Reviewer #3 (Public review):

Summary:

This is an interesting investigation of the benefits of perceiving control and its impact on the subjective experience of stress. To assess a subjective sense of control, the authors introduce a novel wheel-stopping (WS) task where control is manipulated via size and speed to induce low and high control conditions. The authors demonstrate that the subjective sense of control is associated with experienced subjective stress and individual differences related to mental health measures. In a second experiment, they further show that an increased sense of control buffers subjective stress induced by a trier social stress manipulation, more so than a more typical stress buffering mechanism of watching neutral/calming videos.

Strengths:

There are several strengths to the manuscript that can be highlighted. For instance, the paper introduces a new paradigm and a clever manipulation to test an important and significant question. Additionally, it is a well-powered investigation that allows for confidence in replicability and the ability to show both high internal consistency and high external validity with an interesting set of individual difference analyses. Finally, the results are quite interesting and support prior literature while also providing a significant contribution to the field with respect to understanding the benefits of perceiving control.

Weaknesses:

There are also some questions that, if addressed, could help our readership.

(1) A key manipulation was the high-intensity stressor (Anticipatory TSST signal), which was measured via subjective ratings recorded on a sliding scale at different intervals during testing. Typically, the TSST conducted in the lab is associated with increases in cortisol assessments and physiological responses (e.g., skin conductance and heart rate). The current study is limited to subjective measures of stress, given the online nature of the study. Since TSST online may also yield psychologically different results than in the lab (i.e., presumably in a comfortable environment, not facing a panel of judges), it would be helpful for the authors to briefly discuss how the subjective results compare with other examples from the literature (either online or in the lab). The question is whether the experienced stress was sufficiently stressful given that it was online and measured via subjective reports. The control condition (low intensity via reading recipes) is helpful, but the low-intensity stress does not seem to differ from baseline readings at the beginning of the experiment.

(2) The neutral videos represent an important condition to contrast with WS, but it raises two questions. First, the conditions are quite different in terms of experience, and it is interesting to consider what another more active (but not controlled per se) condition would be in comparison to the WS performance. That is, there is no instrumental action during the neutral video viewing (even passive ratings about the video), and the active demands could be an important component of the ability to mitigate stress. Second, the subjective ratings of the stress of the neutral video appear equivalent to the win condition. Would it have been useful to have a high arousal video (akin to the loss condition) to test the idea that experience of control will buffer against stress? That way, the subjective stress experience of stress would start at equivalent points after WS3.

(3) For the stress relief analysis, the authors included time points 2 and 3 (after the stressor and debrief) but not a baseline reading before stress. Given the potential baseline differences across conditions, can this decision be justified in the manuscript?

(4) Is the increased control experience during the losses condition more valuable in mitigating experienced stress than the win condition?

(5) The subjective measure of control ("how in control do you feel right now") tends to follow a successful or failed attempt at the WS task. How much is the experience of control mediated by the degree of experienced success/schedule of reinforcement? Is it an assessment of control or, an evaluation of how well they are doing and/or resolution of uncertainty? An interesting paper by Cockburn et al. 2014 highlights the potential for positive prediction errors to enhance the desire for control.

(6) While the authors do a very good job in their inclusion and synthesis of the relevant literature, they could also amplify some discussion in specific areas. For example, operationalizing task controllability via task difficulty is an interesting approach. It would be useful to discuss their approach (along with any others in the literature that have used it) and compare it to other typically used paradigms measuring control via presence or absence of choice, as mentioned by the authors briefly in the introduction.

(7) The paper is well-written. However, it would be useful to expand on Figure 1 to include a) separate figures for study 1 (currently not included) and 2, and b) a timeline that includes the measurements of subjective stress (incorporated in Figure 1). It would also be helpful to include Figure S4 in the manuscript.

Author response:

The following is the authors’ response to the original reviews

Reviewer #1 (Public review):

Summary:

Busch and Hansel present a morphological and histological comparison between mouse and human Purkinje cells (PCs) in the cerebellum. The study reveals species- specific differences that have not previously been reported despite numerous observations of these species. While mouse PCs show morphological heterogeneity and occasional multi-innervation by climbing fibers (CFs), human PCs exhibit a widespread, multi-dendritic structure that exceeds expectations based on allometric scaling. Specifically, human PCs are significantly larger, and exhibit increased spine density, with a unique cluster-like morphology not found in mice.

Strengths:

The manuscript provides an exceptionally detailed analysis of PC morphology across species, surpassing any prior publication. Major strengths include a systematic and thorough methodology, rigorous data analysis, and clear presentation of results. This work is likely to become the go-to resource for quantitation in this field. The authors have largely achieved their aims, with the results effectively supporting their conclusions.

We are grateful to this reviewer for their thoughtful assessment that this work will be a go-to resource for the field.

Weaknesses:

There are a few concerns that need to be addressed, specifically related to details of the methodology as well as data interpretation based on the limits of some experimental approaches. Overall, these weaknesses are minor.

We thank this reviewer for their careful reading of the manuscript and for highlighting limitations and weaknesses in the methodology. We are in full agreement that while interpretation is somewhat limited, there is still value in their description. As detailed below in response to this reviewer’s recommendations, we provide more description of our imaging resolution. This additional detail clarifies that our quantitation is appropriate for the scale of the objects being measured and provides critical information to help readers assess the findings as they may pertain to their own work.

Reviewer #2 (Public review):

Summary:

This manuscript aims to follow up on a previously published paper (Busch and Hansel 2023) which proposed that the morphological variation of dendritic bifurcation in Purkinje cells in mice and humans is indicative of the number of climbing fiber inputs, with dendritic bifurcation at the level of the soma resulting in a proportion of these neurons being multi-innervated. The functional and anatomical climbing fiber data was obtained solely from mice since all human tissue was embalmed and fixed, and the extension of these findings to human Purkinje cells was indirect. The current comparative anatomy study aims to resolve this question in human tissue more directly and to further analyse in detail the properties of adult human Purkinje cell dendritic morphology.

Strengths:

The authors have carried out a meticulous anatomical quantification of human Purkinje cell dendrites, in tissue preparations with a better signal-to-noise ratio than their previous study, comparing them with those from mice. Importantly, they now present immunolabelling results that trace climbing fiber axons innervating human PCs. As well as providing detailed analyses of spine properties and interesting new findings of human PC dendritic length and spine types, the work confirms that human PCs that have two clearly distinct dendritic branches have an approximately x% chance of receiving more than one CF input, segregated across the two branches. Albeit entirely observational, the data will be of widespread interest to the cerebellar field, in particular, those building computational models of Purkinje cells.

We thank this reviewer for their positive and considered assessment of our work. We enthusiastically agree that while these data are descriptive in nature, they may be of interest across modalities of cerebellar research and will provide a more detailed framework for cross-species comparisons and single cell computational modeling, which remains a critical tool to explore the human case given the inaccessibility of physiological experimentation.

Weaknesses:

The work is, by necessity, purely anatomical. It remains to be seen whether there are any functional differences in ion channel expression or functional mapping of granule inputs to human PCs compared with the mouse that might mitigate the major differences in electronic properties suggested.

We are in full agreement with the reviewer that the focused anatomical description of this manuscript could not make strong assertions about function given that cellular and circuit physiology is determined by many additional factors that remain unexamined. We appreciate that the reviewer acknowledges that this is out of necessity as those factors are inaccessible to experimentation at the current time; however, we are enthusiastic that our current findings will motivate future work that will shed light on these critical additional features of the system, both in rodents and humans.

Reviewer 1 (Recommendations for the authors):

PCs are now known to be genetically diverse, with unique PC types found only in humans. Could this cellular diversity contribute to the differences observed between species in this study? This possibility should be at least discussed in the context of the findings.

We agree that this is a fascinating possibility. The perhaps most detailed recent study (Sepp et al., Nature 625, 2024) – in a conservative assessment – describes four developmental PC subtypes in mice that are identical in humans. The study points out that the subtype ratio changes over the course of development, though. Taken together with the possibility of additional human-specific subtypes, a genetic basis for morphological as well as physiological diversity arises. This is now discussed on p. 7. It needs to be kept in mind, however, that other factors, such as push-pull influences during tissue growth, might also play a role.

The human tissue used in this study was obtained from elderly individuals, while the mouse tissue was not. It is unclear whether the age difference might influence the findings, and this warrants further discussion or control.

We share this concern, in particular regarding the spine / spine cluster analysis as here tissue quality and or degenerative effects might play a role. We additionally analyzed a tissue sample from a 37 year-old human, and observed the same spine clusters as in the other human brains. This is now described on p. 4 of the revised manuscript.

The study includes spine size comparisons, but it is not clear if the point spread function (PSF) of the microscope provides the necessary resolution for these quantitative assessments. For instance, are multi-headed spines truly multi-headed, or could this be an artifact of limited resolution?

This is an important point. We addressed it by calculating the Rayleigh limit (more conservative than the Abbe limit) as 248.4nm for the equipment and conditions used (Methods, p. 22). On pages 3-5, we updated our Results section accordingly to point out what quantifications are well supported and discuss the limitations (p. 3-5).

Reviewer 2 (Recommendations for the authors):

This is nice work which must have been very time-consuming. It would be good to make sure that the technical details are properly discussed, to quantify the data properly. Please include details of how you measured the resolution of the microscope used to evaluate spine size.

See our response to the last comment of Referee 1 above.

The figure panels are mostly satisfactory, but they are exceptionally crowded and will probably be difficult to read at the final size. Some work tidying these would be worth it. In Figure 3B, include mention of open and blue triangles in legend. In 3E, the dendritic branches are shown at a different gray scale. You have not done this elsewhere, so probably good to mention it in the legend.

Figure 3 and its legend have been updated / improved accordingly.

The definition of horizontal and vertical is not absolutely clear. Perhaps re-assess this bit of the text. Does it mean that you did not include cells that were neither vertical nor horizontal?

We categorized those PCs as ‘vertical’ that have a >30° angle relative to the PC layer, and those as ‘horizontal’ that have a <30° angle relative to the PC layer. All PCs are covered by these categories. This is now described on p. 5.

Dried neat's tongue is very plainly cow or oxen tongue. It is known to have unique texture, moderate fat content, and high protein content and nutritional value. It is a delicacy in some countries and most popular in Japan. The tongue often weighs 2-4 pounds and is incredibly tender. I personally don't think I could bring myself to eat it no matter the nutritional value my stomach is turning just writing about it.

https://japan-food.guide/en/articles/gyutan https://www.cookinghub.com/recipe-ingredient/beef-tongue/

it's our default for simple configuration, correct. but in advanced config there are 2 more sizes, not sure if it should be mentioned here

1. I would mark B2s as very beginner level.
2. should not we mention other sizes from Bug 819403

Reviewer #2 (Public review):

Summary:

In this manuscript, Li and colleagues study the fate of endothelial cells in a mouse model of ischemic stroke. Using genetic lineage tracing approaches, they found that endothelial cells give rise to non-endothelial cells, which they term "E-pericytes." They further show that depleting these cells exacerbates blood-brain barrier leakage and worsens functional recovery. The authors also provide evidence that endothelial-to-mesenchymal transition, myeloid cell-derived TGFβ1, and endothelial TGFβRII are involved in this process. These are potentially interesting findings, however, the experimental evidence that endothelial cells undergo transdifferentiation to non-endothelial cells is weak, as is the evidence that these cells are pericytes. Addressing this foundational weakness will facilitate the interpretation of the other findings.

Strengths:

(1) The authors address an important question about blood vessel function and plasticity in the context of stroke.

(2) The authors use a variety of genetic approaches to understand cell fate in the context of stroke. Particularly commendable is the use of several complementary lineage tracing strategies, including an intersectional strategy requiring both endothelial Cre activity and subsequent mural cell NG2 promoter activity.

(3) The authors address upstream cellular and molecular mechanisms, including roles for myeloid-derived TGFβ.

Weaknesses:

(1) The authors use Cdh5-CreERT2; Ai47 mice to permanently label endothelial cells and their progeny with eGFP. They then isolate eGFP+ cells from control and MCAO RP7D and RP34D brains, and use single-cell RNA-seq to identify the resulting cell types. Theoretically, all eGFP+ cells should be endothelial cells or their progeny. This is a very powerful and well-conceived experiment. The authors use the presence of a pericyte cluster as evidence that endothelial-to-pericyte transdifferentiation occurs. However, pericytes are also present in the scRNA-seq data from sham mice, as are several other cell types such as fibroblasts and microglia. This suggests that pericytes and these other cell types might have been co-purified (e.g., as doublets) with eGFP+ endothelial cells during FACS and may not themselves be eGFP+. Pericyte-endothelial doublets are common in scRNA-seq given that these cell types are closely and tightly associated. Additionally, tight association (e.g., via peg-socket junctions) can cause fragments of endothelial cells to be retained on pericytes (and vice-versa) during dissociation. Finally, it is possible that after stroke or during the dissociation process, endothelial cells lyse and release eGFP that could be taken up by other cell types. All of these scenarios could lead to the purification of cells that were not derived (transdifferentiated) from endothelial cells. The authors note that the proportion of pericytes increased in the stroke groups, but it does not appear this experiment was replicated and thus this conclusion is not supported by statistical analysis. The results of pseudotime and trajectory analyses rely on the foundation that the pericytes in this dataset are endothelial-derived, which, as discussed above, has not been rigorously demonstrated.

(2) I have the same concern regarding the inadvertent purification of cells that were not derived from endothelial cells in the context of the bulk RNA-seq experiment (Figure S4), especially given the sample-to-sample variability in gene expression in the RP34D, eGFP+ non-ECs-group (e.g., only 2/5 samples are enriched for mesenchymal transcription factor Tbx18, only 1/5 samples are enriched for mural cell TF Heyl). If the sorted eGFP+ non-ECs were pericytes, I would expect a strong and consistent pericyte-like gene expression profile.

(3) The authors use immunohistochemistry to understand localization, morphology, and marker expression of eGFP+ cells in situ. The representative "E-pericytes" shown in Figure 3A-D are not associated with blood vessels, and the authors' quantification also shows that the majority of such cells are not vessel-associated ("avascular"). By definition, pericytes are a component of blood vessels and are embedded within the vascular basement membrane. Thus, concluding that these cells are pericytes ("E-pericytes") may be erroneous.

(4) CD13 flow cytometry and immunohistochemistry are used extensively to identify pericytes. In the context of several complementary lineage tracing strategies noted in Strength #2, CD13 immunohistochemistry is the only marker used to identify putative pericytes (Figure S3J-M). In stroke, CD13 is not specific to pericytes; dendritic cells and other monocyte-derived cells express CD13 (Anpep) in mouse brain after stroke (PMID: 38177281, https://anratherlab.shinyapps.io/strokevis/).

(5) The authors conclude that "EC-specific overexpression of the Tgfbr2 protein by a virus (Tgfbr2) decreases Evans blue leakage, promotes CBF recovery, alleviates neurological deficits and facilitates spontaneous behavioral recovery after stroke by increasing the number of E-pericytes." All data in Figure 10, however, compare endothelial Tgfbr2 overexpression to a DsRed overexpression control. There is no group in which Tgfbr2 is overexpressed but "E-pericytes" are eliminated with DTA (this is done in Figure 9B, but this experiment lacks the Tgfbr2 overexpression-only control). Thus, the observed functional outcomes cannot be ascribed to "E-pericytes"; it remains possible that endothelial Tgfbr2 overexpression affects EB leakage, CBF, and behavior through alternative mechanisms.

(6) Single-cell and bulk RNA-seq data are not available in a public repository (such as GEO). Depositing these data would facilitate their independent reevaluation and reuse.

1. comment 1 comment 2

Fossil es un sistema SCM distribuido, simple y de alta confiabilidad con estas características avanzadas: Es una interfaz web integrada y que permite ser personalizable como son las wiki , documentación integridad. Al clonar Fossil desde uno de sus repositorios autoalojados , obtiene más que solo el código fuente: obtiene este sitio web completo. 1. Interfaz web integrada 2. Gestión de Proyectos 3. Todo en uno 4. Sincronización automática 5. Libre de código abierto 6. Fácil de almacenar (autoalojar)

Reviewer #1 (Public review):

Summary:

Mazer & Yovel 2025 dissect the inverse problem of how echolocators in groups manage to navigate their surroundings despite intense jamming using computational simulations.

The authors show that despite the 'noisy' sensory environments that echolocating groups present, agents can still access some amount of echo-related information and use it to navigate their local environment. It is known that echolocating bats have strong small and large-scale spatial memory that plays an important role for individuals. The results from this paper also point to the potential importance of an even lower-level, short-term role of memory in the form of echo 'integration' across multiple calls, despite the unpredictability of echo detection in groups. The paper generates a useful basis to think about the mechanisms in echolocating groups for experimental investigations too.

Strengths:

(1) The paper builds on biologically well-motivated and parametrised 2D acoustics and sensory simulation setup to investigate the various key parameters of interest

(2) The 'null-model' of echolocators not being able to tell apart objects & conspecifics while echolocating still shows agents successfully emerge from groups - even though the probability of emergence drops severely in comparison to cognitively more 'capable' agents. This is nonetheless an important result showing the direction-of-arrival of a sound itself is the 'minimum' set of ingredients needed for echolocators navigating their environment.

(3) The results generate an important basis in unraveling how agents may navigate in sensorially noisy environments with a lot of irrelevant and very few relevant cues.

(4) The 2D simulation framework is simple and computationally tractable enough to perform multiple runs to investigate many variables - while also remaining true to the aim of the investigation.

Weaknesses:

There are a few places in the paper that can be misunderstood or don't provide complete details. Here is a selection:

(1) Line 61: '... studies have focused on movement algorithms while overlooking the sensory challenges involved' : This statement does not match the recent state of the literature. While the previous models may have had the assumption that all neighbours can be detected, there are models that specifically study the role of limited interaction arising from a potential inability to track all neighbours due to occlusion, and the effect of responding to only one/few neighbours at a time e.g. Bode et al. 2011 R. Soc. Interface, Rosenthal et al. 2015 PNAS, Jhawar et al. 2020 Nature Physics.

(2) The word 'interference' is used loosely places (Line 89: '...took all interference signals...', Line 319: 'spatial interference') - this is confusing as it is not clear whether the authors refer to interference in the physics/acoustics sense, or broadly speaking as a synonym for reflections and/or jamming.

(3) The paper discusses original results without reference to how they were obtained or what was done. The lack of detail here must be considered while interpreting the Discussion e.g. Line 302 ('our model suggests...increasing the call-rate..' - no clear mention of how/where call-rate was varied) & Line 323 '..no benefit beyond a certain level..' - also no clear mention of how/where call-level was manipulated in the simulations.

“野狐禅”（yěhúchán）是禅宗佛教中的一个术语，带有贬义，用来指那些似是而非、不正统、肤浅或误导的禅法或禅悟。它比喻那些没有真正领悟禅宗精髓，却自以为是、以讹传讹，甚至利用禅宗来达到个人目的的人或教法。

要详细介绍“野狐禅”，需要从其字面意义、典故来源、具体特征以及在禅宗语境中的含义等方面进行阐述：

1. 字面意义：

• 野狐 (yěhú): 野生的狐狸。在中国的传统文化中，狐狸常常带有狡猾、诡计多端、善于迷惑人的形象。有时也与妖魅、邪气联系在一起。
• 禅 (chán): 指禅那，是佛教的一种修行方法，强调通过静坐、冥想等方式达到内心的平静和智慧的觉悟。

因此，从字面上理解，“野狐禅”就是指像野狐一样狡猾、不正统的禅法。

2. 典故来源：

“野狐禅”最著名的典故出自禅宗公案集《碧岩录》中的一个故事，讲述的是百丈怀海禅师（749-814）说法时，一位老人在法会结束后没有离开，而是留下来请教。老人自称过去曾是一位住持，有人问他：“大修行底人，还落因果也无？”（大修行的人，还受因果报应吗？）他回答说：“不落因果。”（不落因果。）结果堕入野狐身五百世。

老人恳请百丈禅师为他开示。百丈禅师便在下次说法时，故意问了同样的问题。老人（此时已恢复人形）回答说：“不昧因果。”（不昧因果。）

这个故事的核心在于对“落因果”和“不落因果”的理解。“不落因果”容易被误解为修行高深的人可以超越因果律，而“不昧因果”则强调即使是修行人也要清楚明白因果的道理，但不会被因果所束缚。

这个典故成为了“野狐禅”的经典来源，指那些似是而非、断章取义、容易让人产生错误理解的禅法。那位老人因为一字之差，错误理解了禅理，导致了五百世的野狐之身，警示后人修行禅法要谨慎，避免落入歧途。

3. “野狐禅”的具体特征：

“野狐禅”通常表现出以下一些特征：

• 断章取义，片面理解： 截取禅宗经典或语录中的片段，不结合整体语境进行理解，导致意义的扭曲。
• 执着于文字，不悟真谛： 停留在语言文字的表面，不能透过文字直指禅宗所强调的“心性”或“佛性”。
• 以讹传讹，误导他人： 自己没有真正领悟，却喜欢高谈阔论，传播一些不正确的禅法，误导他人修行。
• 追求神通，故弄玄虚： 偏离禅宗的根本宗旨，追求一些神秘的体验或神通，甚至用一些玄虚的言语来迷惑信众。
• 功利主义，服务私欲： 将禅宗作为达到个人目的的工具，例如追求名利、地位等，失去了禅宗的纯粹性。
• 缺乏实修实证： 只是在口头上谈论禅理，缺乏真正的修行实践和内心的体验证悟。
• 傲慢自大，排斥正法： 认为自己已经领悟了高深的禅理，听不进他人的正确指导，甚至排斥正统的禅法。

4. “野狐禅”在禅宗语境中的含义：

在禅宗中，“野狐禅”是一个非常严肃的批评，它告诫修行者要避免走入歧途，要以正确的态度和方法来学习和实践禅法。真正的禅宗强调的是直指人心、见性成佛，是通过自身的实践体验来获得智慧和解脱，而不是依赖于肤浅的理解或外在的形式。

“野狐禅”与“正禅”（zhèngchán，正统的禅）相对。正禅强调的是师资传承、如法修行、实证体验和智慧的生发。一个真正的禅者应该具备深刻的智慧、慈悲的心怀和清净的行为。

5. 现代语境下的“野狐禅”：

即使在现代社会，“野狐禅”这个概念仍然具有现实意义。它可以用来批评那些：

• 仅仅将禅宗视为一种哲学或心理学，而忽略其宗教和修行本质的人。
• 打着禅宗的旗号，进行商业炒作或个人营销的人。
• 片面理解禅宗的“空”、“无”等概念，将其作为逃避现实或放纵自我的借口的人。
• 热衷于谈论“觉悟”、“开悟”等词汇，但缺乏实际行动和内心改变的人。

总结：

“野狐禅”是禅宗中一个重要的警示性概念，它提醒修行者要警惕那些似是而非、不正统的禅法，强调真正的禅修需要深入的实践、正确的理解和师父的指导。它告诫人们不要停留在表面的理解和空洞的言辞上，而要通过真实的修行来体悟禅宗的真谛。这个概念也提醒我们在学习任何知识或技能时，都要注重其本质和深度，避免浅尝辄止，甚至被一些虚假的表象所迷惑。

Author response:

The following is the authors’ response to the original reviews

eLife Assessment

In this valuable study, García-Vázquez et al. provide solid evidence suggesting that G2 and S phases expressed protein 1 (GTSE1), is a previously unappreciated non-pocket substrate of cyclin D1-CDK4/6 kinases. To this end, this study holds a promise to significantly contribute to an improved understanding of the mechanisms underpinning cell cycle progression. Notwithstanding these clear strengths of the article, it was thought that the study may benefit from establishing the precise role of cyclin D1-CDK4/6 kinase-dependent GTSE1 phosphorylation in the context of cell cycle progression, …

We do not claim, as editors and reviewers appear to have interpreted, that GTSE1 is phosphorylated by cyclin D1-CDK4 in the G1 phase of the cell cycle under normal physiologic conditions.  Indeed, we agree with the existing literature indicating that in cells that do not express high levels of cyclin D1, GTSE1 is expressed predominantly during S and G2 phase (hence the name GTSE1, which stands for G-Two and S phases expressed protein 1) and is phosphorylated by mitotic cyclins in early mitosis.  Even during G1, when the levels of cyclin D1 peak, GTSE1 is not phosphorylated in normal cells.  This could be due to either a higher affinity between GTSE1 and mitotic cyclins as compared to D-type cyclins or to a higher concentration of mitotic cyclins compared to D-type cyclins.  In the current manuscript, we show that higher levels of cyclin D1 can drive the sustained phosphorylation of GTSE1 across all cell cycle points. To reach this conclusion, we do not rely only on the overexpression of exogenous cyclin D1. In fact, we observe similar effect when we deplete endogenous AMBRA1, resulting in the stabilization of endogenous cyclin D1 in all cell cycle phases (see Figure 2G and Figure supplement 3B).  As we had already mentioned in the Discussion section, we propose that GTSE1 is phosphorylated by CDK4 and CDK6 particularly in pathological states, such as cancers displaying overexpression of D-type cyclins (i.e., it is possible that the overexpression overcomes the lower affinity of the cyclin D-GTSE1 complex). In turn, phosphorylation of GTSE1 induces its stabilization, leading to increased levels that, as expected based on the existing literature, contribute to enhanced cell proliferation.  So, the role of the cyclin D1-CDK4/6 kinase-dependent GTSE1 phosphorylation is to stabilize GTSE1 independently of the cell cycle.  In sum, our study suggests that overexpression of cyclin D1, which is often observed in cancers cells beyond the G1 phase, induces phosphorylation of GTSE1 at all points in the cell cycle. 

… obtaining more direct evidence that cyclin D1-CDK4/6 kinase phosphorylate indicated sites on GTSE1 (e.g., S454) …

We show that treatment of cells with palbociclib completely abolished the effect of cyclin D1-CDK4 on the GTSE1 shift observed using Phos-tag gels (Figure 2H).  Moreover, mutagenesis analysis shows that S91, S262, and S724 are phosphorylated in a cyclin D1-CDK4-dependent manner (Figure 2F and Figure supplement 3A). Compared to wild-type GTSE1, a triple mutant (S91A/S262A/S724A) displayed loss of slower-migrating bands upon co-expression of cyclin D1-CDK4, suggesting diminished phosphorylation. Nevertheless, a residual slow-migrating band persisted, prompting further mutations of the triple GTSE1 mutant in S331 and S454 (individually), which do not have a CDK-phosphorylation consensus, but were identified in several published phospho-proteomics studies. From these two quadruple mutants, only the that containing the S454A mutation demonstrated a complete abrogation of any shift in phos-tagTM gels (Figure 2F). These studies suggest that four major sites (S91, S262, S454, and S724) are phosphorylated (either directly and/or indirectly) in a cyclin D1-CDK4-dependent manner.

… and mapping a degron in GTSE1 whose function may be blocked by cyclin D1-CDK4/6 kinase-dependent phosphorylation.

We show that stabilization or overexpression of cyclin D1, which is often observed in human cancers, promotes GTSE1 phosphorylation on S91, S262, S454, and S724, resulting in GTSE1 stabilization.  Similarly, a phospho-mimicking mutant with the 4 serine residues replaced with an aspartate at positions 91, 261, 454, and 724 display increased half-life. While we appreciate the editor’s suggestion and agree on these being interesting questions, we would like to respectfully point out that mapping the GTSE1 degron and understanding how it is affected by cyclin D1-CDK4/6-dependent phosphorylation is outside the scope of the current project and will require an extensive set of experiments and tools. Accordingly, the three reviewers did not ask to map the GTSE1 degron.  We plan on addressing these interesting questions as part of a follow-up study.

Reviewer #1 (public review):

Summary:

García-Vázquez et al. identify GTSE1 as a novel target of the cyclin D1-CDK4/6 kinases. The authors show that GTSE1 is phosphorylated at four distinct serine residues and that this phosphorylation stabilizes GTSE1 protein levels to promote proliferation.

Strengths:

The authors support their findings with several previously published results, including databases. In addition, the authors perform a wide range of experiments to support their findings.

Weaknesses:

I feel that important controls and considerations in the context of the cell cycle are missing. Cyclin D1 overexpression, Palbociclib treatment and apparently also AMBRA1 depletion can lead to major changes in cell cycle distribution, which could strongly influence many of the observed effects on the cell cycle protein GTSE1. It is therefore important that the authors assess such changes and normalize their results accordingly.

We have approached the question of GTSE1 phosphorylation to account for potential cell cycle effects from multiple angles: 

(i) We conducted in vitro experiments with purified, recombinant proteins and shown that GTSE1 is phosphorylated by cyclin D1-CDK4 in a cell-free system (Figure 2A-C). These experiments provide direct evidence of GTSE1 phosphorylation by cyclin D1-CDK4 without the influence of any other cell cycle effectors. 

(ii) We present data using synchronized AMBRA1 KO cells (new Figure 2G and Figure supplement 3B).  In agreement with what we had shown previously (Simoneschi et al., Nature 2021, PMC8875297), AMBRA1 KO cells progress faster in the cell cycle but they are still synchronized as shown, for example, by the mitotic phosphorylation of Histone H3, peaking at 32 hours after serum readdition like in parental cells. Under these conditions we observed that while phosphorylation of GTSE1 in parental cells is evident in the last two time points, AMBRA1 KO cells exhibited sustained phosphorylation of GTSE1 across all cell cycle phases.  This was evident enough when using Phos-tag gels as in the top panel of the old Figure 2G. We now re-run one the biological triplicates of the synchronized cells using higher concentration of Zn⁺²-Phos-tag reagent and lower voltage to allow better separation of the phosphorylated bands.  Under these conditions, GTSE1 phosphorylation is better appreciable (top panel of the new Figure 2G). This experiment provides evidence that high levels of cyclin D1 in AMBRA1 KO cells affect GTSE1 phosphorylation independently of the specific points in the cell cycle. 

(iii) The relative short half-life of GTSE1 (<4 hours) makes its levels sensitive to acute treatments such as Palbociclib or acute AMBRA1 depletion. The effects of these treatments on GTSE1 levels are measurable within a time frame too short to significantly affect cell cycle progression. For example, we used cells with fusion of endogenous AMBRA1 to a mini-Auxin Inducible Degron (mAID) at the N-terminus. This system allows for rapid and inducible degradation of AMBRA1 upon addition of auxin, thereby minimizing compensatory cellular rewiring. Again, we observed an increase in GTSE1 levels upon acute ablation of AMBRA1 (i.e., in 8 hours) (Figure 3B), when no significant effects on cell cycle distribution are observed (please see Simoneschi et al., Nature 2021, PMC8875297 and Rona et al., Mol. Cell 2024, PMC10997477).

Altogether, the above lines of evidence support our conclusion that GTSE1 is a target of cyclin D1-CDK4, independent of cell cycle effects.

In conclusion, we do not claim that GTSE1 is phosphorylated by cyclin D1-CDK4 in the G1 phase of the cell cycle under normal physiologic conditions.  Indeed, we agree with the existing literature indicating that in cells that do not express high levels of cyclin D1, GTSE1 is expressed predominantly during S and G2 phase (hence the name GTSE1, which stands for G-Two and S phases expressed protein 1) and is phosphorylated by mitotic cyclins in early mitosis.  Even during G1, when the levels of cyclin D1 peak, GTSE1 is not phosphorylated in normal cells. This could be due to either a higher affinity between GTSE1 and mitotic cyclins as compared to D-type cyclins or to a higher concentration of mitotic cyclins compared to D-type cyclins.  In the current manuscript, we show that higher levels of cyclin D1 can drive the sustained phosphorylation of GTSE1 across all cell cycle points. To reach this conclusion, we do not rely only on the overexpression of exogenous cyclin D1. In fact, we observe similar effect when we deplete endogenous AMBRA1, resulting in the stabilization of endogenous cyclin D1 in all cell cycle phases (see Figure 2G and Figure supplement 3B).  As we had already mentioned in the Discussion section of the original submission, we propose that GTSE1 is phosphorylated by CDK4 and CDK6 particularly in pathological states, such as cancers displaying overexpression of D-type cyclins (i.e., it is possible that the overexpression overcomes the lower affinity of the cyclin D1-GTSE1 complex). In turn, phosphorylation of GTSE1 induces its stabilization, leading to increased levels that, as expected based on the existing literature, contribute to enhanced cell proliferation.  In sum, our study suggests that overexpression of cyclin D1, which is often observed in cancers cells beyond the G1 phase, induces phosphorylation of GTSE1 at all points in the cell cycle.    

Reviewer #2 (public review):

Summary:

The manuscript by García-Vázquez et al identifies the G2 and S phases expressed protein 1(GTSE1) as a substrate of the CycD-CDK4/6 complex. CycD-CDK4/6 is a key regulator of the G1/S cell cycle restriction point, which commits cells to enter a new cell cycle. This kinase is also an important therapeutic cancer target by approved drugs including Palbocyclib. Identification of substrates of CycD-CDK4/6 can therefore provide insights into cell cycle regulation and the mechanism of action of cancer therapeutics. A previous study identified GTSE1 as a target of CycB-Cdk1 but this appears to be the first study to address the phosphorylation of the protein by Cdk4/6.

The authors identified GTSE1 by mining an existing proteomic dataset that is elevated in AMBRA1 knockout cells. The AMBRA1 complex normally targets D cyclins for degradation. From this list, they then identified proteins that contain a CDK4/6 consensus phosphorylation site and were responsive to treatment with Palbocyclib.

The authors show CycD-CDK4/6 overexpression induces a shift in GTSE1 on phostag gels that can be reversed by Palbocyclib. In vitro kinase assays also showed phosphorylation by CDK4. The phosphorylation sites were then identified by mutagenizing the predicted sites and phostag got to see which eliminated the shift.

The authors go on to show that phosphorylation of GTSE1 affects the steady state level of the protein. Moreover, they show that expression and phosphorylation of GTSE1 confer a growth advantage on tumor cells and correlate with poor prognosis in patients.

Strengths:

The biochemical and mutagenesis evidence presented convincingly show that the GTSE1 protein is indeed a target of the CycD-CDK4 kinase. The follow-up experiments begin to show that the phosphorylation state of the protein affects function and has an impact on patient outcomes.

Weaknesses:

It is not clear at which stage in the cell cycle GTSE1 is being phosphorylated and how this is affecting the cell cycle. Considering that the protein is also phosphorylated during mitosis by CycB-Cdk1, it is unclear which phosphorylation events may be regulating the protein.

Please see point (ii) and the last paragraph in the response to Reviewer #1.  Moreover, we show that, compared to the amino acids phosphorylated by cyclin D1-CDK4, cyclin B1-CDK1 phosphorylates GTSE1 on either additional residues or different sites (Figure 2H). We also show that expression of a phospho-mimicking GTSE1 mutant leads to accelerated growth and an increase in the cell proliferative index (Figure 4B,C and new Figure supplement 4D-E).  Finally, we have evaluated also the cell cycle distributions by flow cytometry (new Figure supplement 4F). These analyses show that the expression of a phospho-mimicking GTSE1 mutant induces a decrease in the percentage of cells in G1 and an increase in the percentage of cells in S, similarly to what observed in AMBRA1 KO cells.

Reviewer #3 (public review)

Summary:

This paper identifies GTSE1 as a potential substrate of cyclin D1-CDK4/6 and shows that GTSE1 correlates with cancer prognosis, probably through an effect on cell proliferation. The main problem is that the phosphorylation analysis relies on the over-expression of cyclin D1. It is unclear if the endogenous cyclin D1 is responsible for any phosphorylation of GTSE1 in vivo, and what, if anything, this moderate amount of GTSE1 phosphorylation does to drive proliferation.

Strengths:

There are few bonafide cyclin D1-Cdk4/6 substrates identified to be important in vivo so GTSE1 represents a potentially important finding for the field. Currently, the only cyclin D1 substrates involved in proliferation are the Rb family proteins.

Weaknesses:

The main weakness is that it is unclear if the endogenous cyclin D1 is responsible for phosphorylating GTSE1 in the G1 phase. For example, in Figure 2G there doesn't seem to be a higher band in the phos-tag gel in the early time points for the parental cells. This experiment could be redone with the addition of palbociclib to the parental to see if there is a reduction in GTSE1 phosphorylation and an increase in the amount in the G1 phase as predicted by the authors' model. The experiments involving palbociclib do not disentangle cell cycle effects. Adding Cdk4 inhibitors will progressively arrest more and more cells in the G1 phase and so there will be a reduction not just in Cdk4 activity but also in Cdk2 and Cdk1 activity. More experiments, like the serum starvation/release in Figure 2G, with synchronized populations of cells would be needed to disentangle the cell cycle effects of palbociclib treatment.   

Please see last paragraph in the response to Reviewer #1.  Concerning the experiments involving palbociclib, we limited confounding effects on the cell cycle by treating cells with palbociclib for only 4-6 hours. Under these conditions, there is simply not enough time for S and G2 cells to arrest in G1.

It is unclear if GTSE1 drives the G1/S transition. Presumably, this is part of the authors' model and should be tested.

We are not claiming that GTSE1 drives the G1/S transition (please see last paragraph in the response to Reviewer #1). GTSE1 is known to promote cell proliferation, but how it performs this task is not well understood.  Our experiments indicate that, when overexpressed, cyclin D1 promotes GTSE1 phosphorylation and its consequent stabilization.  In agreement with the literature, we show that higher levels of GTSE1 promote cell proliferation.  To measure cell cycle distribution upon expressing various forms of GTSE1, we have now performed FACS analyses (new Figure supplement 4F). These analyses show that the expression of a phospho-mimicking GTSE1 mutant induces a decrease in the percentage of cells in G1 and an increase in the percentage of cells in S, similarly to what observed in AMBRA1 KO cells shown in the same panel and in Simoneschi et al. (Nature 2021, PMC8875297).

The proliferation assays need to be more quantitative. Figure 4B should be plotted on a log scale so that the slope can be used to infer the proliferation rate of an exponentially increasing population of cells. Figure 4c should be done with more replicates and error analysis since the effects shown in the lower right-hand panel are modest.

In Figure 4B, we plotted data in a linear scale as done in the past (Donato et al. Nature Cell Biol. 2017, PMC5376241) to better underline the changes in total cell number overtime.  The experiments in Figure 4B were performed in triplicate, statistical significance was determined using unpaired T-tests with p-values<0.05, and error bars represent the mean +/- SEM.  In Figure 4C, error analysis was not included for simplicity, given the complexity of the data.  We have now included the other two sets of experiments (new Figure supplement 4D,E).  While the effects shown in the lower right-hand panel of Figure 4C are modest, they demonstrate the same trend as those observed in the AMBRA KO cells (Figure 4C and Simoneschi et al., Nature 2021, PMC8875297). It's important to note that this effect is achieved through the stable expression of a single phospho-mimicking protein, whereas AMBRA KO cells exhibit changes in numerous cell cycle regulators. Moreover, these effects are obtained by growing cells in culture for only 5 days. A similar impact on cell growth in vivo over an extended period could pose significant risks in the long term.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Figure 1E is referenced before 1D. The authors should consider switching D and E.

Done.

Figure 1D-E: The authors correctly note in the introduction that GTSE1 is encoded by a cell cycle-dependently expressed gene. Given that cell cycle genes are often associated with poor prognosis (e.g., see Whitfield et al., 2006 Nat. Rev. Cancer), this would be expected to correlate with poor prognosis. This should be mentioned in the results section.

We agree that the overexpression of certain (but not all) cell cycle-regulated genes are prognostically unfavorable across various cancer types, and we cited Whitfield et al., 2006 Nat. Rev. Cancer.  However, our data indicate that phosphorylation of GTSE1 induces its stabilization and, consequently, its levels do not oscillate during the cell cycle any longer (new Figure 2G and Figure supplement 3B).  Moreover, analyzing data from the Clinical Proteomic Tumor Analysis Consortium, we observed an enrichment of GTSE1 phospho-peptides (normalized to total protein) within a pan-cancer cohort as opposed to adjacent, corresponding normal tissues (Figure 2I).

Figure 2F: Contrast is too high. Blot images should not contain fully saturated black or white.

We corrected the contrast.

Figure 2G and Figure Supplement 3B: It looks like AMBRA1 KO cells do not synchronize properly in response to serum withdrawal. The cell cycle distribution should be checked by FACS. Otherwise, it is unclear whether changes in GTSE1 (phosphor) levels are only due to indirect changes in the cell cycle distribution.

Synchronization of both parental and AMBRA1 KO cells is demonstrated by the fact that the phosphorylation of Histone H3 peaks at 32 hours after serum readdition in both cases (Figure supplement 3B). 

Figure 2I: It is important that phosphor-GTSE1 levels are normalized to total GTSE1 levels to understand the distinct contribution of changes in GTSE1 levels and from CCND1-CDK4 driven phosphorylation.

Done.

Figure 3A-B: These experiments should also be controlled for cell cycle distribution. Is this effect specific to GTSE1 and other AMBRA1 targets or are other G2/M cell cycle proteins also affected?

The relative short half-life of GTSE1 (<4 hours) makes its levels sensitive to acute treatments such as Palbociclib or acute AMBRA1 depletion. The effects of these treatments on GTSE1 levels are measurable within a time frame too short to significantly affect cell cycle progression. For example, we used cells with fusion of endogenous AMBRA1 to a mini-Auxin Inducible Degron (mAID) at the N-terminus. This system allows for rapid and inducible degradation of AMBRA1 upon addition of auxin, thereby minimizing compensatory cellular rewiring. Again, we observed an increase in GTSE1 levels upon acute ablation of AMBRA1 (i.e., in 8 hours) (Figure 3B), when no significant effects on cell cycle distribution are observed (please see Simoneschi et al., Nature 2021, PMC8875297 and Rona et al., Mol. Cell 2024, PMC10997477).

Figure 4: It should be noted that the correlation with cell proliferation and cell cycle protein expression is expected for any cell cycle protein, including GTSE1.

Actually, the main point of Figure 4 is to show that expression of the phospho-mimicking mutant of GTSE1 promotes cell proliferation. Comparative analysis revealed that cells overexpressing either wild-type GTSE1 or its phospho-deficient form exhibited significantly reduced proliferation rates compared to those expressing the phospho-mimicking mutant (Figure 4B,C). 

The two-decades-old references 33 and 34 are not well suited to support the notion for Cyclin D1 that "the full spectrum of substrates and their impact on cellular function and oncogenesis remain poorly explored." More recent references should be used to show that this is still the case.

We added more recent references.

The authors conclude that their "data indicate that cyclin D1-CDK4 is responsible for the phosphorylation of GTSE1 on four residues (S91, S262, S454, and S724)." However, the authors' data do not exclude a role for their siblings cyclin D2, cyclin D3, and CDK6. Reflecting this, the conclusions should be toned down.

The analysis of the sites phosphorylated in GTSE1 was performed by experimentally co-expressing cyclin D1-CDK4 (Figure 2F, Figure 2H, and Figure supplement 3A), hence our statement.  Yet, we agree that in cells, cyclin D2, cyclin D3, and CDK6 can contribute to GTSE1 phosphorylation. 

The authors claim that they "observed that in human cells, when D-type cyclins are stabilized in the absence of AMBRA1, GTSE1 becomes phosphorylated also in G1." However, the G1-specific data presented by the authors are not controlled for, and it is unclear whether these phosphorylation events actually occur in G1 cells.

We now provide a WB in which GTSE1 phosphorylation is more evident (top panel of the new Figure 2G) (please see point (ii) in the response to the public review of Reviewer #1).  This experiment clearly shows that in AMBRA1 KO cells, GTSE1 is phosphorylated at all points in the cell cycle. Synchronization of both parental and AMBRA1 KO cells is demonstrated by the fact that phosphorylation of Histone H3 peaks at 32 hours after serum re-addition in both cases (Figure supplement 3B). 

Reviewer #2 (Recommendations for the authors):

(1) It is not clear from the presented data at which point in the cell cycle that phosphorylation of GTSE1 may be affecting the steady state level of the protein. The implication that GTSE1 is a target of CycD-CDK4 would suggest that the protein is stabilized at G1/S. Can this effect be observed?

Please see the last paragraph in the response to the public review of Reviewer #1.

(2) Considering the previous study showing that GTSE1 is also phosphorylated during mitosis by CycB-Cdk1, do levels of GTSE1 protein change during the cell cycle? Do changes in GTSE1 levels correlate with phosphorylation during the cell cycle? Cell synchronization experiments such as double thymidine and subsequent phostag analysis could shed some light on these questions.

Please see the last paragraph in the response to the public review of Reviewer #1.

(3) The authors show that the phosphomimetic mutants of GTSE1 confer a growth advantage on cells. The mechanism of this growth advantage is unclear. Is this effect due to a shorter cell cycle, enhanced survival, or another mechanism?

We did not observe increased cell survival when the phosphomimetic mutants of GTSE1 is expressed.  We show that phosphorylation of GTSE1 induces its stabilization, leading to increased levels that, as expected based on the existing literature, contribute to enhanced cell proliferation.  So, the role of the cyclin D1-CDK4/6 kinase-dependent phosphorylation of GTSE1 is to stabilize GTSE1. 

(4) Other minor points - all of the presented immunoblots do not show molecular weight markers. The IF images require scale bars.

To prevent overcrowding of the Figures, the sizes of blotted proteins are indicated in the uncropped scans of each blot. Uncropped scans have been deposited in Mendeley at:  https://data.mendeley.com/datasets/xzkw7hrwjr/1. Scale bars have been added to the IF images.

Reviewer #2 (Public review):

Summary:

The work aims to further understand the role of macrophages in lung precancer/lung cancer evolution

Strengths:

(1) The use of single-cell RNA seq to provide comprehensive characterisation.

(2) Characterisation of cross-talk between macrophages and the lung precancerous cells.

(3) Functional validation of the effects of S100a4+ cells on lung precancerous cells using in vitro assays.

(4) Validation in human tissue samples of lung precancer / invasive lesions.

Weaknesses identified previously:

(1) The authors need to provide clarification of several points in the text.

(2) The authors need to carefully assess their assumptions regarding the role of macrophages in angiogenesis in precancerous lesions.

(3) The authors should discuss more broadly the current state of anti-macrophage therapies in the clinic.

Comments on revised version:

The authors have adequately addressed all of my comments.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:

Reviewer #1 (Public review):

Summary:

In this paper, the authors have leveraged Single-cell RNA sequencing of the various stages of the evolution of lung adenocarcinoma to identify the population of macrophages that contribute to tumor progression. They show that S100a4+ alveolar macrophages, active in fatty acid metabolic activity, such as palmitic acid metabolism, seem to drive the atypical adenomatous hyperplasia (AAH) stage. These macrophages also seem to induce angiogenesis promoting tumor growth. Similar types of macrophage infiltration were demonstrated in the progression of the human lung adenocarcinomas.

Strengths:

Identification of the metabolic pathways that promote angiogenesis-dependent progression of lung adenocarcinomas from early atypical changes to aggressive invasive phenotype could lead to the development of strategies to abort tumor progression.

We are grateful for your constructive comments. These comments are very helpful for revising and improving our paper and have provided important guiding significance to our study. We have made revisions according to your comments and have provided point-by-point responses to your concerns.

Weaknesses:

(1) Can the authors demonstrate what are the functional specialization of the S100a4+ alveolar macrophages that promote the progression of the AAH to the more aggressive phenotype? What are the factors produced by these unique macrophages that induce tumor progression and invasiveness?

Thank you for your comments. To more comprehensively characterize the functional specialization of the S100a4⁺ alveolar macrophages, we expanded the macrophage functional gene sets based on relevant literature and databases and performed enrichment analysis. The results showed that all stages of precancerous progression presented activated states of angiogenesis, M2-like and immunosuppressive functions relative to the normal stage (Figure 4B). As we have demonstrated, S100a4⁺ alveolar macrophages predominantly exert pro-angiogenic functions during the AAH phase and may be more biased towards M2-like polarization and immunosuppression during further disease progression. Consistently, S100A4⁺ subset population of macrophages has been proved to exhibit a M2-like phenotype with immunosuppressive properties in tumor progression [PMID: 34145030]. In addition, S100A4 has been reported to be associated with macrophage M2 polarization, angiogenesis, and tumorigenesis [PMID: 39664586, 36895491, 30221056, 32117590]. The functional status of human S100A4⁺ alveolar macrophages is basically the same. The relevant description was added to the Results section as follows: “It was revealed that the capacities for angiogenesis, M2-like polarization, and immunosuppression were found to be stronger in AAH or other precancerous stages relative to the normal stage (Figure 4B). The pro-angiogenic function predominated in the AAH stage, while M2-like and immunosuppressive functions were more prominent in the subsequent precancerous progression.” (page 11, line 262). Our study puts more attention on the functional phenotypic changes of S100a4⁺ alveolar macrophages during the progression from normal to AAH to explain the role of this subpopulation in tumor initiation, and similarly, preliminary coculture experiments could only indicate its role in the early malignant transformation of epithelial cells. In further experimental validation, we will confirm the above functions of the S100a4⁺ alveolar macrophages promoting the progression of AAH to the more aggressive phenotype by in vitro and in vivo experiments. We have extended the limitations and potential experimental designs to the Discussion section as follows: “It is worth noting that our mining of S100a4⁺ alv-macro remains at the precancerous initiation stage, and further experimental designs are needed to verify its specific contribution at more aggressive stages. For example, FACS sorting of the subpopulation at different stages of disease progression, respectively, for precise functional characterization;” (page 19, line 468).

For the factors produced by these unique macrophages during induction of malignant transformation, we assayed culture supernatant of S100a4-OE alveolar macrophages for secreted functional cytokines. The results showed up-regulation of MIP-2, HGF, TNFα, IL-1a, CD27, CT-1, MMP9, 4-1BB, and CD40, and GO enrichment showed angiogenesis and tumorigenesis-related processes (Figure 5L and 5M). We have added the detailed content to the Results section as follows: “Next, we detected tumor-inducing factors secreted by these unique macrophages using Cytokine Antibody Array. We noted the production of macrophage inflammatory protein (MIP)-2, hepatocyte growth factor (HGF), tumor necrosis factor α (TNF-α), IL-1α, MMP9, and CD40, and these cytokine-related biological processes were mainly involved in the regulation of angiogenesis and immune response (Figure 5L and 5M).” (page 13, line 319). Furthermore, changes in these cytokines during subsequent invasive tumor progression will also be continuously monitored. The description in the Discussion section have been added as: “Furthermore, TGF-β and HGF activate vascular endothelial cells and promote proliferation and migration, as well as induce the expression of pro-angiogenic factors such as VEGF (Vimalraj, 2022; Watabe, Takahashi, Pietras, & Yoshimatsu, 2023). Macrophage-derived TNF-α and IL-1α lead tumor cells to produce potent angiogenic factors IL-8 and VEGF, which affect angiogenesis and tumor growth (Torisu et al., 2000). MIP2 and CD40 were also identified as pro-tumor factors associated with angiogenesis (Kollmar, Scheuer, Menger, & Schilling, 2006; Murugaiyan, Martin, & Saha, 2007)…continuous monitoring of the fluctuation of the above factors in bronchoalveolar lavage fluid at corresponding periods;” (page 19, line 461).

All method details covered in this section have been updated in the Materials and methods.

(2) Angiogenic factors are not only produced by the S100a4+ cells but also by pericytes and potentially by the tumor cells themselves. Then, how do these factors aberrantly trigger tumor angiogenesis that drives tumor growth?

Thank you for your comment. In our study, we detected up-regulation of angiogenic factors HIF-1α, VEGF, MMP9, and TGF-β (Figure 5K), and elevation of secreted HGF, IL-1α, and TNF-α (Figure 5L). We provide a detailed description of how these factors are involved in angiogenesis-related tumorigenesis to varying degrees in the Discussion section: “Precancerous lesions of LUAD are angiogenic, and pro-angiogenic factors secreted by cells, including S100a4⁺ alv-macro, induce endothelial cell sprouting and chemotaxis, leaving the angiogenic switch activated, prompting the formation of new blood vessels on the basis of the original ones to supply oxygen and nutrients to sustain tumor initiation (Chen et al., 2024; Kayser et al., 2003; van Hinsbergh & Koolwijk, 2008). Under hypoxic conditions, HIF-1α activates numerous factors that contribute to the angiogenic process, including VEGF, which promotes vascular permeability, and MMP9, which breaks down the ECM, promotes endothelial cell migration, and recruits pericytes to provide structural support (Raza, Franklin, & Dudek, 2010; Sakurai & Kudo, 2011). Cytokines secreted into the microenvironment activate macrophages, which subsequently produce angiogenic factors, further promoting angiogenesis (Sica, Schioppa, Mantovani, & Allavena, 2006). Furthermore, TGF-β and HGF activate vascular endothelial cells and promote proliferation and migration, as well as induce the expression of pro-angiogenic factors such as VEGF (Vimalraj, 2022; Watabe, Takahashi, Pietras, & Yoshimatsu, 2023). Macrophage-derived TNF-α and IL-1α lead tumor cells to produce potent angiogenic factors IL-8 and VEGF, which affect angiogenesis and tumor growth (Torisu et al., 2000)…” (page 19, line 449).

(3) It is not clear how abnormal fatty acid uptake by the macrophages drives the progression of tumors.

Thank you for your comment, which coincides with our mechanistic exploration. The metabolic status of macrophages influences their pro-tumor properties, and lipid metabolism has been shown to determine the functional polarization of macrophages [PMID: 29111350]. In this study, we observed more accumulation of lipid droplets in S100a4-OE MH-S, demonstrating enhanced cellular fatty acid uptake (Figure 6A). The pro-angiogenic ability of S100a4⁺ alv-macro was confirmed by tube formation assay and cytokine assay (Figure 6B and 5M). Cpt1a was thought to play a crucial role in the metabolic paradigm shift of S100a4⁺ alv-macro, we therefore performed functional rescue experiments by inhibiting CPT1A expression in S100a4-OE MH-S by addition of etomoxir (ETO). After culture with conditioned medium of MH-S, the proliferation, migration, and ROS production of MLE12 cells were all restored to lower levels (Figure 6E-G). In addition, ETO treatment significantly reversed the angiogenesis, which supported the regulation of fatty acid metabolism on macrophage function (Figure 6H). Immunoblotting also revealed restoration of expression in related proteins (Figure 6I and 6J), these findings reinforced previous analyses of the association of fatty acid metabolism with pro-angiogenesis and M2-like function in S100a4⁺ alv-macro. The involvement of PPAR-γ in the regulation of metabolic state was also confirmed. Taken together, we suggest that S100a4⁺ alv-macro promotes fatty acid metabolism through the CPT1A-PPAR-γ axis, enhances its ability to promote angiogenesis, and thus drives tumor occurrence. The corresponding contents were added in the Results section S100a4⁺ alv-macro drove angiogenesis by promoting Cpt1a-mediated fatty acid metabolism (page 13, line 327) and Discussion section: “We demonstrated the regulation of fatty acid metabolism by CPT1A in S100a4⁺ alv-macro as well as the involvement of PPAR-γ. Nevertheless, the molecular mechanism that drives the acquisition of metabolic and functional switching properties specific to this cell state still requires further characterization in the context of precancerous lesions. It has been reported that CD36 is the main effector of the S100A4/PPAR-γ pathway, and its mediated fatty acid uptake plays an important role in the tumor-promoting function of macrophages (S. Liu et al., 2021).” (page 18, line 433).

All method details covered in this section have been supplemented in the Materials and methods.

(4) Does infusion or introduction of S100a4+ polarized macrophages promote the progression of AAH to a more aggressive phenotype?

Thank you for your comment. We performed intratracheal instillation of lentivirus-infected S100a4-OE MH-S and culture supernatant in A/J and BALB/c mice, respectively, but no aggressive pathological phenotype was observed so far, possibly due to the lack of time required for lesions or the imperfection of experimental conditions. We will continue to explore the instillation dose and frequency for long-term monitoring and will simultaneously evaluate the availability of primary alveolar macrophages. We have discussed as follows: “It is worth noting that our mining of S100a4⁺ alv-macro remains at the precancerous initiation stage, and further experimental designs are needed to verify its specific contribution at more aggressive stages…and intratracheal instillation of primary S100a4⁺ alv-macro to observe the pathological progression of precancerous lesions.” (page 19, line 468).

(5) How does Anxa and Ramp1 induction in inflammatory cells induce angiogenesis and tumor progression?

Thank you for your comment. ANXA2 is an important member of annexin family of proteins expressed on surface of endothelial cells, macrophages, and tumor cells [PMID: 30125343]. ANXA2 was reported to regulate neoangiogenesis in the tumor microenvironment and most likely due to overproduction of plasmin. As a well-established receptor for plasminogen (PLG) and tissue plasminogen activator (tPA) on the cell surface, ANXA2 converts PLG into plasmin. Plasmin plays a critical role in the activation of cascade of inactive proteolytic enzymes such as metalloproteases (pro-MMPs) and latent growth factors (VEGF and bFGF) [PMID: 12963694, 11487021]. Activated forms of MMPs and VEGF then induce extracellular matrix remodeling facilitating angiogenesis and tumor development [PMID: 15788416]. Sharma et al. suggested administration of ANXA2-antibody inhibited tumor angiogenesis and growth concurrent with plasmin generation [PMID: 22044461], the role of ANXA2 in plasmin activation thus explains it’s importance in tumor-related angiogenesis. We verified the simultaneous upregulation of ANXA2 and PLG in S100a4-OE MH-S and cocultured HUVEC and MLE12 by immunoblotting (Figure 6D). The relevant description was added to the Results section as follows: “ANXA2 is considered to be a cellular receptor for plasminogen (PLG), often expressed on the surface of endothelial cells, macrophages, and tumor cells, which activates a cascade of pro-angiogenic factors by promoting the conversion of PLG to plasmin, thereby promoting angiogenesis and tumor progression (Semov et al., 2005; Sharma, 2019). We found synergistic upregulation of ANXA2 and PLG expression in S100a4-OE MH-S and cocultured HUVEC and MLE12, which may help explain how ANXA2 induction was involved in angiogenesis and malignant transformation (Figure 6D).” (page 14, line 338).

Recent studies showed that S100A4 is associated with tumor angiogenesis and progression by the interaction with ANXA2. ANXA2 is the endothelial receptor for S100A4 and that their interaction triggers the functional activity directly related to pathological properties of S100A4, including angiogenesis [PMID: 18608216]. It has been proved that S100A4 induces angiogenesis through interaction with ANXA2 and accelerated plasmin formation [PMID: 15788416, 25303710]. In addition, it is generally believed that ANXA2 participates in malignant cell transformation [PMID: 28867585]. Therefore, we speculate that ANXA2 may promote plasmin production by binding to S100A4, thus promoting angiogenesis and tumor initiation, and we have discussed accordingly: “The role of ANXA2 in angiogenesis has been widely recognized, and it may facilitate plasmin production by binding to S100A4 and then trigger angiogenesis and malignant cell transformation (Grindheim, Saraste, & Vedeler, 2017; Y. Liu, Myrvang, & Dekker, 2015).” (page 18, line 446).

In our study, the primary target of our validation was ANXA2 rather than RAMP1, even though its relationship with angiogenesis had been established [PMID: 20596610], so we weakened the relevant description in the manuscript.

(6) For the in vitro studies the authors might consider using primary tumor cells and not cell lines.

Thank you for your suggestion, which was in our initial experimental plan. However, since S100A4 is not expressed on the cell surface, FACS sorting of primary subset of alveolar macrophages presents technical limitations. We have also attempted overexpression in primary macrophages, but the current overexpression efficiency and cell status are not sufficient to support a subsequent series of experiments. For all these reasons, the alveolar macrophage cell line MH-S and the lung epithelial cell line MLE12 were selected to ensure the consistency and stability of the coculture system.

In addition, we are optimizing the experimental conditions to achieve coculture of primary macrophages and epithelial cells, and will also establish transgenic mouse models for simultaneous validation. The Discussion has been added as: “Besides, as our previous in vitro results were obtained based on cell lines, we will optimize the experimental conditions to achieve coculture of primary macrophage subset and epithelial cells and establish transgenic mouse models for in vivo validation.” (page 19, line 475).

Reviewer #2 (Public review):

Summary:

The work aims to further understand the role of macrophages in lung precancer/lung cancer evolution

Strengths:

(1) The use of single-cell RNA seq to provide comprehensive characterisation.

(2) Characterisation of cross-talk between macrophages and the lung precancerous cells.

(3) Functional validation of the effects of S100a4+ cells on lung precancerous cells using in vitro assays.

(4) Validation in human tissue samples of lung precancer / invasive lesions.

We are grateful for your constructive comments. These comments are very helpful for revising and improving our paper and have provided important guiding significance to our study. We have made revisions according to your comments and have provided point-by-point responses to your concerns.

Weaknesses:

(1) The authors need to provide clarification of several points in the text.

Thank you for your comment. We have clarified these points in the manuscript and responded to all your concerns in detail. Please see the responses to Recommendations for the authors.

(2) The authors need to carefully assess their assumptions regarding the role of macrophages in angiogenesis in precancerous lesions.

Thank you for your comment. We have cited relevant literature to support the occurrence of angiogenesis in precancerous lesions, and demonstrated the contribution of S100a4⁺ alveolar macrophages by tube formation assay and cytokine assay. In addition, we have discussed the relevant limitations of this study and aimed to provide more robust evidence. Please see the responses to Recommendations for the authors.

(3) The authors should discuss more broadly the current state of anti-macrophage therapies in the clinic.

Thank you for your suggestion. We have provided extensive discussion of the clinical state of anti-macrophage therapies. Please see the responses to Recommendations for the authors.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The text has grammatical and syntax errors that need to be corrected accordingly.

Thank you for your suggestion. We have corrected the grammatical and syntactic errors and asked a native English speaker in the field to help polish the full text.

Reviewer #2 (Recommendations for the authors):

This work provides an important contribution to our further understanding of the role of macrophages in lung precancer/lung cancer evolution. I have several comments regarding how the manuscript could be improved:

Introduction:

The authors may consider citing the following work to enhance their work:

(1) At line 78, where they talk about precancerous lesions being reversible, they should cite recent work on this in lung cancer: Teixeria et al 2019 PMID: 30664780, and Pennycuik et al 2020 PMID: 32690541.

Thank you for your suggestion. We have cited the above references in the corresponding paragraph (page 4, line 76).

(2) At line 96, where they talk about developing medicines for precancerous lesions, the authors should cite comprehensive review articles where this concept has been discussed in depth, for example: Reynolds et al 2023 PMID: 37067191, and Asad et al 2012 PMID: 23151603.

Thank you for your suggestion. We have cited the above references in the corresponding paragraph (page 5, line 94).

Results:

(1) Line 142, the authors say "mice were feed for 12-16 months" - do they mean the mice were maintained for 12-16 months?

Thank you for your comment. To best mimic the process of human lung cancer development, A/J mice with the highest incidence of spontaneous lung tumors, which increases substantially with age, were selected. The corresponding description has been modified as: “A/J mice have the highest incidence of spontaneous lung tumors among various mouse strains, and this probability significantly increased with age (Landau, Wang, Yang, Ding, & Yang, 1998). To more comprehensively mirror the tumor initiation and progression process of human lung cancer, A/J mice were maintained for 12-16 months for spontaneous lesions, which resulted in three recognizable precancerous lesions in the lung.” (page 7, line 138).

(2) Line 143, the authors claim to have seen "three recognizable precancerous and cancerous lesions in the lung" but then, they only go on to describe AAH, adenoma, and AIS, lesions which are all commonly recognized as precancers. What was the cancerous (i.e. invasive) lesion they identified?

Thank you for your comment. We apologize for this misstatement and will include cancerous lesions from mice for simultaneous analysis in subsequent study. The corresponding description has been revised as: “To more comprehensively mirror the tumor initiation and progression process of human lung cancer, A/J mice were maintained for 12-16 months for spontaneous lesions, which resulted in three recognizable precancerous lesions in the lung.” (page 7, line 140).

(3) Line 172, the authors say that the "proportion of cell types across the four stages showed a dynamic trend" ... what does this mean? A trend towards what exactly?

Thank you for your comment. Our intention was to highlight heterogeneous changes, and the description has been corrected: “The proportion of cell types across the four stages showed irregular changes, while transcriptional homogeneity was reduced with precancerous progression, illustrating the importance of heterogeneity in tumorigenesis and also proving the reliability of the sampling in this study.” (page 8, line 169).

(4) Line 193, the authors say cell communication "showed a tendency to malignant transformation." What does this statement mean? If they mean more cell communication occurred in the malignant lesions than the precancerous, then there is a flaw in the logic because AAH, adenoma, and AIS are all precancerous lesions. What is the sequence of evolution to malignancy the authors are assuming? Do they mean AIS is a more advanced stage of precancerous malignancy than adenoma, and adenoma is more advanced than AAH (albeit they are all precancerous lesions).

Thank you for your comments. The malignant transformation process involves multiple stages, and histological AAH is regarded as the beginning of this process. Precancerous lesions of LUAD in mice are believed to develop stepwise from AAH, adenoma, to AIS, even if the process is not necessarily completely consistent [PMID: 11235908, 32707077]. What we meant to describe was a gradual increase in the frequency of cell communication during this process. The corresponding description has been modified as: “At the evolutionary stages of precancerous LUAD, despite possible sample heterogeneity and other interference, we observed increased interactions between epithelial cells and surrounding stromal and immune cells in the microenvironment, indicating gradually frequent cell-cell communication during this process” (page 8, line 187).

(5) Immunofluorescence images in Figure 3G and Figure 4F are captured at low magnification, making it very difficult to evaluate the colocalisation data. Suggest authors provide higher magnification images.

Thank you for your suggestion. We have replaced the immunofluorescence images in Figure 3G and Figure 4F with higher magnification images.

(6) Line 284 when referencing the cell line here, the author should make it clear in the text that cells were transfected with a construct expressing S100A4. If possible, would be good to understand if the level of S100A4 expression achieved is less, similar, or greater than that seen in these cells in vivo.

Thank you for your suggestion. We have amended the text to make it clear: “S100a4-overexpressed (OE) alveolar macrophages were established by transfection of the mS100a4 vector into the murine MH-S cell line, and empty vector was transfected as negative control (NC) cells” (page 12, line 284), and it will be clarified in the following exploration whether the level of S100a4 expression achieved is less, similar, or greater than that seen in these cells in vivo.

(7) Line 285 - when the authors first refer to OE cells that have been transfected, they should also inform the reader what NC cells are i.e. negative control cells?

Thank you for your suggestion. We have revised the relevant content as follows: “S100a4-overexpressed (OE) alveolar macrophages were established by transfection of the mS100a4 vector into the murine MH-S cell line, and empty vector was transfected as negative control (NC) cells” (page 12, line 284).

(8) Line 324 - the authors claim they have demonstrated that the macrophages promote angiogenesis through upregulation of fatty acid metabolism. Whilst they may have demonstrated changes in fatty acid metabolism, no experiments assessing the effect of the macrophages in angiogenesis assays are included in the paper, so the authors should modify this statement.

Thank you for your comments. The relevant experiments have been added based on your suggestions. Firstly, we demonstrated in vitro the up-regulation of fatty acid metabolism in S100a4⁺ alv-macro and uncovered the contribution of CPT1A to angiogenesis and cell transformation through rescue experiments; Then, HUVEC tube formation assay and cytokine assay confirmed the pro-angiogenic effect of S100a4⁺ alv-macro. We have added the Results section S100a4⁺ alv-macro drove angiogenesis by promoting Cpt1a-mediated fatty acid metabolism (page 13, line 327) and added the Discussion as: “We demonstrated the regulation of fatty acid metabolism by CPT1A in S100a4⁺ alv-macro as well as the involvement of PPAR-γ. Nevertheless, the molecular mechanism that drives the acquisition of metabolic and functional switching properties specific to this cell state still requires further characterization in the context of precancerous lesions. It has been reported that CD36 is the main effector of the S100A4/PPAR-γ pathway, and its mediated fatty acid uptake plays an important role in the tumor-promoting function of macrophages (S. Liu et al., 2021).” (page 18, line 433).

All method details covered in this section have been supplemented in the Materials and methods.

(9) Regarding angiogenesis in precancerous lesions and the role of macrophages in this process: is there even any evidence that precancerous LUAD lesions are angiogenic? Don't these lesions typically have a lepidic pattern, wherein the cancer cells merely co-opt pre-existing alveolar capillaries without the need to generate new vessels?

Thank you for your comments. As you mentioned, pathologically, precancerous LUAD lesions mainly show a lepidic growth pattern, characterized by the growth of type II alveolar epithelial cells along pre-existing alveolar walls [PMID: 29690599], but this does not mean that this process does not require the formation of new blood vessels. There are multiple patterns of tumor angiogenesis. Some studies have shown that increased angiogenesis can be observed in certain precancerous lesions, which suggests that angiogenesis may play an important role in the early stages of lung cancer development. Microvessel density (MVD) was increased in AAH and AIS compared to normal lung tissue, indicating that new blood vessels are forming to provide essential nutrients and oxygen to tumor cells to support their growth. The expression level of pro-angiogenic factors such as VEGF is usually upregulated, which promotes the formation of new blood vessels by stimulating endothelial cell proliferation and migration. [PMID: 39570802, 14568684] In addition, the infiltration of macrophages into precancerous areas in response to cytokines has been shown to trigger a tumor angiogenic switch and maintain tumor-associated continuous angiogenesis [PMID: 35022204]. Our in vitro tube formation assay and cytokine assay also demonstrated angiogenesis induced by S100a4⁺ alv-macro. We have discussed the relevant content (page 19, line 449) and will provide more sufficient evidence in future work.

Discussion:

Perhaps the authors can cite any literature pertaining to the current wave of anti-macrophage therapies currently being tested in the clinic. Moreover, have these therapies been tested in lung cancer, and if so, what were the results?

Thank you for your suggestion. At present, the clinical trials of anti-macrophage therapies mainly involve Gaucher's disease and hematological malignancies, and the two tests related to lung cancer have no valid data posted. Nevertheless, there are some preclinical studies worth learning from. We have cited the relevant literature and discussed in detail: “With the elaborate resolution of TME, macrophage-related therapy is considered to be promising. So far, macrophage-targeted therapy has demonstrated clinical efficacy in Gaucher's disease and advanced hematological malignancies (Barton et al., 1991; Ossenkoppele et al., 2013). In lung cancer, an attempt to enhance anti-PD-1 therapy in NSCLC by depleting myeloid-derived suppressor cells with gemcitabine was prematurely terminated because of insufficient data collected; another clinical trial of TQB2928 monoclonal antibody promoting macrophage phagocytosis of tumor cells in combination with a third-generation EGFR TKI for advanced NSCLC is now recruiting. Moreover, preclinical studies on macrophage-targeted therapy combined with immune checkpoint inhibitors are being extensively conducted in NSCLC, and it was suggested that blockade of purine metabolism can reverse macrophage immunosuppression, and a synergetic effect can be achieved when combined with anti-PD-L1 therapy, which inspired the direction of our early intervention strategies (H. Wang, Arulraj, Anbari, & Popel, 2024; Yang et al., 2025).” (page 20, line 479).

Methods:

Further description of how lesions were classified as precancerous (AAH, adenoma, AIS) or cancerous by the pathologist should be defined (or cite appropriate reference where this is described).

Thank you for your suggestion. We have cited relevant references in the Methods section (page 21, line 528) on how lesions were classified by the pathologists [PMID: 21252716, 28951454, 32707077, 24811831].

Reviewer #1 (Public review):

Summary:

Fernandez et al. investigate the influence of maternal behavior on bat pup vocal development in Saccopteryx bilineata, a species known to exhibit vocal production learning. The authors performed detailed longitudinal observations of wild mother-pup interactions to ask whether non-vocal maternal displays during juvenile vocal practice, or 'babbling', affect vocal production. Specifically, the study examines the durations of pup babbling events and the developmental babbling phase, in relation to female display rates, as well as pup age and the number of nearby singing adult males. Furthermore, the authors examine pup vocal repertoire size and maturation in relation to maternal display rates encountered during babbling. Statistical models identify female display behavior as a predictor of i) babbling bout duration, ii) the length of the babbling phase, iii) song composition and iv) syllable maturation. Notably, these outcomes were not influenced by the number of nearby adult males (the pups' source of song models) and were largely independent of general maturation (pup age). These findings highlight the impact of non-vocal aspects of social interactions in guiding mammalian vocal development.

Strengths:

Historically, work on developmental vocal learning has focused on how juvenile vocalizations are influenced by the sounds produced by nearby adults (often males). In contrast, this study takes the novel approach of examining juvenile vocal ontogeny in relation to non-vocal maternal behavior, in one of the few mammals known to exhibit vocal production learning. The authors collected an impressive dataset from multiple wild bat colonies in two Central American countries. This includes longitudinal acoustic recordings and behavioral monitoring of individual mother-pup pairs, across development.

The identified relationships between maternal behavior and bat pup vocalizations have intriguing implications for understanding the mechanisms that enable vocal production learning in mammals, including human speech acquisition. As such, these findings are likely be relevant to a broad audience interested in the evolution and development of social behavior as well as sensory-motor learning.

Weaknesses:

The authors qualitatively describe specific patterns of female displays during pup babbling, however, subsequent quantitative analyses are based on aggregate measures of female behavior that pool across display types. Consequently, it remains unclear how certain maternal behaviors might differentially influence pup vocalizations (e.g. through specific feedback contingencies or more general modulation of pup behavioral states).

Comments on revisions:

(1) More detailed analyses of female behavior may be beyond the scope of this study, given the nature of the dataset/recordings. I look forward to the authors' future work on this aspect.

By addressing the important distinction between display number vs. display rate, the authors have provided more direct support for the claim that babbling behavior is related to female displays.

(2) The additional information regarding exposure to adult male song is appreciated.

(3) Added discussion of pup sex differences provides useful context and intriguing speculation about the role of female pup babbling.

(4) The authors' additions have significantly improved the clarity of their acoustic terminology and syllable analyses.

Author response:

The following is the authors’ response to the original reviews

Public Reviews:  

Reviewer #1 (Public Review):

Strengths:

The manuscript utilizes a previously reported misfolding-prone reporter to assess its behaviour in ER in different cell line models. They make two interesting observations:

(1) Upon prolonged incubation, the reporter accumulates in nuclear aggregates.

(2) The aggregates are cleared during mitosis. They further provide some insight into the role of chaperones and ER stressors in aggregate clearance. These observations provide a starting point for addressing the role of mitosis in aggregate clearance. Needless to say, going ahead understanding the impact of aggregate clearance on cell division will be equally important.

Weaknesses:

The study almost entirely relies on an imaging approach to address the issue of aggregate clearance. A complementary biochemical approach would be more insightful. The intriguing observations pertaining to aggregates in the nucleus and their clearance during mitosis lack mechanistic understanding. The issue pertaining to the functional relevance of aggregation clearance or its lack thereof has not been addressed. Experiments addressing these issues would be a terrific addition to this manuscript.

We have performed protein blotting and proteomics to characterize ER-FlucDM-eGFP expressing cells. We have also provided evidence to support the role of ER reorganization in regulating aggregate clearance. Our proteomic analysis provided a global view of the cellular state of cells expressing ER-FlucDM-eGFP, which potentially revealed functional relevance of ER-FlucDM-eGFP. Details are explained in the following comments. 

Reviewer #2 (Public Review):

Summary:

The authors provide an interesting observation that ER-targeted excess misfolded proteins localize to the nucleus within membrane-entrapped vesicles for further quality control during cell division. This is useful information indicating transient nuclear compartmentalization as a quality control strategy for misfolded ER proteins in mitotic cells, although endogenous substrates of this pathway are yet to be identified.

Strengths:

This microscopy-based study reports unique membrane-based compartments of ERtargeted misfolded proteins within the nucleus. Quarantining aggregating proteins in membrane-less compartments is a widely accepted protein quality control mechanism. This work highlights the importance of membrane-bound quarantining strategies for aggregating proteins. These observations open up multiple questions on proteostasis biology. How do these membrane-bound bodies enter the nucleus? How are the singlelayer membranes formed? How exactly are these membrane-bound aggregates degraded? Are similar membrane-bound nuclear deposits present in post-mitotic cells that are relevant in age-related proteostasis diseases? Etc. Thus, the observations reported here are potentially interesting.

Weaknesses:

This study, like many other studies, used a set of model misfolding-prone proteins to uncover the interesting nuclear-compartment-based quality control of ER proteins. The endogenous ER-proteins that reach a similar stage of overdose of misfolding during ER stress remain unknown.

We have included a previous study that showed accumulation of BiP aggregates in the nucleus upon overexpression of BiP (Morris et al., 1997; DOI: 10.1074/jbc.272.7.4327) in the discussion (Line 299).

The mechanism of disaggregation of membrane-trapped misfolded proteins is unclear. Do these come out of the membrane traps? The authors report a few vesicles in living cells. This may suggest that membrane-untrapped proteins are disaggregated while trapped proteins remain aggregates within membranes.

We initially made mStayGold-Sec61β to image the ER structures and ER-FlucDM-eGFP aggregates. However, we could not obtain convincing time-lapse images to show the release of ER-FlucDM-eGFP aggregates from the ER membrane as there are abundant ER structures present close to the aggregates during mitosis, preventing the differentiation of the membrane encapsulating aggregates from the ER structures. 

The authors figure out the involvement of proteasome and Hsp70 during the disaggregation process. However, the detailed mechanisms including the ubiquitin ligases are not identified. Also, is the protein ubiquitinated at this stage?

We performed cycloheximide chase experiments in cells released from the G2/M and found that ER-FlucDM-eGFP protein level did not fluctuate significantly when cells progressed through mitosis and cytokinesis. Thus, we did not consider protein ubiquitination and degradation of ER-FlucDM-eGFP as a major mechanism for its clearance. We have included this observation in the results (Figure S7A; Line 266) and in the discussion (Line 324) of the revised manuscript.

This paper suffers from a lack of cellular biochemistry. Western blots confirming the solubility and insolubility of the misfolded proteins are required. This will also help to calculate the specific activity of luciferase more accurately than estimating the fluorescence intensities of soluble and aggregated/compartmentalized proteins. 

We performed solubility test in cells expressing ER-FlucDM-eGFP and detected insoluble ERFlucDM-eGFP after heat stress (Figure S1E; Line 102). We have also performed protein blotting to detect ER-FlucDM-eGFP to normalize the luciferase activity (Line 609). We have updated the method section for luciferase measurement (Line 494).   

Microscopy suggested the dissolution of the membrane-based compartments and probably disaggregation of the protein. This data should be substantiated using Western blots. Degradation can only be confirmed by Western blots. The authors should try time course experiments to correlate with microscopy data. Cycloheximide chase experiments will be useful.

We performed cycloheximide chase experiments in cells released from the G2/M and found that ER-FlucDM-eGFP protein level did not fluctuate significantly when cells progressed through mitosis and cytokinesis (Figure S7A to S7C). Also, live-cell imaging of cells released from the G2/M indicated no significant change of total fluorescence intensity of ER-FlucDMeGFP (Figure S7D). Thus, we do not think that protein degradation of ER-FlucDM-eGFP is the major mechanism for its clearance. 

The cell models express the ER-targeted misfolded proteins constitutively that may already reprogram the proteostasis. The authors may try one experiment with inducible overexpression.

We have re-transduced fresh MCF10A cells with lentiviral particles to induce expression of ER-FlucDM-eGFP. The aggregates started to form after 24 h post-transduction. We made similar observations as described in the manuscript (e.g. aggregate clearance) two days after re-transduction.

It is clear that a saturating dose of ER-targeted misfolded proteins activates the pathway.

The authors performed a few RT-PCR experiments to indicate the proteostasis-sensitivity.

Proteome-based experiments will be better to substantiate proteostasis saturation.

We have performed proteomic analysis in cells expressing ER-FlucDM-eGFP and observed up-regulation of multiple proteins involved in the ER stress response, indicating that cells expressing ER-FlucDM-eGFP experience proteostatic stress (Figure S4A; Line 179).  

The authors should immunostain the nuclear compartments for other ER-membrane resident proteins that span either the bilayer or a single layer. The data may be discussed.

We have co-expressed ER-FlucDM-mCherry and mStayGold-Sec61β and detected mStayGold- Sec61β around ER-FlucDM-mCherry aggregates (Figure 1B).  

All microscopy figures should include control cells with similarly aggregating proteins or without aggregates as appropriate. For example, is the nuclear-targeted FlucDM-EGFP similarly entrapped? A control experiment will be interesting. Expression of control proteins should be estimated by western blots.

We targeted FlucDM-eGFP to the nucleus by expressing NLS-FlucDM-eGFP (Figure S1A). We found that the nuclear FlucDM-eGFP did not co-localize with the ER-FlucDM-mCherry aggregates (Figure S1B; Line 96). We have also determined the expression levels of NLSFlucDM-eGFP and ER-FlucDM-mCherry (Figure S1C and S1D).

There are few more points that may be out of the scope of the manuscript. For example, how do these compartments enter the nucleus? Whether similar entry mechanisms/events are ever reported? What do the authors speculate? Also, the bilayer membrane becomes a single layer. This is potentially interesting and should be discussed with probable mechanisms. Also, do these nuclear compartments interfere with transcription and thereby deregulate cell division? What about post-mitotic cells? Similar deposits may be potentially toxic in the absence of cell division. All these may be discussed.

Thank you for interesting suggestions for our study. We speculated that ER-FlucDM-eGFP aggregates may derive from the invagination of the inner nuclear membrane given that the aggregates are in close proximity to the inner nuclear membrane in interpase cells (Line 299). We have included a previous study that reported a similar aggregate upon BiP overexpression (Morris et al., 1997; DOI: 10.1074/jbc.272.7.4327; Line 300). Our proteomic analysis showed that cells expressing ER-FlucDM-eGFP have several up-regulated proteins related to cell cycle regulation (Figure S4A; Line 346).  

Reviewer #3 (Public Review):

Summary:

This paper describes a new mechanism of clearance of protein aggregates occurring during mitosis.

The authors have observed that animal cells can clear misfolded aggregated proteins at the end of mitosis. The images and data gathered are solid, convincing, and statistically significant. However, there is a lack of insight into the underlying mechanism. They show the involvement of the ER, ATPase-dependent, BiP chaperone, and the requirement of Cdk1 inactivation (a hallmark of mitotic exit) in the process. They also show that the mechanism seems to be independent of the APC/C complex (anaphase-promoting complex). Several points need to be clarified regarding the mechanism that clears the aggregates during mitosis:

• What happens in the cell substructure during mitosis to explain the recruitment of BiP towards the aggregates, which seem to be relocated to the cytoplasm surrounded by the ER membrane.

We have included images to show that BiP co-localizes with ER-FlucDM-eGFP aggregates in interphase cells (Figure S5C). We think that BiP participates in the formation of ER-FlucDMeGFP during interphase instead of getting recruited to the aggregates during mitosis.  

• How the changes in the cell substructure during mitosis explain the relocation of protein aggregates during mitosis.

We provided evidence to show that clearance of ER-FlucDM-eGFP aggregates involves the ER remodeling process. We depleted ER membrane fusion proteins ATL2 and ATL3 to perturb the distribution of ER sheets or tubules and found that cells were defective in clearing the aggregates (Figure 7A and B; Line 278). 

• Why BiP seems to be the main player of this mechanism and not the cyto Hsp70 first described to be involved in protein disaggregation.

In our proteomic analysis, we found that BiP (HSPA5) but not other Hsp70 family members were up-regulated in cells expressing ER-FlucDM-eGFP (Line 352; Figure S4A). This explains why BiP is the main player of the ER-FlucDM-eGFP aggregate clearance.  

Strengths:

Experimental data showing clearance of protein aggregates during mitosis is solid, statistically significant, and very interesting.

Weaknesses:

Weak mechanistic insight to explain the process of protein disaggregation, particularly the interconnection between what happens in the cell substructure during mitosis to trigger and drive clearance of protein aggregates.

In our revised manuscript, we now provided evidence to show that ER-FlucDM-eGFP aggregate clearance involved remodeling of the ER structures during mitotic exit. This is added as a new Figure 7 in the revised manuscript and is described in the result section (Line 278) and in the discussion section (Line 323). We believe that this addition has provided mechanistic insights into ER-FlucDM-eGFP aggregate clearance.

Recommendations for the authors:

Reviewing Editor comments:

I have read these reviews in detail and would like to recommend that the authors perform the experiments according to the reviewers' suggestions, as well as provide the appropriate controls raised by the reviewers.

I think there are not that many requests and they all seem very reasonable and easily doable. I would recommend that the authors carry out the suggested experiments to develop a stronger story where the evidence transitions from being incomplete presently to a "more complete" standard.

We have addressed questions raised by three reviewers and updated our manuscript (labeled in red in the main text).

Reviewer #1 (Recommendations For The Authors):

The manuscript makes exciting observations about the accumulation of reporter protein aggregates in the nucleus and its clearance during mitosis. It also provides some insight into the role of chaperons in aggregate clearance. These observations provide a good platform to perform in-depth analysis of the underlying mechanism and its functional relevance which perhaps the authors will plan over the long term. However, the below suggestions will help improve the current version of the manuscript:

(1) Although it is assumed that the aggregates are cleared by the protein degradation mechanism, clear evidence supporting this assumption in the author's experiments is lacking and needs to be provided. Is it possible that mitosis induces disassembly of these aggregates instead of degradation?

We performed two experiments to verify whether ER-FlucDM-eGFP aggregates are cleared by the protein degradation mechanism. In the first experiment, we treated cells expressing ER-FlucDM-eGFP released from the G2/M boundary with cycloheximide (CHX) and found that ER-FlucDM-eGFP did not decrease in protein abundance in cells progressing through mitosis (Figure S7A to S7C). In the second experiment, we measured the intensity of ERFlucDM-eGFP in early dividing cells and late dividing cells after release from the G2/M boundary and found that there was no significant difference between early and late dividing cells (Figure S7D). Thus, we concluded that protein degradation of ER-FlucDM-eGFP is not the primary mechanism of its clearance during cell division (Line 324). Furthermore, we included new data to show that the ER-FlucDM-eGFP aggregate clearance depends on ER reorganization during cell division, so mitotic exit induces disassembly of the aggregates instead of protein degradation.

(2) It is intriguing that the aggregates are nuclear. Is the nuclear localization mediated by localization to ER? A time course analysis would reveal this and would provide credence to the idea that the reporter was originally expressed in the ER. It is currently unclear if the reporter ever gets expressed in ER.

We showed that in interphase cells, ER-FlucDM-eGFP co-localizes with mStayGold-Sec61β, which labels the ER structures (Figure 1B). So, ER-FlucDM-eGFP is expressed and present in the ER network and invaginates into the inner nuclear membrane as aggregates. We attempted to image ER-FlucDM-eGFP for its formation; however it was technically challenging as the aggregates appeared very small and not too visible after clearance under our microscopy system.  

(3) It would be expected that the persistence of these aggregates would impact cell division and cellular health. An experiment addressing this hypothesis would be very useful in establishing the functional relevance of this observation in the context of the current study.

We have performed proteomic analysis on cell expressing ER-FlucDM-eGFP and found that multiple proteins involved in the ER stress response were up-regulated (Figure S4A). Additionally, proteins related to cell cycle regulation were up-regulated upon expression of ER-FlucDM-eGFP (Figure S4A). The increase of these proteins may indicate a perturbed cellular health (Line 344). 

(4) A recent report (PMID: 34467852) identified the role of ER tubules in controlling the size of certain misfolded condensates. Would specific ER substructures affect the nuclear localization and/or clearance of the FlucDM aggregates? This is tied to point#2 and would provide insights into the connection between ER and the nuclear aggregates.

Thank you for your suggestions. We perturbed the ER remodeling process by knocking down ATL2 and ATL3, which are ER membrane fusion proteins, and found that clearance of ER-FlucDM-eGFP aggregates was affected (Figure 7A and B). Hence, perturbation of the distribution of ER tubules and ER sheets affects ER-FlucDM-eGFP aggregate clearance. We have also added the recent paper about ER tubule size in regulating the sizes of misfolded condensates in the discussion (Line 321)

Reviewer #2 (Recommendations For The Authors):

I expect that the images indicate z-sections. Should be indicated in legends as applicable.

We have indicated whether the images are Z-stack or single Z-slices in the figure legends.  

Small point: the control region (outside inclusion) that was bleached in 2c may be clearly indicated. 

We have added the explanation in the figure legend of Figure 2C.

Author response:

The following is the authors’ response to the previous reviews

Reviewer #1 (Public review):

Summary:

The authors investigate the neuroprotective effect of reserpine in a retinitis pigmentosa (P23H-1) model, characterized by a mutation in the rhodopsin gene. Their results reveal that female rats show better preservation of both rod and cone photoreceptors following reserpine treatment compared to males.

Strengths:

This study effectively highlights the neuroprotective potential of reserpine and underscores the value of drug repositioning as a strategy for accelerating the development of effective treatments. The findings are significant for their clinical implications, particularly in demonstrating sex-specific differences in therapeutic response.

We sincerely appreciate the reviewer’s comments.

Weaknesses:

The main limitation is the lack of precise identification of the specific pathway through which reserpine prevents photoreceptor death.

We acknowledge that the exact pathway through which reserpine exerts its protective effects on photoreceptors remains undetermined, yet our findings provide critical insights into potential mechanisms. Together with our previous report [PMID: 36975211], the studies being presented here validate proteostasis (including autophagy) and p53 signaling as the key pathways underlying reserpine-mediated survival of photoreceptors in retinal disease models. We also go a step further by showing an influence of the biological sex.

We emphasize that the primary aim of this study was to demonstrate the effectiveness of reserpine in a different retinal degeneration model—specifically, the autosomal dominant RP model—which shares a retinal disease phenotype with the model used for initial screening but involves different genetic and molecular mechanisms of degeneration.

Reviewer #2 (Public review):

Summary:

In the manuscript entitled "Sex-specific attenuation of photoreceptor degeneration by reserpine in a rhodopsin P23H rat model of autosomal dominant retinitis pigmentosa" by Beom Song et al., the authors explore the transcriptomic differences between male and female wild-type (WT) and P23H retinas, highlighting significant gene expression variations and sex-specific trends. The study emphasizes the importance of considering biological sex in understanding inherited retinal degeneration and the impact of drug treatments on mutant retinas.

Strengths:

(1) Relevance to Clinical Challenges: The study addresses a critical limitation in inherited retinal degeneration (IRD) therapies by exploring a gene-agnostic approach. It emphasizes sex-specific responses, which aligns with recent NIH mandates on sex as a biological variable.

(2) Multi-dimensional Methodology: Combining electroretinography (ERG), optical coherence tomography (OCT), histology, and transcriptomics strengthens the study's findings.

(3) Novel Insights: The transcriptomic analysis uncovers sex-specific pathways impacted by reserpine, laying the foundation for personalized approaches to retinal disease therapy.

We are grateful for highlighting the strengths of our work.

Weaknesses:

Dose Optimization

The study uses a fixed dose (40 µM), but no dose-response analysis is provided. Sex-specific differences in efficacy might be influenced by suboptimal dosing, particularly considering potential differences in metabolism or drug distribution.

We acknowledge the limitation of using a fixed dose (40 µM) of reserpine in this study without conducting a comprehensive dose-response analysis. In the primary screens, the EC₅₀ of reserpine was approximately 20 µM. We doubled the concentration for injection to account for the potential loss of reserpine during the in vivo procedures. As we observed the rescue effect of reserpine in mice, we used the same concentration for rats. The fixed-dose approach was chosen to maintain consistency with previous studies evaluating reserpine in retinal degeneration models and to facilitate comparison across studies. Efforts to identify optimal dosing were deprioritized, as the primary goal was different and this information cannot be directly translated to clinical applications.

We also agree that sex-specific differences in efficacy might be influenced by suboptimal dosing, particularly given potential variations in metabolism, drug distribution, and pharmacokinetics between male and female rats. However, recent pharmacokinetic studies on systemically administered reserpine in rats reported no statistically significant covariates, including body weight, age, breed, or sex, affecting pharmacokinetic (PK) or pharmacodynamic (PD) parameters (Alfosea-Cuadrado, G. M., Zarzoso-Foj, J., Adell, A., Valverde-Navarro, A. A., González-Soler, E. M., Mangas-Sanjuán, V., & Blasco-Serra, A. (2024). Population Pharmacokinetic–Pharmacodynamic Analysis of a Reserpine-Induced Myalgia Model in Rats. Pharmaceutics, 16(8), 1101. https://doi.org/10.3390/pharmaceutics16081101). Furthermore, no evidence of sex-specific differences in reserpine pharmacokinetics has been previously identified in available databases (National Center for Biotechnology Information (2025). PubChem Compound Summary for CID 5770, Reserpine. Retrieved January 13, 2025 from https://pubchem.ncbi.nlm.nih.gov/compound/Reserpine). Importantly, the drug in this study was administered intravitreally, where the ocular compartments are relatively isolated from systemic metabolism or excretion. Under these conditions, where absorption, distribution, metabolism, and excretion have minimal impact, we observed sex differences in efficacy using the same dose of drug.

Nonetheless, we agree with the reviewer and plan to pursue dose-response and other studies in future investigations.

Statistical Analysis

In my opinion, there is room for improvement. How were the animals injected? Was the contralateral eye used as control? (no information in the manuscript about it!, line 390 just mentions the volume and concentration of injections). If so, why not use parametric paired analysis? Why use a non-parametric test, as it is the Mann-Whitney U? The Mann-Whitney U test is usually employed for discontinuous count data; is that the case here?<br /> Therefore, please specify whether contralateral eyes or independent groups served as controls. If contralateral controls were used, paired parametric tests (e.g., paired t-tests) would be statistically appropriate. Alternatively, if independent cohorts were used, non-parametric Mann-Whitney U tests may suffice but require clear justification.

We apologize for the lack of clarity. In line 124, we described the injection as “bilateral intravitreal injections of 5 µL of either vehicle or 40 µM reserpine,” and in Figure 1A, we annotated the bilateral injection as DMSO for both eyes and RSP for both eyes. To address this uncertainty, we added the clarification, “with each group receiving bilateral injections of either vehicle or reserpine” (lines 404–405). Since the results are not paired and involve continuous data for which the normality assumption cannot be confidently met or verified, we used the Mann-Whitney U test for statistical analysis.

Sex-Specific Pathways

The authors do identify pathways enriched in female vs. male retinas but fail to explicitly connect these to the changes in phenotype analysed by ERG and OCT. The lack of mechanistic validation weakens the argument.

The study does not explore why female rats respond better to reserpine. Potential factors such as hormonal differences, retinal size, or differential drug uptake are not discussed.

It remains open, whether observed transcriptomic trends (e.g., proteostasis network genes) correlate with sex-specific functional outcomes.

We acknowledge that, while we identified pathways enriched in female versus male retinas, we did not explicitly connect these findings to the functional phenotypes measured by ERG and OCT. Although our transcriptomic data suggest that reserpine differentially influences pathways such as proteostasis and p53 signaling, we did not conduct mechanistic experiments to validate a causal relationship between these pathways and the observed outcomes.

In practice, designing a study to validate the mechanisms of a small molecule modulating multiple pathways presents significant challenges. If the pathways cannot be specifically modulated or if modulation could result in irreversible outcomes, the mechanistic validation becomes difficult to achieve. Drugs demonstrating mutation-agnostic efficacy are often investigated primarily through outcome measures and the analysis of affected pathways rather than through direct mechanistic validation (Leinonen, H., Zhang, J., Occelli, L. M., Seemab, U., Choi, E. H., L P Marinho, L. F., Querubin, J., Kolesnikov, A. V., Galinska, A., Kordecka, K., Hoang, T., Lewandowski, D., Lee, T. T., Einstein, E. E., Einstein, D. E., Dong, Z., Kiser, P. D., Blackshaw, S., Kefalov, V. J., Tabaka, M., … Palczewski, K. (2024). A combination treatment based on drug repurposing demonstrates mutation-agnostic efficacy in pre-clinical retinopathy models. Nature communications, 15(1), 5943. https://doi.org/10.1038/s41467-024-50033-5).

As recommended, we added potential factors that might influence the differential response to reserpine, based on other studies (lines 353–362) highlighting differences in dopamine storage capacity and estrogen independence. We also added a discussion on the possibility of sex-related differences in basal ERG response levels (lines 363–366).

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The study presents compelling findings on the neuroprotective effects of reserpine in a well-established model of retinitis pigmentosa (P23H-1). The use of ERG, optomotor assays, OCT, immunohistochemistry, and transcriptomic techniques provides a good exploration of the treatment's effects, particularly highlighting the differential response in females. The study underscores the potential of drug repurposing to expedite the availability of therapeutic interventions for patients.

Thanks for your generous comments.

While the manuscript presents an important contribution, I would like to highlight a few points that need clarification or further elaboration to strengthen the work:

(1) Please include the photopic a-wave data in your analysis or provide a justification for its omission. Specifically, it would be valuable to know whether there is an improvement in this parameter under reserpine treatment.

We appreciate the reviewer’s suggestion to include photopic a-wave data in our analysis and acknowledge the importance of this parameter in evaluating cone photoreceptor function. However, we did not analyze the photopic a-wave amplitude in our study because we found the photopic a-wave has low amplitude and high variability, consistent with findings in other studies with P23H-1 rats (Orhan E, Dalkara D, Neuillé M, Lechauve C, Michiels C, et al. (2015) Genotypic and Phenotypic Characterization of P23H Line 1 Rat Model. PLOS ONE 10(5): e0127319. https://doi.org/10.1371/journal.pone.0127319) or even with wild type rats (V.L. Fonteille, J. Racine, S. Joly, A.L. Dorfman, S. Rosolen, P. Lachapelle; Do Rats Generate a Photopic a–Wave? . Invest. Ophthalmol. Vis. Sci. 2005;46(13):2246). We added the description (lines 435-437) explaining why the photopic a-wave was not analyzed. Studies with P23H-1 did not analyze the photopic a-wave, probably for similar reasons.

(2) In Figure 1, it would be helpful to include data from normal control animals to provide a benchmark for retinal degeneration in P23H-1 animals and to better contextualize the effects of reserpine treatment.

Thanks. As suggested, we have included data from normal control animals to Figure 1.

(3) The manuscript states that "Treated female retinas have significantly higher expression of the gene for P62 (SQSTM1), indicating a potential key route for reserpine's activity" (Line 331). Please explain how this difference in expression might translate into a better photoreceptor response in females compared to males.

The difference in P62 (SQSTM1) expression between treated female and male retinas could have important implications for the photoreceptor response. We have identified in our previous study that reserpine increased P62 that mediates proteome balance between ubiquitin-proteasome system (UPS) and autophagy. Together with the role of P62 in the regulation of oxidative stress, P62 might be important for photoreceptor survival and function. Higher expression of P62 in treated females could suggest more efficient cellular maintenance and a better ability to cope with stress, leading to improved photoreceptor survival and function.

(4) Numerous studies have shown that animal models of Parkinson's disease (e.g., those treated with MPTP or rotenone) or retinal tissue from Parkinson's patients exhibit dopaminergic cell death and associated vision loss. Please discuss how these findings relate to your results. Can you hypothesize how dopamine depletion by reserpine may lead to improved photoreceptor responses in your model?

We appreciate the reviewer’s insightful comments. Both MPTP and rotenone act via inhibition of complex I of the respiratory chain, causing cell death and leading to dopamine depletion. In contrast, reserpine acts by inhibiting the vesicular monoamine transporter, depleting catecholamines by preventing their storage and facilitating their metabolism by monoamine oxidase. Although reserpine and other agents can induce animal models of Parkinson's disease, reserpine differs from the others in several aspects: (i) reserpine do not induce neurodegeneration and protein aggregation; (ii) motor performance, monoamine content, and TH staining are partially restored after treatment interruption; and (iii) reserpine lacks specificity regarding dopaminergic neurotransmission (Leão, A. H., Sarmento-Silva, A. J., Santos, J. R., Ribeiro, A. M., & Silva, R. H. (2015). Molecular, Neurochemical, and Behavioral Hallmarks of Reserpine as a Model for Parkinson's Disease: New Perspectives to a Long-Standing Model. Brain pathology (Zurich, Switzerland), 25(4), 377–390. https://doi.org/10.1111/bpa.12253). We have discussed the various effects of catecholamine depletion on retinal diseases (lines 331–337). Both dopamine receptor antagonists and agonists, as well as catecholamine depletion, can exert protective effects on the retina. The reduction in scotopic b-wave amplitude observed at P54, followed by a lack of further progression in degeneration, may support the hypothesis that reduced neuronal activity due to catecholamine depletion could have mitigated damage to retinal neurons.

(5) For readers who may not be familiar with the P23H-1 mutation, it would be beneficial to include a brief description of the timeline and progression of retinal degeneration in this model.

As the progression varies among studies, we have provided our description on observations from the same facility where the animals were housed. The timeline and progression of retinal degeneration are briefly described in the results section (lines 112–115) and Supplementary Figure 1.

(6) Do you have any data on the effects of reserpine treatment in older animals? If available, this could provide additional insight into the potential applicability of reserpine in later stages of disease progression.

Unfortunately, we do not have data from older animals. As described in the results section (lines 116–124), we set the timepoint for interventions before functional impairment peaked, aiming to harness the remaining potential for rescue and promote functional improvement. Our approach focused on developing a gene-agnostic therapy that can delay disease progression and be delivered at an earlier stage than AAV-based therapies, using FDA-approved drugs.

(7) Molecular Basis of Sex Differences: The molecular mechanisms underlying the differential responses in males and females should be elaborated upon. If possible, include a discussion or hypothesis that addresses these sex-specific differences at the molecular level.

We thank the reviewer for highlighting the importance of addressing the molecular basis of sex-specific differences. In our study, we observed distinct transcriptomic responses to reserpine between male and female rats, particularly in molecular pathways related to proteostasis and p53 signaling. While the sex-specific differences in these molecular pathways remain to be fully evaluated, we have added a discussion on sex differences in reserpine responses, incorporating findings from other studies (lines 353–366).

Reviewer #2 (Recommendations for the authors):

(1) There is no mention in the manuscript about the fact that the transgene rats have several copies of rhodopsin and how this can affect these sex differences. Would it be the same in the P23H KO mouse? Or in other models with a single copy of the mutation?

We have described in the Materials and Methods section how they were bred, but we did not specifically mention the allele status in the manuscript. Hemizygous P23H-1 rats used in this study carry a single P23H transgene allele with a transgene copy number of 9, in addition to the normal two wild-type opsin alleles. We added this description to clear the uncertainty (lines 384-387.

(2) This sentence: in abstract lines 26 to 29: "Recently, we identified reserpine as a lead molecule for maintaining rod survival in mouse and human retinal organoids as well as in the rd16 mouse, which phenocopy Leber congenital amaurosis caused by mutations in the cilia-centrosomal gene CEP290 (Chen et al. eLife 2023;12:e83205. DOI: https://doi.org/10.7554/eLife.83205)", to my vew, does not belong to the abstract, maybe in the introduction as stage of art.

Thank you for asking. According to the guidelines for the research advance articles (that follow previously published studies), a reference to the original eLife article should be included in the abstract. As specified in the guidelines, we have updated the citation format to (author, year) for referencing eLife articles (line 29).

(3) Lines 167-170: "Histologic evaluation of the retinas also demonstrated more prominent ONL thinning in the dorsal retina and increased ONL thickness in the dorsal retina measured at 1,000, 1,250, and 1,500 µm distant from the optic nerve head in reserpine-treated group compared with control group (Figure 3C)". I do not understand this sentence. Is it a more prominent thinning or an increased thickness?

We apologize for the confusion caused by this sentence. The histological evaluation showed that ONL thinning was more pronounced in the dorsal retina of control group, which was consistent with OCT findings in Figure 3A. Reserpine treatment increased the ONL thickness in the dorsal retina at specific distances from the optic nerve head (1,000, 1,250, and 1,500 µm). We have revised the sentence for clarity (lines 165-168).

(4) Lines 182-185 and Figure 4B: FL is not the best approach to quantify rhodopsin levels. Since the DAPI staining is overexposed, it is hard to evaluate the staining of RHO in the ONL. From the visible staining in the OS, it is only possible to affirm that the OS are longer in RSP-treated retinas... more is not to be affirmed based on these figures. I suggest using WB.

We acknowledge the reviewer’s concern regarding the use of fluorescence imaging to quantify rhodopsin levels. While our current data highlight structural preservation, such as the length of the outer segments, we agree that drawing conclusions about rhodopsin levels from fluorescence staining is limited. As we do not have samples for WB and fluorescence imaging cannot quantify rhodopsin, we have revised the description (lines 180-184).

(5) Lines 188-190 and Figure 4C: The images in 4C showed an extreme divergence between treated and untreated retina concerning the amount of stained cones, which is not observed at the quantification at 1000µm statistic. Are the images not representative?

We agree with the reviewer that the images in Figure 4C may not adequately represent the quantified data. To address this, we have changed the figure to reflect the quantification results accurately.

(6) Figures 6C-6D and 6G. Why do the authors not use any statistical analysis? Or are the differences not statistically significant? Why do authors use only WT and DMSO controls? What about untreated P23H controls (no DMSO)?

Thanks for checking, and we apologize for the oversight. We have updated figures 5, 6 and S5 to include adjusted p-value in relevant plots. In addition, details of significance threshold are available in supplementary tables. Regarding controls, untreated P23H retinas (without DMSO) were not included in the current analysis, as our experience shows that DMSO injection itself does not cause functional or structural changes. The key data demonstrating the effect of reserpine involve a comparison between the group treated with reserpine and the control group treated with DMSO, as the only difference between these groups is the involvement of the drug.

(7) Validation of findings by testing key genes (e.g., p62/SQSTM1, Nrf2) using qPCR or immunohistochemistry will strengthen the findings.

We appreciate the reviewer’s suggestion to validate key findings using qPCR or immunohistochemistry, as such experiments are crucial for further strengthening our conclusions. While this was not feasible in the current study due to various constraints, we fully recognize their importance and plan to incorporate these in our follow-up studies.

Author response:

The following is the authors’ response to the previous reviews

Response to Public Reviews:

We would like to thank the reviewers and editors once more for their time and effort in reviewing our manuscript. Below we discuss specifically our response to the recommendations of Reviewer 2, which were the only substantial changes we made to the manuscript.

Reviewer 2 recommendation:

"My only remaining suggestion is that the authors acknowledge and cite the work of other groups which have similarly found different subsets of LADs based on various molecular/epigenetic features:

(1) doi.org/10.1101/2024.12.20.629719

(2) PMID: 25995381

(3) PMID: 36691074

(4) PMID: 23124521 (fLADs versus cLADs, as described by the authors themselves) The exact subtypes of LADs might be different based on the features examined, but others have found/implicated the existence of different types of LADs. Hence, the pwv-LAD should be contextualized within these findings (which they do relative to v-fiLADs)."

We thank the reviewer for this suggestion and for these references. We think that the best place to go into depth about how our work relates to these references would be in an appropriate review article.

However, we did read these references carefully and responded, as described below, by adding additional clarifying text in the manuscript as well as mention of articles specifically relevant to our description of our results.

(1) Reviewer 2 wrote specifically, "Hence, the pwv-LAD should be contextualized within these findings (which they do relative to v-fiLADs)"

We are not sure exactly what Reviewer 2 means here. In this manuscript we defined p-w-v iLADs, not LADs. So, it would be inappropriate to compare a subset of iLAD regions with different types of LADs.

If this was the meaning of Reviewer 2, then other readers might have similar confusion. Therefore, we added the following clarifying text in red:

"Several previous studies have used varying approaches to subdivide LADs further into distinct subsets of LADs with different biochemical and/or functional properties (Martin et al., 2024; Meuleman et al., 2013; Shah et al., 2023; Zheng et al., 2015). However, in this Section we focused instead on asking whether regions specifically within iLADs might show differential localization relative to the lamina and/or nucleoli and, if so, whether these regions would show different levels of gene expression. More specifically, analogously to how gene expression hot-zones appeared as local maxima in speckle TSA-seq with early DNA replication timing, we asked whether iLAD regions that appeared as local maxima in lamina proximity mapping signals would correspond to iLAD regions with locally reduced gene expression levels and later DNA replication timing relative to their flanking iLAD sequences. Our rationale was that these iLAD regions might represent chromatin domains that together with their flanking iLAD regions would typically localize well within the nuclear interior but in a fraction of the cell population would loop back and attach at the nuclear periphery."

(2) We also added the following text near the end of the section about p-w-v iLADs to place them in the context of one class of "LADs" identified by ChIP-seq rather than DamID. We use quotation marks since the approach used produced a segmentation that included a nearly 50/50 mix of iLAD and LAD regions, as identified by DamID, for this class of domains.

"We note that in a previous study a three-state Hidden Markov Model (HMM) segmented lamin B ChIP-seq data into two chromatin domain states with extensive overlap with LADs defined by lamina DamID (Shah et al., 2023). Whereas the late replicating, low gene density/expression "T1 LAD" state showed very high overlap (98%) with LADs defined by DamID, the intermediate replicating, intermediate gene expression "T2 LAD" state showed only 47% overlap with LADs defined by DamID. This was partly a result of the HMM segmentation algorithm but also due to substantial differences between the lamina ChIPseq versus DamID signals for reasons that remain unclear. The subset of p-w-v iLADs included in T2 comprise only a small percentage of the total T2 LAD coverage, which includes both other iLAD and LAD regions. Thus, the p-w-v iLADs we identified here represent a novel and distinct class of iLAD chromatin domains, not previously described."

(3) Alternatively, what Reviewer 2 might be suggesting implicitly is that we should start with the regions identified as p-w-v iLADs in one cell type and then identify all of those p-w-v iLADs which instead exist as LADs in a second cell type. Once we have identified their LAD equivalents in a second cell type we could then ask whether they possess special characteristics such that they correspond to a specific type of LAD subset. Finally, we could then ask how that specific type of LAD subset compared to the different subtypes of LADs identified by other groups and, in particular, the references Reviewer 2 provided.

We agree that would be an interesting future direction, but we consider that as outside the scope of this current manuscript. We note that we did no such analysis of the characteristics of LADs which existed as p-w-v iLADs in a different cell line. We save that for a possible future analysis, ideally in the same cell types as used in the cited references to allow a more direct comparison.

(4) Finally, we added text in the Discussion that relates our analysis of the differential SON and LMNB1 TSA-seq signals for different LAD regions, and how these correlate with different histone modifications, with results from the recent preprint cited by Reviewer 2. Note that we could not directly correlate our results from human cells with the three classes of LADs described in MEFs by this preprint.

"Fourth, we show how LAD regions showing different histone marks- either enriched in H3K9me3, H3K9me2 plus H2A.Z, H3K27me3, or none of these marks- can differentially segregate within nuclei. These results support the previous suggestion of different "flavors" of LAD regions, based on the sensitivity of the autonomous targeting of BAC transgenes to the lamina to different histone methyltransferases (Bian et al., 2013). Differential nuclear localization also was recently inferred by the appearance of different Hi-C Bsubcompartments, which similarly were differentially enriched in either H3K9m3, H3K27me3, or the combination of H3K9me2 and H2A.Z (Spracklin et al., 2023). More recently, and while this paper was in revision, a new study described segmenting mouse embryonic fibroblast LADs into three clusters using histone modification profiling (Martin et al., 2024). Interestingly, these three LAD clusters also most notably differed by their dominant enrichment of either H3K9me3, H3K9me2, or H3K27me3. Thus, three orthogonal approaches have converged on identifying different LAD regions showing differential enrichment either of H3K9me3, H3K9me2, or H3K27me3. Here, our use of TSA-seq directly measures and assigns the intranuclear localization of these different LAD regions to different nuclear locales."

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Gray and colleagues describe the identification of Integrator complex subunit 12 (INTS12) as a contributor to HIV latency in two different cell lines and in cells isolated from the blood of people living with HIV. The authors employed a high-throughput CRISPR screening strategy to knock down genes and assess their relevance in maintaining HIV latency. They had used a similar approach in two previous studies, finding genes required for latency reactivation or genes preventing it and whose knockdown could enhance the latency-reactivating effect of the NFκB activator AZD5582. This work builds on the latter approach by testing the ability of gene knockdowns to complement the latency-reactivating effects of AZD5582 in combination with the BET inhibitor I-BET151. This drug combination was selected because it has been previously shown to display synergistic effects on latency reactivation.

The finding that INTS12 may play a role in HIV latency is novel, and the effect of its knockdown in inducing HIV transcription in primary cells, albeit in only a subset of donors, is intriguing. However, there are some data and clarifications that would be important to include to complement the information provided in the current version of the manuscript.

We have now added the requested data and clarifications. In particular, we show that knockout of INTS12 has no effect on cell proliferation (new data added in Figure 2—figure supplement 3)), we clarify how the degree of knockout and the complementation were accomplished, we clarify the differences between the RNA-seq and the activation scores, and we have bolstered the claim that INTS12 affected transcription elongation by performing CUT&Tag on Ser2 phosphorylation of the C-terminal tail of RNAPII along the length of the provirus (new data added in Figure 5C) Please see detailed responses below.

Reviewer #2 (Public review):

Summary:

Identifying an important role for the Integrator complex in repressing HIV transcription and suggesting that by targeting subunits of this complex specifically, INTS12, reversal of latency with and without latency reversal agents can be enhanced.

Strengths:

The strengths of the paper include the general strategy for screening targets that may activate HIV latency and the rigor of exploring the mechanism of INTS12 repression of HIV transcriptional elongation. I found the mechanism of INTS12 interesting and maybe even the most impactful part of the findings.

Weaknesses:

I have two minor comments:

There was an opportunity to examine a larger panel of latency reversal agents that reactivate by different mechanisms to determine whether INTS12 and transcriptional elongation are limiting for a broad spectrum of latency reversal agents.

I felt the authors could have extended their discussion of how exquisitely sensitive HIV transcription is to pausing and transcriptional elongation and the insights this provides about general HIV transcriptional regulation.

We have now added data on latency reversal agents of different mechanisms of action. We show that INTS12 affects HIV latency reversal from agents that affect the non-canonical NF-kB pathway (AZD5582), the canonical NF-kB pathway (TNF-alpha), activation via the T-cell receptor (CD3/CD28 antibodies), through bromodomain inhibition (I-BET151), and through a histone deacetylase inhibitor (SAHA). This additional data has been added to the manuscript in Figure 7, panels B and C as well as adding text to the discussion.

We appreciate the suggestion to extend the discussion to emphasize how important pausing and elongation are to HIV transcription. Additionally, to further support our claim that INTS12KO with AZD5582 & I-BET151 leads to an increase in elongation, that we previously showed with CUT&Tag data showing an increase in total RNAPII seen in within HIV (Figure 5B), we measured RNAPII Ser2 phosphorylation (Figure 5C) and RNAPII Ser5 phosphorylation (Figure 5—figure supplement 2) and added these findings to the manuscript. Upon measuring Ser2 phosphorylation, a marker associated with elongation, we observed evidence of elongation-competent RNAPII in our AZD5582 & I-BET151 condition as well as our INTS12 KO with AZD5582 & I-BET151 condition, as we saw an increase of Ser2 phosphorylation within HIV. Despite seeing elongation-competent RNAPII in both conditions, we only saw a dramatic increase in total RNAPII for our INTS12 KO and AZD5582 & I-BET151 condition (Figure 5B), which supports that there are more elongation events and that an elongation block is overcome specifically with INTS12 KO paired with AZD5582 & I-BET151. This claim is further supported by our data showing an increase in virus in the supernatant only with the INTS12 KO with AZD5582 & I-BET151 condition in cells from PLWH (Figure 6C). We did not observe any statistically significant differences between RNAPII Ser5 phosphorylation, which might be expected as this mark is not associated with elongation (Figure 5—figure supplement 2).

Reviewer #3 (Public review):

Summary:

Transcriptionally silent HIV-1 genomes integrated into the host`s genome represent the main obstacle to an HIV-1 cure. Therefore, agents aimed at promoting HIV transcription, the so-called latency reactivating agents (LRAs) might represent useful tools to render these hidden proviruses visible to the immune system. The authors successfully identified, through multiple techniques, INTS12, a component of the Integrator complex involved in 3' processing of small nuclear RNAs U1 and U2, as a factor promoting HIV-1 latency and hindering elongation of the HIV RNA transcripts. This factor synergizes with a previously identified combination of LRAs, one of which, AZD5582, has been validated in the macaque model for HIV persistence during therapy (https://pubmed.ncbi.nlm.nih.gov/37783968/). The other compound, I-BET151, is known to synergize with AZD5582, and is a inhibitor of BET, factors counteracting the elongation of RNA transcripts.

Strengths:

The findings were confirmed through multiple screens and multiple techniques. The authors successfully mapped the identified HIV silencing factor at the HIV promoter.

Weaknesses:

(1) Initial bias:

In the choice of the genes comprised in the library, the authors readdress their previous paper (Hsieh et al.) where it is stated: "To specifically investigate host epigenetic regulators involved in the maintenance of HIV-1 latency, we generated a custom human epigenome specific sgRNA CRISPR library (HuEpi). This library contains sgRNAs targeting epigenome factors such as histones, histone binders (e.g., histone readers and chaperones), histone modifiers (e.g., histone writers and erasers), and general chromatin associated factors (e.g., RNA and DNA modifiers) (Fig 1B and 1C)".

From these figure panels, it clearly appears that the genes chosen are all belonging to the indicated pathways. While I have nothing to object to on the pertinence to HIV latency of the pathways selected, the authors should spend some words on the criteria followed to select these pathways. Other pathways involving epigenetic modifications and containing genes not represented in the indicated pathways may have been left apart.

(2) Dereplication:

From Figure 1 it appears that INTS12 alone reactivates HIV -1 from latency alone without any drug intervention as shown by the MACGeCk score of DMSO-alone controls. If INTS12 knockdown alone shows antilatency effects, why, then were they unable to identify it in their previous article (Hsieh et al., 2023)? The authors should include some words on the comparison of the results using DMSO alone with those of the previous screen that they conducted.

(3) Translational potential:

In order to propose a protein as a drug target, it is necessary to adhere to the "primum non nocere" principle in medicine. It is therefore fundamental to show the effects of INTS12 knockdown on cell viability/proliferation (and, advisably, T-cell activation). These data are not reported in the manuscript in its current form, and the authors are strongly encouraged to provide them.

Finally, as many readers may not be very familiar with the general principles behind CRISPR Cas9 screening techniques, I suggest addressing them in this excellent review: https://pmc.ncbi.nlm.nih.gov/articles/PMC7479249/.

(1) The CRISPR library used was more completely described in a previous publication (Hsieh et al, PLOS Pathogens, 2023). However, we now more explicitly refer the reader to information about the pathways targeted in the library. We also point out how initial hits in the library lead to finding genes outside of the starting library as in the follow-up screen in Figure 7 where each of the members of the INT complex are interrogated even though only INTS12 was the only member in the initial library.

(2) We understand the confusion between the hits in this paper and a previous publication. Indeed, INTS12 was observed in Hsieh et al., PLOS Pathogens, 2023 as a hit in the Venn diagram of Figure 3B of that paper, and in Figure 5A, right panel of that paper. However, it was not followed up on in the previous paper since that paper focused on a hit that was unique to increasing the potency of one particular LRA. We added text to the present manuscript to make it clear that the screens identified many of the same hits. We have also added additional data here on hit validation to underscore the reliability of the CRISPR screen. In one of the cell lines (5A8), EZH2 was a strong hit (Figure 1B). We have now added data that shows that an inhibitor to EZH2 augments the latency reversal of AZD5582/I-BET151 as predicted from the screen. This data has been added to Figure 1, figure supplement 1.

(3) We appreciate the concern that for INTS12 to be a drug target, it should not be essential to cell viability. We now show that knockout of INTS12 has no effect on cell proliferation (new data added in Figure 2—figure supplement 3). In addition, the discussion now adds additional literature references that describe how knockout of INTS12 has relatively minor effects on cell functions in comparison to knockout of other INT members which supports that the proposal that modulation of INTS12 may be more specific than targeting the catalytic modules of Integrator. Nonetheless, we completely agree with the reviewer that many other aspects of how INTS12 affects T cell functions have not been addressed as well as other potential detrimental effect of INTS12 as a drug target in vivo. We now more explicitly describe these caveats in the discussion but feel that the present manuscript is a first step with a long path ahead before the translational potential might be realized.

(4) We now cite the review of CRISPR screens suggested by the reviewer.

Responses to recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) The authors report in the legend of Figure 2 (and similarly in other figures) that there was "a calculated INTS12 knockout score of 76% (for the one guide used) and 69% (for one of three guides used), respectively." However, it would be helpful to show representative data on the efficiency of INTS12 knockdown in cell lines and primary cells, as well as data on the efficiency of the complementation (Figure 2C).

The knockout scores cited are the genetic assays for the efficiency based on sequence files. As the knockouts are done with multiple guides the knockout for each guide is an underestimate of the total knockout. The complementation, however, was done by adding back INTS12 in a lentiviral vector that also contains a drug resistance marker (puromycin). Cells were then selected for puromycin resistance, and therefore, all of them contain the complemented gene. What one would ideally like is a Western blot to quantify the amount of INTS12 remaining in the knockout pools. Unfortunately, despite obtaining multiple different commercial sources of INTS12 antibodies, we were unable to identify one that was suitable for Western blotting (as opposed to two that did work for CUT&Tag). Nonetheless, the functional data in primary T cells from PLWH and in J-Lat cells lines does show the even if the knockout is suboptimal, we find activation after INTS12 knockout (e.g., Figure 6).

(2) Flow cytometry methods are not reported, but was a viability dye included when testing GFP reactivation (Figure S2)? More broadly, showing data on the viability of cells post-knockdown and drug treatments would help, as cell mortality is inherently associated with latency reactivation in J-Lat cells. For the same reason, reporting viability data would be important for primary cells, as the electroporation procedure can lead to significant mortality.

We did not include viability dyes in the data for GFP activation. However, as described in the public response, we have done growth curves in J-Lat 10.6 cells with and without INTS12 knockout and find no effects on cell proliferation (Figure 2—figure supplement 3). As the reviewer points out, it is not possible to do these experiments in primary cells since the electroporation itself causes a degree of cell death. Nonetheless, we do see effects on HIV activation in these primary cells (Figure 6).

(3) Figure S2 shows a relatively high baseline expression (approximately 15%) of HIV-GFP, which is not unusual for the J-Lat 10.6 clone. However, Figure 3 appears to show no HIV RNA reads in the control condition of this same cell clone. How do the authors reconcile this discrepancy?

We believe that the discrepancies in the flow cytometry versus RNA-seq assays are due to differences in the sensitivity of the assays, the linear range of the assays especially at the lower end, and the different half-lives of RNA versus protein. We now clarify that Figure 3 does not show “no” HIV RNA at baseline, but rather values of ~30 copies per million read counts. This increases to ~800 copies per million read counts when INTS12 knockout cells are treated with AZD5582/I-BET151. These values have the same fold change predicted in Figure 4, and more closely resemble the trend in Figure 2—figure supplement 1.

(4) The combination of AZD5582 and I-BET151 consistently reactivates HIV latency (including GFP protein expression), as previously reported and as shown here by the authors. However, in Figure 5B, RPB3/RNAPII occupancy in the DMSO control appears higher than in the AAVS1KO + AZD5582 and I-BET151 samples. This should be discussed, as it could raise concerns about the robustness of RPB3/RNAPII occupancy results as a proxy for provirus elongation.

As addressed in the public comments, in order to strengthen our claims about transcriptional elongation control, we measured RNAPII Ser2 and Ser5 phosphorylation levels. We see evidence of elongation with Ser2 in the condition of concern (AAVS1 KO + AZD5582 & I-BET151) as well as our main condition of interest (INTS12 KO + AZD5582 & I-BET151) and no change in Ser5 for any condition. With both the Ser2 phosphorylation and total RNAPII as well as our virus release and transcription data we believe that we are seeing evidence of increased elongation with INTS12 KO with AZD5582 & I-BET151. One potential nuance that may not be gathered from the CUT&Tag data is the turnover rate of the polymerase. Despite the levels of RNAPII appearing lower in the condition of concern (AAVS1 KO + AZD5582 & I-BET151) compared to DMSO it is possible that low levels of elongation are occurring but that in our INTS12 KO + AZD5582 & I-BET151 condition there is more rapid elongation and this is why we can observe more RNAPII within HIV. This new data is added in Figure 5C and Figure 5—supplement 2 and its implications are now described in more detail in the discussion.

(5) The authors write that "Degree of reactivation was correlated with reservoir size as donors PH504 (star symbol) and PH543 (upside down triangle) have the largest HIV reservoirs (supplemental Figure S2)." I could not find mention of the reservoir size of these donors in the figure provided.

This confusion was caused by mislabeling of the supplement number, which we fixed, and we added additional labeling to make finding the reservoir size even more clear as this is an important part of the manuscript. This is now found in Supplemental file S4.

Reviewer #3 (Recommendations for the authors):

(1) The MAGeCK gene score is a feature that is essential for the interpretation of the results in Figure 1. The authors do quote the Li et al. paper where this score was described for the first time (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-014-0554-4), however, they may understand that not all readers may be familiar with this score. Therefore a didactic short description of this score should be done when introducing the results in Figure 1.

We have added a short description to the paper to address this.

(2) Figure 4. The authors write: "Among the host genes most prominently affected by INTS12 knockout with AZD5582 & I-BET151 are MAFA, MAFB, and ID2 (full list of genes in supplemental file S3)." I am a bit confused. In the linked Excel file there is only a list of a few genes. The differentially expressed genes appear to be many more from Figure 4. The full list should be uploaded.

We believe there was a mistake in our original uploading and naming of the supplements. We have now double-checked numbering on the supplements and added in text clarification of which excel tabs hold the desired information.

(3) Figure 6: The authors are right in highlighting that there is a high level of variability in viral RNA in supernatants in the early stages of viral reactivation. It is therefore advisable to repeat measurements at Day 7, at which variability decreases and data are more reliable (please, see: https://www.thelancet.com/journals/ebiom/article/PIIS2352-3964(23)00443-7/fulltext).

While it would have been nice to prolong these measurements, our current assay conditions are not optimal for longer term growth of the cells. We note that the measurements were all done in biological triplicates (independent knockouts) and in different individuals. Because the number of activatable latent proviruses is variable and the number of cells tested is limiting, the variability in the assays is expected.

(4) Figure 7: The main genes outside the INTS family should be identified, also.

We include the full list in supplemental file S5 and sort by most enriched.

(5) Methods: A statistical paragraph should be added in the Methods section, detailing the data analysis procedures and the key parameters utilized (for example, which is the MAGeCK gene score threshold that they used to consider knockdown efficacy on HIV latency?).

There is no MAGeCK score threshold that we use to determine efficacy on HIV latency. In a previous publication using CRISPR screens for HIV Dependency Factors (Montoya et al, mBio 2023), we showed that there is a relationship between the MAGeCK and the effect of that gene knockout on HIV replication (Figure 5 that paper). However, it is a continuum rather than a strict threshold and we believe that the effects on HIV latency would respond similarly. In the current paper, we have focused on the top hits rather than a comprehensive analysis of all the entire list. In case the reviewer is referring to the average and standard deviation of the non-targeting controls, we have added this to the figure legend and methods.

Reviewer #2 (Public review):

This paper shows and analyzes an interesting phenomenon. It shows that when people are exposed to sequences of moving dots (That is moving dots in one direction, followed by another direction etc.), that showing either the starting movement direction, or ending movement direction causes a coarse-grained brain response that is similar to that elicited by the complete sequence of 4 directions. However, they show by decoding the sensor responses that this brain activity actually does not carry information about the actual sequence and the motion directions, at least not on the time scale of the initial sequence. They also show a reverse reply on a highly-compressed time scale, which is elicited during the period of elevated activity, and activated by the first and last elements of the sequence, but not others. Additionally, these replays seem to occur during periods of cortical ripples, similar to what is found in animal studies.

These results are intriguing. They are based on MEG recordings in humans, and finding such replays in humans is novel. Also, this is based on what seems to be sophisticated statistical analysis. The statistical methodology seems valid, but due to its complexity it is not easy to understand. The methods especially those described in figures 3 and 4 should be explained better.

Author response:

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

Public Reviews: 

Reviewer #1 (Public review): 

Summary:  

The study identifies two types of activation: one that is cue-triggered and nonspecific to motion directions, and another that is specific to the exposed motion directions but occurs in a reversed manner. The finding that activity in the medial temporal lobe (MTL) preceded that in the visual cortex suggests that the visual cortex may serve as a platform for the manifestation of replay events, which potentially enhance visual sequence learning.  

Strengths: 

Identifying the two types of activation after exposure to a sequence of motion directions is very interesting. The experimental design, procedures, and analyses are solid. The findings are interesting and novel. 

Weaknesses: 

It was not immediately clear to me why the second type of activation was suggested to occur spontaneously. The procedural differences in the analyses that distinguished between the two types of activation need to be a little better clarified.  

We thank the reviewer for his/her summary and constructive feedback on our study. We appreciate the recognition of the strengths of our study.

The second type of activation, namely the replay of feature-specific reactivations, is considered spontaneous because it reflects internally driven neural processes rather than responses directly triggered by external stimuli. Unlike responses evoked by stimuli, spontaneous replay is not time-locked to stimulus onset. Instead, it arises from the brain's intrinsic activity, typically observed during offline periods (e.g., rest or blank period) when external stimuli are absent. This allows the neural system to reactivate and consolidate prior experiences without interference from ongoing external stimuli.

Replay is believed to be a key mechanism underlying various cognitive functions, such as memory consolidation (Gillespie et al., 2021; Gridchyn et al., 2020), learning (Igata et al., 2021), prediction and planning (Ólafsdóttir et al., 2018). Furthermore, the hippocampus and related cortical areas engage in replay to extract abstract relationships from sequential experiences, forming a "template" that can generalize across contexts (Liu et al., 2019). In our study, the feature-specific replay observed during blank periods likely reflects this process, supporting the integration of exposed motion direction sequences into cohesive memory representations and facilitating visual sequence learning.

We have extended the Discussion section to incorporate this explanation (Lines 440 - 447).

Regarding the second question, the procedural differences between the two types of activations lie in the classifiers used for the two analyses: a multiclass classifier for non-specific elevated responses and binary classifiers for feature-specific replay. 

For the non-feature-specific elevated responses, we trained a five-class (with the labels of the four RDKs and the ITI (inter-stimulus interval)) classifier on the localizer data and tested on the blank period in the main phase. We attempted to decode motion direction information at each time point at the group level. However, the results revealed no feature-specific information at the group level during the blank period.

For the feature-specific replay, we employed the temporal delayed linear modeling (TDLM) to examine whether individual motion direction information was encoded in a sequential and spontaneous manner. Here, we first needed to train four binary classifiers, each was sensitive to only one motion direction (i.e., 0°, 90°, 180°, or 270°), as our aim was to quantify the evidence of feature-specific sequence in the subsequent analyses. For each classifier, positive instances were trials where the corresponding feature (e.g., 0°) was presented, while negative instances included trials with other features (e.g., 90°, 180°, and 270°) and an equivalent amount of null data from the ITI period (1–1.5 s).

We have clarified these methodological details in the Methods section (Pages 34 – 41).

Reviewer #2 (Public review): 

This paper shows and analyzes an interesting phenomenon. It shows that when people are exposed to sequences of moving dots (that is moving dots in one direction, followed by another direction, etc.), showing either the starting movement direction or ending movement direction causes a coarse-grained brain response that is similar to that elicited by the complete sequence of 4 directions. However, they show by decoding the sensor responses that this brain activity actually does not carry information about the actual sequence and the motion directions, at least not on the time scale of the initial sequence. They also show a reverse reply on a highly compressed time scale, which is elicited during the period of elevated activity, and activated by the first and last elements of the sequence, but not others. Additionally, these replays seem to occur during periods of cortical ripples, similar to what is found in animal studies. 

These results are intriguing. They are based on MEG recordings in humans, and finding such replays in humans is novel. Also, this is based on what seems to be sophisticated statistical analysis. However, this is the main problem with this paper. The statistical analysis is not explained well at all, and therefore its validity is hard to evaluate. I am not at all saying it is incorrect; what I am saying is that given how it is explained, it cannot be evaluated. 

We thank the reviewer’s detailed evaluation as well as the acknowledgment of the novelty of our study.

To address the concern about the statistical analysis, in the revised manuscript, we have modified the Methods section to provide a more detailed explanation of the analytical pipeline, particularly for several important aspects such as decoding probability and TDLM. (Lines 646 – 657, Lines 682 – 734). 

Below, we provide point-by-point responses to further elaborate on these revisions and address the reviewer’s comments.

Recommendations for the authors: 

Reviewer #1 (Recommendations for the authors): 

I have questions.  

(1) Participants were exposed to a predefined sequence of motion directions either clockwise or counterclockwise. Is it possible that the observed replay is related to the activation of MST neurons? If a predetermined sequence is not in either clockwise or counterclockwise but is randomly determined like 0{degree sign}->180{degree sign}->270{degree sign}->90{degree sign}, would the same result be obtained?  

We thank the reviewer for these thoughtful questions.

First, regarding the potential involvement of MST neurons, it is plausible that the observed replay might involve activity in motion-sensitive brain regions, including the medial superior temporal (MST) and even middle temporal (MT) areas. MST neurons, located in the extrastriate visual cortex, are highly direction-selective and are known for their sensitivity to complex motion patterns, such as rotations and expansions (Duffy & Wurtz, 1991; Saito et al., 1986). In our experiment, the use of RDKs with four distinct motion directions might elicit responses in MST neurons. However, due to the limited spatial resolution of MEG, we cannot provide direct evidence for this claim. 

Second, regarding the impact of randomly ordered sequences, we believe that the replay patterns would still occur even if the sequences were randomly ordered (e.g., 0° → 180° → 270° → 90°). After a sequence is repeatedly exposed, the hippocampus has the capacity to encode abstract relationships in the sequence. Evidence supporting this view comes from previous studies. For example, Liu et al., (2019) showed that replay does not merely recapitulate visual experience but can also follow a sequence implied by learned abstract knowledge. In their study, participants were instructed that viewing pictures C→D, B→C, and A→B implies a true sequence of A→B→C→D. During subsequent testing, they observed replay events following this learned true sequence, even with novel visual stimuli, indicating that the brain maintains sequence knowledge independent of specific stimuli. Similarly, Ekman et al., (2023) showed that prediction-based neural responses could be observed when moving dots were presented in a random order rather than in a clockwise or counterclockwise order, which correspond to the four motion directions in our study. 

Together, these studies suggest that replay mechanisms in the brain are flexible and can encode and reproduce abstract relationships between sequential stimuli, regardless of their specific spatial contents. Therefore, we believe that even if the sequence were randomly ordered, the same backward replay pattern would still be observed.

(2) Is it possible that the motion direction non-specific responses actually reflect the replay of another feature of the exposed sequence, namely, the temporally rhythmic presentations of the sequence, rather than suggested in the discussion?  

We thank the reviewer for raising this insightful possibility.

There is substantial evidence that rhythmic stimulation can entrain neural oscillations, which in turn facilitates predictions about future inputs and enhances the brain's readiness for incoming stimuli (Barne et al., 2022; Herrmann et al., 2016; Lakatos et al., 2008, 2013). In our study, the temporally rhythmic presentation of the motion sequence may have entrained oscillatory activity in the brain, leading to periodic activation of sensory cortices. This rhythmic entrainment could account for the observed nonspecific responses by reflecting the brain's temporal predictions rather than specific feature replay. 

It is important to note that, however, this interpretation is in line with our initial explanation that the non-feature-specific elevated responses likely reflect a general facilitation of neural processes for any upcoming stimuli, rather than being tied to specific stimuli. The rhythmic entrainment mechanism provides another way to understand how the temporal structure in the sequences might contribute to the non-feature-specific elevated responses.

We have revised the Discussion section to incorporate this interpretation, providing a more comprehensive account for the non-feature-specific elevated responses (Lines 428 – 439).

Reviewer #2 (Recommendations for the authors): 

The main problem with the paper is that the sophisticated statistical methodology is not explained well and therefore its validity is hard to evaluate. I am not at all saying it is incorrect, what I am saying is that given how it is explained, it cannot be evaluated.  

See below for detailed point-by-point responses.  

The first part is clear. There are 4 directions of motion, and there can also be a blank screen. The random decoding accuracy would be 20%. The decoding methods from the sensors yielded a little above 50% accuracy. This is clearly about chance, but much less than one would get from electrode recording of motion-selective cells in the cortex. However, the concept and methods used here seem clear, in contrast to what comes next.  

Indeed, in the first step, we aimed to validate the reliability of our decoding model by applying a leave-one-out cross validation scheme to the localizer data. Our results showed that the decoding accuracy exceeded 50%, demonstrating robust decoding performance. However, due to the noninvasive nature of MEG and its low spatial resolution, the recorded signals represent population-level activity that inherently includes more noise compared to electrode recordings of motion-selective neurons. Therefore, the decoding accuracy in our study is understandably lower than that obtained with electrode recordings.

Next, and most of the paper relies on this concept, they use the term decoding probability (Figure 2). What is the decoding probability measure (Turner 2023)? This is not explained in the methods section. I scanned the Turner et al 2023 paper referenced and could not find the term decoding probability there. In short, I have no idea what this means. What are these numbers between 0-0.3? How does this relate to accuracies above 50% reported? This is an important concept here, and it is used throughout the paper, so it makes it hard to evaluate the paper.  

We apologize for the lack of clarity in our explanation of the term "decoding probability." Specifically, we used a one-versus-rest Lasso logistic regression model trained on the localizer data to decode the MEG signal patterns elicited by each motion direction during the main phase. The trained model could be used to predict a single label at each time point for each trial (e.g., labels 1 – 4 correspond to the four motion directions and label 5 corresponds to the ITI period). By comparing the predicted label with the true label across test trials, we could compute the time-resolved decoding accuracy as final reports.

Alternatively, rather than predicting a single label for each time point and each trial, the model can also output the probabilities associated with each label/class (e.g., we used the predict_proba function in scikit-learn). This results in a 5-column output, where each column represents the probability of the corresponding class, and the sum of the probabilities across the five columns equals 1. Finally, at each time point, averaging these probabilities across trials yields five values that indicate the likelihood of the predicted stimulus belonging to each class.

For example, Figure 2 in the manuscript depicts the decoding probabilities for the four RDKs (the probabilities for the ITI class are not shown in the figure). The number in a cell (between 0 and 0.3) indicates the probability of each class at a given time point (Figure 2A). The decoding probability does not have a direct relationship with the decoding accuracy. However, since there are five classes, the chance level of the decoding probability is 0.2. The highest probability among the five classes at a given time point determines the decoded label when computing the decoding accuracy.

For illustration, in the left panel of Figure 2B, at the onset of the first RDK (0 s), the mean decoding probabilities for the classes 0°, 90°, 180°, 270°, and the blank ITI are 5%, 4.1%, 4.0%, 4.5%, and 82.4%, respectively. Thus, the decoded label should be the blank ITI. In contrast, 0.4 s after the onset of the first RDK, the mean decoding probabilities for the five classes are 28.0%, 19.0%, 22.8%, 21.2%, and 9.0%, respectively. Therefore, the decoded label should be 0°.

We have revised the Methods section to explain this issue (Lines 646 – 657).

They did find compressed reversed reply events (Figures 3-4). This is again confusing for several reasons. First, because they use the same unexplained decoding probability measure. Second, the optimal time point defined above depends on the start time of a stimulus, but here the start time is random. Third, the TDLM algorithm is hard to understand. For example, what are the reactivation probabilities of Figure 3C? They do make an effort to explain this in the methods section (lines 652-697) but it's not clear enough from the outset. For example, what does the state X_j is this a vector of activity of sensors? Are these decoding probabilities of the different directions? What is it? Also, what is X_i vs X_i(\Delta t)? Frankly, despite their efforts, I am very confused. Additionally, the figures use the term reactivation probability, where is it defined? So again, the results seem interesting, but the methods are not explained well at all.  

This paper must better explain the statistical methods so that they can be evaluated. This is not easy, these are relatively complex methods, but they must be explained much better so the validity of the paper can be examined.  

Regarding the optimal time point, we defined it as the time point with the highest decoding accuracy, determined during the validation of the localizer data using a leave-one-out cross-validation scheme. This optimal time point was participant- and motion-direction-specific, as the latency to achieve the peak decoding accuracy varied across individuals and motion directions. For group-level visualization, we circularly shifted the data over time, aligning each optimal time point to a common reference point (arbitrarily set at 200 ms after stimulus onset). Importantly, however, these time points are unrelated to the data in the main phase, as the models were trained using the independent localizer data and then applied to each time point during the blank period in the main phase.

Regarding the TDLM algorithm, detailed descriptions of the algorithm have been provided in the revised Methods section (Line 683 – 735). Furthermore, we have included explanatory notes in the main text and figure legend to provide immediate context for terms such as "reactivation probability" (Lines 247 – 248, Lines 275 – 276).

This paper uses MEG in humans, a non-invasive technique. This allows for such results in humans. Indeed (if the methods are correct) these units can be decoded to provide statistically significant estimates of motion direction. Note, however, that the spatial resolution of MEG is limited. The decoding accuracies of above 50% are way above chance. Note however that if actual motion-sensitive neurons (e.g. area MT) were recorded, and even if the motion is far from 100% coherence, the decoding accuracy would approach 100%. 

We agree with the reviewer that decoding accuracy would approach 100% if single-neuron data from motion-sensitive areas (e.g., area MT) were recorded, given the exceptionally high signal-to-noise ratio (SNR) of such data. However, two considerations inform the methodology of our study.

First, while single-neuron recordings provide invaluable insights, acquiring such data in humans is both ethically challenging and logistically impractical.

Non-invasive MEG, by contrast, offers a practical alternative that can achieve robust decoding of population-level activity with a reasonable SNR.

Second, the primary goal of our study was not merely to achieve high decoding accuracy but also to examine the replay of an exposed motion sequence in the human visual cortex. To achieve this, we first needed to train feature-specific models that can be used to decode the spontaneous reactivations of the four motion directions during the blank period. The ability to distinguish representations of the four motion directions was essential for calculating the “sequenceness” of the exposed motion sequence in the TDLM algorithm. While the absolute decoding accuracy of MEG data may not match that of single-neuron data, an important outcome was the successful construction of feature-specific models for the four motion directions (Figure 3B in the manuscript). These models provided a robust foundation for investigating sequential replay in the brain. These results also align with the broader goal of leveraging MEG data to study dynamic neural processes in humans, even in the face of its spatial resolution limitation.

Minor:  

(1) Line 246 - there is no figure S2A, subplots are not labeled.  

We have corrected this in the revised manuscript.

(2) Is Figure 3B referred to in the text? Same for 3C. This figure is there for explaining the statistical models used, but it is not well utilized.

We have modified the text to clarify this issue in the revised manuscript.

(3) English:  

There are problems with the use of English in the paper, this should be corrected in the next version. A few examples are below.  

Noises -> noise  

- "along the motion path in visual cortex" What does this sentence mean? Is this referring to motion-sensitive areas in the brain? Please clarify.  

There are many other examples. This is minor, but should be corrected.

We have corrected these errors in the revised manuscript.

References

Barne, L. C., Cravo, A. M., de Lange, F. P., & Spaak, E. (2022). Temporal prediction elicits rhythmic preactivation of relevant sensory cortices. European Journal of Neuroscience, 55(11–12), 3324–3339. https://doi.org/10.1111/ejn.15405

Ekman, M., Kusch, S., & de Lange, F. P. (2023). Successor-like representation guides the prediction of future events in human visual cortex and hippocampus. eLife, 12, e78904. https://doi.org/10.7554/eLife.78904

Gillespie, A. K., Maya, D. A. A., Denovellis, E. L., Liu, D. F., Kastner, D. B., Coulter, M. E., Roumis, D. K., Eden, U. T., & Frank, L. M. (2021). Hippocampal replay reflects specific past experiences rather than a plan for subsequent choice. Neuron, 109(19), 3149-3163.e6. https://doi.org/10.1016/j.neuron.2021.07.029

Gridchyn, I., Schoenenberger, P., O’Neill, J., & Csicsvari, J. (2020). AssemblySpecific Disruption of Hippocampal Replay Leads to Selective Memory Deficit. Neuron, 106(2), 291-300.e6. https://doi.org/10.1016/j.neuron.2020.01.021

Herrmann, B., Henry, M. J., Haegens, S., & Obleser, J. (2016). Temporal expectations and neural amplitude fluctuations in auditory cortex interactively influence perception. NeuroImage, 124, 487–497. https://doi.org/10.1016/j.neuroimage.2015.09.019

Igata, H., Ikegaya, Y., & Sasaki, T. (2021). Prioritized experience replays on a hippocampal predictive map for learning. Proceedings of the National Academy of Sciences, 118(1), e2011266118. https://doi.org/10.1073/pnas.2011266118

Lakatos, P., Karmos, G., Mehta, A. D., Ulbert, I., & Schroeder, C. E. (2008). Entrainment of Neuronal Oscillations as a Mechanism of Attentional Selection. Science, 320(5872), 110–113. https://doi.org/10.1126/science.1154735

Lakatos, P., Musacchia, G., O’Connel, M. N., Falchier, A. Y., Javitt, D. C., & Schroeder, C. E. (2013). The Spectrotemporal Filter Mechanism of Auditory Selective Attention. Neuron, 77(4), 750–761. https://doi.org/10.1016/j.neuron.2012.11.034

Liu, Y., Dolan, R. J., Kurth-Nelson, Z., & Behrens, T. E. J. (2019). Human Replay Spontaneously Reorganizes Experience. Cell, 178(3), 640-652.e14. https://doi.org/10.1016/j.cell.2019.06.012

Ólafsdóttir, H. F., Bush, D., & Barry, C. (2018). The Role of Hippocampal Replay in Memory and Planning. Current Biology, 28(1), R37–R50. https://doi.org/10.1016/j.cub.2017.10.073

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This study by Zimyanin et al. examines the role of the C. elegans chromokinesin KLP-19 in the formation and architecture of the anaphase central spindle in C. elegans zygotes. Through a combination of electron and light microscopy, along with RNAi-mediated perturbations, the authors propose that KLP-19 influences central spindle stiffness by regulating microtubule dynamics.

In Figure 5, the effect of KLP-19 depletion on central spindle microtubules appears unconvincing. The FRAP results show no significant difference with or without KLP-19, and overall microtubule density does not consistently respond to its depletion. Additionally, the double klp-19; gpr-1/2 (RNAi) condition does not exhibit a strong increase in microtubule density, though a statistical test is missing for this condition. Furthermore, the spd-1; gpr-1/2 double depletion produces a similar increase in microtubule density to most klp-19 depletion conditions, suggesting that the effect cannot be solely attributed to the absence of KLP-19.

Figure 5A shows that depletion of KLP-19 leads to an increase in tubulin signal in the spindle midzone. The reviewer is correct, that there are differences in the microtubule density between KLP-19 depletion alone and KLP-19 + GPR-1/2 depletion. While depletion of KLP-19 alone leads to a significant increase, co-depletion of GPR-1/2 and KLP-19 leads to a slight, but not significant increase. Along this line, we have added Supplementary Table 1 that contains all p-Values for the different conditions for Figure 5A. However, depletion of GPR-1/2 alone does not affect the microtubule density in the midzone, arguing that changes in pulling forces do not affect the microtubule density in the midzone. It is possible, that the double RNAi leads to a decrease in efficiency and thus a reduced effect on microtubule intensity. We will demonstrate the RNAi efficiency by western blot. Another possibility is that there are some feedback mechanisms that responds to presence/ absence of pulling forces and some of our data (not from this manuscript) hints in this direction, but we have not yet worked out the details of this. We are planning to publish this in a follow up publication.

• *

In response to the spd-1 + gpr-1/2 (RNAi), the reviewer is correct, that the microtubule density in the midzone is not significantly different from klp-19 (RNAi) conditions and we think it is interesting to note that spd-1 + gpr-1/2 (RNAi) leads to an increased microtubule density in the midzone. This could be, as above mentioned caused by some feedback mechanisms that responds to pulling forces, or also due to some functions of SPD-1 that affects microtubules in the midzone. Interestingly, our data also shows that metaphase spindles are significantly shorter in the absence of SPD-1 in comparison to spindles in control embryos, suggesting that SPD-1 plays a role in regulating microtubules or force transmission. We are currently working on understanding SPD-1's role in this process.

• *

We also agree that there is no significant effect on the microtubule turn-over as shown in Figure 5B and we have stated this in the text. Our data does show a trend to a decreased turn-over, but the difference is not significant. This could be due to the low sample number.

• *

Overall, we think our data, the light microscopy and even more so the EM data does show a clear effect on midzone microtubules.

• *

The use of hcp-6 depletion to argue that KLP-19 depletion affects central spindle elongation independently of stretched chromatin is problematic. hcp-6 encodes a component of the Condensin II complex in C. elegans, and its depletion leads to chromatin decompaction rather than the stretched, dense chromatin observed in the midzone during anaphase in klp-19 (RNAi) embryos. These conditions are not equivalent and do not effectively rule out the possibility that chromatin stretching contributes to the observed phenotype.

We agree with the reviewer that the HCP-6 experiments do not entirely rule out effects from lagging chromosomes. Proving that the reduced spindle and chromosome separation is not due to lagging chromosomes is challenging. Most of the depletions that lead to lagging chromosomes are based on defective kinetochore microtubule connections, such as depletion of KNL-1, NDC-80 or CLS-2 (CLASP). In C. elegans, this leads to the mass of Chromosomes staying behind in anaphase and increased spindle pole separation, which is not comparable to KLP-19 depletion. Perturbations that do not affect kinetochore microtubules but still lead to lagging chromosomes are often targeting cohesin or condensin. Ultimately none of these conditions are directly comparable.

A probably better way to test this would be to deplete KLP-19 only after metaphase to prevent its effect on chromosome alignment. However, this is currently not possible as the time window is about 1 minute or less. We currently do not have the tools to conduct this type of experiment. As other reviewers also criticized this experiment and its significance for the paper, we have removed this entirely and have added the following part to the discussion about the potential effect of lagging chromosomes:

" *We can not unambiguously rule out that failure to properly align chromosomes and the resulting lagging chromosomal material could also lead to some of the observed effects on spindle dynamics, such as slow chromosome segregation and pole separation rates as well as preventing spindle rupture in absence of SPD-1. However, several observations argue in favor of KLP-19 actively changing the midzone cytoskeleton network and thus affecting spindle dynamics. *

Most of the protein depletions in C. elegans that lead to lagging chromosomes are based on defective kinetochore microtubule connections, such as depletion of CeCENP-A, CeCENP-C, KNL-1 or NDC-80 (70-72). This mostly leads to the Chromosome mass staying behind in anaphase and increased spindle pole separation (70-72), which is not comparable to KLP-19 depletion. Perturbations that do not affect kinetochore microtubules but still lead to lagging chromosomes are often targeting cohesin or condensin, which depletion leads to chromatin decompaction (73-74) rather than the stretched, dense chromatin as observed in the midzone during anaphase in klp-19 (RNAi) embryos. Ultimately none of these conditions are directly comparable, making it difficult to completely rule out an effect of lagging chromosomes. A better way to test this would be to deplete KLP-19 only after metaphase to prevent its effect on chromosome alignment. However, this is currently not possible as the time window is about 1 minute or less and we do not have the tools to conduct this type of experiment.

*Based on our results we hypothesize that the observed spindle dynamics in absence of KLP-19 are not only caused by lagging chromosomes. Instead, KLP-19 RNAi results in a global rearrangement of the spindle and leads to a significant reduction of the spindle size, microtubule overlap, growth rate, and stability. Furthermore, the increase of microtubule interactions after klp-19 (RNAi) could also contribute to lagging of chromosomes and exacerbation of fragmented extrachromosomal material." *

Additionally, the authors report that KLP-19 influences astral microtubule dynamics (Figure 5E), yet in Figure 3E, they show that KLP-19 localizes exclusively to kinetochores and spindle microtubules, excluding astral microtubules and spindle poles. How do they reconcile this apparent contradiction?

We think that KLP-19 localizes also to astral Microtubules. Our KLP-19 GFP CRISPR line is very dim and this makes it hard to see. We are proposing to use a TIRF approach to image KLP-19 GFP on the C. elegans cortex, which we will include in the revised version. In addition, in support of our hypothesis of KLP-19 binding to astral Microtubules as well we would like to note that there is a PhD thesis available from Jack Martin in Josana Rodriguez Sanchez's Lab in Newcastle (LINK, will lead to a download of the thesis! ) that has reported KLP-19s localization to cortical Microtubules in C. elegans. In this thesis the author also reports an effect on astral microtubule growth.

Figure legends lack consistency and do not adhere to standard C. elegans nomenclature conventions (e.g., protein names should not be capitalized, and genetic perturbations should be italicized). Standardizing these elements would improve clarity and readability.

We have checked our figure legend and to our best knowledge the legends adhere to the C. elegans nomenclature. All RNAi conditions are lower case italicized and Protein names are capitalized as it is standard in other C. elegans publications. We have however noticed some variation in our Figures, i.e. EB-2 instead of EBP-2 and have corrected this in all figures.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Zimyanin et al, Chromokinesin Klp-19 regulates microtubule overlap and dynamics during anaphase in C. elegans.

The authors used a myriad of techniques, including confocal live-cell imaging, 2-photon microscopy, second harmonic generation imaging, FRAP, microfluidic-coupled TIRF, EM-tomography, to study spindle midzone assembly dynamics in C. elegans one-cell stage embryos. In particular, they illuminated the role of kinesin-4 KLP-19 in maintaining proper midzone length and organization. Inhibition of KLP-19 results in longer more stable midzones, implying KLP-19 functions in depolymerizing microtubules.

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Minor comments:

Fig 3E / There is an unusual diagonal line bisecting the embryo. Visually this does not affect viewing of the His::GFP and KLP-19::GFP signals. However, when these signals are quantified and normalized (as in Fig 3F), the diagonal bisect displaying different background signal may impact the measurements.

We are very sorry about this line in the images. The line is due to a defect in the camera chip of the spinning disc. We will acquire new images for this Figure using our new spinning disc microscope.

Fig 4B / While the kymographs clearly show KLP-19::GFP motility on microtubules, they also show that the majority of KLP(-::GFP do not move. Perhaps some quantification and discussion of this result is appropriate?

The reviewer is correct that only a small fraction small fraction of molecules, maybe ~10%, moves. We will add this quantification to the paper and discussion. This could be due to several reasons: Many of the non-moving particles are not visibly colocalized with microtubules, which could mean they are sticking non-specifically to the surface (or sticking to small tubulin aggregates that aren't long enough to support movement). In addition, as this experiment is done in a lysate it is hard to interpret if the immobile KLP-19 is not moving because other proteins are bound along the microtubule blocking its way or if the KLP-19 requires some activation (i.e. phosphorylations) to become mobiles. We think this could be very interesting and will follow up on this in the future.

• *

Reviewer #2 (Significance (Required)):

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary:

The anaphase spindle midzone is an essential structure for cell division. It consists of antiparallel overlapping microtubules organized by the antiparallel microtubule bundler PRC1, molecular motors and other regulatory proteins. This manuscript investigates the role of KLP-19 (C. elegans ortholog of human kinesin-4 KIF4A) and SPD-1 (C. elegans ortholog of PRC1) for spindle midzone organization in the C. elegans embryo and its relevance for proper spindle function. Advanced fluorescence microscopy, 3D electron tomography, and a fluorescence microscopy-based single molecule assay in embryo lysate are used in a unique combination. The authors confirm several aspects of PRC1 and KIF4A function in anaphase, as reported in previous work, mostly in human cells and Drosophila embryos and also in C. elegans embryos. Measurements are mostly very quantitative and to a high quality standard. The main difference to previous conclusions is that here, the authors propose that KLP-19 does not interact with SPD-1, in contrast to what has been established for other animal kinesin-4s and PRC1, and instead localizes to the spindle midzone independently of PRC1 by a mechanism that remains unknown. The authors provide evidence that KLP-19 nevertheless controls microtubule overlap length as in other species and that it produces outward forces sliding midzone microtubules apart a movement that SPD-1 counteracts (presumably by friction). The manuscript presents a rich resource of carefully measured quantitative structural and dynamic C. elegans anaphase spindle data.

Major comments:

Key conclusions convincing?

(1) The key conclusions that the length of the central anaphase spindle microtubule overlap remains constant as the C.elegans spindle elongates (Fig. 1), that PRC1 indeed localizes quite precisely to these overlaps as previously assumed based on its in vitro (purified protein) behavior (Fig. 3B) and that the kinesin-4 KLP-19 controls overlap length as in other species (Fig. 3B) are all convincingly shown. What's missing are quantitative KLP-19 together with microtubule polarity profiles in the presence and absence of SPD-1, leaving it unclear to which extent this kinesin localizes to microtubule overlaps in the two situations. Such data seem crucial, given the authors' claim that KLP-19 localizes to the midzone and that this localization of KLP-19 is mostly unaffected by the absence of SPD-1.

If we understand this correctly the reviewer is asking for second harmonic imaging (SHG) together with imaging of KLP-19 GFP. This is currently not possible due to the way this imaging must be done (2-photon of GFP-Tubulin followed by the SHG). The only thing we can do is provide KLP-19 GFP profiles for control and SPD-1 depleted embryos. We can also use the line co-expressing SPD-1 Halo-tag and KLP-19 GFP to plot their respective localizations in control conditions. We are happy to provide such plots. Generally, we see KLP-19 going to the midzone in absence of SPD-1 and the SHG data does show that the overlap is increased. If KLP-19 specifically localizes to microtubule overlap (rather to i.e. microtubule ends) can currently not be distinguished in the spindle midzone. In vitro data from other labs and our in vitro assay suggests that KLP-19 does not specifically bind to antiparallel overlaps but rather microtubules in general.

(2) 'Normalized KLP-19 intensities' are used to demonstrate that the total amount of this kinesin localizing to the spindle midzone does not depend on the presence of SPD-1 (Fig. 3F). Given that this claim represents a major novelty of the study, the efficiency of the SPD-1 knock-down should be documented, ideally by western blot and fluorescence microscopy.

We agree with the reviewer and will provide western blots.

(3) The authors show convincingly that the kinesin KLP-19 contributes to outward microtubule sliding (and can contribute to spindle rupture in the absence of SPD-1) (Fig. 2), which is interesting and in line with the author's main claim.

(4) The interaction between KIF4a and PRC1 is well established in other species and has been clearly demonstrated both in cells and in vitro (with purified proteins). The authors claim that this concept does not apply to the C. elegans orthologs. To show 'in vitro' (outside of the spindle) that the C. elegans homologs KLP-19 and SPD-1 do not interact, the authors use a novel microfluidic fluorescence-based single-molecule assay in lysate (Fig. 4). Although very original, these experiments do not reach the biochemical standard of previous experiments with purified proteins without appropriate controls. Given that the lysate setup is fairly novel, it's advisable to present at least one positive control demonstrating that interactions between soluble proteins can indeed be detected using this assay. It would also be useful to show the absence of interaction between KLP-19 and SPD-1 by a more conventional method like co-IP, again with a positive control, to support the authors' claim. Eventually, experiments with purified proteins will have to unequivocally demonstrate whether KLP-19 and SPD-1 indeed do not interact - something which appears, however, to be beyond the scope of this study. Without additional experimental proof, the authors may want to indicate that these results are of more preliminary nature.

*We agree with the reviewer, and we will conduct co-IPs of SPD-1 and KLP-19. We will also add CYK-4 as a positive control as previous publications have shown the interaction of CYK-4 with SPD-1. We are now generating lines co-expressing CYK-4 GFP and SPD-1 Halo-tag for the co-IP experiments. *

(5) Unfortunately, the authors do not propose an alternative mechanism for KLP-19 localization to the midzone in SPD-1 depleted embryos, limiting the conceptual advance. Does KLP-19 bind directly to antiparallel microtubules, for example in the assay presented in Fig. 4 (where signs of microtubule crosslinking are shown for SPD-1)? If not, how would it accumulate in the midzone (if it does) in the C. elegans embryo anaphase spindle? The authors do also not propose a mechanism explaining why central antiparallel microtubule overlap length does not change as the spindle elongates in anaphase. Moreover, there is no discussion regarding the potential mechanism leading to KLP-19 controlling microtubule dynamics globally instead of locally where the motor accumulates, indicating limitations in mechanistic insight.

*We agree with the reviewer and will add these points to the discussion. *

*To address some of the points: *

*How does KLP-19 end up in the midzone? : Our data shows that localization of KLP-19 does depend on AIR-2 and BUB-1 as previously reported. However, those proteins primarily affect the formation of the midzone. The in vitro assay does not suggest that KLP-19 has a preference for overlaps, unlike SPD-1, but rather binds microtubules in general. One possible mechanism of midzone localization could be microtubule end-tagging, as has been suggested for PRC1 (SPD-1 homolog). This could lead to an accumulation of KLP-19 in the midzone. *

Why does the central overlap stay constant? : This can be explained by constant microtubule growth at the plus-ends why maintaining the overlap length. Alternatively, this could be explained by some (sophisticated) rearrangements of microtubules that ensure the overlap length stays the same. Generally, this is a very interesting question, as each of this scenario still requires that the overlap length is tightly regulated. Our data suggests that this is correlated with microtubule length in the midzone, as KLP-19 depletion leads to longer microtubules and overlap. This suggests that the regulation of microtubule dynamics might be an important factor in this process. We will add this to the discussion.

• *

Potential mechanism leading to KLP-19 controlling microtubule dynamics globally: We think that KLP-19 localizes to spindle and astral microtubules and regulates the dynamics on all of those, leading to a global regulation. By increasing it's concentration locally, microtubule dynamics can be regulated in the midzone. We will add data showing the localization of KLP-19 to astral microtubules.

Claims justified/preliminary and clearly presented?

The observation that the spindle length remains constant throughout anaphase in C. elegans is based on elegant, but unconventional fluorescence microscopy data (Fig. 1A & B). It would be helpful to add images of SHG and two-photon microscopy to help the reader understand the graphs. Measurements are presented based on distances between the poles. It is unclear why the distances between 15-20 µm were chosen and how they translate to anaphase progression. Can measurements be carried out across the entire duration of cell division to demonstrate that the overlap's 'constant length' property is unique to anaphase? (This could demonstrate already in Fig. 1 that the method in principle is capable of measuring different overlap lengths.)

We agree with the reviewer and have moved the SHG images from supplementary Fig. 6 to the main Figure 1A for better visibility. In addition, we have added a plot as an inset in (now) Figure 1B and C explanation of how the used spindle pole distances related to the progression through anaphase. Unfortunately, we can only acquire a single timepoint and not a live movie during the SHG.

Even though the manuscript contains an impressive amount of data, it appears to be lengthy, the motivation for several experiments is not clearly described, and the order of data presentation can probably be improved. For example, it is unclear why SPD-1 profiles are presented late and why KLP-19 profiles are missing - one would expect to see them early on as an essential characterization of the system under study. The motivation of the paragraph investigating the relation of KLP-19 and SPD-1 to HCP-6 is especially unclear (more than 1 page of text describing supplementary material).

We will go through our text again and will revise the order of presented experiments. As stated above, we have removed the HCP-6 data.

The absence of interaction between KLP-19 and SPD-1 is not demonstrated to the same quality standard as the presence of interaction between the orthologs in the literature, which should at least be mentioned.

Additional experiments essential to support the claims of the paper?

KLP-19 profiles in the presence and absence of SPD-1 seem to be essential.

We agree with the reviewer and will add this.

A co-IP of KLP-19 and SPD-1 (including positive control) to prove that the proteins are not interacting would help to support the claim.

We agree with the reviewer and will add this

Data and methods presented so that they can be reproduced? Yes.

Experiments adequately replicated and statistical analysis adequate? Yes.

Minor comments:

Generating cellular electron tomography data is very laborious. It is a pity that no raw data is provided; for example, a slice of a reconstructed tomogram or a video of whole volumes without segmentation would be an informative addition and allow assessment of the data quality.

We agree with the reviewer and will add movies of the raw electron microscopy data.

The clear evidence for direct interaction between PRC1 and kinesin-4 in other species should be correctly acknowledged throughout the text.

We agree with the reviewer and have corrected this

The average (mean or median?) values and STDs reported in the text do not appear to match those in Fig. 1D.

*We thank the reviewer for pointing this out and have corrected the figure. The violin lot showed the mean and percentiles, we have now changed the plot to show mean and STD. *

• *

The kymograph of spd-1 RNAi in Fig. 2A seems stretched, and the size based on the scale bar does not fit the values stated in the text.

We thank the reviewer for pointing this out and have corrected the figure.

The figure numbering, as stated in the text, does not seem to agree with those in Supplementary Figure 8.

*We thank the reviewer for pointing this out and have corrected the text. *

Page numbers and/or line numbers and figure numbers on the figures would help the reader to navigate the manuscript more easily.

We agree with the reviewer and have added this.

Reviewer #3 (Significance (Required)):

The manuscript is a rich resource of quantitative measurements of C.elegans' structural and dynamic spindle properties, using advanced light microscopy and 3D electron microscopy imaging. In large parts, the work confirms previous conclusions of the function of PRC1 and kinesin-4 in the anaphase spindle, but also reports some interesting differences, namely that the C.elegans proteins differ from their orthologs in that they do not interact with each other, raising the question of how the kinesin-4 KLP-19 localizes to the central spindle in this organism. This work is of interest for researchers studying cell division, and specifically spindle architecture, dynamics, and function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Summary:

The anaphase spindle midzone is an essential structure for cell division. It consists of antiparallel overlapping microtubules organized by the antiparallel microtubule bundler PRC1, molecular motors and other regulatory proteins. This manuscript investigates the role of KLP-19 (C. elegans ortholog of human kinesin-4 KIF4A) and SPD-1 (C. elegans ortholog of PRC1) for spindle midzone organization in the C. elegans embryo and its relevance for proper spindle function. Advanced fluorescence microscopy, 3D electron tomography, and a fluorescence microscopy-based single molecule assay in embryo lysate are used in a unique combination. The authors confirm several aspects of PRC1 and KIF4A function in anaphase, as reported in previous work, mostly in human cells and Drosophila embryos and also in C. elegans embryos. Measurements are mostly very quantitative and to a high quality standard. The main difference to previous conclusions is that here, the authors propose that KLP-19 does not interact with SPD-1, in contrast to what has been established for other animal kinesin-4s and PRC1, and instead localizes to the spindle midzone independently of PRC1 by a mechanism that remains unknown. The authors provide evidence that KLP-19 nevertheless controls microtubule overlap length as in other species and that it produces outward forces sliding midzone microtubules apart a movement that SPD-1 counteracts (presumably by friction). The manuscript presents a rich resource of carefully measured quantitative structural and dynamic C. elegans anaphase spindle data.

Major comments:

Key conclusions convincing?

1. The key conclusions that the length of the central anaphase spindle microtubule overlap remains constant as the C.elegans spindle elongates (Fig. 1), that PRC1 indeed localizes quite precisely to these overlaps as previously assumed based on its in vitro (purified protein) behavior (Fig. 3B) and that the kinesin-4 KLP-19 controls overlap length as in other species (Fig. 3B) are all convincingly shown. What's missing are quantitative KLP-19 together with microtubule polarity profiles in the presence and absence of SPD-1, leaving it unclear to which extent this kinesin localizes to microtubule overlaps in the two situations. Such data seem crucial, given the authors' claim that KLP-19 localizes to the midzone and that this localization of KLP-19 is mostly unaffected by the absence of SPD-1.
2. 'Normalized KLP-19 intensities' are used to demonstrate that the total amount of this kinesin localizing to the spindle midzone does not depend on the presence of SPD-1 (Fig. 3F). Given that this claim represents a major novelty of the study, the efficiency of the SPD-1 knock-down should be documented, ideally by western blot and fluorescence microscopy.
3. The authors show convincingly that the kinesin KLP-19 contributes to outward microtubule sliding (and can contribute to spindle rupture in the absence of SPD-1) (Fig. 2), which is interesting and in line with the author's main claim.
4. The interaction between KIF4a and PRC1 is well established in other species and has been clearly demonstrated both in cells and in vitro (with purified proteins). The authors claim that this concept does not apply to the C. elegans orthologs. To show 'in vitro' (outside of the spindle) that the C. elegans homologs KLP-19 and SPD-1 do not interact, the authors use a novel microfluidic fluorescence-based single-molecule assay in lysate (Fig. 4). Although very original, these experiments do not reach the biochemical standard of previous experiments with purified proteins without appropriate controls. Given that the lysate setup is fairly novel, it's advisable to present at least one positive control demonstrating that interactions between soluble proteins can indeed be detected using this assay. It would also be useful to show the absence of interaction between KLP-19 and SPD-1 by a more conventional method like co-IP, again with a positive control, to support the authors' claim. Eventually, experiments with purified proteins will have to unequivocally demonstrate whether KLP-19 and SPD-1 indeed do not interact - something which appears, however, to be beyond the scope of this study. Without additional experimental proof, the authors may want to indicate that these results are of more preliminary nature.
5. Unfortunately, the authors do not propose an alternative mechanism for KLP-19 localization to the midzone in SPD-1 depleted embryos, limiting the conceptual advance. Does KLP-19 bind directly to antiparallel microtubules, for example in the assay presented in Fig. 4 (where signs of microtubule crosslinking are shown for SPD-1)? If not, how would it accumulate in the midzone (if it does) in the C. elegans embryo anaphase spindle? The authors do also not propose a mechanism explaining why central antiparallel microtubule overlap length does not change as the spindle elongates in anaphase. Moreover, there is no discussion regarding the potential mechanism leading to KLP-19 controlling microtubule dynamics globally instead of locally where the motor accumulates, indicating limitations in mechanistic insight.

Claims justified/preliminary and clearly presented?

The observation that the spindle length remains constant throughout anaphase in C. elegans is based on elegant, but unconventional fluorescence microscopy data (Fig. 1A & B). It would be helpful to add images of SHG and two-photon microscopy to help the reader understand the graphs. Measurements are presented based on distances between the poles. It is unclear why the distances between 15-20 µm were chosen and how they translate to anaphase progression. Can measurements be carried out across the entire duration of cell division to demonstrate that the overlap's 'constant length' property is unique to anaphase? (This could demonstrate already in Fig. 1 that the method in principle is capable of measuring different overlap lengths.)

Even though the manuscript contains an impressive amount of data, it appears to be lengthy, the motivation for several experiments is not clearly described, and the order of data presentation can probably be improved. For example, it is unclear why SPD-1 profiles are presented late and why KLP-19 profiles are missing - one would expect to see them early on as an essential characterization of the system under study. The motivation of the paragraph investigating the relation of KLP-19 and SPD-1 to HCP-6 is especially unclear (more than 1 page of text describing supplementary material).

The absence of interaction between KLP-19 and SPD-1 is not demonstrated to the same quality standard as the presence of interaction between the orthologs in the literature, which should at least be mentioned.

Additional experiments essential to support the claims of the paper?

KLP-19 profiles in the presence and absence of SPD-1 seem to be essential.

A co-IP of KLP-19 and SPD-1 (including positive control) to prove that the proteins are not interacting would help to support the claim.

Data and methods presented so that they can be reproduced? Yes.

Experiments adequately replicated and statistical analysis adequate? Yes.

Minor comments:

Generating cellular electron tomography data is very laborious. It is a pity that no raw data is provided; for example, a slice of a reconstructed tomogram or a video of whole volumes without segmentation would be an informative addition and allow assessment of the data quality.

The clear evidence for direct interaction between PRC1 and kinesin-4 in other species should be correctly acknowledged throughout the text.

The average (mean or median?) values and STDs reported in the text do not appear to match those in Fig. 1D.

The kymograph of spd-1 RNAi in Fig. 2A seems stretched, and the size based on the scale bar does not fit the values stated in the text.

The figure numbering, as stated in the text, does not seem to agree with those in Supplementary Figure 8.

Page numbers and/or line numbers and figure numbers on the figures would help the reader to navigate the manuscript more easily.

Significance

The manuscript is a rich resource of quantitative measurements of C.elegans' structural and dynamic spindle properties, using advanced light microscopy and 3D electron microscopy imaging. In large parts, the work confirms previous conclusions of the function of PRC1 and kinesin-4 in the anaphase spindle, but also reports some interesting differences, namely that the C.elegans proteins differ from their orthologs in that they do not interact with each other, raising the question of how the kinesin-4 KLP-19 localizes to the central spindle in this organism. This work is of interest for researchers studying cell division, and specifically spindle architecture, dynamics, and function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Zimyanin et al, Chromokinesin Klp-19 regulates microtubule overlap and dynamics during anaphase in C. elegans.

The authors used a myriad of techniques, including confocal live-cell imaging, 2-photon microscopy, second harmonic generation imaging, FRAP, microfluidic-coupled TIRF, EM-tomography, to study spindle midzone assembly dynamics in C. elegans one-cell stage embryos. In particular, they illuminated the role of kinesin-4 KLP-19 in maintaining proper midzone length and organization. Inhibition of KLP-19 results in longer more stable midzones, implying KLP-19 functions in depolymerizing microtubules.

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Minor comments:

Fig 3E / There is an unusual diagonal line bisecting the embryo. Visually this does not affect viewing of the His::GFP and KLP-19::GFP signals. However, when these signals are quantified and normalized (as in Fig 3F), the diagonal bisect displaying different background signal may impact the measurements.

Fig 4B / While the kymographs clearly show KLP-19::GFP motility on microtubules, they also show that the majority of KLP(-::GFP do not move. Perhaps some quantification and discussion of this result is appropriate?

Significance

Indeed, much of the results in the current study are consistent with previously published results elsewhere. Nevertheless, the current work represents a tour-de-force showcase of diverse and state-of-the-art technology application to address spindle assembly dynamics. How KLP-19 functions to define microtubule length at the midzone is still not known. But the current work, with diverse and solid data, serves to highlight where future work should focus.

Prueba

Reviewer #1 (Public review):

Overall I found the approach taken by the authors to be clear and convincing. It is striking that the conclusions are similar to those obtained in a recent study using a different computational approach (finite state controllers), and lends confidence to the conclusions about the existence of an optimal memory duration. There are a few questions that could be expanded on in future studies:

(1) Spatial encoding requirements

The manuscript contrasts the approach taken here (reinforcement learning in a gridworld) with strategies that involve a "spatial map" such as infotaxis. However, the gridworld navigation algorithm has an implicit allocentric representation, since movement can be in one of four allocentric directions (up, down, left, right), and wind direction is defined in these coordinates. Future studies might ask if an agent can learn the strategy without a known wind direction if it can only go left/right/forward/back/turn (in egocentric coordinates). In discussing possible algorithms, and the features of this one, it might be helpful to distinguish (1) those that rely only on egocentric computations (run and tumble), (2) those that rely on a single direction cue such as wind direction, (3) those that rely on allocentric representations of direction, and (4) those that rely on a full spatial map of the environment.

(2) Recovery strategy on losing the plume

The authors explore several recovery strategies upon losing the plume, including backtracking, circling, and learned strategies, finding that a learned strategy is optimal. As insects show a variety of recovery strategies that can depend on the model of locomotion, it would be interesting in the future to explore under which conditions various recovery strategies are optimal and whether they can predict the strategies of real animals in different environments.

(3) Is there a minimal representation of odor for efficient navigation?

The authors suggest that the number of olfactory states could potentially be reduced to reduce computational cost. They show that reducing the number of olfactory states to 1 dramatically reduces performance. In the future it would be interesting to identify optimal internal representations of odor for navigation and to compare these to those found in real olfactory systems. Does the optimal number of odor and void states depend on the spatial structure of the turbulence as explored in Figure 5?

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Overall I found the approach taken by the authors to be clear and convincing. It is striking that the conclusions are similar to those obtained in a recent study using a different computational approach (finite state controllers), and lends confidence to the conclusions about the existence of an optimal memory duration. There are a few questions that could be expanded on in future studies:

(1) Spatial encoding requirements

The manuscript contrasts the approach taken here (reinforcement learning in a gridworld) with strategies that involve a "spatial map" such as infotaxis. However, the gridworld navigation algorithm has an implicit allocentric representation, since movement can be in one of four allocentric directions (up, down, left, right), and wind direction is defined in these coordinates. Future studies might ask if an agent can learn the strategy without a known wind direction if it can only go left/right/forward/back/turn (in egocentric coordinates). In discussing possible algorithms, and the features of this one, it might be helpful to distinguish (1) those that rely only on egocentric computations (run and tumble), (2) those that rely on a single direction cue such as wind direction, (3) those that rely on allocentric representations of direction, and (4) those that rely on a full spatial map of the environment.

We agree that the question of what orientation skills are needed to implement an algorithm is interesting. We remark that our agents do not use allocentric directions in the sense of north, east, west and east relative to e.g. fixed landmarks in the environment. Instead, directions are defined relative to the mean wind, which is assumed fixed and known. (In our first answer to reviewers we used “north east south west relative to mean wind”, which may have caused confusion – but in the manuscript we only use upwind downwind and crosswind).

(2) Recovery strategy on losing the plume

The authors explore several recovery strategies upon losing the plume, including backtracking, circling, and learned strategies, finding that a learned strategy is optimal. As insects show a variety of recovery strategies that can depend on the model of locomotion, it would be interesting in the future to explore under which conditions various recovery strategies are optimal and whether they can predict the strategies of real animals in different environments.

Agreed, it will be interesting to study systematically the emergence of distinct recovery strategies and compare to living organisms.

(3) Is there a minimal representation of odor for efficient navigation?

The authors suggest that the number of olfactory states could potentially be reduced to reduce computational cost. They show that reducing the number of olfactory states to 1 dramatically reduces performance. In the future it would be interesting to identify optimal internal representations of odor for navigation and to compare these to those found in real olfactory systems. Does the optimal number of odor and void states depend on the spatial structure of the turbulence as explored in Figure 5?

We agree that minimal odor representations are an intriguing question. While tabular Q learning cannot derive optimal odor representations systematically, one could expand on the approach we have taken here and provide more comparisons. It will be interesting to follow this approach in a future study.

Reviewer #2 (Public review):

Summary:

The authors investigate the problem of olfactory search in turbulent environments using artificial agents trained using tabular Q-learning, a simple and interpretable reinforcement learning (RL) algorithm. The agents are trained solely on odor stimuli, without access to spatial information or prior knowledge about the odor plume's shape. This approach makes the emergent control strategy more biologically plausible for animals navigating exclusively using olfactory signals. The learned strategies show parallels to observed animal behaviors, such as upwind surging and crosswind casting. The approach generalizes well to different environments and effectively handles the intermittency of turbulent odors.

Strengths:

* The use of numerical simulations to generate realistic turbulent fluid dynamics sets this paper apart from studies that rely on idealized or static plumes.

* A key innovation is the introduction of a small set of interpretable olfactory states based on moving averages of odor intensity and sparsity, coupled with an adaptive temporal memory.

* The paper provides a thorough analysis of different recovery strategies when an agent loses the odor trail, offering insights into the trade-offs between various approaches.

* The authors provide a comprehensive performance analysis of their algorithm across a range of environments and recovery strategies, demonstrating the versatility of the approach.

* Finally, the authors list an interesting set of real-world experiments based on their findings, that might invite interest from experimentalists across multiple species.

Weaknesses:

* Using tabular Q-learning is both a strength and a limitation. It's simple and interpretable, making it easier to analyze the learned strategies, but the discrete action space seems somewhat unnatural. In real-world biological systems, actions (like movement) are continuous rather than discrete. Additionally, the ground-frame actions may not map naturally to how animals navigate odor plumes (e.g. insects often navigate based on their own egocentric frame).

We agree with the reviewer, and will look forward to study this problem further to make it suitable for meaningful comparisons with animal behavior.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

The authors have addressed my major concerns and I support publication of this interesting manuscript. A couple of small suggestions:

(1) In discussing performance in different environments (line 328-362) it might be easier to read if you referred to the environments by descriptive names rather than numbers.

Thank you for the suggestion, which we implemented

(2) Line 371: measurements of flow speed depend on antennae in insects. Insects can measure local speed and direct of flow using antennae, e.g. Bell and Kramer, 1979, Suver et al. 2019. Okubo et al. 2020,

Thank you for the references

(3) line 448: "Similarly, an odor detection elicits upwind surges that can last several seconds" maybe "Similarly, an odor detection elicits upwind surges that can outlast the odor by several seconds"?

Thank you for the suggestion

Reviewer #2 (Recommendations for the authors):

I commend the authors for their revisions in response to reviewer feedback.

While I appreciate that the manuscript is now accompanied by code and data, I must note that the accompanying code-repository lacks proper instructions for use and is likely incomplete (e.g. where is the main function one should run to run your simulations? How should one train? How should one recreate the results? Which data files go where?).

For examples of high-quality code-release, please see the documentation for these RL-for-neuroscience code repositories (from previously published papers):

https://github.com/ryzhang1/Inductive_bias

https://github.com/BruntonUWBio/plumetracknets

The accompanying data does provide snapshots from their turbulent plume simulations, which should be valuable for future research.

Thank you for the suggestions for how to improve clarity of the code. The way we designed the repository is to serve both the purpose of developing the code as well as sharing. This is because we are going to build up on this work to proceed further. Nothing is missing in the repository (we know it because it is what we actually use).

We do plan to create a more user-friendly version of the code, hopefully this will be ready in the next few months, but it wont be immediate as we are aiming to also integrate other aspects of the work we are currently doing in the Lab. The Brunton repository is very well organized, thanks for the pointer.

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

Reviewer #1 (Public review):

Overall I found the approach taken by the authors to be clear and convincing. It is striking that the conclusions are similar to those obtained in a recent study using a different computational approach (finite state controllers), and lend confidence to the conclusions about the existence of an optimal memory duration. There are a few points or questions that could be addressed in greater detail in a revision:

(1) Discussion of spatial encoding

The manuscript contrasts the approach taken here (reinforcement learning in a grid world) with strategies that involve a "spatial map" such as infotaxis. The authors note that their algorithm contains "no spatial information." However, I wonder if further degrees of spatial encoding might be delineated to better facilitate comparisons with biological navigation algorithms. For example, the gridworld navigation algorithm seems to have an implicit allocentric representation, since movement can be in one of four allocentric directions (up, down, left, right). I assume this is how the agent learns to move upwind in the absence of an explicit wind direction signal. However, not all biological organisms likely have this allocentric representation. Can the agent learn the strategy without wind direction if it can only go left/right/forward/back/turn (in egocentric coordinates)? In discussing possible algorithms, and the features of this one, it might be helpful to distinguish<br /> (1) those that rely only on egocentric computations (run and tumble),<br /> (2) those that rely on a single direction cue such as wind direction,<br /> (3) those that rely on allocentric representations of direction, and<br /> (4) those that rely on a full spatial map of the environment.

As Referee 1 points out, even if the algorithm does not require a map of space, the agent is still required to tell apart directions relative to the wind direction which is assumed known. Indeed, although in the manuscript we labeled actions allocentrically as “ up down left and right”, the source is always placed in the same location, hence “left” corresponds to upwind; “right” to downwind and “up” and “down” to crosswind right and left. Thus in fact directions are relative to the mean wind, which is therefore assumed known. We have better clarified the spatial encoding required to implement these strategies, and re-labeled the directions as upwind, downwind, crosswind-right and crosswind-left.

In reality, animals cannot measure the mean flow, but rather the local flow speed e.g. with antennas for insects, with whiskers for rodents and with the lateral line for marine organisms. Further work is needed to address how local flow measures enable navigation using Q learning.

(2) Recovery strategy on losing the plume

While the approach to encoding odor dynamics seems highly principled and reaches appealingly intuitive conclusions, the approach to modeling the recovery strategy seems to be more ad hoc. Early in the paper, the recovery strategy is defined to be path integration back to the point at which odor was lost, while later in the paper, the authors explore Brownian motion and a learned recovery based on multiple "void" states. Since the learned strategy works best, why not first consider learned strategies, and explore how lack of odor must be encoded or whether there is an optimal division of void states that leads to the best recovery strategies? Also, although the authors state that the learned recovery strategies resemble casting, only minimal data are shown to support this. A deeper statistical analysis of the learned recovery strategies would facilitate comparison to those observed in biology.

We thank Referee 1 for their remarks and suggestion to give the learned recovery a more prominent role and better characterize it. We agree that what is done in the void state is definitely key to turbulent navigation. In the revised manuscript, we have further substantiated the statistics of the learned recovery by repeating training 20 times and comparing the trajectories in the void (Figure 3 figure supplement 3, new Table 1). We believe however that starting with the heuristic recovery is clearer because it allows to introduce the concept of recovery more clearly. Indeed, the learned “recovery” is so flexible that it ends up mixing recovery (crosswind motion) to aspects of exploitation (surge): we defer a more in-depth analysis that disentangles these two aspects elsewhere. Also, we added a whole new comparison with other biologically inspired recoveries both in the native environment and for generalization (Figure 3 and 5).

(3) Is there a minimal representation of odor for efficient navigation?

The authors suggest (line 280) that the number of olfactory states could potentially be reduced to reduce computational cost. This raises the question of whether there is a maximally efficient representation of odors and blanks sufficient for effective navigation. The authors choose to represent odor by 15 states that allow the agent to discriminate different spatial regimes of the stimulus, and later introduce additional void states that allow the agent to learn a recovery strategy. Can the number of states be reduced or does this lead to loss of performance? Does the optimal number of odor and void states depend on the spatial structure of the turbulence as explored in Figure 5?

We thank the referee for their comment. Q learning defines the olfactory states prior to training and does not allow a systematic optimization of odor representation for the task. We can however compare different definitions of the olfactory states, for example based on the same features but different discretizations. We added a comparison with a drastically reduced number of non-empty olfactory states to just 1, i.e. if the odor is above threshold at any time within the memory, the agent is in the non-void olfactory state, otherwise it is in the void state. This drastic reduction in the number of olfactory states results in less positional information and degrades performance (Figure 5 figure supplement 5).

The number of void states is already minimal: we chose 50 void states because this matches the time agents typically remain in the void (less than 50 void states results in no convergence and more than 50 introduces states that are rarely visited).

One may instead resort to deep Q-learning or to recurrent neural networks, which however do not provide answers as for what are the features or olfactory states that drive behavior (see discussion in manuscript and questions below).

Reviewer #2 (Public review):

Summary:

The authors investigate the problem of olfactory search in turbulent environments using artificial agents trained using tabular Q-learning, a simple and interpretable reinforcement learning (RL) algorithm. The agents are trained solely on odor stimuli, without access to spatial information or prior knowledge about the odor plume's shape. This approach makes the emergent control strategy more biologically plausible for animals navigating exclusively using olfactory signals. The learned strategies show parallels to observed animal behaviors, such as upwind surging and crosswind casting. The approach generalizes well to different environments and effectively handles the intermittency of turbulent odors.

Strengths:

(1) The use of numerical simulations to generate realistic turbulent fluid dynamics sets this paper apart from studies that rely on idealized or static plumes.

(2) A key innovation is the introduction of a small set of interpretable olfactory states based on moving averages of odor intensity and sparsity, coupled with an adaptive temporal memory.

(3) The paper provides a thorough analysis of different recovery strategies when an agent loses the odor trail, offering insights into the trade-offs between various approaches.

(4) The authors provide a comprehensive performance analysis of their algorithm across a range of environments and recovery strategies, demonstrating the versatility of the approach.

(5) Finally, the authors list an interesting set of real-world experiments based on their findings, that might invite interest from experimentalists across multiple species.

Weaknesses:

(1) The inclusion of Brownian motion as a recovery strategy, seems odd since it doesn't closely match natural animal behavior, where circling (e.g. flies) or zigzagging (ants' "sector search") could have been more realistic.

We agree that Brownian motion may not be biologically plausible -- we used it as a simple benchmark. We clarified this point, and re-trained our algorithm with adaptive memory using circling and zigzaging (cast and surge) recoveries. The learned recovery outperforms all heuristic recoveries (Figure 3D, metrics G). Circling ranks second, and achieves these good results by further decreasing the probability of failure and paying slightly in speed. When tested in the non-native environments 2 to 6, the learned recovery performs best in environments 2, 5 and 6 i.e. from long range more relevant to flying insects; whereas circling generalizes best in odor rich environments 3 and 4, representative of closer range and close to the substrate (Figure 5B, metrics G). In the new environments, similar to the native environment, circling favors convergence (Figure 5B, metrics f⁺) over speed (Figure 5B, metrics g⁺ and τ_min/τ), which is particularly deleterious at large distance.

(2) Using tabular Q-learning is both a strength and a limitation. It's simple and interpretable, making it easier to analyze the learned strategies, but the discrete action space seems somewhat unnatural. In real-world biological systems, actions (like movement) are continuous rather than discrete. Additionally, the ground-frame actions may not map naturally to how animals navigate odor plumes (e.g. insects often navigate based on their own egocentric frame).

We agree with the reviewer that animal locomotion does not look like a series of discrete displacements on a checkerboard. However, to overcome this limitation, one has to first focus on a specific system to define actions in a way that best adheres to a species’ motor controls. Moreover, these actions are likely continuous, which makes reinforcement learning notoriously more complex. While we agree that more realistic models are definitely needed for a comparison with real systems, this remains outside the scope of the current work. We have added a remark to clarify this limitation.

(3) The lack of accompanying code is a major drawback since nowadays open access to data and code is becoming a standard in computational research. Given that the turbulent fluid simulation is a key element that differentiates this paper, the absence of simulation and analysis code limits the study's reproducibility.

We have published the code and the datasets at

- code: https://github.com/Akatsuki96/qNav

- datasets: https://zenodo.org/records/14655992

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Line 59-69: In comparing the results here to other approaches (especially the Verano and Singh papers), it would also be helpful to clarify which of these include an explicit representation of the wind direction. My understanding is that both the Singh and Verano approaches include an explicit representation of wind direction. In Singh wind direction is one of the observations that inputs to the agent, while in Verano, the actions are defined relative to the wind direction. In the current paper, my understanding is that there is no explicitly defined wind direction, but because movement directions are encoded allocentrically, the agent is able to learn the upwind direction from the structure of the plume- is this correct? I think this information would be helpful to spell out and also to address whether an agent without any allocentric direction sense can learn the task.

Thank you for the comment. In our algorithm the directions are defined relative to the mean wind, which is assumed known, as in Verano et al. As far as we understand, Singh et al provide the instantaneous, egocentric wind velocities as part of the input.

(1) Line 105: "several properties of odor stimuli depend on the distance from the source" might cite Boie...Victor 2018, Ackles...Schaefer, 2021, Nag...van Breugel 2024.

Thank you for the suggestions - we have added these references

(2) Line 130: "we first define a finite set of olfactory states" might be helpful to the reader to state what you chose in this paragraph rather than further down.

We have slightly modified the incipit of the paragraph. We first declare we are setting out to craft the olfactory states, then define the challenges, finally we define the olfactory states.

(3) Line 267: "Note that the learned recovery strategy resembles casting behavior observed in flying insects" Might note that insects seem to deploy a range of recovery strategies depending on locomotor mode and environment. For example, flying flies circle and sink when odor is lost in windless environments (Stupski and van Breugel 2024).

Thank you for your comment. We have included the reference and we now added comparisons to results using circling and cast & surge recovery strategies.

(4) Line 289: "from positions beyond the source, the learned strategy is unable to recover the plume as it mostly casts sideways, with little to no downwind action" This is curious as many insects show a downwind bias in the absence of odor that helps them locate the plumes in the first place (e.g. Wolf and Wehner, 2000, Alvarez-Salvado et al. 2018). Is it possible that the agent could learn a downwind bias in the absence of odor if given larger environments or a longer time to learn?

The reviewer is absolutely correct – Downwind motion is not observed in the recovery simply because the agent rarely overshoots the source. Hence overall optimization for that condition is washed out by the statistics. We believe downwind motion will emerge if an agent needs to avoid overshooting the source – we do not have conclusive results yet but are planning to introduce such flexibility in a further work. We added this remark and refs.

(5) Line 377-391: testing these ideas in living systems. Interestingly, Kathman..Nagel 2024 (bioRxiv) shows exactly the property predicted here and in Verano in fruit flies- an odor memory that outlasts the stimulus by a duration of several seconds, appropriate for filling in "blanks." Relatedly, Alvarez-Salvado et al. 2018 showed that fly upwind running reflected a temporal integration of odor information over ~10s, sufficient to avoid responding to blanks as loss of odor.

Indeed, we believe this is the most direct connection between algorithms and experiments. We are excited to discuss with our colleagues and pursue a more direct comparison with animal behavior. We were aware of the references and forgot to cite them, thank you for your careful reading of our work !

Reviewer #2 (Recommendations for the authors):

Suggestions

(1) The paper does not clearly specify which type of animals (e.g., flying insects, terrestrial mammals) the model is meant to approximate or not approximate. The authors should consider clarifying how these simulations are suited to be a general model across varied olfactory navigators. Further, it isn't clear how low/high the intermittency studied in this model is compared to what different animals actually encounter. (Minor: The Figure 4 occupancy circles visualization could be simplified).

Environment 1 represents the lower layers of a moderately turbulent boundary layer. Search occurs on a horizontal plane ~half meter from the ground. The agent is trained at distances of about 10 meters and also tested on longer distances  ~ 17 meters (environment 6), lower heights ~1cm from the ground (environments 3-4), lower Reynolds number (environment 5) and higher threshold of detection (environment 2 and 4). Thus Environments 1,2,5 and 6 are representative of conditions encountered by flying organisms (or pelagic in water), and Environments 3 and 4 of searches near the substrate, potentially involved in terrestrial navigation (benthic in water). Even near the substrate, we use odor dispersed in the fluid, and not odor attached to the substrate (relevant to trail tracking).

Also note that we pick Schmidt number Sc = 1 and this is appropriate for odors in air but not in water. However, we expect a weak dependence on the Schmidt number as the Batchelor and Kolmogorov scales are below the size of the source and we are interested in the large scale statistics Falkovich et al., 2001; Celani et al., 2014; Duplat et al., 2010.

Intermittency contours are shown in Fig 1C, they are highest along the centerline, and decay away from the centerline, so that even within the plume detecting odor is relatively rare. Only a thin region near the centerline has intermittency larger than 66%; the outer and most critical bin of the plume has intermittency under 33%; in the furthest point on the centerline intermittency is <10%. For reference, experimental values in the atmospheric boundary layer report intermittency 25% to 20% at 2 to 15m from the source along the centerline (Murlis and Jones, 1981).

We have more clearly labeled the contours in Fig 1C and added these remarks.

We included these remarks and added a whole table with matching to real conditions within the different environments.

(2) Could some biological examples and references be added to support that backtracking is a biologically plausible mechanism?

Backtracking was observed e.g. in ants displaced in unfamiliar environments (Wystrach et al, P Roy Soc B, 280,  2013), in tsetse flies executing reverse turns uncorrelated to wind, which bring them back towards the location where they last detected odor (Torr, Phys Entom, 13, 1988, Gibson & Brady Phys Entom 10, 1985) and in coackroaches upon loss of contact with the plume (Willis et al, J. Exp. Biol. 211, 2008). It is also used in computational models of olfactory navigation (Park et al, Plos Comput Biol, 12:e1004682, 2016).

(3) Hand-crafted features can be both a strength and a limitation. On the one hand, they offer interpretability, which is crucial when trying to model biological systems. On the other hand, they may limit the generality of the model. A more thorough discussion of this paper's limitations should address this.

(4) The authors mention the possibility of feature engineering or using recurrent neural networks, but a more concrete discussion of these alternatives and their potential advantages/disadvantages would be beneficial. It should be noted that the hand-engineered features in this manuscript are quite similar to what the model of Singh et al suggests emerges in their trained RNNs.

Merged answer to points 3 and 4.

We agree with the reviewer that hand-crafted features are both a strength and a limitation in terms of performance and generality. This was a deliberate choice aimed at stripping the algorithm bare of implicit components, both in terms of features and in terms of memory. Even with these simple features, our model performs well in navigating across different signals, consistent with our previous results showing that these features are a “good” surrogate for positional information.

To search for the most effective temporal features, one may consider a more systematic hand crafting, scaling up our approach. In this case one would first define many features of the odor trace; rank groups of features for their accuracy in regression against distance; train Q learning with the most promising group of features and rank again. Note however that this approach will be cumbersome because multiple factors will have to be systematically varied: the regression algorithm; the discretization of the features and the memory.

Alternatively, to eliminate hand crafting altogether and seek better performance or generalization, one may consider replacing these hand-crafted features and the tabular Q-learning approach with recurrent neural networks or with finite state controllers. On the flip side, neither of these algorithms will directly provide the most effective features or the best memory, because these properties are hidden within the parameters that are optimized for. So extra work is needed to interrogate the algorithms and extract these information. For example, in Singh et al, the principal components of the hidden states in trained agents correlate with head direction, odor concentration and time since last odor encounter. More work is needed to move beyond correlations and establish more systematically what are the features that drive behavior in the RNN.

We have added these points to the discussion.

(5) Minor: the title of the paper doesn't immediately signal its focus on recovery strategies and their interplay with memory in the context of olfactory navigation. Given the many other papers using a similar RL approach, this might help the authors position this paper better.

We agree with the referee and have modified the title to reflect this.

(6) Minor: L 331: "because turbulent odor plumes constantly switch on and off" -- the signal received rather than the plume itself is switching on and off.

Thank you for the suggestion, we implemented it.

Reviewer #2 (Public review):

Summary:

In this study by Sánchez-León and colleagues, the authors attempted to determine the influence of neuronal orientation on the efficacy of cerebellar tDCS in modulating neural activity. To do this, the authors made recordings from Purkinje cells, the primary output neurons of the cerebellar cortex, and determined the inter-dependency between the orientation of these cells and the changes in their firing rate during cerebellar tDCS application.

Strengths:

(1) A major strength is the in vivo nature of this study. Being able to simultaneously record neural activity and apply exogenous electrical current to the brain during both an anesthetized state and during wakefulness in these animals provides important insight into physiological underpinnings of tDCS.<br /> (2) The authors provide evidence that tDCS can modulate neural activity in multiple cell types. For example, there is a similar pattern of modulation in Purkinje cells and non-Purkinje cells (excitatory and inhibitory interneurons). Together, these data provide wholistic insight into how tDCS can affect activity across different populations of cells, which is important implications for basic neuroscience, but also clinical populations where there may be non-uniform or staged effects of neurological disease on these various cell types.<br /> (3) There is systematic investigation into the effects of tDCS on neural activity across multiple regions of the cerebellum. The authors demonstrate that the pattern of modulation is dependent on the target region. These findings have important implications for determining the expected neuromodulatory effects of tDCS when applying this technique over different target regions non-invasively in animals and humans.<br /> (4) The authors provide a thorough background, rationale, and interpretation regarding the expected and observed influence of neuronal orientation on excitability modulation by electrical stimulation.

The OSI model, created by the International Standards Organization, explains how network communication works in seven layers. The physical layer handles the hardware and sends raw bits. The data-link layer manages frames and fixes errors. The network layer breaks data into packets and routes them using logical addresses. The transport layer ensures messages are delivered in order and controls traffic. The session layer manages connections between programs. The presentation layer handles data formats and conversions. Finally, the application layer interacts with users and supports tasks like file sharing, emails, and remote access.

To communicate over a network, systems use host names and process IDs. Since computers prefer numbers (IP addresses), DNS is used to convert names to IPs. DNS checks name servers step-by-step to find the correct address and uses caching to speed things up. Before DNS, all host info was stored in one shared file, which became hard to manage. Now, each domain manages its own data.

In Java, the InetAddress class can look up IP addresses. The program uses getByName() to find the IP and getHostAddress() to print it. If the host isn’t found, it shows an error.

v1.4 Update

The following references have been added:

1. Food and Drug Administration, Citus Pharmaceuticals. LYMPHIR (denileukin diftitox-cxdl) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview.process&ApplNo=761312 Accessed 1/10/25

2. Foss FM, Kim YH, Prince H, Kuzel TM, Yannakou CK, Ooi CE, Xing D, Sauter N, Singh P, Czuczman M, Duvic M. Efficacy and safety of E7777 (improved purity Denileukin diftitox [ONTAK]) in patients with relapsed or refractory cutaneous T-cell lymphoma: results from pivotal study 302. Blood. 2022 Nov 15;140(Supplement 1):1491-2.

3. Food and Drug Administration, Genmab. EPKINLY (epcoritamab) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=BasicSearch.process Accessed 1/10/25

4. Linton KM, Vitolo U, Jurczak W, Lugtenburg PJ, Gyan E, Sureda A, Christensen JH, Hess B, Tilly H, Cordoba R, Lewis DJ. Epcoritamab monotherapy in patients with relapsed or refractory follicular lymphoma (EPCORE NHL-1): a phase 2 cohort of a single-arm, multicentre study. The Lancet Haematology. 2024 Jun 15.

5. Food and Drug Administration, Juno Therapeutics. BREYANZI (lisocabtagene maraleucel) prescribing information. Available: https://www.fda.gov/vaccines-blood-biologics/cellular-gene-therapy-products/breyanzi-lisocabtagene-maraleucel Accessed 1/10/25

6. Food and Drug Administration, Genentech. GAZYVA (obinutuzumab) prescribing information. Available: https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/125486s034lbl.pdf

7. Zinzani PL, Mayer J, Flowers CR, Bijou F, De Oliveira AC, Song Y, Zhang Q, Merli M, Bouabdallah K, Ganly P, Zhang H. ROSEWOOD: a phase II randomized study of zanubrutinib plus obinutuzumab versus obinutuzumab monotherapy in patients with relapsed or refractory follicular lymphoma. Journal of Clinical Oncology. 2023 Nov 20;41(33):5107-17.

8. Food and Drug Administration, Genentech. COLUMVI (glofitamab) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=BasicSearch.process Accessed 1/10/25

9. Food and Drug Administration, Genentech. POLIVY (polatuzumab vedotin) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=BasicSearch.process Accessed 1/10/25

10. Tilly H, Morschhauser F, Sehn LH, Friedberg JW, Trněný M, Sharman JP, Herbaux C, Burke JM, Matasar M, Rai S, Izutsu K. Polatuzumab vedotin in previously untreated diffuse large B-cell lymphoma. New England Journal of Medicine. 2022 Jan 27;386(4):351-63.

11. Food and Drug Administration, Genentech. LUNSUMIO (mosunetuzumab) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=BasicSearch.process Accessed 1/10/25

12. Morschhauser F, Le Gouill S, Feugier P, Bailly S, Nicolas-Virelizier E, Bijou F, et al. Obinutuzumab combined with lenalidomide for relapsed or refractory follicular B-cell lymphoma (GALEN): a multicentre, single-arm, phase 2 study. The Lancet Haematology. 2019;6(8):e429-e37.

13. Gurumurthi A, Chin CK, Feng L, Fowler NH, Strati P, Hagemeister FB, Fayad LE, Westin JR, Obi C, Arafat J, Nair R. Safety and activity of lenalidomide in combination with obinutuzumab in patients with relapsed indolent non-Hodgkin lymphoma: a single group, open-label, phase 1/2 trial. EClinicalMedicine. 2024 Aug 1;74.

14. Herrera AF, LeBlanc M, Castellino SM, Li H, Rutherford SC, Evens AM, et al. Nivolumab+ AVD in advanced-stage classic Hodgkin’s lymphoma. New England Journal of Medicine. 2024;391(15):1379-89.

15. Moskowitz AJ, Shah G, Schöder H, Ganesan N, Drill E, Hancock H, et al. Phase II trial of pembrolizumab plus gemcitabine, vinorelbine, and liposomal doxorubicin as second-line therapy for relapsed or refractory classical Hodgkin lymphoma. Journal of Clinical Oncology. 2021;39(28):3109-17.

16. Mei MG, Lee HJ, Palmer JM, Chen R, Tsai N-C, Chen L, et al. Response-adapted anti-PD-1–based salvage therapy for Hodgkin lymphoma with nivolumab alone or in combination with ICE. Blood, The Journal of the American Society of Hematology. 2022;139(25):3605-16.

17. Bryan LJ, Casulo C, Allen PB, Smith SE, Savas H, Dillehay GL, et al. Pembrolizumab added to ifosfamide, carboplatin, and etoposide chemotherapy for relapsed or refractory classic Hodgkin lymphoma: a multi-institutional phase 2 investigator-initiated nonrandomized clinical trial. JAMA oncology. 2023;9(5):683-91.

18. Zinzani PL, Santoro A, Gritti G, Brice P, Barr PM, Kuruvilla J, Cunningham D, Kline J, Johnson NA, Mehta-Shah N, Lisano J. Nivolumab combined with brentuximab vedotin for R/R primary mediastinal large B-cell lymphoma: a 3-year follow-up. Blood Advances. 2023 Sep 26;7(18):5272-80.

19. Food and Drug Administration, ADC Therapeutics. ZYNLONTA (loncastuximab tesirine) prescribing information. Available: https://www.accessdata.fda.gov/scripts/cder/daf/index.cfm?event=BasicSearch.process Accessed 1/10/25

20. Lisocabtagene maraleucel for patients with relapsed or refractory large B-cell lymphomas (TRANSCEND NHL 001): a multicentre seamless design study https://pubmed.ncbi.nlm.nih.gov/32888407

v1.4 Update

In addition to the approved immunotherapies described in this section, since guideline publication the following approvals and practice-changing data have been reported:

Diffuse large B cell lymphoma

Epcoritamab received an accelerated approval for relapsed or refractory DLBCL NOS, including DLBCL arising from indolent lymphoma, and HGBL after two or more lines of systemic therapy in May 2023. Approval was based on the primary endpoint ORR from EPCORE NHL-1 (NCT03625037), a single-arm phase I/II trial. The ORR for epcoritamab (n=148) was 61% (95% CI 52.5% to 68.7%), with a 9-month DOR rate of 63% (95% CI 51.5% to 72.4%). Any grade CRS occurred in 41% of patients (2.5% were grade 3), and ICANS occurred in less than 10% of patients treated with epcoritamab. [Ref 244].

Glofitamab received an accelerated approval for relapsed or refractory DLBCL, NOS or LBCL arising from FL, after two or more lines of systemic therapy in June 2023. Approval was based on the primary endpoints of ORR and DOR from NP30179 (NCT03075696), a single-arm phase I/II trial. The ORR for glofitamab (n=132) was 56% (95% CI 47% to 65%), and the median DOR was 18.4 months (95% CI 11.4 to NE). Any-grade CRS occurred in 70% of patients (4.1% were grade 3 or higher), and any-grade ICANS occurred in 4.8% of patients. [Ref 249].

Polatuzumab vedotin with a rituximab product, cyclophosphamide, doxorubicin, and prednisone ([pola-]R-CHP) was approved for adult patients who have previously untreated DLBCL, NOS or HGBL and who have an IPI score of 2 or greater in April 2023. Approval was based on the primary endpoint of PFS from POLARIX (NCT03274492), a randomized phase III trial assessing pola-R-CHP (n=440) compared to R-CHOP (n=439). PFS in the pola-R-CHP group was significantly improved (HR 0.73; 95% CI 0.57 to 0.95; p=0.0177). Serious AEs occurred in 34% of patients who received pola-R-CHP and 30.6% of patients who received R-CHOP [Ref 250, 251].

Mantle cell lymphoma

Lisocabtagene maraleucel was approved for adult patients with relapsed or refractory MCL who have received at least two prior lines of systemic therapy, including a BTKi in May 2024. Approval was based on the primary endpoint of ORR from TRANSCEND-MCL (NCT02631044), a single-arm phase I trial that assessed treatment with lisocabtagene maraleucel (n=68). The ORR was 85.3% (95% CI 74.6% to 92.7%). At a median follow-up of 22.2 months (95% CI 16.7 to 22.8), the median DOR was 13.3 months (95% CI 6.0 to 23.3). Serious AEs occurred in 53% of patients, and any-grade CRS occurred in 61% of patients (1.1% grade 4 or higher). [Ref 261].

Follicular lymphoma

Epcoritamab received an accelerated approval for adult patients with relapsed or refractory FL after two or more lines of systemic therapy in June 2024. Approval was based on the primary endpoints ORR and DOR data from EPCORE NHL-1 (Study GCT3013-01; NCT03625037), a single-arm phase II trial. The ORR for epcoritamab (n=127) was 82% (95% CI 74.1% to 88.2%), and the estimated 12-month DOR rate was 68.4% (95% CI 57.6% to 77.0%). ICANS occurred in 6.0% of patients, and serious infections occurred in 40% of patients. Grade 1–2 CRS events occurred in 49% of patients who received the 3-step dosage [Ref 244, 245].

Lisocabtagene maraleucel received an accelerated approval for adults with relapsed or refractory FL who have received two or more prior lines of systemic therapy in May 2024. Approval was based on the primary endpoint of ORR from TRANSCEND-FL (NCT04245839), a single-arm phase I/II trial. The ORR for lisocabtagene maraleucel (n=94) was 95.7% (95% CI 89%.5 to 98.8%), with a 12-month DOR rate of 80.9% (95% CI 71.0% to 87.7%). Serious AEs occurred in 26% of patients, and any-grade CRS occurred in 59% of patients (0.9% grade 3 or higher). [Ref 246].

Zanubrutinib with obinutuzumab received an accelerated approval for relapsed or refractory FL after two or more lines of systemic therapy in March 2024. Approval was based on the primary endpoints ORR and DOR from Study BGB-3111-212 (ROSEWOOD; NCT03332017), a randomized phase II trial. The ORR in patients treated with zanubrutinib with obinutuzumab (n=145) was 69% (95% CI 61% to 76%) versus 46% (95% CI 34% to 58%) in patients treated with obinutuzumab alone (n=72). The median DOR was NE (95% CI 25.3 to NE) versus 14.0 months (95% CI 9.2 to 25.1), respectively. Serious AEs occurred in 35% of patients with FL who received zanubrutinib with obinutuzumab [Ref 247, 248].

Mosunetuzumab received an accelerated approval for adult patients with relapsed or refractory FL after two or more lines of systemic therapy in December 2022. Approval was based on the primary endpoint of ORR from GO29781 (NCT02500407), a single-arm phase I/II trial. The ORR for mosunetuzumab (n=90) was 80% (95% CI 70% to 88%). Serious AEs occurred in 47% of patients, and grade 2 CRS occurred in 15% of patients, grade 3 in 2%, and grade 4 in 0.5%. [Ref 252].

Follicular lymphoma

Although not FDA-approved, data on lenalidomide + obinutuzumab for relapsed or refractory FL have been reported in the single-arm phase I/II GALEN trial (NCT01582776). The overall response was 79% (95% CI 69% to 87%) for the combination (n=86). Serious AEs were reported in 34% of patients [Ref 251]. Lenalidomide plus obinutuzumab was also assessed in the NCT01995669 trial, which included patients with indolent non-Hodgkin lymphoma (n=60; n=4 with MZL). [Ref 253, 254]

Primary mediastinal large B cell lymphoma

Although not FDA-approved, data on BV-nivolumab for the treatment of R/R PMBCL have been reported in the phase I/II singe-arm CheckMate 436 (NCT02581631) trial. At median follow up of 39.6 months, the ORR for BV-nivolumab (n=30) was 73.3% (95% CI 54.1% to 87.7%), median PFS was 26.0 months (95% CI 2.6 to NR), and median OS was NR. The most frequently occurring grade 3-4 treatment-related AE was neutropenia (53.3%). [Ref 259]

Cutaneous T cell lymphoma

Denileukin diftitox-cxdl was approved for the treatment of patients with stage 1 to 3 relapsed/refractory CTCL after at least 1 prior systemic therapy in August 2024. Approval was based on the primary endpoint of ORR from Study 302 (NCT01871727), a single-arm phase III trial. The ORR for denileukin diftitox-cxdl (n = 69) was 36.2% (95% CI 25.0% to 48.7%). Serious AEs occurred in 38% of patients. [Ref 242, 243]

v1.3 Update

Lisocabtagene maraleucel was approved for the treatment of adult patients with R/R large B cell lymphomas, including DLBCL not otherwise specified, DLBCL arising from indolent lymphoma, high-grade B cell lymphoma, primary mediastinal large B cell lymphoma, and FL grade 3B, who have refractory disease to first-line chemoimmunotherapy, or relapse within 12 months of first-line chemoimmunotherapy as well as later relapse in transplant-ineligible patients in June 2022. Lisocabtagene maraleucel was also approved as third-line or later treatment. [Ref 246]

Axicabtagene ciloleucel was FDA-approved for second-line treatment of large B-cell lymphoma and the treatment of R/R large B cell lymphomas (including DLBCL, PMBCL, HGBCL, and transformed FL) after two or more prior lines of systemic therapy in April 2022. [Ref 111]

Tisagenlecleucel was FDA-approved for the treatment of R/R large B cell lymphomas (including DLBCL not otherwise specified, HGBCL, and DLBCL arising from FL) and FL after two or more prior lines of systemic therapy in May 2022. [Ref 113]

Axicabtagene ciloleucel was FDA-approved for the treatment of R/R FL after two or more lines of systemic therapy in March 2021. [Ref 111]

v1.2 Update

Loncastuximab tesirine received an accelerated approval for the treatment of adult patients with relapsed or refractory large B-cell lymphoma after two or more lines of systemic therapy, including DLCBCL not otherwise specified, DLBCL arising from low-grade lymphoma, and HGBCL in April 2021. [Ref 260]

Update v1.4

Table 4 - Food and Drug Administration black box warnings for lymphoma immunotherapies has been updated to include the black box warnings of agents approved since publication of the guideline. See here for revised Table 4.

v1.4 Update

Although not FDA-approved, data for nivolumab + AVD versus BV-AVD for patients with Hodgkin lymphoma have been reported in the phase III SWOG 1826 trial (NCT03907488). For patients treated with nivolumab-AVD (n=487), the median PFS (primary endpoint) was significantly improved in the nivolumab-AVD group (HR for disease progression or death 0.48; 99% CI 0.27 to 0.87; two-sided p=0.001). High-grade AEs were more frequent with BV+AVD, except neutropenia. Hypothyroidism and hyperthyroidism occurred at a rate of 7% and 3%, respectively, in the N+AVD group, compared to <1% and 0%, respectively, in the BV+AVD group. [Ref 255].

Although not FDA-approved, data for nivolumab + chemotherapy and pembrolizumab + chemotherapy for patients with relapsed or refractory Hodgkin lymphoma have been reported in the single-arm phase II trials NCT03016871, NCT03618550, and NCT03077828. In NCT03016871, patients treated with nivolumab (n=34) experienced a CR rate of 71%, while patients treated with nivolumab plus chemotherapy (n=9) experienced a CR rate of 89%. There were no unexpected toxicities observed following nivolumab or nivolumab plus chemotherapy. In NCT03618550, following treatment with pembrolizumab plus gemcitabine, vinorelbine, and liposomal doxorubicin (n=38), the CR rate was 95% (95% CI 82% to 99%). Most AEs were grade 1–2 with this treatment. In NCT03077828, following treatment with pembrolizumab in combination with ifosfamide, carboplatin, and etoposide (n=7) the CR rate was 86.5% (95% CI 71.2% to 95.5%). AEs attributed to pembrolizumab treatment occurred in 81% of patients, and grade 3-4 AEs occurred in 52.4% of patients. [Ref 256–258].

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

In the study "Re-focusing visual working memory during expected and unexpected memory tests" by Sisi Wang and Freek van Ede, the authors investigate the dynamics of attentional re-orienting within visual working memory (VWM). Utilizing a robust combination of behavioral measures, electroencephalography (EEG), and eye tracking, the research presents a compelling exploration of how attention is redirected within VWM under varying conditions. The research question addresses a significant gap in our understanding of cognitive processes, particularly how expected and unexpected memory tests influence the focus and re-focus of attention. The experimental design is meticulously crafted, enabling a thorough investigation of these dynamics. The figures presented are clear and effectively illustrate the findings, while the writing is concise and accessible, making the complex concepts understandable. Overall, this study provides valuable insights into the mechanisms of visual working memory and attentional re-orienting, contributing meaningfully to the field of cognitive neuroscience. Despite the strengths of the manuscript, there are several areas where improvements could be made.

We thank the reviewer for this summary and positive appraisal of our study and our findings. In addition, we are of course grateful for the excellent suggestions for improvements that we have embraced to further strengthen our article. 

Microsaccades or Saccades?

In the manuscript, the terms "microsaccades" and "saccades" are used interchangeably. For instance, "microsaccades" are mentioned in the keywords, whereas "saccades" appear in the results section. It is crucial to differentiate between these two concepts. Saccades are large, often deliberate eye movements used for scanning and shifting attention, while microsaccades are small, involuntary movements that maintain visual perception during fixation. The authors note the connection between microsaccades and attention, but it is not well-recognized that saccades are directly linked to attention. Despite the paradigm involving a fixation point, it remains unclear whether large eye movements (saccades) were removed from the analysis. The authors mention the relationship between microsaccades and attention but do not clarify whether large eye movements (saccades) were excluded from the analysis. If large eye movements were removed during data processing, this should be documented in the manuscript, including clear definitions of "microsaccades" and "saccades." If such trials were not removed, the contribution of large eye movements to the results should be shown, and an explanation provided as to why they should be considered.

We thank the reviewer for raising this relevant point. Before turning to this relevant distinction, we first wish to clarify how, for our main aim of tracking the dynamics of ‘re-orienting in working memory’, any spatial modulation in gaze – be it driven by micro- or macro-saccades – suits this purpose. Having made this explicit, we also fully agree that disambiguating the nature of the saccade bias during internal focusing has additional value.

Because it is notoriously challenging (or at least inherently arbitrary) to draw an absolute fixed boundary between macro- and microsaccades, we instead decided to adopt a two-stage approach to our analysis (building on prior studies from our lab, e.g., de Vries et al., 2023; Liu et al., 2023; Liu et al., 2022). In the first step, we analysed spatial biases in all detected saccades no matter their size (hence our labelling of them as “saccades” when describing these analyses). In a second step, we decomposed and visualized the saccade-rate effect as a function of saccade size in degrees. This second stage directly exposed the ‘nature’ of the saccade bias, as we visualized in Figure 2c (with time on the x axis, saccade size on the y axis, and the spatial modulation color coded). Because these visualizations directly address this major comment, we have now made these key set of results much clearer in our work (we agree that our original visualization of this key aspect of our data was suboptimal). In addition, we have added similar plot for the saccade data in the test-phase in Supplementary Figure S2b.

These complementary analyses show how the saccade bias (more toward than away saccades) is indeed predominantly driven by small saccades (hence are labelling as “micro-saccades” when interpreting our findings), and less so by larger saccades associated with looking back all the way to the location where the memory item had been presented at encoding (positioned at 6 degrees). This is important as it helps to arbitrate between fixational/micro-saccadic eye-movement biases (previously associated with covert and internal attention shifts; cf. de Vries et al., 2023; Engbert and Kliegl, 2003; Hafed and Clark, 2002; Liu et al., 2023; Liu et al., 2022) vs. larger eye movements back to the original locations of the item (previously associated with ‘looking at nothing’ during memory retrieval and imagery; cf. Brandt and Stark, 1997; Ferreira et al., 2008; Johansson and Johansson, 2014; Laeng et al., 2014; Martarelli and Mast, 2013; Spivey and Geng, 2001). By adopting this visualization, we can show this while preserving the richness of our data, and without having to a-priori set an (inherently arbitrary) threshold for classifying saccades as either “macro” or “micro”.

Having explained our rationale, we nevertheless agree with the reviewer that it is worth showing how our time course results hold up when only considering fixational eye movements below 2 visual degrees, which we consider “fixational” provided that our memory stimuli at encoding were presented at 6 visual degrees from central fixation. We show this in Supplementary Figure S1. As can be seen below, our main saccade bias results stay almost the same when restricting our analyses exclusively to fixational saccades within 2 degrees, both when considering our data after the retrocue (Supplementary Figure S1a) as well as after the memory test (Supplementary Figure S1b).

Because we agree this is important complementary data, we have now added this as supplementary figures. In addition, we have added the results to our article. We also point to these additional corroborating findings at key instances in our article:  

Page 5 (Results)

“As in prior studies from our lab with similar experimental set-ups, internal attentional focusing was predominantly driven by fixational micro-saccades (small, involuntary eye-movements around current fixation). To reveal this in the current study, we decomposed and visualized the observed saccade-rate effect as a function of saccade size (Figure 2c), following the same procedure as we have adopted in other recent studies on this bias (de Vries et al., 2023; Liu et al., 2023; Liu et al., 2022). As shown in the saccade-size-over-time plots in Figure 2c, also in the current study, the difference between toward and away saccades (with red colours denoting more toward saccades) was predominantly driven by fixational saccades in the micro-saccades range (< 2°).”

“Moreover, as shown in Supplementary Figure S1a, complementary analyses show that our time course (saccade bias) results hold even when exclusively considering eye movements below 2 visual degrees that we defined as “fixational” provided that the memory items were presented 6 visual degrees from the fixation during encoding. This further corroborates that the bias observed during internal attentional focusing was predominantly driven by fixational micro-saccades rather than looking back to the encoded location of the memory items (cf. Johansson and Johansson, 2014; Richardson and Spivey, 2000; Spivey and Geng, 2001; Wynn et al., 2019).”

Page 7 (Results):

“As shown in the corresponding saccade-size-over-time plots in Supplementary Figure S2b, consistent with what we observed following the cue, the difference between toward and away saccades following the test was again predominantly driven by saccades in the fixational microsaccade range (< 2°), and the time course (saccade bias) results hold even when exclusively considering fixational eye movements below 2 visual degrees (Supplementary Figure S1b). Thus, just like mnemonic focusing after the cue, re-orienting after the memory test was also predominantly reflected in fixational micro-saccades, and not looking back at the original location of the memory items that were encoded at 6 degrees away from central fixation.”

Alpha Lateralization in Attentional Re-orienting

In the attentional orienting section of the results (Figure 2), the authors effectively present EEG alpha lateralization results with time-frequency plots and topographic maps. However, in the attentional reorienting section (Figure 3), these visualizations are absent. It is important to note that the time period in attentional orienting differs from attentional re-orienting, and consequently, the time-frequency plots and topographic maps may also differ. Therefore, it may be invalid to compute alpha lateralization without a clear alpha activity difference. The authors should consider including timefrequency plots and topographic maps for the attentional re-orienting period to validate their findings.

We thank the reviewer also for this constructive suggestion. The reason we did not expand on the time-frequency maps and topographies at the test-stage was the relative lack of alpha effects at the test stage (compared to the clearer alpha modulations after the retrocue). Nevertheless, we agree that including these data will increase transparency and the comprehensiveness of our article. We now added time-frequency plots and topographic maps for alpha lateralization in response to the workingmemory test in Supplementary Figure S2. As can be seen, the time-frequency plots and topographies in the re-focusing period after the working-memory test were consistent with our time-series plots in Figure 3a – reinforcing how alpha lateralization is generally not clear following the working-memory test. In accordance with this relevant addition, we added the following in the revised manuscript:

Page 7 (Results):

“For complementary time-frequency and topographical visualizations, see Supplementary Figure S2a.”

Onset and Offset Latency of Saccade Bias

The use of the 50% peak to determine the onset and offset latency of the saccade bias is problematic. For example, if one condition has a higher peak amplitude than another, the standard for saccade bias onset would be higher, making the observed differences between the onset/offset latencies potentially driven by amplitude rather than the latencies themselves. The authors should consider a more robust method for determining saccade bias onset and offset that accounts for these amplitude differences.

We thank the reviewer for raising this valuable point. We agree that the calculation of onset and offset latencies of the saccade bias could be influenced by the peak amplitude of the waveforms. Thus, we further conducted the Fractional Area Latency (FAL) analysis on the comparison of the saccade bias following the working-memory test between valid cue (expected test) and invalid cue (unexpected test) trials. The FAL analysis has been commonly applied to Event-Related Potentials (ERPs) to estimate the latency of ERP components (Hansen and Hillyard, 1980; Luck, 2005). Instead of relying on the peak latency, the FAL method calculates latency based on a predefined fraction of the area under the waveform. This can provide a more robust measure of component latency. Prompted by this comment, we now also applied FAL analysis to our saccade bias waveforms. This corroborated our original conclusion. Because we believe this is an important complement, we now added these additional outcomes to our article: 

Page 9 (Results): 

“We additionally conducted Fractional Area Latency (FAL) analysis on the comparison of the saccade bias following the memory test between valid- and invalid-cue trials to rule out the potential contribution of peak amplitude differences into the onset and offset latency differences (Hansen and Hillyard, 1980; Kiesel et al., 2008; Luck, 2005). Consistent with our jackknife-based latency analysis, the FAL analysis revealed a significantly prolonged saccade bias following the unexpected tests (the invalid-cue trials) vs. expected tests (the valid-cue trials) in both 80% and 60% cue-reliability conditions (411 ms vs. 463 ms, t₍₁₄₎ = 2.358, p = 0.034; 417 ms vs. 468 ms, t₍₁₅₎ = 2.168, p = 0.047; for 80% and 60%, respectively). Again, there was no significant difference in onset latency following unexpected vs. expected tests. (346 ms vs. 374 ms, t₍₁₄₎ = 2.052, p = 0.060; 353 ms vs. 401 ms, t₍₁₅₎ = 1.577, p = 0.136; for 80% and 60%, respectively).”

In accordance, we also added the following to our Methods:

Page 18 (Methods): 

“In addition to the jackknife-based latency analysis, we further applied a Fractional Area Latency (FAL) method to the saccade bias comparison between validly and invalidly cued memory tests to rule out the contribution of the peak amplitude difference into the onset and offset latency difference (Hansen and Hillyard, 1980; Kiesel et al., 2008; Luck, 2005). We first defined the onset and offset latency of the saccade bias as the first time point at which 25% or 75% of the total area of the component has been reached, relative to a lower boundary of a difference of 0.3 Hz between toward and away saccades (to remove the influence of noise fluctuations in our difference time course below this lower boundary). The extracted onset and offset latency for all participants was then compared using paired-samples t-tests.”

Control Analysis for Trials Not Using the Initial Cue

The control analysis for trials where participants did not use the initial cue raises several questions:

(1) The authors claim that "unlike continuous alpha activity, saccades are events that can be classified on a single-trial level." However, alpha activity can also be analyzed at the single-trial level, as demonstrated by studies like "Alpha Oscillations in the Human Brain Implement Distractor Suppression Independent of Target Selection" by Wöstmann et al. (2019). If single-trial alpha activity can be used, it should be included in additional control analyses.

We agree with the reviewer that alpha activity can also be analyzed at the single-trial level. However, because alpha is a continuous signal, single-trial alpha activity will necessarily be graded (trials with more or less alpha power). This is still different from saccades, that are not continuous signals but true ‘events’ (either a saccade was made, or no saccade was made, with no continuum in between). Because of this unique property, it is possible to sort trials by whether a saccade was present (and, if present, by its direction), in an all-or-none way that is not possible for alpha activity that can only be sorted by its graded amplitude/power. This is the key distinction underlying our motivation to sort the trials based on saccades, as we now make clearer: 

Page 10 (Results): 

“Although alpha can also be analyzed as the single trial level (e.g. Macdonald et al., 2011; Wöstmann et al., 2019; for a review, see Kosciessa et al., 2020), saccades offer the unique opportunity to split trials not by graded amplitude fluctuations but by discrete all-or-none events.” 

In addition, please note how our saccade markers were also more reliable/sensitive, especially in the subsequent memory-test-phase of interest. This is another reason we decided to focus this control analysis on saccades and not alpha activity. 

(2) The authors aimed to test whether the re-orienting signal observed after the test is not driven exclusively by trials where participants did not use the initial cue. They hypothesized that "in such a scenario, we should only observe attention deployment after the test stimulus in trials in which participants did not use the preceding retro cue." However, if the saccade bias is the index for attentional deployment, the authors should conduct a statistical test for significant saccade bias rather than only comparing toward-saccade after-cue trials with no-toward-saccade after-cue trials. The null results between the two conditions do not immediately suggest that there is attention deployment in both conditions.

We thank the reviewer for bringing up this important point. We fully agree and, in fact, we had conducted the relevant statistical analysis for each of the conditions separately (in addition to their comparison). Upon reflection, we came to realize that in our original submission it was easy to overlook this point, and therefore thank the reviewer for flagging this. To make this clearer, we now also added the relevant statistical clusters in Figure 4a,b and more clearly report them in the associated text: 

Page 10 (Results):

“As we show in Figure 4a,b, we found clear gaze signatures of attentional deployment in response to expected (valid) memory tests, no matter whether we had pre-selected trials in which we had also seen such deployment after the cue in gaze (cluster P: 0.115, 0.041, 0.027, <0.001 for 80%-valid, 60%-valid, 80%-invalid, 60%-invalid trials, respectively), or not (cluster P: 0.016, 0.009, 0.001, <0.001 for 80%-valid, 60%-valid, 80%-invalid, 60%-invalid trials, respectively).”

(3) Even if attention deployment occurs in both conditions, the prolonged re-orienting effect could also be caused by trials where participants did not use the initial cue. Unexpected trials usually involve larger and longer brain activity. The authors should perform the same analysis on the time after the removal of trials without toward-saccade after the cue to address this potential confound.

We thank the reviewer for raising this. It is crucial to point out, however, that after any given 80% or 60% reliable cue, the participants cannot yet know whether the subsequent memory test in that trial will be expected (valid cue) or unexpected (invalid cue). Accordingly, the prolonged re-orienting after unexpected vs. expected memory tests cannot be explained by differential use of the cue (i.e., differential cue-use cannot be a “confound” for differential responses to expected and unexpected memory tests, as observed within the 80 and 60% cue-reliability conditions). 

Reviewer #2 (Public Review):

Summary:

This study utilized EEG-alpha activity and saccade bias to quantify the spatial allocation of attention during a working memory task. The findings indicate a second stage of internal attentional deployment following the appearance of a memory test, revealing distinct patterns between expected and unexpected test trials. The spatial bias observed during the expected test suggests a memory verification process, whereas the prolonged spatial bias during the unexpected test suggests a reorienting response to the memory test. This work offers novel insights into the dynamics of attentional deployment, particularly in terms of orienting and re-orienting following both the cue and memory test.

Strengths:

The inclusion of both EEG-alpha activity and saccade bias yields consistent results in quantifying the attentional orienting and re-orienting processes. The data clearly delineate the dynamics of spatial attentional shifts in working memory. The findings of a second stage of attentional re-orienting may enhance our understanding of how memorized information is retrieved.

Weaknesses:

Although analyses of neural signatures and saccade bias provided clear evidence regarding the dynamics of spatial attention, the link between these signatures and behavioral performance remains unclear. Given the novelty of this study in proposing a second stage of 'verification' of memory contents, it would be more informative to present evidence demonstrating how this verification process enhances memory performance.

We thank the reviewer for the positive summary of our work and for highlighting key strengths. We also appreciate the constructive suggestions, such as addressing the link between our observed refocusing signals and behavioral performance in our task. We now performed these additional analyses and added their outcomes to the revised article, as we detail in response to comment 2 below.  

Reviewer #2 (Recommendations For The Authors):

(1) Figure 2 shows graded spatial modulations in both EEG-alpha activity and saccade bias. However, while the imperative 100% cue conditions and 100% validity conditions largely overlap in EEG-alpha activity, a clear difference is present between these two conditions in saccade bias. The cause of the difference in saccade bias is unclear.

We thank the reviewer for pointing out this interesting difference. At this stage, it is hard to know with certainty whether this reflects a genuine difference in our 100% reliable and 100% imperative cue conditions that is selectively picked up by our gaze but not alpha marker. Alternatively, this may reflect differential sensitivity of our two markers to different sources of noise. Either way, we agree that this observation is worth calling out and reflecting on when discussing these results: 

Page 6 (Results):  

“It’s worth noting that while alpha lateralization shows very comparable amplitudes in the imperative-100% and 100% conditions, the saccade bias was larger following imperative-100% vs. 100% reliable cues. This may reflect a difference between these two cueing conditions that is selectively picked up by our gaze marker (though it may also reflect differential sensitivity of our two markers to different sources of noise). […]”

(2) Figure 3 shows signatures of attentional re-orienting after the memory test presented at the center. When the cue was not 100% valid, a noticeable saccade bias towards the memorized location of the test item was observed. This finding was explained as reflecting a re-orienting to the mnemonic contents. To strengthen this interpretation, I suggest providing evidence for the link between the attentional re-orienting signatures and memory performance.

We thank the reviewer for this constructive suggestion. We now sorted trials by behavioral performance using a median split on RT (fast-RT vs. slow-RT trials) and reproduction error (highaccuracy vs. low-accuracy trials).  Because we believe the outcomes of these analyses increase transparency as well as the comprehensiveness of our article, we have now included them as Supplementary Figure S3.

As shown below, we were able to link the saccade bias following the memory test to subsequent performance, but this reached significance only for the 80% valid-cue trials when splitting by RT (cluster P = 0.001). For the other conditions, we could not establish a reliable difference by our performance splits. Possibly this is due to a lack of sensitivity, given the relatively large number of conditions we had and, consequently, the relatively small number of trials we therefore had per condition (particularly in the invalid-cue condition with unexpected memory tests). We now bring forward these additional outcomes at the relevant section in our Results: 

Page 7 (Results):

“We also sorted patterns of gaze bias after the memory test by performance but could only establish a link between this gaze bias and RT following expected memory tests in our 80% cuereliability condition (cluster P = 0.001, Supplementary Figure S3). The lack of significant statistical differences in the remaining conditions may possibly reflect a lack of sensitivity (insufficient trial numbers) for this additional analysis.”

(3) When comparing the time course of attentional re-orienting after the memory test, a prolonged attentional re-orienting was observed for unexpected memory tests compared to the expected ones. While the onset latency was similar for unexpected and expected memory tests, the offset latency was prolonged for the unexpected memory test. Could this be attributed to the learned tendency to saccade toward the expected location in more valid trials? In this case, the prolonged re-orienting may indicate increased efforts in suppressing the previously learned tendency.

We thank the reviewer for bringing up this interesting possibility. In our original interpretation, this prolonged signal reflects a longer time needed to bring the unexpected memory content ‘back in focus’ before being able to report its orientation. At the same time, we agree that there are alternative explanations possible, such as the one raised by the reviewer. We now make this clearer when discussing this finding: 

Page 14 (Discussion): 

“[…] attentional deployment did become prolonged when re-focusing the unexpected memory item, likely reflecting prolonged effort to extract the relevant information from the memory item that was not expected to be tested. However, there may also be alternative accounts for this observation, such as suppressing a learned tendency to saccade in the direction of the expected item following an unexpected memory test.”

(4) To test whether the re-orienting signature is predominantly influenced by trials where participants delayed the use of cue information until the memory test appeared, the authors sorted the trials based on saccade bias after the initial cue. However, it would be more informative to depict the reorienting patterns by sorting trials based on memory performance. The rationale is that in trials where participants delayed using the initial retro-cue, memory performance (e.g., measured by reproduction error) might be less precise due to the extended memory retention period. Compared to saccade bias for initial orienting, memory performance could provide more reliable evidence as it represents a more independent measure.

We thank the reviewer for this suggestion. As delineated in response to comment 2, we now conducted this additional analysis and added the relevant outcomes to our article.  

(5) While the number of trials was well-balanced across blocks (~ 240 trials), how did the authors address the imbalance between valid and invalid trials, especially in the 80% cue validity block?

We thank the reviewer for raising this point.  First, we wish to point out that while trial numbers will indeed impact the sensitivity for finding an effect, trial numbers do not bias the mean – and therefore also not the comparison between means. In this light, it is vital to appreciate that our findings do not reflect a significant effect in valid trials but no significant effect in invalid trials (which we agree could be due to a difference in trial numbers), but rather a statistical difference between valid and invalid trials. This significant difference in the means between valid and invalid true cannot be attributed to a difference in trial numbers between these conditions. 

Having clarified this, we nevertheless agree that it is also worthwhile to empirically validate this assertion and show how our findings hold even when carefully matching the number of trials between valid and invalid conditions (i.e., between expected and unexpected memory tests). To do so, we ran a sub-sampling analysis where we sub-sampled the number of valid trials to match the number of invalid trials available per condition (and averaged the results across 1000 random sub-samplings to increase reliability). As anticipated, this replicated our findings of robust differences between the gaze bias following expected and unexpected memory tests in both our 80 and 60% cue-reliability conditions. We now present these additional outcomes in Supplementary Figure S4.

Because we agree this is an important re-assuring control analysis, we have now added this to our article:

Page 9 (Results):

“To rule out the possibility that the saccade-bias differences following expected and unexpected memory tests are caused by uneven trial numbers (200 vs. 50 trials in the 80% cuereliability condition, 150 vs. 100 trials in the 60% cue-reliability condition), we ran a subsampling analysis where we sub-sampled the number of valid trials to match the number of invalid trials available per condition (averaging the results across 1000 random sub-samplings to increase reliability). As shown in Supplementary Figure S4, this complementary subsampling analysis confirmed that our observed differences between the saccade bias following expected and unexpected memory tests in both 80% and 60% cue-reliability conditions are robust even when carefully matching the number of trials between validly cued (expected) and invalidly cued (unexpected) memory test.”

Reviewer #3 (Public Review):

Summary:

Wang and van Ede investigate whether and how attention re-orients within visual working memory following expected and unexpected centrally presented memory tests. Using a combination of spatial modulations in neural activity (EEG-alpha lateralization) and gaze bias quantified as time courses of microsaccade rate, the authors examined how retro cues with varying levels of reliability influence attentional deployment and subsequent memory performance. The conclusion is that attentional reorienting occurs within visual working memory, even when tested centrally, with distinct patterns following expected and unexpected tests. The findings provide new value for the field and are likely of broad interest and impact, by highlighting working memory as an action-bound process (in)dependent on (an ambiguous) past.

Strengths:

The study uniquely integrates behavioral data (accuracy and reaction time), EEG-alpha activity, and gaze tracking to provide a comprehensive analysis of attentional re-orienting within visual working memory. As typical for this research group, the validity of the findings follows from the task design that effectively manipulates the reliability of retro cues and isolates attentional processes related to memory tests. The use of well-established markers for spatial attention (i.e. alpha lateralization) and more recently entangled dependent variable (gaze bias) is commendable. Utilizing these dependent metrics, the concise report presents a thorough analysis of the scaling effects of cue reliability on attentional deployment, both at the behavioral and neural levels. The clear demonstration of prolonged attentional deployment following unexpected memory tests is particularly noteworthy, although there are no significant time clusters per definition as time isn't a factor in a statistical sense, the jackknife approach is convincing. Overall, the evidence is compelling allowing the conclusion of a second stage of internal attentional deployment following both expected and unexpected memory tests, highlighting the importance of memory verification and re-orienting processes.

Weaknesses:

I want to stress upfront that these weaknesses are not specific to the presented work and do not affect my recommendation of the paper in its present form.

The sample size is consistent with previous studies, a larger sample could enhance the generalizability and robustness of the findings. The authors acknowledge high noise levels in EEG-alpha activity, which may affect the reliability of this marker. This is a general issue in non-invasive electrophysiology that cannot be handled by the authors but an interested reader should be aware of it. Effectively, the sensitivity of the gaze analysis appears "better" in part due to the better SNR. The latter also sets the boundaries for single-tiral analyses as the authors correctly mention. In terms of generalizability, I am convinced that the main outcome will likely generalize to different samples and stimulus types. Yet, as typical for the field future research could explore different contexts and task demands to validate and extend the findings. The authors provide here how and why (including sharing of data and code).

We thank the reviewer for summarising our work and for carefully delineating its strengths. We also appreciate the mentioning of relevant generic limitations and agree that important avenues for future studies will be to expand this work with larger sample sizes, complementary measurement techniques, and complementary task contexts and stimuli.    

Reviewer #3 (Recommendations For The Authors):

In the conclusion, Wang and van Ede successfully demonstrate that attentional re-orienting occurs within visual working memory following both expected and unexpected tests. The conclusions are supported by the data and analyses applied, showing that attentional deployment is by the reliability of retro cues. Centrally presented memory tests can invoke either a verification or a revision of internal focus, the latter thus far not considered in both theory and experimental design in cognitive neuroscience.

I don't have any recommendations that will significantly change the conclusions.

We thank the reviewer for having carefully evaluated our work and hope the reviewer will also perceive the changes we made and the additional analyses we added in responses to the other two reviewers as further strengthening our article.

Reference

Brandt SA, Stark LW. 1997. Spontaneous eye movements during visual imagery reflect the content of the visual scene. J Cogn Neurosci 9. doi:10.1162/jocn.1997.9.1.27

de Vries E, Fejer G, van Ede F. 2023. No obligatory trade-off between the use of space and time for working memory. Communications Psychology.

Engbert R, Kliegl R. 2003. Microsaccades uncover the orientation of covert attention. Vision Res 43. doi:10.1016/S0042-6989(03)00084-1

Ferreira F, Apel J, Henderson JM. 2008. Taking a new look at looking at nothing. Trends Cogn Sci 12. doi:10.1016/j.tics.2008.07.007

Hafed ZM, Clark JJ. 2002. Microsaccades as an overt measure of covert attention shifts. Vision Res 42. doi:10.1016/S0042-6989(02)00263-8

Hansen JC, Hillyard SA. 1980. Endogeneous brain potentials associated with selective auditory attention. Electroencephalogr Clin Neurophysiol 49. doi:10.1016/0013-4694(80)90222-9

Johansson R, Johansson M. 2014. Look Here, Eye Movements Play a Functional Role in Memory Retrieval. Psychol Sci 25. doi:10.1177/0956797613498260

Kiesel A, Miller J, Jolicœur P, Brisson B. 2008. Measurement of ERP latency differences: A comparison of single-participant and jackknife-based scoring methods. Psychophysiology 45. doi:10.1111/j.1469-8986.2007.00618.x

Kosciessa JQ, Grandy TH, Garrett DD, Werkle-Bergner M. 2020. Single-trial characterization of neural rhythms: Potential and challenges. Neuroimage 206. doi:10.1016/j.neuroimage.2019.116331

Laeng B, Bloem IM, D’Ascenzo S, Tommasi L. 2014. Scrutinizing visual images: The role of gaze in mental imagery and memory. Cognition 131. doi:10.1016/j.cognition.2014.01.003

Liu B, Alexopoulou SZ, van Ede F. 2023. Jointly looking to the past and the future in visual working memory. Elife.

Liu B, Nobre AC, van Ede F. 2022. Functional but not obligatory link between microsaccades and neural modulation by covert spatial attention. Nat Commun 13. doi:10.1038/s41467-022-312173

Luck S. 2005. Ten Simple Rules for Deisgning ERP Experiments. Event-related potentials: A methods handbook.

Macdonald JSP, Mathan S, Yeung N. 2011. Trial-by-trial variations in subjective attentional state are reflected in ongoing prestimulus EEG alpha oscillations. Front Psychol 2. doi:10.3389/fpsyg.2011.00082

Martarelli CS, Mast FW. 2013. Eye movements during long-term pictorial recall. Psychol Res 77. doi:10.1007/s00426-012-0439-7

Richardson DC, Spivey MJ. 2000. Representation, space and Hollywood Squares: Looking at things that aren’t there anymore. Cognition 76. doi:10.1016/S0010-0277(00)00084-6

Spivey MJ, Geng JJ. 2001. Oculomotor mechanisms activated by imagery and memory: Eye movements to absent objects. Psychol Res 65. doi:10.1007/s004260100059

van Ede F, Chekroud SR, Nobre AC. 2019. Human gaze tracks attentional focusing in memorized visual space. Nat Hum Behav. doi:10.1038/s41562-019-0549-y

Wöstmann M, Alavash M, Obleser J. 2019. Alpha oscillations in the human brain implement distractor suppression independent of target selection. Journal of Neuroscience 39. doi:10.1523/JNEUROSCI.1954-19.2019

Wynn JS, Shen K, Ryan JD. 2019. Eye movements actively reinstate spatiotemporal mnemonic content. Vision (Switzerland) 3. doi:10.3390/vision3020021

Reviewer #2 (Public review):

Summary:

Carabalona and colleagues investigated the role of the membrane-deforming cytoskeletal regulator protein Abba (MTSS1L/MTSS2) in cortical development to better understand the mechanisms of abnormal neural stem cell mitosis. The authors used short hairpin RNA targeting Abba20 with a fluorescent reporter coupled with in utero electroporation of E14 mice to show changes to neural progenitors. They performed flow cytometry for in-depth cell cycle analysis of Abba-shRNA impact to neural progenitors and determined an accumulation in S phase. Using culture rat glioma cells and live imaging from cortical organotypic slides from mice in utero electroporated with Abba-shRNA, the authors found Abba played a prominent role in cytokinesis. They then used a yeast-two-hybrid screen to identify three high confidence interactors: Beta-Trcp2, Nedd9, and Otx2. They used immunoprecipitation experiments from E18 cortical tissue coupled with C6 cells to show Abba requirement for Nedd9 localization to the cleavage furrow/cytokinetic bridge. The authors performed an shRNA knockdown of Nedd9 by in utero electroporation of E14 mice and observed similar results as with the Abba-shRNA. They tested a human variant of Abba using in utero electroporation of cDNA and found disorganized radial glial fibers and misplaced, multipolar neurons, but lacked the impact of cell division seen in the shRNA-Abba model.

Strengths:

Fundamental question in biology about the mechanics of neural stem cell division.<br /> Directly connecting effects in Abba protein to downstream regulation of RhoA via Nedd9.<br /> Incorporation of human mutation in ABBA gene.<br /> Use of novel technologies in neurodevelopment and imaging.

Weaknesses:

Unexplored components of the pathway (such as what neurogenic populations are impacted by Abba mutation) and unleveraged aspects of their data (such as the live imaging) limit the scope of their findings and left significant questions about the effect of ABBA on radial glia development.

(1) Claim of disorganized radial glial fibers lacks quantifications.<br /> -On page 11, the authors claim that knockdown of Abba lead to changes in radial glial morphology observed with vimentin staining. Here they claim misoriented apical processes, detached end feet, and decreased number of RGP cells in the VZ. However, they no not provide quantification of process orientation to better support their first claim. Measurements of radial glia fiber morphology (directionality, length) and of angle of division would be metrics that can be applied to data. Some of these analysis could be done in their time-lapse microscopy images, such as to quantify the number of cell division during their period of analysis (though that is short-15 hours).

(2) Unclear where effect is:<br /> -in RG or neuroblasts? Is it in cell cleavage that results in accumulation of cells at VZ (as sometimes indicated by their data like in Fig 2A or 4D)? Interrogation of cell death (such as by cleaved caspase 3) would also help. Given their time lapse, can they identify what is happening to the RG fiber? The authors describe a change in "migration" but do not show evidence for this for either progenitor or neuroblast populations. Given they have nice time-lapse imaging data, could they visualize progenitor versus young neuron migration? Analysis of neuroblasts (such as with doublecortin expression in the tissue) would also help understand any issues in migration (of neurons v stem cells).<br /> -at cleaveage furrow? In abscission? There is high resolution data that highlights the cleavage furrow as the location of interest (fig 3A), however there is also data (fig 3B) to suggest Abba is expressed elsewhere as well and there is an overall soma decrease. More detail of the localization of Abba during the division process would be helpful-for example, could cleavage furrow proteins, such as Aurora B, co-localization (and potentially co-IP) help delineate subpopulations of Abba protein? Furthermore, the FRET imaging is unique way to connect their mutation with function-could they measure/quantify differences at furrow compared to rest of soma to further corroborate that Abba-associated RhoA effect was furrow-enriched?<br /> -The data highlights nicely that a furrow doesn't clearly form when ABBA expression and subsequent RhoA activity are decreased (in Fig 3 or 5A). Does this lead to cells that can't divide because of poor abscission, especially since "rounding" still occurs? Or abnormal progenitors (with loss of fiber or inability to support neuroblast migration)? Or abnormal progression of progenitors to neuroblasts?

(3) Limited to a singular time point of mouse cortical development<br /> On page 13, the authors outline the results of their Y2H screen with the identification of three high confidence interactors. Notably, they used a E10.5-E12.5 mouse brain embryo library rather than one that includes E14, the age of their in utero electroporation mice. Many of the authors' claims focus on in utero electroporation of shRNA-Abba of E14 mice that are then evaluated at E16-18. Justification for the focus on this age range should be included to support that their findings can then be applied to all of mouse corticogenesis.

(4) Detail of the effect of the human variant of the ABBA mutation in mouse is lacking.<br /> Their identification of the R671W mutation is interesting and the IUE model warrants more characterization, as they did with their original KD experiments.<br /> -Could they show that Abba protein levels are decreased (in either cell lines or electroporated tissue)?<br /> -While time-lapse morphology might not have been performed, more analysis on cell division phenotype (such as plane of division and radial glia morphology) would be helpful.

The resubmission has addressed many of the questions raised.

I have a few comments that should be addressed:

(1) The authors maintain a deficit in "migration of immature neurons" which remains unsubstantiated. In their resonse, they state: "we believe that the data showing the accumulation of migrating electroporated cells in the ventricular (V) and subventricular (SV) zones provide compelling evidence of abnormal migration in ABBA-shRNA electroporated cells. "<br /> -Firstly, they do not demonstrate that it's immature neurons, not RGs, that are affected. Secondly, accumulation of cells at the V-SVZ could be due to soley the inability for the RGC to undergo mitosis, therefore remaining stuck"<br /> The commentary of migration, especially of neurons, should be modified.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

The manuscript investigates the role of the membrane-deforming cytoskeletal regulator protein Abba in cortical development and its potential implications for microcephaly. It is a valuable contribution to the understanding of Abba's role in cortical development. The strengths and weaknesses identified in the manuscript are outlined below:

Clinical Relevance:

The authors identified a patient with microcephaly and a patient with an intellectual disability harboring a mutation in the Abba variant (R671W) adding a clinically relevant dimension to the study.

Mechanistic Insights:

The study offers valuable mechanistic insights into the development of microcephaly by elucidating the role of Abba in radial glial cell proliferation, radial fiber organization, and the migration of neuronal progenitors. The identification of Abba's involvement in the cleavage furrow during cell division, along with its interaction with Nedd9 and positive influence on RhoA activity, adds depth to our understanding of the molecular processes governing cortical development. Though the reported results establish the novel interaction between Abba and Nedd9, the authors have not addressed whether the mutant protein loses this interaction and whether that results in the observed effects.

We appreciate the reviewer’s observation and fully agree that our study does not provide direct evidence that the phenotypes induced by the R671W mutant are mediated through NEDD9. We sincerely apologize if the manuscript inadvertently conveyed this impression.

While we show that the interaction with NEDD9 plays a role in the action of ABBA, our findings suggest that NEDD9 and RhoA activation have a minor influence on the phenotypes induced by this mutation, as highlighted by the evidence we presented.

We would like to point out that we have previously addressed this point in the discussion section of the manuscript. For clarity, below is an excerpt from that section:

“heterozygous expression of the human R671W variant would exert a dominant negative effect on ABBA's role in brain development, leading to microcephaly and cognitive delay. This notion is supported by recent work disclosing additional patient carrying the R671W variant42. In the same study the significant neurological phenotypes were observed in a drosophila model where the ortholog of human MTSS2 and MTSS1 mim was deleted.   However, from a clinical genetics’ standpoint, it is unlikely to find patients with the recurrent R671W mutation without any homozygous or compound heterozygous loss-of-function mutations elsewhere in the ABBA gene. This could also suggest a gain-of-function effect of the R671W mutation. Supporting this notion, overexpressing ABBA-R671W in cells expressing the wild-type Abba in this study did not result in a dominant-negative decrease in RhoA activation, nor did it affect the expression of PH3 in vivo. These findings make it plausible to suggest that a mechanism responsible for the phenotype associated with overexpression of the human variant may primarily involve post-cell division processes, such as cell migration. “

We have made corrections to the new version of the manuscript to emphasize this further.

In Vivo Validation:

The overexpression of mutant Abba protein (R671W) resulting in phenotypic similarities to Abba knockdown effects supports the significance of Abba in cortical development.

Reviewer #2 (Public Review):

Summary:

Carabalona and colleagues investigated the role of the membrane-deforming cytoskeletal regulator protein Abba (MTSS1L/MTSS2) in cortical development to better understand the mechanisms of abnormal neural stem cell mitosis. The authors used short hairpin RNA targeting Abba20 with a fluorescent reporter coupled with in-utero electroporation of E14 mice to show changes to neural progenitors. They performed flow cytometry for in-depth cell cycle analysis of Abba-shRNA impact on neural progenitors and determined an accumulation in the S phase. Using culture rat glioma cells and live imaging from cortical organotypic slides from mice in utero electroporated with Abba-shRNA, the authors found Abba played a prominent role in cytokinesis. They then used a yeast-two-hybrid screen to identify three high-confidence interactors: Beta-Trcp2, Nedd9, and Otx2. They used immunoprecipitation experiments from E18 cortical tissue coupled with C6 cells to show Abba's requirement for Nedd9 localization to the cleavage furrow/cytokinetic bridge. The authors performed a shRNA knockdown of Nedd9 by in-utero electroporation of E14 mice and observed similar results as with the Abba-shRNA. They tested a human variant of Abba using in-utero electroporation of cDNA and found disorganized radial glial fibers and misplaced, multipolar neurons, but lacked the impact of cell division seen in the shRNA-Abba model.

Strengths:

A fundamental question in biology about the mechanics of neural stem cell division.

Directly connecting effects in Abba protein to downstream regulation of RhoA via Nedd9.

Incorporation of human mutation in ABBA gene.

Use of novel technologies in neurodevelopment and imaging.

Weaknesses:

Unexplored components of the pathway (such as what neurogenic populations are impacted by Abba mutation) and unleveraged aspects of their data (such as the live imaging) limit the scope of their findings and leave significant questions about the effect of ABBA on radial glia development.

(1) The claim of disorganized radial glial fibers lacks quantifications.

On page 11, the authors claim that knockdown of Abba leads to changes in radial glial morphology observed with vimentin staining. Here they claim misoriented apical processes, detached end feet, and decreased number of RGP cells in the VZ. However, they do not provide quantification of process orientation to better support their first claim. Measurements of radial glia fiber morphology (directionality, length) and angle of division would be metrics that can be applied to data.

In the corrected version of the manuscript, we provide new qualification of changes in dispersion of vimentin immunostaining (Supplementary Figure 1).

Some of these analyses could be done in their time-lapse microscopy images, such as to quantify the number of cell divisions during their period of analysis (though that is short-15 hours).

This is indeed a very good idea. We have reanalyzed the recordings to follow cell division. Unfortunately, the number of cells that we were able to follow was low, making statistical analysis of the data unreliable.  As the reviewer alluded in the comment longer recording times than 15h are required to make reliable conclusion. Instead, we have performed live-cell imaging using Aniling-GFP coelectroporeted with RFP as a marker of mitotic progression . We monitored the distribution of cells showing accumulation of Anillin-GFP in control (Scramble) and ABBA-shRNA3 conditions (this data was added to new Supplementary Figure 3). Anillin has been shown to be an efficient tool to monitor cell division in vivo as in particular as it displays accumulation and correlated increase intensity of Anillin-GFP ((Hesse et al Nature Com. 2012, DOI: 10.1038/ncomms2089).

(2) It is unclear where the effect is:

-In RG or neuroblasts? Is it in cell cleavage that results in the accumulation of cells at VZ (as sometimes indicated by their data like in Figure 2A or 4D)?

The data suggest that radial glial (RG) cells are indeed blocked prior to abscission. This phenomenon might contribute to the accumulation of cells at the ventricular zone (VZ), as indicated by observations such as those in Figure 2A and 4D. The interruption in cell cleavage likely prevents the proper progression of division, causing RG cells to remain at the VZ rather than proceeding with their normal differentiation or migration processes. This finding highlights a potential mechanistic link between disrupted abscission and cell accumulation in the VZ.

Interrogation of cell death (such as by cleaved caspase 3) would also help.  

Caspase-3 cleavage is widely used as a marker for apoptosis; however, it may not be the most reliable tool for monitoring apoptosis during brain cortical development. The developing brain is a highly dynamic environment where caspase-3 activation can be transient and involved in non-apoptotic processes, such as synaptic pruning and neuronal remodeling. This makes it challenging to distinguish caspase-3 activity associated with apoptosis from its roles in physiological processes.

In contrast, monitoring overall cell survival provides a more reliable measure of developmental outcomes, as it reflects the net balance of cell death and survival mechanisms. By focusing on cell survival e.g. quantification of number of RGP, we can better assess the functional consequences of apoptosis and its interplay with neurogenesis and other developmental processes.  In line with this we have added more data on the quantification of RGPC as well as their distribution in new Supplementary Figure 3. 

Given their time-lapse, can they identify what is happening to the RG fiber?

Both apical and basal endfeet appear to detach and retract prior to radial glial (RG) cell death. This is evident in Figure 1D, as well as from our observation of cellular bodies located far from the ventricular surface (VS), as demonstrated in the new Supplementary Figure 3.

The authors describe a change in "migration" but do not show evidence for this for either progenitor or neuroblast populations. Given they have nice time-lapse imaging data, could they visualize progenitor versus young neuron migration? Analysis of neuroblasts (such as with doublecortin expression in the tissue) would also help understand any issues in migration (of neurons v stem cells).

This is an excellent question that arises from the extensive data presented in this study. Addressing it would require repeating a significant portion of the experiments. We fully agree with the reviewer that these are important and obvious questions that warrant a dedicated study to answer them thoroughly. Additionally, we believe that the data showing the accumulation of migrating electroporated cells in the ventricular (V) and subventricular (SV) zones provide compelling evidence of abnormal migration in ABBA-shRNA electroporated cells.

-At cleavage furrow? In abscission? There is high-resolution data that highlights the cleavage furrow as the location of interest (Figure 3A), however, there is also data (Figure 3B) to suggest Abba is expressed elsewhere as well and there is an overall soma decrease. More detail of the localization of Abba during the division process would be helpful for example, could cleavage furrow proteins, such as Aurora B, co-localization (and potentially co-IP) help delineate subpopulations of Abba protein? Furthermore, the FRET imaging is a unique way to connect their mutation with function - could they measure/quantify differences at furrow compared to the rest of soma to further corroborate that the Abba-associated RhoA effect was furrow-enriched?

In the corrected version of the manuscript, we include new quantification of RhoA activity in the region corresponding to the cleavage furrow (New Figure 5), This new data show similar results as the previous and indicate that the changes observed are primarily derived from the cleavage furrow region. In the future a detailed dissection of the molecules involved in the mechanism would be highly desirable. These notions are now included in the discussion. 

-The data highlights nicely that a furrow doesn't clearly form when ABBA expression and subsequent RhoA activity are decreased (in Figure 3 or 5A). Does this lead to cells that can't divide because of poor abscission, especially since "rounding" still occurs? Or abnormal progenitors (with loss of fiber or inability to support neuroblast migration)? Or abnormal progression of progenitors to neuroblasts?

Our findings, combined with previous results, suggest multiple mechanisms through which ABBA depletion and subsequent Nedd9 and RhoA signaling disruptions could impact progenitor cells and neuroblasts. Below is a detailed response to each question: 

(1) Do cells fail to divide due to poor abscission?

Nedd9 is a key regulator of RhoA signaling, which could be essential for cleavage furrow ingression and abscission. Reduced Nedd9 expression may leads to non-activation of RhoA, thereby impairing cleavage furrow ingression. Furthermore, since RhoA deactivation is critical for successful abscission, any disruption in this signaling pathway could compromise the final stages of cytokinesis. While we do not directly observe failed abscission, the impaired furrow formation in Figure 3 and 5A aligns with the hypothesis that some cells may struggle to complete division due to defects in RhoA-mediated abscission. 

(2) Are abnormal progenitors generated (e.g., loss of fiber or inability to support neuroblast migration)?

Disrupted Nedd9 expression not only affects cell cycle progression but also influences the structural integrity of radial glial progenitors (RGPs). RGPs with impaired cleavage furrow ingression may exhibit detachment of apical and basal endfeet (Supplementary Figure 3), leading to abnormalities in their scaffold function. This structural disruption likely contributes to the accumulation of electroporated cells in the ventricular (V) and subventricular (SV) zones (Figure 5A), supporting the idea that abnormal progenitors fail to support proper neuroblast migration. 

(3) Is there abnormal progression of progenitors to neuroblasts?

Given that Nedd9 triggers cells to enter mitosis, its impaired function may prevent progenitors from properly progressing through the cell cycle, causing cell cycle arrest and eventual decrease survival. This would directly impact the ability of progenitors to transition into neuroblasts. Moreover, the abnormal membrane composition and PI(4,5)P2 enrichment we hypothesize during cytokinesis could disrupt ABBA recruitment and its interaction with Nedd9. This disruption would impair RhoA activation, further compromising the progression of progenitors to neuroblasts. 

In conclusion, our findings suggest that impaired ABBA expression disrupts Nedd9 and RhoA signaling, leading to poor cleavage furrow ingression, abnormal progenitor structure, and defective neuroblast migration. These processes collectively contribute to developmental defects in the cortex. Future studies focusing on live imaging of cytokinesis and cell fate mapping will help elucidate better these mechanisms further.

(3) Limited to a singular time point of mouse cortical development

On page 13, the authors outline the results of their Y2H screen with the identification of three high-confidence interactors. Notably, they used an E10.5-E12.5 mouse brain embryo library rather than one that includes E14, the age of their in-utero electroporation mice. Many of the authors' claims focus on in-utero electroporation of shRNA-Abba of E14 mice that are then evaluated at E16-18. Justification for the focus on this age range should be included to support that their findings can then be applied to all mouse corticogenesis.

We thank the reviewer to point this out. Indeed, the data suggest that the interaction between ABBA and Nedd9 occurs before E14. The reason to address the questions at E14 is that in earlier work, we have shown that ABBA is mainly expressed through E10.5-12.5 in the floorplate structure formed by radial glia. The radial glia-specific expression was confirmed through double staining with radial glial (RC2) and neuronal (Tuj1) markers at E12.5 (see Saarikangas et al. J. Cell Sci. 121:1444-1454, 2008). Thus, we consider the Y2H library relevant for identifying ABBA's interactors within radial glia. We have specified this better in the corrected manuscript.

(4) Detail of the effect of the human variant of the ABBA mutation in mice is lacking.

Their identification of the R671W mutation is interesting and the IUE model warrants more characterization, as they did with their original KD experiments.

We have now included addition data in the corrected manuscript showing R671W dependent changes in INM (Supplementary Figure 3 )

Could they show that Abba protein levels are decreased (in either cell lines or electroporated tissue)?

Estimation of ABBA expression in cell expressing ABBA R671W as in Supplemental Figure 5 did not show significant change.

-While time-lapse morphology might not have been performed, more analysis on cell division phenotype (such as plane of division and radial glia morphology) would be helpful. 

This would be indeed very informative, but we were not able to perform these analysis in the existing dataset.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Here are some suggestions for targeting some of the weaknesses by additional experiments:

Regional Demarcation in Radial Glial Cell Population:

While the authors demonstrate a decrease in overall RFP-positive cells in response to Abba knockdown, the distinction between different regions should be demarcated using cortical layer-specific markers (e.g., CUX1/BRN2 for the upper layer and CTIP2/FOXP2). Quantification based on regional markers would enhance accuracy and meaningful interpretation.

In order to harmonize the quantification during the different developmental stages we have used a broader definition of the cortical regions that may not be entirely fitting with the regions identified with the staining of Cux1 and CTIP2. We have now however included in the supplementary figure 1 with the staining for Cux1 and CTIP2 showing the corresponding regions defined in the manuscript. Supplementary Figure 1.

Mitotic Stage Marker and BrdU Staining:<br /> The discrepancy between no changes in staining with the mitotic stage marker PH3 and a reported decrease in Ki67 staining calls for further clarification. Additionally, the use of BrdU staining could distinguish the effects on dividing cells after Abba knockdown. The authors are encouraged to explore these aspects further, including their applicability to NEDD9 knockdown and Abba mutant overexpression.

As suggested by the reviewer elsewhere, we made use of life imaging. We monitored the distribution of cells showing accumulation of Anillin-GFP in control (Scramble) and ABBA-shRNA3 conditions (this data has been added to the new Supplementary Figure 3). Anillin has been shown to be an efficient tool for monitoring cell cycle stages in vivo (Hesse et al Nature Com. 2012, DOI: 10.1038/ncomms2089). Interestingly, we observed an increase in cells displaying accumulated Anillin in ABBA-shRNA3 treated cells, which is consistent with an arrest of progression of mitosis.  

Quantification of Cytokinesis Effects:

The brain slices illustrating the effects of Abba knockdown on cytokinesis would benefit from a quantification depicting changes in interkinetic nuclear migration and the number of successful mitosis events. This would enhance the clarity and interpretation of the observed effects.

In the revised manuscript we have included new data in Supplementary Figure 3 were we report the quantification of the distance of the RGC from the ventricle to address the reviewer’s comments. We were not entirely sure about comment about quantification of successful mitosis events, but as specified above, we have included new data from the monitoring of anillin. We hope to perform more detailed experiments and analysis in future studies. 

Loss of Interaction and NEDD9 Localization:

The manuscript lacks an exploration of the loss or decrease in interaction between Abba and NEDD9 in the case of the pathogenic patient-derived mutation in Abba. Addressing this aspect is crucial, as it may shed light on the underlying causes of the observed effects. Furthermore, investigating changes in NEDD9 localization following overexpression of the Abba mutant would provide additional insights.

We fully agree with the reviewer’s comment. Unfortunately the anti NEDD9 antibody had a poor performance in slice immunohistochemistry, which hampered further reliable investigation of expression and distribution changes in vivo. Resolving this issue and providing a more detailed characterization of the mechanism of Abba-NEDD9 interaction will be important in future studies.

Overall, I believe that with minor revisions and additional contextualization, the manuscript has the potential to make a significant contribution to the field. I recommend acceptance pending the incorporation of the suggested revisions.

Reviewer #2 (Recommendations For The Authors):

The manuscript is generally well-organized. We hope that given their nice experimental systems, many of the comments and questions can be addressed with their data already on hand.

Minor Comments

• For Figure 6E A closeup of the vimentin would be helpful - hard to visualize radial glia morphology at the current magnification.

This has been corrected in the new version of the manuscript

• For the in utero electroporation what was their rationale for 2-4 day interval before evaluation? For example, waiting for more cortical plate development to be able to manifest long-term effects.

We observed a massive cell death at E18, in only few of those brains we were able to still observe RFP cells. We have also tried P6 animals but none of them had significant reminding electroporated cells that’s why we have decided to focus at E17, 3 days after the electroporation to have still enough expression of the shRNA.

• Figure 4E-F lacks images of controls for comparison of effect.

This has been corrected in the revised version of the manuscript

Author response:

Reviewer #1:

The manuscript Xu et al. explores the regulation of the microtubule minus end protein CAMSAP2 localization to the Golgi by the Serine/threonine-protein kinase MARK2 (PAR1, PAR1B). The authors utilize immunofluorescence and biochemical approaches to demonstrate that MARK2 is localized at the Golgi apparatus via its spacer domain. They show that depletion of this protein alters Golgi morphology and diminishes CAMSAP2 localization to the Golgi apparatus. The authors combine mass spectroscopy and immunoprecipitation to show that CAMSAP2 is phosphorylated at S835 by MARK2, and that this phosphorylation regulates localization of CAMSAP2 at Golgi membranes. Further, the authors identify USO1 (p115) as the Golgi resident protein mediating CAMSAP2 recruitment to the Golgi apparatus following S835 phosphorylation. The authors would need to address the following queries to support their conclusions.

We sincerely thank the reviewer for their valuable time and effort in evaluating our manuscript. We deeply appreciate the constructive feedback and insightful suggestions, which have been instrumental in improving the quality and clarity of our study. We have carefully considered all the comments and have made the necessary revisions to address the concerns raised.

Major Comments 

(1) Dynamic localization of CAMSAP2 during Golgi reorientation

- The authors use fixed wound edges assays and co-localization analysis to describe changes in CAMSAP2 positioning during Golgi reorientation in response to polarizing cues (a free wound edge in this case). In Figure 1C, they present a graphical representation of quantified immunofluorescence images, using color coding to to describe the three states of Golgi reorientation in response to a wound (green, blue, red indicating non-polarised, partial and complete Golgi reorientation, respectively). They then use these 'colour coded' classifications to quantitate CAMSAP2/GM130 co-localization.It is unclear why the authors have not just used representative immunofluorescence images in the main figures. Transparent, color overlays could be placed over the cells in the representative images to indicate which of the three described states each cell is currently exhibiting. However, for clarity, I would recommend changing the color coded 'states' to a descriptor rather than a color. i.e. Figure 1D x axis labels should be 'complete' and 'partial', instead of 'red' and 'blue'. 

Thank you for this insightful suggestion. We have added representative immunofluorescence images with transparent color overlay to indicate the three Golgi orientation states. These images are included in Supplementary Figure 2B-C, providing a clear visual reference for the quantitative data. Additionally, we have revised the x-axis labels in Figure 1E from "Red" and "Blue" to "Complete" and "Partial" to ensure clarity and consistency with the descriptive terminology in the text. These changes are described in the Results section (page 7, lines 15-19) and the figure legend (page 29, lines 27-29).

We believe these updates improve the clarity and accessibility of our figures and hope they address the reviewer’s concerns.

- note- figure 2 F-G, is semi quantitative, why did the authors not just measure Golgi angle using the nucleus and Golgi distribution?

We appreciate the reviewer’s comment on this point. Following the recommendation, we have performed an additional analysis measuring Golgi orientation angles based on the nucleus-Golgi distribution. This quantitative approach complements our initial semi-quantitative analysis and provides a more precise assessment of Golgi orientation during cell migration.

The new data have been incorporated into Supplementary Figure 1F-H. These results clearly demonstrate the consistency between the quantitative and semi-quantitative methods, further validating our findings and highlighting the dynamic changes in Golgi orientation during cell migration. These changes are described in the Results section (page 6, lines 24-31).

- While it is established that the Golgi is dispersed during reorientation in wound edge migration, the Golgi apparatus also becomes dispersed/less condensed prior to cell division. As the authors have used fixed images - how are they sure that the Golgi morphology or CAMSAP2 localization in 'blue cells' are indicative of Golgi reorientation and not division? Live imaging of cells expressing CAMSAP2, and an additional Golgi marker could be used to demonstrate that the described changes in Golgi morphology and CAMSAP2 localization are occurring during the rear-to-front transition of the Golgi.

Thank you for raising this important question. To address this concern, we carefully examined the nuclear morphology of dispersed Golgi cells and found no evidence of mitotic features, indicating that these cells are not undergoing division (Figure 1A, Supplemental Figure 2A). Furthermore, during the scratch wound assay, we use 2% serum to culture the cells, which helps minimize the impact of cell division. This analysis has been added to the Results section (page7, lines 19-22 in the revised manuscript).

Additionally, we conducted live-cell imaging, as suggested, using cells expressing a Golgi marker. This approach confirmed that Golgi dispersion occurs transiently during reorientation in cell migration. The new live-cell imaging data have been incorporated into Supplementary Figure 2A, and the corresponding description has been updated in the Results section (page 7, lines 2-5).

Finally, considering that overexpression of CAMSAP2 can lead to artifactually condensed Golgi structures, we used endogenous staining to observe CAMSAP2 localization at different stages of migration. These observations provide a clearer understanding of CAMSAP2 dynamics during Golgi reorientation and are now presented in revised Figure 1A-B. This information has been described in the Results section (page 7, lines 5-10).

We hope these additions and clarifications address the reviewer’s concerns. Once again, we are deeply grateful for this constructive feedback, which has greatly improved the robustness of our study.

(2) MARK2 localization to the Golgi apparatus

- The authors investigated the positioning of endogenous MARK2 via immunofluorescence staining, and exogenous flag-tagged MARK2 in a KO background. The description of the protocol required to visualize Golgi localization of MARK2 is inconsistent between the results and methods text. The results text reads as through the 2% serum incubation occurs as a blocking step following fixation. Conversely, the methods section describes the 2% serum incubation as occurring just prior to fixation as a form of serum starvation. The authors need to clarify which of these protocols is correct. Further, whilst I can appreciate that the mechanistic understanding of why serum starvation is required for MARK2 Golgi localization is beyond the scope of the current work, the authors should at a minimum speculate in the discussion as to why they think it might occur.

We sincerely thank the reviewer for the constructive feedback on the localization of MARK2 at the Golgi. Due to the complexity and variability of this phenomenon, we decided to remove the related data from the current manuscript to maintain the rigor of our study. However, we have included a discussion of this phenomenon in the Discussion section (page 13, lines 31-39 and page 14, 1-6in the revised manuscript) and plan to further investigate it in future studies.

The localization of MARK2 at the Golgi was initially observed in experiments following serum starvation, where cells were fixed and stained (The data is not displayed). This observation was supported by the loss of Golgi localization in MARK2 knockdown cells, indicating the specificity of the antibody (The data is not displayed). However, this phenomenon was not consistently observed across all cells, likely due to its transient nature.We speculate that the localization of MARK2 to the Golgi depends on its activity and post-translational modifications. For example, phosphorylation at T595 has been reported to regulate the translocation of MARK2 from the plasma membrane to the cytoplasm (Hurov et al., 2004). Serum starvation might induce modifications or conformational changes in MARK2, leading to its temporary Golgi localization. Additionally, we hypothesize that this localization may coincide with specific Golgi dynamics, such as the transition from dispersed to ribbon-like structures during cell migration.

We also acknowledge the inconsistency in the Results and Methods sections regarding serum starvation. We confirm that serum starvation was performed prior to fixation as an experimental condition, rather than as a blocking step in immunostaining. This clarification has been incorporated into the revised Methods section (page 24, lines 11-12).

We hope this clarification, along with our planned future studies, adequately addresses the reviewer’s concerns. Once again, we deeply appreciate the reviewer’s valuable comments, which have provided important insights for our ongoing work. References：

Hurov, J.B., Watkins, J.L., and Piwnica-Worms, H. (2004). Atypical PKC phosphorylates PAR-1 kinases to regulate localization and activity. Curr Biol 14 (8): 736-741.

- The authors should strengthen their findings by using validated tools/methods consistent with previous publications. i.e. Waterman lab has published two MARK2 constructs- Apple and eGFP tagged versions (doi.org/10.1016/j.cub.2022.04.088), and the localization of MARK2 in U2Os cells (using the same antibody (Anti- MARK2 C-terminal, ABCAM Cat# ab136872). The authors should (1) image the cells live using eGFP-tagged MARK2 during serum starvation to show the dynamics of this localization, (2) image U2Os cells using the abcam ab136872 antibody +/- 2% serum starve. Two MARK2 antibodies are listed in Table 2. Does abcam (ab133724) show a similar localisation?

- The Golgi localization of MARK2 occurs in the absence of the T structural domain, but not when full length MARK2 is expressed. The authors conclude the T- domain is likely inhibitory. When combined with the requirement for serum starvation for this interaction to occur, the authors should clarify the physiological relevance of these observations.

We sincerely thank the reviewer for their valuable suggestions regarding the use of tools and methods and the physiological relevance of MARK2 localization to the Golgi. Regarding the question of how MARK2 itself localizes to the Golgi, we are currently unable to fully elucidate the underlying mechanism. Therefore, we have removed the discussion of MARK2’s Golgi localization from the manuscript to ensure scientific accuracy. However, Below, we provide our detailed response as soon as possible:

First, regarding the suggestion to use tools and methods developed by the Waterman lab to strengthen our findings, we have carefully evaluated their applicability. In our live-cell imaging experiments, we found that full-length MARK2 does not stably localize to the Golgi, even under serum starvation conditions. However, truncated MARK2 mutants lacking the Tail (T) domain exhibit robust Golgi localization. Furthermore, our immunofluorescence staining results indicate that the Spacer domain is the minimal region required for MARK2 localization at the Golgi. Based on these findings, we believe that live-cell imaging of EGFP-tagged full-length MARK2 may not effectively reveal the dynamics of its Golgi localization. However, we plan to focus on the truncated constructs in future studies to better explore the mechanisms underlying MARK2's dynamic behavior. 

Regarding the use of the ab136872 antibody to stain U2OS cells with and without serum starvation, we note that the protocol described by the Waterman lab involves pre-fixation and permeabilization steps, which are not compatible with live-cell imaging. Additionally, we observed that MARK2 Golgi localization appears to be condition-dependent and may coincide with specific Golgi dynamics, such as transitions from dispersed stacks to intact ribbon structures. These events are likely brief and challenging to capture consistently. Nevertheless, we recognize the value of this experimental design and plan to adapt the staining conditions in future work to validate our results further. As for the ab133724 antibody listed in Table 2, we clarify that it has only been validated for Western blotting in our study and does not yield reliable results in immunofluorescence experiments. For this reason, all immunofluorescence staining in this study relied exclusively on ab136872. This distinction has been clarified in the revised Table 2 .

Regarding the hypothesis that the Tail domain of MARK2 is inhibitory, our observations showed that truncated MARK2 mutants lacking the T domain stably localized to the Golgi, whereas fulllength MARK2 did not. Literature evidence supports this hypothesis, as studies on the yeast homolog Kin2 indicate that the C-terminal region (including the Tail domain) binds to the Nterminal catalytic domain to inhibit kinase activity (Elbert et al., 2005). We speculate that serum starvation disrupts this intramolecular interaction, relieving the inhibition by the T domain, activating MARK2, and promoting its localization to the Golgi. Moreover, we hypothesize that the transient nature of MARK2 localization to the Golgi may be related to specific Golgi remodeling processes, such as the transition from dispersed stacks to intact ribbon structures during cell migration or polarity establishment. 

References：

Elbert, M., Rossi, G., and Brennwald, P. (2005). The yeast par-1 homologs kin1 and kin2 show genetic and physical interactions with components of the exocytic machinery. Mol Biol Cell 16 (2): 532-549.

(3) Phosphorylation of CAMSAP2 by MARK2

- The authors examined the effects of MARK2 phosphorylation of CAMSAP2 on Golgi architecture through expression of WT-CAMSAP2 and two CAMSAP2 S835 mutants in CAMSAP2 KO cells. They find that CAMSAP2 S835A (non-phosphorylatable) was less capable of rescuing Golgi morphology than CAMSAP2 S835D (phosphomimetic). Golgi area has been measured to demonstrate this phenomenon. Representative immunofluorescence images in Fig. 4D appear to indicate that this is the case. However, quantification in Fig. 4E does not show significance between HA-CAMSAP2 and HA-CAMSAP2A that would support the initial claim. The authors could analyze other aspects of Golgi morphology (e.g. number of Golgi fragments, degree of dispersal around the nucleus) to capture the clear structural defects demonstrated in HACAMSAP2A cells.

We sincerely thank the reviewer for their valuable feedback and for pointing out potential areas of improvement in our analysis of Golgi morphology. We apologize for any misunderstanding caused by our description of the results in Figure 4E.

The quantification indeed shows a significant difference between HA-CAMSAP2 and HACAMSAP2A in terms of Golgi area, as indicated in the figure by the statistical annotations (pvalue provided in the legend). To ensure clarity, we have revised the figure legend (page 32, lines 19-23 in the revised manuscript) to explicitly describe the statistical significance, and the method used for quantification.

Because the quantification indeed shows a significant difference between HA-CAMSAP2 and HA-CAMSAP2A in terms of Golgi area, and to maintain consistency throughout the manuscript, we did not further analyze other aspects of Golgi morphology.

We hope this clarification, along with the additional analyses, will address the reviewer’s concerns. Once again, we are deeply grateful for these constructive comments, which have helped us improve the quality and robustness of our study.

- Wound edge assays are used to capture the difference in Golgi reorientation towards the leading edge between CAMSAP2 S835A and CAMSAP2 S835D. However, these studies lack comparison to WT-CAMSAP2 that would support the role of phosphorylated CAMSAP2 in reorienting the Golgi in this context.

We sincerely thank the reviewer for their insightful suggestion. In response, we have added a comparison between CAMSAP2 S835A/D and WT-CAMSAP2, in addition to HT1080 and MARK2 KO cells, to better evaluate the role of phosphorylated CAMSAP2 in Golgi reorientation.

The results, now shown in Figure 5A-C, indicate that in the absence of MARK2, there is no significant difference in Golgi reorientation between WT-CAMSAP2 and CAMSAP2 S835A. This observation supports the conclusion that MARK2-mediated phosphorylation of CAMSAP2 at S835 is essential for effective Golgi reorientation.

To enhance clarity, we have updated the corresponding Results section (page 9, lines 37-40 and page 10, line 1 in the revised manuscript) to describe this additional comparison. We believe this analysis strengthens our findings and provides a clearer understanding of the role of phosphorylated CAMSAP2 in Golgi dynamics.

We hope this additional data addresses the reviewer’s concerns. Once again, we are grateful for the constructive feedback, which has helped improve the clarity and robustness of our study.

(4) Identification of CAMSAP2 interaction partners

- Quantification of interaction ability between CAMSAP2 and CG-NAP, CLASP2, or USO1 in Fig. 5D, 5F and 5J respectively, lack WT-CAMSAP2 comparisons.

We sincerely thank the reviewer for their valuable suggestion. In response, we have included WT-CAMSAP2 data in the quantification of interaction ability between CAMSAP2 and CG-NAP, CLASP2, and USO1. These results, now shown in revised Figures 5 D-G and Figures 6 C-D, provide a direct comparison that further validates the differential interaction abilities of CAMSAP2 mutants.

The inclusion of WT-CAMSAP2 allows us to better contextualize the effects of specific mutations on CAMSAP2 interactions and strengthens our conclusions regarding the role of these interactions in Golgi dynamics.

We hope this addition addresses the reviewer’s concerns and enhances the clarity and robustness of our study. We deeply appreciate the constructive feedback, which has been instrumental in improving our manuscript.

- The CG-NAP immunoblot presented in Fig. 5C shows that the protein is 310 kDa, which is the incorrect molecular weight. CG-NAP (AKAP450) should appear at around 450 kDa. Further, no CG-NAP antibody is included in Table 2 - Information of Antibodies. The authors need to explain this discrepancy.

We sincerely apologize for the lack of clarity in our annotation and description, which may have caused confusion regarding the CG-NAP immunoblot presented in Figure 5C (Figure 5D in the revised manuscript). To clarify, CG-NAP (AKAP450) is indeed a 450 kDa protein, and the marker at 310 kDa represents the molecular weight marker’s upper limit, above which CG-NAP is observed. This has been clarified in the figure legend (page 33, lines 21-23 in the revised manuscript).

Regarding the CG-NAP antibody, it was custom-made and purified in our laboratory. Polyclonal antisera against CG-NAP, designated as αEE, were generated by immunizing rabbits with GSTfused fragments of CG-NAP (aa 423–542). This antibody has been validated extensively in our previous research, demonstrating its specificity and reliability (Wang et al., 2017). The details of the antibody preparation are included in the footnote of Table 2 for reference.

We hope this clarification, along with the additional context regarding the antibody validation, resolves the reviewer’s concerns. We are deeply grateful for the reviewer’s attention to detail, which has helped us improve the clarity and rigor of our manuscript.

References：

Wang, J., Xu, H., Jiang, Y., Takahashi, M., Takeichi, M., and Meng, W. (2017). CAMSAP3dependent microtubule dynamics regulates Golgi assembly in epithelial cells. Journal of genetics and genomics = Yi chuan xue bao 44 (1): 39-49.

Minor Comments

- Authors should change immunofluorescence images to colorblind friendly colors. The current presentation of merged overlays makes it really difficult to interpret- I would strongly encourage inverted or at a minimum greyscale individual images of key proteins of interest.

We sincerely thank the reviewer for their valuable suggestion regarding the presentation of immunofluorescence images. In response, we have converted the images in Figure 1C to greyscale individual images for each key protein of interest. This adjustment ensures that the figures are more accessible and interpretable, including for readers with color vision deficiencies.

We hope this modification addresses the reviewer’s concern and improves the clarity of our data presentation. We are grateful for the constructive feedback, which has helped us enhance the overall quality of our figures.

- On p. 8 text should be amended to 'Previous literature has documented MARK2's localization to the microtubules, microtubule-organizing center (MTOC), focal adhesions..'

We sincerely thank the reviewer for their comment regarding the text on page 8. Considering the reasoning provided in response to question 2, where we clarified that MARK2's Golgi localization is not fully understood, we have decided to remove this section from the manuscript to maintain the accuracy and rigor of our study.

We appreciate the reviewer’s attention to detail and constructive feedback, which has helped us improve the clarity and focus of our manuscript. 

- In Fig.1A scale bars are not shown on individual channel images of CAMSAP or GM130

We sincerely thank the reviewer for pointing out the omission of scale bars in the individual channel images of CAMSAP and GM130 in Figure 1A (Figure 1C in the revised manuscript). In response, we have added a scale bar (5 μm) to the CAMSAP2 channel, as shown in the revised Figure 1C. These updates have been described in the figure legend (page 29, line 21).

We hope this modification addresses the reviewer’s concern and improves the accuracy and clarity of our figure presentation. We greatly appreciate the reviewer’s constructive feedback, which has helped enhance the quality of our manuscript.

- In Fig. 1B the title should be amended to 'Colocalization of CAMSAP2/GM130'

We sincerely thank the reviewer for their suggestion to amend the title in Figure 1B (Figure 1D in the revised manuscript). In response, we have updated the title to "Colocalization of CAMSAP2/GM130," as shown in the revised Figure 1D.

We hope this modification addresses the reviewer’s concern and improves the clarity and accuracy of the figure. We greatly appreciate the reviewer’s valuable feedback, which has helped us refine the presentation of our results.

- In Fig. 2F, 5A, and Sup Fig 3C scale bars have been presented vertically

We sincerely thank the reviewer for pointing out the issue with the vertical orientation of scale bars in Figures 2F (Figure 2D in the revised manuscript), 5A, and Supplementary Figure 3C. In response, we have modified the scale bars in revised Figures 2D and 5A to a horizontal orientation for improved consistency and clarity. Additionally, Supplementary Figure 3C has been removed from the revised manuscript.

We hope these adjustments address the reviewer’s concerns and enhance the overall presentation quality of the figures. We greatly appreciate the reviewer’s constructive feedback, which has helped us refine our manuscript.

- Panels are not correctly aligned, and images are not evenly spaced or sized in multiple figures - Fig. 2F, 4D, Sup Fig. 1F, Sup Fig. 2C, Sup Fig. 3E, Sup Fig. 4C

We sincerely thank the reviewer for pointing out the misalignment and uneven spacing or sizing of panels in multiple figures, including Figures 2F, 4D, Supplementary Figures 1F, 2C, 3E, and 4C (Figure 2D, 4D, Supplementary Figures 1F, 2C, and 3H in the revised manuscript.

Supplementary Figure 3E was removed from our manuscript). In response, we have standardized the spacing and sizing of all panels throughout the manuscript to ensure consistency and improve visual clarity.

We hope this modification addresses the reviewer’s concerns and enhances the overall presentation quality of our figures. We greatly appreciate the reviewer’s constructive feedback, which has helped us improve the organization and professionalism of our manuscript.

- An uncolored additional data point is present in Fig. 3F

We sincerely thank the reviewer for pointing out the presence of an uncolored additional data point in Figure 3F. In response, we have removed this data point from the revised figure to ensure accuracy and clarity.

We hope this adjustment resolves the reviewer’s concern and improves the overall quality of the figure. We greatly appreciate the reviewer’s careful review and constructive feedback, which have helped us refine our manuscript.

- In Fig. 3A 'GAMSAP2/GM130' in the vertical axis label should be amended to 'CAMSAP2/GM130'

We sincerely thank the reviewer for pointing out the error in the vertical axis label of Figure 3A. In response, we have corrected "GAMSAP2/GM130" to "CAMSAP2/GM130," as shown in the revised Figure 3I.

We hope this correction resolves the reviewer’s concern and improves the accuracy of our figure. We greatly appreciate the reviewer’s careful review and constructive feedback, which have helped us refine our manuscript.

- In Fig 5A the green label should be amended to 'GFP-CAMSAP2' instead of 'GFP'

We sincerely apologize for the confusion caused by our labeling in Figure 5A. To clarify, the green label “GFP” refers to the antibody used, while “GFP-CAMSAP2” is indicated at the top of the figure to specify the construct being analyzed.

We hope this explanation resolves the misunderstanding and provides clarity regarding the labeling in Figure 5A. We greatly appreciate the reviewer’s feedback, which has allowed us to address this issue and improve the precision of our figure annotations.

- The repeated use of contractions throughout the manuscript was distracting, I would strongly encourage removing these.

We sincerely thank the reviewer for pointing out the distracting use of contractions in the manuscript. In response, we have removed and replaced all contractions with their full forms to improve the clarity and formal tone of the text.

We hope this modification addresses the reviewer’s concern and enhances the readability and professionalism of our manuscript. We greatly appreciate the reviewer’s constructive feedback, which has helped us refine the quality of our writing.

Reviewer #2: 

Summary  

This work by the Meng lab investigates the role of the proteins MARK2 and CAMSAP2 in the Golgi reorientation during cell polarisation and migration. They identified that both proteins interact together and that MARK2 phosphorylates CAMSAP2 on the residue S835. They show that the phosphorylation affects the localisation of CAMSAP2 at the Golgi apparatus and in turn influences the Golgi structure itself. Using the TurboID experimental approach, the author identified the USO1 protein as a protein that binds differentially to CAMSAP2 when it is itself phosphorylated at residue 835. Dissecting the molecular mechanisms controlling Golgi polarisation during cell migration is a highly complex but fundamental issue in cell biology and the author may have identified one important key step in this process. However, although the authors have made a genuine iconographic effort to help the reader understand their point of view, the data presented in this study appear sometimes fragile, lacking rigour in the analysis or over-interpreted. Additional analyses need to be conducted to strengthen this study and elevate it to the level it deserves.

We sincerely thank the reviewer for their thoughtful evaluation and recognition of our study's significance in understanding Golgi reorientation during cell migration. We appreciate the constructive feedback regarding data robustness, clarity, and interpretation. In response, we have conducted additional analyses, revised data presentation, and ensured cautious interpretation throughout the manuscript. These changes aim to address the reviewer’s concerns comprehensively and strengthen the scientific rigor of our study.

Major comments

In order to conclude as they do about the putative role of USO1, the authors need to perform a siRNA/CRISPR of USO1 to validate its role in anchoring CAMSAP2 to the Golgi apparatus in a MARK2 phosphorylation-dependent manner. In other words, does depletion of USO1 affect the recruitment of CAMSAP2 to the Golgi apparatus?

We sincerely thank the reviewer for their insightful suggestion regarding the role of USO1 in anchoring CAMSAP2 to the Golgi apparatus. In response, we performed USO1 knockdown using siRNA and quantified the Pearson correlation coefficient of CAMSAP2 and GM130 colocalization in control and USO1-knockdown cells.

The results show that CAMSAP2 localization to the Golgi is significantly reduced in USO1knockdown cells, confirming that USO1 plays a critical role in recruiting CAMSAP2 to the Golgi apparatus. These results are now presented in Figures 6 E–G, and corresponding updates have been incorporated into the Results section (page 10, lines 36-37 in the revised manuscript).

We hope this additional experiment addresses the reviewer’s concern and strengthens our conclusions regarding the role of USO1. We are grateful for the reviewer’s constructive feedback, which has greatly improved the robustness of our study.  

It is not clear from this study exactly when and where MARK2 phosphorylates CAMSAP2. What is the result of overexpression of the two proteins in their respective localisation to the Golgi apparatus? As binding between CAMSAP2 and MARK2 appears robust in the immunoprecipitation assay, this should be readily investigated. 

We sincerely thank the reviewer for their insightful comments and questions. To address the role of MARK2 in regulating CAMSAP2 localization to the Golgi apparatus, we overexpressed GFPMARK2 in cells and compared its effects on CAMSAP2 localization to the Golgi with control cells overexpressing GFP alone. Our results show that CAMSAP2 localization to the Golgi is significantly increased in GFP-MARK2-overexpressing cells, as shown in Supplementary Figures 3C and 3E. Corresponding updates have been incorporated into the Results section (page 8, lines 25-27 in the revised manuscript).

Regarding the question of how MARK2 itself localizes to the Golgi, we are currently unable to fully elucidate the underlying mechanism. Therefore, we have removed the discussion of MARK2’s Golgi localization from the manuscript to ensure scientific accuracy. Consequently, we have not conducted experiments to assess the effects of CAMSAP2 overexpression on MARK2’s localization to the Golgi.

We hope this explanation clarifies the reviewer’s concerns. We are grateful for the reviewer’s constructive feedback, which has guided us in improving the clarity and focus of our study.

To strengthen their results, can the author map the interaction domains between CAMSAP2 and MARK2? The authors have at their disposal all the constructs necessary for this dissection.

We sincerely thank the reviewer for their insightful suggestion to map the interaction domains between CAMSAP2 and MARK2. In response, we performed immunoprecipitation experiments using truncated constructs of CAMSAP2. Our results reveal that MARK2 interacts specifically with the C-terminus (1149F) of CAMSAP2, as shown in Supplementary Figures 3A and 3B. Corresponding updates have been incorporated into the Results section (page 7, lines 41-42 and page 8, line 1 in the revised manuscript).

We hope this additional analysis addresses the reviewer’s suggestion and further strengthens our conclusions. We greatly appreciate the reviewer’s constructive feedback, which has helped improve the depth of our study.

Minor comments

Sup-fig1  

H: It is not clear if the polarisation experiment has been repeated three times (as it should) and pooled or is just the result of one experiment?

We sincerely apologize for the lack of clarity regarding the experimental details for Supplementary Figure 1H. To clarify, the polarization experiment was repeated three times, and the results were pooled to generate the data presented. We have updated the figure legend for Supplementary Figure 1H to explicitly state this information (page 35, lines 27-29 in the revised manuscript).

We hope this clarification resolves the reviewer’s concern. We greatly appreciate the reviewer’s careful review and constructive feedback, which have helped us improve the accuracy and transparency of our manuscript.

Sup-fig2  

C: "Immunofluorescence staining plots" formula used in the legend is not clear. Which condition is presented in the panel, parental HT1080 or CAMSAP2 KO cells?  

We thank the reviewer for pointing out the lack of clarity regarding the conditions presented in Supplementary Figure 2C. To clarify, the immunofluorescence staining plots shown in this panel are from parental HT1080 cells. We have updated the figure legend to include this information (page 36, line 14 in the revised manuscript).

We hope this clarification resolves the reviewer’s concern and improves the transparency of our data presentation. We greatly appreciate the reviewer’s feedback, which has helped us refine the manuscript.

Figure 1  

D: In the plot, the colour of the points for the "red cells" are red but the one for the "blue cells" are green, this is confusing.

E: Once again, the colour choice is confusing as blue cells (t=0.5h) are quantified using red dots and red cells (t=2h) quantified using green dots. The t=0h condition should be quantified as well and added to the graph.  

F: Representative CAMSAP2 immunofluorescence pictures for the three time points should be provided in addition to the drawings.  

We thank the reviewer for their valuable comments regarding Figure 1D (revised Figure 1E), Figure 1E (revised Figure 1B), and Figure 1F (revised Supplementary Figure 2C).

- Figure 1D (revised Figure 1E): we have modified the x-axis labels and adjusted the color scheme of the data points to ensure consistency and avoid confusion.

- Figure 1E (revised Figure 1B): we have updated the x-axis and included the quantification of the t=0h condition, which has been added to the graph.

- Figure 1F (revised Supplementary Figure 2C): we have provided representative immunofluorescence images of CAMSAP2 for the three-time points to complement the schematic drawings.

We hope these revisions address the reviewer’s concerns and improve the clarity and completeness of our data presentation. We greatly appreciate the reviewer’s constructive feedback, which has significantly contributed to enhancing our manuscript.

Figure 2  

A: No methodology in the material and methods is provided for this analysis.  

B: Can the authors be more precise regarding the source of the CAMSAP2 interactants? Can the author provide the citation of the publication describing the CAMSAP2-MARK2 interaction?  

D: Genotyping for the MARK2 KO cell line should be provided the same way it was provided for the CAMSAP2 cell line in Sup-fig1. "MARK2 was enriched around the Golgi apparatus in a  significant proportion of HT1080 cells": which proportion of the cells?  

F: The time point of fixation is missing  

G: It is not clear if the polarisation experiment has been repeated three times (as it should) and pooled or is just the result of one experiment?  

We thank the reviewer for their detailed comments and suggestions regarding Figure 2. Below, we provide clarifications and outline the modifications made:

- Figure 2A: The methodology for this analysis has been added to section 5.14 (Data statistics). Specifically, we have stated: “GO analysis of proteins was plotted using https://www.bioinformatics.com.cn, an online platform for data analysis and visualization” (page 26 lines 5-6 in the revised manuscript).

- Figure 2B: The CAMSAP2 interactants were derived from the study by Wu et al., 2016, which provides the source of these interactants. The interaction between CAMSAP2 and MARK2 is referenced from Zhou et al., 2020. These citations have been added to the relevant sections of the manuscript (page 30, lines 10-11 and 13-14).

- Figure 2D (removed in the revised manuscript): Genotyping for the MARK2 KO cell line has been provided in the same format as for the CAMSAP2 KO cell line in Figure 2G. Additionally, as the MARK2 Golgi localization discussion cannot yet be fully elucidated, we have removed this portion from the manuscript.

- Figure 2F (revised Figure 2D): The time point of fixation, which occurred 2 hours after the scratch wound assay, has been added to the figure legend (page 30, lines 15-16).

- Figure 2G (revised Figure 2E-F): The polarization experiment was repeated three times, and the results were pooled. This information has been included in the figure legend (page 30, lines 26 and 29).

We hope these updates address the reviewer’s concerns and improve the clarity and completeness of the manuscript. We are grateful for the reviewer’s constructive feedback, which has greatly enhanced the rigor of our study. References：

Wu, J., de Heus, C., Liu, Q., Bouchet, B.P., Noordstra, I., Jiang, K., Hua, S., Martin, M., Yang, C., Grigoriev, I., et al. (2016). Molecular Pathway of Microtubule Organization at the Golgi Apparatus. Dev Cell 39 (1): 44-60.

Sup-fig3  

E: Although colocalisation between CAMSAP2 and MARK2 is clear in your serum conditions in HT1080 and RPE1 cells, the deletion domain analysis appears weak and insufficient to implicate the role of the spacer domain. This part should be deleted or strengthened, but the data do not satisfactorily support your conclusion as it stands.  

We sincerely thank the reviewer for their critical comments regarding the deletion domain analysis of MARK2 and its role in colocalization with CAMSAP2. As the current data do not satisfactorily support our conclusions, we have removed all related content on MARK2 and the deletion domain analysis from the manuscript to maintain scientific rigor.

We appreciate the reviewer’s valuable feedback, which has helped us refine and improve the quality and focus of our study.

Figure 3  

A: Can the reduced CAMSAP2 Golgi localisation phenotype be rescued by the overexpression of MARK2 cDNA in the MARK2 KO cells?  

F: Presence of a white dot on the HT1080 plot  

G: The composition of the homogenization buffer is not indicated in the material and methods  

We thank the reviewer for their valuable comments and suggestions regarding Figure 3. Below, we detail the modifications made:

- Figure 3A: To address whether the reduced CAMSAP2 Golgi localization phenotype can be rescued, we overexpressed MARK2 cDNA in MARK2 KO cells. Our results show that overexpression of MARK2 successfully rescues the reduced CAMSAP2 localization to the Golgi, as demonstrated in Supplementary Figures 3C and 3E (page 8, lines 5-7).

- Figure 3F: We have removed the white dot on the HT1080 plot to ensure clarity and accuracy.

- Figure 3G: The composition of the homogenization buffer used in the experiment has been added to the Materials and Methods section for completeness (page 24, lines 34-41 and page 25, lines 1-10).

We hope these revisions address the reviewer’s concerns and enhance the clarity and rigor of our study. We are grateful for the reviewer’s constructive feedback, which has significantly improved the quality of our manuscript.

Figure 4  

B: Quantification of the effect of the S835A mutation should be provided  

D: Top left panel: Why Ha antibody stains Golgi structure in absence of Ha-CAMSAP2 transfection ? IF the Ha antibody has unspecific affinity towards the Golgi apparatus, may be it is not the good tag to use in this assay?  

E: The number of cells studied should be standardized. 119 cells were analyzed in the CAMSAP KO vs only 35 cells in the CAMSAP2 KO (HA-CAMSAP2-S835D) conditions. This could introduce strong bias to the analysis. Furthermore the CAMSAP2 S835A seems to provide a certain level of rescue. It would be interesting to see what is the result of the T test between the HT1080 and HA-CAMSAP S835A conditions.  

We thank the reviewer for their thoughtful comments and suggestions regarding Figure 4. Below, we detail the revisions and clarifications made:

- Figure 4B: The S835A mutation renders CAMSAP2 non-phosphorylatable by MARK2. This conclusion is based on our experimental observations and previously reported mechanisms.

- Figure 4D: The HA antibody does not exhibit non-specific affinity toward the Golgi apparatus. The observed labeling in the top left panel was due to an error in our annotation. We have corrected the label, replacing "HA" with "CAMSAP2" to accurately reflect the experimental conditions.

- Figure 4E: To standardize the number of cells analyzed across conditions, we reduced the number of CAMSAP2 KO cells analyzed to 50 and balanced the sample sizes for comparison. Additionally, we performed a t-test between the HT1080 and HACAMSAP2 S835A conditions. The results support that CAMSAP2 S835A provides partial rescue, as reflected in the updated analysis (page 32, lines 19-23).

We hope these revisions address the reviewer’s concerns and improve the accuracy and reliability of our results. We greatly appreciate the reviewer’s constructive feedback, which has significantly enhanced the quality of our study.

Figure 6  

6A: The wound position should be indicated on the picture.  

6B: Given that microtubule labelling is present on the vast majority of the cell surface, this type of quantification provides very little information using conventional light microscopy and should not be used to conclude any change in the microtubule network using Pearson's coefficient.  The text describing the figure 6A and 6B needs re written as I do not understand what the author want to say. "In cells located before the wound edge..." : I do not understand how a cell could be located before the wound edge. Which figure corresponds to the trailing edge of the wounding?

We thank the reviewer for their valuable comments on Figure 6A (revised Supplementary Figure 6E) and Figure 6B (revised Supplementary Figure 6F). Below, we detail the modifications made:

- Figure 6A (revised Supplementary Figure 6E), we have added arrows to indicate the wound position, providing clearer guidance for interpreting the image.

- Figure 6B (revised Supplementary Figure 6F), we revised our quantification method based on the approach used in literature (Wu et al., 2016). Specifically, we analyzed the relationship between microtubules and the Golgi apparatus in cells at the leading edge of the wound. The x-axis represents the distance from the Golgi center, while the y-axis shows the normalized radial fluorescence intensity of microtubules and the Golgi apparatus.

Additionally, we revised the accompanying text for clarity and accuracy. The original description:

“In cells located before the wound edge, the Golgi apparatus maintained a ribbon-like shape, with a higher density of microtubules. In contrast, at the trailing edge of the wounding, the Golgi apparatus appeared more as stacks around the nucleus, with fewer microtubules”  was replaced with:

“Finally, to comprehensively understand the dynamics between non-centrosomal microtubules and the Golgi apparatus during Golgi reorientation, we conducted cell wound-healing experiments (Supplementary Figure 6 E-F). Our observations revealed notable changes in the Golgi apparatus and microtubule network distribution in relation to the wounding. These findings corroborate our earlier results and suggest a highly dynamic interaction between the Golgi apparatus and microtubules during Golgi reorientation” (Revised manuscript page 11 lines 3-10).

We hope these changes address the reviewer’s concerns and improve the clarity and robustness of our study. We greatly appreciate the reviewer’s constructive feedback, which has significantly enhanced the presentation and interpretation of our data. References：

Wu, J., de Heus, C., Liu, Q., Bouchet, B.P., Noordstra, I., Jiang, K., Hua, S., Martin, M., Yang, C., Grigoriev, I., et al. (2016). Molecular Pathway of Microtubule Organization at the Golgi Apparatus. Dev Cell 39 (1): 44-60.

Reviewer #3:  

Summary  

In this study, Xu et al. analyzed the wound healing process of HT1080 cells to elucidate the molecular mechanisms by which the Golgi apparatus exhibits transient dispersion before reorienting to the wound edge in the compact assembly structure. They focused on the role of the microtubule minus-end binding protein CAMSAP2, which mediates the linkage between microtubules and the Golgi membrane. At first, they noticed that CAMSAP2 transiently lost Golgi colocalization during the initial phase of the wound healing process. They further found that the cell polarity-regulating kinase MARK2 binds and phosphorylates S835 of CAMSAP2, thereby enhancing the interaction between CAMSAP2 and the Golgi protein Uso1. Together with the phenotypes of CAMSAP2, MARK2, and Uso1 KO cells, these authors argue that the MARK2dependent phosphorylation of CAMSAP2 plays an important role in the reassembly and reorientation of the Golgi apparatus after a transient dispersion observed during the wound healing process.

We sincerely thank the reviewer for their thoughtful summary of our study and constructive feedback. Your comments have been invaluable in refining our research and enhancing the clarity and impact of our manuscript.

Major comments

(1) The premise of this study was that during the wound healing process, the Golgi apparatus exhibits transient dispersion before reorientation to the front of the nucleus.  

In the first place, this claim has not been well established in previous studies or this paper. Therefore, the authors should present a proof of this claim in a clearer manner.  

To introduce this cellular event, the authors cite several papers in the introduction (page 4) and the results (page 6) sections. However, many papers cited are review articles, and some of them do not describe this change in the Golgi assembly structure before reorientation. Only two original articles discussed this phenomenon (Bisel et al. 2008 and Wu et al. 2016), and direct evidence was provided by only one paper (Wu et al. 2016) in which changes in the Golgi apparatus in wound-healing RPE1 cells were recorded by live imaging (Fig.7A in Wu et al. 2016).

Furthermore, it should be noted that this previous paper demonstrated that depletion of CAMSAP2 inhibits Golgi dispersion. Obviously, this conclusion is inconsistent with their statement to introduce this study (page4) that ‟This emphasizes CAMSAP2's role in sustaining Golgi integrity during critical cellular events like migration." In addition, it also contradicts the authors' model of the present paper (Fig. 6E), which argued that disruption of the Golgi association of CAMSAP2 facilitates the Golgi dispersion.  

We sincerely thank the reviewer for their detailed comments and for providing us with the opportunity to clarify the premise and conclusions of our study. Below, we address the main concerns raised:

First, to provide direct evidence of Golgi apparatus changes during the wound-healing process, we conducted live-cell imaging experiments. Our observations, presented in revised Supplementary Figure 2A, clearly demonstrate that the Golgi apparatus exhibits a transient dispersion state before reorienting toward the leading edge of the nucleus during migration.

Regarding the interpretation of previous studies, we acknowledge the reviewer’s concerns about the citation of review articles. To address this, we have revisited the literature and clarified that the phenomenon of Golgi dispersion during reorientation has been directly demonstrated in Wu et al (Wu et al., 2016), where live imaging of wound-healing RPE1 cells showed this dynamic behavior. Furthermore, we note that in Wu et al paper explicitly demonstrates that CAMSAP2 depletion promotes Golgi dispersion, contrary to the reviewer’s interpretation that "depletion of CAMSAP2 inhibits Golgi dispersion."

Our model focuses on the role of CAMSAP2 in restoring the Golgi from a transiently dispersed structure back to an intact ribbon-like structure during reorientation. Specifically, we propose that during this process, the disruption of CAMSAP2’s association with the Golgi affects this restoration, rather than directly promoting Golgi dispersion as suggested by the reviewer. We believe this distinction aligns with our data and the existing literature.

To strengthen the background of our study, we have revised the introduction and results sections (page 6, lines 6-13 and page 7, lines 1-17) to minimize reliance on review articles and have provided more explicit citations to original research papers. We hope this addresses the reviewer’s concern about the sufficiency of the cited literature.

We trust these clarifications and revisions resolve the reviewer’s concerns and enhance the robustness of our study. Once again, we are grateful for the reviewer’s constructive feedback, which has greatly helped refine our manuscript. References：

Wu, J., de Heus, C., Liu, Q., Bouchet, B.P., Noordstra, I., Jiang, K., Hua, S., Martin, M., Yang, C., Grigoriev, I., et al. (2016). Molecular Pathway of Microtubule Organization at the Golgi Apparatus. Dev Cell 39 (1): 44-60.

The authors did not provide experimental data for this temporal change in the Golgi assembly structures during the wound-healing process of HT1080 that they analyzed. They only provide an illustration of wound-healing cells (Fig.1F), in which cells are qualitatively discriminated and colored based on the Golgi states, without indicating the experimental basis of the discrimination.

According to their ambiguous descriptions in the text (page7), the reader can speculate that Fig. 1F is illustrated based on the images in Supplementary Fig. 2C. However, because of the low quality and presentation style of these data, it is impossible to recognize the assembly structures of the Golgi apparatus in wound-edge cells.  

If the authors hope to establish this premise claim for the present paper, they should provide their own data corresponding to the present Supplementary Fig. 2C in more clarity and present qualitative data verifying this claim, as Wu et al. did in Fig. 7A in their paper.

We sincerely thank the reviewer for their constructive feedback and the opportunity to address the concern regarding the lack of experimental data supporting the temporal changes in Golgi assembly during the wound-healing process.

To establish this premise, we conducted live-cell imaging experiments to observe the dynamic changes in the Golgi apparatus during directed cell migration. Our data, now presented in Supplementary Figure 2A, clearly demonstrate that the Golgi apparatus undergoes a transient dispersed state before reorganizing into an intact structure. These findings provide direct experimental evidence supporting our claim.

In addition, we have revised the data originally presented in Supplementary Figure 2C and enhanced its quality and presentation style. This supplementary figure now includes clearer images and annotations to better illustrate the Golgi assembly structures in wound-edge cells. The improved data presentation aligns with the standards set by Wu et al reported (Wu et al., 2016) and provides qualitative support for our observations.

We hope these additions and revisions address the reviewer’s concerns and strengthen the scientific rigor and clarity of our manuscript. We are grateful for the reviewer’s valuable suggestions, which have significantly improved the quality of our study. References：

Wu, J., de Heus, C., Liu, Q., Bouchet, B.P., Noordstra, I., Jiang, K., Hua, S., Martin, M., Yang, C., Grigoriev, I., et al. (2016). Molecular Pathway of Microtubule Organization at the Golgi Apparatus. Dev Cell 39 (1): 44-60.

(2) In Fig.1A-D, the authors claim that CAMSAP2 dissociates from the Golgi apparatus in cells "that have not yet completed Golgi reorientation and exhibit a transitional Golgi structure, characterized by relative dispersion and loss of polarity (page7)." However, I these analyses, they do not analyze the initial stage (0.5h after wound addition) of cells facing the wound edge, as they do in Supplementary Fig. 2C. Instead, they analyze cells separated from the wound edge at 2 h after wound addition when the wound-edge cells complete their polarization. These data are highly misleading because there is no evidence that the cells separated from the wound edge are really in the transitional state before polarization.  

In this regard, Fig. 1E shows the analysis of the wound-edge cells at 0.5 and 2 h after the addition of wound, which provides suitable data to verify the authors' claim. However, the corresponding legend indicates that these statistical data are based on the illustration in Fig. 1F, which is probably based on highly ambiguous data in Supplementary Fig. 2C (see above).  

Taken together, I strongly recommend the authors to remove Fig.1A-D. Instead, they should include the improved figure corresponding to the present Supplementary Fig.2C and present its statistical analysis similar to the present Fig.1E for this claim.

We sincerely thank the reviewer for their constructive feedback and recommendations. Below, we address the concerns raised regarding Figure 1A-D and Supplementary Figure 2C.

To provide stronger evidence for the transitional state of the Golgi apparatus during reorientation and the dynamic regulation of CAMSAP2 localization, we conducted live-cell imaging experiments. These results, now presented in Supplementary Figure 2A, clearly demonstrate that the Golgi apparatus undergoes a transitional state characterized by dispersion before reorienting toward the leading edge.

Additionally, we analyzed fixed wound-edge cells at different time points during directed migration to observe CAMSAP2’s colocalization with the Golgi apparatus. The results, shown in Figures 1A and 1B, reveal dynamic changes in CAMSAP2 localization, confirm its regulation during Golgi reorientation, and include a corresponding statistical analysis (page 7, lines 1-17).

These updates ensure that our claims are supported by robust and unambiguous data.

We hope these revisions address the reviewer’s concerns and provide clear and reliable evidence for the transitional state of the Golgi apparatus and CAMSAP2’s dynamic regulation. We are grateful for the reviewer’s constructive suggestions, which have greatly improved the quality and focus of our manuscript.

(3) In Supplementary Fig. 5 and Fig. 4, the authors claim that MARK2 phosphorylates S835 of CAMSAP2.  

There are many issues to be addressed. Otherwise, the above claim cannot be assumed to be reliable.  

First, the descriptions (in the text and method sections) and figures (Supplementary Fig.5) concerning the in vitro kinase assay and subsequent phosphoproteomic analysis are too immature and contain many errors.  

Legend to Supplementary Fig. 5 is too immature for comprehension. It should be completely rewritten in a more comprehensive manner. The figure in Supplementary Fig. 5C is also too immature for understanding. They simply paste raw mass spectrometric data without any modification for presentation.  

We sincerely apologize for the lack of clarity and inaccuracies in the original descriptions and figure legends for the in vitro kinase assay and phosphoproteomic analysis. We greatly appreciate the reviewer’s detailed comments, which have allowed us to address these issues comprehensively.

To improve clarity and accuracy, we have rewritten the figure legend for the original Supplementary Figure 5 (now Supplementary Figure 4) as follows:

(A): CBB staining of a gel with GFP-CAMSAP2, GST, and GST-MARK2. GFP-CAMSAP2 was expressed in Sf9 cells and purified. GST and GST-MARK2 were expressed in E. coli and purified.

(B): Western blot analysis of an in vitro kinase assay. GST or GST-MARK2 was incubated with GFP-CAMSAP2 in kinase buffer (50 mM Tris-HCl pH 7.5, 12.5 mM MgCl2, 1 mM DTT, 400 μM ATP) at 30°C for 30 minutes. Reactions were stopped by boiling in the loading buffer.

(C): Detection of phosphorylation at S835 in CAMSAP2 by mass spectrometry. The observed mass increases in b4, b5, b6, b7, b8, b10, b11, and b12 fragments indicate phosphorylation at Ser835.

(D): Kinase assay samples analyzed using Phos-tag SDS-PAGE. HEK293 cells were cotransfected with the indicated plasmids. Band shifts of CAMSAP2 mutants were examined via western blot. Phos-tag was used in SDS-PAGE, and arrowheads indicate the shifted bands caused by phosphorylation.

To address the reviewer’s concern about Supplementary Figure 5C, we have reformatted the mass spectrometry data to improve readability and presentation quality. The revised figure includes clearer annotations and graphical representations of the mass spectrometric evidence for phosphorylation at S835.

We believe these updates enhance the comprehensibility and reliability of our data, providing robust support for our claim that MARK2 phosphorylates CAMSAP2 at S835. We hope these

revisions address the reviewer’s concerns and demonstrate our commitment to improving the quality of our manuscript.

The readers cannot understand how the authors purified GFP-CAMSAP2 for the kinase assay.

The method section incorrectly states that the product was purified using Ni-resin.  

We thank the reviewer for their comment regarding the purification of GFP-CAMSAP2 for the kinase assay. We would like to clarify that GFP-CAMSAP2 carries a His-tag, which allows for purification using Ni-resin, as described in the Methods section (page 23, Lines 32-40). Therefore, the description in the Methods section is correct.

To avoid any potential misunderstanding, we have revised the Methods section to provide more detailed and precise descriptions of the purification process. Specifically, GFP-CAMSAP2 was cloned into the pOCC6_pOEM1-N-HIS6-EGFP vector, which includes a His-tag, and was expressed in Sf9 cells. The His-GFP-CAMSAP2 protein was purified using Ni-resin chromatography. Relevant details have been added to the Methods section (page 21, Lines 34-36:

“CAMSAP2 was cloned into the pOCC6_pOEM1-N-HIS6-EGFP vector expressed in Sf9, purified as His-GFP-CAMSAP2.”; page 23, Lines 32-33: “His-GFP-CAMSAP2 was cotransfected with bacmids into Sf9 cells to generate the passage 1 (P1) virus.”).

We hope these clarifications and revisions address the reviewer’s concern and improve the comprehensibility of our experimental details. We appreciate the reviewer’s feedback, which has helped us refine the manuscript.

In this relation, GST and GST-MARK2 are described as having been purified from Sf9 insect cells in the text section (page9) and legend to Supplementary Fig. 5, but from E. coli in the method section. Which is correct?  

We thank the reviewer for pointing out the inconsistencies in the descriptions regarding the source of GST and GST-MARK2. To clarify, both GST and GST-MARK2 were purified from E. coli, as stated in the Methods section (page 23, Lines 26-31). We have corrected the erroneous descriptions in the main text (page 8, Lines 35-36) and the legend to Supplementary Figure 4 to ensure consistency.

Additionally, we have updated the legend for Supplementary Figure 4A to state the sources of each protein explicitly:

“GFP-CAMSAP2 were expressed in Sf9 cells and purified. GST and GST-MARK2 were expressed in E. coli and purified.” (page 38, Lines 2-3)

These revisions ensure that the experimental details are accurate and consistent across the manuscript, eliminating any potential confusion. We appreciate the reviewer’s careful review and constructive feedback, which have helped us improve the clarity and reliability of our study.

Because the phosphoproteomic data (Supplementary Fig. 5C) are not provided clearly, the experimental data for Fig.4A, in which possible CAMSAP2 phosphorylation sites are illustrated, are completely unknown. For me, it is highly strange that only the serine residues are listed in Fig. 4A.

We sincerely thank the reviewer for raising this important point regarding Figure 4A and the phosphoproteomic data in Supplementary Figure 5C.

- Phosphorylation Sites in Figure 4A

The phosphorylation sites illustrated in Figure 4A are derived from our analysis of the original mass spectrometry data. These sites were included based on their high confidence scores and data reliability. Importantly, only serine residues met the stringent criteria for inclusion, as no threonine or tyrosine residues had sufficient evidence for phosphorylation. To clarify this, we have updated the figure legend for Figure 4A (page 32, Lines3-7).

- Improvements to Supplementary Figure 5C (Supplementary Figure 4D in the revised manuscript)

To enhance transparency and clarity, we have reformatted Supplementary Figure 4D to include clearer annotations. The revised figure highlights the phosphopeptides used to identify the phosphorylation sites and provides a more comprehensive presentation of the mass spectrometry data. To clarify this, we have updated the figure legend for Supplementary Figure 4D (page 38, Lines 11-13).

- Data Availability

We will follow the journal’s guidelines by uploading the raw mass spectrometry data to the required public database upon manuscript acceptance. This ensures that the data are accessible and reproducible in compliance with journal standards.

We hope these clarifications and updates address the reviewer’s concerns and improve the reliability and comprehensibility of our data presentation. We greatly appreciate the reviewer’s constructive feedback, which has helped us enhance the rigor and clarity of our manuscript.

Considering the crude nature of the GST-MARK2 sample used for the in vitro kinase assay (Supplementary Fig. 5A), it is unclear whether MARK2 is responsible for all phosphorylation sites on CAMSAP2 detected in the phosphoproteomic analysis. Furthermore, if GFP-CAMSAP2 was purified from Sf9 insect cells, these sites might have been phosphorylated before incubation for the in vitro kinase assay. The authors should address these issues by including a negative control using the kinase-dead mutant of MARK2 in their in vitro kinase assay.

We sincerely thank the reviewer for raising these important points regarding the potential prephosphorylation of GFP-CAMSAP2 and the role of MARK2 in the phosphorylation sites detected in our analysis.

To address the possibility that GFP-CAMSAP2 may have been pre-phosphorylated during its expression in Sf9 insect cells, we conducted an in vitro comparison. Specifically, we compared the band shifts observed in GST-MARK2 + GFP-CAMSAP2 versus GST + GFP-CAMSAP2 under identical conditions. As shown in Supplementary Figure 4B, the GST-MARK2 + GFP-CAMSAP2 group exhibited a clear upward band shift compared to the GST + GFP-CAMSAP2 group, indicating additional phosphorylation events induced by MARK2.

Regarding the inclusion of a kinase-dead MARK2 mutant as a negative control, we acknowledge this as a valuable suggestion for further confirming the specificity of MARK2 in phosphorylating CAMSAP2. While this experiment is not currently included, we plan to conduct it in our future studies to strengthen our findings.

We hope this clarification and the provided evidence address the reviewer’s concerns. We are grateful for this constructive feedback, which has helped us critically evaluate and refine our experimental approach.

(4) In Supplementary Fig.6A-C and Fig.5A-B, the authors claim that the phosphorylation of CAMSAP2 S835 is required for restoring the reduced reorientation of the Golgi in wound-healing cells and the delay in wound closure observed in MARK2 KO cells.  

If the aforementioned claim is adequately supported by experimental data, it indicates that the defects in Golgi repolarization and wound closure in MARK2 KO cells can be mainly attributed to the reduced phosphorylation of S835 of CAMSAP2 in HT1080. Considering the presence of many well-known substrates of MARK2 for regulating cell polarity, this claim is highly striking.  

However, to strongly support this conclusion, the authors should first perform a rescue experiment using MARK2 KO cells exogenously expressing MARK2. This step is essential for determining whether the defects observed in MARK2 KO cells are caused by the loss of MARK2 expression, but not by other artificial effects that were accidentally raised during the generation of the present MARK2 KO clone.  

We sincerely thank the reviewer for their insightful suggestion regarding the rescue experiment to confirm that the defects observed in MARK2 KO cells are specifically caused by the loss of MARK2 expression.

To address this, we performed a rescue experiment in MARK2 KO HT1080 cells by exogenously expressing GFP-MARK2. Our results, presented in Supplementary Figures 3C-E, demonstrate that GFP-MARK2 expression successfully restores the localization of CAMSAP2 on the Golgi apparatus in MARK2 KO cells.

These findings strongly support the conclusion that the defects in Golgi architecture and CAMSAP2 Golgi localization are directly attributable to the loss of MARK2 expression, rather than any artificial effects potentially introduced during the generation of the MARK2 KO clone.

We hope these additional experimental results address the reviewer’s concerns and provide robust evidence for the role of MARK2 in regulating Golgi reorientation and wound closure. We are grateful for the reviewer’s constructive feedback, which has significantly improved the rigor and clarity of our study.

In addition, to evaluate the impact of the rescue effect of CAMSAP2, the authors should include the data of wild-type HT1080 and MARK2 KO cells in Fig. 5B to reliably demonstrate the aforementioned claim.  

We thank the reviewer for their valuable suggestion to include data from wild-type HT1080 and MARK2 KO cells in Figure 5A-C to better evaluate the rescue effects of CAMSAP2.

In response, we have incorporated data from wild-type HT1080 and MARK2 KO cells into Figure 5A-C. These additions provide a comprehensive comparison and further demonstrate the impact of CAMSAP2-S835A and CAMSAP2-S835D on Golgi reorientation relative to the wild-type and MARK2 KO conditions.

These changes are reflected in Figures 5A-C.

We hope these updates address the reviewer’s concerns and strengthen the reliability of our conclusions. We greatly appreciate the reviewer’s constructive feedback, which has significantly enhanced the robustness of our study.

Principally, before checking the rescue effects in MARK2 KO cells, the authors should examine the rescue activity of the CAMSAP2 S835 mutants in restoring the reduced reorientation of the Golgi in wound-healing cells and the delay in wound closure observed in CAMSAP2 KO cells (Supplementary Fig.1F-H and Supplementary Fig.2A, B). These experiments are more essential experiments to substantiate the authors' claim.

We thank the reviewer for their insightful suggestion to examine the rescue activity of CAMSAP2 S835 mutants in CAMSAP2 KO cells to further substantiate our claims.

In Figure 4D-F, we observed significant differences between CAMSAP2 S835 mutants in their ability to restore Golgi structure and localization, indicating functional differences between these mutants. To better reflect the regulatory role of MARK2-mediated phosphorylation of CAMSAP2, we performed scratch wound-healing experiments in MARK2 KO cells by establishing stable cell lines expressing CAMSAP2 S835 mutants. These experiments allowed us to assess Golgi reorientation during wound healing and are presented in Figure 5A-C.

We also attempted to generate stable cell lines expressing GFP-CAMSAP2 and its mutants in CAMSAP2 KO cells. Unfortunately, these cells consistently failed to survive, preventing successful construction of the cell lines.

We hope these experiments and explanations address the reviewer’s concerns. We are grateful for the reviewer’s constructive feedback, which has helped us refine and improve our study.

(5) The data presented in Fig. 6A and B are not sufficient to support the authors' notion that "our observation revealed notable changes in the Golgi apparatus and microtubule network distribution in relation to the wounding. (page 11)"  

Fig. 6A, which includes only a single-cell image in each panel, does not demonstrate the general state of microtubules and the Golgi in the wound-edge cells. The reader cannot even know the migration direction of each cell.  

Fig.6 B are not suitable to quantitatively support the authors' claim. The authors should find a way to quantitatively estimate the microtubule density around the Golgi and the shape and compactness of the Golgi in each cell facing the wound, not estimating the colocalization of microtubules and the Golgi, as in the present Fig. 6B.  

We sincerely apologize for the confusion caused by our unclear descriptions and presentation.

Here, we clarify the purpose and improvements made to address the reviewer’s concerns. In this study, we primarily aimed to observe the relationship between microtubules and the Golgi apparatus in cells at the leading edge of the wound during directed migration. In Figure 6A (now Supplementary Figure 6E), the images represent cells located at the wound edge at different time points. To improve clarity, we have added arrows indicating the migration direction and updated the figure legend to describe these details (page 40 lines 13-14).

To better quantify the relationship between microtubules and the Golgi apparatus, we revised our analysis by referring to the quantitative method used in Figure 3F of the paper Molecular Pathway of Microtubule Organization at the Golgi Apparatus. Specifically, we performed a radial analysis of fluorescence intensity in cells at the wound edge, measuring the distance from the Golgi center (x-axis) and the normalized radial fluorescence intensity of microtubules and the Golgi (y-axis). These results are now presented in Supplementary Figure 6E and 6F.

We hope these improvements address the reviewer’s concerns and provide stronger evidence for the changes in the Golgi apparatus and microtubule network distribution in relation to wound healing. We greatly appreciate the reviewer’s constructive feedback, which has significantly enhanced the clarity and rigor of our study.

The legends to Fig. 6A and B indicate that they compared immunofluorescent staining of cells at the edge of the wound after 0.5h and 2 h of migration. However, the authors state in the text that they compared "the cells located before the wound" and "the cells at the trailing edge of the wounding (page 11)."Although this description is highly ambiguous and misleading, if they compared the wound-edge cells and the cells separated from the wound edge at 2 h after cell migration here, they should improve the experimental design as I pointed out in the 2nd major comment.  

We thank the reviewer for their detailed feedback regarding the experimental design and the need to clarify our descriptions. We have addressed these concerns as follows:

- Clarification of descriptions:

We recognize that the previous description in the text regarding "the cells located before the wound" and "the cells at the trailing edge of the wounding" was ambiguous and potentially misleading. We have revised this text to accurately describe the experimental design. Specifically, we compared cells at the leading edge of the wound at different time points (0.5h and 2h post-migration). These corrections are reflected in figure legends (Supplementary Figure 6E and 6F ) and the Results section (page 11,lines 3-8).

- Improved experimental design:

To better support our conclusions, we performed live-cell imaging to observe the dynamic changes in the Golgi apparatus during directed migration. As shown in Supplementary Figure 2A, our results confirm that the Golgi apparatus undergoes a transient dispersed state before reorganizing into an intact structure.

Additionally, we performed fixed-cell staining at different time points to analyze the colocalization of CAMSAP2 with the Golgi apparatus in cells at the leading edge of the wound. The colocalization analysis, presented in Figures 1A-C, further demonstrates the dynamic regulation of CAMSAP2 during Golgi reorientation.

We hope these updates address the reviewer’s concerns and provide a clearer and more robust foundation for our conclusions. We are grateful for the reviewer’s constructive feedback, which has greatly enhanced the clarity and rigor of our study.

Minor comments  

(1) In Fig. 2 and Supplementary Fig. 3, the authors claim that MARK2 is enriched around the Golgi. However, this claim was based on immunofluorescent images of single cells and single-line scans.  

It is better to present the statistical data for Pearson's coefficient as shown in Figs. 1D and E. To demonstrateMARK2 enrichment around Golgi, but not localization in Golgi, the authors should find a way to quantify the specific enrichment of MARK2 signals in the Golgi region.  

We thank the reviewer for raising this important point regarding the enrichment of MARK2 around the Golgi apparatus. Upon further consideration, we acknowledge that our current data do not provide sufficient evidence to fully elucidate the mechanism of MARK2 localization to the Golgi.

To maintain the scientific rigor of our study, we have removed this claim and the corresponding content from the manuscript, including original Figures 2 and Supplementary Figure 3 that specifically discuss MARK2 enrichment. These changes do not affect the primary conclusions of the study, which focus on the role of MARK2-mediated phosphorylation of CAMSAP2.

We hope this clarification addresses the reviewer’s concerns. In the future, we plan to investigate the precise mechanism of MARK2 localization using additional experimental approaches. We are grateful for the reviewer’s constructive feedback, which has helped us refine the scope and focus of our manuscript.

(2) In Fig. 3 and Supplementary Fig. 4, the authors report that CAMSAP2 localization on the Golgi is reduced in cells lacking MARK2.  

Essentially, the present results support this claim. However, the authors should analyze the Golgi localization of CAMASP2 with the same quantification parameter because they used Pearson's coefficient in Fig. 1D, E and Supplementary Fig.4D but Mander's coefficient in Fig. 3C and Fig.4F.  

We thank the reviewer for their insightful comment regarding the consistency of quantification parameters used in our analysis of CAMSAP2 localization on the Golgi apparatus.

To address this concern, we have revised Figure 3C to use Pearson’s coefficient for consistency with Figure 1D, 1E (Figure 1B and 1E in the revised manuscript), and Supplementary Figure 4D (Supplementary Figure 3I in the revised manuscript). This ensures uniformity in the quantification parameters across these analyses.

For Figure 4F, we have retained Mander’s coefficient, as it accounts for variability in expression levels due to overexpression in individual cells. We believe this approach provides a more accurate reflection of CAMSAP2 localization under the experimental conditions shown in Figure 4F.

We hope these adjustments clarify our analysis and address the reviewer’s concerns. We greatly appreciate the reviewer’s constructive feedback, which has helped improve the consistency and accuracy of our study.

(3) In Fig.4D-F, the authors claim that S835 phosphorylation of CAMSAP2 is essential for its localization to the Golgi apparatus and for restoring the Golgi dispersion induced by CAMASAP2 depletion.  

Fig.4E indicates that the S835A mutant of CAMSAP2 significantly restores the compact assembly of the Golgi apparatus, and the differences in the rescue activities of the wild type, S835A, and S835D are rather small. These data contradict the authors' conclusions regarding the pivotal role of MARK2-mediated phosphorylation at the S835 site of CAMSAP2 in maintaining the Golgi architecture (page 9). The authors should remove the phrase "MARK2-mediated" from the sentence unless addressing the aforementioned issues (see 3rd major comment) and describe the role of S835 phosphorylation in more subdued tone.  

We thank the reviewer for their constructive feedback regarding the conclusions drawn about the role of MARK2-mediated phosphorylation of CAMSAP2 at S835.

In response, we have revised the relevant sentence to reflect a more nuanced interpretation of the data. Specifically, the original statement:

“These observations indicate that the phosphorylation of serine 835 in CAMSAP2 is essential for its proper localization to the Golgi apparatus.”

has been updated to:

“These observations indicate that MARK2 phosphorylation of serine at position 835 of CAMSAP2 affects the localization of CAMSAP2 on the Golgi and regulates Golgi structure” (page 9, Lines 27-29).

We hope this modification addresses the reviewer’s concerns. We are grateful for the feedback, which has helped us refine our conclusions and enhance the clarity of our manuscript.

(4) In Figs. 5I, J and Supplementary Fig.7A-E, the authors claim that the S835 phosphorylationdependent interaction of CAMSAP2 with Uso1 is essential for its localization to the Golgi apparatus.  

This claim was made based on immunofluorescent images of single cells and single-line scans, and was not sufficiently verified (Supplementary Fig.7B, C). Because this is a crucial claim for the present paper, the authors should present statistical data for Pearson's coefficient, as shown in Fig. 1D and E, to quantitatively estimate the Golgi localization of CAMSAP2.  

We thank the reviewer for their suggestion to present statistical data using Pearson's coefficient for a more robust quantification of the Golgi localization of CAMSAP2.

In response, we have revised the statistical analysis for Supplementary Figures 7B-C (Revised Figures 6F and 6G) to use Pearson's coefficient. This change ensures consistency with the quantification methods used in Figures 1D and 1E (Revised Figures 1B and 1E), allowing for a more standardized evaluation of CAMSAP2’s localization to the Golgi apparatus.

We hope this modification addresses the reviewer’s concerns and strengthens the quantitative support for our claims. We are grateful for the reviewer’s constructive feedback, which has helped improve the rigor of our study.

(5) The signal intensities of the immunofluorescent data in Fig. 4D, Fig. 5A, Sup-Fig. 3C and E, and Sup-Fig. 7S are very weak for readers to clearly estimate the authors' claims. They should be improved appropriately.  

We thank the reviewer for highlighting the need to improve the clarity of the immunofluorescent data presented in several figures.

In response, we have enhanced the signal intensities in Figures 4D, 5A, and Supplementary Figure 7D (Revised Supplementary Figure 6A) to make the signals clearer for readers, while ensuring that the adjustments do not alter the integrity of the original data. Supplementary Figures 3C and 3E was remove from our manuscript.

Additionally, to improve consistency and readability across the manuscript, we have standardized the quantification methods for similar analyses:

For CAMSAP2 localization to the Golgi, Pearson's coefficient has been used throughout the manuscript. Figure 3C has been updated to use Pearson's coefficient for consistency.

For Golgi state analysis in wound-edge cells, we have used the Golgi position relative to the nucleus as a uniform metric. This has been applied to Supplementary Figures 1F and 1G, Figures 2D and 2E, and Figures 5A and 5B.

We hope these adjustments address the reviewer’s concerns and improve the clarity and consistency of our study. We greatly appreciate the reviewer’s constructive feedback, which has significantly enhanced the quality of our manuscript.

(6) As indicated above, the authors frequently change the parameters or methods for quantifying the same phenomena (for example, the localization of CAMSAP on the Golgi and Golgi state in wound edge cells) in each figure. This is highly confusing. They should unify them.  

We thank the reviewer for their valuable feedback regarding the inconsistency in quantification methods across the manuscript.

To address this concern, we have carefully reviewed the entire manuscript and standardized the methods used for quantifying similar phenomena:

- CAMSAP2 localization on the Golgi: 

Pearson's coefficient is now consistently used throughout the manuscript. For example, Figure 3C has been updated to use Pearson's coefficient to align with other figures, such as Figures 1B and 1E.

- Golgi state in wound-edge cells: 

The Golgi state is now uniformly measured based on the position of the Golgi relative to the nucleus. This method has been applied to Supplementary Figures 1F and 1G, Figures 2D and 2E, and Figures 5A and 5B.

We believe these changes significantly improve the clarity and consistency of the manuscript, ensuring that readers can easily interpret the data. We are grateful for the reviewer’s constructive feedback, which has greatly helped us enhance the quality and rigor of our study.

(7) The legends frequently fail to clearly indicate the number of independent experiments on which each statistical analysis was based.  

We thank the reviewer for highlighting the need to clearly indicate the number of independent experiments for each statistical analysis.

In response, we have carefully reviewed the entire manuscript and updated the figure legends to include the number of independent experiments for every statistical analysis. This ensures transparency and allows readers to better evaluate the reliability of the data.

We hope these updates address the reviewer’s concerns and improve the clarity and rigor of the manuscript. We appreciate the reviewer’s constructive feedback, which has helped us enhance the quality of our work.

(8) Supplemental Figs. 4E and 4F are not cited in the text.  

We thank the reviewer for pointing out that Supplemental Figures 4E and 4F were not cited in the text.

To address this, we have updated the manuscript to cite these figures (Revised Figures 2H and 2I) in the appropriate section (page 8, lines 1-5).

“the absence of MARK2 can also influence the orientation of the Golgi apparatus during cell wound healing and cause a delay in wound closure (Figure 2 D-I and Figure 3 D).”

We hope this revision resolves the reviewer’s concern and improves the clarity and completeness of the manuscript. We appreciate the reviewer’s feedback, which has helped us refine our work.

(9) The data in Fig. 3 analyzed MARK2 knockout cells (not knockdown cells). The caption should be corrected.  

We thank the reviewer for pointing out the incorrect use of "knockdown" in the caption of Figure 3.

To address this, we have revised the title of Figure 3 from:

“MARK2 knockdown reduces CAMSAP2 localization on the Golgi apparatus.”

to:

“MARK2 affects CAMSAP2 localization on the Golgi apparatus.”

This updated caption reflects the inclusion of both MARK2 knockout and knockdown cell lines analyzed in Figure 3.

We hope this correction resolves the reviewer’s concern and ensures the accuracy of our manuscript. We greatly appreciate the reviewer’s attention to detail, which has helped us improve the clarity and consistency of our work.

(10) The present caption in Fig. 6 disagrees with the content of the figure.  

We thank the reviewer for pointing out the inconsistency between the caption and the content of Figure 6.

To address this issue, we have revised the content of Figure 6 to ensure it aligns accurately with the caption. The updated figure now reflects the description provided in the caption, eliminating any discrepancies and improving clarity for the readers.

We appreciate the reviewer’s constructive feedback, which has helped us enhance the accuracy and presentation of our manuscript.

(11) What do "CS" indicate in Fig. 4B and Supplementary Fig. 5D? The style used to indicate point mutants of CAMSAP2 should be unified. 835A or S835A?  

We thank the reviewer for pointing out the inconsistency in the naming of CAMSAP2 mutants.

To address this, we have revised all relevant figures and text to use the consistent format "S835A" and "S589A" for CAMSAP2 mutants. Specifically, in Figure 4B and Supplementary Figure 5D (now Supplementary Figure 4C), we have replaced the abbreviation "CS2" with "CAMSAP2" and updated the mutant names from "835A" and "589A" to "S835A" and "S589A," respectively. We hope these updates resolve the reviewer’s concerns and ensure clarity and consistency throughout the manuscript. We are grateful for the reviewer’s attention to detail, which has helped us improve the quality of our work.

(12) Uso1 is not a Golgi matrix protein.  

We thank the reviewer for pointing out the incorrect description of Uso1 as a Golgi matrix protein.

In response, we have revised the manuscript to replace all references to “USO1 as a Golgi matrix protein” with “USO1 as a Golgi-associated protein.” This correction ensures that the terminology used in the manuscript is accurate and consistent with current scientific understanding.

We appreciate the reviewer’s attention to detail, which has helped us improve the accuracy and quality of our manuscript.

Reviewer #1 (Public review):

Summary

In this manuscript, De La Forest Divonne et al. build a repertory of hemocytes from adult Pacific oysters combining scRNAseq data with cytologic and biochemical analyses. Three categories of hemocytes were described previously in this species (i.e. blast, hyalinocyte and granulocytes). Based on scRNAseq data, the authors identified 7 hemocyte clusters presenting distinct transcriptional signatures. Using Kegg pathway enrichment and RBGOA, the authors determined the main molecular features of the clusters. In parallel, using cytologic markers, the authors classified 7 populations of hemocytes (i.e. ML, H, BBL, ABL, SGC, BGC, and VC) presenting distinct sizes, nucleus sizes, acidophilic/basophilic, presence of pseudopods, cytoplasm/nucleus ratio and presence of granules. Then, the authors compared the phenotypic features with potential transcriptional signatures seen in the scRNAseq. The hemocytes were separated in a density gradient to enrich for specific subpopulations. The cell composition of each cell fraction was determined using cytologic markers and the cell fractions were analysed by quantitative PCR targeting major cluster markers (two per cluster). With this approach, the authors could assign cluster 7 to VC, cluster 2 to H, and cluster 3 to SGC. The other clusters did not show a clear association with this experimental approach. Using phagocytic assays, ROS, and copper monitoring, the authors showed that ML and SGC are phagocytic, ML produces ROS, and SGC and BGC accumulate copper. Then with the density gradient/qPCR approach, the authors identified the populations expressing anti-microbial peptides (ABL, BBL, and H). At last, the authors used Monocle to predict differentiation trajectories for each subgroup of hemocytes using cluster 4 as the progenitor subpopulation.

The manuscript provides a comprehensive characterisation of the diversity of circulating immune cells found in Pacific oysters.

Strengths

The combination of scRNAseq, cytologic markers and gradient based hemocyte sorting offers an integrative view of the immune cell diversity.<br /> Hemocytes represent a very plastic cell population that has key roles in homeostatic and challenged conditions. Grasping the molecular features of these cells at the single-cell level will help understand their biology.<br /> This type of study may help elucidate the diversification of immune cells in comparative studies and evolutionary immunology.

Weaknesses

Several figures show inconsistency leading to erroneous conclusions and some conclusions are poorly supported. Moreover, the manuscript remains highly descriptive with limited comparison with the available literature.

Reviewer #2 (Public review):

Summary:

This work provides a comprehensive understanding of cellular immunity in bivalves. To precisely describe the hemocytes of the oyster C. gigas, the authors morphologically characterized seven distinct cell groups, which they then correlated with single-cell RNA sequencing analysis, also resulting in seven transcriptional profiles. They employed multiple strategies to establish relationships between each morphotype and the scRNAseq profile. The authors correlated the presence of marker genes from each cluster identified in scRNAseq with hemolymph fractions enriched for different hemocyte morphotypes. This approach allowed them to correlate three of the seven cell types, namely hyalinocytes (H), small granule cells (SGC), and vesicular cells (VC). A macrophage-like (ML) cell type was correlated through the expression of macrophage-specific genes and its capacity to produce reactive oxygen species. Three other cell types correspond to blast-like cells, including an immature blast cell type from which distinct hematopoietic lineages originate to give rise to H, SGC, VC, and ML cells. Additionally, ML cells and SGCs demonstrated phagocytic properties, with SGCs also involved in metal homeostasis. On the other hand, H cells, non-granular cells, and blast cells expressed antimicrobial peptides. This study thus provides a complete landscape of oyster hemocytes with functional validation linked to immune activities. This resource will be valuable for studying the impact of bacterial or viral infections in oysters.

The main strength of this study lies in its comprehensive and integrative approach, combining single-cell RNA sequencing, cytological analysis, cell fractionation and functional assays to provide a robust characterization of hemocyte populations in Crassostrea gigas.

(1) The innovative use of marker genes, quantifying their expression within specific cell fractions, allows for precise annotation of different cellular clusters, bridging the gap between morphological observations and transcriptional profiles.

(2)The study provides detailed insights into the immune functions of different hemocyte types, including the identification of professional phagocytes, ROS-producing cells, and cells expressing antimicrobial peptides.

(3) The identification and analysis of transcription factors specific to different hemocyte types and lineages offer crucial insights into cell fate determination and differentiation processes in oyster immune cells.

(4) The authors significantly advance the understanding of oyster immune cell diversity by identifying and characterizing seven distinct hemocyte transcriptomic clusters and morphotypes.

These strengths collectively make this study a significant contribution to the field of invertebrate immunology, providing a comprehensive framework for understanding oyster hemocyte diversity and function.

Conclusion:

The authors largely achieved their primary objective of providing a comprehensive characterization of oyster immune cells. They successfully integrated multiple approaches to identify and describe distinct hemocyte types. The correlation of these cell types with specific immune functions represents a significant advancement in understanding oyster immunity. The authors are aware of the limitations of their study, particularly with regards to the pseudotime analysis, which provides a conceptual framework for understanding lineage relationships but requires further experimental validation to confirm its findings.

This study is likely to have a significant impact on the field of invertebrate immunology, particularly in bivalve research. It provides a new standard for comprehensive immune cell characterization in invertebrates. The identification of specific markers for different hemocyte types will facilitate future research on oyster immunity. The proposed model of hemocyte lineages, while requiring further validation, offers a framework for studying hematopoiesis in bivalves.

Author response:

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

Reviewer #1 (Public review):

Summary:

In this manuscript, De La Forest Divonne et al. build a repertory of hemocytes from adult Pacific oysters combining scRNAseq data with cytologic and biochemical analyses. Three categories of hemocytes were described previously in this species (i.e. blast, hyalinocyte, and granulocytes). Based on scRNAseq data, the authors identified 7 hemocyte clusters presenting distinct transcriptional signatures. Using Kegg pathway enrichment and RBGOA, the authors determined the main molecular features of the clusters. In parallel, using cytologic markers, the authors classified 7 populations of hemocytes (i.e. ML, H, BBL, ABL, SGC, BGC, and VC) presenting distinct sizes, nucleus sizes, acidophilic/basophilic, presence of pseudopods, cytoplasm/nucleus ratio and presence of granules. Then, the authors compared the phenotypic features with potential transcriptional signatures seen in the scRNAseq. The hemocytes were separated in a density gradient to enrich for specific subpopulations. The cell composition of each cell fraction was determined using cytologic markers and the cell fractions were analysed by quantitative PCR targeting major cluster markers (two per cluster). With this approach, the authors could assign cluster 7 to VC, cluster 2 to H, and cluster 3 to SGC. The other clusters did not show a clear association with this experimental approach. Using phagocytic assays, ROS, and copper monitoring, the authors showed that ML and SGC are phagocytic, ML produces ROS, and SGC and BGC accumulate copper. Then with the density gradient/qPCR approach, the authors identified the populations expressing anti-microbial peptides (ABL, BBL, and H). At last, the authors used Monocle to predict differentiation trajectories for each subgroup of hemocytes using cluster 4 as the progenitor subpopulation.

The manuscript provides a comprehensive characterisation of the diversity of circulating immune cells found in Pacific oysters.

Strengths:

The combination of the two approaches offers a more integrative view.

Hemocytes represent a very plastic cell population that has key roles in homeostatic and challenged conditions. Grasping the molecular features of these cells at the single-cell level will help understand their biology.

This type of study may help elucidate the diversification of immune cells in comparative studies and evolutionary immunology.

Weaknesses:

The study should be more cautious about the conclusions, include further analyses, and inscribe the work in a more general framework.

Reviewer #1 (Recommendations for the authors):

The manuscript provides a comprehensive characterisation of the diversity of circulating immune cells found in Pacific oysters.

Major comments:

(1) The introduction would benefit from a clear description of what is known about immune cell development and diversity in this model. The bibliography on the three subtypes origins and properties (i.e. blast, hyalinocyte, and granulocytes) should be described in the introduction.

We thank Reviewer #1 for their valuable comments, which have allowed us to further improve our manuscript. We have enriched the introduction with the following addition (line 79 to 82):

“Blast-like cells are considered as undifferentiated hemocyte types (20), hyalinocytes (21) seem to be more involved in wound repair, and granulocytes, more implicated in immune surveillance. The latter are considered as the main immunocompetent hemocyte types (22).”

(2) The authors mentioned a previous scRNAseq dataset produced in another oyster species. They should compare the two datasets to show the robustness of the molecular signatures determined in the present study. In addition, the authors do not mention markers identified in the literature that could be relevant to characterize the clusters (e.g. inflammatory pathway PMID: 29751033, proliferative markers PMID: 36591234/ PMID: 29317231, granulocyte markers PMID: 30633961 ... list not exhaustive). Overall, the comparison of this manuscript dataset and the available literature is too partial

We appreciate the reviewer’s suggestion to compare our dataset with previously published scRNAseq data and to integrate markers from the literature. Below, we address these points in detail.

The transcription factors involved in hematopoiesis, such as Tal1, Sox, Runx, and GATA, are highly conserved across metazoans. These markers were identified in our dataset, consistent with findings in other species (1–3), including the previously mentioned scRNA-seq dataset in C. hongkongensis (4). However, defining robust and specific markers for distinct hemocyte types remains an ambitious goal that requires validation across diverse biological contexts - work that is beyond the scope of the present study. Additionally, meaningful comparisons between datasets are constrained by differences in annotation frameworks and the absence of a standardized system for defining hemocyte subtypes. These limitations underscore the need for harmonization efforts to facilitate robust cross-study comparisons. Nonetheless, our dataset provides a strong foundation for future comparative analyses once such standardization is achieved.

In response to the reviewer’s comment, we have added a paragraph to the discussion (lines 747 - 760) detailing that we identified conserved transcription factor markers in C. gigas and C. hongkongensis.

(3) The authors sequenced 3000 cells without providing more comprehensive information/rationale on the analysed population. What is the number of hemocytes found in an adult? What proportion of the whole hemocyte population does this analysis represent? Does it include the tissue-interacting hemocytes? Also, what is the rationale for choosing that specific stage?

We thank the reviewer for their insightful questions regarding the analyzed hemocyte population.

Adult 18-month-old Crassostrea gigas contain approximately 1 million circulating hemocytes per mL of hemolymph, with an average of 1 mL of hemolymph per individual. Thus, this represents approximately 1 million circulating hemocytes per oyster. For our scRNA-seq analysis, we sampled 3,000 hemocytes, which corresponds to 0.3% of the total circulating hemocyte population.

The number of cells processed was optimized to minimize the occurrence of doublets during scRNAseq. Following 10x Genomics Chromium guidelines, we loaded 4,950 cells to successfully recover a target of 3,000 cells, with a doublet rate of 2.4%, well below the target threshold of 2.5%. This information has been added on line 125 of the document. The target was 3,000 cells, and as reported in Supplementary Table S1, the estimated number of cells after STAR-solo alignment was 2,937. This ensures the reliability and accuracy of single-cell transcriptomic data.

We selected 18-month-old oysters for two key reasons: (i) to facilitate hemolymph collection, as hemocyte counts are more stable and sufficient at this stage, enabling us to collect enough cells for all planned experiments, including functional and cytological analyses; and (ii) to use oysters that are not susceptible to OsHV-1 μVar herpesvirus, which predominantly affects younger animals. This ensured that the hemocyte populations analyzed were not influenced by viral infections or related immune responses.

Our study focused on circulating hemocytes collected from hemolymph, which does not include tissue-interacting hemocytes. While these cells may represent an additional population of interest, they fall outside the scope of our current investigation.

By carefully selecting the animal stage and optimizing cell sampling, we ensured that the scRNA-seq dataset provides a robust representation of circulating hemocyte diversity while maintaining high data quality.

(4) For the GO term enrichment analysis, the authors included all genes presenting a cluster enrichment above L2FC>0.25. This seems extremely low to find distinct functions for each cluster. The risk is to call "cluster specific GO term" GO terms for which the genes are poorly enriched in the cluster. For the most important GO term mentioned in the text, the authors should show the expression levels of the genes (with DotPlot similar to Fig1D) to illustrate the specificity of the GO term. At last, the GO enrichment scores were apparently calculated using the whole genome as background. The analysis, aiming at finding differences between hemocyte subgroups, should use the genes detected in the dataset as background.

We appreciate the reviewer's concerns regarding the threshold used for GO term enrichment analysis and the choice of background genes. Below, we provide clarification on these points.

For nuanced comparisons, such as those between activation states of the same cell type, lower thresholds for log2FC (e.g., ≥0.25) are commonly used to detect subtle regulatory shifts. In single-cell RNA sequencing (scRNA-seq) analyses, it is typical to use a log2FC threshold between 0.25 and 0.5 to ensure that biologically relevant, yet subtle, changes are captured. For our analysis, this threshold was chosen to maintain sensitivity to such shifts, particularly given the diversity and functional specialization of hemocyte clusters.

To address the reviewer's suggestion, we will include DotPlot representations (similar to Fig. 1D) for the most significant GO terms highlighted in the text. This will illustrate the expression levels of the associated genes across clusters and demonstrate their specificity to the identified GO terms.

Regarding the background used in the GO enrichment analysis, we employed the Rank Based Gene Ontology Analysis (RBGOA) approach, which explicitly states in its documentation: "It is important to have the latter two tables representing the whole genome (or transcriptome) — at least the portion that was measured — rather than some select group of genes since the test relies on comparing the behavior of individual GO categories to the whole." Our analysis was conducted in agreement with these initial recommendations, ensuring that the results are consistent with the methodology outlined for RBGOA.

(5) The authors reannotated the genes of C. gigas to reach 73.1% annotation. What are the levels of annotations found prior to the reannotation? What do the scores/scale bars from the RBGOA analysis mean in Figures 2B-D?

Thank you for your comment. The original annotation for C. gigas was based on the work of Penaloza et al. (5), which provided GO annotations for 18,750 out of 30,724 genes, corresponding to 61% annotation. Following our reannotation efforts, we were able to increase the annotation coverage to 73.1%, enhancing the resolution of downstream analyses. In response to the reviewer’s comment, we have updated the results section (line 211 and 216) to explicitly include the original annotation coverage of 61% from the work of Penaloza et al., followed by details on our newly achieved annotation percentage of 73.1%.

Thank you for pointing this out. We apologize for the oversight regarding the scale bar in Figures 2BD. The colors in the original figure correspond to a z-score calculated from the gene ratio, which was not clearly explained and may have caused confusion. In the revised version of the manuscript, we propose a new representation to facilitate understanding and improve the clarity of the data presentation (Figure 2B).

(6) The authors describe first the result of the Kegg enrichment analysis and then of the RBGOA. To gain fluidity, I would suggest merging the results of both Kegg and RBGOA for each cluster.

Thank you for the suggestion. To enhance the fluidity of the results section, we have redesigned the KEGG/RBGOA figure (see figure 2A and 2B) to present the results for each cluster in an integrated manner. This revised approach aims to provide a clearer and more cohesive representation of the findings.

(7) The authors make correlations between gradient fraction containing multiple hemocyte populations and qPCR expression levels of cluster-specific markers to associated cytologic features with specific clusters. If feasible, I would recommend validating the association of several markers with hemocyte subgroups using in situ hybridisation or immunolabelling.

Cytological identification of hemocytes in our study relies on MCDH staining, which provides detailed morphological and cytological information. Unfortunately, the fixation methods required for in situ hybridization (ISH) or immunolabeling are not compatible with those used for MCDH staining. We attempted to combine these approaches but found that the fixation protocols necessary for ISH or immunolabeling compromised the quality of the cytological features observed with MCDH staining. Consequently, such validation was not feasible within the constraints of our experimental setup.

(8) Anti-microbial peptides are mentioned as enriched in agranular cells based on the gradient/qPCR analysis (Figure 6). Are these AMPs regulated by inflammatory pathways? Are any inflammatory pathways enriched in any scRNAseq cluster? In addition, without validating the data by directly labelling AMP in the different populations, it seems hard to conclude that AMP are expressed only by agranular cells.

In oysters, two families of antimicrobial peptides/proteins appear to be transcriptionally regulated in hemocytes in response to an infection. The first is that of Cg-BigDefs (6). A 2020 article indicates that the expression of CgBigDef1 is regulated by CgRel, an ortholog of the NFkB transcription factor, which also control the expression of the proinflammatory cytokine CgIL17 (7). Cg-BPI is induced in response to infection but its regulatory pathways remain unknown (8). The last well characterized family of antimicrobial peptides is Cg-Defs. It exhibits constitutive expression in hemocytes.

In our scRNA-seq analysis, CgRel (G12420) shows an increased expression in cluster 5, with a log2FC of 0.4 (equivalent to a 1.32-fold change or 32% higher expression compared to other clusters). Cluster 5 corresponds to blast-like cells, which are transcriptionally distinct and predominantly found in fractions 1, 2, and 3. These same fractions exhibit the highest CgBigDef expression, as demonstrated by qPCR.

From our qPCR results, we see no expression of the three AMP families in cell-sorted granular cells while the cell-sorted agranular cells are positive for the three AMP families, even for inducible ones. Still, we agree that labelling of cell sorted hemocyte populations would reinforce our data. We now specify in the text that further staining would be necessary to confirm these transcriptomic results (Discussion, lines 695 to 296).

(9) The authors should play down some statements concerning cluster identity. In the absence of a true lineage tracing approach, it is possible that those clusters represent states rather than true cell subtypes. Immune cells are very plastic in nature and able to adapt to the environment, even in conditions that are considered homeostatic.

We appreciate the reviewer’s insightful comment regarding the plasticity of immune cells and the potential for clusters to represent states rather than distinct cell subtypes. We agree that, in the absence of a lineage tracing approach, definitive classification of clusters as fixed subtypes is challenging. Immune cells, including those in invertebrates, are known for their high degree of plasticity and adaptability to environmental cues.

In response to the reviewer’s comment, we have revised the Discussion section to include a statement clarifying that these clusters may represent dynamic states rather than fixed subtypes, thereby acknowledging the plasticity of immune cells (lines 766 to 770).

(10) Related to the above issue, there is no indication of stem cells being present in the cell population. Is there any possibility to look for proliferative or progenitor markers? In homeostatic and in challenged conditions (for example Zymosan treatment)? This would provide some hints into the cellular pathways involved in the response. Perhaps determining the number/fraction of phagocytic cells in challenged conditions would help as well, in the absence of time-lapse assays.

Thank you for highlighting the possibility of stem cells or progenitor markers in our hemocyte populations. In our current analysis, we did not detect any known stem cell or proliferative markers, nor evidence of a clearly defined hematopoiesis site in the hemolymph. Indeed, previous work suggests that oyster hematopoiesis may occur in tissues such as the gills, implying that stem or progenitor cells might not circulate in the hemolymph under homeostatic conditions. Consequently, it is plausible that our observation of no proliferative cell populations partly reflects their absence in hemolymph, especially in naïve (unstimulated) oysters. To conclusively identify potential progenitor cells and their proliferative activity, further approaches involving deliberate perturbation of hemocyte homeostasis - such as immunological challenge (e.g., Zymosan treatment) combined with lineagetracing or proliferation assays - would be necessary. These future investigations would not only clarify whether proliferative cells emerge in the hemolymph in response to environmental or pathological stimuli but also help elucidate the broader cellular pathways underlying oyster immune responses.

In response to the reviewer’s comment, we have revised the Discussion (lines 742 to 745) and added : “Nevertheless, we did not detect any canonical stem or progenitor cell populations in our dataset, underscoring the need for future investigations - potentially involving immunological challenges and lineage-tracing assays - to clarify whether proliferative cells circulate in the hemolymph or instead reside primarily in tissue compartments.”

(11) Could the authors discuss the phagocytic hemocytes in light of scavenger receptor expression?

We thank the reviewer for this insightful question. Our study identifies macrophage-like cells and small granule cells as the principal phagocytes in Crassostrea gigas, capable of robust pathogen engulfment. Transcriptomic data reveal that these cell types express markers associated with endocytosis and immune defense pathways, such as CLEC and LACC24, which are integral to their phagocytic functionality.

Interestingly, our single-cell RNA sequencing analysis indicates that cluster 3, corresponding to small granule cells, expresses the scavenger receptor cysteine-rich (SRCR) gene G3876, annotated as an Low-density lipoprotein receptor-related protein with a Log2 fold change (Log2FC) of 0.77. This finding directly links small granule cells to scavenger receptor-mediated functions, supporting their role as professional phagocytes. Scavenger receptors, including SRCR proteins, are known for their ability to bind and internalize diverse ligands, including pathogens, and their presence in small granule cells highlights a potential mechanism for pathogen recognition and clearance.

Additionally, scavenger receptors are significantly expanded in oysters, as shown in Wang et al. (9). These receptors exhibit dynamic upregulation in hemocytes upon pathogen exposure, particularly following stimulation with pathogen-associated molecular patterns (PAMPs) such as lipopolysaccharide (LPS). This evidence suggests that SRCR proteins, including the one identified in our study, play a pivotal role in the phagocytic activities of hemocytes by facilitating pathogen recognition and internalization.

We propose to add this paragraph (lines 610 to 618) in the Discussion : “Interestingly, our scRNA-seq analysis indicates that SGC (cluster 3) expresses the scavenger receptor cysteine-rich (SRCR) gene G3876, annotated as an Low-density lipoprotein receptor-related protein with a Log2 fold change (Log2FC) of 0.77 linking them to scavenger receptor-mediated pathogen recognition and clearance. This aligns with findings by Wang et al. (9), who demonstrated significant expansion and dynamic regulation of SRCR genes in response to pathogen-associated molecular patterns. “

(12) I am not convinced by the added value of the lineage analysis and the manuscript could stand without it. There is no experimental validation to substantiate the filiation between the clusters. In addition, rooting the lineage to cluster 4 is poorly justified (enrichment in the ribosomal transcript). Cluster 6 is also enriched in ribosomal transcripts and this enrichment can be caused by the low threshold used for the selection of cluster-specific genes (L2FC >0.25). At last, cluster 4 > VC and cluster 4 >SGC belong to the same lineage according to Figure 7 FH.

We thank the reviewer for their detailed comments regarding the lineage analysis. We acknowledge the limitations in experimentally validating the proposed filiation between clusters, as hemocytes in Crassostrea gigas cannot currently be cultivated ex-vivo, and we lack the ability to isolate cells specifically from cluster 4 for further functional assays. Consequently, our lineage analysis is based solely on transcriptomic data and pseudo-time trajectory analysis.

Hematopoietic stem cells (HSCs) are a population of stem cells that are largely cell-cycle-quiescent (G0 phase) with low biosynthetic activity. Upon stimulation and stress HScs undergo proliferation and differentiation and produce all lineages of hemocytes.

Ribosomal proteins play a multifaceted role in preserving the balance between stem cell quiescence and activation. By ensuring precise regulation of protein synthesis, they allow stem cells to maintain their undifferentiated state while remaining poised for activation when needed. Furthermore, ribosomal proteins contribute to the cellular stress response, safeguarding stem cells from oxidative damage and other stressors that could compromise their functionality. Importantly, ribosomal biogenesis and the dynamic assembly of ribosomes provide a regulatory mechanism that fine-tunes the transition from self-renewal to differentiation, a critical feature of hematopoietic stem cells (HSCs) and other stem cell types. These mechanisms collectively highlight the indispensable role of ribosomal proteins in stem cell biology, underscoring their relevance to our study's findings.

In vertebrate, the maintenance of hematopoietic stem cells (HSCs) and hematopoietic homeostasis is widely acknowledged to rely on the proper regulation of ribosome function and protein synthesis (10). This process necessitates the coordinated expression of numerous genes, including genes that encode ribosomal proteins (RP genes) and those involved in regulating ribosome biogenesis and protein translation. Disruptions or mutations in these critical genes are associated with the development of congenital disorders (11). Among these, Rpl22 (found in cluster 4 with a Log2FC of 1.59) has been shown to play a pivotal role in HSC maintenance by balancing ribosomal protein paralog activity, which is critical for the emergence and function of HSCs (12).

Regarding the justification for rooting the lineage to cluster 4, our decision was informed by the enrichment of ribosomal transcripts and functional annotations suggesting a role in translation and cell proliferation, consistent with a precursor-like state. The use of a log2 fold-change (L2FC) threshold of >0.25, while conservative, allowed us to include subtle but meaningful transcriptional shifts essential for resolving lineage transitions.

Finally, the lineage progression from cluster 4 to vesicular cells (VC), macrophage-like cells (ML), and ultimately small granule cells (SGC) is supported by trajectory analysis (Figure 7FH), which consistently places VC and ML as intermediates in the differentiation process toward SGC. Although experimental validation is currently not feasible, these findings provide a conceptual framework for future investigations when cell isolation and functional validation tools become available.

(13) The figures containing heatmaps (Figure 7, Figure 2, Figure S10) or too many subpanels (Figure S5) and Table S5 are hardly readable.

Thank you for highlighting the issues related to the clarity of the heatmaps (Figures 2, 7, and S10), the multi-panel figure (Figure S5), and Table S5. In response to your feedback, we have revised all of these elements to enhance readability and comprehension. Specifically, we increased font sizes, optimized color scales, and reorganized the layout of the subpanels to emphasize the key findings. We also updated Table S5 to ensure that the data are presented in a clear and easily interpretable format.

We trust that these modifications address the concerns raised and improve the overall clarity of the figures and table.

(14) A number of single-cell analyses are now available in different species and the authors allude to similar pathways/transcription factors being involved. Perhaps the authors could expand on this in the discussion section.

Transcription factors involved in hematopoiesis, such as Tal1, Runx and GATA, are highly conserved across metazoans. Consistent with findings in other species, our dataset identifies these markers, reinforcing the evolutionary conservation of these pathways. Furthermore, these markers are also reported in the previous scRNA-seq dataset for C. hongkongensis (4), supporting the robustness of our molecular signatures. However, defining specific and robust markers for distinct hemocyte types remains an ambitious task, requiring additional validation in diverse biological and experimental contexts. This validation is beyond the scope of the present study.

In addition, meaningful comparisons between scRNA-seq datasets are constrained by differences in annotation frameworks and the absence of standardized definitions for hemocyte subtypes. Harmonizing these datasets to enable robust cross-species comparisons is a critical challenge for future studies. Nonetheless, the insights provided by our dataset establish a strong foundation for such comparative analyses when these standardization efforts are realized.

In crayfish (1), 16 transcriptomic clusters were identified corresponding to three hemocyte types, with markers such as integrin prominently expressed in hyalinocytes, consistent with our identification of integrin-related genes in hemocytes. In shrimp (1), 11 transcriptomic clusters were described, with markers of hemocytes in immune-activated states, that we observed also in our dataset. For Anopheles gambiae (2), 8 transcriptomic clusters were identified, including clusters with high ribosomal activity, analogous to those we described in our study. Finally, in Bombyx mori (3), 20 transcriptomic clusters were reported, corresponding to five cytological hemocyte types. Transcription factors such as bHLH, myc, and runt were identified in granulocytes and oenocytoid, showing parallels with markers identified in our dataset.

Despite these similarities, cross-species comparisons are hindered by variability in genome availability and annotation quality, which complicates the precise identification and functional characterization of genes across datasets. Notably, we did not detect pro-phenoloxidase genes in our dataset, unlike shrimp and crayfish, suggesting potential species-specific differences in immune mechanisms.

Regarding the previously published C. hongkongensis scRNA-seq dataset (4), we observe overlap in markers such as runx and GATA. However, direct comparisons remain limited due to differences in dataset annotations and definitions of hemocyte subtypes. This underscores the need for standardized frameworks to facilitate cross-study comparisons. While we emphasize that robust cross-species validation was beyond the scope of this study, our findings contribute valuable insights into the molecular signatures of oyster hemocytes and provide a framework for future comparative research.

We have expanded our discussion to include comparisons with available scRNAseq data from other invertebrate specie (lines 747 to 760)

Minor comments:

(1) Figure 2A-D: to increase the readability of the figure, the authors should display only the GO terms mentioned in the text and keep the full list in supplementary data.

To enhance the fluidity of the results section, we have redesigned the KEGG/RBGOA figure to present the results for each cluster in an integrated manner (See figure 2A and 2B).

(2) Line 223: the authors mention that cluster 1 is characterized by its morphology without providing an explanation or evidence.

We have revised the description of Cluster 1 to remove references to morphology, ensuring consistency with the data presented at this stage of the manuscript (lines 227 to 229) : ”Cluster 1, comprising 27.6 % of cells, is characterized by GO-terms related to myosin complex, lamellipodium, membrane and actin cytoskeleton remodelling, as well as phosphotransferase activity.”

(3) Line 306: the authors mentioned expression levels and associated them with Log2FC, which represents an enrichment, not the level of expression.

Thank you for pointing this out. We agree that log2FC represents enrichment rather than absolute expression levels. We have revised the text in the manuscript to clarify this distinction (line 309). The corrected text now states that log2FC reflects the degree of enrichment or depletion of a gene in a specific cluster relative to others, rather than its absolute expression level.

(4) Figure 4B: the figure shows the distribution of all hemocytes subgroups for each fraction. To better appreciate the distribution of the subgroups in the different fractions, it would be good to have the number of cells of each subtype in the fractions.

We thank the reviewer for their suggestion to include the number of cells of each subtype in the fractions. While we do not have the exact total number of cells per fraction, we systematically performed hemocyte counts for each fraction as part of our methodology. These counts provide a robust estimation of hemocyte distributions across fractions.

Including these counts in the figure could be an alternative approach; however, we believe it would not significantly enhance the interpretability of the data, as the focus of this analysis is on the relative proportions of hemocyte subtypes rather than absolute numbers. The current representation provides a clear and concise overview of subtype distribution patterns, which aligns with the goals of the study.

Nevertheless, if the reviewer considers it essential, we are open to integrating the hemocyte counts into the figure or supplementing the information in the text or supplementary materials to provide additional context.

(5) Line 487-488: the authors mentioned that monocle 3 can deduce the differentiation pathway from the mRNA splice variant. I did not find this information in the publication associated with the statement.

Thank you for pointing this out. We acknowledge the inaccuracy in our statement regarding Monocle3's capabilities. Monocle3 does not deduce differentiation pathways based on mRNA splice variants, as was erroneously suggested in the manuscript. Instead, Monocle3 performs trajectory inference using gene expression profiles. It calculates distances between cells based on their transcriptomic profiles, where cells with similar profiles are positioned closer together, and those with distinct profiles are farther apart. This method enables the construction of potential differentiation trajectories by identifying paths between transcriptionally related cells.

We revise the text in the manuscript to accurately describe this process and remove the incorrect reference to mRNA splice variants (lines 495 to 497).

(6) Figures 6C-H display heatmaps with two columns representing the beginning and the end of the lineage predicted. It would be more talkative to show the whole path presented in Figure S10.

Thank you for pointing out that Figures 7C–H currently only show the beginning and end of the predicted lineage, limiting the clarity of the intermediate stages. In response to your suggestion, we have revised these figures to include the full trajectory as presented in Figure S10, ensuring that the intermediate transitions are more clearly visualized. We believe these modifications offer a more comprehensive overview of the entire lineage and enhance the interpretability of our results.

Bibliography:

(1) F. Xin, X. Zhang, Hallmarks of crustacean immune hemocytes at single-cell resolution. Front. Immunol. 14 (2023).

(2) H. Kwon, M. Mohammed, O. Franzén, J. Ankarklev, R. C. Smith, Single-cell analysis of mosquito hemocytes identifies signatures of immune cell subtypes and cell differentiation. eLife 10, e66192 (2021).

(3) M. Feng, L. Swevers, J. Sun, Hemocyte Clusters Defined by scRNA-Seq in Bombyx mori: In Silico Analysis of Predicted Marker Genes and Implications for Potential Functional Roles. Front. Immunol. 13 (2022).

(4) J. Meng, G. Zhang, W.-X. Wang, Functional heterogeneity of immune defenses in molluscan oysters Crassostrea hongkongensis revealed by high-throughput single-cell transcriptome. Fish & Shellfish Immunology 120, 202–213 (2022).

(5) C. Peñaloza, A. P. Gutierrez, L. Eöry, S. Wang, X. Guo, A. L. Archibald, T. P. Bean, R. D. Houston, A chromosome-level genome assembly for the Pacific oyster Crassostrea gigas. GigaScience 10, giab020 (2021).

(6) R. D. Rosa, A. Santini, J. Fievet, P. Bulet, D. Destoumieux-Garzón, E. Bachère, Big Defensins, a Diverse Family of Antimicrobial Peptides That Follows Different Patterns of Expression in Hemocytes of the Oyster Crassostrea gigas. PLOS ONE 6, e25594 (2011).

(7) Y. Li, J. Sun, Y. Zhang, M. Wang, L. Wang, L. Song, CgRel involved in antibacterial immunity by regulating the production of CgIL17s and CgBigDef1 in the Pacific oyster Crassostrea gigas. Fish & Shellfish Immunology 97, 474–482 (2020).

(8) Evidence of a bactericidal permeability increasing protein in an invertebrate, the Crassostrea gigas Cg-BPI | PNAS. https://www.pnas.org/doi/abs/10.1073/pnas.0702281104.

(9) L. Wang, H. Zhang, M. Wang, Z. Zhou, W. Wang, R. Liu, M. Huang, C. Yang, L. Qiu, L. Song, The transcriptomic expression of pattern recognition receptors: Insight into molecular recognition of various invading pathogens in Oyster Crassostrea gigas. Developmental & Comparative Immunology 91, 1–7 (2019).

(10) R. A. J. Signer, J. A. Magee, A. Salic, S. J. Morrison, Haematopoietic stem cells require a highly regulated protein synthesis rate. Nature 509, 49–54 (2014).

(11) A. Narla, B. L. Ebert, Ribosomopathies: human disorders of ribosome dysfunction. Blood 115, 3196–3205 (2010).

(12) Y. Zhang, A.-C. E. Duc, S. Rao, X.-L. Sun, A. N. Bilbee, M. Rhodes, Q. Li, D. J. Kappes, J. Rhodes, D. L. Wiest, Control of Hematopoietic Stem Cell Emergence by Antagonistic Functions of Ribosomal Protein Paralogs. Developmental Cell 24, 411–425 (2013).

Reviewer #2 (Public review):

Summary:

This work provides a comprehensive understanding of cellular immunity in bivalves. To precisely describe the hemocytes of the oyster C. gigas, the authors morphologically characterized seven distinct cell groups, which they then correlated with single-cell RNA sequencing analysis, also resulting in seven transcriptional profiles. They employed multiple strategies to establish relationships between each morphotype and the scRNAseq profile. The authors correlated the presence of marker genes from each cluster identified in scRNAseq with hemolymph fractions enriched for different hemocyte morphotypes. This approach allowed them to correlate three of the seven cell types, namely hyalinocytes (H), small granule cells (SGC), and vesicular cells (VC). A macrophage-like (ML) cell type was correlated through the expression of macrophage-specific genes and its capacity to produce reactive oxygen species. Three other cell types correspond to blast-like cells, including an immature blast cell type from which distinct hematopoietic lineages originate to give rise to H, SGC, VC, and ML cells. Additionally, ML cells and SGCs demonstrated phagocytic properties, with SGCs also involved in metal homeostasis. On the other hand, H cells, nongranular cells, and blast cells expressed antimicrobial peptides. This study thus provides a complete landscape of oyster hemocytes with functional validation linked to immune activities. This resource will be valuable for studying the impact of bacterial or viral infections in oysters.

Strengths:

The main strength of this study lies in its comprehensive and integrative approach, combining single-cell RNA sequencing, cytological analysis, cell fractionation, and functional assays to provide a robust characterization of hemocyte populations in Crassostrea gigas.

(1) The innovative use of marker genes, quantifying their expression within specific cell fractions, allows for precise annotation of different cellular clusters, bridging the gap between morphological observations and transcriptional profiles.

(2) The study provides detailed insights into the immune functions of different hemocyte types, including the identification of professional phagocytes, ROS-producing cells, and cells expressing antimicrobial peptides.

(3) The identification and analysis of transcription factors specific to different hemocyte types and lineages offer crucial insights into cell fate determination and differentiation processes in oyster immune cells.

(4) The authors significantly advance the understanding of oyster immune cell diversity by identifying and characterizing seven distinct hemocyte transcriptomic clusters and morphotypes.

These strengths collectively make this study a significant contribution to the field of invertebrate immunology, providing a comprehensive framework for understanding oyster hemocyte diversity and function.

Weaknesses:

(1) The authors performed scRNAseq/lineage analysis and cytological analysis on oysters from two different sources. The methodology of the study raises concerns about the consistency of the sample and the variability of the results. The specific post-processing of hemocytes for scRNAseq, such as cell filtering, might also affect cell populations or gene expression profiles. It's unclear if the seven hemocyte types and their proportions were consistent across both samples. This inconsistency may affect the correlation between morphological and transcriptomic data.

We thank the reviewer for highlighting the importance of sample consistency and potential variability, and we acknowledge the need for clarification regarding the use of oysters from two different sources.

Oysters from La Tremblade (known pathogen-free in standardized conditions) were used to establish the hemocyte transcriptomic atlas through scRNA-seq and for cytological analyses. Oysters from the Thau Lagoon (Bouzigues) were used for cytological, functional, and fractionation experiments. These oysters were sampled during non-epidemic periods and monitored under Ifremer’s microbiological surveillance to ensure pathogen free status.

The cytological results (hemocytograms) presented in Figure 3 and Supplementary Figure S3 were derived from Thau Lagoon oysters. To clarify, we updated The Table 3 in Figure 3 and Supplementary Figure S3 to explicitly display hemocyte counts for oysters from both La Tremblade and Thau Lagoon. These data confirm consistent proportions of hemocyte types across both sources, with no significant differences (p > 0.05).

Hemocyte isolation and filtering protocols were rigorously optimized to preserve cell viability and morphology during scRNA-seq library preparation. Viability assays and cytological evaluations confirmed that these procedures did not significantly alter hemocyte populations or their proportions. Sample processing times were minimized to ensure that the scRNA-seq results accurately reflect the native state of the hemolymph.

Taken together, our results confirm that variability between oyster sources or methodological processes did not compromise our findings. This ensures that the correlations between morphological and transcriptomic data are reliable and robust.

(2) The authors claim to use pathogen-free adult oysters (lines 95 and 119), but no supporting data is provided. It's unclear if the oysters were tested for bacterial and viral contaminations, particularly Vibrio and OsHV-1 μVar herpesvirus.

The oysters used in this study were sourced from two distinct origins. First, the animals (18 months old) utilized for scRNA-seq and cytological analyses were obtained from the Ifremer controlled farm located in La Tremblade, France (GPS coordinates: 45.7981624714465, -1.150171788447683). This facility exclusively produces standardized oysters bred in controlled conditions with filtered seawater, entirely isolated from environmental known pathogens. The oysters from this source are certified “pathogen-free” upon arrival at the laboratory, following Ifremer's stringent quality control protocols. We have replaced the term 'pathogen-free' with 'known pathogen-free’ (line 123) to accurately reflect the animals' true status.

Second, for the fractionation experiments and functional tests, oysters were either sourced from the aforementioned Ifremer farm or from a producer located in the Thau Lagoon, France (GPS coordinates: 43.44265228308842, 3.6359883059292057). The Thau Lagoon is subject to comprehensive environmental and microbiological surveillance by the Ifremer monitoring network and the regional veterinary laboratory. For these experiments, we specifically selected oysters aged 18 months - an age associated with reduced susceptibility to OsHV-1 μVar herpesvirus - and ensured that sampling occurred outside of any detected epidemic periods. Furthermore, prior to experimentation, hemocyte samples from all oysters were examined. Oysters showing signs of contamination or exhibiting abnormal hemocyte profiles were excluded from the study.

These measures ensured that the oysters used in this work were of high health status and minimized the likelihood of bacterial or viral contamination, including Vibrio and OsHV-1 μVar.

(3) The KEGG and Gene Ontology analyses, while informative, are very descriptive and lack interpretation. The use of heatmaps with dendrograms for grouping cell clusters and GO terms is not discussed in the results, missing an opportunity to explore cell-type relationships. The changing order of cell clusters across panels B, C, and D in Figure 2 makes it challenging to correlate with panel A and to compare across different GO term categories. The dendrograms suggest proximity between certain clusters (e.g., 4 and 1) across different GO term types, implying similarity in cell processes, but this is not discussed. Grouping GO terms as in Figure 2A, rather than by dendrogram, might provide a clearer visualization of main pathways. Lastly, a more integrated discussion linking GO term and KEGG pathway analyses could offer a more comprehensive view of cell type characteristics. The presentation of scRNAseq results lacks depth in interpretation, particularly regarding the potential roles of different cell types based on their transcriptional profiles and marker genes. Additionally, some figures (2B, C, D, and 7C to H) suffer from information overload and small size, further hampering readability and interpretation.

Thank you for your valuable suggestions regarding the presentation and interpretation of our KEGG and Gene Ontology (GO) analyses. In response, we revised Figure 2 to enhance clarity and provide deeper insights into cell-type relationships and biological processes.

The revised figure 2 reorganizes GO term analysis into a more intuitive layout, grouping related biological processes and pathways in a structured manner. This approach replaces the dendrogram organization and provides a clearer visualization of key pathways for each cell cluster.

(4) The pseudotime analysis presented in the study provides modest additional information to what is already manifest from the clustering and UMAP visualization. The central and intermediate transcriptomic profile of cluster 4 relative to other clusters is apparent from the UMAP and the expression of shared marker genes across clusters (as shown in Figure 1D). The statement by the authors that 'the two types of professional phagocytes belong to the same granular cell lineage' (lines 594-596) should be formulated with more caution. While the pseudotime trajectory links macrophage-like (ML) and small granule-like (SGC) cells, this doesn't definitively establish a direct lineage relationship. Such trajectories can result from similarities in gene expression induced by factors other than lineage relationships, such as responses to environmental stimuli or cell cycle states. To conclusively establish this lineage relationship, additional experiments like cell lineage tracing would be necessary, if such tools are available for C. gigas.

We appreciate the reviewer’s detailed feedback on the pseudotime analysis and its interpretation. While we acknowledge that the clustering and UMAP visualization provide valuable insights, the pseudotime analysis offers a complementary approach by highlighting significantly expressed genes, including key transcription factors, that might otherwise be overlooked in differential expression analysis based solely on Log2FC between clusters. In our study, the pseudotime analysis revealed transcription factors known to play crucial roles in hemocyte differentiation, providing additional depth to our understanding of hemocyte lineage relationships and functional specialization.

Regarding the statement on lines 594 - 596, we agree that the evidence provided by pseudotime trajectories does not definitively establish a direct lineage relationship between macrophage-like (ML) and small granule-like (SGC) cells. Instead, these trajectories suggest potential developmental connections that warrant further investigation. We propose the following revised sentence (lines 616 to 618) :

"The pseudotime trajectory linking macrophage-like (ML) and small granule-like (SGC) cells suggests a potential developmental relationship within the granular cell lineage; however, this hypothesis requires further validation."

We also concur with the reviewer that additional experiments, such as cell lineage tracing, would be necessary to definitively establish this relationship. Unfortunately, the long-term cultivation of hemocytes in C. gigas is currently not feasible. However, we are planning to develop FACS-based approaches to separate the seven hemocyte subtypes, which will allow us to refine their ontology and explore their potential lineage relationships more precisely.

(6) Given the mention of herpesvirus as a major oyster pathogen, the lack of discussion on genes associated with antiviral immunity is a notable omission. While KEGG pathway analysis associated herpesvirus with cluster 1, the specific genes involved are not elaborated upon.

Thank you for your valuable observation regarding the lack of discussion on genes associated with antiviral immunity, particularly in the context of herpes virus infection. The KEGG pathway analysis indeed identified a weak signature associated with herpesvirus in Cluster 1, primarily involving genes encoding beta integrins. In humans, beta integrins have been described as receptors facilitating herpesvirus entry (1). However, in the case of naive oysters used in this study, the KEGG signature was subtle, likely reflecting the absence of active viral infection. Additionally, beta integrins are multifunctional molecules that also play critical roles in processes such as cell adhesion, a function attributed to hyalinocytes, as highlighted in our results.

Given the naive status of the oysters and the weak antiviral signature observed, we chose not to discuss these findings in detail in this study. However, ongoing work in our laboratory aims to further investigate the specific hemocyte populations targeted by OsHV-1, which may shed light on the role of integrins in antiviral immunity in oysters.

We hope this clarifies our approach and the context of the KEGG findings. Thank you for bringing this important perspective to our attention.

(7) The discussion misses an opportunity for comparative analysis with related species. Specifically, a comparison of gene markers and cell populations with Crassostrea hongkongensis, could highlight similarities and differences across systems.

In response to the reviewer’s comment, we have added a comparative analysis between C. hongkongensis and C. gigas hemocyte populations, situating our findings within the broader context of invertebrate immune cell diversity and specialization (lines 747 to 760)

Reviewer #2 (Recommendations for the authors):

(1) Lines 92-93: The authors should add references associated with transcriptomic studies of C. gigas hemocytes.

Thank you for pointing this out. In the revised manuscript, we have added references to previous transcriptomic studies of C. gigas hemocytes (line 83).

(2) Line 121 and 127: The authors should clarify whether 3,000 represents the number of cells loaded or their target for analysis.

The number of cells processed was optimized to minimize the occurrence of doublets during scRNAseq. Following 10x Genomics Chromium guidelines, we loaded 4,950 cells to successfully recover a target of 3,000 cells, with a doublet rate of 2.4%, well below the target threshold of 2.5%. This information has been added on line 125 of the document. The target was 3,000 cells, and as reported in Supplementary Table S1, the estimated number of cells after STAR-solo alignment was 2,937. This ensures the reliability and accuracy of single-cell transcriptomic data.

(3) Line 129: "Supp. Table 1" in the text and "Supp. Table S1" in the figure title should be edited.

The inconsistency between "Supp. Table 1" in the text and "Supp. Table S1" in the figure title has been corrected for uniformity throughout the manuscript (line 134).

(4) Line 138-139: The authors should clarify that the heatmap displays the top 10 positively enriched marker genes for each cluster, as identified by Seurat's differential expression analysis. It is important to note that the analysis does not explicitly show under-represented transcripts, but rather highlights the contrast between cluster-specific overexpressed genes and their lower expression in other clusters.

We have clarified that the heatmap displays the top 10 positively enriched marker genes for each cluster, as identified by Seurat's differential expression analysis, and that the analysis highlights cluster-specific overexpressed genes rather than explicitly showing under-represented transcripts (lines 143 - 145).

(5) Figure 1: The authors should consider improving or potentially removing Figure 1C. The gene IDs are not readable due to their small size, which significantly reduces the informative value of the figure. In addition, the data presented in this heatmap is largely redundant with the more informative and readable dot plot in Figure 1D, which shows both expression levels and the percentage of cells expressing each gene.

Thank you for your suggestion regarding Figure 1C. In the revised manuscript, we have removed the original panel C from the main figure and transferred it to Supplementary Figure S1K, which improves readability while retaining the relevant data. We have also renumbered the remaining panels for clarity, with the former panel D now designated as panel C. We believe these adjustments address the reviewer’s concerns and streamline the presentation of the data.

(6) Table 1: The authors should clarify in the legend the statistical significance criteria (adjusted p-value) for the genes listed.

As requested, we have added the adjusted p-value threshold (adj. p-value < 0.05) to the legend of Table 1.

(7) Line 188: The authors should align the text description of the KEGG pathways in cluster 7 with Figure 2A, describing Wnt signaling pathway and clarifying the terminology "endosome pathway" to ensure consistency.

In the revised text, we have aligned our description with Figure 2A by explicitly mentioning the Wnt signaling pathway in cluster 7 (lines 193 to 194).

The endo-lysosomal pathway encompasses a series of membrane-bound compartments and trafficking events responsible for the uptake of macromolecules from the extracellular environment, their subsequent sorting in endosomes, and eventual degradation in lysosomes. This pathway is tightly regulated, ensuring not only the breakdown of macromolecules but also the recycling of membrane components and signaling receptors essential for maintaining cellular homeostasis (2). In our study, the KEGG signatures of cluster 7 highlight the involvement of the endo-lysosomal pathway.

(8) Line 223: The authors should revise the description of cluster 1, avoiding references to morphology at this point in the manuscript, as no morphological data has been presented yet.

We have revised the description of Cluster 1 to remove references to morphology, ensuring consistency with the data presented at this stage of the manuscript (lines 227 to 229) : ”Cluster 1, comprising 27.6 % of cells, is characterized by GO-terms related to myosin complex, lamellipodium, membrane and actin cytoskeleton remodelling, as well as phosphotransferase activity.”

(9) Figure 2: The authors should revise Figure 2 to improve the clarity. For Figure 2A, they should address the redundancy in the "Global and overview maps" category by removing overlapping pathways such as carbon metabolism and biosynthesis of amino acids, which are likely represented in more specific metabolic categories (glycolysis, pentose). They could consider grouping similar pathways together, such as combining "Amino acid metabolism" with "Metabolism of other amino acids," and separating metabolic pathways from cellular processes for easier interpretation. They should also address the surprising absence of certain expected pathways like lipid metabolism, nucleotide metabolism, and cofactor/vitamin metabolism, as well as cellular processes such as cell growth and chromatin modeling. Even if these pathways are not enriched in specific clusters, mentioning their absence could provide valuable context for the reader.

In the revised version of the manuscript, we propose a new representation to facilitate understanding and improve the clarity of the data presentation.

(10) For Figures 2B, C, and D, the authors should significantly increase the font size of text and numbers, ensuring readability at 100% scale in PDF format. They could also add labels directly on each graph to clearly indicate the type of GO terms represented, (Biological Process, Cellular Component, or Molecular Function).

In the revised version of the manuscript, we propose a new representation to facilitate understanding and improve the clarity of the data presentation.

(11) Line 247-250: The authors should revise their description of cell types to follow the same order as presented in Figure 3A.

We have revised the description of cell types in the manuscript to follow the same order as presented in Figure 3A, as requested.

(12) Line 265-266: The authors should develop the significance of the nucleo-cytoplasmic ratio in hemocyte morphology and identification.

We thank the editor for bringing this to our attention and apologize for the discrepancy between the terminology used in the text and the results presented in Figure 3. The text refers to the nuclear-tocytoplasmic ratio (N/C), while the figure mistakenly displays the inverse ratio, cytoplasmic-to-nuclear ratio (C/N). We recognize that this inversion may cause confusion and will ensure consistency between the text and the figure.

To address this, we propose correcting the figure legend and labels in Figure 3 to align with the terminology used in the text (N/C ratio). This will prevent confusion and maintain clarity throughout the manuscript.

The nuclear-to-cytoplasmic (N:C) ratio, also known as the nucleus:cytoplasm ratio or N/C ratio, is a well-established measurement in cell biology that reflects the relative size of the nucleus to the cytoplasm. This ratio is frequently used as a morphologic feature in the diagnosis of atypia and malignancy in human cells, underscoring its diagnostic value. In the context of our study, we use the N:C ratio to provide a more precise and quantitative description of hemocyte types in Crassostrea gigas. Specifically, the N:C ratio allows us to distinguish between different hemocyte morphotypes, such as blasts and granular cells, and to enrich the characterization of their functional specialization. This quantitative measure supports the morphological classification and enhances the reproducibility and clarity of hemocyte identification.

(13) Line 286-294: The authors should review and correct the legend for Figure 3. It seems that the description of results related to Figure 3C has been mistakenly inserted into the legend.

We thank the reviewer for pointing out this issue with the legend of Figure 3. The description of results related to Figure 3C has now been removed from the legend. The revised legend focuses solely on the figure elements, improving clarity and consistency. We believe this adjustment addresses the reviewer's comment effectively.

(14) Figure 3: The authors should revise the legend for Figure 3A to provide more detailed and explicit descriptions of the "Size, shape and particularities" of the ML, SGC, BGC, and VC hemocyte types.

We thank the reviewer for their insightful suggestion to provide more explicit descriptions in the legend for Figure 3A. We have revised the legend to include detailed explanations of the "Size, shape, and particularities" for the ML, SGC, BGC, and VC hemocyte types. Specifically, we have clarified that size refers to the average granule diameter, shape describes the morphology of the granules (e.g., spherical or elongated), and particularities highlight distinguishing features such as granule color or fluorescence properties observed under specific staining or imaging conditions. We believe this updated legend provides the level of detail requested and enhances the clarity of the figure (lines 294 - 297).

(15) Figure 4: The authors should clarify the method used for calculating relative gene expression in Figure 4A and Figure 6. They should explicitly state in the figure legend that the expression was normalized to the Cg-rps6 reference gene, as mentioned in line 835. The authors should also provide details on the calculation method used (e.g., 2-ΔCt method) and confirm whether the reference gene was expressed at similar levels across all clusters.

We thank the reviewer for pointing out the need for additional clarity regarding the calculation of relative gene expression in Figures 4A and 6. To address this, we have revised the legends for both figures to explicitly state that gene expression levels were normalized to the reference gene Cg-rps6 and calculated using the 2^-ΔCt method. We have also confirmed that Cg-rps6 was stably expressed across all hemocyte clusters and explicitly mentioned this in the revised legends. These changes ensure greater transparency and address the reviewer’s concerns (lines 342 to 346).

(16) The authors could consider removing or modifying Figure 4B, as it appears to be redundant with Figure 3C. Both figures show the average percentage of each hemocyte type in the seven Percoll gradient fractions.

We thank the reviewer for highlighting potential redundancy between Figures 3C and 4B. While both figures present the distribution of hemocyte types across Percoll gradient fractions, Figure 4B serves a distinct and critical purpose in the manuscript. Specifically, it provides the numerical data necessary to understand the correlations shown in Figure 4A, where we analyze the relationship between gene expression levels and the distribution of hemocyte types. These detailed percentages are essential for interpreting the statistical robustness and biological relevance of the correlation matrix, which could not be derived solely from the qualitative visualization in Figure 3C.

(17) Figure 5: The authors should address the redundancy between Figure S7B and Figure 5B, as they appear to present the same data. In Figure S7B, "SGC" is incorrectly abbreviated as "G".

In the revised version of the manuscript, we addressed the redundancy between the two figures and we corrected the incorrectly abbreviated SGC.

(18) Line 412: The authors should correct the typographical error, changing "Pecoll" to "Percoll".

In the revised version of the manuscript, we correct this typographical error (line 417).

(19) Line 417: The statement about the inhibitor apocynin likely refers to Figure 5D, not Figure 5C.

In the revised version of the manuscript, we have corrected this reference error to accurately refer to Figure 5D (line 422).

(20) Line 441-444: The authors should provide references to support their annotation of cluster 1 as macrophage-like cells based on macrophage-specific genes. These references should cite established literature on known macrophage gene markers, particularly in bivalves or related species if available. They need to clarify whether specific gene markers exist for each of the hemocyte morphotypes they have identified. If such markers are known from previous studies, they should be mentioned and referenced.

We propose to modify lines 446 to 449 to address the reviewer's concerns. Cluster 1, which we have termed "macrophage-like" due to its pronounced phagocytic activity and reactive oxygen species (ROS) production, is enriched in Angiopoietin-1 receptor expression (Table 1). Angiopoietin receptors belong to the Tie receptor family, which is expressed in a subset of macrophages known as Tie2-expressing monocytes (TEMs) in humans (3–5). While our analysis reveals a strong overexpression of the Angiopoietin-1 receptor, we acknowledge that this receptor is not an exclusive marker for macrophages.

In bivalves, including oysters, no definitive molecular markers have been established for macrophagelike cells as they are defined functionally in this study. Consequently, the identification of such cells relies on their functional characteristics rather than strict marker expression. To clarify, we propose the following revision to the sentence:

Furthermore, this cluster expresses macrophage-related genes, including the macrophage-expressed gene 1 protein (G30226) (Supp. Data S1), along with maturation factors for dual oxidase, an enzyme involved in peroxide formation (Supp. Fig. S8), supporting its designation as macrophage-like based on functional characteristics.

(21) Figure 7: For Figures 7C to 7H, the authors should increase the font size of gene names and descriptions to ensure legibility in both printed versions and digital formats. To simplify these figures, the authors could consider displaying less differentially expressed genes for each lineage, along with the top genes for each differentiation pathway. If detailed gene information is crucial, they could move the full list to a supplementary table and reference it in the figure legend. Regarding Figure 7I, the authors should reorder the transcription factor genes by cluster and specificity to improve visualization and interpretation, like in Figure 1D.

Thank you for these valuable suggestions regarding Figure 7. We have revised Figures 7C–H to ensure improved readability. Furthermore, we have simplified these panels by highlighting fewer differentially expressed genes for each lineage. In Figure 7I, we have reordered the transcription factor genes by cluster and specificity, following a layout similar to Figure 1D, to facilitate clearer visualization and interpretation of the data.

(22) Line 490: The authors should provide more precise references to the specific GO terms and figure panels they are discussing.

To address this comment, we have revised the sentence and provided additional information in the text to clearly indicate where the corresponding figure panels can be found in the manuscript (line 499)

(23) Line 510: The authors state that "5 cell lineages could be defined," but the subsequent text and Figure 7C to H actually present 6 distinct lineages.

We have corrected in the manuscript. 6 lineages could be defined (line 521).

(24) Line 534: The authors should consider further investigating the pluripotent potential of cluster 4 cells by exploring known or potential stem cell markers in their scRNAseq data.

Thank you for highlighting the possibility of pluripotent potential of cluster 4. In our current analysis, we did not detect any known stem cell or proliferative markers, nor evidence of a clearly defined hematopoiesis site in the hemolymph. Indeed, previous work suggests that oyster hematopoiesis may occur in tissues such as the gills, implying that stem or progenitor cells might not circulate in the hemolymph under homeostatic conditions. Consequently, it is plausible that our observation of no proliferative cell populations partly reflects their absence in hemolymph, especially in naïve (unstimulated) oysters. To conclusively identify potential progenitor cells and their proliferative activity, further approaches involving deliberate perturbation of hemocyte homeostasis - such as immunological challenge (e.g., Zymosan treatment) combined with lineage-tracing or proliferation assays - would be necessary. These future investigations would not only clarify whether proliferative cells emerge in the hemolymph in response to environmental or pathological stimuli but also help elucidate the broader cellular pathways underlying oyster immune responses.

In response to the reviewer’s comment, we have revised the Discussion (lines 695 to 696) and added : “Nevertheless, we did not detect any canonical stem or progenitor cell populations in our dataset, underscoring the need for future investigations - potentially involving immunological challenges and lineage-tracing assays - to clarify whether proliferative cells circulate in the hemolymph or instead reside primarily in tissue compartments.”

(25) Figure S10: The authors should significantly improve the readability of Figure S10 by increasing the font size. Currently, the small font size makes it impossible for readers to discern the information presented.

Thank you for highlighting the readability concerns regarding Figure S10. In response to your comment, we have increased the overall size and font of the figure, ensuring that all labels and legends are clearly legible in both printed and digital formats. We believe these adjustments will allow readers to more easily interpret the information presented.

(26) Line 896: The authors should correct the typographical error on line 896 by deleting the additional bracket.

In the revised version of the manuscript, we correct this typographical error.

(27) Figure S12: The authors should address the absence of any reference to Figure S12 in the main text of the manuscript.

The reference to Supp. Figure S12 has been corrected. It was a referencing error between Supp. Figure S11(in the discussion, line 670) and Supp. Figure S12.

Bibliography:

(1) G. Campadelli-Fiume, D. Collins-McMillen, T. Gianni, A. D. Yurochko, Integrins as Herpesvirus Receptors and Mediators of the Host Signalosome. Annual Review of Virology 3, 215–236 (2016).

(2) J. P. Luzio, P. R. Pryor, N. A. Bright, Lysosomes: fusion and function. Nat Rev Mol Cell Biol 8, 622–632 (2007).

(3) A. S. Harney, E. N. Arwert, D. Entenberg, Y. Wang, P. Guo, B.-Z. Qian, M. H. Oktay, J. W. Pollard, J. G. Jones, J. S. Condeelis, Real-Time Imaging Reveals Local, Transient Vascular Permeability, and Tumor Cell Intravasation Stimulated by TIE2hi Macrophage-Derived VEGFA. Cancer Discov 5, 932–943 (2015).

(4) M. De Palma, R. Mazzieri, L. S. Politi, F. Pucci, E. Zonari, G. Sitia, S. Mazzoleni, D. Moi, M. A. Venneri, S. Indraccolo, A. Falini, L. G. Guidotti, R. Galli, L. Naldini, Tumor-targeted interferon-alpha delivery by Tie2-expressing monocytes inhibits tumor growth and metastasis. Cancer Cell 14, 299–311 (2008).

(5) M. De Palma, M. A. Venneri, R. Galli, L. Sergi Sergi, L. S. Politi, M. Sampaolesi, L. Naldini, Tie2 identifies a hematopoietic lineage of proangiogenic monocytes required for tumor vessel formation and a mesenchymal population of pericyte progenitors. Cancer Cell 8, 211–226 (2005).

Reviewer #3 (Public review):

The paper addresses pivotal questions concerning the multifaceted functions of oyster hemocytes by integrating single-cell RNA sequencing (scRNA-seq) data with analyses of cell morphology, transcriptional profiles, and immune functions. In addition to investigating granulocyte cells, the study delves into the potential roles of blast and hyalinocyte cells. A key discovery highlighted in this research is the identification of cell types engaged in antimicrobial activities, encompassing processes such as phagocytosis, intracellular copper accumulation, oxidative bursts, and antimicrobial peptide synthesis.

A particularly intriguing aspect of the study lies in the exploration of hemocyte lineages, warranting further investigation, such as employing scRNA-seq on embryos at various developmental stages.

In the opinion of this reviewer, the discussion should compare and contrast the transcriptome characteristics of hemocytes, particularly granule cells, across the three species of bivalves, aligning with the published scRNA-seq studies in this field to elucidate the uniformities and variances in bivalve hemocytes.

Reviewer #3 (Recommendations for the authors):

Minor Concerns:

(1) In the context of C. gigas, the notable expansion of stress and immune-related genes in its genome stands out. It is anticipated that the article will discuss the expression patterns of classical immune-related genes like TLR and RLR across different cell clusters.

We appreciate the reviewer's interest in the expression patterns of classical immune-related genes, such as Toll-like receptors (TLRs) and RIG-I-like receptors (RLRs), across different cell clusters in Crassostrea gigas. In our single-cell RNA sequencing (scRNA-seq) analysis, we did not detect significant expression of TLR or RLR genes. This absence can be attributed to several factors. First, technical limitations of scRNA-seq: The droplet-based scRNA-seq technology employed in our study captures only a fraction of the transcripts present in each cell approximately 10–20% (https://kb.10xgenomics.com/hc/en-us/articles/360001539051-What-fraction-of-mRNA-transcriptsare-captured-per-cell). This inherent limitation often results in the underrepresentation of genes with low expression levels. Consequently, TLRs and RLRs, which may be expressed at low levels in certain hemocytes, could be undetected due to this capture inefficiency. TLRs are typically expressed at low basal levels under resting conditions and are upregulated in response to specific stimuli or pathogenic challenges (1, 2). Given that our study analyzed hemocytes in their basal state, the expression levels of these receptors may have been below the detection threshold of the scRNA-seq platform. Furthermore, as highlighted by De Lorgeril et al. (3) the expression of these immune receptors varies depending on the resistance of the oyster. This variability further underscores the dynamic and context-dependent nature of TLR and RLR expression

To comprehensively assess the expression patterns of TLRs and RLRs across different hemocyte clusters, future studies could incorporate targeted enrichment strategies, such as bulk RNA-seq or single-cell technologies with higher capture efficiencies. Additionally, analyzing hemocytes under stimulated conditions or comparing oysters with varying levels of resistance could provide insights into the inducible and context-specific expression of these immune receptors.

(2) Clarification is needed in lines 265-266 regarding the nucleo-cytoplasmic ratio (N/C) terminology to prevent confusion, considering the discrepancy with the results presented in Figure 3.

We thank the editor for bringing this to our attention and apologize for the discrepancy between the terminology used in the text and the results presented in Figure 3. The text refers to the nuclear-tocytoplasmic ratio (N/C), while the figure mistakenly displays the inverse ratio, cytoplasmic-to-nuclear ratio (C/N). We recognize that this inversion may cause confusion and will ensure consistency between the text and the figure.

To address this, we propose correcting the figure legend and labels in Figure 3 to align with the terminology used in the text (N/C ratio). This will prevent confusion and maintain clarity throughout the manuscript.

(3) The selection of cluster 4 as the root for pseudotime analysis based on high ribosomal protein expression raises questions. It would be beneficial to elaborate on the inclusion of other genes, such as cell cycle or mitotic-related genes, to validate the pseudotime analysis outcomes.

We appreciate the reviewer’s insightful comment on the significance of ribosomal proteins in stem cell maintenance.

Hematopoietic stem cells (HSCs) are a population of stem cells that are largely cell-cycle-quiescent (G0 phase) with low biosynthetic activity. Upon stimulation and stress HScs undergo proliferation and differentiation and produce all lineages of hemocytes.

Ribosomal proteins play a multifaceted role in preserving the balance between stem cell quiescence and activation. By ensuring precise regulation of protein synthesis, they allow stem cells to maintain their undifferentiated state while remaining poised for activation when needed. Furthermore, ribosomal proteins contribute to the cellular stress response, safeguarding stem cells from oxidative damage and other stressors that could compromise their functionality. Importantly, ribosomal biogenesis and the dynamic assembly of ribosomes provide a regulatory mechanism that fine-tunes the transition from self-renewal to differentiation, a critical feature of hematopoietic stem cells (HSCs) and other stem cell types. These mechanisms collectively highlight the indispensable role of ribosomal proteins in stem cell biology, underscoring their relevance to our study's findings.

In vertebrate, the maintenance of hematopoietic stem cells (HSCs) and hematopoietic homeostasis is widely acknowledged to rely on the proper regulation of ribosome function and protein synthesis (4). This process necessitates the coordinated expression of numerous genes, including genes that encode ribosomal proteins (RP genes) and those involved in regulating ribosome biogenesis and protein translation. Disruptions or mutations in these critical genes are associated with the development of congenital disorders (5). Among these, Rpl22 (found in cluster 4 with a Log2FC of 1.59) has been shown to play a pivotal role in HSC maintenance by balancing ribosomal protein paralog activity, which is critical for the emergence and function of HSCs (6).

(4) What is the resolution of the cell clustering employed in the study? Given that cluster 1 potentially encompasses two distinct cell types, Macrophage-Like and Big Granule cells, further sub-clustering efforts and correlation analyses between cluster markers and cell morphologies could aid in their differentiation.

Thank you for your inquiry regarding the resolution of our cell clustering. As described in the Materials and Methods section, we used the Seurat FindClusters function with a resolution parameter of r = 0.1 for the scRNA-seq dataset. We performed sub-clustering within Cluster 1, resulting in four distinct subclusters. However, despite analyzing various specific markers, we did not identify any marker uniquely associated with the Big Granule Cell (BGC) morphology. Notably, LACC24 specifically marks a subset of cells within Cluster 1, as shown in Supplementary Figure S8, although this gene alone was insufficient to definitively distinguish a distinct BGC population.

(5) Line 78's statement regarding the primary identification of three hemocyte cell types in C. gigas-blast, hyalinocyte, and granulocyte cells would benefit from including references to substantiate this claim.

We thank Reviewer #1 for their valuable comments, which have allowed us to further improve our manuscript. We have enriched the introduction with the following addition (lines 79 to 82):

“Blast-like cells are considered undifferentiated hemocyte types (Donaghy et al., 2010), hyalinocytes appear to play a key role in wound repair (de la Ballina et al., 2020), and granulocytes are primarily involved in immune surveillance. Among these, granulocytes are regarded as the main immunocompetent hemocyte type (Wang et al., 2017).”

Conclusion:

The authors largely achieved their primary objective of providing a comprehensive characterization of oyster immune cells. They successfully integrated multiple approaches to identify and describe distinct hemocyte types. The correlation of these cell types with specific immune functions represents a significant advancement in understanding oyster immunity. However, certain aspects of their objectives have not been fully achieved. The lineage relationships proposed on the basis of pseudotime analysis, while interesting, require further experimental validation. The potential of antiviral defense mechanisms, an important aspect of oyster immunity, has not been discussed in depth.

This study is likely to have a significant impact on the field of invertebrate immunology, particularly in bivalve research. It provides a new standard for comprehensive immune cell characterization in invertebrates. The identification of specific markers for different hemocyte types will facilitate future research on oyster immunity. The proposed model of hemocyte lineages, while requiring further validation, offers a framework for studying hematopoiesis in bivalves.

Bibliography:

(1) J. Chen, J. Lin, F. Yu, Z. Zhong, Q. Liang, H. Pang, S. Wu, Transcriptome analysis reveals the function of TLR4-MyD88 pathway in immune response of Crassostrea hongkongensis against Vibrio Parahemolyticus. Aquaculture Reports 25, 101253 (2022).

(2) Y. Zhang, X. He, F. Yu, Z. Xiang, J. Li, K. L. Thorpe, Z. Yu, Characteristic and Functional Analysis of Toll-like Receptors (TLRs) in the lophotrocozoan, Crassostrea gigas, Reveals Ancient Origin of TLR-Mediated Innate Immunity. PLOS ONE 8, e76464 (2013).

(3) J. de Lorgeril, B. Petton, A. Lucasson, V. Perez, P.-L. Stenger, L. Dégremont, C. Montagnani, J.M. Escoubas, P. Haffner, J.-F. Allienne, M. Leroy, F. Lagarde, J. Vidal-Dupiol, Y. Gueguen, G.

Mitta, Differential basal expression of immune genes confers Crassostrea gigas resistance to Pacific oyster mortality syndrome. BMC Genomics 21, 63 (2020).

(4) R. A. J. Signer, J. A. Magee, A. Salic, S. J. Morrison, Haematopoietic stem cells require a highly regulated protein synthesis rate. Nature 509, 49–54 (2014).

(5) A. Narla, B. L. Ebert, Ribosomopathies: human disorders of ribosome dysfunction. Blood 115, 3196–3205 (2010).

(6) Y. Zhang, A.-C. E. Duc, S. Rao, X.-L. Sun, A. N. Bilbee, M. Rhodes, Q. Li, D. J. Kappes, J. Rhodes, D. L. Wiest, Control of Hematopoietic Stem Cell Emergence by Antagonistic Functions of Ribosomal Protein Paralogs. Developmental Cell 24, 411–425 (2013).

Reviewer #2 (Public review):

Summary:

This manuscript reconsiders the "general form" of Hamilton's rule, in which "benefit" and "cost" are defined as regression coefficients. It points out that there is no reason to insist on Hamilton's rule of the form -c+br>0, and that, in fact, arbitrarily many terms (i.e. higher-order regression coefficients) can be added to Hamilton's rule to reflect nonlinear interactions. Furthermore, it argues that insisting on a rule of the form -c+br>0 can result in conditions that are true but meaningless and that statistical considerations should be employed to determine which form of Hamilton's rule is meaningful for a given dataset or model.

Strengths:

The point is an important one. While it is not entirely novel-the idea of adding extra terms to Hamilton's rule has arisen sporadically (Queller 1985, 2011; Fletcher & Zwick 2006; van Veelen et al. 2017)--it is very useful to have a systematic treatment of this point. I think the manuscript can make an important contribution by helping to clarify a number of debates in the literature. I particularly appreciate the heterozygote advantage example in the SI.

Weaknesses:

Although the mathematical analysis is rigorously done and I largely agree with the conclusions, I feel there are some issues regarding terminology, some regarding the state of the field, and the practice of statistics that need to be clarified if the manuscript is truly to resolve the outstanding issues of the field. Otherwise, I worry that it will in some ways add to the confusion.

(1) The "generalized" Price equation: I agree that the equations labeled (PE.C) and (GPE.C) are different in a subtle yet meaningful way. But I do not see any way in which (GPE.C) is more general than (PE.C). That is, I cannot envision any circumstance in which (GPE.C) applies but (PE.C) does not. A term other than "generalized" should be used.

(2) Regression vs covariance forms of the Price equation

I think the author uses "generalized" in reference to what Price called the "regression form" of his equation. But to almost everyone in the field, the "Price Equation" refers to the covariance form. For this reason, it is very confusing when the manuscript refers to the regression form as simply "the Price Equation".

As an example, in the box on p. 15, the manuscript states "The Price equation can be generalized, in the sense that one can write a variety of Price-like equations for a variety of possible true models, that may have generated the data." But it is not the Price equation (covariance form) that is being generalized here. It is only the regression that Price used that is being generalized.

To be consistent with the field, I suggest the term "Price Equation" be used only to refer to the covariance form unless it is otherwise specified as in "regression form of the Price equation".

(3) Sample covariance: The author refers to the covariance in the Price equation as "sample covariance". This is not correct, since sample covariance has a denominator of N-1 rather than N (Bessel's correction). The correct term, when summing over an entire population, is "population covariance". Price (1972) was clear about this: "In this paper we will be concerned with population functions and make no use of sample functions". This point is elaborated on by Frank (2012), in the subsection "Interpretation of Covariance".

Of course, the difference is negligible when the population is large. However, the author applies the covariance formula to populations as small as N=2, for which the correction factor is significant.

The author objects to using the term "population covariance" (SI, pp. 8-9) on the grounds that it might be misleading if the covariance, regression coefficients, etc. are used for inference because in this case, what is being inferred is not a population statistic but an underlying relationship. However, I am not convinced that statistical inference is or should be the primary use of the Price equation (see next point). At any rate, avoiding potential confusion is not a sufficient reason to use incorrect terminology.

Relatedly, I suggest avoiding using E for the second term in the Price equation, since (as the ms points out), it is not the expectation of any random variable. It is a population mean. There is no reason not to use something like Avg or bar notation to indicate population mean. Price (1972) uses "ave" for average.

I should add, however, that the distinction between population statistics vs sample statistics goes away for regression coefficients (e.g. b, c, and r in Hamilton's rule) since in this case, Bessel's correction cancels out.

(4) Descriptive vs. inferential statistics

When discussing the statistical quantities in the Price Equation, the author appears to treat them all as inferential statistics. That is, he takes the position that the population data are all generated by some probabilistic model and that the goal of computing the statistical quantities in the Price Equation is to correctly infer this model.

It is worth pointing out that those who argue in favor of the Price Equation do not see it this way: "it is a mistake to assume that it must be the evolutionary theorist, writing out covariances, who is performing the equivalent of a statistical analysis." (Gardner, West, and Wild, 2011); "Neither data nor inferences are considered here" (Rousset 2015). From what I can tell, to the supporters of the Price equation and the regression form of Hamilton's rule, the statistical quantities involved are either population-level *descriptive* statistics (in an empirical context), or else are statistics of random variables (in a stochastic modeling context).

In short, the manuscript seems to argue that Price equation users are performing statistical inference incorrectly, whereas the users insist that they are not doing statistical inference at all.

The problem (and here I think the author would agree with me) arises when users of the Price equation go on to make predictive or causal claims that would require the kind of statistical analysis they claim not to be doing. Claims of the form "Hamilton's rule predicts.." or use of terms like "benefit" and "cost" suggest that one has inferred a predictive or causal relationship in the given data, while somehow bypassing the entire theory of statistical inference.

There is also a third way to use the Price equation which is entirely unobjectionable: as a way to express the relationship between individual-level fitness and population-level gene frequency change in a form that is convenient for further algebraic manipulation. I suspect that this is actually the most common use of the Price equation in practice.

For a paper that aims to clarify these thorny concepts in the literature, I think it is worth pointing out these different interpretations of statistical quantities in the Price equation (descriptive statistics vs inferential statistics vs algebraic manipulation). One can then critique the conclusions that are inappropriately drawn from the Price equation, which would require rigorous statistical inference to draw. Without these clarifications, supporters of the Price equation will again argue that this manuscript has misunderstood the purpose of the equation and that they never claimed to do inference in the first place.

(5) "True" models

Even if one accepts that the statistical quantities in the Price equation are inferential in nature, the author appears to go a step further by asserting that, even in empirical populations, there is a specific "true" model which it is our goal to infer. This assumption manifests at many points in the SI when the author refers to the "true model" or "true, underlying population structure" in the context of an empirical population.

I do not think it is necessary or appropriate, in empirical contexts, to posit the existence of a Platonic "true" model that is generating the data. Real populations are not governed by mathematical models. Moreover, the goal of statistical inference is not to determine the "true model" for given data but to say whether a given statistical model is justified based on this data. Fitting a linear model, for example, does not rule out the possibility there may be higher-order interactions - it just means we do not have a statistical basis to infer these higher-order interactions from the data (say, because their p-scores are insignificant), and so we leave them out.

What we can say is that if we apply the statistical model to data generated by a probabilistic model, and if these models match, then as the number of observations grows to infinity, the estimators in the statistical model converge to the parameters of the data-generating one. But this is a mathematical statement, not a statement about real-world populations.

A resolution I suggest to points 3, 4, and 5 above is:<br /> *A priori, the statistical quantities in the Price Equation are descriptive statistics, pertaining only to the specific population data given.<br /> *If one wishes to impute any predictive power, generalizability, or causal meaning to these statistics, all the standard considerations of inferential statistics apply. In particular, one must choose a statistical model that is justified based on the given data. In this case, one is not guaranteed to obtain the standard (linear) Hamilton's rule and may obtain any of an infinite family of rules.<br /> *If one uses a model that is not justified based on the given data, the results will still be correct for the given population data but will lack any meaning or generalizability beyond that.<br /> *In particular, if one considers data generated by a probabilistic model, and applies a statistical model that does not match the data-generating one, the results will be misleading, and will not generalize beyond the randomly generated realization one uses.

Of course, the author may propose a different resolution to points 3-5, but they should be resolved somehow. Otherwise, the terminology in the manuscript will be incorrect and the ms will not resolve confusion in the field.

Reviewer #2 (Public review):

In my previous review, I considered the contributions of the authors to be substantial because they have nearly doubled the number of genome sequences for chitons, and their newly sequenced genomes apparently are very well annotated. I would even extend these strengths now that I have had a chance to better review recent literature on marine animal genomes. Their contribution has helped make the chitons one of the best available marine taxa for comparative genomic studies. However, I still am unconvinced by the authors' claims to have demonstrated an unusually high rate of large-scale genome rearrangements across chitons. Their best argument seems to be comparisons drawn within a couple of similarly ancient bivalve lineages that were used to identify the conserved genomic regions in the first place, specifically the 20 molluscan linkage groups (MLGs). Perhaps it is safest to conclude that these MLGs are mostly conserved in conchiferans. Their main comparison with other molluscan classes is presented in tables 4 and 5 in the supplement, where they report a somewhat higher mean translocation rate for chitons (45.48) than for bivalves (41.10) or gastropods (41.87) but does this justify the implications of the title or the claims made in the Summary? I am not sure, partly because these summary tables are not made in a way that separates the gastropod or bivalve species listed into subtaxa separated by LCAs with estimated age, so the mean value across each class is not especially helpful. I still feel that the authors were not convincing in their arguments that chiton chromosomes have been subject to an unexpected history of rearrangement when contrasting quite ancient chitons lineages. This does not include impressive rearrangements documented for the likely geologically recent rearrangements seen within the genus, Acanthochitona, and separately within the subfamily Acanthopleurinae. These are quite impressive events that occurred recently within lineages of shallow-water chiton taxa, not deep still waters.

By the authors' estimates, some sequenced chiton genomes represent lineages that share a last common ancestor (LCA) as much as over 300 million years before present. This is like comparing a frog genome with a human genome. I suspect that the authors would agree that the pace of chiton genome rearrangements is not nearly as great as that observed for younger taxa such as mammals or particular insect orders known to have a history of genome shuffling. For example, according to Damas et al. (2022; https://www.pnas.org/doi/full/10.1073/pnas.2209139119) for comparisons within mammals, "94 inversions, 16 fissions, and 14 fusions that occurred over 53 My differentiated the therian from the descendent eutherian ancestor."

It is more interesting to me how the chiton genome rearrangements compare with other molluscan classes or for comparisons of other marine taxa genomes that share a similarly ancient LCA, but this is difficult to dig out of the authors' presentation. As far as I am aware, there are relatively few such comparisons of genome rearrangements available for marine animals. Attempting to do my own search for any comparison I could make, I noticed in that in a recent compilation of "high quality"* genomes (Martínez-Redondo 2024; https://doi.org/10.1093/gbe/evae235), this included genomes for 84 (mostly insect) arthropods, 67 vertebrates, 31 mollusks, 15 annelids, 12 nematodes, and 6 cnidarians, but the numbers drop off to 1-4 for many phyla, e.g., echinoderms. If there are really so few marine taxa available for comparison to the last 300 My of chiton genome rearrangements and fusions, then I would like to see a better presentation of the contrasts of the 20 molluscan linkage groups (MLGs) across molluscan classes. I found it very difficult to evaluate beyond the assertion that these are relatively conserved in two bivalve taxa. I remain unconvinced whether the amount of genome rearrangement observed by the authors for chitons is either especially rapid or slow. Certainly the genome browsers I have seen do not make it easy to compare, for example, marine gastropod or bivalve genomes (e.g., https://www.ncbi.nlm.nih.gov/cgv/9606 or https://genome.ucsc.edu/cgi-bin/hgGateway).

An unrelated topic that I also brought up in my earlier review is the ancestral reconstruction of molluscan chromosome numbers. The authors' explanation does nothing to justify how they ended up with an optimization of 20 for the ancestor of Mollusca. The outgroups included two annelids, Owenia [12 chromosomes] and Paraescarpia [14], plus the very distant chordate, Branchiostoma [19] (but the tunicate, Oikopleura has 6). Do the authors not understand that outgroups are critical for the optimization of character states at an ancestral node, with the most proximal outgroups being most important? How did they end up with an ancestral reconstruction of the chiton LCA with 16 chromosomes when there is no chiton with more than 13? Given the number of chromosomes in annelids, which is clearly the most proximal outgroups included with chromosome numbers available, it is more parsimonious to postulate that there was an increase in chromosome number for the conchiferan lineage. Related, they should have rooted that tree figure (Fig. 2) with the deuterostome, Branchiostoma, not a monophyletic grouping of all outgroups.

A recent study by Lewin et al. (2024; https://doi.org/10.1093/molbev/msae172) comparing annelid genomic rearrangements suggests to me that annelids probably have a more striking history of rearrangements than for chitons, but I found it difficult to evaluate. I do tend to agree with the overall conclusion of Lewin et al: "All animals with bilateral symmetry inherited a genome structure from their last common ancestor that has been highly conserved in some taxa but seemingly unconstrained in others." That is also my impression so far but the authors have done little to summarize what is known. One study that implies that at least deuterostomes have conserved elements of an ancestral chromosomal arrangement is provided by Lin et al. (2024; https://doi.org/10.1371/journal.pbio.3002661), who sequenced two genomes representing crown group hemichordates (LCA about 504 My).

Even if my general impression is wrong that the history of chiton genome rearrangement is not especially remarkable, or at least we still do not have a great idea of how rapid it is, I still think the authors could have done a better job of demonstrating their claims. This is important if they are going to make big claims about the pace of chiton chromosomal rearrangements. There is very little discussion of other similarly ancient marine animal taxa. I do not especially have a problem with excluding better known terrestrial mammalian or insect genomes as perhaps not a very relevant contrast, but am I supposed to be impressed with the comparisons made with bivalves and gastropods in Tables 4 and 5 of the Supplement? Where do the authors present a detailed comparison of how these estimates compare to conchiferans? Is this amount of genome rearrangement observed for chitons surprising for an extant taxon that has diversified for over 300 My? This is claimed in the title and summary of the manuscript as the take-home for the contribution, but I am left with the impression that there is too little attempt to justify that the pace across Polyplacophora (Neoloricata) is in any way remarkable. Apparently, there are few other lineages of marine taxa within which there are sequenced and well annotated genomes that can be compared, and this confounds the extent of conclusions that can be made.

* "high quality" genomes defined as follows by Martínez-Redondo (2024): "...we lowered the threshold used to consider a data set as high quality to 70% C + F (complete plus fragmented) BUSCO score (Manni et al. 2021), as the original 85% threshold was too restrictive when prioritizing a wide taxonomic sampling and the inclusion of biologically interesting species that are not widely studied."

Author response:

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

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

This paper provides a compelling analysis of chiton genomes, revealing extensive genomic rearrangements despite the group's apparent morphological stasis. By examining five reference-quality genomes, the study identifies 20 conserved molluscan linkage groups that are subject to significant rearrangements, fusions, and duplications in chitons, particularly in the basal Lepidopleurida clade. The high heterozygosity observed adds complexity to genome assembly but also highlights notable genetic diversity.

We also note the comment from this reviewer that “more information is needed to clarify how this affects genome assembly and evolutionary outcomes.” We strongly agree; although it is outside the scope of this study, this may help develop future work on that topic.

The research challenges the assumption that morphological stability implies genomic conservatism, suggesting that dynamic genome structures may play a role in species diversification. Although limited by the small number of molluscan genomes available for comparison, this study offers valuable insights into evolutionary processes and calls for further genomic exploration across molluscan clades. Some minor comments need to be tackled:

(1) Line 39: 'major changes'. Please, better explain what you mean here?

Clarified as major morphological change

(2) Lines 70-73: refer to 'extant' cephalopods.

Corrected

(3) There is an inconsistency in the use of "Callochitonida" (lines 76, 85, 140, 145, Table S3, Figure S3) and "Chitonida s.l." (Figures 2, 3, and 4) throughout the text, figures, and supplementary material. To maintain clarity and avoid confusion, I recommend choosing one taxon and using it consistently across all sections of the manuscript. This will ensure coherence and help readers follow the discussion without ambiguity.

An explanation has been added to the introduction and other instances in the text changed to Chitonida s.l. for consistency

(4) Overall, the conclusions introduce several important topics and additional information that were not addressed earlier in the paper. It would enhance the coherence and impact of the study to introduce these points in the introduction, as they highlight the broader significance and relevance of the research. Integrating these key aspects earlier on would better frame the study's objectives and provide readers with a clearer understanding of its importance from the outset.

The paragraph about chiton natural history and some additional lines have been moved to the introduction

(5) Lines 242-245 and 254-256: While I agree with the authors on the remarkable results found in molluscs, particularly in polyplacophorans, I suggest toning down the comparisons with lepidopterans. The current framing may come across as dismissive towards butterflies, which does not seem necessary. It's true that biases exist in studying taxa that are more charismatic due to factors like diversity or aesthetic appeal, but the goal should be to emphasize the value of polyplacophorans without downplaying the significance of butterfly research. Instead, the focus should be on highlighting chitons as an exciting new model for understanding key evolutionary processes like synteny, polyploidy, and genome evolution. This shift would underscore the importance of polyplacophorans in a positive light without diminishing the value of lepidopteran studies.

This sentence has been rephrased to adjust the tone of this paragraph

(6) Figure 3: should be read 'Polyplacophora'.

Corrected

Reviewer #2 (Recommendations for the authors):

I hope these comments by line number are helpful, despite my lack of experience with comparative genomics:

We note the general comment from this reviewer that “most chiton genomes seem to be relatively conserved” may be  a misunderstanding from our presentation; we have added some additional notes in the first part of the discussion to ensure that this is clear to all readers.

The reviewer also pointed out that “geologically recent events that do not especially represent the general pattern of genome evolution across this ancient molluscan taxon”. To clarify, the (limited) phylogenetic evidence suggests these changes are a longer term pattern throughout chiton evolution, since chromosomal rearrangements are found when comparing congeneric species (Acanthochitona spp., Fig 4C) and also across orders (Fig 4B). This has been added to the conclusions, as this is clearly an important point that was not adequately explained in the original text.

(1) Line 72: It is true that adaptive radiations occur and are an interesting general model for how diversification can lead to species-rich taxa. However, there are other "non-adaptive" processes that can lead to geographically isolated species that are not much differentiated in their ecological or morphological diversity. The sentence here implies that such adaptive radiation is a necessary correlation of species richness. I agree that chitons have hardly frozen in time since the Paleozoic.

This is clarified by moving some additional natural history aspects of chitons to the introduction, also as suggested by the first reviewer

(2) L113: I am curious about how this character optimization was accomplished to allow the authors to reconstruct the HAM (hypothetical ancestral mollusc) chromosome number as 20 when the range of variation in Polyplacophora is 6 to 16 (mode 11), and chitons are part of the sister taxon to conchiferans. Is this dependent on the chromosome numbers found in the outgroup?

We inferred ancestral linkage groups (“chromosomes”) based on comparison with other gastropods and bivalves noted in the methods; the other study cited (Simakov et al. 2022) used a broader selection of metazoans and also predicted an ancestral Mollusca karyotype of 1N=20.

(3) L116: "Using five chromosome-level genome assemblies for chitons, we reconstructed the ancestral karyotype for Polyplacophora (more strictly the taxonomic order Neoloricata), and all intermediate phylogenetic nodes to demonstrate the stepwise fusion and rearrangement of gene linkage groups during chiton evolution (Fig. 3)."

This is probably fine, but I had to struggle to understand what genome events happened between the Acanthochitona species. Are the chromosomes merely ordered and numbered by chromosome size and the switch in position between chromosomes 1 and 3 just has to do with the chromosomes 4+5, so they become the largest chromosome, and the former 1 is now 3? Confusing! The way it is drawn it seems like this implies more genome rearrangement than occurred, whereas if the order was maintained it would be more obvious that there were simply two chromosome fusions.

The linkage groups are numbered in order of size, which is the typical way they would each be presented if the taxon was illustrated alone. Here this allows the reader to understand how the fusions or rearrangements have shifted the volume of genetic information between groups especially in comparison to the molluscan or polyplacophoran ancestor. In Fig 4 we instead decided to present the linkage groups in a revised form, so that each transition from the nearest ancestor is visible in more detail. We have added these points in the figure caption for Fig 3 which should make it easier for new readers to understand the presentation.

(4) L481: Typo: A. rubrolineatain should be A. rubrolineata.

Corrected

(5) Figure 4: I am a little confused with what is meant by an "Ancestor" in these diagrams. For example, for comparing the two species of Acanthochitona with a hypothetical ancestor, it seems that the ancestor should be like one of the two, not different from both.

I am looking at Ancestor "3" compared with the Acanthochitona rubrolineata "3" and A. discrepans "4". Again, I assume that the latter is "4" because it is slightly smaller than a new "3" and now the new "3" corresponds to "1" in the other Acanthochitona. This figure does help interpret Figure 3.

To the point about reconstructing ancestral types; the two species both descended from a common ancestor. In morphology it is sometimes clear that one lineage retains more plesiomorphic character states; but in this case we must assume equal probability of change in any direction. The ancestor is a compromise that estimates the shortest distance to both descendants.

We understand how the numbers were unclear and potentially distracting. This has been added to the figure caption, we are grateful for the feedback that will certainly help future readers.

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

The study investigates protein-protein interactions (PPIs) within the nuage, a germline-specific organelle essential for piRNA biogenesis in Drosophila melanogaster, using AlphaFold2 to predict interactions among 20 nuage-localizing proteins. The authors identify five novel interaction candidates and experimentally validate three of them, including Spindle-E and Squash, through co-immunoprecipitation assays. They confirm the functional significance of these interactions by disrupting salt bridges at the Spn-E_Squ interface. The study further expands its scope to analyze approximately 430 oogenesis-related proteins, validating three additional interaction pairs. A comprehensive screen of around 12,000 Drosophila proteins for interactions with the key piRNA pathway player, Piwi, identifies 164 potential binding partners. Overall, the research demonstrates that in silico approaches using AlphaFold2 can link bioinformatics predictions with experimental validation, streamlining the identification of novel protein interactions and reducing the reliance on extensive experimental efforts. The manuscript is commendably clear and easy to follow; however, areas for improvement should be addressed to enhance its clarity and rigor.

Major Concerns:

(1) While AlphaFold2 was developed and trained primarily for predicting protein structures and their interactions, applying it to predict protein-protein interactions is an extrapolation of its intended use. This introduces several important considerations and risks. First, it assumes that AlphaFold's accuracy in structure prediction extends to interactions, despite not being explicitly trained for this task. Additionally, the assumption that high-scoring models with structural complementarity imply biologically relevant interactions is not always valid. Experimental validation is essential to address these uncertainties, as over-reliance on computational predictions without such validation can lead to false positives and inaccurate conclusions. The authors should expand on the assumptions, limitations, and risks associated with using AlphaFold2 for predicting protein-protein interactions.

We appreciate the reviewer's point. The prediction of protein-protein interactions using AlphaFold2 relies on the number of conserved homologous sequences and previous conformational data(8) (Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021)). We added sentences explaining the limitations and risks of the AlphaFold2 prediction method in Introduction and the end of Result and Discussion of the revised manuscript, respectively.

Page 5, Line 67;

“AlphaFold2 requires sequence homology information to predict protein-protein interactions and the complex structure model. The reliability of these predictions is basically dependent on the strength of co-evolutionary signals(9).”

Page 6, Line 84;

“AlphaFold2 was initially trained to predict the structure of individual proteins(8). Its application to complex prediction is an extrapolative use beyond its original intended scope, and its accuracy remains unverified. Even high-confidence predictions may not correspond to actual interactions, necessitating experimental validation to confirm whether predicted protein dimers truly bind.”

Page 21, Line 361;

“This study identifies several potential protein interactions, but AlphaFold2 predictions require caution. Protein-protein interactions involve conformational changes and dependencies on ligands, ions, and cofactors, which AlphaFold2 does not consider, potentially reducing prediction accuracy. Notably, the presence of a high-scoring model in terms of structural complementarity does not guarantee that the interaction is biologically significant.”

(2) The authors experimentally validated three interactions, out of five predicted interactions, using co-immunoprecipitation (co-IP). They attributed the lack of validation for the other two predictions to the limitations of the co-IP method. However, further clarification on the potential limitations of the co-immunoprecipitation behind the negative results would strengthen the conclusions. While co-IP is a widely used technique, it may not detect weak or transient interactions, which could explain the failure to validate some predictions. Suggesting alternative validation methods such as FRET or mass spectrometry could further substantiate the results. On the other hand, AlphaFold2 predictions are not infallible and may generate false positives, particularly when dealing with structurally plausible but biologically irrelevant interactions. By acknowledging both the potential limitations of co-IP and the possibility of false positives from AlphaFold2, the authors can provide a more balanced interpretation of their findings.

We appreciate the reviewer's point of view. We have used the co-IP method to detect interactions in this study. However, as the reviewer pointed out, it is likely that weak and transient interactions may not be detected. We added a note on the detection limits of the co-IP method and the possibility that AlphaFold2 method produces false positives in the revised manuscript.

Page 12, Line 197;

“While co-immunoprecipitation is a widely used method, it may not always detect weak or transient interactions. Other validation methods, such as FRET or co-localization assay in culture cells, could offer further insights to support the results. It is also important to note that AlphaFold2's predictions are not definitive and may lead to false positives, particularly when analyzing a large number of interactions.”

(3) In line 143, the authors state that "This approach identified 13 pairs; seven of these were already known to form complexes, confirming the effectiveness of AlphaFold2 in predicting complex formations (Table 2). The highest pcScore pair was the Zuc homodimer, possibly because AlphaFold2 had learned from Zuc homodimer's crystal structure registered in the database." While the authors mentioned the presence of the Zuc homodimer's crystal structure, they do not provide a systematic bioinformatics analysis to evaluate pairwise sequence identity or check for the presence of existing structures for all the proteins or protein pairs (or their homologs) in databases such as the Protein Data Bank (PDB) or Swiss-Model. Conducting such an analysis is critical, as it significantly impacts the novelty and reliability of AlphaFold2 predictions. For instance, high sequence identity between the query proteins could lead to high-scoring models for biologically irrelevant interactions. Including this information would strengthen the conclusions regarding the accuracy and utility of the predictions.

We appreciate the reviewer's critical point. The AlphaFold2 method generates a high confidence score when the 3D structure of the protein of interest, or of proteins with very similar sequences, is solved. We investigated whether the proteins used in this study are included in the 3D structure database (PDB) and added the information as a supplemental table S2. The following sentences were added to explain the structural references that AlphaFold2 has learned in the revised manuscript.

Page 9, Line 150;

The structures of the 20 proteins used in this study have been analyzed to varying extents in previous studies (Supplementary Table S2). A complex of Vas and the Lotus domain of Osk has been reported(20), and based on this complex structure, the interaction between Vas and Tej Lotus domain was predicted with a high score. Although the conformational analyses of the RNA helicase domain and the eTud domain have been reported previously, many of those cover only a subset of the regions and unlikely to affect our predictions in this study.

The predicted 3D structures and the Predicted Aligned Error (PAE) plots for the 12 pairs, are shown in Fig. 1C.

(4) While the manuscript successfully identifies novel protein interactions, the broader biological significance of these interactions remains underexplored. The manuscript could benefit from elaborating on how these findings may contribute to understanding the piRNA pathway and its implications on germline development, transposon repression, and oogenesis.

We added to the revise manuscript the potential biological significance of the novel protein-protein interactions presented in this manuscript as follows;

Page 16, Line 268;

“In this study, three novel protein-protein interactions were predicted and experimentally confirmed. AlphaFold2 also predicted the 3D structure of these complexes, providing insight into the important regions involved in complex formation. These predictions will provide fundamental information to elucidate nuage assembly. Nuage is thought to form by liquid-phase separation; however, direct protein-protein interactions likely occur within protein-dense nuage, facilitating RNA processing. Although the precise roles of individual interactions require further study, characterization of protein-protein interactions within nuage will help clarify the mechanism of piRNA production.”

Reviewer #1 (Recommendations for the authors):

Minor Concerns:

(1) In the Materials and Methods section, the authors thoroughly describe the computational infrastructure (SQUID at Osaka University) and the use of AlphaFold2. However, it would greatly benefit the readers to include a detailed breakdown of the computational cost. Understanding the computational cost (in terms of time, CPU/GPU hours, or other relevant metrics) for predicting 3D structures, especially for 400 protein pairs, would provide valuable insight into the efficiency and scalability of the approach. This would enhance the practical relevance of the methodology section and offer a better understanding of the resources required, beyond just the infrastructure description.

Thank you for your valuable suggestion. The following descriptions were added in the revised manuscript.

Page 24, Line 403;

“The calculation of the MSA took on average 2-4 hours per protein, with the more homologs of the protein in query, the longer it took.”

Page 24, Line 409;

“Prediction of dimer structure took approximately 1-2 hours per pair on average, depending on protein size. Each user can compute 100~200 pairs of calculations per day, but since the supercomputer is shared, job availability varies with overall demand.”

(2) The manuscript will benefit from a review for grammatical accuracy and clarity, especially in complex explanations. For example, in Line 160: "The predicted dimer structures of Me31B_Tral and Cup_Me31B showed the score of 0.74 and 0.68, respectively (Table 2)." could be revised to "The predicted dimer structures of Me31B_Tral and Cup_Me31B showed scores of 0.74 and 0.68, respectively.

Thank you very much for pointing it out. Correction has been made to the text pointed out (Page 10, Line 170).

(3) For alphafold3 webserver, please use (https://alphafoldserver.com/) instead of (https://golgi.sandbox.google.com/about).

Thank you very much for pointing it out. The URL has been changed in the revised manuscript (Page 25, Line 422).

Reviewer #2 (Public review):

Summary:

In this paper, the authors use AlphaFold2 to identify potential binding partners of nuage localizing proteins.

Strengths:

The main strength of the paper is that the authors experimentally verify a subset of the predicted interactions.

Many studies have been performed to predict protein-protein interactions in various subsets of proteins. The interesting story here is that the authors (i) focus on an organelle that contains quite some intrinsically disordered proteins and (ii) experimentally verify some (but not all) predictions.

Weaknesses:

Identification of pairwise interactions is only a first step towards understanding complex interactions. It is pretty clear from the predictions that some (but certainly not all) of the pairs could be used to build larger complexes. AlphaFold easily handles proteins up to 4-5000 residues, so this should be possible. I suggest that the authors do this to provide more biological insights.

We thank the reviewer for his kind suggestions. In this study, protein dimers were screened on the assumption that the two proteins bind 1:1; in some cases, multiple binding partners were predicted for a single protein. For example, Spn-E was predicted to bind Tej and Squ, respectively. Therefore, for Spn-E_Squ_Tej, we used the latest AlphaFold3 to predict the trimeric structure, which has already been described in the first manuscript. In addition, as suggested by the reviewer, other possible trimer results were also added in the revised manuscript as follows;

Page 15, Line 249;

“In addition to the Spn-E_Squ_Tej complex, 1:1 dimer prediction described above further suggested potential trimers (Fig. 1; Supplemental Fig. S4). For example, Tej protein is predicted to bind both Vas and Spn-E, and AlfaFold3 indeed further predicted a Vas_Tej_Spn-E trimer, where Tej’s Lotus and eTud domains interact with Vas and Spn-E, respectively. However, Lin et al. reported that Tej binds exclusively either with Vas or Spn-E, but not simultaneously(17), in Drosophila ovary, suggesting that the predicted trimers may be weak or transient. Similarly, the BoYb_Vret_Shu and the Me31B_Cup_Tral trimers remain hypothetical and require experimental verification (Supplemental Fig. S4).”

Another weakness is the use of a non-standard name for "ranking confidence" - the author calls it the pcScore - while the name used in AlphaFold (and many other publications) is ranking confidence.

“pcScore” has been changed to “ranking confidence”

Reviewer #2 (Recommendations for the authors):

(1) The pcScore is actually what is called RankingConfidence. Also, many other measures have been developed by other groups (based on PAE for instance) - these could be compared.

Thank you for your valuable suggestions. While other indicators are being developed, we have computed the affinity of the complex based on the predicted three-dimensional structure by using PRODIGY web server. The description was added in the revised manuscript as follows;

Page 18, Line 300;

“The ranking confidence score reflects the reliability of AlphaFold2's predicted structure but does not always ensure accuracy. Therefore, we assessed complex affinity based on the predicted three-dimensional structures (Supplemental Table S6). Most dimers with high ranking confidence scores exhibited low Kd values indicative of high affinity, while some showed high Kd values indicating weak interactions (Supplemental Table S6). For example, the Baf_Vas complex had a high AlphaFold2 ranking confidence score (0.85) but a relatively high Kd value (1.1E-4 M), indicating low affinity. Consistently, Baf_Vas binding was not detected in Co-IP experiments (Fig. S5C). Although accurate Kd prediction may be limited due to insufficient structural optimization, it could serve as a valuable secondary screening tool following AlphaFold2 predictions.”

(2) A statistical estimate of FDR for binding to the PIWI protein needs to be estimated. It is possible that 1.6% of random proteins (from another species for instance) also obtain ranking confidence over 0.6, i.e. how trustful are the predictions?

Thank you for the insightful comments. Unfortunately, it is difficult to infer the FDR from the value of ranking confidence. Presumably, the accuracy will vary depending on the target protein, since the number of homologs and known conformational information will differ. In the case of Piwi, the FDR is expected to be relatively low since the conformation of the protein on its own has been experimentally determined. However, even for Piwi complexes with high values of ranking confidence, the estimated affinity varied from high to low (Supplemental Table S6). Therefore, it may be useful to conduct further secondary evaluation for AlphaFold2 predictions with high ranking confidence.

(3) Identification of pairwise interactions is only a first step towards understanding complex interactions. It is pretty clear from the predictions that some (but certainly not all) of the pairs could be used to build larger complexes. AlphaFold easily handles proteins up to 4-5000 residues, so this should be possible. I suggest that the authors do this to provide more biological insights.

Already mentioned above.

(4) The comparisons of ranking confidence vs ipTM/pTM are less interesting (by definition ranking confidence is virtually identical to ipTM).

Thank you for the thoughtful comment. As the reviewer pointed out, there is not much difference between ranking confidence and ipTM shown in Fig. 1A. A high value of pTM (firmly folding) tends to increase ranking confidence, while a low value of pTM (many disorder regions) tends to decrease ranking confidence. Therefore, it may be useful to change the threshold for confidence for each protein pair.

git status 能展示 general 的更改信息，能看到哪个文件被修改了，在此之上 git diff 能展示更多的信息。通常来说，git diff 能够回答： - 更改了但是没有 staged 的内容 - staged 但是没有提交的内容

案例：起始状态是，当前有一个更改并 staged 的文件，一个只更改没有 staged 的文件

《图1》

如果想要查看没有 staged 的更改，直接使用 git diff（比较的是工作目录和 staging area）

图2

如果想要知道 staged 了什么可以提交，使用 git diff --staged（比较的是 staged files 和上一次提交）

图3

重要的是记住，git diff 本身比较的是工作目录和暂存区，如果所有更改都 staged 了，git diff 也就没有啥输出了。

如果更改了文件的一部分，staged 了，然后再更改一部分，就会看到这样的输出：

图4

然后使用 git diff 查看没有 staged 的部分，使用 git diff --cached (等同于 --staged) 查看 staged 的那部分。

Author response:

We thank the reviewers for the detailed evaluations and thoughtful comments, which have improved the clarity and readability of this manuscript. We have responded to all reviewer comments and incorporated their suggested changes into the text and figures. We have also included new experimental results suggested by reviewer 2, which further strengthen our main conclusion.

Point-by-point description of the revisions

Reviewer #1:

(1) Introduction, page 3: The statement "Single dimeric kinesin moves processively along microtubules in a hand-over-hand manner by alternately moving the two heads in an 8-nm step toward the plus-end of the microtubule" is inaccurate. The kinesin heads take ~16 nm steps, while the center of mass advances in ~8 nm increments. Please adjust the wording accordingly.

(2) Introduction, page 5: In the sentence "These results are consistent with the closed and open conformations of the nucleotide-binding pocket in the rear and front heads of microtubule-bound kinesin dimers observed in cryo-electron microscopy (cryo-EM) studies," I recommend changing the order to align with the previous sentence. The correct order would be "These results are consistent with the open and closed conformations of the nucleotide-binding pocket in the front and rear heads."

We thank the reviewer for pointing out our misunderstandings. We have corrected these sentences accordingly (lines 45-47 and lines 111-112).

Reviewer #2:

MAJOR CONCERNS

Limitations of this study: The authors need to discuss the limitations of their work. 1) They used a cys-lite kinesins mutant and introduced new surface-exposed cysteines. These mutants have lower kcat values than WT. 2) They used fluorescently labeled ATP molecules, which are hydrolyzed 10 times slower than unlabeled nucleotides. 3) They still observe crosslinking under reducing conditions and partial (but almost complete) crosslinking under oxidized conditions. 4)They assumed that cysteine crosslinked orientation mimics the orientation of the neck-linker in the front and rear conditions. The authors clearly pointed to these issues in the Results section. While these assumptions are also supported by several control experiments, the authors need to acknowledge some of these limitations in the Discussion as well.

We have now reiterated some of the key caveats in the Discussion, and newly described in the Results section those points not mentioned in the original manuscript that do not affect the conclusion. We also added a summary of the limitations and caveats into the first paragraph of the Discussion section (lines 425-431).

(1) We added a sentence in the Results section to describe that the ATP-binding kinetics of the Cys-light mutant remained consistent with previous studies as follows: “First, we demonstrated that k₊₁ and k_-1 of the wild-type head without Cys-modification were unchanged after oxidization (Table 1) and were comparable to those previously reported (Cross, 2004)” (lines 163-166). The reduced kcat values of cysteine pair-added mutants before crosslinking were primarily due to reduced microtubule association rate (data not included in this manuscript). We have added a sentence in the Results section describing the kcat results as follows: “The reduced ATPase activity primarily results from a decreased microtubule association rate (data to be presented elsewhere) with little change in ATP binding or microtubule dissociation rates (Table 1).” (lines 144-146).

(2) Fluorescently-labeled ATP was used to determine the ATP off-rates of the E236A mutant monomer and E236A rear head of the E236A/WT heterodimer. Two caveats in these measurements could lead to underestimating the ATP off-rate: 1) The off rate of Alexa-ATP from the head may be reduced compared to unmodified ATP, as Alexa-ATP driven motility showed a 10-fold reduce velocity. 2) The ATP off-rate of the E236A mutant may differ from that of the rear head in the wild-type dimer, since the E236A mutant likely stabilizes the neck linker-docked state more strongly than in the rear head of the wild-type dimer. These points are crucial for evaluating the results of ATP off-rate and the affinity for ATP, so we have added sentences in the Discussion section as follows: “We note, however, that this K_d of ATP may somewhat underestimate the true value in wild-type kinesin for two reasons: first, the E236A mutation likely stabilizes the neck linker-docked, closed state more than in the rear head of the wild-type dimer (Rice et al., 1999), and second, the Alexa-ATP used to measure the ATP off-rate of E236A head showed ~10-fold smaller velocity compared to unmodified ATP, partly due to a slower ATP off-rate (Figure 2-figure supplement 3).” (lines 449-454).

(3) Under reducing condition, the rear head crosslink contained 30% crosslinked species, while under oxidized condition, the front head crosslink contained 11% un-crosslinked species (Figure 1-figure supplement 1). These heterogeneities likely affect the rate constants of K_-1 for rear head crosslink and K₂ for front head crosslink, as crosslinked and un-crosslinked species showed significantly different rate constants. However, we did not use the rear head crosslink result to determine K_-1, since ATP hydrolysis likely occurred before reversible ATP dissociation. Instead, we used E236A monomer to estimate the K_-1 of the rear head. In addition, the result for K₂ of the front head crosslink was further validated using the E236A/WT heterodimer, which will be described in the next section.

(4) This is an important point, and therefore, we conducted experiments using the E236A/WT heterodimer (including new experimental results of ATP binding kinetics of the front head) and obtained consistent results. To address this point, we have revised the following sentences in the Discussion: “In the front head, backward orientation of the neck linker has little effect on ATP binding and dissociation rates, both when measured for a monomer crosslink (Figure 2A, B) and for the front head of a E236A-WT heterodimer (Figure 4B, C, F).” (lines 432-433); “However, we found that the ATP-induced detachment rates from microtubule (K₂) were similarly reduced for both the front head crosslink (7.0 s^-1; Figure 3A) and the front WT head of the E236A/WT heterodimer (6.3 s^-1; Figures 6D), suggesting that a step subsequent to ATP binding is gated in the front head.” (lines 437-441).

Line 238, the authors wrote that "forward constraint on the neck linker in the rear head does not significantly accelerate the detachment from the microtubule." Can the authors comment on why the read-head-like construct has a low affinity for microtubules even in the absence of ATP (Line 220)? I believe that the low affinity of the head in this conformation is more striking (and potentially more important) than the changes they observe in detachment rates. The authors should also consider that they might not be able to reliably measure the changes in the dissociation rate in single molecule assays of this construct (especially if the release rate of the rear head in the oxidized condition increases a lot higher than that of WT). The kymographs show infrequent and brief events, which raises doubts about how reliably they can measure the release rates under those imaging conditions. Higher motor concentrations and faster imaging rates may address this concern.

The low microtubule affinity of the rear-head-like crosslink stems from an extremely slow ADP release rate upon microtubule binding, not from a fast microtubule-detachment rate. Using stopped-flow measurements of microtubule-binding kinetics (microtubule-stimulated mant-ADP release and microtubule association rates), we found that the rear-head-crosslink resulted in a 2,000-fold decrease in the microtubule-stimulated ADP-release rate. This finding also explains the reduced ATPase of the rear-head-crosslink (Figure 1E). Since this low microtubule-affinity state occurs in the ADP-bound state rather than the ATP-bound state, we hypothesized that the neck-linker docked ADP-bound state cannot effectively bind to microtubules, requiring neck-linker undocking for microtubule binding (Mattson-Hoss et al., Proc. Natl. Acad. Sci., 111, 7000-7005 (2014)). While we acknowledge that understanding slow microtubule binding in the neck linker docked state is important for elucidating the mechanism and regulation of microtubule-binding of the head, this paper focuses specifically on the mechanism and regulation of “microtubule-detachment”. We plan to present these microtubule-binding kinetics data in a separate manuscript currently in preparation.

To explain the low microtubule affinity of the rear-head-crosslink, we added this explanation to the text; “because this constraint on the neck linker dramatically reduces the microtubule-activated ADP release rate (data to be presented elsewhere), creating a weak microtubule binding state” (lines 226-228).

Although the rear head crosslinking construct under oxidative condition showed fewer fluorescent spots per kymographs (images) due to its low microtubule binding rate, we collected more than one hundred spots by recording additional microscope movies (N=140; Figure 3-figure supplement 2B), ensuring sufficient data for statistical analysis.

Figure 2: How do the rates shown in Figure 2A-B compare to the previous kinetics studies in the field? The authors compare the dissociation rate of WT measured in rapid mixing experiments to that of E236A in smFRET assays. It is not clear whether these comparisons can be made reliably using different assays. Can the authors perform rapid mixing of E236A or try to determine the rate for the WT from smFRET trajectories?

The results of ATP on/off rates are comparable to the previous stopped flow measurements of ATP binding to monomeric kinesin-1 on microtubule, which are 2-5 µM^-1s^-1 and ~150 s^-1, respectively (summarized in the review by Cross (2004)). We added a sentence as follows: “First, we demonstrated that K₊₁ and K_-1 of the wild-type head without Cys-modification were unchanged after oxidization (Table 1) and were comparable to those previously reported (Cross, 2004).” (lines 163-166).

As the reviewer pointed out, the rapid mixing and smFRET data cannot be directly compared due to the differences in temporal resolution and fluorescent probe used. In Figure 2E (2F in the revised version), we measured ATP dissociation rate for both WT and E236A using smFRET. Due to the lower temporal resolution, we could not accurately determine ATP binding rate using smFRET. Therefore, to compare the ATP binding rate between WT and E236A heads, we now have added stopped-flow measurements of mant-ATP binding to the E236A monomer, as shown in Fig. 2C and Figure 2-supplement 2, and described in the text (lines 182-185).

Line 396: One of the most significant conclusions of this work is that the backward orientation of the neck linker has little effect on ATP binding to the front head. This is only supported by the results shown in Fig. 2A-B. Can the authors perform/analyze smFRET assays on the E236A/WT heterodimer to directly show whether the ATP binding rate to the WT head is affected or not affected by the orientation of the neck linker of the WT head?

We agree with the reviewer that our finding about ATP binding to the front head is potentially significant in the kinesin field, as it has been widely believed that ATP-binding is suppressed in the front head. In our original manuscript, this conclusion was supported only by the measurement of ATP on-rate of the front-head-crosslink, which may differ from the front head of a dimer in which the backward orientation of the neck linker is maintained by the backward strain. Although the reviewer suggested performing smFRET experiments using E236A/WT heterodimer, smFRET have relatively low temporal resolution (50-100 fps) and cannot accurately measure the frequency of ATP binding, so we used this technique only to determine ATP off rates. In this revised manuscript, we now have added stopped-flow experiments to separately measure the ATP binding to the front and rear heads of the E236A/WT heterodimer. By labeling the rear E236A head with a fluorophore to quench the mant-ATP signal bound to the rear head, we successfully measured mant-ATP binding rate to the front head. We found that the ATP-binding rate to the front head was comparable to that of an unconstrained monomer head, providing direct evidence for our conclusion. The revised version includes Fig. 4 A-C (with Figure 4-supplement 2; Figs. 4 and 5 are swapped in order) showing the kinetics of ATP binding to the front and rear heads of the E236A/WT heterodimer, with corresponding text in the result section (lines 315-324).

MINOR CONCERNS

Lines 31 and 32: I recommend replacing "ATP affinity" with "ATP binding rate" or "the dissociation of ATP" to be more specific. This is because they do not directly measure the affinity (Kd), but instead measure the on or off rates.

Line 41: Replace "cellar" with "cellular".

Line 83: The authors should cite Andreasson et al. here.

We have corrected these sentences accordingly (lines 31, 40, 85).

Lines 83-86: It seems this sentence belongs to the next paragraph. It also needs a citation(s).

This statement lacks experimental evidence and may confuse readers, so we have removed it for clarity.

Line 151: It would be helpful to add a conclusion sentence at the end of this paragraph to explain what these results mean to the reader.

A conclusion sentence of this paragraph has been added: “These results demonstrate that neck linker constraints in both forward and rearward orientations inhibit specific steps in the mechanochemical cycle of the head (lines 151-153)”.

Lines 175-180: I recommend combining and shortening these sentences, as follows, to avoid confusing the reader: "To detect the ATP dissociation event of the rear head, we employed a mutant kinesin with a point mutation of E236A in the switch II loop, which almost abolishes ATPase hydrolysis and traps in the microtubule-bound, neck-linker docked state,"

We have corrected these sentences accordingly (line 179-181).

Line 314: "which was rarely observed ...". This is out of place and confusing as is. I recommend moving this sentence after the sentence that ends in Line 295.

This sentence explains how the dark-field microscopy data was analyzed to determine whether the labeled head was in the leading or trailing position before detaching from the microtubule, but the explanation needs clarification. We removed the phrase “which was rarely observed for E236A-WT heterodimer” and simplified this sentence as follows: “Moreover, these observations allow us to distinguish whether the gold-labeled WT head was in the leading or trailing position just before microtubule detachment; the backward displacement of the detached head indicates that the labeled WT head occupied the leading position prior to detachment (Figure 5-figure supplement 1).” (lines 347-351).

Line 300: Can the authors comment on why E236A/WT has a substantially lower ATPase rate than WT homodimer? Is it possible to determine which step in the catalytic cycle is inhibited?

We demonstrated that the k₂ (microtubule-detachment rate) of the front head matched the ATP turnover rate of the E236A/WT heterodimer (Figure 6 B and E), suggesting that the inhibited step occurs after ATP binding in the front head. In contrast, the rear E236A head showed virtually no ATP hydrolysis activity, since in high-speed dark field microscopy, we observed forward step caused by rear E236A head detachment from microtubule only rarely, approximately once every few seconds (Figure 5-figure supplement 1). We added a sentence in the text as follows: “As described later, the reduced ATPase rate results from suppressed microtubule detachment of the front WT head, while the rear E236A head is virtually unable to detach from microtubules” (lines 311-313).

Line 323: Is the unbound dwell time unchanged?

The unbound dwell time exhibited a weak ATP-dependence, which we described only in Figure 5-supplement 2 (Figure 4-supplement 2 in the old version). We observed three distinct phases in the unbound dwell time based on mobility differences, with ATP dependence appearing only in the third phase. This finding suggests that ATP binding to the microtubule-bound E236A head is sometimes necessary for the detached WT head to rebind to the forward-tubulin binding site, indicating that the microtubule-bound E236A head occasionally releases ATP during the one-head-bound state (without the forward neck linker strain). To describe the ATP-dependence of the unbound dwell time, we added a sentence in the main text as follows: “In contrast, the dwell time of the unbound state of the gold-labeled WT head showed weak ATP dependence (Figure 5-figure supplement 2), indicating that the rear E236A head occasionally releases ATP when the front head detaches from the microtubule and the neck linker of E236A head becomes unconstrainted. This finding further supports the idea that forward neck linker strain plays a crucial role in reducing the reversible ATP release rate.” (lines 372-377).

Line 331: I recommend replacing "ATP-induced detachment" with "nucleotide-induced detachment" for clarity.

We have revised the phrase accordingly (line 371).

Line 344: I recommend replacing "affinity" with "forward strain prevents the release of the nucleotide" or similar to avoid confusion. Forward strain reduces the off-rate of the bound nucleotide, rather than allowing ATP to bind more efficiently to the rear head.

We agree to the reviewer’s comment and have corrected this sentence accordingly (line 338).

Lines 376-385: G7-12 constructs are introduced in Figure 6, but the results in this paragraph are shown in Figure 5. They should be moved to Figure 6 to avoid confusion.

To improve the readability, we have reorganized Figures 4-6, such that all the figure panels related to the neck linker extended mutants are shown in Figure 6; Figure 5D has been moved to Figure 6F.

Line 421: delete "not" before "does not".

We have corrected this typo.

Lines 433-441: Unless I am mistaken, more recent work in the kinesin field showed that backward trajectories of kinesin 1 reported by Carter and Cross are due to slips from the microtubule rather than backward processive runs of the motor.

The slip motion demonstrated by Sudhakar et al. (2021) differs from the backstep motion reported by Carter and Cross (and many other laboratories). Slip motion occurs after kinesin detaches from the microtubule and continues until the bead returns to the trap center. In contrast, backstep motion occurs during processive movement when the trap force either exceeds or approaches the stall force. The kinetics of these motions also differ significantly: slip steps occur with a dwell time of 71 µs and are independent of ATP concentration, while backsteps take ~0.3 s (at 1 mM ATP) and depend on ATP concentration. These differences indicate that slip motion is phenomenologically distinct from backsteps occurring under supra-stall or near-stall force.

Line 474: Replace "suppresses" with "suppressed".

We have corrected this typo.

Figure 4E: I would plot these results with increasing ATP concentration on the x-axis.

We formatted Figure 4E to match Figure 4b from Isojima et al. (Nature Chem. Biol. 2015), to emphasize the difference in ATP dependence of the front and rear head.

Figure 4B: The authors should explain how they distinguish between bound and unbound states in the main text or figure legends. For example, it is not clear how the authors score when the motor rebinds to the microtubule in the first unbinding event shown in Figure 4B (displacement plot).

The method was described in the Materials and Methods section, but we have now described how to distinguish between bound and unbound states in the main text as follows: “Unlike the unbound trailing head of wild-type dimer that showed continuous mobility (Isojima et al., 2016), the unbound WT head of E236A-WT heterodimer exhibited a low-fluctuation state in the middle (Figure 5B, s.d. trace). This low-fluctuation unbound state was distinguishable from the typical microtubule-bound state, having a shorter dwell time of ~5 ms compared to the bound state and positioning backward, closer to the E236A head, relative to the bound state (Figure 5-figure supplement 2).” (lines 351-356).

Reviewer #3:

Minor Issues:

- Line 22, Abstract - The phrase "move in a hand-over-hand manner" could be clearer if phrased as "move in a hand-over-hand fashion" to improve readability.

We changed the word “manner” to “process” (line 23).

- Abstract - Neck linker conformation in the leading head: The sentence "We demonstrate that the neck linker conformation in the leading kinesin head increases microtubule affinity without altering ATP affinity" would benefit from defining this conformation as "backward" for clarity.

- Abstract - Neck linker conformation in the trailing head: The sentence "The neck linker conformation in the trailing kinesin head increases ATP affinity by several thousand-fold compared to the leading head, with minimal impact on microtubule affinity" should also clarify that this conformation is "forward."

We have corrected these sentences accordingly (line 30, 32).

- Abstract - Conformation-specific effects: The authors mention conformation-specific effects in the neck linker structure but do not define the neck linker's conformation or the motor domain's (MD) conformation. Clarifying these conformational changes would improve the explanation of how they promote ATP hydrolysis and dissociation of the trailing head before the leading head detaches from the microtubule, thereby providing a kinetic basis for kinesin's coordinated walking mechanism.

We have revised the last sentence of the abstract accordingly by specifying the neck linker’s conformation as follows: “In combination, these conformation-specific effects of the neck linker favor ATP hydrolysis and dissociation of the rear head prior to microtubule detachment of the front head, thereby providing a kinetic explanation for the coordinated walking mechanism of dimeric kinesin.” (lines 34-37).

- Line 306 - Use of ATP in the E236A-WT heterodimer: In discussing the "ATP-induced detachment rate of the WT head in the E236A-WT heterodimer," the authors should consider justifying their choice of ATP over ADP for inducing microtubule (MT) dissociation. Since ATP typically promotes tighter MT binding and ATP turnover is reduced in forward-positioned WT heads, it may be unclear to some readers why ATP was chosen.

We measured the ATP-induced detachment rate k₂ of the front head of the E236A-WT heterodimer to validate our findings from the front-head-crosslinked monomer experiments, which demonstrated reduced k₂ after oxidation. To clarify this point, we have now included ATP binding kinetics measurements for both front and rear heads of the E236A-WT heterodimer, as suggested by reviewer 2. These additional data demonstrate consistency between the results from the crosslinked monomer and E236A-WT heterodimer experiments.

- Discussion - Backward-oriented neck linker in the front head: The discussion mentions that the backward-oriented neck linker in the front head reduces its ATP-induced detachment rate, suggesting that a step after ATP binding (e.g., isomerization, ATP hydrolysis, or phosphate release) is gated in the front head. However, the authors do not clarify that the backward neck linker orientation would imply the nucleotide pocket should be open or at least not fully closed, thus inhibiting ATP turnover. This is important because, as demonstrated in other studies, full closure of the nucleotide pocket is linked to neck linker docking. This point should be addressed earlier in the discussion.

We have addressed this point by revising this sentence as follows: “These results are consistent with an inability of the front head to fully close its nucleotide pocket to promote ATP hydrolysis and Pi release (Benoit et al., 2023), as will be discussed later.” (lines 441-443)

Reviewer #1 (Public review):

Summary:

Cell metabolism exhibits a well-known behavior in fast-growing cells, which employ seemingly wasteful fermentation to generate energy even in the presence of sufficient environmental oxygen. This phenomenon is known as Overflow Metabolism or the Warburg effect in cancer. It is present in a wide range of organisms, from bacteria and fungi to mammalian cells.

In this work, starting with a metabolic network for Escherichia coli based on sets of carbon sources, and using a corresponding coarse-grained model, the author applies some well-based approximations from the literature and algebraic manipulations. These are used to successfully explain the origins of Overflow Metabolism, both qualitatively and quantitatively, by comparing the results with E. coli experimental data.

By modeling the proteome energy efficiencies for respiration and fermentation, the study shows that these parameters are dependent on the carbon source quality constants K_i (p.115 and 116). It is demonstrated that as the environment becomes richer, the optimal solution for proteome energy efficiency shifts from respiration to fermentation. This shift occurs at a critical parameter value K_A(C).<br /> This counterintuitive result qualitatively explains Overflow Metabolism.

Quantitative agreement is achieved through the analysis of the heterogeneity of the metabolic status within a cell population. By introducing heterogeneity, the critical growth rate is assumed to follow a Gaussian distribution over the cell population, resulting in accordance with experimental data for E. coli. Overflow metabolism is explained by considering optimal protein allocation and cell heterogeneity.

The obtained model is extensively tested through perturbations: 1) Introduction of overexpression of useless proteins; 2) Studying energy dissipation; 3) Analysis of the impact of translation inhibition with different sub-lethal doses of chloramphenicol on Escherichia coli; 4) Alteration of nutrient categories of carbon sources using pyruvate. All model perturbations results are corroborated by E. coli experimental results.

Strengths:

In this work, the author effectively uses modeling techniques typical of Physics to address complex problems in Biology, demonstrating the potential of interdisciplinary approaches to yield novel insights. The use of Escherichia coli as a model organism ensures that the assumptions and approximations are well-supported in existing literature. The model is convincingly constructed and aligns well with experimental data, lending credibility to the findings. In this version, the extension of results from bacteria to yeast and cancer is substantiated by a literature base, suggesting that these findings may have broad implications for understanding diverse biological systems.

Weaknesses:

The author explores the generalization of their results from bacteria to cancer cells and yeast, adapting the metabolic network and coarse-grained model accordingly. In the previous version this generalization was not completely supported by references and data from the literature. This drawback, however, has been treated in this current version, where the authors discuss in much more detail and give references supporting this generalization.

Comments on revisions:

I have no specific comments for the authors. My previous comments were all addressed, discussed and explained.

Reviewer #1 (Public review):

Summary:

This paper presents a method for reconstructing videos from mouse visual cortex neuronal activity using a state-of-the-art dynamic neural encoding model. The authors achieve high-quality reconstructions of 10-second movies at 30 Hz from two-photon calcium imaging data, reporting a 2-fold increase in pixel-by-pixel correlation compared to previous methods. They identify key factors for successful reconstruction including the number of recorded neurons and model ensembling techniques.

Strengths:

(1) A comprehensive technical approach combining state-of-the-art neural encoding models with gradient-based optimization for video reconstruction.

(2) Thorough evaluation of reconstruction quality across different spatial and temporal frequencies using both natural videos and synthetic stimuli.

(3) Detailed analysis of factors affecting reconstruction quality, including population size and model ensembling effects.

(4) Clear methodology presentation with well-documented algorithms and reproducible code.

(5) Potential applications for investigating visual processing phenomena like predictive coding and perceptual learning.

Weaknesses:

The main metric of success (pixel correlation) may not be the most meaningful measure of reconstruction quality:

High correlation may not capture perceptually relevant features.

Different stimuli producing similar neural responses could have low pixel correlations The paper doesn't fully justify why high pixel correlation is a valuable goal

Comparison to previous work (Yoshida et al.) has methodological concerns: Direct comparison of correlation values across different datasets may be misleading; Large differences in the number of recorded neurons (10x more in the current study); Different stimulus types (dynamic vs static) make comparison difficult; No implementation of previous methods on the current dataset or vice versa.

Limited exploration of how the reconstruction method could provide insights into neural coding principles beyond demonstrating technical capability.

The claim that "stimulus reconstruction promises a more generalizable approach" (line 180) is not well supported with concrete examples or evidence.

The paper would benefit from addressing how the method handles cases where different stimuli produce similar neural responses, particularly for high-speed moving stimuli where phase differences might be lost in calcium imaging temporal resolution.

Reviewer #2 (Public review):

This is an interesting study exploring methods for reconstructing visual stimuli from neural activity in the mouse visual cortex. Specifically, it uses a competition dataset (published in the Dynamic Sensorium benchmark study) and a recent winning model architecture (DNEM, dynamic neural encoding model) to recover visual information stored in ensembles of the mouse visual cortex.

This is a great project - the physiological data were measured at a single-cell resolution, the movies were reasonably naturalistic and representative of the real world, the study did not ignore important correlates such as eye position and pupil diameter, and of course, the reconstruction quality exceeded anything achieved by previous studies. Overall, it is great that teams are working towards exploring image reconstruction. Arguably, reconstruction may serve as an endgame method for examining the information content within neuronal ensembles - an alternative to training interminable numbers of supervised classifiers, as has been done in other studies. Put differently, if a reconstruction recovers a lot of visual features (maybe most of them), then it tells us a lot about what the visual brain is trying to do: to keep as much information as possible about the natural world in which its internal motor circuits may act consequently.

While we enjoyed reading the manuscript, we admit that the overall advance was in the range of those that one finds in a great machine learning conference proceedings paper. More specifically, we found no major technical flaws in the study, only a few potential major confounds (which should be addressable with new analyses), and the manuscript did not make claims that were not supported by its findings, yet the specific conceptual advance and significance seemed modest. Below, we will go through some of the claims, and ask about their potential significance.

(1) The study showed that it could achieve high-quality video reconstructions from mouse visual cortex activity using a neural encoding model (DNEM), recovering 10-second video sequences and approaching a two-fold improvement in pixel-by-pixel correlation over attempts. As a reader, I am left with the question: okay, does this mean that we should all switch to DNEM for our investigations of the mouse visual cortex? What makes this encoding model special? It is introduced as "a winning model of the Sensorium 2023 competition which achieved a score of 0.301... single-trial correlation between predicted and ground truth neuronal activity," but as someone who does not follow this competition (most eLife readers are not likely to do so, either), I do not know how to gauge my response. Is this impressive? What is the best achievable score, in theory, given data noise? Is the model inspired by the mouse brain in terms of mechanisms or architecture, or was it optimized to win the competition by overfitting it to the nuances of the data set? Of course, I know that as a reader, I am invited to read the references, but the study would stand better on its own if clarified how its findings depended on this model.

(2) Along those lines, two major conclusions were that "critical for high-quality reconstructions are the number of neurons in the dataset and the use of model ensembling." If true, then these principles should be applicable to networks with different architectures. How well can they do with other network types?

(3) One major claim was that the quality of the reconstructions depended on the number of neurons in the dataset. There were approximately 8000 neurons recorded per mouse. The correlation difference between the reconstruction achieved by 1 neuron and 8000 neurons was ~0.2. Is that a lot or a little? One might hypothesize that ~7,999 additional neurons could contribute more information, but perhaps, those neurons were redundant if their receptive fields were too close together or if they had the same orientation or spatiotemporal tuning. How correlated were these neurons in response to a given movie? Why did so many neurons offer such a limited increase in correlation?

(4) On a related note, the authors address the confound of RF location and extent. The study resorted to the use of a mask on the image during reconstruction, applied during training and evaluation (Line 87). The mask depends on pixels that contribute to the accurate prediction of neuronal activity. The problem for me is that it reads as if the RF/mask estimate was obtained during the very same process of reconstruction optimization, which could be considered a form of double-dipping (see the "Dead salmon" article, https://doi.org/10.1016/S1053-8119(09)71202-9). This could inflate the reconstruction estimate. My concern would be ameliorated if the mask was obtained using a held-out set of movies or image presentations; further, the mask should shift with eye position, if it indeed corresponded to the "collective receptive field of the neural population." Ideally, the team would also provide the characteristics of these putative RFs, such as their weight and spatial distribution, and whether they matched the biological receptive fields of the neurons (if measured independently).

(5) We appreciated the experiments testing the capacity of the reconstruction process, by using synthetic stimuli created under a Gaussian process in a noise-free way. But this further raised questions: what is the theoretical capability for the reconstruction of this processing pipeline, as a whole? Is 0.563 the best that one could achieve given the noisiness and/or neuron count of the Sensorium project? What if the team applied the pipeline to reconstruct the activity of a given artificial neural network's layer (e.g., some ResNet convolutional layer), using hidden units as proxies for neuronal calcium activity?

(6) As the authors mentioned, this reconstruction method provided a more accurate way to investigate how neurons process visual information. However, this method consisted of two parts: one was the state-of-the-art (SOTA) dynamic neural encoding model (DNEM), which predicts neuronal activity from the input video, and the other part reconstructed the video to produce a response similar to the predicted neuronal activity. Therefore, the reconstructed video was related to neuronal activity through an intermediate model (i.e., SOTA DNEM). If one observes a failure in reconstructing certain visual features of the video (for example, high-spatial frequency details), the reader does not know whether this failure was due to a lack of information in the neural code itself or a failure of the neuronal model to capture this information from the neural code (assuming a perfect reconstruction process). Could the authors address this by outlining the limitations of the SOTA DNEM encoding model and disentangling failures in the reconstruction from failures in the encoding model?

(7) The authors mentioned that a key factor in achieving high-quality reconstructions was model assembling. However, this averaging acts as a form of smoothing, which reduces the reconstruction's acuity and may limit the high-frequency content of the videos (as mentioned in the manuscript). This averaging constrains the tool's capacity to assess how visual neurons process the low-frequency content of visual input. Perhaps the authors could elaborate on potential approaches to address this limitation, given the critical importance of high-frequency visual features for our visual perception.

Reviewer #2 (Public review):

Earhart et al. investigated the role of the complement system in trained innate immunity (TII) in alveolar macrophages (AM). They used a WT and C3 knockout murine model primed with locally administered heat-killed P. aeruginosa (HKPA). Additionally, they employed ex vivo AM training models using C3 knockout mice, where reconstitution of C3 and blockade of C3R were performed. The study concluded that the C3-C3R axis is essential for inducing TII in macrophages in the ex vivo model. The manuscript is well-written and easy to follow. However, I have the following major concerns.

(1) The secondary challenge to assess the reprogramming of innate cells in the BAL was conducted 14 days after the initial exposure to HKPA. However, no evidence is provided to confirm that homeostasis was re-established following the primary exposure. Demonstrating the resolution of acute inflammation is essential to ensure that the observed responses to the secondary challenge are not confounded by persistent inflammation from the initial exposure.

(2) In Figure 1D, cytokine production by BAL cells from WT and C3KO mice after HKPA exposure and LPS challenge is shown. However, it is unclear whether the reduced response in trained C3KO mice is due to a defect in trained immunity or an intrinsic inability of C3KO cells to respond to LPS. To clarify this, the response of trained C3KO cells should also be compared to untrained C3KO controls after the LPS challenge. This comparison is necessary to determine if the reduction is specifically related to innate immune memory or a broader impairment in LPS responsiveness. Such control should be included in all ex vivo training and LPS stimulation experiments as well.

(3) The data presented provide evidence of alterations in the functional and metabolic activities of innate cells in the lung, indicating the induction of innate immune memory in a C3-C3R axis-dependent pathway. However, it remains to be established whether such changes can lead to altered disease outcomes. Therefore, the impact of these changes should be demonstrated, for instance, through an infection model to support the claim made in the study that C3 modulates trained immunity in AMs through C3aR signalling.

(4) Figure 3, panels B and C - stats should be shown for comparing WT-HKPA-trained and C3KO HKPA-trained.

(5) In Figure 4, where the proper untrained C3KO is included, the data presented in Figure 4C show an increase in basal and maximum glycolysis in trained C3KO compared to their untrained control counterparts. Statistical analysis should be provided for this comparison. Based on these data, it appears that metabolic reprogramming occurs even in the absence of C3. Furthermore, C3KO cells intrinsically exhibit reduced glycolytic capacity compared to WT. These observations challenge the conclusions made in the manuscript. Therefore, without the proper control (untrained C3KO) included in all experimental approaches, it is impossible to draw an evidence-based conclusion that the C3-C3R axis plays a role in the induction of innate immune memory.

(6) The Results and Discussion sections should be separated, and the results should be thoroughly analyzed in the context of published literature. Separating these sections will allow for a clearer presentation of findings and ensure that the discussion provides a comprehensive interpretation of the data.

Author response:

We thank both reviewers for their suggestions on improving our manuscript, which is focused on demonstrating that the C3a-C3aR axis modulates trained immune responses in alveolar macrophages. The Short Report format precludes separating the Results and Discussion sections. However, we will work towards a clearer presentation of findings and providing a more comprehensive interpretation of the data in the Revision, by addressing the points brought up by both Reviewers.

We agree with the suggestions from Reviewer 1 that (1) other cell types such as dendritic cells, neutrophils, and endothelial cells can also be involved in immune training, and (2) macrophages have other activities beyond releasing inflammatory cytokines, and will clarify both these points in the Revision. The mechanism of C3 being cleaved intracellularly and binding to lysosomal C3aR involves cathepsin-dependent cleavage of C3 to C3a and has been experimentally proven (Liszewski et al. Immunity 2013). However, we will clarify this mechanism in the revision. We also acknowledge that the observations need to be validated in human-based models. Currently, we do not have access to an adequate representation of human alveolar macrophages for our ex vivo testing to account for individual-level variation in immune responses. However, we anticipate this work will form the basis of these future studies.

We also appreciate Reviewer 2’s suggestions regarding demonstrating the resolution of acute inflammation after the initial exposure to heat-killed Pseudomonas. We will address this critique by performing additional experiments, which will be included in the Revision. We also agree that the responses of trained C3-deficient cells should be compared to untrained C3-deficient controls after the LPS challenge. We will include this data in the Revision, in addition to the requested data for Figures 3 and 4. We would like to clarify that we do not observe baseline differences between untrained C3-sufficient (wildtype) and C3-deficient alveolar macrophages, even in their glycolytic capacity, and thus, anticipate that our revised data will strengthen the conclusions from the original manuscript.

Reviewer #1 (Public review):

Summary:

The authors aimed to characterize neurocomputational signals underlying interpersonal guilt and responsibility. Across two studies, one behavioral and one fMRI, participants made risky economic decisions for themselves or for themselves and a partner; they also experienced a condition in which the partners made decisions for themselves and the participant. The authors also assessed momentary happiness intermittently between choices in the task. Briefly, results demonstrated that participants' self-reported happiness decreased after disadvantageous outcomes for themselves and when both they and their partner were affected; this effect was exacerbated when participants were responsible for their partner's low outcome, rather than the opposite, reflecting experienced guilt. Consistent with previous work, BOLD signals in the insula correlated with experienced guilt, and insula-right IFG connectivity was enhanced when participants made risky choices for themselves and safe choices for themselves and a partner.

Strengths:

This study implements an interesting approach to investigating guilt and responsibility; the paradigm in particular is well-suited to approach this question, offering participants the chance to make risky v. safe choices that affect both themselves and others. I appreciate the assessment of happiness as a metric for assessing guilt across the different task/outcome conditions, as well as the implementation of both computational models and fMRI.

Weaknesses:

In spite of the overall strengths of the study, I think there are a few areas in which the paper fell a bit short and could be improved.

(1) While the framing and goal of this study was to investigate guilt and felt responsibility, the task implemented - a risky choice task with social conditions - has been conducted in similar ways in past research that were not addressed here. The novelty of this study would appear to be the additional happiness assessments, but it would be helpful to consider the changes noted in risk-taking behavior in the context of additional studies that have investigated changes in risky economic choice in social contexts (e.g., Arioli et al., 2023 Cerebral Cortex; Fareri et al., 2022 Scientific Reports).

(2) The authors note they assessed changes in risk preferences between social and solo conditions in two ways - by calculating a 'risk premium' and then by estimating rho from an expected utility model. I am curious why the authors took both approaches (this did not seem clearly justified, though I apologize if I missed it). Relatedly, in the expected utility approach, the authors report that since 'the number of these types of trials varied across participants', they 'only obtained reliable estimates for [gain and loss] trials in some participants' - in study 1, 22 participants had unreliable estimates and in study 2, 28 participants had unreliable estimates. Because of this, and because the task itself only had 20 gains, 20 losses, and 20 mixed gambles per condition, I wonder if the authors can comment on how interpretable these findings are in the Discussion. Other work investigating loss aversion has implemented larger numbers of trials to mitigate the potential for unreliable estimates (e.g., Sokol-Hessner et al., 2009).

(3) One thing seemingly not addressed in the Discussion is the fact that the behavioral effect did not replicate significantly in study 2.

(4) Regarding the computational models, the authors suggest that the Reponsibility and Responsibility Redux models provided the best fit, but they are claiming this based on separate metrics (e.g., in study 1, the redux model had the lowest AIC, but the responsibility only model had the highest R^2; additionally, the basic model had the lowest BIC). I am wondering if the authors considered conducting a direct model comparison to statistically compare model fits.

(5) In the reporting of imaging results, the authors report in a univariate analysis that a small cluster in the left anterior insula showed a stronger response to low outcomes for the partner as a result of participant choice rather than from partner choice. It then seems as though the authors performed small volume correction on this cluster to see whether it survived. If that is accurate, then I would suggest that this result be removed because it is not recommended to perform SVC where the volume is defined based on a result from the same whole-brain analysis (i.e., it should be done a priori).

Author response:

Reviewer #1 (Public review):

Summary:

The authors aimed to characterize neurocomputational signals underlying interpersonal guilt and responsibility. Across two studies, one behavioral and one fMRI, participants made risky economic decisions for themselves or for themselves and a partner; they also experienced a condition in which the partners made decisions for themselves and the participant. The authors also assessed momentary happiness intermittently between choices in the task. Briefly, results demonstrated that participants' self-reported happiness decreased after disadvantageous outcomes for themselves and when both they and their partner were affected; this effect was exacerbated when participants were responsible for their partner's low outcome, rather than the opposite, reflecting experienced guilt. Consistent with previous work, BOLD signals in the insula correlated with experienced guilt, and insula-right IFG connectivity was enhanced when participants made risky choices for themselves and safe choices for themselves and a partner.

Strengths:

This study implements an interesting approach to investigating guilt and responsibility; the paradigm in particular is well-suited to approach this question, offering participants the chance to make risky v. safe choices that affect both themselves and others. I appreciate the assessment of happiness as a metric for assessing guilt across the different task/outcome conditions, as well as the implementation of both computational models and fMRI.

We thank Reviewer 1 for their positive assessment of our manuscript.

Weaknesses:

In spite of the overall strengths of the study, I think there are a few areas in which the paper fell a bit short and could be improved.

We are looking forward to improving our manuscript based on the Reviewers’ comments. According to eLife’s policy, here are our provisional replies as well as plans for changes.

(1) While the framing and goal of this study was to investigate guilt and felt responsibility, the task implemented - a risky choice task with social conditions - has been conducted in similar ways in past research that were not addressed here. The novelty of this study would appear to be the additional happiness assessments, but it would be helpful to consider the changes noted in risk-taking behavior in the context of additional studies that have investigated changes in risky economic choice in social contexts (e.g., Arioli et al., 2023 Cerebral Cortex; Fareri et al., 2022 Scientific Reports).

We certainly agree that several previously published studies have relied on risky choice tasks with social conditions. We will happily refer to the studies mentioned when discussing changes in risk-taking behaviour in our revised manuscript.

(2) The authors note they assessed changes in risk preferences between social and solo conditions in two ways - by calculating a 'risk premium' and then by estimating rho from an expected utility model. I am curious why the authors took both approaches (this did not seem clearly justified, though I apologize if I missed it). Relatedly, in the expected utility approach, the authors report that since 'the number of these types of trials varied across participants', they 'only obtained reliable estimates for [gain and loss] trials in some participants' - in study 1, 22 participants had unreliable estimates and in study 2, 28 participants had unreliable estimates. Because of this, and because the task itself only had 20 gains, 20 losses, and 20 mixed gambles per condition, I wonder if the authors can comment on how interpretable these findings are in the Discussion. Other work investigating loss aversion has implemented larger numbers of trials to mitigate the potential for unreliable estimates (e.g., Sokol-Hessner et al., 2009).

We agree that we have not clearly justified why we have taken two approaches to assess risk preferences. In short, both approaches have advantages and inconveniences when applied to our experiment. We will happily detail our reasons in the revised manuscript. Regarding the second point of this comment: the small number of reliable estimates is one of the reasons that we have used another approach to assess risk preferences. We would certainly have obtained more reliable estimates if we had implemented more trials. We will discuss the interpretability of all the risk preference estimates we used in the revised Discussion.

(3) One thing seemingly not addressed in the Discussion is the fact that the behavioral effect did not replicate significantly in study 2.

We agree that we could have discussed more the fact that there were (slight but significant) differences in risk preferences between the Solo and Social conditions in Study 1 but not in Study 2. While the absence of a significant difference in Study 2 is helpful to compare the neural mechanisms involved in making decisions for oneself vs. for oneself and another person (because any differences could not be explained by differences in risk preferences), we certainly should expand our discussion of the differences in findings between the two studies, which we will do in the revised manuscript.

(4) Regarding the computational models, the authors suggest that the Reponsibility and Responsibility Redux models provided the best fit, but they are claiming this based on separate metrics (e.g., in study 1, the redux model had the lowest AIC, but the responsibility only model had the highest R^2; additionally, the basic model had the lowest BIC). I am wondering if the authors considered conducting a direct model comparison to statistically compare model fits.

We agree that we should run formal, direct model comparison tests using for example chi-square or log-likelihood-ratio tests. We will do so in the revised manuscript.

(5) In the reporting of imaging results, the authors report in a univariate analysis that a small cluster in the left anterior insula showed a stronger response to low outcomes for the partner as a result of participant choice rather than from partner choice. It then seems as though the authors performed small volume correction on this cluster to see whether it survived. If that is accurate, then I would suggest that this result be removed because it is not recommended to perform SVC where the volume is defined based on a result from the same whole-brain analysis (i.e., it should be done a priori).

As indicated in the manuscript, the small insula cluster centered at [-28 24 -4] and shown in Figure 4F survived corrections for multiple tests within the anatomically-defined anterior insula (based on the anatomical maximum probability map described in Faillenot et al., 2017), which is independent of the result of our analysis. We agree that one should not (and we did not) perform multiple corrections based on the results one is correcting – that would indeed be circular and misleading “double-dipping”. The anterior insula is one of the regions most frequently associated with guilt (see the explanations in our Introduction, which refers for example to Bastin et al., 2016; Lamm & Singer, 2010; Piretti et al., 2023). Thus we feel that performing small-volume correction within the anatomically-defined anterior insula is an acceptable approach to correct for multiple tests in this case. We fully acknowledge that, independently of any correction, the effect and the cluster are small. We will clarify these explanations in the revised manuscript.

Reviewer #2 (Public review):

Summary

This manuscript focuses on the role of social responsibility and guilt in social decision-making by integrating neuroimaging and computational modeling methods. Across two studies, participants completed a lottery task in which they made decisions for themselves or for a social partner. By measuring momentary happiness throughout the task, the authors show that being responsible for a partner's bad lottery outcome leads to decreased happiness compared to trials in which the participant was not responsible for their partner's bad outcome. At the neural level, this guilt effect was reflected in increased neural activity in the anterior insula, and altered functional connectivity between the insula and the inferior frontal gyrus. Using computational modeling, the authors show that trial-by-trial fluctuations in happiness were successfully captured by a model including participant and partner rewards and prediction errors (a 'responsibility' model), and model-based neuroimaging analyses suggested that prediction errors for the partner were tracked by the superior temporal sulcus. Taken together, these findings suggest that responsibility and interpersonal guilt influence social decision-making.

Strengths

This manuscript investigates the concept of guilt in social decision-making through both statistical and computational modeling. It integrates behavioral and neural data, providing a more comprehensive understanding of the psychological mechanisms. For the behavioral results, data from two different studies is included, and although minor differences are found between the two studies, the main findings remain consistent. The authors share all their code and materials, leading to transparency and reproducibility of their methods.

The manuscript is well-grounded in prior work. The task design is inspired by a large body of previous work on social decision-making and includes the necessary conditions to support their claims (i.e., Solo, Social, and Partner conditions). The computational models used in this study are inspired by previous work and build on well-established economic theories of decision-making. The research question and hypotheses clearly extend previous findings, and the more traditional univariate results align with prior work.

The authors conducted extensive analyses, as supported by the inclusion of different linear models and computational models described in the supplemental materials. Psychological concepts like risk preferences are defined and tested in different ways, and different types of analyses (e.g., univariate and multivariate neuroimaging analyses) are used to try to answer the research questions. The inclusion and comparison of different computational models provide compelling support for the claim that partner prediction errors indeed influence task behavior, as illustrated by the multiple model comparison metrics and the good model recovery.

We thank Reviewer 2 very much for their comprehensive description of our study and the positive assessment of our study and approach.

Weaknesses

As the authors already note, they did not directly ask participants to report their feelings of guilt. The decrease in happiness reported after a bad choice for a partner might thus be something else than guilt, for example, empathy or feelings of failure (not necessarily related to guilt towards the other person). Although the patterns of neural activity evoked during the task match with previously found patterns of guilt, there is no direct measure of guilt included in the task. This warrants caution in the interpretation of these findings as guilt per se.

We fully agree that not directly asking participants about feelings of guilt is a clear limitation of our study. While we already mention this in our Discussion, we will happily expand our discussion of the consequences on interpretation of our results along the lines described by the reviewer in the revised manuscript. We would like to thank Reviewer 2 for proposing these lines of thought.

As most comparisons contrast the social condition (making the decision for your partner) against either the partner condition (watching your partner make their decision) or the solo condition (making your own decision), an open question remains of how agency influences momentary happiness, independent of potential guilt. Other open questions relate to individual differences in interpersonal guilt, and how those might influence behavior.

We fully agree that the way agency influences happiness has not been much discussed in our manuscript so far, and we would happily do so in the revised manuscript. The same goes for individual differences in interpersonal guilt which we have not investigated due to our relatively small sample sizes but would certainly be worth investigation in subsequent work.

This manuscript is an impressive combination of multiple approaches, but how these different approaches relate to each other and how they can aid in answering slightly different questions is not very clearly described. The authors could improve this by more clearly describing the different methods and their added value in the introduction, and/or by including a paragraph on implications, open questions, and future work in the discussion.

We again thank the reviewer for their praise of our approach and fully agree that we can improve the description of the benefit of combining methods in the Introduction, which we will do in the revised manuscript. We will also include a paragraph on implications, open questions, and future work in the Discussion of the revised manuscript.

However, taken together, this study provides useful insights into the neural and behavioral mechanisms of responsibility and guilt in social decision-making, and how they influence behavior.

We again thank Reviewer 2 for their attentive reading and thoughtful comments and look forward to submitting our revised and improved manuscript.

Author response:

Reviewer 1:

(1) We appreciate the reviewer’s suggestion to test a multi-attribute attentional drift-diffusion model (maaDDM) that does not constrain the taste and health weights to the range of 0 and 1 and will test such a model.

(2) Similarly, we will follow the reviewer’s suggestion to address potential demand effects. First, we will add “order” (binary: hungry-sated or sated hungry) as a predictor to our GLMM, to test for potential systematic effects of order on choices and response times. Second, we will split the participants by “order” and examine whether we see group differences of tasty and healthy decisions within the first testing session. Note that we already anticipate that looking at only 50% of the data and testing for a between-subject rather than within-subject effect is likely to reduce effect size and statistical sensitivity.

(3) We thank the reviewer for their observant remark about faster tasty choices and potential markers in the drift rate. While our starting point models show that there might be a small starting point bias towards the taste boundary which result in faster decisions, we will take a closer look at the simulated value differences as obtained in our posterior predictive checks to see if the drift rate is systematically more extreme for tasty choices.

(4) Regarding the mtDDM, we will verify that the relative starting time (rst) effects are minuscule. While we will follow the recommendation of correlating first fixations with rst, we would like to point out that a majority of fixations (see Figure 3b) and first fixations (see Figure S6b) are on food images. We will also provide a parameter recovery of the mtDDM.

Reviewer 2:

(1) We would like to verify the reviewer’s interpretation that hungry people in negative calorie balance simply prefer more calories and would like to point to our supplementary analyses, in which we show that hunger state also increases the probability of higher wanted and higher caloric decisions (see SOM4, SOM5, Figure S4). Moreover, we agree that high caloric items might not be unhealthy and are happy to demonstrate the correlations between health ratings and objective caloric content, to demonstrate the strong negative correlation in our dataset, which our principal component analyses hints at, too.

Reviewer 3:

(1) We agree that choosing tasty over healthy options under hunger may be evolutionarily adaptive. We will address the adaptiveness of this hunger driven mechanism in our discussion, reiterating the differentiation made in the introduction that this system no longer be adaptive in our obesogenic environment, leading to suboptimal decisions.

(2) We will address alternative explanations of the observed effects in our discussion with respect to the macro-nutritional content of the Shake and potential placebo effects arising from the shake vs no shake manipulation.

Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Revision Plan (Response to Reviewers)

1. General Statements [optional]

Response: We are pleased the reviewers appreciate the power of this novel proteomics methodology that allowed us to uncover new depths on the complexity of the ribosome ubiquitination code in response to stress. We also appreciate that the reviewers think that this is a “very timely” study and “interesting to a broad audience” that can change the models of translation control currently adopted in the field. Characterizing complex cellular processes is critical to advance scientific knowledge and our work is the first of its kind using targeted proteomics methods to unveil the integrated complexity of ribosome ubiquitin signals in eukaryotic systems. We also appreciate the fairness of the comments received and below we offer a comprehensive revision plan substantially addressing the main points raised by the reviewers. According to the reviewers’ suggestions, we will also expand our studies to two additional E3 ligases (Mag2 and Not4) known to ubiquitinate ribosomes, which will create an even more complete perspective of ubiquitin roles in translation regulation.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

        __Response: __We appreciate the evaluation of our work and that the power of our proteomics method established a good foundation for the study. We also understand the reviewer’s concerns and we will detail below a plan to enhance quantification and increase systematic comparisons. The experiments presented here were conducted with biological replicates, but in several instances, we focused on presence and absence of bands, or their pattern (mono vs poly-ub) because of the semi-quantitative nature of immunoblots. We will revise the figures and present their quantification and statistical analyses. In additional, we did not intend to use these stressors interchangeably, but instead, to use select conditions to highlight the complexity the stress response. In particular, we followed up with H2O2 *versus* 4-NQO because both chemicals are considered sources of oxidative stress. Even though it is unfeasible to compare every single stress condition in every strain background, in the revised version, we will include additional controls to increase the cohesion of the narrative, and expand the comparison between MMS, H2O2, and 4-NQO, as suggested. Details below.

To strengthen the work, the following revisions are essential:

R1.1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.

        __Response: __As requested, we will display quantification and statistical analysis of the suggested and new immunoblots that will be conducted during the revision period.

R1.3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.

        __Response: __We will follow the reviewers’ suggestion and redesign the analysis to increase consistency and prioritize data under identical conditions. To increase confidence in the mRNA data analysis, we intend to perform follow up experiments and analyze protein abundance of *ARG proteins* and *CTT1 *under different conditions. The remaining data using non-parallel comparisons will be moved to supplemental material and de-emphasized in the final version of the manuscript.

R1.4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

        __Response: __To ensure a better comparison across strains and conditions, we will re-run several experiments and focus on our main stress conditions. Specifically:

• 3D: We plan to re-run this experiment and include MMS

• 3E: We plan to perform the same panel of experiments in rad6D ,and display WT data as main figure.

• 4A-B: We plan to perform translation output (HPG incorporation) experiments with MMS as suggested

• 4C: We plan to re-run blots for p-eIF2a under MMS for improved comparison.

Reviewer #1 (Significance (Required)):

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript the authors use a new target proteomics approach to quantify site-specific ubiquitin modification across the ribosome before and after oxidative stress. Then they validate their findings following in particular ubiquitination of Rps20 and Rps3 and extend their analysis to different forms of oxidative stress. Finally they question the relevance of two known actors of ribosome ubiquitination, Hel2 and Rad6. It is not easy to summarize the observations because in fact the major finding is that the patterns of ribosome ubiquitination occur in a stresser and enyzme specific manner (even when considering only oxidative stress). However, the complexity revealed by this study is very relevant for the field, because it underlies that the ubiquitination code of ribosomes is not easy to interpret with regard to translation dynamics and responses to stress or players involved. It suggests that some of the models that have generally been adopted probably need to be amended or completed. I am not a proteomics expert, so I cannot comment on the validity of the new proteomics approach, of whether the methods are appropriately described to reproduce the experiments. However, for the follow up experiments, the results following Rps20 and Rps3 ubiquitination are well performed, nicely controlled and are appropriately interpreted.

Maybe what one can regret is that the authors have limited their analysis to the study of Hel2 and Rad6, and not included other enyzmes that have already been associated with regulation of ribosome ubiquitination, to get a more complete picture. It may not take that much time to test more mutants, but of course there is the risk that rather than enable to make a working model it might make things even more complex.

        __Response: __We value the positive evaluation of our work. We also appreciate the notion that it meaningfully expands the knowledge on the complexity of the ribosome ubiquitination code, challenges the current models of translation control, and conducted well-performed, and nicely controlled experiments. To address the main concern of the reviewer, we will expand our work by studying two additional enzymes involved in ribosome ubiquitination (Mag2 and Not4) and provide a more comprehensive picture of this integrated system. Specifically, we will generate yeast strains deleted for *MAG2* and *NOT4*, and evaluate their impact in ribosome ubiquitination under our main conditions of stress. We will investigate the role of these additional E3s in translation output (HPG incorporation), and in inducing the integrated stress response via phosphorylated eIF2α and Gcn4 expression. Additional follow up experiments will be performed according to our initial results.

Reviewer #2 (Significance (Required)):

In recent years, regulation of translation elongation dynamics has emerged as a much more relevant site of control of gene expression that previously envisonned. The ribosome has emerged as a hub for control of stress responses. Therefore this study is certainly very timely and interesting for a broad audience. However, it does fall short of giving any simple picture, and maybe the only point one can question is whether it is interesting to publish a manuscript that concludes that regulation is complicated, without really being able to provide any kind of suggestive model.

My feeling is nevertheless that it will impact how scientists in the field design their experiments and what they will conclude. It will certainly also drive new experiments and approaches, and lead to investigations on how all the different players in regulation of ribosome modification talk to each other and signal to signaling pathways.

        __Response: __We appreciate the comments and the balanced view that studies like ours will still be impactful and contribute to a number of fields in multiple and meaningful ways. With the new experiments proposed here, and used of additional mutants and strains, we intend to propose and provide a more unified model that explain this complex and dynamic relationship.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

R3.1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.

        __Response: __We understand the importance of the suggested experiment. We have already requested and kindly received strains expressing these mutations, which will reduce the time required to successfully address this point. We will perform our translation and ISR assays such as the one referred by the reviewer in Figs. 4A-C and 5E, and results will determine the role of individual ribosome ubiquitination sites in translation control.

R3.2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

        __Response: __As was also requested by reviewer 1 and discussed above (point R1.1), we will conduct quantification and display statistical analyses for our immunoblots. In addition, we will re-run the aforementioned experiments to improve quantification following the reviewers’ request (same gel & diluted control samples).

Reviewer #3 (Significance (Required)):

• General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

• Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

• Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

• The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

__ Response:__ We appreciate that our work will be valuable to a number of fields in protein dynamics and that our method advances the field by measuring ribosome ubiquitination relatively and stoichiometrically in response to stress.

3. Description of the revisions that have already been incorporated in the transferred manuscript

Response: All requested changes require experiments and data analyses, and a complete revision plan is delineated above in section #2.

• *

4. Description of analyses that authors prefer not to carry out

• *

R1.2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.

        __Response: __Although we understand the interest on the proposed result for consistency, this is the only requested experiment that we do not intend to conduct. Because of the lack of overall ubiquitination of ribosomal proteins in *rad6**D* in response to H2O2 (e.g., Silva et al., 2015, Simoes et al., 2022), we believe that this PRM experiment in unlikely to produce meaningful insight on the ubiquitination code. In this context, we expected that sites regulated by Hel2 will be the ones largely modified in rad6*D *and we followed up on them via immunoblot. Moreover, this experiment would not be time or cost-effective, and resources and efforts could be used to strengthen other important areas of the manuscript, such as including the E3’s Mag2 and Not4 into our work.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Recent studies have shown that the ubiquitination of uS3 (Rps3) is crucial for the quality control of nonfunctional rRNA, specifically in the process known as 18S noncoding RNA degradation (NRD). Additionally, the ubiquitination of uS10 (Rps20) plays a significant role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress are not yet fully understood.

In this study, the authors developed a targeted proteomics method to quantify the dynamics of ribosome ubiquitination in response to oxidative stress, both relatively and stoichiometrically. They identified 11 ribosomal sites that exhibited increased ubiquitin modification after exposure to hydrogen peroxide (H2O2). This included two known targets: uS10 and uS3 (of Hel2), which recognize collided ribosomes and initiate the processes of 18S NRD and translation quality control (RQC). Using isotope-labeled peptides, the researchers demonstrated that these modifications are non-stoichiometric and display significant variability among different peptides.

Furthermore, the authors explored how specific enzymes in the ubiquitin system affect these modifications and their impact on global translation regulation. They found that uS3 (Rps3) and uS10 (Rps20) were modified differently by various stressors, which in turn influenced the Integrated Stress Response (ISR). The authors suggest that different types of stressors alter the pattern of ubiquitinated ribosomes, with Rad6 and Hel2 potentially competing for specific subpopulations of ribosomes.

Overall, this study emphasizes the complexity of the ubiquitin ribosomal code. However, further experiments are necessary to validate these findings before publication.

Major Comments:

I consider the additional experiments essential to support the claims of the paper.

1. To understand the roles of ribosome ubiquitination at the specific sites, the authors must perform stressor-specific suppression of global translation, as demonstrated in Figures 4 and 5. This should include the uS10-K6R/K8R and uS3-K212R mutants.
2. It is crucial to ensure that experiments are adequately replicated and that statistical analysis is thorough, with precise quantification. For a more accurate comparison between wild-type (WT) and Hel2 deletion mutants regarding ribosome ubiquitination, the authors should quantify the ubiquitinated ribosomes in both WT and Hel2 mutants under stress conditions. This quantification should be conducted on the same blot, using diluted control samples. Similarly, in Figures 3F and 4C, for an accurate comparison between WT and Hel2 or Rad6 deletion mutants, the authors should quantify the ubiquitinated ribosomes across these conditions. Again, this quantification should be performed on the same blot with the dilution of control samples.

Significance

General assessment:

Recent studies reveal that the ubiquitination of uS3 (Rps3) is essential for the quality control of nonfunctional rRNA (18S NRD), while the ubiquitination of uS10 (Rps20) plays a crucial role in ribosome-associated quality control (RQC). However, the dynamics of ribosome ubiquitination in response to oxidative stress remain unclear.

Advance:

In this study, the authors developed a targeted proteomics method to quantify ribosome ubiquitination dynamics in response to oxidative stress, both relatively and stoichiometrically. By utilizing isotope-labeled peptides, they demonstrated that these modifications are non-stoichiometric and exhibit significant variability across different peptides. They identified 11 ribosomal sites that showed increased ubiquitin modification following H2O2 exposure, including two known targets of Hel2, which recognize collided ribosomes and induce translation quality control (RQC).

Audience: This information will be of interest to a specialized audience in the fields of translation, ribosome function, quality control, ubiquitination, and proteostasis.

The field: Translation, ribosome function, quality control, ubiquitination, and proteostasis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

However, the study's focus shifts in a manner that introduces several limitations. Following the rigorous PRM-based analyses, the reliance on Western blotting without replication or quantification (e.g., single-experiment data in Figs. 3-5) significantly weakens the evidence. Experimental design becomes inconsistent, with variable combinations of stressors (H₂O₂, MMS, 4-NQO) and genetic backgrounds (WT, hel2Δ, rad6Δ) that preclude systematic comparisons. For instance, Fig. 3C/E and Fig. 4 omit critical controls (e.g., MMS in Fig. 4, rad6Δ in Fig. 3E), while Fig. 5 conflates distinct variables by comparing H₂O₂-treated rad6Δ with MMS-treated hel2Δ-a design that obscures causal relationships. Furthermore, Fig. 3F highlights that 4-NQO and MMS elicit divergent responses in hel2Δ, undermining the rationale for using these stressors interchangeably. These inconsistencies culminate in a fragmented narrative; attempts to link ISR activation or ribosome stalling to RP ubiquitylation become impossible, leaving the primary takeaway as "stress responses are complex" rather than advancing mechanistic insight.

To strengthen the work, the following revisions are essential:

1. Repeat and quantify immunoblots: All Western blotting data require biological replicates and statistical analysis to support claims.
2. Leverage the PRM platform: Apply the established quantitative proteomics approach to validate or extend findings in Fig. 3 (e.g., RAD6-dependent ubiquitylation), ensuring methodological consistency.
3. Remove non-parallel comparisons: The mRNA expression analysis in Fig. 5, which compares dissimilar conditions (e.g., rad6Δ + H₂O₂ vs. hel2Δ + MMS), should be omitted or redesigned to enable direct, strain- and stressor-matched contrasts.
4. Standardize experimental variables: Restructure the study to maintain identical genetic backgrounds and stressors across all figures, enabling systematic interrogation of enzyme- or stress-specific effects on the ubiquitin code.

Significance

The authors present a potentially powerful proteomics platform using parallel reaction monitoring (PRM) to quantitatively profile ribosomal protein (RP) ubiquitylation, with a focus on yeast under hydrogen peroxide (H₂O₂) stress. This approach robustly identifies both known and novel RP modifications, including basal ubiquitylation events previously undetected, and identifies Hel2-dependent mechanisms. The data support the conclusion that RPs are regulated by a multifaceted ubiquitin code, establishing a good foundation for the study.

Reviewer #1 (Public review):

This work employs both in vitro and in vivo/transplant methods to investigate the contribution of BDNF/TrkB signaling to enhancing differentiation and dentin-repair capabilities of dental pulp stem cells in the context of exposure to a variety of inflammatory cytokines. A particular emphasis of the approach is the employment of dental pulp stem cells in which BDNF expression has been enhanced using CRISPR technology. Transplantation of such cells is said to improve dentin regeneration in a mouse model of tooth decay.

The study provides several interesting findings, including demonstrating that exposure to several cytokines/inflammatory agents increases the quantity of (activated) phospho-Trk B in dental pulp stem cells.

However, a variety of technical issues weaken support for the major conclusions offered by the authors. These technical issues include the following:

(1) It remains unclear exactly how the cytokines tested affect BDNF/TrkB signaling. For example, in Figure 1C, TNF-alpha increases TrkB and phospho-TrkB immunoreactivity to the same degree, suggesting that the cytokine promotes TrkB abundance without stimulating pathways that activate TrkB, whereas in Figure 2D, TNF-alpha has little effect on the abundance of TrkB, while increasing phospho-TrkB, suggesting that it affects TrkB activation and not TrkB abundance.

(2) I find the histological images in Figure 3 to be difficult to interpret. I would have imagined that DAPI nuclear stains would reveal the odontoblast layer, but this is not apparent. An adjacent section labeled with conventional histological stains would be helpful here. Others have described Stro-1 as a stem cell marker that is expressed on a minority of cells associated with vasculature in the dental pulp, but in the images in Figure 3, Stro-l label is essentially co-distributed with DAPI, in both control and injured teeth, indicating that it is expressed in nearly all cells. Although the authors state that the Stro-1-positive cells are associated with vasculature, but I see no evidence that is true.

(3) The data presented convincingly demonstrate that they have elevated BDNF expression in their dental pulp stem cells using a CRISPR-based approach I have a number of questions about these findings. Firstly, nowhere in the paper do they describe the nature of the CRISPR plasmid they are transiently transfecting. Some published methods delete segments of the BDNF 3'-UTR while others use an inactivated Cas9 to position an active transactivator to sequences in the BDNF promoter. If it is the latter approach, transient transfection will yield transient increases in BDNF expression. Also, as BDNF employs multiple promoters, it would be helpful to know which promoter sequence is targeted, and finally, knowing the identity of the guide RNAs would allow assessment for the potential of off-target effects I am guessing that the investigators employ a commercially obtained system from Santa Cruz, but nowhere is this mentioned. Please provide this information.

(4) Another question left unresolved is whether their approach elevated BDNF, proBDNF, or both. Their 28 kDa western blot band apparently represents proBDNF exclusively, with no mature BDNF apparent, yet only mature BDNF effectively activates TrkB receptors. On the other hand, proBDNF preferentially activates p75NTR receptors. The present paper never mentions p75NTR, which is a significant omission, since other investigators have demonstrated that p75NTR controls odontoblast differentiation.

(5) In any case, no evidence is presented to support the conclusion that the artificially elevated BDNF expression has any effect on the capability of the dental pulp stem cells to promote dentin regeneration. The results shown in Figures 4 and 5 compare dentin regeneration with BDNF-over-expressing stem cells with results lacking any stem cell transplantation. A suitable control is required to allow any conclusion about the benefit of over-expressing BDNF.

(6) Whether increased BDNF expression is beneficial or not, the evidence that the BDNF-overexpressing dental pulp stem cells promote dentin regeneration is somewhat weak. The data presented indicate that the cells increase dentin density by only 6%. The text and figure legend disagree on whether the p-value for this effect is 0.05 or 0.01. In either case, nowhere is the value of N for this statistic mentioned, leaving uncertainty about whether the effect is real.

(7) The final set of experiments applies transcriptomic analysis to address the mechanisms mediating function differences in dental pulp stem cell behavior. Unfortunately, while the Abstract indicates " we conducted transcriptomic profiling of TNFα-treated DPSCs, both with and without TrkB antagonist CTX-B" that does not describe the experiment described, which compared the transcriptome of control cells with cells simultaneously exposed to TNF-alpha and CTX-B. Since CTX-B blocks the functional response of cells to TNF-alpha, I don't understand how any useful interpretation can be attached to the data without controls for the effect of TNF alone and CTX-B alone.

Reviewer #2 (Public review):

Summary:<br /> In this manuscript, the authors investigate the potential for overexpressing BDNF in dental pulp stem cells to enhance dentin regeneration. They suggest that in the inflammatory environment of injured teeth, there is increased signaling of TrkB in response to elevated levels of inflammatory molecules.

Strengths:<br /> The potential application to dentin regeneration is interesting.

Weaknesses:<br /> There are a number of concerns with this manuscript to be addressed.

(1) Insufficient citation of the literature. There is a vast literature on BDNF-TrkB regulating survival, development, and function of neurons, yet there is only one citation (Zhang et al 2012) which is on Alzheimer's disease.

(2) There are several incorrect statements. For example, in the introduction (line 80) TrkA is not a BDNF receptor.

(3) Most important - Specific antibodies must be identified by their RRID numbers. To state that "Various antibodies were procured:... from BioLegend" is unacceptable, and calls into question the entire analysis. Specifically, their Western blot in Figure 4B indicates a band at 28 kDa that they say is BDNF, however the size of BDNF is 14 kDa, and the size of proBDNF is 32 and 37 kDa, therefore it is not clear what they are indicating at 28 kDa. The validation is critical to their analysis of BDNF-expressing cells.

(4) Figure 2 indicates increased expression of TrkB and TrkA, as well as their phosphorylated forms in response to inflammatory stimuli. Do these treatments elicit increased secretion of the ligands for these receptors, BDNF and NGF, respectively, to activate their phosphorylation? Or are they suggesting that the inflammatory molecules directly activate the Trk receptors? If so, further validation is necessary to demonstrate that.

(5) Figure 7 - RNA-Seq data, what is the rationale for treatment with TNF+ CTX-B? How does this identify any role for TrkB signaling? They never define their abbreviations, but if CTX-B refers to cholera toxin subunit B, which is what it usually refers to, then it is certainly not a TrkB antagonist.

Reviewer #3 (Public review):

In general, although the authors interpret their results as pointing towards a possible role of BDNF in dentin regeneration, the results are over-interpreted due to the lack of proper controls and focus on TrkB expression, but not its isoforms in inflammatory processes. Surprisingly, the authors do not study the possible role of p75 in this process, which could be one of the mechanisms intervening under inflammatory conditions.

(1) The authors claim that there are two Trk receptors for BDNF, TrkA and TrkB. To date, I am unaware of any evidence that BDNF binds to TrkA to activate it. It is true that two receptors have been described in the literature, TrkB and p75 or NGFR, but the latter is not TrkA despite its name and capacity to bind NGF along with other neurotrophins. It is crucial for the authors to provide a reference stating that TrkA is a receptor for BDNF or, alternatively, to correct this paragraph.

(2) The authors discuss BDNF/TrkB in inflammation. Is there any possibility of p75 involvement in this process?

(3) The authors present immunofluorescence (IF) images against TrkB and pTrkB in the first figure. While they mention in the materials and methods section that these antibodies were generated for this study, there is no proof of their specificity. It should be noted that most commercial antibodies labeled as anti-TrkB recognize the extracellular domain of all TrkB isoforms. There are indications in the literature that pathological and excitotoxic conditions change the expression levels of TrkB-Fl and TrkB-T1. Therefore, it is necessary to demonstrate which isoform of TrkB the authors are showing as increased under their conditions. Similarly, it is essential to prove that the new anti-p-TrkB antibody is specific to this Trk receptor and, unlike other commercial antibodies, does not act as an anti-phospho-pan-Trk antibody.

(4) I believe this initial conclusion could be significantly strengthened, without opening up other interpretations of the results, by demonstrating the specificity of the antibodies via Western blot (WB), both in the presence and absence of BDNF and other neurotrophins, NGF, and NT-3. Additionally, using WB could help reinforce the quantification of fluorescence intensity presented by the authors in Figure 1. It's worth noting that the authors fixed the cells with 4% PFA for 2 hours, which can significantly increase cellular autofluorescence due to the extended fixation time, favoring PFA autofluorescence. They have not performed negative controls without primary antibodies to determine the level of autofluorescence and nonspecific background. Nor have they indicated optimizing the concentration of primary antibodies to find the optimal point where the signal is strong without a significant increase in background. The authors also do not mention using reference markers to normalize specific fluorescence or indicating that they normalized fluorescence intensity against a standard control, which can indeed be done using specific signal quantification techniques in immunocytochemistry with a slide graded in black-and-white intensity controls. From my experience, I recommend caution with interpretations from fluorescence quantification assays without considering the aforementioned controls.

(5) In Figure 2, the authors determine the expression levels of TrkA and TrkB using qPCR. Although they specify the primers used for GAPDH as a control in materials and methods, they do not indicate which primers they used to detect TrkA and TrkB transcripts, which is essential for determining which isoform of these receptors they are detecting under different stimulations. Similarly, I recommend following the MIQE guidelines (Minimum Information for Publication of Quantitative Real-Time PCR experiments), so they should indicate the amplification efficiency of their primers, the use of negative and positive controls to validate both the primer concentration used, and the reaction, the use of several stable reference genes, not just one.

(6) Moreover, the authors claim they are using the same amounts of cDNA for qPCRs since they have quantified the amounts using a Nanodrop. Given that dNTPs are used during cDNA synthesis, and high levels remain after cDNA synthesis from mRNA, it is not possible to accurately measure cDNA levels without first cleaning it from the residual dNTPs. Therefore, I recommend that the authors clarify this point to determine how they actually performed the qPCRs. I also recommend using two other reference genes like 18S and TATA Binding Protein alongside GAPDH, calculating the geometric mean of the three to correctly apply the 2^-ΔΔCt formula.

(7) Similarly, given that the newly generated antibodies have not been validated, I recommend introducing appropriate controls for the validation of in-cell Western assays.

(8) The authors' conclusion that TrkB levels are minimal (Figure 2E) raises questions about what they are actually detecting in the previous experiments might not be the TrkB-Fl form. Therefore, it is essential to demonstrate beyond any doubt that both the antibodies used to detect TrkB and the primers used for qPCR are correct, and in the latter case, specify at which cycle (Ct) the basal detection of TrkB transcripts occurs. Treatment with TNF-alpha for 14 days could lead to increased cell proliferation or differentiation, potentially increasing overall TrkB transcript levels due to the number of cells in culture, not necessarily an increase in TrkB transcripts per cell.

(9) Overall, there are reasonable doubts about whether the authors are actually detecting TrkB in the first three images, as well as the phosphorylation levels and localization of this receptor in the cells. For example, in Figure 3 A to J, it is not clear where TrkB is expressed, necessitating better resolution images and a magnified image to show in which cellular structure TrkB is expressed.

(10) In Figure 4, the authors indicate they have generated cells overexpressing BDNF after recombination using CRISPR technology. However, the WB they show in Figure 4B, performed under denaturing conditions, displays a band at approximately 28kDa. This WB is absolutely incorrect with all published data on BDNF detection via this technique. I believe the authors should demonstrate BDNF presence by showing a WB with appropriate controls and BDNF appearing at 14kDa to assume they are indeed detecting BDNF and that the cells are producing and secreting it. What antibodies have been used by the authors to detect BDNF? Have the authors validated it? There are some studies reporting the lack of specificity of certain commercial BDNF antibodies, therefore it is necessary to show that the authors are convincingly detecting BDNF.

(11) While the RNA sequencing data indicate changes in gene expression in cells treated with TNFalpha+CTX-B compared to control, the authors do not show a direct relationship between these genetic modifications with the rest of their manuscript's argument. I believe the results from these RNA sequencing assays should be put into the context of BDNF and TrkB, indicating which genes in this signaling pathway are or are not regulated, and their importance in this context.

Reviewer #1 (Public review):

Summary:

This paper proposes a new set of local synaptic plasticity rules that differs from classic rules in two regards: First, working under the assumption that signals coming into synapses change smoothly over time and thus have temporal correlations such that immediate activity is positively correlated with subsequent activity, it proposes both fast plasticity that immediately corrects errors as well as slower plasticity. Second, it derives these rules from optimal, Bayesian control theory principles that, even without the fast component of plasticity, are shown to provide more accurate performance than classic, non-Bayesian plasticity rules. As a proof of principle, it applies these to a simple cerebellar learning example that demonstrates how the proposed rules lead to learning performance that exceeds that achieved with classic cerebellar learning rules. The work also provides a potential normative explanation for post-climbing fiber spike pauses in Purkinje cell firing and proposes testable predictions for cerebellar experiments. Overall, I found the idea to be compelling and potentially broadly applicable across many systems. Further, I thought the work was a rare, very beautiful display of the application of optimal control theory to fundamental problems in neuroscience. My comments are all relatively minor and more expressions of interest than criticism.

Comments:

(1) The algorithm assumes, reasonably, that inputs are relatively smooth. However, I was wondering if this could make additional experimental predictions for the system being exceptionally noisy or otherwise behaving in signature ways if one were able to train a real biological network to match a rapidly changing or non-smooth function that does not align with the underlying assumptions of the model.

(2) The algorithm assumes that one can, to a good approximation, replace individual input rates by their across-synapse average. How sensitive is the learning to this assumption, as one might imagine scenarios where a neuron is sensitive to different inputs for different tasks or contexts so that a grand average might not be correct? Or, the functional number of inputs driving the output might be relatively low or otherwise highly fluctuating and less easily averaged over.

(3) On the cerebellar example, it is nice that the Bayesian example provides a narrower PF-CF interval for plasticity than the classical rules, but the window is not nearly as narrow as the Suvrathan et al. 2016 paper cited by the authors. Maybe this is something special about that system having well-defined, delayed feedback, but (optional) further comments or insights would be welcome if available.

(4) In the discussion, I appreciated the comparison with the Deneve work which has fast and slow feedback components. I was curious whether, although non-local, there were also conceptual similarities with FORCE learning in which there is also an immediate correction of activity through fast changing of synaptic weights, which then aids the slow long-term learning of synaptic weights.

Reviewer #2 (Public review):

Summary:

Hurst et al. developed a new Tol2-based transgenesis system ImPaqT, an Immunological toolkit for PaqCl-based Golden Gate Assembly of Tol2 Transgenes, to facilitate the production of transgenic zebrafish lines. This Golden Gate assembly-based approach relies on only a short 4-base pair overhang sequence in their final construct, and the insertion construct and backbone vector can be assembled in a single-tube reaction using PaqCl and ligase. This approach can also be expandable by introducing new overhang sequences while maintaining compatibility with existing ImPaqT constructs, allowing users to add fragments as needed.

Strengths:

The generation of several lines of transgenic zebrafish for the immunologic study demonstrates the feasibility of the ImPaqT in vivo. The lineage tracing of macrophages by LPS injection shows this approach's functionality, validating its usage in vivo.

Weaknesses:

(1) There is no quantitative data analysis showing the percentage of off-target based on these 4-bp overhang sequences.

(2) There is no statement for the upper limitation of the expandability.

(3) There is no data about any potential side effect on their endogenous function of promoter/protein of interest with the ImPaqT method.

Reviewer #1 (Public review):

Summary:

Optogenetic tools enable very precise spatiotemporal control of the signaling pathway. The authors developed an optimized light-regulated PKC epsilon, Opto-PKCepsilon using AlphaFold for rational design. Interactome and phosphoproteome studies of light-activated Opto-PKCepsilon confirmed a high similarity of interaction partners to PMA-stimulated wild-type PKCepsilon and high specificity for PKCepsilon substrates. Light-dependent recruitment of Opto-PKCepsilon to the plasma membrane revealed the specific phosphorylation of the insulin receptor at Thr 1160 and recruitment to mitochondria the phosphorylation of the complex I subunit NDUFS4 correlating with reduced spare respiratory capacity, respectively. The interactome and phosphoproteome studies confirm the functionality of Opto-PKCepsilon.

Strengths:

AlphaFold simulations enable the design of an optimized Opto-PKCepsilon with respect to dark-light activity. Opto-PKCepsilon is a versatile tool to study the function of PKCepsilon in a precisely controlled manner.

Weaknesses:

Light-controlled PCKepsilon was recently reported by Gada et al. (2022). Ong et al. developed an optimized Opto-PKCepsilon and presented in their manuscript the potential of this tool for controlling signaling pathways. However, some data have to be improved and appropriate controls are still missing for some experiments.

Major comments:

(1) The group of proteins detected as phosphorylated PKC substrates (phospho-Ser PKC substrate antibody) induced by Opto-PKCepsilon varies significantly between Figure 1 C and Figure 2 C. Have the authors any explanation for this? Do both figures show similar areas of the membrane? The size marker indicates that this is not the case.

(2) The ratio of endogenous and exogeneous PCKepsilon is quite different in the experiments shown in Figure 1 C and Figure 2 C. What is the reason for this effect?

(3) In addition to the overall phosphorylation of PKC substrates, the PKCepsilon mutants should be tested for phosphorylation of a known PKCepsilon substrate. The phosphorylation of the insulin receptor at Thr 1160 by Opto-PKCepsilon (see Figure 6) is very convincing and would provide clearer results for comparing the mutants.

(4) The quality of the fluorescence images shown in Figure 5 is poor and should be improved. In addition, a MitoTracker dye for mitochondria labeling should be included to confirm the mitochondrial localization of Opto-PKCepsilon.

(5) Figure S6 shows a light experiment in the absence of insulin, as stated in the headline of the figure legend and in the main text. Does this mean that Figure 6B shows an experiment in which the cells were exposed to light in the presence of insulin? If so, this should be mentioned in the legend of the figure and in the main text. What influence does insulin have on IR phosphorylation at Thr 1160?

(6) The signal of NDUSF4 phosphorylation induced by Opto-PKCepsilon is weak in the experiment shown in Figure 7E. What about the effect of shorter and longer exposure times? How many times was this experiment repeated?

Reviewer #2 (Public review):

Summary:

The authors developed an optogenetic tool (Opto-PKCε) and demonstrated spatiotemporal control of optoPKCε at different subcellular compartments such as the plasma membrane or mitochondria. Signaling outcomes of optoPKCε were characterized by phosphoproteomics and biochemical analysis of downstream signaling effectors.

Strengths:

(1) Conventional strategy to activate PKC often involves activation of multiple downstream signaling pathways. This work showcases an alternative strategy that could help dissect the effect of specific PKC-elicited signaling outcomes.

(2) The differential phosphoproteomic analysis of PKC substrates between PMA stimulation and optoPKCε activation is insightful. A follow-up question is whether co-transfection of CIBN-GFP-CaaX and optoPKCε increases the pool of substrate compared to optoPKCε only, or optoPKCε activation at the plasma membrane is more effective in phosphorylating its substrates?

(3) The finding that PKC activation at the plasma membrane is required for insulin receptor activation is interesting. Why does Thr1160 phosphorylation lead to a reduction of Thr1158/1162/1163? Does "insulin-stimulated" imply that insulin was administrated in the culture during optogenetic stimulation? Also, did the author observe any insulin receptor endocytosis upon optoPKCε activation?

Weaknesses:

(1) When citing the previous work on optogenetics, the reviewer believes a broader scope of papers (reviews) and recent research articles should be cited, especially those that used similar strategies, i.e., membrane translocation followed by oligomerization (of cryptochrome), as reported in this work.

(2) In terms of molecular modeling, how would the author enable AlphaFold3 structure prediction of activated optoPKCε (or the blue-light stimulated state of cryptochrome)? Current methods only describe that "To generate models of the monomer, an amino acid sequence corresponding to Opto-PKCɛ, 2 ATPs and 1 FAD were used as input whereas for the tetramer, copies of Opto-PKCɛ, 8 ATPs and 4 FADs were used as input" (likely missing "four" between "tetramer" and "copies"). However, simply putting four monomers would not ensure that each monomer is in the "activated" state, which involves excitation of the FAD cofactor and likely conformational changes in cryptochrome.

(3) It would be helpful if the authors could help interpret some results. For example, Figure S1: Was the puncta of mCherry-PKCε on the plasma membrane or within the cytosol? Also, why does optoPKCε only work when PKCε is fused at the C-terminus? When screening for the optoPKCε system with the largest light-to-dark contrast, the AGC domain was truncated. What is the physiological function of AGC? Does AGC removal limit PKC's access to its endogenous substrates?

Reviewer #2 (Public review):

The provided evidence in the study by MacQueen and colleagues is convincing, albeit some methodological challenges still exist. The authors rightly state that different subpopulations are likely to have evolved distinct patterns of GxE. It has been recently shown that the genetic architecture for adaptive traits differs across subpopulations (Lopez-Arboleda et al. 2021), hence this effect should be even more pronounced for GxE. How to best account for this in a statistical framework is not utterly clear. Here the authors describe their efforts to asses these interactions and to estimate the magnitude of the respective effects. Building on the statistical framework described, it could be possible to translate their findings from switchgrass to other species. A plus of the study is the effort to use an independent pseudo-F2 population to confirm the found associations.<br /> The manuscript is written coherently and all data and code used is freely available and explained in detail in the supplementary information.

Nevertheless, I feel that there are some points in the data analysis that could be clarified some more.

(1) Dividing GxE interactions into discrete, measurable GxWeather terms is a nice idea to gain a reliable measurement of E. I also appreciate the effort to create date-related values as a summary function of a weather variable across a specified date range. Using cumulative data the week prior to flowering seems like a good choice to associate weather patterns to this phenotype, but there are many - including non-linear ways - to accumulate these data. Additionally, weather parameters like temperature and precipitation can show interaction effects. I wonder if there is a way to consider these.

(2) As pointed out in Section S1, a trait measured in eight common gardens could be modeled at eight genetically correlated traits. To assess the genetic correlation one would need to estimate the genetic variance within each trait and 28 genetic covariance structures. Here model convergence would be painful given the sample sizes. There are different statistical solutions for this including the mash algorithm the authors choose. I highly appreciate the effort in how the rationale is described in the supplementary information, but to me, it is still not completely clear how 'strong' and random effects have been selected from GWAS. How sensitive is the model to a selection of different effects? Could one run permutations to assess this? Why is the number of total markers different for different phenotypes and subsets and does this affect statistical power?

(3) The mash model chooses different covariance matrices for the different analyses. Although I do understand the rationale for this, I am not sure how this will impact the respective analysis and how comparable the results are. Would one not like to have the same covariance matrices selected for all analyses?

(4) Although the observed pattern of different GxE in different subpopulations is intriguing, it remains a little unclear what we actually learn apart from the fact that GxE in adaptive traits is complex. Figure 3 divides GxE into sign and magnitude effects. Interestingly the partition differs significantly between Greenup date and Flowering Date. Still, the respective QTLs in Figure 4 do - at least partially - overlap (e.g. on CHR05N). What is the interpretation of these? Here, I would appreciate a more detailed discussion and hearing the thoughts of the authors.

(5) Figure 4 states that Stars indicate QTLs with significant enrichment for SNPs in the 1% mash tail. The shown Rug plots indicate this, but unfortunately, I am missing the respective stars. Is there a way to identify what is underlying these QTLs?

To summarize, the manuscript nicely shows the complex nature of GxE in different switchgrass subpopulations. The goal now would be to identify the causative alleles for these phenomena and understand how these have evolved. Here the provided study paves the way for further analyses in this perspective.

This image explains how modern evolutionary theory interprets prosocial behavior—actions that benefit others.

Key idea: • Modern evolutionary theory proposes that natural selection operates at the level of genes, not necessarily at the level of the whole individual. • In this view, behaviors that increase the likelihood of a gene’s survival will be favored by evolution—even if they don’t benefit the individual directly.

How does this relate to helping others?

Helping others can increase the chances that your genes survive, especially if: 1. You help close relatives (they share many of your genes). 2. You help others who help you in return (reciprocal altruism). 3. You work cooperatively in groups, which improves survival for everyone involved. 4. You build a good reputation, which can bring future social or material benefits.

So even though helping others might seem selfless, it can still serve your genetic self-interest in the long run.

Let me know if you’d like a real-world example or a simpler summary!

This work has been published in GigaByte Journal under a CC-BY 4.0 license (https://doi.org/10.46471/gigabyte.152). These reviews (including a protocol review) are as follows.

Reviewer 1. Jeet Sukumaran

Is the documentation provided clear and user friendly? Yes. Excellent documentation. A pleasure to read. Are there (ideally real world) examples demonstrating use of the software? No.

Reviewer 2. Ziqi Deng

Is installation/deployment sufficiently outlined in the paper and documentation, and does it proceed as outlined? Yes. I'm able to run all the tests and used CompactTree c++ correctly except for encounter issue installation installing Python Wrapper via pip install CompactTree.

Are there (ideally real world) examples demonstrating use of the software? Yes. CompactTree has provided examples of simulated trees for testing comparing to other peer packages. In the meanwhile it mentioned its ability to load the ~22M nodes greengenes2 tree. It would be great to see the test workflow so users can verify.

Additional Comments: CompactTree is aimed at a very specific task, that of loading large phylogenetic trees with millions of nodes. The result shows that it is significantly faster than the other peer tools not only in loading but also in traversing trees, with less peak memory usage. It also includes the test workflow for users to repeat the test in comparison with other peer tools.

Reviewer 3. Giorgio Bianchini

Is the language of sufficient quality? Yes. It is slightly confusing that the paper is written using plural pronouns ("We"), when there is a single author.

Is there a clear statement of need explaining what problems the software is designed to solve and who the target audience is? No. The statement of need is present; however, it does not clearly explain what kinds of problems the software will be able to solve, beyond generic statements about addressing scalability issues. The aims of the library should be explored in more detail: as noted by the author, this library offers great speed and efficiency, but at the cost of reduced flexibility and functionality compared to other tools. Speed and efficiency are always good things, but what does the library actually do? A very fast library that does nothing is not particularly useful. So, what specific analyses does CompactTree allow, that would be impractical using other tools? For example, they could select a case study from the literature, where the analyses were limited by the algorithm, and use their library to extend the analysis to a larger dataset. The author mentions clustering, ancestral state reconstruction, and transmission risk prediction as examples of analyses that involve tree traversals, so they could start here (although I am not convinced that the efficiency of the tree representation is the computational bottleneck in these cases). The results should also be briefly mentioned in the abstract. Furthermore, the author mentions a number of packages used to analyse trees, but these are all Python packages. Since CompactTree is presented as a C++ library, this seems odd; other tools and programming languages should be mentioned/compared. For example, “ape” and “phytools” are very popular R packages, while “Bio++” is another C++ library; a literature review (or a simple web search) may reveal other such libraries. Also, the reference given for bp (“[4]”) is incorrect.

Is installation/deployment sufficiently outlined in the paper and documentation, and does it proceed as outlined? Yes. Everything works fine if the header is included in a single source file, but if multiple distinct files contain the #include statement, a compilation error will occur due to the multiple definitions. In a real-world application, the library would reasonably need to be included in multiple source files, so this should be fixed.

Is the documentation provided clear and user friendly? Yes. The documentation "Cookbook" is very nicely organised.

Have any claims of performance been sufficiently tested and compared to other commonly-used packages? No. While the author compares CompactTree to a number of Python packages, no comparison is made against tools that use other programming languages. In particular, the author states that there is no C++ library for loading and traversing phylogenetic trees; however, as I mentioned, at least Bio++ exists and appears to be reasonably well cited. Furthermore, the memory plot does not consider the baseline memory usage. This is evident in the first two datapoints (n=100 and n=1000) for each tool, which show a very small difference, despite the leaf count increasing by an order of magnitude. If the first datapoint is subtracted from all subsequent datapoints, the memory plot looks quite similar to the other plots. If you re-run the benchmarks to include other tools, I would suggest including a “control” datapoint with a very small n (or even, loading the library without opening a tree), and subtracting this from all other datapoints; this will provide an estimate of the memory actually used to load the trees.

Are there (ideally real world) examples demonstrating use of the software? No. As I mentioned above, having at least one example demonstrating an analysis that is significantly improved by the use of this library would be beneficial. Discussion of the improvements should also consider usability trade-offs in a real-world scenario.

Is automated testing used or are there manual steps described so that the functionality of the software can be verified? No.

Additional Comments: The library looks promising and is reasonably well documented, the only two things that are really missing are a real-world practical application and a comparison with other relevant alternatives (especially Bio++). A large portion of the manuscript is spent describing how the library could be improved, rather than what it can currently do. This could be summarised in just one or two sentences, thus leaving more space for describing the real-world example.

La columna de la izquierda son inhibidores y la de la derecha, inductores

Een richtlijn kan alleen rechtstreekse werking hebben als: 1. De implementatietermijn is verstreken zonder correcte omzetting 2. De bepaling voldoende duidelijk, precies en onvoorwaardelijk is 3. Het gaat om een verticale verhouding (burger tegen staat of overheidsinstantie, niet tussen burgers) – zie Faccini Dori.

Application: Verticale verhouding? Ja, Mario staat tegenover de overheid (justitiële autoriteiten).

Implementatietermijn verstreken? Ja, de richtlijn moest uiterlijk op 25 mei 2019 zijn omgezet.

Duidelijk en onvoorwaardelijk? Nee. Artikel 4 vereist dat autoriteiten beoordelen of:

Mario onvoldoende middelen heeft; De belangen van de rechtspleging rechtsbijstand vereisen. ➤ Dit zijn open normen met beoordelingsruimte voor de lidstaat. ➤ Het Hof oordeelt in vergelijkbare gevallen dat zulke bepalingen te vaag zijn voor rechtstreekse werking (vgl. Kücükdeveci en Dominguez).

考的就是TM的边界值

最好的就是左闭右开

解法：

1. 先统计绕几圈，是奇数还是偶数

如果偶数圈，那么不用管后续； 如果奇数圈，最后记得赋值最中心的格子

2. 再设置偏移的offset，一圈转完，横纵都进行偏移

3. while offset<maxoffset:

(1) 第一行赋值 (2) 最后一列赋值 (3) 最后一行赋值 (4) 第一列赋值

O STF admitiu a possibilidade de habeas corpus coletivo.

O habeas corpus se presta a salvaguardar a liberdade. Assim, se o bem jurídico ofendido é o direito de ir e vir, quer pessoal, quer de um grupo determinado de pessoas, o instrumento processual para resgatá-lo é o habeas corpus, individual ou coletivo.

A ideia de admitir a existência de habeas corpus coletivo está de acordo com a tradição jurídica nacional de conferir a maior amplitude possível ao remédio heroico (doutrina brasileira do habeas corpus).

Apesar de não haver uma previsão expressa no ordenamento jurídico, existem dois dispositivos legais que, indiretamente, revelam a possibilidade de habeas corpus coletivo. Trata-se do art. 654, § 2º e do art. 580, ambos do CPP.

O art. 654, § 2º estabelece que compete aos juízes e tribunais expedir ordem de habeas corpus de ofício. O art. 580 do CPP, por sua vez, permite que a ordem concedida em determinado habeas corpus seja estendida para todos que se encontram na mesma situação.

Assim, conclui-se que os juízes ou Tribunais podem estender para todos que se encontrem na mesma situação a ordem de habeas corpus concedida individualmente em favor de uma pessoa.

Existem mais de 100 milhões de processos no Poder Judiciário, a cargo de pouco mais de 16 mil juízes, exigindo do STF que prestigie remédios processuais de natureza coletiva com o objetivo de emprestar a máxima eficácia ao mandamento constitucional da razoável duração do processo e ao princípio universal da efetividade da prestação jurisdicional.

Diante da inexistência de regramento legal, o STF entendeu que se deve aplicar, por analogia, o art. 12 da Lei nº 13.300/2016, que trata sobre os legitimados para propor mandado de injunção coletivo.

Assim, possuem legitimidade para impetrar habeas corpus coletivo:

1) o Ministério Público;

2) o partido político com representação no Congresso Nacional;

3) a organização sindical, entidade de classe ou associação legalmente constituída e em funcionamento há pelo menos 1 (um) ano;

4) a Defensoria Pública.

STF. 2ª Turma.HC 143641/SP. Rel. Min. Ricardo Lewandowski, julgado em 20/2/2018 (Info 891).

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #3

Evidence, reproducibility and clarity

Using human pluripotent stem cell-based Gel-3D model of amniogenesis this study investigates the transcriptional dynamics of amnion differentiation at single cell level. Seven cell clusters are identified that emerge over four days of differentiation, including progressive amniotic fates precursors, among them CLDN10 progenitor located on the boundary of amnion and epiblast, and primordial germ cell-like fate. Mutational studies support the role of CLDN10 in promoting amniotic but limiting primordial germ cell-like.

Major Comments:

This generally clearly presented study significantly advances our understanding of human/NHP amniogenesis and should be of broad interest with relevance to human reproduction. However, the there are several questions about how experiments were performed, analyzed, presented and interpreted that need to be answered.

1. The presented antibody stainings while beautiful are presented without sufficient quantification. A single representative (?) cyst is shown. Please provide information about how many cysts on average have been analyzed in how many experiments, expression levels of should be quantified to support conclusions such as: Line 116-118 "a subset of cells within the cysts displays reduced expression of NANOG, while TFAP2A expression becomes weakly activated"; or Line 129 "the transition from pluripotent to amnion cell types occurs progressively over the cyst, starting from focal initiation sites".
2. As the the scRNA-seq experiment is one of the main advances of this study and it explores the temporal dynamics and transitional cell populations during amniogenesis this experiment should be performed with two independent biological replicates to investigate the variability of the amniogenesis in this model in terms of the proportion of the 7 distinct cell populations the authors identified in this analysis.
3. Another interesting parallel between the amnion model and the CS7 human gastrula is most Tyser "Epiblast" cells are seen in the "pluripotency-exiting" population of the amnion model. However, pluripotency exit is a hallmark of epiblast as it initiates gastrulation and primitive streak formation/mesendoderm differentiation. This should be analyzed and discussed further, especially that the authors see in the amnion model some cells expressing TBXT at low level.
4. How do the authors explain/interpret the difference in CLDN10 expression at RNA and protein level?
5. Two hESC CLDN10 mutant lines are presented in Figure S4, which are transheterozygous for framesfhit mutations. However, it is not clear how (guideRNAs), in which position of the gene these mutations were generated and what is predicted mutant protein product of each allele. Please provide, gene structure, gRNA position and predicted protein product cartoons. As we do not know the antigen recognized by CLDN10 antibody, these are critical considerations.
6. What are the consequences of these mutations on CLDN10 transcript? qPCR and also scRNA-seq data the authors have.
7. Please indicate in the experiments using CLDN10 mutant lines, which KO line has been used for specific experiment and whether same/different results have been obtained with the two lines.
8. The excess of PGCL cells in CLDN10 KO Gel-3D amnion model is an important observation, but not fully supported by the data. We are presented with single images of mutant cysts at different stages of amniogenesis. Additional data and the number of SOX17+ cells in WT and mutant cysts at should be provided.
9. The authors propose an interesting concept of CLDN10 at the boundary between the amnion and the epiblast promoting amniogenesis and limiting hHPGLC formation. They speculate about the role of tight junction in this process in agreement with increased hHPGLC formation upon ZO1 reduction in another hPSC model. However, surprisingly little discussion is provided about signaling implications of the reported amniogenic transcriptional cascade, and signals emanating from the different amnion progression cell types. Given the important role of BMP in the formation of amnion and hPGCs, notable is increasing expression of BAMBI in progenitor cell types and high expression in specified and maturing clusters. The expression of signaling pathway components should be analyzed and discussed in more depth.

Additional comments:

1. It is not easy to discern the numbers of the seven populations that are detected at D1-D4 from Figure 1C. A panel in Figure 1 illustrating this would be informative.
2. The similarity of the "Ectoderm" cluster from the CS7 human gastrula Tyser et al., 2021 to extraembryonic cell type with amnion/trophectoderm characteristics in hESC 2D-gastruloid model has been reported by Minn et al., Stem Cell Reports, 2021 and this should be acknowledged.

Referees cross-commenting

There is consensus among the reviewers that this is a novel and important work, but additional experiments and their rigorous quantification is needed. Attending to the reviewers comments will significantly elevate this exciting work.

Significance

Occurring upon implantation of human blastocyst, amniogenesis, or formation of the amniotic sac from the pluripotent epiblast, is still poorly understood but essential process of human embryogenesis. The key morphogenetic aspects of amniogenesis, i.e. epithelial polarization of epiblast into a cyst and subsequently differentiation of the portion of the cyst abutting the trophectoderm proximal to the uterus into squamous epithelium is in part modeled by the hESC-based amnion models in which BMP stimulation plays a crucial role. In the Gel-3D amnion model model deployed here, no exogenous BMP is added, however, BMP signaling is activated in the cells by a mechanosensitive cue provided by the soft substrate; hESCs initially form a cyst of epithelial cells expressing pluripotent markers that initiate transcriptional cascade and within 4 days of culture differentiate into a cyst of squamous-amnion-like epithelium.

This work expands on the previous studies by investigating the transcriptional dynamics of amnion differentiation at single cell level combined with additional antibody stainings and compare their findings to distinct cell types in a Carnegie stage 7 human embryo (Tyser et al., 2021) and relevant non-human primate datasets. Based on the resulting data the authors posit contiguous amniogenic cell states: pluripotency-exiting, early progenitor, late progenitor, specified and maturing. Moreover, they also uncover that this model of amniogenesis also produces primordial germ cell-like (hPGC-L) and mesoderm-like cells. A notable finding is that high levels of CLDN10 mark a later transient progenitor state, but CLDN10 expression is downregulated more differentiated cells. Moroever, the authors posit that CLDN10 is a marker of the progenitor population, expression of which is restricted to the boundary between the amnion and the epiblast of the cynomolgus macaque peri-gastrula. Functional interrogation of CLDN10 using hESC mutant lines in the Gel-3D amnion model shows reduced amniogenesis and excess of hPGC-L cells. The authors propose that the CLDN10 the amnion-epiblast boundary is a site of active amniogenesis but limits hPGC-L. This work advances our understanding of amniogenesis, strengthens the concept that amnion and PGC progressing cells initially share acommon intermediate lineage, provides a valuable transcriptomic dataset and should be of broad interest with relevance to human development and reproduction.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Sekulovski et al characterize the transcriptome of human pluripotent stem cells differentiating to an amnion fate in 3D, using single-cell RNA sequencing. This leads to the identification of CLDN10+ cells as amnion progenitors. When CLDN10 is eliminated, amniogenesis is compromised. Moreover, analysis of CLDN10 localization in cynomolgus macaque embryos reveals that this progenitor population is located at the boundary between the epiblast and the amnion.

Major comments:

The key results are convincing and supported by clear experiments. However, additional controls, quantifications, and clarifications are needed as follows:

• The authors identify five amnion-progressing states in vitro and mention that each of these states also shows transcriptional similarities to cell types in a CS7 embryo (Tyser et al, 2021). How do the authors interpret this result? Would this mean that there are amnion cells at all different maturation stages present at a specific time point in development? Given that the available in vivo reference is derived from a single human embryo, it is more likely that the true in vivo counterpart of these states is not captured in the embryo data.
• The authors stain the 3D amnion model at different stages and conclude that "amniogenesis initiates focally and spreads laterally". This cannot be concluded from the data provided. The images in Figure 1 simply show heterogeneity in the levels of TFAP2A. To support their claims, the authors would need to perform time-lapse experiments using a TFAP2A reporter line.
• The authors conclude that CLDN10+ cells give rise to amnion during gastrulation of cynomolgus macaque embryos. The data provided does not prove that CLDN10+ cells are the amnion progenitors in vivo.
• CLDN10 KO cells form amnion cysts like control cells by day 3. However, by day 4 the cysts lose expression of the amnion marker ISL1 and become disorganized. To characterize the epithelial (or lack of) phenotype, the authors should include membrane/polarity/adhesion immunostainings. Is the disorganization observed at day 4 associated with the progressive changes in cell identity, or is it a time-dependent phenotype? The authors should include human PSC cysts as a control. This would allow them to determine whether the role of CLDN10 is specific to amnion cells.
• Figure 2: is there a correlation between the levels of CLDN10 and TFAP2A based on the scRNAseq data and the immunofluorescence stainings? The IF data would benefit from quantifications.
• Figure 4: the experiment has not been quantified. What is the % of PGCLCs in WT and KO cells? What are the levels of ISL1 in WT and KO cells? What is the localization of epithelial determinants in WT and KO cells? Is there an anti-correlation between CLDN10 and ISL1?

Referees cross-commenting

I think there is a general consensus that additional quantifications and careful analyses are needed before this paper is accepted for publication. I agree with the comments raised by the other reviewers.

Significance

This manuscript is a follow-up work of Sekulovski et al, 2024. In this recent manuscript, the authors already provided a temporally resolved transcriptomic characterization of in vitro amniogenesis. The key difference between the two articles is that while Sekulovski et al, 2024 performed a bulk RNAseq experiment, in the current manuscript a single-cell RNAseq experiment has been done. It is fundamental to clearly define what new findings have been obtained thanks to the single-cell experiment, which could not have been obtained using the bulk transcriptomics data. This is a particularly important point given the robustness and synchrony of the model. For example, the authors had already identified five amnion states in vitro in their previous publication. Is CLDN10 differentially expressed in the progenitor population based on the bulk RNAseq data? Are the same dynamics of expression recapitulated? The title of the manuscript does not mention CLDN10 but rather focuses on transcriptional profiling at the single-cell level. In my opinion, the key novelty of this manuscript is the identification of CLDN10 and the role it plays during amniogenesis. Focusing the manuscript on the dynamic transcriptional profile diminishes the novelty, as this had already been done by the authors at the bulk level. Globally, this manuscript provides additional information of the poorly understood process of amniogenesis that will be interesting for those working on early human embryogenesis.

My area of expertise is early mammalian embryo development and stem cells. I do not have the computational background to evaluate the bioinformatic analyses of the manuscript in-depth.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This is a novel and important and interesting study that uses one of the best amniogeneisis form PSCs int he field. The authors do scRNA-seq during 4 day course to understand different populations emerging during amniogeneisis, and they identify CLDN10 as a marker for newly emerging new amion cells, and then use their model and monkey real embryos to prove the CLDN10+ population at the amnion-epiblast border. In the final part, the authors knockout CLDN10 and claim it compromises amniogenesis and favours formation.

Significance

This is a well conducted study, and conclusions are novel and super exciting and IMPORTANT!!!. I have one-2 major comments to strengthen conclusions in the last part, and will help make this excellent study become superb and a landmark study.

1. it is not really clear what is the phenotype of CLDN10 KO cells. is amniogenesis totally inhibited? can the authors do scRNA-seq on the KO cells and compare them to WT cells? There is no quantitation to amnion or PGC formation efficiency ? how many structures where analyzed?
2. in continuation with the above The claim that PGC formation is enhanced in KO is not strong. PGCs should be stained for NANOS3 and blimp1 specific marker and not only SOX17 which can also be a Pre marker. Then quantification should be properly done.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This work shows that a specific adenosine deaminase protein in Dictyostelium generates the ammonia that is required for tip formation during Dictyostelium development. Cells with an insertion in the ADGF gene aggregate but do not form tips. A remarkable result, shown in several different ways, is that the ADGF mutant can be rescued by exposing the mutant to ammonia gas. The authors also describe other phenotypes of the ADGF mutant such as increased mound size, altered cAMP signalling, and abnormal cell type differentiation. It appears that the ADGF mutant has defects in the expression of a large number of genes, resulting in not only the tip defect but also the mound size, cAMP signalling, and differentiation phenotypes.

Strengths:

The data and statistics are excellent.

Weaknesses:

(1) The key weakness is understanding why the cells bother to use a diffusible gas like ammonia as a signal to form a tip and continue development.

Diffusion of a gas can affect the signalling process of the entire colony of cells and will be quicker than other signaling mechanisms. A number of findings suggest that ammonia acts as both a local and long-range regulatory signal, integrating environmental and cellular cues to coordinate multicellular development. Ammonia serves as a crucial signalling molecule, influencing both multicellular organization and differentiation in Dictyostelium (Francis, 1964; Bonner et al., 1989; Bradbury and Gross, 1989). By raising the pH of the intracellular acidic vesicles of prestalk cells (Poole and Ohkuma, 1981; Gross et al, 1983), and the cytoplasm, ammonia is known to increase the speed of chemotaxing amoebae (Siegert and Weijer 1989; Van Duijn and Inouye, 1991), triggering multicellular movement (Bonner et al., 1988, 1989) to favor tipped mound development. The slug tip is known to release ammonia while the slime sheath at the back of the slug prevents diffusion thus maintaining high ammonia levels to (Bonner et al., 1989) promote pre-spore differentiation (Newell et al., 1969). Ammonia has been found to favor slug migration rather than fruiting (Schindler and Sussman, 1977) and thus, tip-derived ammonia may stimulate synchronized development of the entire colony. The tip exerts negative chemotaxis towards ammonia, potentially directing the slugs away from each other to ensure equal spacing of fruiting bodies (Feit and Sollitto, 1987).  

Ammonia released in pulses acts as a long-distance signalling molecule between colonies of yeast cells indicating depletion of nutrient resources and promoting synchronous development (Palkova et al., 1997; Palkova and Forstova, 2000). A similar mechanism may be at play to influence neighbouring Dictyostelium colonies. Furthermore, ammonia produced in millimolar concentrations (Schindler and Sussman, 1977) may also ward off predators in soil as observed in Streptomyces symbionts of leaf-cutting ants to inhibit fungal pathogens (Dhodary and Spiteller, 2021). Additionally, ammonia may be recycled into amino acids, within starving Dictyostelium cells to supporting survival and differentiation as observed in breast cancer cells (Spinelli et al., 2017). Therefore, using a diffusible gas like ammonia as a signalling molecule is likely to have bioenergetic advantages. Ammonia is a natural metabolic byproduct of amino acid catabolism and other cellular processes, making it readily available without requiring additional energy for synthesis. Instead of producing a dedicated signalling molecule, cells can exploit an existing by-product for developmental regulation.

(2) The rescue of the mutant by adding ammonia gas to the entire culture indicates that ammonia conveys no positional information within the mound.

Ammonia is known to influence rapid patterning of Dictyostelium cells confined in a restricted environment (Sawai et al., 2002). Both neutral red staining (a marker for prestalk and ALCs) (Fig. S2) and the prestalk marker ecmA/ ecmB expression (Fig. 8C) in the adgf mutants suggest that the mounds have differentiated prestalk cells but are blocked in development. The mound arrest phenotype can be reversed by exposing the adgf mutant mounds to ammonia.  

Based on cell cycle phases, there exists a dichotomy of cell types, that biases cell fate to prestalk or prespore (Weeks and Weijer, 1994; Jang and Gomer, 2011). Prestalk cells are enriched in acidic vesicles, and ammonia, by raising the pH of these vesicles and the cytoplasm (Davies et al 1993; Van Duijn and Inouye 1991), plays an active role in collective cell movement (Bonner et al., 1989). Thus, ammonia reinforces or maintains the positional information by elevating cAMP levels, favouring prespore differentiation (Bradbury and Gross, 1989; Riley and Barclay, 1990; Hopper et al., 1993). 

(3) By the time the cells have formed a mound, the cells have been starving for several hours, and desperately need to form a fruiting body to disperse some of themselves as spores, and thus need to form a tip no matter what.

When the adgf mutants were exposed to ammonia just after tight mound formation, tips developed within 4 h (Fig. 6). In contrast, adgf mounds not exposed to ammonia remained at the mound stage for at least 30 h. This demonstrates that starvation alone is not sufficient to drive tip development and ammonia serves as a cue that promotes the transition from mound to tipped mound formation. 

Many mound arrest mutants are blocked in development and do not proceed to form fruiting bodies (Carrin et al., 1994). Furthermore, not all the mound arrest mutants tested in this study were rescued by ADA enzyme (Fig. S3 A), and they continue to stay as mounds without dispersing as spores, suggesting that mound arrest in Dictyostelium can result from multiple underlying defects, whereas ammonia is an important factor controlling transition from mound to tip formation.

(4) One can envision that the local ammonia concentration is possibly informing the mound that some minimal number of cells are present (assuming that the ammonia concentration is proportional to the number of cells), but probably even a minuscule fruiting body would be preferable to the cells compared to a mound. This latter idea could be easily explored by examining the fate of the ADGF cells in the mound - do they all form spores? Do some form spores?

Or perhaps the ADGF is secreted by only one cell type, and the resulting ammonia tells the mound that for some reason that cell type is not present in the mound, allowing some of the cells to transdifferentiate into the needed cell type. Thus, elucidating if all or some cells produce ADGF would greatly strengthen this puzzling story.

A fraction of adgf mounds form bulkier spore heads by the end of 36 h as shown in Fig. 3. This late recovery may be due to the expression of other ADA isoforms. Mixing WT and adgf mutant cell lines results in a slug with the mutants occupying the prestalk region (Fig. 9) suggesting that WT ADGF favours prespore differentiation. However, it is not clear if ADGF is secreted by a particular cell type, as adenosine can be produced by both cell types, and the activity of three other intracellular ADAs may vary between the cell types. To address whether adgf expression is cell type-specific, we will isolate prestalk and prespore cells, and thereafter examine adgf expression in each population.

ADGF activity is likely to be higher in the tip to remove excess adenosine, the tip-inhibiting molecule (Wang and Schaap, 1985). Moreover, our results show that adgf^- cells with high adenosine preferentially migrate to the prestalk rather than the prespore region when mixed with WT cells. Ammonia generated from adenosine deamination could thus drive tip development and prespore differentiation.

Reviewer #2 (Public review):

Summary:

The paper describes new insights into the role of adenosine deaminase-related growth factor (ADGF), an enzyme that catalyses the breakdown of adenosine into ammonia and inosine, in tip formation during Dictyostelium development. The ADGF null mutant has a pre-tip mound arrest phenotype, which can be rescued by the external addition of ammonia. Analysis suggests that the phenotype involves changes in cAMP signalling possibly involving a histidine kinase dhkD, but details remain to be resolved.

Strengths:

The generation of an ADGF mutant showed a strong mound arrest phenotype and successful rescue by external ammonia. Characterization of significant changes in cAMP signalling components, suggesting low cAMP signalling in the mutant and identification of the histidine kinase dhkD as a possible component of the transduction pathway. Identification of a change in cell type differentiation towards prestalk fate

Weaknesses:

(1) Lack of details on the developmental time course of ADGF activity and cell type type-specific differences in ADGF expression.

ADGF expression was examined at 0, 8, 12, and 16 h (Fig. 1), and the total ADA activity was assayed at 12 and 16 h (Fig. 4). As per the reviewer’s suggestion, we have now included the 12 h data (Fig. 4A) to provide additional insights into the kinetics of ADGF activity. The adgf expression was found to be highest at 16 h and hence, the ADA assay was carried out at that time point. However, the ADA assay will not exclusively reflect ADGF activity since it reports the activity of the three other isoforms as well.

A fraction of adgf^- mounds form bulkier spore heads by the end of 36 h as shown in Fig. 3. This late recovery may be due to the expression of the other ADA isoforms. Mixing WT and adgf mutant cell lines results in a slug with the mutants occupying the prestalk region (Fig. 9), suggesting that WT adgf favours prespore differentiation.

However, it’s not clear if ADGF is secreted by a particular cell type, as adenosine can be produced by both cell types, and the activity of the other three intracellular ADAs may vary between the cell types. To address whether adgf expression is cell typespecific, we will isolate prestalk and prespore cells, and thereafter examine adgf expression in each population.

ADGF activity is likely to be higher in the tip to remove excess adenosine, the tipinhibiting molecule (Wang and Schaap, 1985). Moreover, our results show that adgf^- cells with high adenosine preferentially migrate to the prestalk rather than the prespore region when mixed with WT cells.

(2) The absence of measurements to show that ammonia addition to the null mutant can rescue the proposed defects in cAMP signalling.

The cAMP levels were measured at two time points 8 h and 12 h in the mutant. The adgf mutant has lower ammonia levels (Fig. 6), diminished acaA expression (Fig. 7) and reduced cAMP levels (Fig. 7) in comparison to WT at both 12 and 16 h of development. Since ammonia is known to increase cAMP levels (Riley and Barclay, 1990; Feit et al., 2001), addition of ammonia addition to the mutant is likely to increase acaA expression, thereby rescuing the defects in cAMP signalling.

(3) No direct measurements in the dhkD mutant to show that it acts upstream of adgf in the control of changes in cAMP signalling and tip formation.

The histidine kinases dhkD and dhkC are reported to modulate phosphodiesterase RegA activity, thereby maintaining cAMP levels (Singleton et al., 1998; Singleton and Xiong, 2013). By activating RegA, dhkD ensures proper cAMP distribution within the mound, which is essential for the patterning of prestalk and prespore cells, as well as for tip formation (Singleton and Xiong, 2013). Therefore, ammonia exposure to dhkD mutants is likely to regulate cAMP signalling and thereby tip formation. We will address this issue by measuring cAMP levels in the dhkD mutant.

Author response:

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

Overview:

We appreciate all the constructive comments from the reviewer and the reviewing editor, as their suggestions have significantly improved our manuscript. In response to their comments, we have made several key revisions: First, we have performed new colocalization analyses between the active zone marker UNC-10::GFP and all UNC-13L variants (UNC13L, UNC-13L^HK, UNC-13L^D1-5N, and UNC-13L^HK+D1-5N, all tagged with mApple). These results confirm that the mutations do not affect synaptic localization. Second, we have provided a clearer explanation of the “gain-of-function” term used in this study, emphasizing that it reflects an increased SV release due to C1-C2B module dysfunction rather than a single mechanistic state. Third, we have expanded the discussion on the physiological implications of the C1-C2B model, particularly its role in regulating synaptic transmission under varying neuronal activity conditions. Finally, to improve clarity and focus, we have removed unnecessary speculative discussions, ensuring that the revised manuscript centers on the most relevant findings.

We have reorganized the manuscript to incorporate these new results into the figures and text. Full responses to all reviewer comments are provided below. We hope that the reviewer and the editor find these revisions satisfactory and that our manuscript is now suitable for publication in eLife.

Joint Public Review:

Summary:

In this manuscript, the authors investigate how different domains of the presynaptic protein UNC-13 regulate synaptic vesicle release in the nematode C. elegans. By generating numerous point mutations and domain deletions, they propose that two membrane-binding domains (C1 and C2B) can exhibit "mutual inhibition," enabling either domain to enhance or restrain transmission depending on its conformation. The authors also explore additional Nterminal regions, suggesting that these domains may modulate both miniature and evoked synaptic responses. From their electrophysiological data, they present a "functional switch" model in which UNC-13 potentially toggles between a basal state and a gain-of-function state, though the physiological basis for this switch remains partly speculative.

Strengths:

(1) The authors conduct a thorough exploration of how mutations in the C1, C2B, and other regulatory domains affect synaptic transmission. This includes single, double, and triple mutations, as well as domain truncations, yielding a large, informative dataset.

(2) The study includes systematically measuring both spontaneous and evoked synaptic currents at neuromuscular junctions, under various experimental conditions (e.g., different Ca²⁺ levels), which strengthens the reliability of their functional conclusions.

(3) Findings that different domain disruptions produce distinct effects on mEPSCs, mIPSCs, and evoked EPSCs suggest UNC-13 may adopt an elevated functional state to regulate synaptic transmission.

Weaknesses:

It remains unclear whether the various domain alterations truly converge on a single "gain-offunction" state or instead represent multiple pathways for enhancing UNC-13 activity. Different mutations selectively affect spontaneous or evoked release, suggesting that each variant may not share the same underlying mechanism. Moreover, many conclusions rely on combining domain deletions or point mutations, yet the electrophysiological data show distinct outcomes across EPSCs, IPSCs, mini, and evoked responses. This raises questions about whether these manipulations all act on the same pathway and whether their observed additivity or suppression genuinely reflects a single mechanistic process. A unifying model-or at least a clearer explanation of why the authors infer one mechanistic state across different domain manipulations would strengthen the paper's conclusions.

We appreciate the comment and understand the potential confusion regarding the use of the term "gain-of-function" in the manuscript. To clarify, the gain-of-function state described in this study does not refer to a single specific mechanistic change in UNC-13 but rather to a high synaptic vesicle (SV) release state achieved by disrupting the C1-C2B module - either through dysfunction of the C1 domain or the C2B domain (as seen with the HK and DN mutations).

Our findings support a "seesaw" model in which the C1 and C2B domains maintain a dynamic balance in their interaction with the plasma membrane, binding to DAG and PIP2. This balance may increase the energy barrier for SV release, preventing excessive neurotransmitter release under basal conditions. However, the C1-C2B toggle may be disrupted by high neuronal activity and act in an unbalanced state, thereby enhancing synaptic transmission (i.e., the gain-of-function state). To address these concerns, we have provided a clearer explanation of this functional switch in the revised version of the manuscript (page 27).

Regarding the differences between spontaneous and evoked neurotransmitter release, our previous studies have revealed that these two forms of release do not always respond similarly to various unc-13 mutations. This is a common phenomenon observed in other synaptic protein mutants, including synaptotagmin, tomosyn, and complexin, which indicates distinct yet partially overlapping regulatory mechanisms. Our model is well supported by most of the electrophysiological results from HK, DN, and HK+DN mutations across different unc-13 isoforms (UNC-13L, UNC-13S, UNC-13R, UNC-13ΔC2A, UNC-13ΔX). The main exception is that in UNC-13ΔX^HK+DN mutants, the changes in mEPSCs and mIPSCs differ from those observed in evoked EPSCs. This suggests that the mechanisms regulating the functional switch of unc-13 may differ slightly between spontaneous and evoked release. Since the X region of unc-13 and Munc13 remains largely uncharacterized, our findings provide intriguing insights into its potential functional role.

The manuscript proposes that UNC-13 toggles from a basal to a "gain-of-function" state under normal synaptic activity. However, it does not address when or how this switch might occur in vivo, since it is demonstrated principally via artificial mutations. Providing direct evidence or additional discussion of such switching under physiological conditions would be particularly informative.

What is the physiological significance of the proposed gain-of-function state? The data suggest that certain mutants (e.g., HK+D1-5N) lacking the gain-of-function state can still support synaptic transmission at wild-type levels. How do the authors reconcile this with the idea that the gain-of-function state plays a critical role at the synapse?

We appreciate these comments. While our model is mainly based on the dysfunction of the C1-C2B module (through HK and DN mutations), it provides a potential physiological framework for understanding how the structural balance of C1-C2B relates to the variability of synaptic transmission in the nervous system. In the CNS, synaptic transmission is highly variable, and the temporal pattern of the presynaptic activity may require dynamic switching of the fusion machinery, including UNC-13, between different functional modes, thereby triggering synaptic transmission at various levels. Our model suggests that under conditions of high neuronal activity, the C1-C2B module may transition from a balanced to an unbalanced state (gain-of-function state), thereby enhancing synaptic transmission.

Regarding the physiological significance of the gain-of-function state, we acknowledge that certain mutants (e.g., HK+D1-5N) lacking this state can still support wild-type levels of synaptic transmission. This observation suggests that the gain-of-function state may not be strictly required for baseline synaptic function but rather plays a modulatory role under specific conditions, such as heightened neuronal activity or synaptic plasticity. Further investigations will be needed to determine the precise in vivo triggers and functional consequences of this switch under physiological conditions. Moreover, we will focus on several linker regions (between C1 and C2B, C2B and MUN) to investigate their potential roles in regulating synaptic transmission and their broader functional significance in UNC-13 dynamics.

The authors determined the fluorescence intensity of mApple-tagged UNC-13 variants (Figure 1J-K and Figure 7J-K), finding no significant changes compared to the wild-type. However, a more detailed analysis of the density or distribution of fluorescent puncta in axons could clarify whether certain mutations alter the localization of UNC-13 at synapses. Demonstrating colocalization with wild-type UNC-13 (or another presynaptic marker) would help rule out mislocalization effects.

We appreciate the comment. In response, we have included a more detailed analysis of the synaptic localization of both wild-type and mutated UNC-13L in the revised manuscript. Our data show that in all scenarios, UNC-13 proteins exhibit strong colocalization with the active zone marker UNC-10::GFP (Figure 1L). Along with the fluorescence intensity data in Figure 1J, our findings indicate that the C1 and C2B mutations do not affect the expression level or the localization of UNC-13 at synapses. These results have been incorporated into the revised manuscript (page 8) and in Figure 1L.

The study mainly relies on extrachromosomal transgenes, which can show variable copy numbers and expression levels among individual worm strains. This variability might complicate interpretation, as differences in expression could mask or exaggerate certain phenotypes.

We agree that the expression levels of synaptic proteins can influence synaptic transmission levels. However, given the large number of mutations and truncations employed in this study, generating single-copy rescue lines for all transgenic strains would be a significant undertaking. On average, we need to microinject 50-100 worms to obtain one single-copy line, whereas injecting only 5-10 worms allows us to generate at least three independent extrachromosomal arrays. Based on our previous work, we found that the synaptic transmission levels are comparable between various extrachromosomal rescue arrays of unc13 and their single-copy rescue lines (e.g., UNC-13L, UNC-13S, UNC-13R, UNC-13ΔC2A, UNC-13ΔC2B, etc.). In future studies, we aim to use single-copy expression or CRISPRbased methods to induce deletions or mutations in various synaptic proteins.

Finally, the discussion is somewhat diffused. Streamlining the text to focus on the most direct connections would help readers pinpoint the key conclusions and open questions.

We appreciate the comment. As suggested, we have refined the discussion section. Specifically, we have removed the last part of the discussion (Functional roles of the linkers in UNC-13).  

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

(1) Clarify the "Gain-of-Function" State. Provide stronger justification or explicit discussion of whether all manipulations that enhance SV release truly correspond to the same mechanistic state or if multiple conformational states might be at play.

The “gain-of-function” state in this manuscript refers to a specific conformational status of UNC-13 that enhances synaptic vesicle (SV) release probability (both spontaneous and evoked) as a result of mutations (HK and DN) in the C1 and C2B domains. This effect is observed across multiple UNC-13 isoforms, including UNC-13L, UNC-13S, and UNC-13R. Prior studies from our group and others have demonstrated that C1 and C2B exhibit conserved functions in regulating synaptic transmission (Li et al., 2019, Cell Reports; Liu et al., 2021, Cell Reports; Michelassi et al., 2017, Neuron), supporting the idea that these domains share a common mechanism for modulating SV release. Given that C1 and C2B act as a functional unit (Michelassi et al., 2017, Neuron; and this study), we define all synaptic states induced by the dysfunction of these two domains as the "gain-of-function" mode.

However, it is important to note that this classification does not apply to high-release probability states induced by mutations in other domains.

The concept of a gain-of-function state due to C1 and C2B dysfunction has been previously proposed in studies of Munc13. Basu et al. (2007, Journal of Neuroscience) demonstrated that the H567K mutation in Munc13-1 C1 increases both spontaneous and evoked release probability, leading to a gain-of-function mode. Similarly, work from the Südhof group showed that KW and DN mutations in Munc13-1 C2B also enhance release probability, thereby inducing a gain-of-function state (Shin et al., 2010, Nature Structural & Molecular Biology). Our recent findings further support this idea, showing that UNC-13 C2B D3,4N (Li et al., 2019, Cell Reports; Liu et al., 2021, Cell Reports; Michelassi et al., 2017, Neuron) and the newly identified D1-5N mutation (this study) significantly elevate SV release, consistent with the D1,2N mutations reported by Shin et al.

Overall, our study integrates and extends previous findings, providing strong evidence that the C1 and C2B domains function as a regulatory switch between a basal physiological mode, a gain-of-function mode (enhanced release), and a loss-of-function mode (impaired release). This framework advances our understanding of how C1 and C2B dysfunction affects synaptic transmission and plasticity.

(2) Add comparisons to wild-type UNC-13L: When presenting data for deletions/mutants as "controls," include a visual reference (e.g., dashed line in figures) showing wild-type UNC13L levels. This will help readers see whether each construct is above or below the normal activity baseline.

As suggested, a dashed line showing the level of UNC-13L has been added to the bar graphs of all evoked EPSCs. The functional switch model is well supported by the results of the evoked EPSCs.

(3) Mutant and wild-type UNC-13 colocalization analysis: Demonstrating whether each mutant localizes robustly to synapses, in comparison to wild-type UNC-13, would bolster the interpretation of electrophysiological changes. If the authors have these data, adding them would address the possibility of mislocalization.

We agree with the reviewer that there would be value to address the possibility of mislocalization. However, in our experience working with UNC-13 mutant colocalization, we have found that neither deleting the X, C1 and C2B domains in UNC-13L  nor deleting C1 and C2B domain in UNC-13MR or UNC-13R altered the synaptic colocalization with the active zone protein UNC-10/RIM (Li 2019, Liu 2021), suggesting that C1 and C2B domains in UNC-13 are not involved in the regulation of protein localization. Thus, the mutations in the C1 and C2B domains are unlikely leading to protein mislocalization in the synaptic region.

(4) If possible, adding analysis using single-copy transgenes to confirm that extrachromosomal array expression variability does not qualitatively change the conclusions.

We strongly agree with the reviewer that single-copy transgenes would provide more stable protein expression levels and further consolidate our conclusions. However, several factors give us confidence that the extrachromosomal array rescue approach does not introduce significant variability in our results: First, our prior research has shown that SV release levels are generally comparable between extrachromosomal arrays carrying various unc13 transgenes and their corresponding single-copy rescue lines (e.g., UNC-13L, UNC-13S, UNC-13R, UNC-13ΔC2A, and UNC-13ΔC2B). Second, the major conclusions in this study are drawn from highly consistent and robust changes in SV release between different rescue lines (e.g., UNC-13L^HK+DN vs UNC-13L^DN; UNC-13S^HK+DN vs UNC-13S^HK or UNC-13S^DN ). Third, our imaging data indicate that the protein levels are indistinguishable between different unc-13 rescue arrays carrying C1 and C2B mutations, further supporting the validity of our findings.

Additionally, due to our recent relocation to a new institute, we are still in the process of setting up our microinjection system. Generating single-copy transgenes for all the extrachromosomal arrays used in this study would require significant time. We appreciate the reviewer’s understanding of our current situation. For our future studies regarding unc-13 and other synaptic proteins, we will prefer to use single-copy expression rather than extrachromosomal arrays.

(5) Reduce the length and speculation in the Discussion. A concise discussion that focuses on the most direct implications of the present findings will help improve the readability of this paper.

We appreciate the comment. As suggested, we have refined the discussion section.

Specifically, the last part of the discussion (Functional roles of the linkers in UNC-13) was removed.

(6) Minor formatting detail: In Figure 5C (left panel), adjust the y-axis label to ensure it aligns properly and improves clarity.

We appreciate the reviewer’s suggestion and have adjusted the y-axis label accordingly in the revised version (see revised Figure 5).

Reviewer #2 (Public review):

Summary:

The authors present a new model for animal pose estimation. The core feature they highlight is the model's stability compared to existing models in terms of keypoint drift. The authors test this model across a range of new and existing datasets. The authors also test the model with two mice in the same arena. For the single animal datasets the authors show a decrease in sudden jumps in keypoint detection and the number of undetected keypoints compared with DeepLabCut and SLEAP. Overall average accuracy, as measured by root mean squared error, generally shows generally similar but sometimes superior performance to DeepLabCut and better performance compared to SLEAP. The authors confusingly don't quantify the performance of pose estimation in the multi (two) animal case instead focusing on detecting individual identity. This multi-animal model is not compared with the model performance of the multi-animal mode of DeepLabCut or SLEAP.

Strengths:

The major strength of the paper is successfully demonstrating a model that is less likely to have incorrect large keypoint jumps compared to existing methods. As noted in the paper, this should lead to easier-to-interpret descriptions of pose and behavior to use in the context of a range of biological experimental workflows.

Weaknesses:

There are two main types of weaknesses in this paper. The first is a tendency to make unsubstantiated claims that suggest either model performance that is untested or misrepresents the presented data, or suggest excessively large gaps in current SOTA capabilities. One obvious example is in the abstract when the authors state ADPT "significantly outperforms the existing deep-learning methods, such as DeepLabCut, SLEAP, and DeepPoseKit." All tests in the rest of the paper, however, only discuss performance with DeepLabCut and SLEAP, not DeepPoseKit. At this point, there are many animal pose estimation models so it's fine they didn't compare against DeepPoseKit, but they shouldn't act like they did. Similar odd presentation of results are statements like "Our method exhibited an impressive prediction speed of 90{plus minus}4 frames per second (fps), faster than DeepLabCut (44{plus minus}2 fps) and equivalent to SLEAP (106{plus minus}4 fps)." Why is 90{plus minus}4 fps considered "equivalent to SLEAP (106{plus minus}4 fps)" and not slower? I agree they are similar but they are not the same. The paper's point of view of what is "equivalent" changes when describing how "On the single-fly dataset, ADPT excelled with an average mAP of 92.83%, surpassing both DeepLabCut and SLEAP (Figure 5B)" When one looks at Figure 5B, however, ADPT and DeepLabCut look identical. Beyond this, oddly only ADPT has uncertainty bars (no mention of what uncertainty is being quantified) and in fact, the bars overlap with the values corresponding to SLEAP and DeepPoseKit. In terms of making claims that seem to stretch the gaps in the current state of the field, the paper makes some seemingly odd and uncited statements like "Concerns about the safety of deep learning have largely limited the application of deep learning-based tools in behavioral analysis and slowed down the development of ethology" and "So far, deep learning pose estimation has not achieved the reliability of classical kinematic gait analysis" without specifying which classical gait analysis is being referred to. Certainly, existing tools like DeepLabCut and SLEAP are already widely cited and used for research.

The other main weakness in the paper is the validation of the multi-animal pose estimation. The core point of the paper is pose estimation and anti-drift performance and yet there is no validation of either of these things relating to multi-animal video. All that is quantified is the ability to track individual identity with a relatively limited dataset of 10 mice IDs with only two in the same arena (and see note about train and validation splits below). While individual tracking is an important task, that literature is not engaged with (i.e. papers like Walter and Couzin, eLife, 2021: https://doi.org/10.7554/eLife.64000) and the results in this paper aren't novel compared to that field's state of the art. On the other hand, while multi-animal pose estimation is also an important problem the paper doesn't engage with those results either. The two methods already used for comparison in the paper, SLEAP and DeepPoseKit, already have multi-animal modes and multi-animal annotated datasets but none of that is tested or engaged with in the paper. The paper notes many existing approaches are two-step methods, but, for practitioners, the difference is not enough to warrant a lack of comparison. The authors state that "The evaluation of our social tracking capability was performed by visualizing the predicted video data (see supplement Videos 3 and 4)." While the authors report success maintaining mouse ID, when one actually watches the key points in the video of the two mice (only a single minute was used for validation) the pose estimation is relatively poor with tails rarely being detected and many pose issues when the mice get close to each other.

Finally, particularly in the methods section, there were a number of places where what was actually done wasn't clear. For example in describing the network architecture, the authors say "Subsequently, network separately process these features in three branches, compute features at scale of one-fourth, one-eight and one-sixteenth, and generate one-eight scale features using convolution layer or deconvolution layer." Does only the one-eight branch have deconvolution or do the other branches also? Similarly, for the speed test, the authors say "Here we evaluate the inference speed of ADPT. We compared it with DeepLabCut and SLEAP on mouse videos at 1288 x 964 resolution", but in the methods section they say "The image inputs of ADPT were resized to a size that can be trained on the computer. For mouse images, it was reduced to half of the original size." Were different image sizes used for training and validation? Or Did ADPT not use 1288 x 964 resolution images as input which would obviously have major implications for the speed comparison? Similarly, for the individual ID experiments, the authors say "In this experiment, we used videos featuring different identified mice, allocating 80% of the data for model training and the remaining 20% for accuracy validation." Were frames from each video randomly assigned to the training or validation sets? Frames from the same video are very correlated (two frames could be just 1/30th of a second different from each other), and so if training and validation frames are interspersed with each other validation performance doesn't indicate much about performance on more realistic use cases (i.e. using models trained during the first part of an experiment to maintain ids throughout the rest of it.)

Editors' note: None of the original reviewers responded to our request to re-review the manuscript. The attached assessment statement is the editor's best attempt at assessing the extent to which the authors addressed the outstanding concerns from the previous round of revisions.

Author response:

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

eLife Assessment

This study introduces a useful deep learning-based algorithm that tracks animal postures with reduced drift by incorporating transformers for more robust keypoint detection. The efficacy of this new algorithm for single-animal pose estimation was demonstrated through comparisons with two popular algorithms. However, the analysis is incomplete and would benefit from comparisons with other state-of-the-art methods and consideration of multi-animal tracking.

First, we would like to express our gratitude to the eLife editors and reviewers for their thorough evaluation of our manuscript. ADPT aims to improve the accuracy of body point detection and tracking in animal behavior, facilitating more refined behavioral analyses. The insights provided by the reviewers have greatly enhanced the quality of our work, and we have addressed their comments point-by-point.

In this revision, we have included additional quantitative comparisons of multi-animal tracking capabilities between ADPT and other state-of-the-art methods. Specifically, we have added evaluations involving homecage social mice and marmosets to comprehensively showcase ADPT’s advantages from various perspectives. This additional analysis will help readers better understand how ADPT effectively overcomes point drift and expands its applicability in the field.

Reviewer #1:

In this paper, the authors introduce a new deep learning-based algorithm for tracking animal poses, especially in minimizing drift effects. The algorithm's performance was validated by comparing it with two other popular algorithms, DeepLabCut and LEAP.The accessibility of this tool for biological research is not clearly addressed, despite its potential usefulness. Researchers in biology often have limited expertise in deep learning training, deployment, and prediction. A detailed, step-by-step user guide is crucial, especially for applications in biological studies.

We appreciate the reviewers' acknowledgment of our work. While ADPT demonstrates superior performance compared to DeepLabCut and SLEAP, we recognize that the absence of a user-friendly interface may hinder its broader application, particularly for users with a background solely in biology. In this revision, we have enhanced the command-line version of the user tutorial to provide a clear, step-by-step guide. Additionally, we have developed a simple graphical user interface (GUI) to further support users who may not have expertise in deep learning, thereby making ADPT more accessible for biological research.

The proposed algorithm focuses on tracking and is compared with DLC and LEAP, which are more adept at detection rather than tracking.

In the field of animal pose estimation, the distinction between detection and tracking is often blurred. For instance, the title of the paper "SLEAP: A deep learning system for multi-animal pose tracking" refers to "tracking," while "detection" is characterized as "pose estimation" in the body text. Similarly, "Multi-animal pose estimation, identification, and tracking with DeepLabCut" uses "tracking" in the title, yet "detection" is also mentioned in the pose estimation section. We acknowledge that referencing these articles may have contributed to potential confusion.

To address this, we have clarified the distinction between "tracking" and "detection" Results section under " Anti-drift pose tracker." (see lines 118-119). In this paper, we now explicitly use “track” to refer to the tracking of all body points or poses of an individual, and “detect” for specific keypoints.

Reviewer #1 recommendations:

(1) DLC and LEAP are mainly good in detection, not tracking. The authors should compare their ADPT algorithm with idtracker.ai, ByteTrack, and other advanced tracking algorithms, including recent track-anything algorithms.

(2) DeepPoseKit is outdated and no longer maintained; a comparison with the T-REX algorithm would be more appropriate.

We appreciate the reviewer's suggestion for a more comprehensive comparison and acknowledge the importance of including these advanced tracking algorithms. However, we have not yet found suitable publicly available datasets for such comparative testing. We appreciate this insight and will consider incorporating T-REX into future comparisons.

(3) The authors primarily compared their performance using custom data. A systematic comparison with published data, such as the dataset reported in the paper "Multi-animal pose estimation, identification, and tracking with DeepLabCut," is necessary. A detailed comparison of the performances between ADPT and DLC is required.

In the previous version of our manuscript, we included the SLEAP single-fly public dataset and the OMS_dataset from OpenMonkeyStudio for performance comparisons. We recognize that these datasets were not comprehensive. In this revision, we have added the marmoset dataset from "Multi-animal pose estimation, identification, and tracking with DeepLabCut" and a customized homecage social mice dataset to enhance our comparative analysis of multi-animal pose estimation performance. Our comprehensive comparison reveals that ADPT outperforms both DLC and SLEAP, as discussed in the Results section under "ADPT can be adapted for end-to-end pose estimation and identification of freely social animals.". (Figure 1, see lines 303-332)

(4) Given the focus on biological studies, an easy-to-use interface and introduction are essential.

In this revision, we have not only developed a GUI for ADPT but also included a more detailed tutorial. This can be accessed at https://github.com/tangguoling/ADPT-TOOLBOX

Reviewer #2:

The authors present a new model for animal pose estimation. The core feature they highlight is the model's stability compared to existing models in terms of keypoint drift. The authors test this model across a range of new and existing datasets. The authors also test the model with two mice in the same arena. For the single animal datasets the authors show a decrease in sudden jumps in keypoint detection and the number of undetected keypoints compared with DeepLabCut and SLEAP. Overall average accuracy, as measured by root mean squared error, generally shows similar but sometimes superior performance to DeepLabCut and better performance compared to SLEAP. The authors confusingly don't quantify the performance of pose estimation in the multi (two) animal case instead focusing on detecting individual identity. This multi-animal model is not compared with the model performance of the multi-animal mode of DeepLabCut or SLEAP.

We appreciate the reviewer's thoughtful assessment of our manuscript. Our study focuses on addressing the issue of keypoint drift prevalent in animal pose estimation methods like DeepLabCut and SLEAP. During the model design process, we discovered that the structure of our model also enhances performance in identifying multiple animals. Consequently, we included some results related to multi-animal identity recognition in our manuscript.

In recent developments, we are working to broaden the applicability of ADPT for multi-animal pose estimation and identity recognition. Given that our manuscript emphasizes pose estimation, we have added a comparison of anti-drift performance in multi-animal scenarios in this revision. This quantifies ADPT's capability to mitigate drift in multi-animal pose estimation.

Using our custom Homecage social mice dataset, we compared ADPT with DeepLabCut and SLEAP. The results indicate that ADPT achieves more accurate anti-drift pose estimation for two mice, with superior keypoint detection accuracy. Furthermore, we also evaluated pose estimation accuracy on the publicly available marmoset dataset, where ADPT outperformed both DeepLabCut and SLEAP. These findings are discussed in the Results section under "ADPT can be adapted for end-to-end pose estimation and identification of freely social animals."

The first is a tendency to make unsubstantiated claims that suggest either model performance that is untested or misrepresents the presented data, or suggest excessively large gaps in current SOTA capabilities. One obvious example is in the abstract when the authors state ADPT "significantly outperforms the existing deep-learning methods, such as DeepLabCut, SLEAP, and DeepPoseKit." All tests in the rest of the paper, however, only discuss performance with DeepLabCut and SLEAP, not DeepPoseKit. At this point, there are many animal pose estimation models so it's fine they didn't compare against DeepPoseKit, but they shouldn't act like they did.

We appreciate the reviewer's feedback regarding unsubstantiated claims in our manuscript. Upon careful review, we acknowledge that our previous revisions inadvertently included statements that may misrepresent our model's performance. In particular, we have revised the abstract to eliminate the mention of DeepPoseKit, as our comparisons focused exclusively on DeepLabCut and SLEAP.

In addition to this correction, we have thoroughly reviewed the entire manuscript to address other instances of ambiguity and ensure that our claims are well-supported by the data presented. Thank you for bringing this to our attention; we are committed to maintaining the integrity of our claims throughout the paper.

In terms of making claims that seem to stretch the gaps in the current state of the field, the paper makes some seemingly odd and uncited statements like "Concerns about the safety of deep learning have largely limited the application of deep learning-based tools in behavioral analysis and slowed down the development of ethology" and "So far, deep learning pose estimation has not achieved the reliability of classical kinematic gait analysis" without specifying which classical gait analysis is being referred to. Certainly, existing tools like DeepLabCut and SLEAP are already widely cited and used for research.

In this revision, we have carefully reviewed the entire manuscript and addressed the instances of seemingly odd and unsubstantiated claims. Specifically, we have revised the statements "largely limited" to "limited" to ensure accuracy and clarity. Additionally, we thoroughly reviewed the citation list to ensure proper attribution, incorporating references such as "A deep learning-based toolbox for Automated Limb Motion Analysis (ALMA) in murine models of neurological disorders" to better substantiate our claims and provide a clearer context.

We have also added an additional section to comprehensively discuss the applications of widely-used tools like DeepLabCut and SLEAP in behavioral research. This new section elaborates on the challenges and limitations researchers encounter when applying these methods, highlighting both their significant contributions and the areas where improvements are still needed.

The other main weakness in the paper is the validation of the multi-animal pose estimation. The core point of the paper is pose estimation and anti-drift performance and yet there is no validation of either of these things relating to multi-animal video. All that is quantified is the ability to track individual identity with a relatively limited dataset of 10 mice IDs with only two in the same arena (and see note about train and validation splits below). While individual tracking is an important task, that literature is not engaged with (i.e. papers like Walter and Couzin, eLife, 2021: https://doi.org/10.7554/eLife.64000) and the results in this paper aren't novel compared to that field's state of the art. On the other hand, while multi-animal pose estimation is also an important problem the paper doesn't engage with those results either. The two methods already used for comparison in the paper, SLEAP and DeepPoseKit, already have multi-animal models and multi-animal annotated datasets but none of that is tested or engaged with in the paper. The paper notes many existing approaches are two-step methods, but, for practitioners, the difference is not enough to warrant a lack of comparison.

We appreciate the reviewer's insights regarding the validation of multi-animal pose estimation in our paper. While our primary focus has been on pose estimation and anti-drift performance, we recognize the importance of validating these aspects within the context of multi-animal videos.

In this revision, we have included a comparison of ADPT's anti-drift performance in multi-animal pose estimation, utilizing our custom Homecage social mouse dataset (Figure 1A). Our findings indicate that ADPT achieves more accurate pose estimation for two mice while significantly reducing keypoint drift, outperforming both DeepLabCut and SLEAP. (see lines 311-322). We trained each model three times, and this figure presents the results from one of those training sessions. We calculated the average RMSE between predictions and manual labels, demonstrating that ADPT achieved an average RMSE of 15.8 ± 0.59 pixels, while DeepLabCut (DLC) and SLEAP recorded RMSEs of 113.19 ± 42.75 pixels and 94.76 ± 1.95 pixels, respectively (Figure 1C). ADPT achieved an accuracy of 6.35 ± 0.14 pixels based on the DLC evaluation metric across all body parts of the mice, while DLC reached 7.49 ± 0.2 pixels (Figure 1D). ADPT achieved 8.33 ± 0.19 pixels using the SLEAP evaluation Metric across all body parts of the mice, compared to SLEAP’s 9.82 ± 0.57 pixels (Figure 1E).

Furthermore, we have conducted pose estimation accuracy evaluations on the publicly available marmoset dataset from DeepLabCut, where ADPT also demonstrated superior performance compared to DeepLabCut and SLEAP. These results can be found in the "ADPT can be adapted for end-to-end pose estimation and identification of freely social animals" section of the Results. (see lines 323-329)

We acknowledge the existing literature on multi-animal tracking, such as the work by Walter and Couzin (2021). While individual tracking is crucial, our primary focus lies in the effective tracking of animal poses and minimizing drift during this process. This dual emphasis on pose tracking and anti-drift performance distinguishes our work and aligns with ongoing advancements in the field. Engaging with relevant literature, highlights the importance of contextualizing our results within the broader tracking literature, demonstrating that while our findings may overlap with existing methods, the unique focus on improving tracking stability and reducing drift presents valuable contributions to the field. Thank you for your valuable feedback, which has helped us improve the robustness of our manuscript.

The authors state that "The evaluation of our social tracking capability was performed by visualizing the predicted video data (see supplement Videos 3 and 4)." While the authors report success maintaining mouse ID, when one actually watches the key points in the video of the two mice (only a single minute was used for validation) the pose estimation is relatively poor with tails rarely being detected and many pose issues when the mice get close to each other.

We acknowledge that there are indeed challenges in pose estimation, particularly when the two mice get close to each other, leading to tracking failures and infrequent detection of tails in the predicted videos. The reasons for these issues can be summarized as follows:

Lack of Training Data from Real Social Scenarios: The training data used for the social tracking assessment were primarily derived from the Mix-up Social Animal Dataset, which does not fully capture the complexities of real social interactions. In future work, we plan to incorporate a blend of real social data and the Mix-up data for model training. Specifically, we aim to annotate images where two animals are in close proximity or interacting to enhance the model's understanding of genuine social behaviors.

Challenges in Tail Tracking in Social Contexts: Tracking the tails of mice in social situations remains a significant challenge. To validate this, we have added an assessment of tracking performance in real social settings using homecage data. Our findings indicate that using annotated data from real environments significantly improves tail tracking accuracy, as demonstrated in the supplementary video.

We appreciate your feedback, which highlights critical areas for improvement in our model.

Finally, particularly in the methods section, there were a number of places where what was actually done wasn't clear.

We have carefully reviewed and revised the corresponding parts to clarify the previously incomprehensible statements. Thank you for your valuable feedback, which has helped enhance the clarity of our methods.

For example in describing the network architecture, the authors say "Subsequently, network separately process these features in three branches, compute features at scale of one-fourth, one-eight and one-sixteenth, and generate one-eight scale features using convolution layer or deconvolution layer." Does only the one-eight branch have deconvolution or do the other branches also?

We apologize for the confusion this has caused. Upon reviewing our manuscript, we identified an error in the diagram. In the revised version, we have clarified that the model samples feature maps at multiple resolutions and ultimately integrates them at the 1/8 resolution for feature fusion. Specifically, the 1/4 feature map from ResNet50's stack 2 is processed through max-pooling and convolution to generate a 1/8 feature map. Additionally, the 1/4 feature map from ResNet50's stack 2 is also transformed into a 1/8 feature map using a convolution operation with a stride of 2. Finally, both the input and output of the transformer are at the 1/16 resolution, which can be trained on a 2080Ti GPU. The 1/16 feature map is then upsampled to produce the final 1/8 feature map. We have updated the manuscript to reflect these changes, and we also modified the model architecture diagram for better clarity.

Similarly, for the speed test, the authors say "Here we evaluate the inference speed of ADPT. We compared it with DeepLabCut and SLEAP on mouse videos at 1288 x 964 resolution", but in the methods section they say "The image inputs of ADPT were resized to a size that can be trained on the computer. For mouse images, it was reduced to half of the original size." Were different image sizes used for training and validation? Or Did ADPT not use 1288 x 964 resolution images as input which would obviously have major implications for the speed comparison?

For our inference speed evaluation, all models, including ADPT, used images with a resolution of 1288 x 964. In ADPT's processing pipeline, the first layer is a resizing layer designed to compress the images to a scale determined by the global scale parameter. For the mouse images, we set the global scale to 0.5, allowing our GPU to handle the data at that resolution during transformer training.

We recorded the time taken by ADPT to process the entire 15-minute mouse video, which included the time taken for the resizing operation, and subsequently calculated the frames per second (FPS). We have clarified this process in the manuscript, particularly in the "Network Architecture" section, where we specify: "Initially, ADPT will resize the images to a390 scale (a hyperparameter, consistent with the global scale in the DLC configuration)."

Similarly, for the individual ID experiments, the authors say "In this experiment, we used videos featuring different identified mice, allocating 80% of the data for model training and the remaining 20% for accuracy validation." Were frames from each video randomly assigned to the training or validation sets? Frames from the same video are very correlated (two frames could be just 1/30th of a second different from each other), and so if training and validation frames are interspersed with each other validation performance doesn't indicate much about performance on more realistic use cases (i.e. using models trained during the first part of an experiment to maintain ids throughout the rest of it.)

In our study, we actually utilized the first 80% of frames from each video for model training and the remaining 20% for testing the model's ID tracking accuracy. We have revised the relevant description in the manuscript to clarify this process. The updated description can be found in the "Datasets" section under "Mouse Videos of Different Individuals."

Author response:

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

Public Reviews:

Reviewer #1 (Public review):

Summary:

This study aims to uncover molecular and structural details underlying the broad substrate specificity of glycosaminoglycan lyases belonging to a specific family (PL35). They determined the crystal structures of two such enzymes, conducted in vitro enzyme activity assays, and a thorough structure-guided mutagenesis campaign to interrogate the role of specific residues. They made progress towards achieving their aims but I see significant holes in data that need to be determined and in the authors' analyses.

Impact on the field:

I expect this work will have a limited impact on the field, although, with additional experimental work and better analysis, this paper will be able to stand on its own as a solid piece of structure-function analysis.

Strengths:

The major strengths of the study were the combination of structure and enzyme activity assays, comprehensive structural analysis, as well as a thorough structure-guided mutagenesis campaign.

Weaknesses:

There were several weaknesses, particularly:

(1) The authors claim to have done an ICP-MS experiment to show Mn2+ binds to their enzyme but did not present the data. The authors could have used the anomalous scattering properties of Mn2+ at the synchrotron to determine the presence and location of this cation (i.e. fluorescence spectra, and/or anomalous data collection at the Mn2+ absorption peak).

Thank you for your kind comment and suggestion. Many studies utilized ICP-MS for the detection of metal ions within proteins (doi: 10.1016/j.jbc.2023.103047; doi: 10.1074/jbc.RA119.011790), so we utilized this method to determine the type of atoms within GAGases. In the revised manuscript, the data of ICP-MS experiment has been presented in “Supplemental Table S1”

(2) The authors have an over-reliance on molecular docking for understanding the position of substrates bound to the enzyme. The docking analysis performed was cursory at best; Autodock Vina is a fine program but more rigorous software could have been chosen, as well we molecular dynamics simulations. As well the authors do not use any substrate/product-bound structures from the broader PL enzyme family to guide the placement of the substrates in the GAGases, and interpret the molecular docking models.

Thank you for your kind comments. The interaction between the enzyme and ligand should be confirmed by resolving the structure of enzyme-ligand complex. Unfortunately, we tried to prepare the co-crystals of GAGases with various oligosaccharide substrates but ultimately failed. Thus, we tried to use docking to explain the catalytic mechanism of polysaccharide lyases using Autodock Vina although this method may be questionable. In the revised manuscript, we predicted the substrate binding site of GAGase II using Caver Web 1.2 and performed molecular docking near the substrate binding site simultaneously using Molecular Operating Environment (MOE) to verify the accuracy of the docking results (Figure 6, Supplemental Figure S4). In addition, a series of enzyme-substrate complex structures of identified PL family enzymes with structural similarities to the GAGases are showed in Supplemental Figure S2, and the positions of the catalytic cavities and the substrate binding modes are similar to those of the molecular docking results, which may also corroborate the referability of our molecular docking results in another aspect.

(3) The conclusion that the structures of GAGase II and VII are most similar to the structures of alginate lyases (Table 2 data), and the authors' reliance on DALI, are both questioned. DALI uses a global alignment algorithm, which when used for multi-domain enzymes such as these tends to result in sub-optimal alignment of active site residues, particularly if the active site is formed between the two domains as is the case here. The authors should evaluate local alignment methods focused on the optimization of the superposition of a single domain; these methods may result in a more appropriate alignment of the active site residues and different alignment statistics. This may influence the overall conclusion of the evolutionary history of these PL35 enzymes.

Thank you for your kind question. As your suggestion, multiple structural alignment assays were carried out for the (α/α)_n toroid and the antiparallel β-sheet domain, respectively, based on the structures of GAGs/alginate lyases from PL5, PL8, PL12, PL15, PL17, PL21, PL23, PL36, PL38 and PL39 families. The results showed that the overall structure of GAGases is more similarity to that of PL15, PL17 and PL39 family alginate lyases, which have an (α/α)₆ toroid and an antiparallel β-sheet domain (Table 3). In terms of the toroid and antiparallel β-sheet domains, most of them have an (α/α)₆ toroid and an antiparallel β-sheet as shown in Table 3. We also noticed that GAGases possess such a (α/α)₆ toroid structure rather than a (α/α)₇ toroid structure, and revised the relevant statement in the manuscript.

(4) The data on the GAGase III residue His188 is not well interpreted; substitution of this residue clearly impacts HA and HS hydrolysis as well. The data on the impact on alginate hydrolysis is weak, which could be due to the fact that the WT enzyme has poor activity against alginate to start with.

Thank you very much for your helpful comments and questions. To verify your suggestion that the weak impact of alginate hydrolysis could be due to poor activity of wild type GAGase III, we degraded alginate using different enzyme concentrations (3 to 30 μg) and analyzed the degradation products. The results showed that the alginate-degrading activity of GAGase III-H188A and GAGase III-H188N was abolished, even at a quite high ratio of the mutated enzyme to substrate such as 30 μg enzyme to 30 μg substrate (Supplemental Figure S3A), while their GAG-degrading activity was only partially affected, indicating that this residue plays a more important role for the digestion of alginate than other substrates. Unfortunately, we were unable to confer the ability to GAGase III through the mutation of N191H in GAGase II. Therefore, we suggest that His¹⁸⁸ play a key role in the specificity of alginate degradation by GAGase III, but that other determinants also contribute to this process. We will try more methods to obtain the structure of enzyme-substrate co-crystals and explain its substrate-selective mechanism in future studies.

(5) The authors did not use the words "homology", "homologous", or "homolog" correctly (these terms mean the subjects have a known evolutionary relationship, which may or may not be known in the contexts the authors used these targets); the words "similarity" and "similar" are recommended to be used instead.

Thank you for your helpful suggestions. We have revised the relevant part of the description in the manuscript.

(6) The authors discuss a "shorter" cavity in GAGases, which does not make sense and is not supported by any figure or analysis. I recommend a figure with a surface representation of the various enzymes of interest, with dimensions of the cavity labeled (as a supplemental figure). The authors also do not specifically define what subsites are in the context of this family of enzymes, nor do they specifically label or indicate the location of the subsites on the figures of the GAGase II and IV enzyme structures.

Thank you for your helpful suggestions. Figures (Supplemental Figure S2) with surface representations of the GAGase II and some structurally similar GAGs/alginate lyases with the dimensions of the cavity labeled, were added to the supplementary data as you suggested. Considering the correlation between enzyme specificity and substrate binding sites, we speculated that a shorter substrate binding cavity might allow the enzyme to accommodate a wider variety of substrates, resulting in a smaller restriction of the catalytic cavity to substrate binding, although this speculation needs to be verified by the resolution of the crystal structure of the enzyme-substrate complexes.

Reviewer #2 (Public review):

Summary:

Wei et al. present the X-ray crystallographic structures of two PL35 family glycosaminoglycan (GAG) lyases that display a broad substrate specificity. The structural data show that there is a high degree of structural homology between these enzymes and GAGases that have previously been structurally characterized. Central to this are the N-terminal (α/α)7 toroid domain and the C-terminal two-layered β-sheet domain. Structural alignment of these novel PL35 lyases with previously deposited structures shows a highly conserved triplet of residues at the heart of the active sites. Docking studies identified potentially important residues for substrate binding and turnover, and subsequent site-directed mutagenesis paired with enzymatic assays confirmed the importance of many of these residues. A third PL35 GAGase that is able to turn over alginate was not crystallized, but a predicted model showed a conserved active site Asn was mutated to a His, which could potentially explain its ability to act on alginate. Mutation of the His into either Ala or Asn abrogated its activity on alginate, providing supporting evidence for the importance of the His. Finally, a catalytic mechanism is proposed for the activity of the PL35 lyases. Overall, the authors used an appropriate set of methods to investigate their claims, and the data largely support their conclusions. These results will likely provide a platform for further studies into the broad substrate specificity of PL35 lyases, as well as for studies into the evolutionary origins of these unique enzymes

Strengths:

The crystallographic data are of very high quality, and the use of modern structural prediction tools to allow for comparison of GAGase III to GAGase II/GAGase VII was nice to see. The authors were comprehensive in their comparison of the PL35 lyases to those in other families. The use of molecular docking to identify key residues and the use of site-directed mutagenesis to investigate substrate specificity was good, especially going the extra distance to mutate the conserved Asn to His in GAGase II and GAGase VII.

Weaknesses:

The structural models simply are not complete. A cursory look at the electron density and the models show that there are many positive density peaks that have not had anything modelled into them. The electron density also does not support the placement of a Mn2+ in the model. The authors indicate that ICP-MS was done to identify the metal, but no ICP-MS data is presented in the main text or supplementary. I believe the authors put too much emphasis on the possibility of GAGase III representing an evolutionary intermediate between GAG lyases and alginate lyases based on a single Asn to His mutation in the active site, and I don't believe that enough time was spent discussing how this "more open and shorter" catalytic cavity would necessarily mean that the enzyme could accommodate a broader set of substrates. Finally, the proposed mechanism does not bring the enzyme back to its starting state.

Recommendations for the authors:

Reviewer #1 (Recommendations for the authors):

Minor points:

(1) The number of significant digits used in Table 1 and Figure 3 legend are not justified. The authors should use a maximum of 2 significant digits.

Thank you for your kind suggestion. We have verified the relevant data and retained two significant digits.

(2) The authors should use the words "mutant" or "mutation" only when discussing DNA, but when discussing protein, the words "variant" and "substitution" should be used instead as these are more appropriate.

Thank you for your helpful suggestions. We have revised the relevant description in the manuscript as you suggested.

(3) Lines 102-110 are a long, run-on sentence that should be split into shorter sentences. Similarly, lines 367-378 should be split into shorter sentences.

Thank you for your suggestions. In the revised manuscript, the long sentences in lines 102-110 and 367-378 have been rewritten into shorter ones.

(4) Lines 174-175: His, Tyr, Glu, and Trp are not positively charged residues and this wording should be changed.

Thank you for your suggestions. We have revised the relevant description in the manuscript as you suggested.

(5) Lines 423-426 require a reference.

Thank you for your suggestion. We have provided the reference at the right position and revised the relevant description in the manuscript as you suggested.

(6) Grammar/language:

-line 90 - change "should emerge" to "likely emerged"

-line 145 - delete "Finally"

-line 264 - delete "their"

-line 265 - delete "active sites"

-line 265-266 - change to "To confirm this hypothesis, site-directed mutagenesis followed by enzyme activity assay was performed"

-line 311 - change "residue in the catalytic cavity of GAGase III, which.." to "residue in its catalytic cavity, which..."

-line 318 - change "affect" to "affected"

-line 323 - change to "degrading activity of GAGase II remains to be determined outside of the His188 residue"

-line 345 - delete "assays"

-line 359 - change to "evidence"

-line 397 - change "folds" to "3D fold"

-line 420 - change to "share similar catalytic sites"

-lines 411, 433 - change "conversed" to "conserved"

-line 441 - change to "Mutational analysis showed that the His188.."

-line 450 - delete "which"

Thank you for your suggestions. Grammatical errors in the revised manuscript have been corrected in the revised manuscript.

Reviewer #2 (Recommendations for the authors):

Major Concerns

The electron density in your model clearly does not support the placement of a Mn ion. In the GAGase II structure, the placement of the Mn and the placement of waters around it still results in two density peaks of > 12 rmsd. The manuscript suggests that ICP-MS was done but the results of this are not shown anywhere. Please include your ICP-MS data. I see the structures have already been deposited, and if they have been deposited unchanged, please see if you can modify them to actually finish building the models. I don't find your data in Figure 2B particularly convincing that Mn is necessarily important for activity.

Thank you for your kind comments. As we known, ICP-MS is a common method used for the detection of metal ions within proteins (doi: 10.1016/j.jbc.2023.103047; doi: 10.1074/jbc.RA119.011790), and thus we utilized it to determine the type of atoms within GAGases in this study. In the revised manuscript, the data of ICP-MS experiment has been presented in “Supplemental Table S1”, and the data clearly showed that the content of Mn²⁺ rather than others in test sample is much higher than that in the negative control, suggesting the involvement of Mn²⁺ in the protein. We agree that the addition of Mn²⁺ does not show very strong promotion to the activity of GAGase II just like other tested metal ions, but the addition of EDTA significantly inhibited the enzyme activity (Figure 2), indicating that metal ion such as Mn²⁺ is necessary for the function of GAGases. Regarding the role of metal ion, whether it participates in the catalytic reaction or only stabilize the structure of enzyme remains to be further explored in our further study.

Minor Concerns

(1) Please include CC1/2 in your Table 1.

Thank you for your kind suggestions. CC1/2 parameters have been added in the revised manuscript (Table 1).

(2) If possible please include SDS-PAGE gel images of your purified proteins. Particularly for the point mutations. Ideally, you would have done SEC on your mutants to show that the reduction in activity is not due to aggregation/misfolding, but at the very least I would to see that you have similar levels of purity.

Thank you for your kind suggestions. As your suggestion, we have added SDS-PAGE gel images of purified GAGase II, GAGase III, GAGase VII, and their mutant enzymes to the supplementary data. As shown in Figure S5, site-directed mutagenesis did not affect the soluble expression levels of GAGase II, GAGase III or GAGase VII, indicating that the reduction in activity is not due to aggregation or misfolding. Due to the large number of variants, we used crude enzyme for the activity assay of substrate binding sites, while for some catalytic key residues, we purified the corresponding mutant enzymes and then verified their activities by HPLC.

(3) When referring to your structural predictions, it is not appropriate to say that you used Robetta. Your reference is correct though - you should say that the structures were predicted using RoseTTAfold.

Thank you for your helpful suggestions. We have revised the relevant description in the manuscript.

(4) If possible expand on how the shorter/more open active site cavity would result in broader substrate specificity.

Thank you for your kind comment. In the revised manuscript, figures (Supplemental Figure S2) with surface representations of the GAGase II and some representatively structurally similar GAGs/alginate lyases, with the dimensions of the cavity labeled, were added to the supplementary data. Considering the correlation between enzyme specificity and substrate binding sites, we speculated that a shorter substrate binding cavity might allow the enzyme to accommodate a wider variety of substrates, resulting in a smaller restriction of the catalytic cavity to substrate binding. However, unfortunately, we did not succeed in obtaining co-crystals of GAGases with any of the substrates. We will try to explain the mechanism of substrate selectivity in future studies by culturing and resolving crystals of its enzyme substrate complex or otherwise.

(5) I would put less emphasis on His188 in GAGase III being a strong indicator that this protein represents an evolutionary intermediate between alginate lyases and GAGases.

Thank you for your comment. The His¹⁸⁸ residue, which is unique compared to other GAGases, is essential for the alginate-degrading activity of GAGase III. Regarding why GAGases are thought to represent a possible evolutionary intermediate between alginate lyases and GAG lyases, phylogenetic analysis demonstrated that GAGases show considerable homology with some identified GAG lyases and alginate lyases (DOI: 10.1016/j.jbc.2024.107466). The similarity in primary structure between some GAG lyases, alginate lyases, and GAGases suggests structural similarities, which are further supported by this study. As structure determines function, structural similarity is often used as a key criterion when studying the evolution of proteins, the GAGase III, which shows significant GAGs and alginate-degrading activity, support for this speculation. Of course, in this study, our analysis of the evolutionary relationship between GAGases and identified GAG lyases and alginate lyases, based on structural comparison, is an attempt using existing methods. The conclusions we have drawn remain a hypothesis that still requires further evidence to support and validate.

• Informativo 1168
• HC 232627 / DF
• Órgão julgador: Tribunal Pleno
• Relator(a): Min. GILMAR MENDES
• Julgamento: 11/03/2025 (Virtual)
• Ramo do Direito: Processual Penal, Constitucional
• Matéria: Jurisdição e Competência; Foro Especial por Prerrogativa de Função; Cessação do Exercício da Função/Direitos e Garantias Fundamentais; Poder Judiciário; Competência do Supremo Tribunal Federal

Foro por prerrogativa de função: competência para julgamento de crimes funcionais após a cessação do cargo

Tese fixada - A prerrogativa de foro para julgamento de crimes praticados no cargo e em razão das funções subsiste mesmo após o afastamento do cargo, ainda que o inquérito ou a ação penal sejam iniciados depois de cessado seu exercício.

Resumo - O STF fixou posição mais abrangente sobre a competência dos tribunais para julgar os crimes funcionais praticados por autoridades com prerrogativa de foro (“foro privilegiado”), no sentido de mantê-la mesmo após o término do exercício das respectivas funções. Aprimorou-se a orientação vigente com o intuito de assegurar a imparcialidade, a independência do julgamento e inibir os deslocamentos que resultam em lentidão, ineficiência e até mesmo prescrição das ações penais.

• O ordenamento jurídico prevê o foro especial por prerrogativa de função (CF/1988, art. 102, I, “b”) para proteger o exercício de cargos ou funções estatais de alta relevância constitucional contra ameaças do próprio acusado, manter a estabilidade das instituições democráticas, preservar o funcionamento do Estado e assegurar um julgamento menos suscetível a influências externas (1).
• Essa prerrogativa assegura que determinadas autoridades sejam julgadas por órgãos colegiados de maior hierarquia do Poder Judiciário. Portanto, o foro especial não constitui um privilégio pessoal, mas uma garantia para o adequado exercício das funções públicas.

• No que concerne à problemática do momento de encerramento do direito ao foro privilegiado, a jurisprudência desta Corte oscilou ao definir a sua extensão, ora pela natureza do delito (regra da contemporaneidade e da pertinência temática), ora pelo exercício atual de funções públicas (regra da atualidade), o que gerou uma indefinição quanto à abrangência do instituto.

• Com o cancelamento da Súmula 394/STF (2) — no julgamento da Questão de Ordem no Inquérito nº 687/SP (3) —, esta Corte realizou uma redução teleológica do foro privilegiado ao limitar sua aplicabilidade, de modo que o foro especial não se manteria após a perda do mandato, mesmo na hipótese de crimes cometidos durante o exercício das funções.

• Posteriormente, na Questão de Ordem na Ação Penal nº 937/RJ (4), o Tribunal entendeu que o referido foro se aplicaria apenas aos crimes cometidos durante o exercício do cargo e relacionados às funções desempenhadas. Assim, com exceção das ações cuja fase da instrução processual esteja concluída — hipótese de manutenção da competência, inclusive nos casos de infrações penais não relacionadas ao cargo ou à função exercida — a cessação do exercício das funções ensejaria o declínio da competência para o Juízo de primeiro grau.
• Nesse contexto, nas hipóteses de crimes funcionais, a imposição da remessa dos autos para a primeira instância com o término do exercício funcional subverte a finalidade do foro por prerrogativa de função. Isso ocorre porque, além de ser contraproducente ao causar flutuações de competência (“sobe e desce”) no decorrer das causas criminais e trazer instabilidade ao sistema de Justiça, permite a alteração da competência absoluta ratione personae ou ratione funcionae por ato voluntário do agente público acusado, ao renunciar ao mandato ou à função antes do final da instrução processual.
• Na espécie, esta Corte firmou a perpetuação da competência para o julgamento de crimes funcionais com base em uma interpretação mais ampla do foro especial, centrada na natureza do crime praticado pelo agente, em vez de critérios temporais relacionados à permanência no cargo ou ao exercício atual do mandato, que podem ser manipulados pelo acusado. Ademais, a saída do cargo somente afasta o foro privativo na hipótese de crimes perpetrados antes da investidura no cargo ou que não possuam relação com o seu exercício.
• Com base nesses e em outros entendimentos, o Plenário, por maioria, concedeu a ordem de habeas corpus para (i) assentar a competência do Supremo Tribunal Federal para processar e julgar a ação penal nº 1033998-13.2020.4.01.3900; e (ii) fixar a tese anteriormente mencionada, com o entendimento de que essa nova linha interpretativa deve aplicar-se imediatamente aos processos em curso, ressalvados todos os atos praticados e decisões proferidas pelo STF e pelos demais Juízos com base na jurisprudência anterior, conforme precedentes firmados no QO no INQ 687 e na QO na AP 937.

(1) CF/1988: “Art. 102. Compete ao Supremo Tribunal Federal, precipuamente, a guarda da Constituição, cabendo-lhe: I - processar e julgar, originariamente: (...) b) nas infrações penais comuns, o Presidente da República, o Vice-Presidente, os membros do Congresso Nacional, seus próprios Ministros e o Procurador-Geral da República”. (2) Súmula 394/STF: “Cometido o crime durante o exercício funcional, prevalece a competência especial por prerrogativa de função, ainda que o inquérito ou a ação penal sejam iniciados após a cessação daquele exercício. (Cancelada)”. (3) Precedente citado: QO no INQ 687. (4) Precedente citado: QO na AP 937.

Legislação: CF/1988: art. 102, I, b. Súmula 394/STF.

Precedentes: QO no INQ 687 e QO na AP 937.

Note: This response was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Reply to the reviewers

Dear Review Commons editorial team,

Thank you for coordinating the thorough and careful review of our manuscript. We are especially grateful to the four anonymous reviewers for recognizing the value of our work and for their constructive suggestions on how to improve it.

We are encouraged by the positive reception of our main conclusions on the robustness of adaptation to DNA replication stress and its relevance to multiple fields. All reviewers provided insightful comments, with reviewers #2 and #4 emphasizing that further experimental validation of the hypothesized role of reduced dNTPs in alleviating fitness during constitutive DNA replication stress would strengthen the paper. While the precise molecular mechanisms underlying this suppression are not the primary focus of this manuscript, we are eager to perform additional experiments based on the reviewers’ suggestions.

Below, we present a detailed revision plan in the form of a point-by-point response to their comments.

Reviewer #1 (Evidence, reproducibility and clarity):

This study investigates the compensatory evolutionary response of Saccharomyces cerevisiae to DNA replication stress, focusing on the influence of genotype-environment interactions (GXE). The authors used a range of experimental conditions with varying nutrient levels to assess evolutionary outcomes under replication stress. Their genomic analysis reveals that while glucose levels affect initial adaptation rates, the genetics of adaptation remain robust across all nutritional environments. The research offers new insights into the adaptability of S. cerevisiae, emphasizing the role of the nutritional environment in evolutionary processes related to DNA replication stress. It identifies recurrent advantageous mutations under different macronutrient availabilities and uncovers a novel role for the RNA polymerase II mediator complex in adaptation to replication stress. Overall, this well-designed study adds to the growing recognition of the complexity and robustness of evolutionary responses to environmental stressors. It provides strong evidence that compensatory evolution to replication stress is robust across varying nutritional conditions. It both challenges and reinforces previous findings regarding the resilience of the yeast genetic interaction network to environmental perturbations. The detailed analysis of specific compensatory mutations and their fitness impacts across different conditions offers valuable insights into adaptive dynamics over 1000 generations, contributing a clear empirical framework for understanding how replication-associated stress shapes evolutionary outcomes in diverse environments.

Based on the analysis:

1) The conclusions are generally well-supported by the presented data. The evolution experiments and genomic analyses are robust and provide convincing evidence for the study's main claims. The authors took steps to eliminate bias, such as maintaining an adequate Ne, which, if not done, could have compromised their conclusions by affecting genetic drift and limiting the population's access to beneficial mutations.

2) The figures are well-designed and easy to understand.

3) The methodology is well-described and appears reproducible. The authors provide sufficient details on experimental procedures. Experimental replication is adequate, with multiple evolutionary lines.

4) They also made efforts to validate their observations, such as the validation of mutations, the prediction of interactions in the Med14 structure, and its potential implication in gene regulation, as well as the analysis of the cumulative fitness benefit and the reconstruction of the quadruple mutant.

There are, however, a few results that would benefit from further clarification:

1) The experimental design is strong, offering a diverse range of conditions. However, the high glucose condition (8%) stands out as significantly different from the neutral 2% condition, both in range and margin, compared to the low glucose conditions (0.25-0.5%). While this mainly affects growth profiles and evolvability in the early generations, a brief explanation in the discussion would strengthen the conclusions. Specifically, addressing:

1. a) The rationale behind selecting these particular glucose concentrations.

2. b) How other glucose concentrations might influence the outcomes. Providing this additional context would enhance the reader's understanding of the experimental setup and its potential implications, while also offering insights into the broader applicability of the findings and possible directions for future research.

We thank the reviewer for pointing out the need to clarify the rationale behind the glucose concentrations used in our study, an aspect we agree should have been better explained. In response, we have added the following text detailing the chosen conditions and their established effects on cellular metabolism.

Line 67: “Glucose is the most abundant monosaccharide in nature, and represents the preferred source of energy for most cells.”

Line 110: “...we grew WT and ctf4Δ cells in varying glucose concentrations to induce distinct physiological states. Low glucose levels (0.25% and 0.5%) induce caloric restriction and ultimately glucose starvation (Lin et al 2000, Smith et al. 2009). These conditions elicit increased respiration (Lin et al., 2002), sirtuins expression (Guarente, 2013), autophagy (Bagherniya et al. 2018), DNA repair (Heydari et al., 2007), and reduced recombination at the ribosomal DNA locus (Riesen and Morgan, 2009) ultimately extending lifespan in several organisms (Kapahi et al., 2016). In contrast, standard laboratory conditions typically use 2% glucose, promoting a rapid proliferation environment to which strains have been adapted since laboratory domestication (Lindergren, 1949). Finally, elevated glucose concentrations (such as 8%) result in higher ethanol production (Lin et al., 2012) and reactive oxygen species (ROS) levels (Maslanka et al., 2017).

2) In the discussion section, a more explicit comparison with similar studies in other model organisms would help contextualize the findings within the broader field of evolutionary biology. While the results appear robust, it would be beneficial to explore how they align with or contrast to previous studies on DNA damage, particularly in bacteria or highly complex eukaryotes.

We appreciate this suggestion to better contextualize our findings within the broader literature, as it provides an opportunity to highlight the unique aspects of our work. While many studies have explored how environmental factors shape fitness landscapes and influence evolutionary strategies, to our knowledge, only a few have addressed this in the context of compensatory evolution, where cells must recover fitness lost due to intracellular perturbations. To address this point, we have added a discussion of additional examples involving other model organisms, highlighting their difference with the question asked in this work.

Line 34: “Genotype-by-environment (GxE) interactions are well-documented. For example, several studies on E. coli have demonstrated how different environments influence fitness and epistatic interactions among adaptive mutations in the Lenski Long-Term Evolution Experiment (Ostrowski et al., 2005, 2008; Flynn et al., 2012; Hall et al., 2019). Adaptive mutations in viral genomes similarly exhibit variable fitness effects across different hosts (Lalic and Elena, 2012; Cervera, 2016). Furthermore, interactions between mutations in the Plasmodium falciparum dihydrofolate reductase gene have been shown to predict distinct patterns of resistance to antimalarial drugs (Ogbunugafor et al., 2016). However, the role of environmental factors in shaping evolution within the context of compensatory adaptation, when fitness defects primarily arise from intracellular perturbations, remains much less explored.”

However, if the reviewer have particular additional studies in mind, we welcome further suggestions to include in the final manuscript.

Minor comments:

1) The presentation of data in the figures is clear and informative. However, some figure legends could benefit from more detailed explanations. For example, although the statistical tests used are mentioned in the methods section, it would be helpful to also include them in the figure legends, such as in legend 1acde, as well as in all other figures.

We are now reporting the statistical test used for each comparison also in figure legends.

2) In terms of broader conclusions, here are a few suggestions, though they are, of course, optional:

a) The study could benefit from exploring the potential trade-offs of adaptive mutations in the hypothetical return to environments without replication stress, at least theoretically. This would provide a more comprehensive understanding of the evolutionary constraints.

We thank the reviewer for the suggestion, we had performed the measurements but did not comment on them explicitly. We are now commenting on them as follows:

Line 310: “In the WT background, all mutations were nearly neutral, with only minimal deleterious or advantageous effects on fitness depending on glucose concentrations (Fig S4A).”

Line 468: “The nearly neutral effects on fitness of the core adaptive mutations in WT suggest that they are likely to persist even after the initial replication stress is resolved.”

b) A brief discussion of the potential limitations of using lab strains versus wild isolates of S. cerevisiae would offer valuable context for the generalizability of the findings.

This is an excellent point. While addressing it fully would warrant a separate manuscript, we provide our comments here, along with similar observations raised by this and other reviewers, as follows:

Line 450: “How generalizable are our conclusions about the reproducibility of evolutionary repair to DNA replication stress across other organisms, species, or replication challenges? While dedicated future studies are needed to fully address these important questions, several lines of evidence are encouraging. A recent report demonstrated that the identity of suppressor mutations of lethal alleles was conserved when introduced into highly divergent wild yeast isolates (Paltenghi and van Leeuwen, 2024). Similarly, earlier work showed that even ploidy, which significantly alters the target size for loss- and gain-of-function mutations, affected only the identity of the genes targeted by selection, while the broader cellular modules involved remained consistent (Fumasoni and Murray, 2021). Moreover, divergent organisms experiencing different types of DNA replication stress exhibit some of the adaptive responses described here. For example, the yeast genus Hanseniaspora, which lacks the Pol32 subunit of the replisome, has also been reported to have lost the DNA damage checkpoint (Steenwyk et al., 2019). Human Ewing sarcoma cells carrying the fusion oncogene EWS-FLI1 frequently exhibit adaptive amplification of the cohesin subunit RAD21 (Su et al., 2021). Together, these findings suggest that while the specific details of DNA replication perturbations and the genomic features of organisms may shape the precise targets of compensatory evolution, the overarching principles and cellular modules affected are broadly conserved.”

Furthermore, we plan to search a recently published database of variants found in natural isolates of S. cerevisiae to assess whether similar evolutionary processes to those described in this study may have occurred in wild strains.

c) It would be valuable to present the differences in ploidy in the context of other studies, such as the nutrient-limitation hypothesis (e.g., 'The Evolutionary Advantage of Haploid Versus Diploid Microbes in Nutrient-Poor Environments' by Bessho, 2015), since, as previously demonstrated by the authors of this article that is being reviewed, ploidy may influence the evolutionary trajectories of DNA repair.

d) Interrelating these three terms: nutrient-limitation, ploidy, and DNA repair could be an interesting avenue to explore in the discussion.

In response to comments c and d, we have now commented on the intersection between ploidy and other types of DNA perturbation in the paragraph starting in line 491 (see response above)

3) Specific details:

a) Line 116: To improve clarity, it would be beneficial to refer to the figure right after the statement: 'However, their relative fitness improved compared to the WT reference as the initial glucose levels (Figure X).'

b) Line 404: The statement about antibiotics and cancer progression is somewhat brief here; it might be helpful to provide more context on why this mechanism influences these processes (here or before).

c) Line 418: "were re-suspended in water containing zymolyase (Zymo Research, Irvine, CA, US, 0.025 μ/μL), incubated at". Something is missing in the units.

d) Line 459: "and G2 phases for each genotype was estimated by deriving the the relative cell distribution". The article "the" is repeated.

e) 1a: The x-axis ticks appear misaligned, which makes it difficult to interpret the boxplots. For example, at 0.25, the tick is closer to the orange boxplot than to the black one. In contrast, at 2%, the tick seems well-centered."

f) Figure 3 could benefit from a general legend at the top regarding the colors, as finding it in 2c was not intuitively easy.

The typos and suggestions raised in points 3a-f have now been corrected in the manuscript.

g) I didn't review the code on GitHub.

Reviewer #1 (Significance):

The main strength of the study is that it shows robustness of compensatory evolution across varying nutrient conditions. The study adds to the growing body of literature on DNA replication stress and evolutionary adaptation by showing that compensatory evolution can occur regardless of nutrient availability. This fundamental finding challenges prior assumptions that nutrient conditions significantly alter evolutionary outcomes, contributing to a more nuanced understanding of how cells respond to stress. Furthermore, the discovery of the RNA polymerase II mediator complex's role in this process is particularly novel and opens new lines of investigation.

Advance in the field: The results advance our understanding of evolutionary biology, particularly in the context of DNA replication stress and compensatory evolution. The study demonstrates that evolutionary repair mechanisms are predictable, even under variable environmental conditions, which has key implications for evolutionary biology and therapeutic applications.

Audience:

This paper will be of interest to a specialized audience in evolutionary biology, genomics, and cell biology, particularly those interested in DNA replication stress and adaptive evolution. Researchers studying stress responses in model organisms, such as S. cerevisiae, will find the findings valuable, as will those working in applied fields where stress adaptation is a critical factor (e.g., industrial yeast fermentation, drug development, disease resistance, cancer research, or aging studies).

Expertise:

Evolutionary biology, genomic analysis, and cellular stress responses, with a particular focus on experimental evolution under DNA damage stress in Saccharomyces cerevisiae. Recently graduated and beginner reviewer.

Reviewer #2 (Evidence, reproducibility and clarity):

The paper addresses the effect of sugar availability in shaping compensatory evolution. The first observation of the paper is that cell physiology changes by modulating glucose availability also in strains that come with defective DNA replication (ctf4-null previously studied by the authors). An intriguing result is that ctf4-null grows comparatively better in low concentrations of glucose. This is hypothesized to be a consequence of both the decrease in dNTPs in low glucose, which causes slow down of fork progression, and/or reduced fork collapse at rDNA locus. Hence, wild types and ctf4-null show an opposite trend: in the mutant, the lowest concentration of glucose is the least affected by the mutation; in wild type, the highest concentration is the least affected. Adaptation rate is inversely related with the initial fitness. The effect on physiology and adaptation rate is a starting point for asking the key question: are evolutionary trajectories influnced by the growth conditions? The answer is negative: evolution experiments show the very same core of genetic changes at all sugar concentrations. The result is apparently at odds with previous publications, and the authors conclude that, in this particular setting, availability of carbon sources plays a minor role compared to impaired DNA replication. The different rates of adaptation in WT and mutant is rather explained by the initial fitness at the different glucose concentrations, which, as mentioned, is opposite in WT and ctf4-null mutants. The paper also reports a new mutation in MED14, component of the transcription mediator complex, which rescues the lack of Ctf4 activity. The study is interesting and asks a relevant question. The experiments are well executed and convincing, but the paper can be strengthened by testing some of the hypotheses which are put forward.

Main points

1- The raw data for evolutionary dynamics (Figure S2C) are fitted with the power law suggested by Wiser and Lenski, and return different values of the parameter 'b'. The authors say that the result depends greatly on the initial conditions ("due to the varying initial fitness of ctf4Δ cells across different glucose environments, they display an opposite trend to WT"). Around the initial values, however, the curves are non-monotonic, especially for low glucose availability. Both for WT and ctf4-null there is an initial drop in fitness, after which fitness increases. If one would neglect this initial dynamics, the value of the parameter 'b' would likely be different.

The non-monotonic trend in fitness highlighted by the reviewer is likely due to technical factors: Fitness at Generation 0 was measured with high precision in a low-throughput manner early in the project. In contrast, fitness from Generation 100 to 1000 was measured later in the study in a high-throughput fashion, necessitated by the large number of competitions conducted (96 wells × 4 time points × 6 replicates = 2304 assays). This difference in methodologies may have introduced a slight offset when the datasets were combined at Generation 100. Following the reviewer’s suggestion, we have excluded the data point at Generation 100 responsible for this non-monotonic behavior and re-fitted the curves. While this adjustment has caused minor changes in the parameter ‘b’, the qualitative trends, particularly the opposing trends between WT and ctf4Δ as glucose increases, remain consistent (Figure_rev_only 1). To ensure transparency, we have retained all recorded fitness values in the original figure for reference.

In general, one can question whether curves with this shape are best fitted by the power law proposed by Wiser and Lenski. For example, for the WT 0.25% glucose the linear fit gives a better R2 (why do the authors show the linear fit anyway?). This impression is further reinforced by the observation that Wiser and Lenski fit dynamics that last 50.000 generation, here the curves last 1/50th of it. In conclusion, I would question whether the parameter 'b' is a solid measurement of 'rate of adaptation'. Also, normalizations makes it difficult to appreciate the result shown in Figure 2B. I think the authors should look for a different way to show the different trend in adaptation dynamics for different glucose concentrations between wild types and mutants. For example, they could move Figure S2C in the main text to stress the result shown in Figure 2C, which already shows the difference between WT and mutant. This is especially true if what Figure 2C shows is (evo-anc)/evo. This is not fully clear to me: in the legend it refers to the delta, in the label of the y-axis I read that this is a percentage.

We thank the reviewer for prompting us to clarify our methods for reporting fitness changes over time. The fitness values are reported, throughout the paper, as a percentage change relative to the reference WT strain. The gain in fitness during evolution (reported as Δ) represents the difference between the evolved strain (evo%) and the ancestral strain (anc%), calculated as Δ = evo% - anc%. This represents the absolute gain, rather than the relative gain. This value is still reported as a percentage as it’s the same scale and unit as the two values being subtracted. We have included additional details to clarify this aspect in the figure legend.

“(C) Absolute fitness gains (Δ) at generation 1000 for evolved WT (upper panel, black) and ctf4Δ (lower panel, orange) populations. Box plots show median, IQR, and whiskers extending to 1.5×IQR, with individual data points beyond whiskers considered outliers. Absolute fitness gains were calculated by subtracting the ancestral relative fitness from the relative fitness of the evolved (Δ = evo% - anc%), both calculated as percentages relative to the same reference strain in the same glucose concentration.”

To conclude: the data show a different trend between wild types and mutants, which is interesting. Fitting it with the power law seems to be neither required nor appropriate. I suggest the authors to show the WT vs mutant pattern differently.

We followed the reviewer’s suggestion and moved Figure S2C, which depicts the detailed fitness trajectories over time, into the main manuscript as Figure 2D. We agree that presenting these trajectories alongside the absolute fitness gains (now in Figure S2C) provides a more intuitive and effective depiction of the evolutionary dynamics of WT and ctf4Δ strains without relying solely on the power-law fit. Additionally, we quantified the mean adaptation rate, calculated as the absolute fitness gain (Δ) divided by the total number of generations (now Figure 2B). While no individual method definitively captures the adaptation rates across the experiment, these complementary analyses consistently highlight the same trends noted by the reviewer. We have re-written the main text as follows:

Line 171: “By generation 1000, both WT and ctf4Δ evolved lines achieved, on average, slightly higher fitness in low glucose compared to high glucose conditions (Fig S2B). However, due to the varying initial fitness of ctf4Δ cells across different glucose environments, they recovered the same extent of the original defect (Fig S2C). ctf4Δ lines displayed an opposite trend to WT, with increasing absolute fitness throughout the experiment as glucose concentration rose (Fig S2B vs S2D). The differint absolute fitness gains over the same number of generations highlight distinct mean adaptation rates (Fig 2B). These differences are evident when examining the evolutionary dynamics of the evolved lines over time (Fig 2C). Additionally, we approximated the fitness trajectories using the power law function (Fig 2C, dashed purple lines), previously proposed to describe long-term evolutionary dynamics in constant environments (Wiser et al., 2013). The parameter b in this formula determines the curve's steepness, and can be used to quantify the global adaptation rate over generations (Fig S2E). Collectively, these analyses demonstrate that, unlike WT cells, ctf4Δ lines adapt faster in the presence of high glucose. This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

Overall, these results demonstrate that cells can recover from fitness defects caused by constitutive DNA replication stress regardless of the glucose environment. However, adaptation rates under DNA replication stress exhibit opposing trends compared to WT cells, with faster adaptation yielding greater fitness gains in higher glucose conditions.”

2- In Figure S2C, the individual trajectories for WT at 2% glucose are strangely variable. In this case, plotting the average does not make too much sense. This result is strange, since this is the default condition, where cells are grown without any change of sugar concentration. Can the authors give any rationale? Are there other available results to replace those published in Figure S2C?

We agree with the reviewer that the individual trajectories for WT at 2% glucose are intriguing. However, we do not find these results necessarily “strange” as they could be explained by the following rationale: WT cells have been cultivated in 2% glucose since the 1950s, likely fixing most beneficial mutations for this condition. When many isogenic strains are evolved in parallel, (a) some lines show no improvement due to the scarcity of available beneficial mutations, (b) others exhibit slight decreases in fitness due to genetic drift fixing deleterious mutations, and (c) a few lines discover rare beneficial mutations, leading to fitness increases. In contrast, other conditions represent “newer” environments with larger mutational target sizes, resulting in more consistent outcomes.

Prompted by the reviewer’s comment, we look for other studies reporting detailed fitness measurements of evolved WT strains in standard laboratory media. We downloaded and plotted the fitness data from Johnson et al. 2021, where authors studied the evolution of WT strains over 10.000 generations. Interestingly, we see that in the early phase of the evolution (generations 500-1400) evolved lines show similar levels of variability in fitness as the one reported in our study (Figure_rev_only 2). Of note is that in Johnson et al. 2021 most of the adaptive mutations alleviate the toxicity of the ade2-1 allele. In our WT strain the gene was preemptively restored, furter reducing the target size for adaptation in YPD.

We believe it is important to report these measurements and decided to leave the original data, with the appropriate quantifications of variability, in Figure 2.

3- The molecular explanation given for the rescue of ctf4-null proposes a very relevant role for dNTPs downregulation. Particularly, both for Irx1 and med14-H919P, the authors propose that this happens via Rnr1 downregulation. At this stage, this is only a hypothesis. The molecular verification of the central role of Rnr1 downregulation would make the conclusion much stronger. For example, a preliminary test would imply that duplicating RNR1 in ctf4-null irx1-null and/or ctf4-null med14-H919P would revert the rescue. Any other experiment addressing this point would be useful to improve the paper.

We agree that the experiment suggested by the reviewer, or similar tests, would substantiate our hypotheses and strengthen the paper. Specifically, we plan to perturb dNTP production in both ctf4Δ ixr1Δ and ctf4Δ med14-H919P mutants through genetic manipulation of known factors involved in dNTP synthesis. We will then compare the resulting fitness to the expectations based on our hypotheses: reduced fitness benefits of the double mutants upon increasing dNTP levels and/or increased fitness in ctf4Δ mutants by decreasing dNTP levels through alternative mechanisms.

4- The authors propose from Figure S4B that the rescue of ixr1-null is less evident at low sugar concentration since both conditions trigger a reduction of dNTPs. I think this is interesting, since it would provide a link between glucose concentration and evolutionary trajectories to adaptation, which is what the authors wanted to study. In particular, one would predict that 0.25% glucose would see less ixr1-null than the other glucose conditions. I could not (was not able to) confute this hypothesis from the data shown in the paper. Likewise, for med14-H919P. If the authors have not tested it, it would be worth trying.

We had reported the appearance and frequency of all ‘core adaptive mutations’ (Figure S6C) but did not explicitly test the likelihood of their appearance under different glucose conditions. Following the reviewer’s suggestion, we have now performed χ2 tests (on the presence or absence of mutations) and ANOVA tests (on their mean frequency) to determine whether any mutation is particularly enriched or depleted in a given glucose environment. At first glance, the results do not support the hypothesis proposed by the reviewer. However, we note that although ixr1 mutants are less beneficial in low glucose than in high glucose, they still confer an 8% fitness advantage, which is likely sufficient to drive clones to fixation. We believe the reviewer’s reasoning is correct but is potentially masked by the still elevated fitness advantage of ixr1 in low glucose.

To better convey the results of this analysis, we have included a visual representation of the presence and frequency of the mutations in Figure 6A, and the results of the χ2 and ANOVA tests in Supplementary File 5. We also comment on the analysis as follows:

Line 314: “Similarly, we did not detect differences in the frequency of occurrence (χ2 tests) or average fractions (ANOVA test) achieved by the mutations in the populations evolved under different glucose environments (Fig 6A, Fig S4C and Supplementary File 5. The presence of all mutations in the final evolved lines correlated with their fitness benefits, suggesting how their selection in all glucose conditions was mostly dictated by their relative fitness benefits, rather than the environment (Fig 6A).”

5- The combination of the four genetic adaptation (Fig 6B) would benefit from an experimental verification to show that the different solutions are not mutually exclusive. This is not obvious: if more than one solution acts by reducing dNTPs, maybe their combined effect is less strong than what measured theoretically. The authors could derive some clones at the end of the experiment and Sanger sequencing some of the four genes, to confirm the co-presence of some of them in the same cell.

The co-occurrence of nearly every combination of the four core adaptive mutations we identified can be inferred from their relative frequencies, as revealed by deep whole-genome sequencing of the evolved populations (Fig. S4C). In these data, we observe populations carrying each pairwise combination of mutations at frequencies exceeding 50%, implying their coexistence. Moreover, many combinations of mutations approach or reach fixation. A particularly striking example is ctf4Δ Population 11, evolved in 8% glucose, where all core adaptive mutations are present at 100% frequency. These findings provide robust evidence that the different adaptive solutions are not mutually exclusive and can coexist within the same genetic background.

Nevertheless, we agree that experimentally verifying the compatibility and fitness of the four genetic adaptations described in Figure 6B (now Fig 6C) would further strengthen our conclusions. To this end, we plan to reconstruct all combinations of mutations observed at high frequency in the final evolved populations. We will then measure their fitness and compare it to that of the evolved populations, as well as to the theoretical expectations based on additivity currently presented in Figure 6C.

Minor points

Figures

• S4B: in the legend it should be explained that it is compared to ctf4D

We now report how the values were obtained in the figure legend:

(D = |anc%|-|reconstraucted%|)

-2A: the color code is not fully clear to me: what does green and blue indicate? higher and lower than 2%?

We apogise for not having included an explicit description of the color code in Figure 2A. Throughout the paper blue refers to glucose starvation (light blue for 0,25%, dark blue for 0,5%), while green refers to glucose abundance (light blue for 2%, dark blue for 8%). We now include a detailed description of the color code when it first appears (Fig 1B) and make sure is properly reported in all figure legends.

• S3A: the authors should show the statistical difference between WT and ctf4-null, which is mentioned as non-existent in p.6

The p value is now represented in Fig S3A

Text

• RNR1 is not really the gene with the highest score in Figure 5D, not even close: can you give a rationale for pin-pointing it (see also main point 3)?

The reviewer is correct. Perturbations of the mediator complex, which regulate the expression of most of RNA PolII transcripts, is expected to result in changes in the expression of a large set of genes. However, our focus on dNTPs and RNR1 is based on the following rationale:

1. Gene Ontology Enrichment Analysis: The downregulated genes in our dataset are enriched for the 'nucleotide metabolism' term, which includes pathways critical for dNTP production and directly linked to DNA replication and repair.

2. Role of RNR1: Among the downregulated genes, RNR1 stands out as it encodes the major subunit of ribonucleotide reductase, the rate-limiting enzyme in dNTP synthesis. This enzyme is essential for DNA replication, and cells experiencing constitutive DNA replication stress, as in our system, are particularly sensitive to changes in dNTP levels.

To make this rationale more explicit to the reader, we are adding the following sentence in the discussion:

Line 404: “Nucleotide metabolism, particularly ribonucleotide reductase, is essential for dNTP production. Given the role of dNTPs in regulating DNA replication and repair, the advantage of med14-H919P mutants in the ctf4Δ background may stem from reduced dNTP levels caused by the perturbed TID domain."

In addition, following the reviewers’ suggestions, we are conducting additional experiments to investigate the role of med14-H919P mutants in enhancing fitness under conditions of constitutive DNA replication stress (See response to reviewer #4). We anticipate that the final revised manuscript will offer further insights into the role of dNTPs or present alternative explanations for the observed phenomena.

• The med14-H919P mutation is observed in 22/48 wells. I guess the authors checked already: are some of these wells close to each other in the plate?

Correct. We took significant precautions in our experimental design to prevent cross-contamination, as outlined in the Materials and Methods section. Specifically, rows of ctf4Δ samples were alternated with rows of WT samples. Daily dilutions were then performed row by row using a 12 channels pipette. This approach ensured that any potential carry-over of cells would result in them being placed in wells containing a different genotype, where they would be eliminated by the consistent use of genotype-specific drugs.

As a result of these measures, we do not observe any distinct pattern of core genetic adaptation corresponding to the plate layout (Figure_rev_only 3). The only exception are mutations in IXR1, which appear in all ctf4Δ strains (albeit with different alleles, see supplementary File 3). Moreover, we reasoned that if a highly fit strain had invaded other wells, all the pre-existing mutations from its lineage would have been detected in those wells. However, apart from the recurrent ixr1 and rad9 mutations, which are also strongly adaptive, we find no evidence of shared mutations in wells carrying the med14-H919P allele (Figure_rev_only 4).

• Compensatory evolution of ctf4-null in 2% glucose is the experiment published by Fumasoni and Murray in eLife. In that paper, there is no trace of mutations in MED14. I think the authors should comment on this (different method for detecting putative compensatory mutations?).

We also noticed the absence of MED14 mutations in the eLife study by Fumasoni and Murray and find this discrepancy intriguing. One possible explanation lies in methodological differences. Our current study employed an improved version of the mutational analysis pipeline. However, we have not yet reanalyzed the original data from the previous study to determine whether MED14 mutations were present but undetected.

Interestingly, in the current study, we observed that in 2% glucose, MED14 mutations arose in only 3 out of 12 populations, a frequency lower than in other glucose conditions (Figure S6C). Assuming a similar frequency occurred in the 8 populations evolved in 2% glucose by Fumasoni and Murray (2020), one would expect only 2 populations to carry the mutation. This number falls below the threshold required for our algorithm to detect statistically significant parallelism.

Additionally, two significant experimental differences may also contribute to the observed discrepancy. First, the culture volumes and vessels differed: 10 mL cultures in tubes were used previously, whereas 1.5 mL cultures in 96-well plates were used in the current study.

• I may be mistaken, but Szamecz et al do not actually investigate whether different conditions result in different evolutionary trajectories (i.e., different genetics), and so their results may not be at odds with those presented here.

The reviewer is correct that Szamecz et al. do not explicitly test whether different conditions result in different evolutionary trajectories. However, in the section titled “Compensatory Evolution Generates Diverse Growth Phenotypes across Environments,” they examine how lines evolved in 2% YPD perform across various environments. They report how in roughly 50% of the cases tested, evolved lines showed either no improvement or even some lower fitness than the ancestor (Figure 5A).

While this could be explained by the accumulation of detrimental non-adaptive mutations in specific contexts, it likely implies that the adaptive strategies compensating for the original mutation in one environment do not confer similar benefits in other environments. This observation contrasts with our findings in Figure 6D, where we demonstrate that the main adaptive strategies provide a consistent benefit across diverse environments, including those with glucose, nitrogen, or phosphate abundance or starvation.

We have now modified the introduction, results and discussion to avoid misleading interpretations:

Line 42: “Szamecz and colleagues examined the evolutionary trajectories of 180 haploid yeast gene deletions over 400 generations (Szamecz et al., 2014). They found that, while fitness recovery occurred in the environment where evolution took place, the evolved lines often showed no improvement over their ancestors in other environments. This suggests that compensatory mutations beneficial in one environment often fail to restore fitness in others.”

Line 327: “A previous study in yeast showed how evolved lines which compensate for detrimental defects of gene deletions in standard laboratory conditions often failed to show fitness benefits compared to their ancestor when tested in other environments (Szamecz et al., 2014). We thus investigated the extent to which the core genetic adaptation to DNA replication stress was beneficial under alternative nutrient conditions.”

Line 422: “What could explain the discrepancies between our results, and previous studies on evolutionary repair highlighting the role of the environment in shaping evolutionary trajectories (Filteau et al., 2015), and the heterogeneous behavior of evolved lines in various environments (Szamecz et al., 2014)?”

typos

p.18, line 564 preformed -> performed

1. 6 line 189 with a strongly skew -> with a strong skew ?

Typos are now corrected in the main text

Reviewer #2 (Significance):

This is a well-done paper that could be of interest for the community of evolutionary biologists, scientists working on metabolism and cell division. It addresses an interesting problem, how metabolism affects compensatory evolution. Among the strengths: experiments are well done, the results are novel, the cross-talk between metabolism and evolutionary repair is intriguing. Among the weaknesses, the fact that the molecular explanations for the observations are only hypothesized and not tested experimentally. This is where the authors could improve the manuscript.

Reviewer #3 (Evidence, reproducibility and clarity):

This paper combines phenotypic and genomic data from an experimental evolution study in yeast to assess how repeatable evolution is in response to DNA replication stress. Importantly, the authors ask whether genotype by environment interactions influence repeatability of their evolved lines. To this end, the authors have constructed an elegant highly-replicated experiment in which two yeast genotypes (WT and CTF4 KO) were evolved under a variety of glucose levels for 1,000 generations. Recurrent mutations are found across many replicates, suggesting that repeatability is robust to GxE interactions. Of course, the authors correctly identify that these results are dependent on many particulars, as is always the case in biology, but provide a comprehensive discussion to accompany their results. I do not have any major comments to give, but simply some suggestions and points of clarification.

Major comments: N/A

Minor comments:

L19: I found the definition for compensatory evolution/mutations to be somewhat vague in the introduction (and subsequently throughout the text). It's clear that this was written for a more medical/physiological audience, but without a more explicit explanation of compensatory evolution/mutations, it became difficult to properly weigh some claims/discussions made by the authors later on. Do you define compensatory mutations as those which completely recover WT function/fitness, or are simply of opposite effect to the altered genotype? Others define "compensatory evolution" as simply any epistastically interacting amino acid substitutions (Ivankov et al, 2014). It would be nice to see more explicitly defined.

We thank the reviewer for highlighting the need for a precise definition of compensatory evolution and compensatory mutations. We recognize that the literature encompasses multiple definitions, including the one cited by the reviewer, which emphasizes compensatory mutations within the context of structural biology. This particular definition, prevalent in molecular evolution, was introduced by Kimura (Kimura, 1985) and is frequently used to explain the co-occurrence of amino acid mutations within a protein. These mutations offset each other’s defects, restoring or maintaining protein function. Here, however, we are using an older and broader definition of compensatory mutation, first introduced by Wright (Wright, 1964, 1977, 1982) and frequently used in evolutionary genomics (e.g., Moore et al., 2000; Szamecz et al., 2014; Rajon and Mazel, 2013; Eckartt et al., 2024). This definition includes any mutation in the rest of the genome that compensates (fully or partially) for another mutation's detrimental effects on fitness.

We have now included this definition in the introduction:

Line 19: “Compensatory evolution is a process by which cells mitigate the negative fitness effects of persistent perturbations in cellular processes across generations. This adaptation occurs through spontaneously arising compensatory mutations anywhere in the genome (Wright, 1964, 1977, 1982) that partially or fully alleviate the negative fitness effects of perturbations (Moore et al., 2000). The successive accumulation of compensatory mutations over evolutionary timescales progressively repair the cellular defects, ultimately restoring fitness.”

Line 361: “Our findings demonstrate that while glucose availability significantly affects the physiology and adaptation speed of cells under replication stress, it does not alter the fundamental genome-wide compensatory mutations that drive fitness recovery and evolutionary repair.”

Along these lines, I would have liked to see a more direct comparison/discussion of the degree to which deletion lines recovered. I can see from Fig 2E and Fig S2B that fitness increased quite a bit; would it not be possible to include a figure on the degree of compensation (basically relative fitness of evolved deletion lines - relative fitness of ancestral deletion lines)?

If the reviewer is suggesting calculating the difference between the evolved and ancestor fitness, the data is already in Figure S2B and S2D, defined as ‘Absolute fitness gains Δ’ and calculated as Δ = evo% - anc%.

If instead is suggesting to plot the fitness of evolved deletion lines (Y axis) against the relative fitness of ancestral deletion lines (X axis), we have now produced the plot is Figure S2F.

To better understand the extent of the fitness recovery in Ctf4 strains, we have also calculated and plotted the ‘relative fitness gain’ calculated as |evo%| / |anc%| *100 (Figure S2C)

We are now commenting on these comparisons in the following paragraph:

Line 171: “By generation 1000, both WT and ctf4Δ evolved lines achieved, on average, slightly higher fitness in low glucose compared to high glucose conditions (Fig S2B). However, due to the varying initial fitness of ctf4Δ cells across different glucose environments, they recovered the same extenct of the original defect (Fig S2C), displaying an opposite trend to WT, with increasing absolute fitness throughout the experiment as glucose concentration rose (Fig S2B vs S2D). The differint absolute fitness gains over the same number of generations highlight distinct mean adaptation rates (Fig 2B). These differences are evident when examining the evolutionary dynamics of the evolved lines over time (Fig 2C). Additionally, we approximated the fitness trajectories using the power law function (Fig 2C, dashed purple lines), previously proposed to describe long-term evolutionary dynamics in constant environments (Wiser et al., 2013). The parameter b in this formula determines the curve's steepness, and can be used to quantify the global fitness change over generations (Fig S2E). Collectively, these analyses demonstrate that, unlike WT cells, ctf4Δ lines adapt faster in the presence of high glucose. This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

L57: Another minor nitpick that just comes down to semantics. When discussing "96 parallel populations", it invokes a higher sense of replication than is actually present in the study. I would rephrase this to something along the lines of "12 replicate populations across 8 treatments under conditions of [...]".

We changed the sentence as follows:

Line 66: “We evolved 96 parallel populations of budding yeast, organized into 12 replicate lines, across four conditions of glucose availability (from starvation to abundance) with or without replication stress.”

L185-187: The wording here needs to be clarified. Be explicit in that are examine the ratio (or count) of synonymous to non-synonymous mutations here, otherwise the interpretations appears to be direct contradiction to the (as written) results. Only after viewing the supplemental figure was I able to figure out what exactly was meant here.

We changed the sentence as follows:

Line 212: “We found no significant differences in the numbers of synonymous mutations detected in evolved populations in WT and ctf4∆ populations (Fig. S3A). These results support the hypothesis that replication stress in ctf4∆ lines favors the retention of beneficial mutations, rather than simply increasing the overall mutation rate.”

L349-350: The authors observe higher rates of adaptation in deletion lines than WT lines, and discuss this in adequate detail. Although not explicitly mentioned, this is consistent with a diminishing returns epistasis model (that could be beneficial to discuss, but is not necessary), which has been implicated in modulating the degree of repeatability observed along evolutionary trajectories (Wünsche et al. 2017). Although definitely not required for this already very nice manuscript, I think it would be very rewarding if the authors were to eventually analyze fine-scale dynamics of phenotypic and genomic adaptation to mine for these putative interactions and their influence on repeatability.

We agree with the reviewer on how our results align with a model of diminishing returns epistasis. This pattern is apparent not only between ctf4Δ and WT lines but also among ctf4Δ lines evolved in different glucose conditions. This phenomenon likely arises from the interaction of various adaptive mutations, which we aim to explore further in a dedicated manuscript. However, until we do so, we prefer to refer generally to a pattern of declining adaptability. To explicit this trend we have now included Fig S2F and commented on it in the manuscript:

Line 181: “This evidence aligns with the declining adaptability observed in other studies (Moore et al., 2000; Kryazhimskiy et al., 2014; Couce & Tenaillon, 2015), where low-fitness strains consistently adapt faster than their more fit counterparts (Fig S2F).”

Line 388: "Our results are consistent with declining adaptability, as evidenced by the reduced rates of adaptation observed both between ctf4Δ and WT lines and among ctf4Δ lines evolved in different glucose conditions (Fig S2F)"

Reviewer #3 (Significance):

It is clear to me that a great deal of time and care has been put into this study and the preparation of this manuscript. The science and analyses are appropriate to answer the questions at hand, and it bodes well that whenever I had a question pop up while reading, they were typically answered immediately after. I think that this manuscript will be broadly relevant to both biologists both evolutionary and clinical, and was written in a way to be accessible to both.

As someone with an expertise in repeatable evolution, I felt most excited by the observation of so many parallel substitutions at a single amino acid across deletion lines. As the authors rightfully point out in the results and discussion, it's likely that this degree of robustness is highly dependent on the particular mechanism of disruption that cells experience. The authors then go above and beyond to functionally validate the putative molecular mechanisms of (repeatable) adaptation in this system. While it may not always be possible to accomplish in non-model organisms, such multi-modal approaches will be crucial to advance the field of repeatable evolution.

Reviewer #4 (Evidence, reproducibility and clarity):

The authors investigated the effects of DNA replication stress on adaptation in different nutrient availabilities by passaging wild-type and ctf4Δ Saccharomyces cerevisiae in media with varying levels of glucose over ~1000 generations. The ctf4Δ strain experiences increased DNA replication stress due to the deletion of a non-essential replication fork protein. The authors found differences in evolution between wild-type and ctf4Δ yeast, which held across different growth media. This study identified a compensatory single amino acid variant in Med14, a protein in the mediator complex of RNA polymerase II, that was specifically selected in ctf4Δ strains. The authors conclude that while environmental nutrient availability has implications for cell fitness and physiology, adaptation is largely independent and instead dependent on genetic background. The data provide excellent support for the key aspects of the models, although some details are (to me) overstated.

Major comments:

• A ctf4Δ mutant strain was used to investigate the effects of replication stress. Why was this mutant chosen instead of other deletions that cause different types of replication stress?

We appreciate the opportunity to clarify our rationale for choosing the ctf4Δ mutant. The following are the main reasons why we believe ctf4Δ strains represent an ideal tool to study a global perturbation of the DNA replication program over evolutionary timescales:

1. General replication stress: The absence of Ctf4 perturbs replication fork progression, leading to a spectrum of replication stress-related phenotypes, including DNA damage sensitivity, single-stranded DNA gaps, reversed forks (Abe et al., 2018; Fumasoni et al., 2015), checkpoint activation (Poli et al., 2012), cell cycle delays (Miles and Formosa, 1992), increased recombination (Alvaro et al., 2007), and chromosome instability (Kouprina et al., 1992). This broad disruption makes it an excellent model for observing global perturbations in replication processes. In contrast, other mutants typically affect specific enzymatic (e.g., POL32 and RRM3) or signaling (e.g., MRC1) functions, making them better suited to address specific questions.
2. Constitutive stress: Unlike drug-induced stress (e.g., Hydroxyurea; Krakoff et al., 1968) or conditional depletion systems (e.g., GAL1-POLε; Zhang et al., 2022), which cells can easily circumvent through single mutations, ctf4Δ enforces persistent replication stress. Its deletion cannot be complemented by a single mutation, ensuring a robust and consistent stress environment for evolutionary studies.

We have now modified the main text to convey these advantages in a concise form:

Line 91: “In the absence of Ctf4, cells exhibit multiple defects commonly associated with DNA replication stress, such as single-stranded DNA gaps and altered replication forks (Fumasoni et al., 2015), leading to basal cell cycle checkpoint activation (Poli et al., 2012). These defects result in severe and persistent growth impairments, cell cycle delays, elevated nucleotides pools and chromosome instability (Miles and Formosa, 1992; Kouprina et al., 1992; Poli at al., 2012), making ctf4Δ mutants an ideal model for studying the cellular consequences of general and constitutive replication stress over evolutionary time.”

It's not clear from the study that the effects are generalizable to other forms of replication stress.

As with any method to induce DNA replication stress (including commonly used drugs like HU) each approach inevitably affects replication in a specific manner. Testing the broader applicability of our conclusions would require evolving additional strains with different replisome perturbations. For instance, mutations in ELG1 and CTF18 (affecting the alternative Replication Factor C), POL30 (affecting the sliding clamp PCNA), POL32 (affecting Polε), RRM3 (protective helicase) and (MRC1 (coordinating leading strand activities and signalling to the checkpoint) would have to be taken into account. Furthermore, specific mutant alleles of Ctf4 that disrupt interactions with particular binding partners (Such as ctf4–4E and ctf4–3E, perturbing the interaction with the CMG helicase and accessory factors respectively) will be highly informative on which specific aspects of the replication stress generated by the lack of Ctf4 each adaptive mutation alleviate.

However, accommodating such extensive variability would inflate the sample size to an extent that will become unfeasible within the experimental design focused on capturing parallel evolution over a nutrient gradient (the primary focus of this study). We agree that this is an important question and intend to address it comprehensively in a dedicated future study.

• The authors could be clearer that a (the?) cause of the ctf4∆ fitness defect is spurious upregulation of RNR1. I don't think it is mentioned until the Discussion, but it is highly relevant to Fig 4, and to the adaptations one would expect from ctf4∆.

We thank the reviewer for the opportunity to clarify this aspect. We do not think that the fitness defects of ctf4∆ cells stem solely from the spurious upregulation of RNR1. However, we believe that a major aspect of the evolutionary adaptation is aimed at decreasing dNTP levels, potentially through different mechanisms. We are now mentionig increased dNTPs as major phenotype of ctf4∆ and commenting on the hypothesis more clearly in the discussion.

Line 93: “These defects result in severe and persistent growth impairments, cell cycle delays, elevated nucleotides pools and chromosome instability (Miles and Formosa, 1992; Kouprina et al., 1992; Poli at al., 2012)”

Line 409: “This condition will, in turn, be detrimental when proliferation rates are high (as in WT in high glucose) but beneficial under constitutive DNA replication stress (ctf4Δ), where cells experience spurious upregulation of dNTP production (Poli et al., 2012; Davidson et al., 2012).

• In Figure 1E, there is a very large spread in the relative fitness at 2% and 8% glucose, but this was not commented on. Is this heteroscedasticity expected?

The observed heteroscedasticity is expected. Our competition assays tend to exhibit increased variability when a strain approaches very low fitness levels. Specifically, as one strain nears extinction by the third day of competition, its abundance is estimated based on a much smaller number of events in the flow cytometer. Furthermore, we noticed a small number of reference cells carrying pACT1-yCerulean not showing strong fluorescence in 8% glucose. The nature of this effect is uncertain, and possibly linked to metabolism-linked changes in the cytoplasm. The combination of these two phenomena amplifies the impact of noise inherent to the methodology, leading to increased variability across replicates.

Nontheless, the overall decreasing fitness trend across glucose conditions, combined with the statistical significance observed between high and low glucose levels, collectively convey a roboust phenotype

• The med14-H919P mutant was highly selected in ctf4Δ strains, independent of glucose availability. Is this variant found in any natural yeast strains (i.e., are there environments that select for this variant)? Also, if this variant is found in natural strains, does it co-occur with other mutations that could affect DNA replication?

We agree that this is an intriguing question. To address it, we plan to explore existing databases of variants identified in S. cerevisiae natural isolates. Specifically, we will investigate whether the med14-H919P mutation is present in these strains, identify any potential environmental factors that may select for it, and assess whether it co-occurs with other mutations that could influence DNA replication processes.

• The statement on lines 271-273 is not particularly well-supported. The analysis of the Warfield data suggest that reduced expression of RNR1 could be causal, but the data don't go as far as showing how the med14 mutation is advantageous in ctf4∆. Further experimentation would be necessary to support the possibilities that the authors discuss.

The sentence the reviewer refers to is: “Overall, these results show how an amino acid substitution in the Med14 subunit of the mediator complex, putatively affecting transcription, is strongly selected, and advantageous, in the presence of constitutive DNA replication stress.” We are unsure which aspect of the statement is seen as unsupported. The mutation's strong selection in ctf4∆ is demonstrated in Figures 5A, 6A, and S4C, while its advantageous nature is supported by Figures 5B and S4B. Regarding the mechanism, we have been cautious with our phrasing, describing its effect on transcription as "putative" (Line 272) and suggesting that our observations “are compatible with” reduced dNTP availability in med14-H919P cells due to RNR1 downregulation (Line 361).

The main focus of this study is to explore how nutrient availability influences evolutionary dynamics and compensatory adaptation in cells lacking Ctf4. We believe the identification of a novel selected allele (Fig. 5A) and confirmation of its benefit across glucose conditions (Fig. 5B) serves as an excellent complement to the primary conclusions (present in the title). We invite the reviewer to consider that the molecular basis of such a phenotype is not mentioned in our abstract, as we believe that its precise characterization would require a dedicated study on Med14.

Nonetheless, we are encouraged by the reviewer’s interest in this newly identified compensatory mutant (also noted by Reviewer #2), and we are eager to perform further experiments to better understand the biological processes affected by this mutation. We plan to extend our work as follows:

Based on known phenotypes associated with perturbations of Med14, we propose the following novel hypotheses regarding the mechanism by which med14-H919P alleviates ctf4Δ defects:

1. Decreased replication-transcription conflicts: Conflicts between the transcription machinery and replication forks are known to cause fragile sites, leading to increased chromosome breaks and genomic instability (Garcia-Muse and Aguilera, 2016). A general reduction in PolII transcription during replication, resulting from perturbations of the mediator complex, could reduce these conflicts and mitigate the fitness defects observed in ctf4Δ cells.
2. Increased cohesin loading: We have demonstrated that amplification of the cohesin loader SCC2 is beneficial in the absence of Ctf4. Recent findings (Mattingly et al., 2022) indicate that the mediator complex recruits SCC2 to PolII-transcribed genes. The med14-H919P mutation may enhance the fitness of ctf4Δ cells by facilitating cohesin loading during DNA replication.
3. Decreased dNTP levels: As discussed in the manuscript, perturbations of Med14 subunits in the mediator complex reduce the expression of genes, including those associated with nucleotide metabolism. Notably, these include RNR1, the major subunit of ribonucleotide reductase. The med14-H919P mutation could benefit the ctf4Δ background by counteracting the reported spurious increase in dNTPs, which affects replication fork speed (Poli et al., 2012).

We plan to distinguish between these hypotheses using the following approaches. First, the proposed mechanisms underlying Hypotheses 1 and 3 suggest that med14-H919P is a loss-of-function mutation, while Hypothesis 2 implies a gain-of-function effect. Testing the impact of a heterozygous med14-H919P allele in a homozygous ctf4Δ strain will allow us to differentiate between these two categories of mechanisms. Additionally, we aim to investigate the molecular process affected by the med14-H919P allele by analyzing its genetic interactions with genes involved in replication-transcription conflicts, cohesin loading, and dNTP production (See also response to reviewer #2).

We believe that the results of these experiments will provide further insights on the mechanism of suppression exerted by med14-H919P in the presence of constitutive DNA replication stress, without diverting the reader from the main message of the paper.

• The authors comment that the med14-H919P mutant could have implications for the stability of Med14, based on computational modelling. Verifying the stability of the med14-H919P in vivo would strengthen this discussion.

We believe that in vivo and in vitro structural studies investigating the effect of this mutation on the stability and function of the Mediator complex are beyond the scope of this manuscript. These investigations would be more appropriately addressed in future, dedicated studies focused on these specific aspects.

• In the discussion, the authors propose that the context of the perturbation may influence the robustness of adaptation. A more detailed explanation of this point (including a discussion of the findings of other similar studies investigating different conditions) would be helpful to further bolster this section.

We are now supporting this concept more explicitly by commenting on other studies as follows:

Line 429: “Third, the environment’s influence on compensatory evolution may depend on the specific cellular module perturbed and its genetic interactions with other modules that are significantly influenced by environmental conditions. For example, the actin cytoskeleton, which must rapidly respond to extracellular stimuli, is likely to be more directly influenced by environmental factors (Filateau et al., 2015) compared to the DNA replication machinery, which operates within the nucleus and is relatively insulated from such changes. Supporting this idea, a study examining mutants’ fitness across diverse environments found that conditions such as different carbon sources or TOR inhibition, similar to those used in this study, primarily affected genes involved in vesicle trafficking, transcription, protein metabolism, and cell polarity. In contrast, genes associated with genome maintenance, as well as their epistatic interactions, were largely unaffected (Costanzo et al., 2021)”.

In addition, to further substantiate this hypothesis, we plan to re-analyze published datasets on fitness and epistatic interactions among genes in various environments, testing whether specific cellular modules are more prone to changes following shifts in nutrient conditions.

Minor comments: - Competitions were performed between ctf4Δ strains and a constructed strain with yCerulean integrated at ACT1. Is the fitness of the fluorescent strain comparable to the ancestral wild-type strain (i.e., in a competition between the ancestral WT and the fluorescent strain, does either have an advantage)?

We noticed a slight disadvantage of the reference strain compare to WT, likely due to the costs of the extra fluorescence reporter. However, the disadvantage is minimal, ranging from -0.5 to -2.5 depending on the glucose environment (raw measurments are reported supplementary file 1, sheet 5). To take this into account, all fitness reported in figures are normalized for the WT value measured in the same environment line 613: “Relative fitness of the ancestral WT strain was used to normalize fitness across conditions.​​”

• In Figure 3, the legends for panels B and C appear to be swapped. Discussion of Figure 3 on pages 6 and 7 appear to reference the wrong panels.

We are unsure about this typo. Main text and figure legend seem to refer to the appropriate panels, 3B for mutation fractions and 3C for mutation counts. Perhaps the organization of the panels with B being under A instead of on its right confounds the reader?

• In Figure 4A and B, having the same colour scale between both heatmaps is misleading, as the scales are different. Consider having the same scale across both heatmaps so that enrichments are visually comparable.

Following the reviewer’s suggestion we have have chosen a uniform heatmap to visually represent GO terms enrichment in WT and ctf4∆ genetic backgrounds.

• In Figure 4C, having a legend in the figure for node size would be helpful to understand the actual number of populations with mutations in each gene.

A legend for node size has now being added next to Figure 4C.

Reviewer #4 (Significance):

In this study, a high-throughput evolution experiment uncovered the effects of genetic background on the development of adaptive mutations. The authors were able to identify a single amino acid variant of Med14 (med14-H919P) that was positively selected in ctf4Δ. Furthermore, they demonstrated the causality of med14-H919P in conferring a fitness advantage in ctf4Δ. The novelty of this mechanistic finding opens future avenues of investigation regarding the interaction network of the mediator complex in conditions of DNA replication stress. A limitation of the study is that only one mechanism of replication stress was assessed (ctf4Δ). Other gene mutations that cause replication stress would be interesting to assess and would provide a more thorough investigation of the effects of DNA replication factors on evolvability. This work will be of interest to researchers in the population genetics and genotype-by-environment fields, as it suggests the robustness of evolvability to environmental factors in the specific condition of DNA replication stress. As discussed by the authors, this finding differs from other works that have linked environmental conditions to adaptive evolution to different conditions, and is concordant with work that indicates the robustness of genetic interactions to environmental stresses. Furthermore, the identification of the highly-selected med14-H919P variant will be of interest to the DNA replication field. There is the potential for future work investigating the role of Med14 in mediating the response to DNA replication stress in both yeast and mammalian cell contexts, since the authors note that there are links between altered mediator complex regulation and cancers. Although I suspect that the very different regulation of RNR in mammalian cells makes it unlikely that the kind of upregulation of dNTP pools seen in ctf4∆ would be induced by replication stress in mammalian cells.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

Referee #4