB)-mediated lysosomal biogenesis and accumulation of autophagosomes.

Strengths:

Alcohol-induced liver disease (ALD) and metabolic-associated steatotic liver disease (MASLD) are major global health problems, and polymorphism near the gene encoding MBOAT7 has been associated with these conditions. This paper is timely as it is important to gain insights on how loss of MBOAT function contributes to liver disease as this may eventually lead to therapeutic strategies. -The conclusions of the paper are mostly well supported by data.

We sincerely thank Reviewer #1 for constructive feedback on this work.

Weaknesses:

(1) In regards to circulating levels of MBOAT7 products, a comparison of heavy drinkers with ALD versus heavy drinkers without ALD would be more clinically relevant.

We agree this comparison would be an important comparison to make in future studies, but given the difficulties in accessing well-matched samples such as these we see this as beyond the scope of the current work.

(2) A few typos need to be addressed. For Figure 1 - figure supplement 1, should the second column heading be "Heavy drinkers" instead of "Healthy drinkers"? Also, in the same figure, it is unclear what the "healthy" subcategory under MELD means.

The typographical error was addressed in the main text and in all associated tables and figures.

(3) Some of the data in the tables need to be addressed/discussed. For instance, the white blood cell count (WBC) in Figure 1 - figure supplement 1 for "healthy controls" is 34, compared to 13.51 for drinkers. A WBC of 34 is not at all healthy and should be explained. The vast difference between BMI and also between racial distribution within the two cohorts should also be explained. Is it possible that some of these differences contributed to the different levels of circulating MBOAT7 products that were measured?

Sincere thanks for catching this error. In follow up, we found that some of our patient recruitment sites were using different units to report WBC counts (percent vs 1000/ml) and at this time we cannot retrospectively correct that difference. Therefore, we have incomplete WBC values for the cohort so elected to exclude that information to avoid confusing readers. A revised table is provided in revision reflecting these changes/ If we look at each site separately, values for WBC were in the normal range, so we do not think this is a major limitation of our studies. In regards to BMI and race: Race is not actually significant, but close. For BMI, there are 2 very low BMIs in the Heavy drinkers which bring that average down. We agree with Reviewer # 1 that race and / or BMI could impact MBOAT7, but larger cohorts are needed to detect such potential differences.

(4) The representation of the statistical difference between the bars in the results figures by using alphabets is a bit confusing. For instance, in figure 2C, does that mean all the bars labelled A are significantly different from B? The solid black bar seems to be very similar to the open red bar; please double check.

We apologize for this confusing presentation. Using the letter system, groups not sharing a common superscript differ statistically. Given this concern, we have gone back and reviewed all statistical comparisons and realized that there were several mistakes in the graph Figure 2C, Figure 3F and G, Figure 3-Supplementary Figure 1 F and Figure 3-Supplementary Figure 10H. The graphs themselves were not altered, but the denotation of statistical significance was updated with the correct letter superscripts.

Reviewer #2 (Public Review):

Summary:

The work by Varadharajan et. al. explored a previously known genetic variant and its pathophysiology in the development of alcohol-associated liver injury. It provides a plausible mechanism for how varying levels of MBOAT7 could impact the lipid metabolomics of the cell, leading to a deleterious phenotype in MBOAT7 knockout. The authors further characterized the impact of the lipidomic changes and raised lysosomal biogenesis and autophagic flux as mechanisms of how MBOAT7 deletion causes the progression of ALD.

Strengths:

Connecting the GWAS data on MBOAT7 variants with plausible pathophysiology greatly enhances the translational relevance of these findings. The global lipidomic profiling of ALD mice is also very informative and may lead to other discoveries related to lipid handling pathways.

We sincerely thank Reviewer #1 for constructive feedback on this work.

Weaknesses:

The rationale of why MBOAT7 metabolites are lower in heavy drinkers than in normal individuals is not well explained. MBOAT7 loss of function drives ALD, but unclear if MBOAT7 deletion also drives preference for alcohol or if alcohol inhibits MBOAT7 function. Presuming most individuals studied here were WT and expressed an appropriate level of MBOAT7?

Although we were unable to genotype for the rs641738 SNP in the human subjects studied here, the original study by Buch et al. published in Nature Genetics performed cis expression quantitative trait lock (cis-eQTL) analyses to demonstrate that the minor disease-associated allele was associated with reduced MBOAT7 expression in subjects with alcohol-related cirrhosis. It is important to note that we did not see any evidence that alcohol preference was altered in either myeloid- or hepatocyte-specific Mboat7-knockout mice, given ethanol intake was similar in all genotypes. Additional studies are needed to address the possibility that MBOAT7 loss of function may promote alcohol preference, but we agree that this should be further investigated.

Also, the discussion of mechanisms of MBOAT7-induced dysregulation of lysosomal biogenesis/autophagy, while very interesting, seems incomplete. It is not clear how MBOAT7 an enzyme involved in membrane phospholipid remodeling increases mTOR which leads to decreased TFEB target gene transcription.

Although we agree with Reviewer #2 that mechanistic understanding by which MBOAT7 loss of function impacts mTOR activity and TFEB-driven lysosomal biogenesis is still incomplete, we do feel that the results published here will inform downstream investigation linking phosphatidylinositol remodeling to mTOR and TFEB. The MBOAT7 gene encodes an acyltransferase enzyme that specifically esterifies arachidonyl-CoA to lysophosphatidylinositol (LPI) to generate the predominant molecular species of phosphatidylinositol (PI) in cell membranes (38:4). It is well established that PI-related lipids can regulate membrane dynamics and signal transduction pathways. For instance PI-phosphates (PIPs) are dynamically shaped by PI kinases and phosphatases to play crucial roles in the regulation of a wide variety of cellular processes via specific interactions of PIP-binding proteins. Among PI phosphates, PI 3phosphate (PI3P) regulates vesicular trafficking pathways, including endocytosis, endosome-toGolgi retrograde transport, autophagy and mTOR signaling. Although additional work is needed to understand the molecular details of how MBOAT7-driven LPI acylation impacts mTOR and TFEB, it is not particularly surprising that PI lipid remodeling could broadly impact cell signaling.

Furthermore, given the significant disturbances of global lipidomic profiling in MBOAT7 knockout, many pathways are potentially affected by this deletion. Further in vivo modeling that specifically addresses these pathways (TFEB targeting, mTOR inhibitor) would help strengthen the conclusions of this paper.

We agree that further in vivo studies are needed that are beyond the scope of the current work.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

(1) p values are rather hard to read. For example, Figure 2c, Hepatocyte-specific deletion of Mboat7 resulted in enhanced ethanol-induced increases in liver weight. However, doesn't look like there is a significant difference between the 2 EtOH groups in Figure 2C? Same comment for Figure 2e, not sure if pair-fed groups had a significant difference.

(2) Figure 2 Supp fig 1, what is the top band on the MBOAT7 WB?

We have addressed these statistical comparison comments as described above. Although we cannot be sure, it is likely that the top band on the MBOAT7 Western blot is a non-specific band that shows up with the antibody combination used given there is equal intensity in the Mboat7flox/flox and the MSKO mice (Mboat7flox/flox+LysM-Cre).

eLife assessment

Varadharajan et al. explore the mechanistic basis of MBOAT7 SNP association with steatotic liver disease and link its function in LPI acylation to altered lipidomics of endosomal/lysosomal system and impaired TFEB mediated lysosomal biogenesis. The findings are important with theoretical and practical implications in MAFLD, alcohol-induced hepatic steatosis, and lysosomal diseases. The strength of evidence is convincing using methodology in line with current state-of-the-art.

Reviewer #2 (Public Review):

Summary:

The work by Varadharajan et. al. explored a previously known genetic variant and its pathophysiology in the development of alcohol-associated liver injury. It provides a plausible mechanism for how varying levels of MBOAT7 could impact the lipid metabolomics of the cell, leading to a deleterious phenotype in MBOAT7 knockout. The authors further characterized the impact of the lipidomic changes and raised lysosomal biogenesis and autophagic flux as mechanisms of how MBOAT7 deletion causes the progression of ALD.

Strengths:

Connecting the GWAS data on MBOAT7 variants with plausible pathophysiology greatly enhances the translational relevance of these findings. The global lipidomic profiling of ALD mice is also very informative and may lead to other discoveries related to lipid handling pathways.

Weaknesses:

The rationale of why MBOAT7 metabolites are lower in heavy drinkers than in normal individuals is not well explained. MBOAT7 loss of function drives ALD, but unclear if MBOAT7 deletion also drives preference for alcohol or if alcohol inhibits MBOAT7 function. Presuming most individuals studied here were WT and expressed an appropriate level of MBOAT7?

Also, discussion of mechanisms of MBOAT7-induced dysregulation of lysosomal biogenesis/autophagy, while very interesting, seems incomplete. It is not clear how MBOAT7 an enzyme involved in membrane phospholipid remodeling increases mTOR which leads to decreased TFEB target gene transcription. Furthermore, given the significant disturbances of global lipidomic profiling in MBOAT7 knockout, many pathways are potentially affected by this deletion. Further in vivo modeling that specifically addresses these pathways (TFEB targeting, mTOR inhibitor) would help strengthen the conclusions of this paper.

eLife assessment

The authors build upon prior data implicating the secreted peptidoglycan hydrolase SagA produced by Enterococcus faecium in immunotherapy. Leveraging new strains with sagA deletion/complementation constructs, the investigators reveal that sagA is non-essential, with sagA deletion leading to a marked growth defect due to impaired cell division, and sagA being necessary for the immunogenic and anti-tumor effects of E. faecium. In aggregate, the study utilizes compelling methods to provide both fundamental new insights into E. faecium biology and host interactions and a proof-of-concept for identifying the bacterial effectors of immunotherapy response.

Reviewer #1 (Public Review):

Klupt, Fam, Zhang, Hang, and colleagues present a novel study examining the function of sagA in E. faecium, including impacts on growth, peptidoglycan cleavage, cell separation, antibiotic sensitivity, NOD2 activation, and modulation of cancer immunotherapy. This manuscript represents a substantial advance over their prior work, where they found that sagA-expressing strains (including naturally-expressing strains and versions of non-expressing strains forced to overexpress sagA) were superior in activating NOD2 and improving cancer immunotherapy. Prior to the current study, an examination of sagA mutant E. faecium was not possible and sagA was thought to be an essential gene.

The study is overall very carefully performed with appropriate controls and experimental checks, including confirmation of similar densities of ΔsagA throughout. Results are overall interpreted cautiously and appropriately.

I have only two comments that I think addressing would strengthen what is already an excellent manuscript.

In the experiments depicted in Figure 3, the authors should clarify the quantification of peptidoglycans from cellular material vs supernatants. It should also be clarified whether the sagA need to be expressed endogenously within E. faecium, and whether ambient endopeptidases (perhaps expressed by other nearby bacteria or recombinant enzymes added) can enzymatically work on ΔsagA cell wall products to produce NOD2 ligands?

In the murine experiments depicted in Figure 4, because the bacterial intervention is being performed continuously in the drinking water, the investigators have not distinguished between colonization vs continuous oral dosing of the mice peptidoglycans. While I do not think additional experimentation is required to distinguish the individual contributions of these 2 components in their therapeutic intervention, I do think the interpretation of their results should include this perspective.

Reviewer #2 (Public Review):

Summary:

The gut microbiome contributes to variation in the efficacy of immune checkpoint blockade in cancer therapy; however, the mechanisms responsible remain unclear. Klupt et al. build upon prior data implicating the secreted peptidoglycan hydrolase SagA produced by Enterococcus faecium in immunotherapy, leveraging novel strains with sagA deleted and complemented. They find that sagA is non-essential, but sagA deletion leads to a marked growth defect due to impaired cell division. Furthermore, sagA is necessary for the immunogenic and anti-tumor effects of E. faecium. Together, this study utilizes compelling methods to provide fundamental new insights into E. faecium biology and host interactions, and a proof-of-concept for identifying the bacterial effectors of immunotherapy response.

Strengths:

Klupt et al. provide a well-written manuscript with clear and compelling main and supplemental figures. The methods used are state-of-the-art, including various imaging modalities, bacterial genetics, mass spectrometry, sequencing, flow cytometry, and mouse models of immunotherapy response. Overall, the data supports the conclusions, which are a valuable addition to the literature.

Weaknesses:

Only minor revision recommendations were noted.

eLife assessment

This manuscript presents a valuable, lightweight software package that aims to accelerate the implementation of experiment pipelines running on heterogeneous computer environments. The authors present solid evidence of the advantages of their chosen approach, including demonstrating the flexibility, ease of use, and practical utility of this new system. However, questions remain regarding the maturity of the project and its specific advantages in relation to existing widely adopted software packages.

Reviewer #1 (Public Review):

Summary:<br /> The authors have created a system for designing and running experimental pipelines to control and coordinate different programs and devices during an experiment, called Heron. Heron is based around a graphical tool for creating a Knowledge Graph made up of nodes connected by edges, with each node representing a separate Python script, and each edge being a communication pathway connecting a specific output from one node to an input on another. Each node also has parameters that can be set by the user during setup and runtime, and all of this behavior is concisely specified in the code that defines each node. This tool tries to marry the ease of use, clarity, and self-documentation of a purely graphical system like Bonsai with the flexibility and power of a purely code-based system like Robot Operating System (ROS).

Strengths:<br /> The underlying idea behind Heron, of combining a graphical design and execution tool with nodes that are made as straightforward Python scripts seems like a great way to get the relative strengths of each approach. The graphical design side is clear, self-explanatory, and self-documenting, as described in the paper. The underlying code for each node tends to also be relatively simple and straightforward, with a lot of the complex communication architecture successfully abstracted away from the user. This makes it easy to develop new nodes, without needing to understand the underlying communications between them. The authors also provide useful and well-documented templates for each type of node to further facilitate this process. Overall this seems like it could be a great tool for designing and running a wide variety of experiments, without requiring too much advanced technical knowledge from the users.

The system was relatively easy to download and get running, following the directions and already has a significant amount of documentation available to explain how to use it and expand its capabilities. Heron has also been built from the ground up to easily incorporate nodes stored in separate Git repositories and to thus become a large community-driven platform, with different nodes written and shared by different groups. This gives Heron a wide scope for future utility and usefulness, as more groups use it, write new nodes, and share them with the community. With any system of this sort, the overall strength of the system is thus somewhat dependent on how widely it is used and contributed to, but the authors did a good job of making this easy and accessible for people who are interested. I could certainly see Heron growing into a versatile and popular system for designing and running many types of experiments.

Weaknesses:<br /> The number one thing that was missing from the paper was any kind of quantification of the performance of Heron in different circumstances. Several useful and illustrative examples were discussed in depth to show the strengths and flexibility of Heron, but there was no discussion or quantification of performance, timing, or latency for any of these examples. These seem like very important metrics to measure and discuss when creating a new experimental system.

After downloading and running Heron with some basic test Nodes, I noticed that many of the nodes were each using a full CPU core on their own. Given that this basic test experiment was just waiting for a keypress, triggering a random number generator, and displaying the result, I was quite surprised to see over 50% of my 8-core CPU fully utilized. I don't think that Heron needs to be perfectly efficient to accomplish its intended purpose, but I do think that some level of efficiency is required. Some optimization of the codebase should be done so that basic tests like this can run with minimal CPU utilization. This would then inspire confidence that Heron could deal with a real experiment that was significantly more complex without running out of CPU power and thus slowing down.

I was also surprised to see that, despite being meant specifically to run on and connect diverse types of computer operating systems and being written purely in Python, the Heron Editor and GUI must be run on Windows. This seems like an unfortunate and unnecessary restriction, and it would be great to see the codebase adjusted to make it fully cross-platform-compatible.

Lastly, when I was running test experiments, sometimes one of the nodes, or part of the Heron editor itself would throw an exception or otherwise crash. Sometimes this left the Heron editor in a zombie state where some aspects of the GUI were responsive and others were not. It would be good to see a more graceful full shutdown of the program when part of it crashes or throws an exception, especially as this is likely to be common as people learn to use it. More problematically, in some of these cases, after closing or force quitting Heron, the TCP ports were not properly relinquished, and thus restarting Heron would run into an "address in use" error. Finding and killing the processes that were still using the ports is not something that is obvious, especially to a beginner, and it would be great to see Heron deal with this better. Ideally, code would be introduced to carefully avoid leaving ports occupied during a hard shutdown, and furthermore, when the address in use error comes up, it would be great to give the user some idea of what to do about it.

Overall I think that, with these improvements, this could be the beginning of a powerful and versatile new system that would enable flexible experiment design with a relatively low technical barrier to entry. I could see this system being useful to many different labs and fields.

Reviewer #2 (Public Review):

Summary:<br /> The authors provide an open-source graphic user interface (GUI) called Heron, implemented in Python, that is designed to help experimentalists to<br /> (1) design experimental pipelines and implement them in a way that is closely aligned with their mental schemata of the experiments,<br /> (2) execute and control the experimental pipelines with numerous interconnected hardware and software on a network.

The former is achieved by representing an experimental pipeline using a Knowledge Graph and visually representing this graph in the GUI. The latter is accomplished by using an actor model to govern the interaction among interconnected nodes through messaging, implemented using ZeroMQ. The nodes themselves execute user-supplied code in, but not limited to, Python.

Using three showcases of behavioral experiments on rats, the authors highlighted three benefits of their software design:<br /> (1) the knowledge graph serves as a self-documentation of the logic of the experiment, enhancing the readability and reproducibility of the experiment,<br /> (2) the experiment can be executed in a distributed fashion across multiple machines that each has a different operating system or computing environment, such that the experiment can take advantage of hardware that sometimes can only work on a specific computer/OS, a commonly seen issue nowadays,<br /> (3) the users supply their own Python code for node execution that is supposed to be more friendly to those who do not have a strong programming background.

Strengths:<br /> (1) The software is light-weight and open-source, provides a clean and easy-to-use GUI,<br /> (2) The software answers the need of experimentalists, particularly in the field of behavioral science, to deal with the diversity of hardware that becomes restricted to run on dedicated systems.<br /> (3) The software has a solid design that seems to be functionally reliable and useful under many conditions, demonstrated by a number of sophisticated experimental setups.<br /> (4) The software is well documented. The authors pay special attention to documenting the usage of the software and setting up experiments using this software.

Weaknesses:<br /> (1) While the software implementation is solid and has proven effective in designing the experiment showcased in the paper, the novelty of the design is not made clear in the manuscript. Conceptually, both the use of graphs and visual experimental flow design have been key features in many widely used softwares as suggested in the background section of the manuscript. In particular, contrary to the authors' claim that only pre-defined elements can be used in Simulink or LabView, Simulink introduced MATLAB Function Block back in 2011, and Python code can be used in LabView since 2018. Such customization of nodes is akin to what the authors presented.

(2) The authors claim that the knowledge graph can be considered as a self-documentation of an experiment. I found it to be true to some extent. Conceptually it's a welcoming feature and the fact that the same visualization of the knowledge graph can be used to run and control experiments is highly desirable (but see point 1 about novelty). However, I found it largely inadequate for a person to understand an experiment from the knowledge graph as visualized in the GUI alone. While the information flow is clear, and it seems easier to navigate a codebase for an experiment using this method, the design of the GUI does not make it a one-stop place to understand the experiment. Take the Knowledge Graph in Supplementary Figure 2B as an example, it is associated with the first showcase in the result section highlighting this self-documentation capability. I can see what the basic flow is through the disjoint graph where 1) one needs to press a key to start a trial, and 2) camera frames are saved into an avi file presumably using FFMPEG. Unfortunately, it is not clear what the parameters are and what each block is trying to accomplish without the explanation from the authors in the main text. Neither is it clear about what the experiment protocol is without the help of Supplementary Figure 2A.

In my opinion, text/figures are still key to documenting an experiment, including its goals and protocols, but the authors could take advantage of the fact that they are designing a GUI where this information, with properly designed API, could be easily displayed, perhaps through user interaction. For example, in Local Network -> Edit IPs/ports in the GUI configuration, there is a good tooltip displaying additional information for the "password" entry. The GUI for the knowledge graph nodes can very well utilize these tooltips to show additional information about the meaning of the parameters, what a node does, etc, if the API also enforces users to provide this information in the form of, e.g., Python docstrings in their node template. Similarly, this can be applied to edges to make it clear what messages/data are communicated between the nodes. This could greatly enhance the representation of the experiment from the Knowledge graph.

(3) The design of Heron was primarily with behavioral experiments in mind, in which highly accurate timing is not a strong requirement. Experiments in some other areas that this software is also hoping to expand to, for example, electrophysiology, may need very strong synchronization between apparatus, for example, the record timing and stimulus delivery should be synced. The communication mechanism implemented in Heron is asynchronous, as I understand it, and the code for each node is executed once upon receiving an event at one or more of its inputs. The paper, however, does not include a discussion, or example, about how Heron could be used to address issues that could arise in this type of communication. There is also a lack of information about, for example, how nodes handle inputs when their ability to execute their work function cannot keep up with the frequency of input events. Does the publication/subscription handle the queue intrinsically? Will it create problems in real-time experiments that make multiple nodes run out of sync? The reader could benefit from a discussion about this if they already exist, and if not, the software could benefit from implementing additional mechanisms such that it can meet the requirements from more types of experiments.

(4) The authors mentioned in "Heron GUI's multiple uses" that the GUI can be used as an experimental control panel where the user can update the parameters of the different Nodes on the fly. This is a very useful feature, but it was not demonstrated in the three showcases. A demonstration could greatly help to support this claim.

(5) The API for node scripts can benefit from having a better structure as well as having additional utilities to help users navigate the requirements, and provide more guidance to users in creating new nodes. A more standard practice in the field is to create three abstract Python classes, Source, Sink, and Transform that dictate the requirements for initialisation, work_function, and on_end_of_life, and provide additional utility methods to help users connect between their code and the communication mechanism. They can be properly docstringed, along with templates. In this way, the com and worker scripts can be merged into a single unified API. A simple example that can cause confusion in the worker script is the "worker_object", which is passed into the initialise function. It is unclear what this object this variable should be, and what attributes are available without looking into the source code. As the software is also targeting those who are less experienced in programming, setting up more guidance in the API can be really helpful. In addition, the self-documentation aspect of the GUI can also benefit from a better structured API as discussed in point 2 above.

(6) The authors should provide more pre-defined elements. Even though the ability for users to run arbitrary code is the main feature, the initial adoption of a codebase by a community, in which many members are not so experienced with programming, is the ability for them to use off-the-shelf components as much as possible. I believe the software could benefit from a suite of commonly used Nodes.

(7) It is not clear to me if there is any capability or utilities for testing individual nodes without invoking a full system execution. This would be critical when designing new experiments and testing out each component.

Reviewer #3 (Public Review):

Summary:<br /> The authors present a Python tool, Heron, that provides a framework for defining and running experiments in a lab setting (e.g. in behavioural neuroscience). It consists of a graphical editor for defining the pipeline (interconnected nodes with parameters that can pass data between them), an API for defining the nodes of these pipelines, and a framework based on ZeroMQ, responsible for the overall control and data exchange between nodes. Since nodes run independently and only communicate via network messages, an experiment can make use of nodes running on several machines and in separate environments, including on different operating systems.

Strengths:<br /> As the authors correctly identify, lab experiments often require a hodgepodge of separate hardware and software tools working together. A single, unified interface for defining these connections and running/supervising the experiment, together with flexibility in defining the individual subtasks (nodes) is therefore a very welcome approach. The GUI editor seems fairly intuitive, and Python as an accessible programming environment is a very sensible choice. By basing the communication on the widely used ZeroMQ framework, they have a solid base for the required non-trivial coordination and communication. Potential users reading the paper will have a good idea of how to use the software and whether it would be helpful for their own work. The presented experiments convincingly demonstrate the usefulness of the tool for realistic scientific applications.

Weaknesses:<br /> In my opinion, the authors somewhat oversell the reproducibility and "self-documentation" aspect of their solution. While it is certainly true that the graph representation gives a useful high-level overview of an experiment, it can also suffer from the same shortcomings as a "pure code" description of a model - if a user gives their nodes and parameters generic/unhelpful names, reading the graph will not help much. Making the link between the nodes and the actual code is also not straightforward, since the code for the nodes is spread out over several directories (or potentially even machines), and not directly accessible from within the GUI. The authors state that "[Heron's approach] confers obvious benefits to the exchange and reproducibility of experiments", but the paper does not discuss how one would actually exchange an experiment and its parameters, given that the graph (and its json representation) contains user-specific absolute filenames, machine IP addresses, etc, and the parameter values that were used are stored in general data frames, potentially separate from the results. Neither does it address how a user could keep track of which versions of files were used (including Heron itself).

Another limitation that in my opinion is not sufficiently addressed is the communication between the nodes, and the effect of passing all communications via the host machine and SSH. What does this mean for the resulting throughput and latency - in particular in comparison to software such as Bonsai or Autopilot? The paper also states that "Heron is designed to have no message buffering, thus automatically dropping any messages that come into a Node's inputs while the Node's worker function is still running."- it seems to be up to the user to debug and handle this manually?

As a final comment, I have to admit that I was a bit confused by the use of the term "Knowledge Graph" in the title and elsewhere. In my opinion, the Heron software describes "pipelines" or "data workflows", not knowledge graphs - I'd understand a knowledge graph to be about entities and their relationships. As the authors state, it is usually meant to make it possible to "test propositions against the knowledge and also create novel propositions" - how would this apply here?

eLife assessment

This useful manuscript reports on a new mouse model for LAMA2-MD, a rare but very severe congenital muscular dystrophy; the knockout mice were generated by removing exon3 in the Lama2 gene, which results in a frameshift in exon4 and a premature stop codon. These animals lack any laminin-alpha2 protein and confirm results from previous Lama2 knockout models. Additionally, this study includes transcriptomics data that might be a good resource for the field. However, the experimental evidence supporting the main claims of the manuscript is incomplete, citations of previous Lama2 null mice studies are lacking, and both data presentation and interpretation need improvement.

Reviewer # 1 (Public Review):

Summary:<br /> The paper nicely confirms the phenotype of Lama2 knockout mice and extends the phenotypic description with a set of new molecular studies (transcriptomics) that might serve as a resource for other scientists interested in the LAMA2-MD.

Strengths:<br /> Set of new molecular studies (transcriptomics) that might serve as a resource for other scientists interested in the LAMA2-MD.

Weaknesses:<br /> Some of the figures are of rather poor quality. For example, the H&E and Sirius Red stainings in Figures 3 and 4 are quite poor so it is difficult to see what is going on in the muscles. The authors should take note of another publication on dy3K/dy3K mice of similar age (PMID: 31586140) where such images are of much higher quality. Similarly, the Western blot for laminin-alpha2 (Figure 4B) of the wild-type mouse needs improvement. If the single laminin-alpha2 protein is not detected, there is an issue with the denaturation buffer used to load the protein.

My biggest concern is, however, the many overstatements in the manuscript and the over-interpretation of the data. This already starts with the first sentence in the abstract where the authors write: "Understanding the underlying pathogenesis of LAMA2-related muscular dystrophy (LAMA2-MD) have been hampered by lack of genuine mouse model." This is not correct as the dy3K/dy3K, generated in 1997 (PMID: 9326364), are also Lama2 knockout mice; there are also other strains (dyW/dyW mice) that are severely affected and there are the dy2J/dy2J mice that represent a milder form of LAMA2-MD.

Similarly, the last two sentences of the abstract "This is the first reported genuine model simulating human LAMA2-MD. We can use it to study the molecular pathogenesis and develop effective therapies." are a clear overstatement. The mechanisms of the disease are well studied and the above-listed mouse models have been amply used to develop possible treatment options.

The overinterpretation concerns the results from transcriptomics. The fact that Lama2 is expressed in particular cell types of the brain does not at all imply that Lama2 knockout mice have a defect in the blood-brain barrier as the authors state. If there are no functional data, this cannot be stated. Indications for a blood-brain barrier defect come from work in dy3K/dy3K mice (PMID: 25392494) and this needs to be written like this.

Finally, the bulk RNA-seq data also needs to be presented in a disease context. The authors, again, mix up changes in expression with functional impairment. All gene expression changes are interpreted as direct evidence of an involvement of the cytoskeleton. In fact, changes in the cytoskeleton are more likely a consequence of the severe muscle phenotype and the delay in muscle development. This is particularly possible as muscle samples from 14-day-old mice are compared; a stage at which muscle still develops and grows tremendously. Thus, all the data need to be interpreted with caution.

In summary, the authors need to improve data presentation and, most importantly, they need to tone down the interpretation and they must be fully aware that their work is not as novel as they present it.

Reviewer #2 (Public Review):

Summary:<br /> This manuscript describes the production of a mouse model for LAMA2-CMD. This mouse was produced using CRISPR-Cas9 and deleted exon 3 of the Lama2 gene. The mice exhibit reduced life expectancy, muscle pathology, and disruption of the gliovascular basal lamina assembly leading to defects in the blood-brain barrier. Single-cell RNAseq was used to explore the effect that loss of Laminin-211/221 had on gene expression.

Strengths:<br /> (1) The authors produced a mouse model of LAMA2-CMD using CRISPR-Cas9.

(2) The authors identify cellular changes that disrupt the blood-brain barrier.

Weaknesses:<br /> (1) The major weakness is the manuscript reads like this was the first-ever knockout mouse model generated for LAMA2-CMD. There are in fact many Lama2 knockout mice (dy, dy2J, dy4k, dyW, and more) which have all been extensively studied with publications. It is important for the authors to comment on these other published studies that have generated these well-studied mouse lines. Therefore, there is a lack of background information on these other Lama2 null mice.

(2) The phenotypes of dyH/dyH are similar to, if not identical to dy/dy, dy2J/dy2J, dy4k/dy4k, dyW/dyW including muscle wasting, muscle weakness, compromised blood-brain barrier, and reduced life expectancy. This should be addressed, and a comparison made with Lama2 deficient mice in published literature.

(3) Recent published studies (Chen et al., Development (2023), PMID 36960827) show loss of Itga7 causes disruption of the brain-vascular basal lamina leading to defects in the blood-brain barrier. This should be referenced in the manuscript since this integrin is a major Laminin-211/221 receptor in the brain and the mouse model appears to phenocopy the dyH/dyH mouse model.

eLife assessment

This solid study presents valuable insights into the role of Cers1 on skeletal muscle function during aging, although further substantiation would help to fully establish the experimental assertions. It examines an unexplored aspect of muscle biology that is a relevant opening to future studies in this area of research.

Reviewer #1 (Public Review):

Summary:

The authors identified that genetically and pharmacological inhibition of CERS1, an enzyme implicated in ceramides biosynthesis worsen muscle fibrosis and inflammation during aging.

Strengths:

The study points out an interesting issue on excluding CERS1 inhibition as a therapeutic strategy for sarcopenia. Overall, the article it's well written and clear.

Reviewer #2 (Public Review):

Summary:

The manuscript by Wohlwend et al. investigates the implications of inhibiting ceramide synthase Cers1 on skeletal muscle function during aging. The authors propose a role for Cers1 in muscle myogenesis and aging sarcopenia. Both pharmacological and AAV-driven genetic inhibition of Cers1 in 18-month-old mice lead to reduced C18 ceramides in skeletal muscle, exacerbating age-dependent features such as muscle atrophy, fibrosis, and center-nucleated fibers. Similarly, inhibition of the Cers1 orthologue in C. elegans reduces motility and causes alterations in muscle morphology.

Strengths:

The study is well-designed, carefully executed, and provides highly informative and novel findings that are relevant to the field.

Author Response

eLife assessment

This manuscript provides useful information about the lipid metabolite 15d-PGJ2 as a potential regulator of myoblast senescence. The authors provide experimental evidence that 15d-PGJ2 inhibits myoblast proliferation and differentiation by binding and regulating HRas. However, the manuscript is incomplete in its current form, as it lacks robust support from the data regarding the main conclusions related to senescence and technical concerns related to the senescence models used in this study.

Authors Response- We ae grateful to the editors and the reviewers for their time and comments in sharpening the science and the writing of the manuscript. We have attached a detailed response to emphasize that the manuscript does include robust evidence regarding the claims, which could have been missed during the review process. We have provided a better context for these points now.

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The authors show that upon treatment with Doxorubicin (Doxo), there is an increase in senescence and inflammatory markers in the muscles. They also show these genes get upregulated in C2C12 myoblasts when treated with conditioned media or 15d-PGJ2. 15dPGJ2 induces cell death in the myoblasts, decreases proliferation (measured by cell numbers), and decreases differentiation and fusion. 15d-PGJ2 modified Cys184 of HRas, which is required for its activation as indicated by the FRET analysis with RAF RBD. They also showed that 15d-PGJ2 activates ERK signaling, but not Akt signaling, through the electrophilic center. 15d-PGJ2 inhibits Golgi localization of HRAS (only WT, not C181 or C184 mutant). They also showed that expressing the WT HRas followed by 15d-PGJ2 treatment led to a decrease in the levels of MHC mRNA and protein, and this defect is dependent on C184. This is a well-written manuscript with interesting insights into the mechanism of action of 15d-PGJ2. However, some clarification and experiments will help the paper advance the field significantly.

Strengths:

The data clearly shows that 15d-PGJ2 has a negative role in the myoblast cells and that it leads to modification of HRas protein. Moreover, the induction of biosynthetic enzymes in the PGD2 pathway also supports the induction of 15d-PGJ2 in Doxorubicin-treated cells. Both conditioned media experiments and the 15d-PGJ2 experiments show that 15d-PGJ2 could be the active component secreted by the senescent myoblasts.

Weaknesses:

The genes that are upregulated in the muscles upon injection with Doxo are also markers for inflammation. Since Doxo is also known to induce systemic inflammation, it is important to delineate these two effects (inflammatory cells vs senescent cells). The expression of beta Gal and other markers of senescence in the tissue sections will help to delineate these.

As pointed out Doxo induces systemic inflammation along with inducing DNA damage-mediated senescence. Therefore, along with the inflammatory markers of the SASP (CXCL1/2, TNF1α, IL6, PTGS1/2, PTGDS) we also observed an increase in the mRNA levels of canonical markers of DNA damage-mediated senescence. We observed an increase in the mRNA levels of cell cycle and senescence associated proteins p16 and p21 (Fig. 1C). We also observed an increased nuclear accumulation of p21 (Fig. 1A) and increased levels of phosphorylated H2A.X in the nucleus (Fig. 1B). We will characterize other markers of senescence including senescence-associated β galactosidase in the revised manuscript.

In Figure 2, where the defect in the differentiation of myoblasts upon treatment with 15d-PGJ2 is shown, most of the cells die within 48 hours at higher concentrations, making it difficult to perform the experiments. This also shows that 15d-PGJ2 was toxic to these cells. Lower concentrations show a decrease in the differentiation based on the lower number of nuclei in fibers and low expression of MyoD, MyoG, and MHC. However, it is unclear if this is due to increased cell death or defective differentiation. It would be a lot more informative if the cell count, cell division, and cell death could be plotted for these concentrations of the drug during the experiment.

We only observed the death of cells at higher concentrations of 15d-PGJ2 (5 µM and 10 µM) (Fig. S2A), but not significantly at the 4 µM concentration used in Figure 2. This is the reason 4uM was used, and we should have clarified this. We will include viability data for the low concentration of 15d-PGJ2 (4 µM) in the revised manuscript.

Also, in the myoblast experiments, are the effects of treatment with Dox reversible?

The treatment with Doxorubicin is irreversible as the senescent phenotype was not reversed after withdrawal of Doxorubicin, even after 20 days.

In Figure 3, most of the experiments are done at a high concentration, which induces almost complete cell death within 48 hours.

Figure 3 is an acute experiment for only 1 hour, at which time no cell death was observed. Specifically, we measured the phosphorylation of Erk and Akt proteins after 1 hour of treatment with 15d-PGJ2 (10 µM) during which we did not observe any cell death.

Even at such a high concentration of 15dPGJ2, the increase in ERK phosphorylation is minimal.

We observe a ~30% increase in the phosphorylation of Erk proteins after treatment with 15d-PGJ¬2 in 0.2% serum medium compared to treatment with vehicle (DMSO). This is reproducible and significant.

The experiment Figure 4C shows that C181 and C84 mutants of the HRas show higher levels in Golgi compared with WT. However, this could very well be due to the defect in palmitoylation rather than the modification with 15d-PGJ2.

Our data does not suggest higher levels of C184S mutant in the Golgi compared with WT (Fig. S4A). We observed that the ratio of HRas levels in the Golgi to the HRas levels in the plasma membrane were similar in C2C12 cells expressing HRas C184S and HRas WT (Fig. S4A graph columns 1 and 5).

Though the authors allude to the possibility that intracellular redistribution of HRas by 15d-PGJ2 requires C181 palmitoylation, the direct influence of C184 modification on C181 palmitoylation is not shown. To have a meaningful conclusion, the authors need to compare the palmitoylation and modification with 15d-PGJ2.

Palmitoylation of HRas C181S is required for the localization of HRas at the plasma membrane. The inhibition of palmitoylation of C181, either by mutation (C181S) or treatment with protein palmitoyl transferase inhibitor (2-Bromopalmitate), results in the accumulation of HRas at Golgi(Rocks et al., 2005) (Fig. S4A). Modification of HRas at C184 by 15d-PGJ2 (Fig. 3A) could inhibit the palmitoylation of HRas at C181. However, our data does not support this hypothesis as modification of HRas WT by 15d-PGJ2 does not increase the level of HRas at the Golgi, like in the case of inhibition of cysteine palmitoylation due to C181S mutation.

To test if the inhibition of myoblast differentiation depends on HRas, they overexpressed the HRas and mutants in the C2C12 lines. However, this experiment does not take the endogenous HRAs into consideration, especially when interpreting the C184 mutant. An appropriate experiment to test this would be to knock down or knock out HRas (or make knock-in mutations of C184) and show that the effect of 15d-PGJ2 disappears.

Endogenous HRas (wild type) is present in the C2C12 cells overexpressing the EGFP-tagged HRas constructs. Therefore, we only observe a partial rescue in the differentiation after 15d-PGJ2 treatment in C2C12 cells expressing the C184S mutant (Fig. 4D and E). However, since HRas is expressed under high expression CMV promoter and in the absence of other regulatory elements, the overexpressed constructs do show a dominant effect over the endogenous HRas, showing cysteine mutant dependent inhibition of differentiation of myoblasts after treatment with 15d-PGJ2 (Fig. 4D and E).

Moreover, in this specific experiment, it is difficult to interpret without a control with no HRas construct and another without the 15d-PGJ2 treatment.

The mRNA levels of MyoD, MyoG, and MHC in C2C12 cells expressing HRas constructs after treatment with 15d-PGJ2 were normalized to the mRNA levels in C2C12 cells expressing corresponding constructs and were treated with vehicle (DMSO). mRNA levels in C2C12 cells treated with vehicle were not shown as they were normalized to 1. MHC protein levels in C2C12 cells expressing HRas constructs after 15d-PGJ2 treatment were normalized to that in C2C12 cells treated with vehicle (DMSO). Since the hypothesis to study the effect of HRas cysteine mutations on the differentiation of myoblasts after treatment with 15d-PGJ2, C2C12 cells expressing HRas WT serve as adequate control. Fig. 2 shows the effect of 15d-PGJ2 on muscle differentiation when HRas was not overexpressed.

Moreover, the overall study does not delineate the toxic effects of 15d-PGJ2 from its effect on the differentiation. The inhibition of differentiation in C212 cells after treatment with 15d-PGJ2 cannot be attributed to the general toxicity of 15d-PGJ2 in cells. We show that the inhibition of differentiation of myoblasts after 15d-PGJ2 depends on modification of HRas at C184 i.e. failure to modify HRas at C184 (Fig. 3A) and resultant activation (Fig. 3B) by 15d-PGJ2 rescues this inhibition of differentiation of C2C12 cells (Fig. 4D and E), dissecting the inhibition of differentiation of myoblasts by 15d-PGJ2 from general toxic effects of 15d-PGJ2 on cell physiology.

Please note that the effect of 15d-PGJ2 on cell physiology is context-specific. On one hand, 15d-PGJ2 has been shown to exert tumor-suppressor effects by inhibiting the proliferation of ovarian cancer cells and lung adenocarcinoma cells (de Jong et al., 2011; Slanovc et al., 2024), 15d-PGJ2 also exerts pro-carcinogenic effects by induction of epithelial to mesenchymal transition in breast cancer cells MCF7 and inhibition of tumor-suppressor protein p53 in MCF7 and PC-3 cells (Choi et al., 2020; Kim et al., 2010).

Reviewer #2 (Public Review):

Summary:

In this study, Swarang and colleagues identified the lipid metabolite 15d-PGJ2 as a potential component of senescent myoblasts. They proposed that 15d-PGJ2 inhibits myoblast proliferation and differentiation by binding and regulating HRas, suggesting its potential as a target for restoring muscle homeostasis post-chemotherapy.

Strengths:

The regulation of HRas by 15d-PGJ2 is well controlled.

Weaknesses:

The novelty of the study is compromised as the activation of PGD and 15d-PGJ2, as well as the regulation of HRas and cell proliferation, have been previously reported.

Literature does support this statement, and it is important to clarify this mis-impression for the field as whole

Let us clarify-

Covalent modification of HRas by 15d-PGJ2 has been reported only twice in the literature(Luis Oliva et al., 2003; Yamamoto et al., 2011) in fibroblasts and neurons respectively.

Interaction between HRas and 15d-PGJ2 in skeletal muscles has not been shown before, even though both HRas and 15d-PGJ2 are shown to be key regulators of muscle homeostasis.

Activation of HRas by 15d-PGJ2 was reported first by Luis Oliva et al (Luis Oliva et al., 2003). However, this study does not comment on the functional implications of activation of HRas signaling.

Recently, our lab contributed to a study where the functional implication of activation of HRas signaling due to covalent modification by 15d-PGJ2 was shown in the maintenance of senescence phenotype (Wiley et al., 2021).

15d-PGJ2 was shown to inhibit the differentiation of myoblasts by Hunter et al (Hunter et al., 2001). This study hypothesized that the inhibition of myoblast differentiation is via 15d-PGJ2 mediated activation of the PPARγ signaling, the study also showed inhibition of myoblast differentiation independent of PPARγ activity, suggesting the presence of other mechanisms.

This is the first study to show a molecular mechanism where activation of HRas signaling in skeletal myoblasts due to covalent modification by 15d-PGJ2 at C184 of HRas inhibits the differentiation of skeletal myoblasts.

Additionally, there are major technical concerns related to the senescence models, limiting data interpretation regarding the relevance to senescent cells.

Major concerns:

(1) The C2C12 cell line is not an ideal model for senescence study due to its immortalized nature and lack of normal p16 expression. A more suitable myoblasts model is recommended, with a more comprehensive characterization of senescence features.

C2C12 is a good model for DNA damage based senescence that is used in this manuscript. It is not a models for replicative senescence since it is immortalized. In this study we show that C2C12 cells undergo DNA damage mediated senescence after treatment with Doxo. We also observe similar phenotype in MCF7 breast cancer cells and IMR90 lung fibroblasts after treatment with Doxo (Data will be updated in the supplementary figure 1). Also, several reports in the literature have shown induction of senescence in C2C12 cells. Moiseeva et al 2023 show induction of senescence in C2C12 cells after etoposide mediated DNA damage. Moustogiannis et al 2021 show induction of replicative senescence in C2C12 cells.

(2) The source of increased PGD or its metabolites in the conditioned medium is unclear. Including other senescence models, such as replicative or oncogene-induced senescence, would strengthen the study.

Fig. 1E shows time dependent increase in the expression of PGD2 biosynthetic enzymes in senescent C2C12 cells. Fig. 1F shows increase in the levels of 15d-PGJ2 secreted by senescent C2C12 cells in the conditioned medium. This data shows that senescent C2C12 cells are the source of PGD and its metabolites in the conditioned medium.

Again, C2C12 is not suitable for replicative senescence due to its immortalized status.

We and others have shown that C2C12 cells undergo senescence, and this manuscript only used DNA damage induced senescence.

(3) In the in vivo part, it is unclear whether the increased expression of PTGS1, PTGS2, and PTGDS is due to senescence or other side effects of DOXO.

We concur that this is a limitation of this study and the subsequent work will demonstrate the origin of prostaglandin biosynthesis after treatment with Doxo in vivo.

(4) Figure 2A lacks an important control from non-senescent cells during the measurement of C2C12 differentiation in the presence of a conditioned medium.

Figure 2A tests the effect of prostaglandin PGD2 and its metabolites secreted by the senescent cells on the differentiation of myoblasts. Therefore, we inhibited the synthesis of PGD2 in senescent cells by treatment with AT-56, and then collected the conditioned medium. Conditioned medium collected from senescent C2C12 cells treated with vehicle (DMSO) served as a control for the experiment, whereas differentiation of C2C12 cells without any treatment serves as a positive control.

There is no explanation of how differentiation was quantified or how the fusion index was calculated.

The fusion index was calculated using a published myotube analyzer software (Noë et al., 2022). Appropriate info will be added to the materials and methods section in the revised manuscript.

eLife assessment

This manuscript provides useful information about the lipid metabolite 15d-PGJ2 as a potential regulator of myoblast senescence. The authors provide experimental evidence that 15d-PGJ2 inhibits myoblast proliferation and differentiation by binding and regulating HRas. However, the manuscript is incomplete in its current form, as it lacks robust support from the data regarding the main conclusions related to senescence and technical concerns related to the senescence models used in this study.

Reviewer #1 (Public Review):

Summary:

The authors show that upon treatment with Doxorubicin (Doxo), there is an increase in senescence and inflammatory markers in the muscles. They also show these genes get upregulated in C2C12 myoblasts when treated with conditioned media or 15d-PGJ2. 15dPGJ2 induces cell death in the myoblasts, decreases proliferation (measured by cell numbers), and decreases differentiation and fusion. 15d-PGJ2 modified Cys184 of HRas, which is required for its activation as indicated by the FRET analysis with RAF RBD. They also showed that 15d-PGJ2 activates ERK signaling, but not Akt signaling, through the electrophilic center. 15d-PGJ2 inhibits Golgi localization of HRAS (only WT, not C181 or C184 mutant). They also showed that expressing the WT HRas followed by 15d-PGJ2 treatment led to a decrease in the levels of MHC mRNA and protein, and this defect is dependent on C184. This is a well-written manuscript with interesting insights into the mechanism of action of 15d-PGJ2. However, some clarification and experiments will help the paper advance the field significantly.

Strengths:

The data clearly shows that 15d-PGJ2 has a negative role in the myoblast cells and that it leads to modification of HRas protein. Moreover, the induction of biosynthetic enzymes in the PGD2 pathway also supports the induction of 15d-PGJ2 in Doxorubicin-treated cells. Both conditioned media experiments and the 15d-PGJ2 experiments show that 15d-PGJ2 could be the active component secreted by the senescent myoblasts.

Weaknesses:

The genes that are upregulated in the muscles upon injection with Doxo are also markers for inflammation. Since Doxo is also known to induce systemic inflammation, it is important to delineate these two effects (inflammatory cells vs senescent cells). The expression of beta Gal and other markers of senescence in the tissue sections will help to delineate these.

In Figure 2, where the defect in the differentiation of myoblasts upon treatment with 15d-PGJ2 is shown, most of the cells die within 48 hours at higher concentrations, making it difficult to perform the experiments. This also shows that 15d-PGJ2 was toxic to these cells. Lower concentrations show a decrease in the differentiation based on the lower number of nuclei in fibers and low expression of MyoD, MyoG, and MHC. However, it is unclear if this is due to increased cell death or defective differentiation. It would be a lot more informative if the cell count, cell division, and cell death could be plotted for these concentrations of the drug during the experiment. Also, in the myoblast experiments, are the effects of treatment with Dox reversible?

In Figure 3, most of the experiments are done at a high concentration, which induces almost complete cell death within 48 hours. Even at such a high concentration of 15dPGJ2, the increase in ERK phosphorylation is minimal.

The experiment Figure 4C shows that C181 and C84 mutants of the HRas show higher levels in Golgi compared with WT. However, this could very well be due to the defect in palmitoylation rather than the modification with 15d-PGJ2. Though the authors allude to the possibility that intracellular redistribution of HRas by 15d-PGJ2 requires C181 palmitoylation, the direct influence of C184 modification on C181 palmitoylation is not shown. To have a meaningful conclusion, the authors need to compare the palmitoylation and modification with 15d-PGJ2.

To test if the inhibition of myoblast differentiation depends on HRas, they overexpressed the HRas and mutants in the C2C12 lines. However, this experiment does not take the endogenous HRAs into consideration, especially when interpreting the C184 mutant. An appropriate experiment to test this would be to knock down or knock out HRas (or make knock-in mutations of C184) and show that the effect of 15d-PGJ2 disappears. Moreover, in this specific experiment, it is difficult to interpret without a control with no HRas construct and another without the 15d-PGJ2 treatment.

Moreover, the overall study does not delineate the toxic effects of 15d-PGJ2 from its effect on the differentiation.

Reviewer #2 (Public Review):

Summary:

In this study, Swarang and colleagues identified the lipid metabolite 15d-PGJ2 as a potential component of senescent myoblasts. They proposed that 15d-PGJ2 inhibits myoblast proliferation and differentiation by binding and regulating HRas, suggesting its potential as a target for restoring muscle homeostasis post-chemotherapy.

Strengths:

The regulation of HRas by 15d-PGJ2 is well controlled.

Weaknesses:

The novelty of the study is compromised as the activation of PGD and 15d-PGJ2, as well as the regulation of HRas and cell proliferation, have been previously reported. Additionally, there are major technical concerns related to the senescence models, limiting data interpretation regarding the relevance to senescent cells.

Major concerns:<br /> (1) The C2C12 cell line is not an ideal model for senescence study due to its immortalized nature and lack of normal p16 expression. A more suitable myoblasts model is recommended, with a more comprehensive characterization of senescence features.

(2) The source of increased PGD or its metabolites in the conditioned medium is unclear. Including other senescence models, such as replicative or oncogene-induced senescence, would strengthen the study. Again, C2C12 is not suitable for replicative senescence due to its immortalized status.

(3) In the in vivo part, it's unclear whether the increased expression of PTGS1, PTGS2, and PTGDS is due to senescence or other side effects of DOXO.

(4) Figure 2A lacks an important control from non-senescent cells during the measurement of C2C12 differentiation in the presence of a conditioned medium. There is no explanation of how differentiation was quantified or how the fusion index was calculated.

Reviewer #2 (Public Review):

Summary:

In the manuscript, Yu et al reported a two-sample Mendelian randomization study to evaluate the causation between polyunsaturated fatty acids (PUFA) and cerebral aneurysm, based on summary statistics from published genome-wide association studies. The authors identified that omega-3 fatty acids and Docosahexaenoic acid decreased the risk for intracranial aneurysm (IA) and aneurysmal subarachnoid hemorrhage (aSAH). COLOC analysis suggested that the acids and IA, aSAH likely share causal variants in gene fatty acid desaturase 2.

Strengths:

The methodology is sound, with appropriate sensitivity analysis.

Weaknesses:

The results did not provide significant novel findings. The interpretation of the results is not sound.

eLife assessment

Yu and colleagues used two-sample MR to test the effect of PUFA on cerebral aneurysms. They found that genetically predicted omega-3 and DHA decreased the risk for Intracranial Aneurysm and Subarachnoid Haemorrhage. This work is useful, although the evidence is incomplete as it lacks a robust set of analyses to provide credibility to the findings.

Reviewer #1 (Public Review):

Summary:

The authors performed two-sample MR combined with sensitivity analyses and colocalization to test the effect of PUFA on cerebral aneurysms. They found that genetically predicted omega-3 and DHA decreased the risk for intracranial aneurysm (IA) and subarachnoid haemorrhage (SAH) but not for unruptured IA (uIA).

Strengths:

PUFA on the risk of cerebral aneurysms is of clinical importance; the authors performed multiple sensitivity analyses to ensure MR fulfills its assumptions.

Weakness:

In my opinion, the major weakness is the selection of IVs, the same IVs should be used for each exposure, especially when the outcomes (IA, SAH, and uIA) are closely related. The removal of IVs was inconsistent, for example, why was LPA rs10455872 removed for SAH but not for uIA? (significantly more IVs were used for uIA). The authors should provide more details for the justification of the removal of IVs other than only indicating "confounder" in supplementary tables. The authors should also perform additional analyses including all IVs and IVs from other PUFA GWAS.

In addition, it seems that the SNPs in the FADS locus were driving the MR association, while FADS is a very pleiotropic locus associated with many lipid traits, removing FADS could attenuate the MR effect. The authors should perform a sensitivity analysis to remove this locus.

Instead of removing multiple "confounder" IVs which I think may bias the MR results due to very closely related lipid traits, the authors should perform multivariable MR to identify independent effects of PUFAs to IA, conditioning on other PUFAs and/or other lipids.

Colocalization was not well described, the authors should include the colocalization results for each locus in a supplementary table. They also mentioned "a large PP for H4 (PP.H4 above 0.75) strongly supports shared causal variants affecting both gene expression and phenotype". The authors should make sure that the colocalization was performed using the expression data of each gene or using the GWAS summary of each PUFA locus.

Reviewer #2 (Public Review):

Significance of the findings:

In this study, blood donors were assessed using serology and viral neutralization assays to determine the prevalence of SARS-CoV-2 antibodies. S1 and NCP antibodies were used to distinguish between vaccination and natural infection and virus-specific neut titers were used to determine which variants the antibodies respond to. The study reports almost universal antibody prevalence and increases in antibodies against specific variants at different points corresponding to circulating variants identified phylogenetically in neighbouring countries. The authors propose this approach for settings like Bolivia where genetic sequencing is not readily available. Unfortunately, there are significant limitations to this approach that limit its utility - serological data are available after the fact in a fast-moving pandemic and so are a poor alternative to phylogenetic data. Rather, serological information can supplement phylogenetic data and is most useful in estimating population-level immunity.

(1) Considerations in interpreting the results:

a. Serology provides different information to phylogenetic sequencing of the viruses and so both are important. Viral sequencing provides real-time information on circulating variants and indicates the proportion of each variant in circulation at any point as there are almost always multiple variants spreading but it is the fastest spreading variant that comes to dominate. Importantly serology measures asymptomatic infections as well, providing population estimates of infection that are not available through viral gene sequencing.

b. A major concern in the interpretation of serology is that antibody titers vary markedly over time with rapid declines in the first year post-infection or post-vaccination. However, these declines vary depending on whether hybrid immunity is present. Disentangling this retrospectively is a challenge. A low antibody titer could reflect an infection that occurred a few months ago but may be below the threshold for positivity at the time of testing. There is also substantial individual variability in antibody responses.

c. Serology becomes increasingly difficult to untangle when an individual has had doses of vaccine and multiple natural infections with different variants. Due to the importance of hybrid immunity in population risk to new variants, it would be useful for estimates of hybrid immunity to be generated based on anti-S1 and anti-NCP antibodies. From a population immunity perspective, this could be important in guiding future protection and boosting strategies.

d. Since there is cross-neutralization by the antibodies stimulated by each variant, it is important to establish the sensitivity and specificity of each of the neutralization assays in a panel comprising multiple variants. An assessment of the accuracy of the neut assay for each variant is needed to be confident that it is able to distinguish between variants.

e. Blood donors are notoriously poor representations of the general population in many countries, driven partly by whether donation is financially rewarded. For example, in the USA, drug addicts are disproportionately over-represented in blood donor populations as they use it as a source of money. The authors provide no information on whether the blood donor population in Bolivia is representative of the entire population. Comparison of the prevalence of specific disease markers in the general population and in blood donors could provide a signal of their comparability.

(2) Please provide the sensitivity and specificity of each of the assays so that the reader can assess the degree of accuracy in the assay that claims that the prevalent antibodies are due to, for example, omicron.

(3) Please provide an assessment of the representativity of the blood donor population eg. Is the prevalence of hepatitis B serological markers in the blood donor population comparable with the prevalence of hepatitis B serological markers in the general population from community-based studies?

eLife assessment

This important study examines SARS-CoV-2 seroprevalence in Bolivia and aims to provide insights into the transmission of the virus and the effects of vaccination on population immunity. However, the evidence for the main claims is incomplete because of the uncertainties about the accuracy of the neutralization assays given the cross-neutralization present across variants, as well as the selected population of blood donors tested. These uncertainties need to be addressed to support the premise of the paper.

Reviewer #1 (Public Review):

Summary:

This study provides valuable and comprehensive information about the SARS-CoV-2 seroprevalence during 2021 and 2022 in different regions of Bolivia. Moreover, data on immune responses against the SARS-CoV-2 variants based on neutralization tests denotes the presence of several virus variants circulating in the Bolivian population. Evidence for seroprevalence data provided by the authors is solid, across the study period, while data regarding variant circulation is limited to the early stages of the pandemic.

Strengths:

The major strength of this study is that it provided nationwide seroprevalence estimates from infection and/or vaccination based on antibodies against both spike and the nucleocapsid protein in a large representative sample of sera collected at two time-points from all departments of Bolivia, gaining insight into COVID-19 epidemiology. On the other hand, data from virus neutralization assays inferred the circulation during the study period of four SARS-CoV-2 variants in the population. Overall, the study results provide an overview of the level of viral transmission and vaccination and insights into the spread across the country of SARS-CoV-2 variants.

Weaknesses:

The assessment of a Lambda variant that circulated in several neighboring countries (Peru, Chile, and Argentina), which had a significant impact on the COVID-19 pandemic in the region, may have strengthened the study to contrast Gamma spread. In addition, even though neutralizing antibodies can certainly reveal previous infections of SARSCOV2 variants in the population, it is of limited value to infer from this information some potential timing estimates of specific variant circulation, considering the heterogeneous effects that past infections, vaccinations, or a combination of both could have on the level of variant-specific neutralizing antibodies and/or their cross-neutralization capacity.

An appraisal of whether the authors achieved their aims, and whether the results support their conclusions:

The conclusions of this paper are well supported by data, particularly regarding seroprevalence that reliably reflects the epidemiology of COVID-19 in Bolivia, and seroprevalence trends in other low- and middle-income countries.

A discussion of the likely impact of the work on the field, and the utility of the methods and data to the community:

Since this is the first study that has been conducted to assess indicators of immunity against SARS-CoV-2 in the population of Bolivia at a nationwide scale, seroprevalence data provided by geographic regions at two time-points can be useful as a reference for potential retrospective global meta-analysis and further explore and compare the risk factors for infection, variant distribution, and the impact on infection and vaccination, gaining deeper insights into understanding the evolution of the COVID-19 pandemic in Bolivia and in the region.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

The manuscript offers a commendable exploration into the relationship between plasma omega-6/omega-3 fatty acid ratios and mortality outcomes.

Strengths:

The chosen study design and analytical techniques align well with the research objectives, and the results resonate with existing literature.

Weaknesses:

Lack of information on the selection criteria for participants; 5. The analysis of individual PUFAs is not appropriate; The definition of comorbidities is vague; The rationale of conducting the mediation analysis of blood biomarkers is not given.

Thank you for your insightful feedback and for acknowledging the strengths of our manuscript, particularly regarding the alignment of our study design and analytical methods with our research objectives. Your recognition of how our results resonate with existing literature is greatly appreciated.

Addressing the concerns you've raised:

Selection Criteria for Participants: In the “Methods-Study population” section, we have outlined the exclusion criteria for participant selection. This information provides comprehensive insight into our methodology for selecting the study cohort.

Analysis of Individual PUFAs: We acknowledge your concern regarding the analysis of individual PUFAs due to their inter-correlations in plasma levels. However, the correlations between omega-3% and omega-6% (r = -0.12) and between DHA% and LA% (r = 0.03) are actually low. Because DHA is one of omega-3 PUFAs, we did not include PUFAs in the same model. Similar considerations apply to LA and omega-6. We believe that exploring the effects of individual fatty acids adds valuable depth to our research. Both DHA and LA have been included in the same model due to their low correlation, with careful adjustments for confounding factors to provide a nuanced understanding of their individual impacts on mortality.

Definition of Comorbidities: The definition of comorbidities, including hypertension, diabetes, and longstanding illness, is elaborated under the Methods section. These conditions were identified through self-reported data collected via the Assessment Centre Environment (ACE) touchscreen questionnaire, allowing us to capture a broad range of chronic conditions as reported by participants.

Rationale for Mediation Analysis: Initially, our approach to mediation analysis included various blood biomarkers available in the UK Biobank database to explore the potential underlying pathways. However, upon considering your feedback regarding the overlap of fatty acids with lipid classes or lipid particles in plasma, we have decided to remove these elements from our mediation analysis.

Reviewer #2 (Public Review):

Summary:

This study utilized a large sample from the UK Biobank which enhanced statistical robustness, employed a prospective design to establish clear temporal relationships, used objective biomarkers for assessing plasma omega-6/omega-3 ratio, and investigated various mortality causes including CVD and cancer for a holistic health understanding.

Strengths:

The authors used a large sample size, employed a prospective design, and investigated various mortality.

Weaknesses:

Analyzing n-3 and n-6 PUFAs separately might be less instructive. It might not be methodologically sound to treat TG, HDL, LDL, and apolipoproteins as mediators. It's imperative to exercise caution when drawing causal conclusions from the observed correlations. The manuscript might propose potential research trajectories.

We are grateful for your thoughtful analysis of our study's strengths and for your constructive feedback on areas for improvement.

Response to Weaknesses:

Analyzing n-3 and n-6 PUFAs Separately: We recognize the challenge in analyzing n-3 and n-6 PUFAs separately due to their correlations. However, the correlation between n-3% and n-6% in UK Biobank was actually relatively low (r = -0.12). We include them in one model to test if both are associated with the outcomes after controlling for the effects of the other. Indeed, both were negatively associated with the mortality outcomes in our analysis. We believe our supplemental analysis of n-3 and n-6 PUFAs provides useful information to the readers, in addition to our findings based on the n-6/n-3 ratio.

Mediation Analysis of TG, HDL, LDL, and Apolipoproteins: We appreciate your insight on the methodological considerations of treating these biomarkers as mediators. After careful review and in line with suggestions from another reviewer, we have removed these elements from our mediation analysis. This revision improves the net scientific rigor of our work, ensuring that our conclusions are drawn from the most robust and methodologically sound of our analyses.

Causal Conclusions from Correlations: We fully agree with the need for caution in interpreting correlations in observational studies. To this end, we have avoided implying causality in our manuscript. Terms suggesting causality, like "protective effects," have been replaced with "inverse associations" to more accurately represent our findings. This adjustment enhances the clarity and accuracy of our conclusions.

Proposing Future Research Trajectories: Recognizing the importance of advancing causal and mechanistic understanding in this field, we have called for future studies to further examine causality and characterize molecular mechanisms of the observed associations in our study.

Reviewer #3 (Public Review):

Summary:

The authors are trying to find out whether the levels of omega-6 and omega-3 fatty acids in the blood are linked to the likelihood of dying from anything, of dying from cancer and of dying from cardiovascular disease. They use a large dataset called UK Biobank where fatty acid levels were measured in blood at the start of the study and what happened to the participants over the following years (average of 12.7 years) was followed. They find that both omega-6 AND omega-3 fatty acids were linked with less likelihood of dying from anything, from cancer and from cardiovascular disease. The effects of omega-3s were stronger. They then made a ratio of omega-6 to omega-3 fatty acids and found that as that ratio increased risk of dying also increased,. This supports the idea that omega-3s have stronger effects than omega-6s.

Strengths:

This is a large study (over 85,000 participants) with a good follow up period (average 12.7 years). Using blood levels of fatty acids is superior to using estimated dietary intakes. The authors take account of many variables that could interfere with the findings (confounding variables) - they do this using statistical methods.

Weaknesses:

There are several omega-6 and omega-3 fatty acids - it is not clear which ones were actually measured in this study

Thank you for recognizing the strengths of our study, including the large sample size, the duration of follow-up, and our methodological approach to using blood levels of fatty acids and addressing potential confounders. Regarding the weakness you've highlighted, we understand the importance of specifying which omega-6 and omega-3 fatty acids were analyzed in our study. We have revised the Method section to provide detailed information about how the exposures were measured.

Recommendations for the author:

Reviewer #1 (Recommendations for the Authors):

To elevate the manuscript's scholarly rigor, I propose the following refinements:

(1) The manuscript lacks information on the selection criteria for participants and the representativeness of the UK Biobank cohort. It is important to provide details on how participants were selected and whether it is representative of the general population, which is crucial for assessing the generalizability of the findings.

We appreciate the opportunity to clarify the participant selection criteria and the representativeness of the UK Biobank cohort within our manuscript. In the “Methods-Study population” section, we delineated the exclusion criteria: "Participants with cancer (n=37,736) or CVD (n=100,972), those who withdrew from the study (n=879), and those with incomplete data on the plasma omega-6/omega-3 ratio (n=277,372) were excluded from this study, leaving 85,425 participants, 6,461 died during follow-up, including 2,794 from cancer and 1,668 from CVD." To further address representativeness, we performed a sensitivity analysis, examining the baseline characteristics of participants included in our study relative to those omitted due to lack of exposure information. This analysis, presented in Additional file 2: Table S13, indicates comparable baseline characteristics across both participant groups, bolstering confidence in the representativeness of our study sample with the general UK Biobank participants.

Regarding the UK Biobank's representativeness with the general population, we acknowledge that the cohort does not mirror the broader UK demographic in terms of socioeconomic and health profiles. Participants in the UK Biobank generally exhibit better health and higher socioeconomic status than the average UK resident, potentially influencing the disease prevalence and incidence rates. Nonetheless, the UK Biobank's extensive sample size and comprehensive exposure data enable the generation of valid estimates for exposure-disease associations. These estimates have been corroborated by findings from more demographically representative cohorts, as highlighted in the studies by Batty et al., and Fry et al..

We recognize the importance of this aspect and will incorporate a discussion on the implications of these factors for the generalizability of our findings in the “Discussion-Limitations” section of our manuscript. We are grateful for this insightful comment and believe that this addition will enhance the manuscript's contribution to the field.

Here is what we added in the “Discussion-Limitations” section of our manuscript: “Third, we acknowledged that the cohort did not mirror the broader UK demographic in terms of socioeconomic and health profiles. Participants in the UK Biobank generally exhibited better health and higher socioeconomic status than the average UK resident, potentially influencing the disease prevalence and incidence rates. Nonetheless, the UK Biobank's extensive sample size and comprehensive exposure data enable the generation of valid estimates for exposure-disease associations. These estimates have been corroborated by findings from more demographically representative cohorts47,48.”

References:

Batty, G. D., Gale, C. R., Kivimäki, M., et al. Comparison of risk factor associations in UK Biobank against representative, general population based studies with conventional response rates: prospective cohort study and individual participant meta-analysis. BMJ. 2020; 368: m131.

Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of Sociodemographic and Health-Related Characteristics of UK Biobank Participants With Those of the General Population. Am J Epidemiol. 2017;186(9):1026–34.

(2) The study sample included different ancestries which may introduce confounding from genetic background. As over 90% of the participants were of European ancestry, I recommend excluding individuals of non-European ancestry in the main analysis.

Thank you for raising the concern regarding the inclusion of different ancestries in our study sample and the potential confounding. In our research, we have adhered to the widely accepted practice of including all participants in the study to ensure a comprehensive analysis. Recognizing the predominance of European ancestry within our cohort, which exceeds 90%, we have proactively incorporated ethnicity as a covariate in our statistical models to mitigate confounding influences.

We also considered the feasibility of conducting a stratified analysis for non-European participants. However, the small sample sizes of non-European subgroups do not provide sufficient statistical power to yield reliable or meaningful separate analyses. Consequently, to maintain the integrity and robustness of our findings, we opted to include all participants in the main analysis, adjusting for ethnicity to account for potential confounders.

(3) I noted that a large proportion of participants were excluded due to the lack of data on plasma PUFAs. Were the characteristics of these participants similar to the current analysis sample?

Thank you for raising this very important point. According to UK Biobank, “The EDTA plasma samples were picked randomly and are therefore representative of the 502,543 participants in the full cohort.” (As detailed in Julkunen et al.) Moreover, as noted in our reply to comment #1 above, we performed a sensitivity analysis, examining the baseline characteristics of participants included in our study relative to those omitted due to lack of exposure information.

The results of this analysis are detailed in Additional file 2: Table S13. They demonstrate that the baseline characteristics—such as age, gender, ethnicity, socioeconomic status, and lifestyle habits—are indeed similar between the two groups. This similarity supports the representativeness of our analysis sample and suggests that the exclusion of participants without plasma PUFA data does not introduce a bias that would undermine the validity of our study's findings.

References:

Julkunen H, Cichońska A, Tiainen M, et al. Atlas of plasma NMR biomarkers for health and disease in 118,461 individuals from the UK Biobank. Nat Commun. 2023 Feb 3;14(1):604. doi: 10.1038/s41467-023-36231-7.

(4) The methods section should include a detailed description of the measurement of plasma omega-6/omega-3 fatty acid ratio. It is important to provide information on the analytical techniques used and any quality control measures implemented to ensure the accuracy and reliability of the measurements. Importantly, were repeated measurements done?

Thank you for raising this important point. The details of the metabolomic profiling have been described in previous UK Biobank publications. In this revision, we added a brief description of the measurement process and provided references to previous publications.

Here is what we added in the “Methods- Ascertainment of exposure” section of our manuscript: “Metabolomic profiling of plasma samples was performed with high-throughput nuclear magnetic resonance (NMR) spectroscopy. At the time of this analysis (15 Mar 2023), UK Biobank released the Phase 1 metabolomic dataset, which covered a random selection of 118,461 plasma samples from the baseline recruitment. These samples were collected between 2007 and 2010 and had been stored in −80 °C freezers, while the NMR measurements took place between 2019 and 2020. Detailed descriptions could be found in previous publications about plasma sample preparation, NMR spectroscopy setup, quality control protocols, correction for sample dilution, verification with duplicate samples and internal controls, and comparisons with independent measurements from clinical chemistry assays20-22.”

(5) The analysis of individual PUFAs is not appropriate because plasma levels of these PUFAs, including n-3 PUFAs and n-6 PUFAs, EPA, DHA and AA, are usually correlated. It is hard to differentiate these correlated FAs in Cox model. Whereas the ratio of n-6/n-3 is indeed more comprehensive, and the current analysis demonstrated this ratio as a good marker of mortality. Therefore, the analyses of individual PUFAs can be removed and only focus on the ratio of n-6/n-3.

We resonate with the Reviewer regarding the importance of focusing on the ratio of n-6/n-3. Indeed, the ratio is our focus in this manuscript. We also acknowledge the Reviewer's concern regarding the inclusion of correlated covariates in one statistical model. In that specific analysis, the correlations between omega-3% and omega-6% (r = -0.12) and between DHA% and LA% (r = 0.03) are relatively low. Additionally, we also checked the model for multicollinearity and found that the variance inflation factors (VIFs) were within acceptable ranges. In the fully adjusted model that included omega-3% and omega-6%, all variables had VIFs below 1.13, with omega-3% at a VIF of 1.06 and omega-6% at a VIF of 1.12. Similarly, in the model including DHA% and LA%, all variables also exhibited VIFs under 1.13, with DHA% recording a VIF of 1.07 and LA% a VIF of 1.10. Because DHA is one of omega-3 PUFAs, we did not include them in the same model. We did not include LA and omega-6 in the same model, either. Because the ratio has two components and each component is the sum of multiple individual PUFAs, it is natural to ask which component is more important (e.g., omega-6 or omega-3?), which specific fatty acid is driving the effect of omega-3 PUFAs (e.g., ALA? Or the marine omega-3, EPA and DHA?). We received such feedback frequently when we presented our research previously. Therefore, as an effort to address them, we performed analysis of omega-3, omega-6, DHA, and LA. While we understand the complexities involved in differentiating the effects of individual fatty acids in a Cox model, we believe there is intrinsic value in exploring these relationships further. In our analysis, we have attempted to investigate the effects of individual PUFAs on mortality by including both DHA and LA within the same model due to their low correlation, making adjustments to account for confounding factors (As detailed in Additional file 2: Table S9). Our findings indicate significant inverse associations between both DHA and LA with all-cause, cancer, and cardiovascular disease (CVD) mortality. We agree with the Reviewer that the focus of our manuscript should be the ratio, but also hope the Reviewer will agree with us that keeping the results from individual PUFAs will provide additional useful information to the readers.

(6) The definition of comorbidities (including hypertension, diabetes, and longstanding illness) is vague. Please clarify what diseases longstanding illness includes.

We appreciate the request for clarification regarding the definition of comorbidities in our study, including the categorization of longstanding illness. The information regarding longstanding illnesses was obtained via the Assessment Centre Environment (ACE) touchscreen questionnaire. Participants were asked, "Do you have any long-standing illness, disability, or infirmity?" with the response options being “Yes,” “No,” “Do not know,” and “Prefer not to answer.” For the purposes of our analysis, participants who selected “Yes” were categorized as having a longstanding illness, while the remaining options were grouped as not having a longstanding illness.

This method of classification aligns with our detailed explanation in the “Methods-Ascertainment of covariates” section of the manuscript, where we state that “Comorbidities, including hypertension, diabetes, and longstanding illness, were self-reported at baseline. Longstanding illness refers to any long-standing illness, disability, or infirmity, without other specific information.” It is important to note that this approach is consistent with established precedents in the field. Specifically, the paper by Li et al. in the BMJ utilized a similar definition for comorbidities, reinforcing the validity of our methodology.

References:

Li ZH, Zhong WF, Liu S, et al. Associations of habitual fish oil supplementation with cardiovascular outcomes and all cause mortality: evidence from a large population based cohort study. BMJ. 2020 Mar 4;368:m456.

(7) The rationale of conducting the mediation analysis of blood biomarkers is not given. Since fatty acids can be formed as TG or bound with apolipoproteins in plasma, there is a large overlap of FAs with these biomarkers and thus it is not appropriate to analyze TG, HDL, LDL, and apolipoproteins as mediators.

We are grateful for the insightful feedback regarding the mediation analysis of blood biomarkers. Our mediation analysis aimed to explore the possible biomarkers and biological processes that explain the effects of PUFAs on mortality. Upon reflection, we recognize the complexities introduced by the inherent overlap of fatty acids with different lipid particles and lipid classes in plasma. Considering the potential confounding this overlap presents, and in agreement with your recommendation, we have decided to remove the mediation analyses involving cholesterol, TG, HDL-C, LDL-C, Lp(a), ApoA, and ApoB from our study. We appreciate your guidance on this matter and have updated our manuscript accordingly to reflect these changes.

Reviewer #2 (Recommendations for the Authors):

(1) Analyzing n-3 and n-6 PUFAs separately might be less instructive given the inherent correlations among plasma levels of n-3 PUFAs and n-6 PUFAs. Also, some important specific PUFAs, such as ALA, AA, EPA, etc. were not available in the UK Biobank data though the authors tried to analyze LA and DHA. The n-6/n-3 ratio, as evidenced by the current analysis, offers a more holistic perspective and might be a superior mortality marker. Thus, I recommend shifting the focus solely to this ratio.

Thank you for the thoughtful comment. Reviewer #1 raised a similar point (comment #5 above). We are glad that both reviewers recognized the importance of the omega-6/omega-3 ratio and agreed with us that the ratio should be the focus of the paper. Please also see our more detailed response above. Briefly, our manuscript centered on the ratio, while the supplemental analysis of omega-3%, omega-6%, DHA%, and LA% provided additional useful information. We included omega-3% and omega-6% in the same model because their correlation was relatively low (r = -0.12). We also checked the model for multicollinearity and found that the variance inflation factors (VIFs) for n-3 PUFAs and n-6 PUFAs were within acceptable ranges. In the fully adjusted model that included omega-3% and omega-6%, all variables had VIFs below 1.13, with omega-3% at a VIF of 1.06 and omega-6% at a VIF of 1.12. Similarly, in the model including DHA% and LA%, all variables also exhibited VIFs under 1.13, with DHA% recording a VIF of 1.07 and LA% a VIF of 1.10. Therefore, we decided to keep the content for omega-3 and omega-6 PUFAs. We hope that Reviewer will agree with us that this content only provides additional information to the readers.

(2) It might not be methodologically sound to treat TG, HDL, LDL, and apolipoproteins as mediators. Since the model included comorbidities as covariates, hypercholesteremia and hypertriglyceridemia seemed to have been adjusted in the analysis. Thus, further adjusting these blood biomarkers for mediation analysis which overlapped with comorbidities is redundant.

We appreciate your critical evaluation of our methodological approach. Your point is well-taken, especially in light of the fact that comorbidities such as hypercholesterolemia and hypertriglyceridemia have been accounted for as covariates in our model. This overlap, as you correctly identified, could indeed render the mediation analysis redundant. In concordance with your recommendation, and incorporating the comments of another reviewer, we have now omitted the mediation analysis involving these blood biomarkers from our study. We believe this adjustment strengthens the methodological soundness of our research and are thankful for your contribution to this refinement. We have updated our manuscript to reflect these changes and ensure our analysis remains robust and free from redundancy.

(3) It's imperative to exercise caution when drawing causal conclusions from the observed correlations. The inherent constraints of observational studies, coupled with potential residual confounding or reverse causality, should be acknowledged.

We concur with the caution against implying causality from correlations observed in our study. As such, we have carefully refrained from claiming any causal relationships within our paper. We acknowledge that the term "protective effects" could suggest a causal inference, and we have revised our language to describe these observations as "inverse associations" to more accurately reflect the nature of our findings.

We have also addressed the inherent limitations of observational research in the Discussion section under 'limitations' of our manuscript. There, we recognize that while we have accounted for many confounders, the possibility of residual confounding cannot be entirely excluded. We also agree that reverse causality is a concern in observational studies. To mitigate this, we performed a sensitivity analysis excluding participants who died within the first year of follow-up. The results from this analysis, which are provided in Additional file 2: Table S12, show consistency with our main findings, suggesting that the observed associations are less likely to be predominantly driven by reverse causation. We are grateful for your insights, which have guided us in strengthening our manuscript and ensuring that our conclusions are presented with the appropriate scientific rigor.

(4) To guide subsequent scholarly endeavors, the manuscript might propose potential research trajectories, such as spearheading randomized controlled trials to delve deeper into the causal nexus between plasma omega-6/omega-3 ratios and mortality outcomes or probing the mechanistic underpinnings of the observed correlations.

We agree that conducting randomized controlled trials could illuminate the potential causal relationships between plasma PUFA biomarkers and mortality outcomes. While the primary focus of our manuscript is to report on associations, we acknowledge the importance of causal analysis in advancing the field. In our secondary analysis, we touched upon mediation effects of blood biomarkers, which could serve as a preliminary step towards establishing causality. Although our current work did not delve deeply into causal mechanisms, the results we have presented may indeed stimulate further exploration. By reporting our mediation analysis results, we aim to provide a foundation that other researchers might build upon. We hope that our work will act as a catalyst for more in-depth studies, such as RCTs or mechanistic investigations, to pursue the questions we have begun to explore.

Following this recommendation, we have revised our Conclusion paragraph and added: “Our findings support the active management of a high circulating level of omega-3 fatty acids and a low omega-6/omega-3 ratio to prevent premature death. Future research is warranted to further test the causality, such as Mendelian randomization and randomized controlled trials. Mechanistic research, including comprehensive mediation analysis, in-depth experimental characterization in animal models or cell lines, and intervention studies, is also needed to unravel the molecular and physiological underpinnings.”

Reviewer #3 (Recommendations for the Authors):

(1) Line 32. Delete "a balanced" because a balanced o6:o3 cannot be defined.

Thank you for pointing out the issue with the term "a balanced". Most authors agree with your observation that defining what constitutes a 'balanced' ratio can be ambiguous and potentially misleading. One author, JTB, disagrees that “balance” as a concept is unacceptably ambiguous or misleading. In response, we have removed the words from our manuscript.

(2) In the abstract you should present the findings for omega-6 and omega-3 PUFAs first and then the findings for the ratio.

We appreciate your suggestion to present the findings for omega-6 and omega-3 PUFAs prior to those for the ratio in the abstract. As laid out in the Background section, the ratio was our primary exposure of interest. So, we organized our manuscript by centering on the ratio. We are glad that both Reviewer #1 and #2 expressed a particular interest in the ratio findings and urged us to keep the ratio as the focus. We believe that this emphasis reflects the novel aspects of our research and aligns with the thematic structure of our manuscript.

(3) Line 80. controversial should read uncertain.

Thank you for the suggestion. We have changed “controversial” to “uncertain”.

(4) It is unclear which fatty acids are included in total PUFAs, omega-6 PUFAs and omega-3 PUFAs. It is vital that this is specified.

Thank you very much for your suggestion. We agree that it is important to clarify the specific fatty acids included in the analysis. In the revised manuscript, we emphasized that we analyzed “total omega-6 PUFAs” and “total omega-3 PUFAs”, while “LA is one type of omega-6 PUFAs” and “DHA is one type of omega-3 PUFAs”. We also revised the Method section of “Ascertainment of exposure” to provide more information about how the exposures were measured. Here is what we added in the “Methods- Ascertainment of exposure” section of our manuscript: “Five PUFAs-related biomarkers were directly measured in absolute concentration units (mmol/L), including total PUFAs, total omega-3 PUFAs, total omega-6 PUFAs, docosahexaenoic acid (DHA), and linoleic acid (LA). Of note, DHA is one type of omega-3 PUFAs, and LA is one type of omega-6 PUFAs. Our primary exposure of interest, the omega-6/omega-3 ratio, was calculated based on their absolute concentrations. We also performed supplemental analysis for four exposures, the percentages of omega-3 PUFAs, omega-6 PUFAs, DHA, and LA in total fatty acids (omega-3%, omega-6%, DHA%, and LA%), which were calculated by dividing their absolute concentrations to that of total fatty acids.”