Author Response

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

Reviewer #2 (Public Review)

Weaknesses

1) The usage of young growing mice (8-10 weeks) versus adult mice (>4 months) in the murine mechanical overload experiments. The usage of adult mice would be preferable for these experiments given that maturational growth may somehow affect the outcomes.

The basis for this critique is not clear as it has been shown that the longitudinal growth of bones is complete by ⁓8 weeks of age (e.g., PMID: 28326349, and 31997656). These studies, along with others, also indicate that 8 weeks is a post-pubescent age in mice. For these reasons, 8 weeks of age was viewed as being representative of the human equivalent of when people start to perform resistance exercise with the goal of increasing muscle mass. Also, it’s important to consider that the mice were 10-12 weeks of age when the muscles were collected which would be equivalent to a human in their lower 20’s. In our human study, the mean age of the subjects was 23. Given the above points, it’s hard for us to appreciate why the use of mice that started at 8-10 weeks of age is viewed as a weakness. With that being said, we recognize that there may be age-related changes in mechanisms of mechanical load-induced growth, but it was not our intent to address this topic.

1b) No consideration for biological sex.

We appreciate this point and we agree that sex is an important variable to consider. In this study, we explored an unchartered topic and therefore we wanted to minimize as many known variables as possible. We did that, in part, by focusing specifically on male subjects. In the future, it will certainly be important to explore whether sex (and age) impact the structural adaptations that drive the mechanical load-induced growth of muscle fibers.

2) Information on whether myofibrillogenesis is dependent on hypertrophy induced by loading, or just hypertrophy in general. To provide information on this, the authors could use, for instance, inducible Myostatin KO mice (a model where hypertrophy and force production are not always in lockstep) to see whether hypertrophy independent from load induces the same result as muscle loading regarding myofibrillogenesis.

This is a great suggestion, but it goes beyond the intended scope of our study. Nevertheless, with the publication of our FIM-ID methodology, the answer to this and related questions can now be obtained in a time- and cost-effective manner.

3) Limited information on Type 1 fiber hypertrophy. A "dual overload" model is used for the mouse where the soleus is also overloaded, but presumably, the soleus was too damaged to analyze. Exploring hypertrophy of murine Type 1 fibers using a different model (weight pulling, weighted wheel running, or forced treadmill running) would be a welcome addition.

The point is well taken and further studies that are aimed at determining whether there are differences in how Type I vs. Type II fibers grow would be an excellent subject for future studies.

Reviewer #3 (Public Review)

1) Supplemental Figure 1 is not very clear.

Supplemental Figure 1 is now presented as Supplemental Figure 2. We carefully reexamined this figure and, in our opinion, the key points have been appropriately conveyed. We would be more than happy to revise the figure, but we would need guidance with respect to which aspect(s) of the figure were not clear to the reviewer.

Reviewer #1 (Recommendations For The Authors)

Introduction.

1) I do not think the first paragraph is really necessary. Cell growth is a fundamental property of cell biology that requires no further justification.

We believe that it is essential to remind all readers about the importance of skeletal muscle research. For some, the detrimental impact of skeletal muscle loss on one’s quality of life and the greater burden on the healthcare system may not be known.

2) I prefer "fundamental" over "foundationally".

All mentions of the word “foundational” and “foundationally” have been changed to “fundamental” and “fundamentally.”

3) As usual for the Hornberger lab, the authors do an excellent job of providing the (historical) context of the research question.

Thank you for this positive comment.

4) I prefer “Goldspink” as “Dr. Goldspink” feels too personal especially when you are critical of his studies.

All instances of “Dr.” have been removed when referring to the works of others. This includes Dr. Goldspink and Dr. Tokuyasu.

5) Fourth paragraph, after reference #17. I felt like this discussion was not necessary and did not really add any value to the introduction.

We believe that this discussion should remain since it highlights the widely accepted notion that mechanical loading leads to an increase in the number of myofibrils per fiber, yet there is no compelling data to support this notion. This discussion highlights the need for documented evidence for the increase in myofibril number in response to mechanical loading and, as such, it serves as a major part of the premise for the experiments that were conducted in our manuscript.

6) The authors do a nice job of laying out the challenge of rigorously testing the Goldspink model of myofiber hypertrophy.

Thank you!

Results

1). For the EM images, can the authors provide a representative image of myofibril tracing? From the EM image provided, it is difficult to evaluate how accurate the tracing is.

-Representative images and an explanation of myofibril calculation have been provided in Supplemental Figure 5.

2) In the mouse, how does the mean myofibril CSA compare between EM and FIM-ID?

Author response image 1.

The above figures compare the myofibril CSA and fiber CSA measurements that were obtained with EM and FIM-ID for all analyzed fibers, as well as the same fibers separated according to the fiber type (i.e., Ox vs. Gly). The above figure shows that the FIM-ID measurements of myofibril CSA were slightly, yet significantly, lower than the measurements obtained with EM. However, we believe that it would be misleading to present the data in this manner. Specifically, as shown in Fig. 4C, a positive linear relationship exists between myofibril CSA and fiber CSA. Thus, a direct comparison of myofibril CSA measurements obtained from EM and FIM-ID would only be meaningful if the mean CSA of the fibers that were analyzed were the same. As shown on the panel on the right, the mean CSA of the fibers analyzed with FIM-ID was slightly, yet significantly, lower than the mean CSA of the fibers analyzed with EM. As such, we believe that the most appropriate way to compare the measurements of the two methods is to express the values for the myofibril CSA relative to the fiber CSA and this is how we presented the data in the main figure (i.e., Fig. 4E).

3) Looking at Fig. 3D, how is intermyofibrillar space calculated when a significant proportion of the ROI is odd-shaped myofibrils that are not outlined? It is not clear how the intermyofibrillar space between the odd-shaped myofibrils is included in the total intermyofibrillar space calculation for the fiber.

The area occupied by the intermyofibrillar components is calculated by using our custom “Intermyofibrillar Area” pipeline within CellProfiler. Briefly, the program creates a binary image of the SERCA signal. The area occupied by the white pixels in the binary image is then used to calculate the area that is occupied by the intermyofibrillar components. To help readers, an example of this process is now provided in supplemental figure 4.

4) What is the average percentage of each ROI that was not counted by CP (because a myofibril did not fit the shape criteria)? The concern is that the method of collection is biasing the data. In looking at EM images of myofibrils (from other studies), it is apparent that myofibrils are not always oval; in fact, it appears that often myofibrils have a more rectangular shape. These odd-shaped myofibrils are excluded from the analysis yet they might provide important information; maybe these odd-shaped myofibrils always hypertrophy such that their inclusion might change the overall conclusion of the study. I completely understand the challenges of trying to quantify odd-shaped myofibrils. I think it is important the authors discuss this important limitation of the study.

First, we would like to clarify that myofibrils of a generally rectangular shape were not excluded. The intent of the filtering steps was to exclude objects that exhibited odd shapes because of an incomplete closure of the signal from SERCA. To illustrate this point we have annotated the images from Figure 3B-D with a red arrow which points to a rectangular object and blue arrows which point to objects that most likely consisted of two or more individual myofibrils that were falsely identified as a single object.

Author response image 2.

We appreciate the reviewer's concern that differences in the exclusion rates between groups could have biased the outcomes. Indeed, this was something that we were keeping a careful eye on during our analyses, and we hope that the reviewer will take comfort in knowing that objects were excluded at a very similar rate in both the mouse and human samples (44% vs. 46% for SHAM vs. MOV in mice, and 47% vs. 47% for PRE vs. POST in humans). We realize that this important data should have been included in our original submission and it is now contained with the results section of the revised version of our manuscript. Hopefully the explanation above, along with the inclusion of this data, will alleviate the reviewers concerns that differences between the groups may have been biased by the filtering steps.

Discussion.

1) I think the authors provided a balanced interpretation of the data by acknowledging the limitation of having only one time-point. i.e., not being able to assess the myofibril splitting mechanism.

Thank you!

2) I think a discussion on the important limitation of only quantifying oval-shaped myofibrils should be included in the discussion.

Please refer to our response to comment #4 of the results section.

Reviewer #2 (Recommendations For The Authors)

Overall, this is a thoughtful, clear, and impactful manuscript that provides valuable tools and information for the skeletal muscle field. My specific comments are as follows:

1) In the introduction, I really appreciate the historical aspect provided on myofbrillogenesis. As written, however, I was expecting the authors to tackle the myofibril "splitting" question in greater detail with their experiments given the amount of real estate given to that topic, but this was not the case. Consider toning this down a bit as I think it sets a false expectation.

We acknowledge that the study does not directly address the question about myofibril splitting. However, we believe that it is important to highlight the background of this untested theory since it serves as a major part of the premise for the experiments that were performed.

2) In the introduction, is it worth worth citing this study? https://rupress.org/jcb/articlepdf/111/5/1885/1464125/1885.pdf.

This is a very interesting study but, despite the title, we do not believe that it is accurate to say that this study investigated myofibrillogenesis. Instead (as illustrated by the author in Fig. 9) the study focused on the in-series addition of new sarcomeres at the ends of the pre-existing myofibrils (i.e., it studied in-series sarcomerogenesis). In our opinion, the study does not provide any direct evidence of myofibrillogenesis, and we are not aware of any studies that have shown that the chronic stretch model employed by the authors induces myofibrillogenesis. However, numerous studies have shown that chronic stretch leads to the in-series addition of new sarcomeres.

3) Is there evidence for myofbrillogenesis during cardiac hypertrophy that could be referenced here?

This is a great question, and one would think that it would have been widely investigated. However, direct evidence for myofibrillogenesis during load-induced cardiac hypertrophy is just as sparse as the evidence for myofibrillogenesis during load-induced skeletal muscle hypertrophy.

4) In the introduction, perhaps mention that prolonged fixation is another disadvantage of EM tissue preparation. This typically prevents the usage of antibodies afterwards, whereas the authors have been able to overcome this using their method, which is a great strength.

Thank you for the suggestion. This point has been added the 5th paragraph of the introduction.

5) In the introduction, are there not EM-compatible computer programs that could sidestep the manual tracing and increase throughput? Why could software such as this not be used? https://www.nature.com/articles/s41592-019-0396-9

While we agree that automated pipelines have been developed for EM, such methods require a high degree of contrast between the measured objects. With EM, the high degree of contrast required for automated quantification is rarely observed between the myofibrils and the intermyofibrillar components (especially in glycolytic fibers). Moreover, one of the primary goals of our study was to develop a time and cost-effective method for identifying and quantifying myofibrils. As such, we developed a method that would not require the use of EM. We only incorporated EM imaging and analysis to validate the FIM-ID method. Therefore, utilizing an EM-compatible program to sidestep the manual tracing would have sped up the validation step, but it would not have accomplished one of the primary goals of our study.

6) In the results, specifically for the human specimens, were "hybrid" fibers detected and, if so, how did the pattern of SERCA look? Also, did the authors happen to notice centrallynucleated muscle fibers in the murine plantaris after overload? If so, how did the myofibrils look? Could be interesting.

For the analysis of the human fibers, two distinct immunolabeling methods were performed. One set of sections was stained for SERCA1 and dystrophin, while the other set was stained for SERCA2 and dystrophin. In other words, we did not perform dual immunolabeling for SERCA1 and SERCA2 on the same sections. Therefore, during the analysis of the human fibers, we did not detect the presence of hybrid fibers. Furthermore, while we did not perform nuclear staining on these sections, it should be noted that nuclei do not contain SERCA, and to the best of our recollection, we did not detect any SERCAnull objects within the center of the fibers. Moreover, our previous work has shown that the model of MOV used in this study does not lead to signs of degeneration/regeneration (You, Jae-Sung et al. (2019). doi:10.1096/fj.201801653RR). Therefore, it can be safely assumed that very few (if any) of the fibers analyzed in this study were centrally nucleated.

7) In the Results, fixed for how long? This is important since, at least in my experience, with 24+ hours of fixation, antibody reactivity is significantly reduced unless an antigen retrieval step is performed (even then, not always successful). Also, presumably these tissues were drop-fixed? These details are in the Methods but some additional detail here could be warranted for the benefit of the discerning and interested reader.

For both the mouse and human, the samples were immersion-fixed (presumably the equivalent of “drop-fixed”) in 4% paraformaldehyde in 0.1M phosphate buffer solution for a total of 24 hours (as described in the Methods section). We agree that prolonged aldehyde fixation can affect antibody reactivity; however, the antibodies used for FIM-ID did not require an antigen retrieval step.

8) In the results regarding NADH/FAD autofluorescence imaging, a complimentary approach in muscle was recently described and could be cited here: https://journals.physiology.org/doi/full/10.1152/japplphysiol.00662.2022

We appreciate the reviewer’s recommendation to add this citation for the support of our method for fiber type classification and have added it to the manuscript in the second paragraph under the “Further refinement and validation of the automated measurements with FIM-ID” subsection of the Results as citation number 57.

9) In the results, "Moreover, no significant differences in the mean number of myofibrils per fiber CSA were found when the results from the FIM-ID and EM-based measurements were directly compared, and this point was true when the data from all analyzed fibers was considered..." Nit-picky, but should it be "were considered" since data is plural?

Thanks, this error was corrected.

10) In the discussion, are the authors developing a "methodology" or a "method"? I think it may be the latter.

We agree that “method” is the correct term to use. Instances of the word “methodology” have been replaced with “method.”

11) In the discussion, since the same fibers were not being tracked over time, I'm not sure that saying "radial growth" is strictly correct. It is intuitive that the fibers were growing during loading, of course, but it may be safer to say "larger fibers versus control or the Pre sample" or something of the like. For example, "all the fiber types that were larger after loading versus controls" as opposed to "showed significant radial growth"

While we agree that the fiber size was not tracked over time, the experiments were designed to test for a main effect of mechanical loading. Therefore, we are attributing the morphological adaptations to the mechanical loading variable (i.e., mechanical loadinduced growth). The use of terms like “the induction of radial growth” or “the induction of hypertrophy” are commonly used in studies with the methods employed in this study. Respectfully, we believe that it would be more confusing for the readers if we used the suggested terms like "all the fiber types that were larger after loading versus controls". For instance, if I were the reader I would think to myself… but there fiber types that were larger than others before loading (e.g., Ox vs. Gly), so what are the authors really trying to talk about?

12) I would suggest making a cartoon summary figure to complement and summarize the Methods/Results/Discussion

Thank you for this suggestion. We created a cartoon that summarizes the overall workflow for FIM-ID and this cartoon is now presented in Supplemental Figure 1.

To me, Hobbes is saying that In simple terms, people are generally equally smart. The idea that some are wiser might be because individuals often think they are smarter than others. Hobbes argues that, in reality, people are more equal in their mental abilities than they realize.

This is such an interesting point. We like to assume that we better than, in any aspect, our peers, but when it comes down to it, we are all equal. We all carry that belief that we are better then someone else, believing we are unequal to said person. They may carry that same idea about you, they think they outrank you in some aspect, making them believe they are better then you. Therefore, we are all the same. The same arrogant people.

I think we need to understand the nuances of recommendation algorithms, which is critical in addressing their influence on individual experiences. As these systems are designed to enhance user interaction, they can inadvertently perpetuate biases and present content that may not always align with the best interests or intentions of the users.

In this section, Adam responds to Eve as to why the stars and heavens shine. He explains to her that the sun must shine over all the earth, for those who will inhabit it in the future. They sleep at night, so that they may work harder in the day. He also talks about various "celestial voices"(4. 682) that he has heard at night, praising the glory of God. This heavenly chorus will protect them, as they “divide the night”(4. 688) to keep watch over Adam and Eve, while continuing to exalt their Creator. While reading this section, it seemed to me that the difference between night and day was emphasised heavily. This is an important distinction to make, as God is attributed as the giver of light, and Satan as a bringer of darkness.

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

---

1. General Statements

We thank the reviewers for their excellent work that greatly improved our work. We are very content that reviewer #1 considered our work to be “novel, interesting and important for understanding the mitochondrial biology of PD”. This reviewer also valued our work as “a significant advancement” and suggested further study of the relationship of CISD1 (dimerization) to general mitophagy/autophagy. We addressed this in the already transferred revision (version 1, v1).

Also reviewer #2 considered our work to be “an exciting and well-executed piece of research focusing on the defects in iron homeostasis observed in Parkinson's disease which a wide audience will appreciate”. This reviewer had a very specific suggestion on how to improve our manuscript which makes a lot of sense and is feasible. As the suggested experiments include fly breeding and behavioral analysis, these experiments will be included in the second revision to be uploaded as soon as possible (version 2, v2).

Finally, reviewer #3 gathered that parts of our results “are confirmatory to recently published work” but also appreciated that our results established that iron-depleted apo-Cisd is an important determinant of toxicity which has not been shown before. I would like to comment here, that in contrast to the paper mentioned by this reviewer, our contribution includes data from dopaminergic neurons obtained from human patients suffering from familial Parkinson’s disease that demonstrate the same increase in apo-Cisd levels as the flies. This reviewer mainly suggested that the manuscript would be improved by a more balanced discussion of the strengths and weaknesses of the study and more circumspection in interpretation of data which we did in the revised version of our manuscript. We also added data on the expression levels of Cisd and apo-Cisd in transgenic flies as also suggested.

2. Description of the planned revisions

Reviewer #1 (Evidence, reproducibility and clarity):

Summary: The manuscript focuses on mitochondrial CISD1 and its relationship to two Parkinson's disease (PD) proteins PINK1 and Parkin. Interestingly, CISD1 is a mitochondrial iron sulfur binding protein and an target of Parkin-mediated ubiquitinylation. Disruption of iron metabolism and accumulation of iron in the brain has long since been reported in PD but the involvement of iron sulfur binding is little studied both in vivo and in human stem cell models of PD. This work addresses the relationship between CISD1 and two mitochondrial models of PD (PINK1 and Parkin) making use of in vivo models (Drosophila), PINK1 patient models (iPSC derived neurons) and Mouse fibroblasts. The authors report a complex relationship between CISD1, PINK1 and Parkin, where iron-depleted CISD1 may illicit a toxic gain of function downstream of PINK1 and Parkin.

Major comments:

The conclusions are overall modest and supported by the data. One question remains unaddressed. Is mitochondrial CISD1 a downstream target that specifically mediates PINK1 and Parkin loss of function phenotypes or are the phenotypes being mediated because CISD1 is downstream of mitophagy in general?

It would be interesting to know what happens to CISD1 (dimerization?) upon initiation of mitophagy in wild type cells? Would dissipation of mitochondrial membrane potential be sufficient to induce changes to CISD1 in wild type cells or PINK1 deficient cells? Since iron chelation is a potent inducer of mitophagy (Loss of iron triggers PINK1/Parkin-independent mitophagy. George F G Allen, Rachel Toth, John James, Ian G Ganley. EMBO Reports (2013)14:1127-1135) it would be useful to show one experiment addressing the role of CISD1 dimerization under mitochondrial depolarizing and non-depolarizing conditions in cells.

Based on the overall assumption of the reviewer that our work is “novel, interesting and important for understanding the mitochondrial biology of PD” and “a significant advancement” we understand the word “modest” here as meaning “not exaggerated”. To address this question, we studied CISD1 dimerization in response to more classical activators of mitophagy namely FCCP and antimycin/oligomycin which had no significant effect on dimerization suggesting that this phenotype is more pronounced under iron depletion. These data are shown in the new Fig. 2c.

Alternatively, the authors should discuss the topic of mitophagy (including PINK1-parkin independent mitophagy), the limitation of the present study not being able to rule out a general mitophagy effect and previous work on the role of iron depletion on mitophagy induction in the manuscript.

The data and the methods are presented in such a way that they can be reproduced.

The experiments are adequately replicated and statistical analysis is adequate.

Minor comments:

Show p values even when not significant (ns) since even some of the significant findings are borderline < p0.05.

Here, I decided to leave it as it is, because the figures became very cluttered and less easy to understand. Borderline findings are however indicated and mentioned in the text.

Because the situation for CISD1 is complicated (overexpression, different models etc.) it would be helpful if in the abstract the authors could summarize the role. E.g. as in the discussion that iron-depleted CISD1 could represent a toxic function.

The abstract has been completely rewritten and now mentions the potential toxic function of iron-depleted CISD1.

If there is sufficient iron (accumulation in PD) why would CISD1 be deactivated? Perhaps that could be postulated or discussed in a simplified way?

We actually think that apo-CISD1 without its iron/sulfur cluster is incapable of transferring its Fe/S cluster to IRP1 and IRP2. This then results in increased levels of apo-IRP1/2 and subsequent changes that lead to iron overload. Such a sequence of events would place CISD1 upstream of the changes in iron homeostasis observed in PD and models of PD. This is now discussed in more detail.

In the methods section both reducing and non-reducing gel/Western blotting is mentioned but the manuscript only describes data from blots under reducing conditions. Are there blots under non-reducing conditions that could be shown to see how CISD1 and dimerized CISD1 resolve?

We now show these blots as supplemental data in new supplemental Figure 2.

In the results section, PINK1 mutant flies, it is said that the alterations to CISD1 (dimerization) are analogous to the PINK1 mutation patient neurons. The effect is seen in old but not young flies. Since iPSC-derived neurons are relatively young in the dish, would one not expect that young flies and iPSC-derived neurons have similar CISD1 phenotypes? Could the authors modify the text to reflect that? or discuss the finding in further context.

We only studied one time point in PINK1 mutation patient neurons and controls. It would indeed be interesting whether neuronal aging (as far as this can be studied in the dish) would result in increased CISD1 dimerization. This is now discussed.

Reviewer #1 (Significance):

The strengths of this work are in the novelty of the topic and the use of several well established in vivo and cell models including patient-derived neurons. The findings discussed in the text are honest and avoid over-interpretation. The findings are novel, interesting and important for understanding the mitochondrial biology of PD.

We thank the reviewer for their kind words.

Limitations include the lack of strong phenotypes in the CISD1 models and the lack of robust, sustained and consistent increase in CISD1 dimers in the patient and fly models (just significant because of variability). The relationship of CISD1 (dimerization) to general mitophagy/autophagy is not shown here.

We do not completely agree with the assumption that all CISD1 models lack a strong phenotype. At least the CISD1-deficient fibroblasts exhibit a strong phenotype consisting of fragmented mitochondria and increased oxidative stress. The lack of a strong phenotype in Cisd-deficient flies could actually hint to a potential compensatory mechanism that could also protect the Pink1 mutant x Cisd-deficient double-knockout flies. It is correct that the increase in CISD1/Cisd dimers in the PD models are not overwhelming but – as also mentioned by the reviewer – this could be increased in “older” cultures. This is now discussed in more detail. As suggested by the reviewer, we have now added experiments that study the relationship between CISD1 dimerization and conventional mitophagy as described above.

There is a significant advancement. So far researchers were able to describe the importance of iron metabolism in PD (For example refer to work from the group of Georg Auburger such as PMID 33023155 and discussion of therapeutic intervention such as reviewed by Ma et al. PMID: 33799121) but few papers describe involvement of iron sulfur cluster proteins specifically (such as Aconitase) in relation to PINK1 and parkin (these are cited). The fact that CISD1 is a protein of the mitochondrial outer membrane makes it particularly interesting and further studies looking more closely at the interaction of CISD1 with mitochondrial proteins associated with PD will be of interest.

We thank the reviewer for pointing out these excellent publications. Key et al present an enormous wealth of data on protein dysregulation of wildtype and Pink1-/- fibroblast cell lines upon perturbation of the iron homeostasis (Key et al, 2020). Both cell lines exhibit a downregulation of CISD1 levels upon iron deprivation with the agent 2,2′ -Bipyridine possibly as a compensatory mechanism to limit the toxic gain of function of iron-depleted CISD1. The other paper, Ma et al. is a recent review on changes in iron homeostasis in PD and PD models (Ma et al, 2021). Both papers are now cited in the manuscript.

This paper describes CISD1 as a new and relevant player in PINK1 and Parkin biology. Further work could lead to exploration of whether CISD1 could be a therapeutic target, considering its role in maintaining mitochondrial redox and mitochondrial health. This is of particular interest to mitochondrial biologists and pre-clinical research in PD.

This preprint was reviewed by three scientists whose research focus in the mitochondrial biology underlying Parkinson's disease. The group has a special interest in the functions of the mitochondrial outer membrane. We work with several cell models of Parkinson's disease and work with patient donated samples. We do not have expertise in Drosophila models of PD nor the quantification of iron described in the manuscript.

Reviewer #2 (Evidence, reproducibility and clarity):

Summary: In the paper entitled 'Mitochondrial CISD1 is a downstream target that mediates PINK1 and Parkin loss-of-function phenotypes', Bitar and co-workers investigate the interaction between CISD1 and the PINK1/Parkin pathway. Mutations in PINK1 and PARKIN cause early onset Parkinson's disease and CISD1 is a homodimeric mitochondrial iron-sulphur binding protein. They observed an increase in CISD1 dimer formation in dopaminergic neurons derived from Parkinson's disease patients carrying a PINK1 mutation. Immuno-blots of cells expressing CISD1 mutants that affects the iron sulphur cluster binding and as well as cells treated with iron chelators, showed that the tendency of CISD1 to form dimers is dependent on its binding to iron-sulphur clusters. Moreover, the Iron-depleted apo-CISD1 does not rescue mitochondrial phenotypes observed in CISD1 KO mouse cells. Finally, In vivo studies showed that overexpression of Cisd and mutant apo-Cisd in Drosophila shortened fly life span and, using a different overexpression model, apo-Cisd caused a delay in eclosion. Similar as patient derived neurons, they observed an increase in Cisd dimer levels in Pink1 mutant flies. Additionally, the authors showed that double mutants of Cisd and Pink1 alleviated all Pink1 mutant phenotypes, while double mutants of Prkn and Cisd rescued most Prkn mutant phenotypes.

Major comments:

1) The authors observed an increase in the levels of Cisd dimers in Pink1 mutant flies and removing Cisd in Pink1 mutant background rescues all the mutant phenotypes observed in Pink1 mutant flies, suggesting that the Cisd dimers are part and partial of the Pink1 mutant phenotype. The authors also generated a UAS_C111S_Cisd fly which can overexpress apo-Cisd. Overexpression of the C111S_Cisd construct with Tub-Gal4 showed a developmental delay. Since apo-Cisd forms more dimeric Cisd, my question is: does the strong overexpression (e.g. with Tub-Gal4) of the C111S_Cisd in wild type flies shows any of the Pink1 mutant phenotypes? If not, the authors should mention this and elaborate on it.

We thank the reviewer for their comments. In fact, we only observed very few flies ecclosing after overexpression of wildtype Cisd or C111S Cisd using the strong tubP-Gal4 driver during development. We considered these very few flies to be escapees (also indicated by the rather low induction of Cisd mRNA suggesting compensatory downregulation) and only used them to conduct the analysis shown in Figure 4c-e. This is now mentioned in more detail in the manuscript.

2) Figure 6g: Shows the TEM pictures of the indirect flight muscles of Pink1 mutant flies and Pink1, Cisd double mutants. To me, the Picture of Pink1 mutant mitochondria is not very convincing. We expect swollen (enlarged) mitochondria with disrupted mitochondrial matrix. However, this is not clear in the picture. Moreover, in my opinion, Figure6 g, is missing an EM Picture of the Cisd mutant indirect flight muscles.

We now show exemplary pictures from Pink1 mutant and DKO in a higher magnification which better demonstrate the rounded Pink1 mutant mitochondria and the disrupted cristae structure. EM pictures of all four genotypes in different magnifications are now shown in new supplemental Figure 6.

3) OPTIONAL: The authors suggest that most probably apo-Cisd, assumes a toxic function in Pink1 mutant flies and serves as a critical mediator of Pink1-linked phenotypes. If this statement is correct, we can hypothesize that increasing apo-Cisd in Pink1 mutant background should worsen the pink1 mutant defects.

Therefore, I suggest overexpressing Cisd1 wild type (and/or C111S Cisd) in pink1 mutant flies, as pink1 is on the X chromosome, and mild overexpression of Cisd1 with da is not lethal, these experiments could be done in 3-4 fly crosses and hence within 1.5 - 2 months.

We have set up this experiment and will report in the second revision (v2) of our manuscript.

Since Pink1 mutant flies contain higher levels of endogenous Cisd dimers, we can expect that overexpression of wild type Cisd will result in an even stronger increase of dimers. If these dimers indeed contribute to Pink1 mutant phenotypes we can expect that overexpression of Cisd will result in a worsening of the Pink1 mutant phenotypes.

We have set up this experiment and will report in the second revision of our manuscript.

Minor Comments:

-) In the Introduction (Background) there are some parts without references:

E.g., there is not a single reference in the following part between

'However, in unfit mitochondria with a reduced mitochondrial membrane potential ...&... compromised mitochondria safeguards overall mitochondrial health and function.'

We thank the reviewer for pointing out this flaw. We have now added a suitable reference to the introduction.

-) In the introduction there is some confusion about the nomenclature used in the article: e.g. following comments are made in the text: Cisd2 (in this publication referred to as Dosmit) or fly Cisd2 (in this publication named MitoNEET).

However, the names Dosmit and MitoNEET do not appear in the manuscript (except in references)

The literature and nomenclature for CISD1 are indeed confusing. We have now revised the introduction.

-) Figure 1: I am not sure why some gels are shown in this figure. The two last lanes of figure 1c are redundant and Figure 1c' which is also not mentioned in the text, is also a repetition of figure 1c.

The blots in 1c and 1c’ represent all data points (different patients and different individual differentiations) shown in the quantification in 1d. This is now explained better in the revised manuscript.

-) The authors mention in material and methods that T2A sites are used at the C-terminus of CISD1 to avoid tagging of CISD1. However, this is not entirely true as T2A will leave some amino acids (around 20) after the self-cleaving and therefore CISD1 will be tagged.

This is indeed true and we have now changed the wording in the revised manuscript.

-) In figure 5 P1 is used to abbreviate Pink1 mutants, however P1, to me, refers to pink1 wild type. It would be clearer to abbreviate Pink1 mutants as P1B9 in the graphs as B9 is the name of the mutant pink1 allele.

We thank the reviewer for pointing out this flaw. We have now altered Fig. 5 to be clearer.

-) In figure 7: Parkin is abbreviated both as Prkn and as Park

We thank the reviewer for pointing out this flaw, we indeed mixed up both names because it is complicated. The gene symbol is Prkn, the fly line is called Park25. We have now clarified this in the text and Fig. 7.

-) I suggest changing the title. Recently an article (Ham et al, 2023 PMID: 37626046) was published showing similar genetic interactions between Pink1/Prkn and Cisd. However, the article of Ham et al, 2023 was focused on Pink1/Prkn regulation of ER calcium release, while this article is more related to iron homeostasis. I suggest that the title shows this distinction.

This is indeed a very good suggestion. We have now altered the title to “Iron/sulfur cluster loss of mitochondrial CISD1 mediates PINK1 loss-of-function phenotypes”.

Reviewer #2 (Significance):

In general, this is an exciting and well-executed piece of research focusing on the defects in iron homeostasis observed in Parkinson's disease which a wide audience will appreciate. Very recently, a similar genetic interaction between Cisd and Pink1/Prkn in flies was published (Ham et al, 2023 PMID: 37626046) however, from a different angle. While, Ham et al focused on the role of Pink1/Prkn and Cisd in IP3R related ER calcium release, this manuscript approaches the Pink1/Prkn - Cisd interaction from an iron homeostasis point of view. Since, iron dysregulation contributes to the pathogenesis of Parkinson's disease, the observations in this manuscript are relevant for the disease. Hence, the work is sufficiently novel and deserves publication. However, additional experiments are suggested to strengthen the authors' conclusions.

We thank the reviewer for their kind words. As mentioned above, these additional experiments are on their way and will be included in version 2 of our revised manuscript (v2).

I work on Drosophila models of Parkinson's disease

Referees cross-commenting

I agree with the reviewer number 1 that it would be interesting to investigate CISD1 dimerisation status during mitophagy.

As mentioned above, we now studied CISD1 dimerization in response to more classical activators of mitophagy namely FCCP and antimycin/oligomycin which had no significant effect on dimerization suggesting that this phenotype is more pronounced under iron depletion. These data are shown in the new Fig. 2c.

Reviewer #3 (Evidence, reproducibility and clarity):

Here the authors provide evidence that Cisd is downstream of Parkin/Pink1 and suggest that the levels of apo-Cisd correlate with neurotoxicity. The data presented generally supports the conclusions of the authors and will be useful to those in the field. The manuscript would be improved by a more balanced discussion of the strengths and weaknesses of the study and more circumspection in interpretation of data.

We thank the reviewer for their comments aimed to improve our manuscript. We have now discussed the strengths and weaknesses of our study in more detail.

Introduction. While iron has been implicated in Parkinson's disease, it is an overstatement to say that disruption in iron metabolism contributes significantly to the pathogenesis of the disease.

There is certainly a plethora of data implicating perturbed iron homeostasis in PD as also pointed out by reviewer #1. We have tried to tone down our wording in the text and added a recent review on the topic (Ma et al, 2021) as also suggested by reviewer #1.

Introduction. The discussion of the various names for Cisd2 is important, but confusing as written. Specifically, the use of "this" makes the wording unclear.

We thank the reviewer for pointing out this flaw. We have altered the wording in the introduction.

Methods. It would be preferable to use heterozygous driver lines or a more similar genetic control rather than w-1118.

The exact controls were indeed not well explained in the Methods section, this has been corrected in the revised version. In brief, homozygous driver and UAS lines were indeed used in Fig. 4, this will be addressed in the second revision of our manuscript together with the experiments reviewer #2 suggested. The data shown in Fig. 5, 6, and 7 all used w1118 as control because all other fly strains are on the same genetic background.

Page 10. It appears that the PINK1 lines have been described previously. The authors should clarify this point and ensure that the new data presented in the current manuscript (presumably the mRNA levels, Fig. 1a) is indicated, as well as data that is confirmatory of prior findings (Fig. 1b).

Yes, these PINK1 lines have been described previously as pointed out in the manuscript. The original paper did not quantify the PINK1 mRNA levels shown in Fig. 1a. The blots shown in Fig. 1b are from new differentiations and have also not been shown before but confirm findings published in Jarazo et al. (Jarazo et al, 2022). This has been clarified in the revised version of our manuscript.

Fig. 3 legend. There is a typographical error, "ne-way ANOVA."

We thank the reviewer for pointing out this flaw. This has been corrected in the revised version.

Page 15. The nature of the Pink1-B9 mutant should be specified.

We now added a supplemental Figure 1 that depicts the specific mutation in these flies.

Fig. 4. Levels of mutant and wild type Cisd should be compared in transgenic flies.

We now added a quantification of mutant and wildtype Cisd levels to the new Figure 4d.

Fig. 5b,d. The striking change seems to be the decrease in dimers in young Pink1 mutant animals, not the small increase in dimers in the older Pink1 mutants.

It is always difficult to find a “typical” picture that reflects all changes observed in quantitative data. This Figure actually shows a decrease of total Cisd levels in young flies in Fig. 5c but no difference of the dimer/monomer ratio in Fig. 5d.

Fig. 5f. Caution should be used in interpreting the results. Deferiprone has toxicity to wildtype flies (trend) and may simply be making sick Pink1 mutants sicker.

There is certainly a tendency for wildtype flies to thrive less in food containing deferiprone. To make this more obvious, we have now added the exact p value (0.0764, which we don’t consider borderline but a tendency) to this figure and mention this fact in the text.

Fig. 5e. The data are hard to interpret. The number of animals is very small for a viability study and the strains are apparently in different genetic backgrounds, though this is not clearly specified. The experiment in Supplementary Fig. 1 appears better controlled and supports the Pink1 data; however, a similar concern pertains to Fig. 7. The authors may thus wish to be more circumspect in their interpretation, especially of the Parkin data.

In Fig 5e we quantified total iron levels and the Fe3+/Fe2+ ratio using capillary electrophoresis-inductively coupled plasma mass spectrometry (CE-ICP-MS). Although indeed not so many flies were used in this quantification, the results are highly significant. If the reviewer was referring to Fig. 5f, we agree that this experiment was not well (to be honest, even wrongly explained) which we corrected in the revised version of this manuscript. We thank the reviewer for pointing out this flaw.

Reviewer #3 (Significance):

The major significance of the study is in putting downstream of Parkin/Pink1 (largely confirmatory to recently published work) and suggesting that the levels of apo-Cisd are an important determinant of toxicity. The work will be of interest to those in the field.

3. Description of the revisions that have already been incorporated in the transferred manuscript

The changes already carried out and included in the transferred manuscript (v1) are indicated above in bold orange. All changes pending on ongoing experiments to be included in the second revision of the manuscript are indicated above in bold magenta.

4. Description of analyses that authors prefer not to carry out

All changes suggested by the reviewers were addressed (v1) or will be addressed (v2).

References

Jarazo J, Barmpa K, Modamio J, Saraiva C, Sabaté-Soler S, Rosety I, Griesbeck A, Skwirblies F, Zaffaroni G, Smits LM, et al (2022) Parkinson’s Disease Phenotypes in Patient Neuronal Cultures and Brain Organoids Improved by 2-Hydroxypropyl-β-Cyclodextrin Treatment. Mov Disord 37: 80–94

Key J, Sen NE, Arsović A, Krämer S, Hülse R, Khan NN, Meierhofer D, Gispert S, Koepf G & Auburger G (2020) Systematic Surveys of Iron Homeostasis Mechanisms Reveal Ferritin Superfamily and Nucleotide Surveillance Regulation to be Modified by PINK1 Absence. Cells 9

Ma L, Gholam Azad M, Dharmasivam M, Richardson V, Quinn RJ, Feng Y, Pountney DL, Tonissen KF, Mellick GD, Yanatori I, et al (2021) Parkinson’s disease: Alterations in iron and redox biology as a key to unlock therapeutic strategies. Redox Biol 41: 101896

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

1. General Statements [optional]

__We thank all the reviewers for their time and their constructive criticism, based on which we will revise our manuscript. All our responses are indicated in red. __

2. Description of the planned revisions

Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

__Reviewer #1 (Evidence, reproducibility and clarity (Required)): __

The manuscript by Nguyen and Cheng is investigating the timing and mechanism of cessation of neuroblasts in the pupal optic lobe. Previous studies by several groups have determined the spatial and temporal factors required for the neuroepithelial to neuroblast transition and neuroblast to neural/glycogenesis in third instar larvae such that neuroblasts are eliminated. The mechanism of elimination of neuroblasts in the VNC or mushroom bodies have been investigated, but the mechanism(s) and the timing of elimination of medulla neuroblasts has not been investigated. The authors suggest that medulla neuroblasts are eliminated via a combination of mechanisms including apoptosis, prospero induced size symmetric terminal differentiation and a switch to gliogenesis by gcm expression. Expression of Tailless also was found to affect the timing of medulla neuroblast termination. They also ruled out several mechanisms such as ecdysone pulses.

Major comments

Clearly written and logical flow to experiments and results not over interpreted.

Clearly show that the neuroblast number and size decrease (12 to 18 hrs) and are eliminated by 30 hours

Figure 2a Marking of the Neuroepithelium. Would be more convincing if shown by PatJ expression and is clonal analysis. While the following panels use PatJ in clones suggesting are NE and NBs present it is more difficult to put into the context in the higher magnification images (Figure 2 D- M) and the Miranda expression in F' seems to be the entire lobe and it is not clear if would be any NE which does not agree with what is shown in panel A.

We will perform clonal analysis using MARCM to show that the elimination of medulla NBs (marked by Dpn) is accompanied by the depletion of NE (marked by PatJ). For Figure 2 D, E, I, L, we will change the images to the whole lobes to clearly show the shift in the NE-NB transition upon Notch OE/KD.

Is difficult to see the neuroblasts in Figure 2 D D" and E. The figure does not match what is stated in the results in that the neuroblasts are difficult to observe. If the point is that there is fewer NE cells and more neuroblasts then this is hard to see. It has been previously shown that with Notch RNAi clones prematurely extrude form the NE (Egger 20210; Keegan 2023) and could be expressing more Neuroblast markers but this is not visible in the panels as shown. Are the images single focal plane or maximum projections? Imaging more deeply in the brain or viewing in cross section would account for these possibilities. The possibility that are more neuroblasts but not all at the surface of the OL should be addressed as this could also alter the overall results.

Figure 2 is key to first point of the paper so needs to be addressed.

The images are single focal plane of the superficial layer of the medulla. We will specify this information in the figure legends. We will include cross-section of the notch RNAi clones to show the delamination of precocious NBs.

Minor comments

Why express volume of DPN in clone volume. Would make the point more clear and more strong be to express as number of NB in the 3-D volume of the clone. This measurement occurs in several figures.

We will redo the quantification as suggested.

Use of Miranda to mark NBs is unclear in Figure 2. Perhaps more clear in B&W.

We will redo the staining with Dpn instead of Mira to mark medulla NBs. Figures will be presented in B&W as suggested.

Make clear in figures (or figure legend) if single focal plane or projections.

We will do so.

It is unclear what percentage of NB the Gal4 line eyR16F10 are expressed in. Veen 2023 state that the GAL4 is also expressed in neurons and at different levels whether deeper within the brain or superficially on the surface of the brain. At 16 APF it is expressed but it is not clear whether it is in all cells at a low level or only within a few cells

We will further characterize the expression of eyR16F10-GAL4 in the pupal medulla as suggested.

Some RNAi lines referenced as previously validated and other are not. For example: EcR, Oxphos, Med27, Notch need references or confirmation of specificity to the intended target (qRT)

We will perform RT-qPCR to validate the use of UAS-med27 RNAi. For RNAi stocks such as UAS-EcR RNAi, UAS-Atg1 RNAi, UAS-notch RNAi that have been previously used in other publications, we will provide appropriate references.

At least 2 animals per genotype were used. While I appreciate the technical difficulty of working in pupae this seems a bit low in terms of number of samples and data would be more robust with more numbers.

Any experiments in which less than 3 animals were used, we will redo the experiments.

Reviewer #1 (Significance (Required)):

This provides mechanism and timing for the elimination of neuroblasts (NE to NB) that arise from the medulla. As these are most similar to mammalian brain development (Radial glial to NSC) this information provides more context to interpret the formation of glial and neurons in the adult optic lobe given the effect on timing and mechanisms of elimination.

This paper would be of interest to developmental biologist who work with Drosophila or mice who are looking at neural development. An understanding of how neural diversity is achieved and the mechanisms behind this that can be dysfunctional in terms of etiology of neural diseases. Is a well done study for the most part that would be improved by clarifying some data and provided more replicates for robustness of the data.

I am a developmental biologist working with Drosophila in larval and adult neural development.

__Reviewer #2 (Evidence, reproducibility and clarity (Required)): __

Lineages of neural stem cells are of great interest to understand how many neural types are generated. They produce very diverse neurons, often in a highly stereotyped series. However, they must terminate their life when the animal becomes functional or if neurons need time to become mature before birth.

In the Drosophila optic lobes, neural stem cells are produced over a period of several days by a wave of neurogenesis that transforms a neuroepithelium into neural stem cells that undergo a series of temporal patterning steps. It has been reported that they finish their life when a symmetric division generates glial cells. The authors however analyze the end of a particular lineage, that of the latest born neural stem cells of the medulla.

The paper shows that neural stem cells stop being produced when the neuroepithelium is consumed. But how do the latest born neural stem cells stop their lineage?

The results show that they do so by several means, which is quite unexpected: they may die from apoptosis, or autophagy, by becoming glioblasts or by a terminal symmetric division.

There are no major issues affecting the conclusions

• The paper shows that the end of production of neural stem cells occurs the neuroepithelium is completely transformed. The experiments performed by the authors are fine and show that, if the transition is delayed, neural stem cells terminate their life later, and vice versa. However, the lifespan of the neural stem cells is not affected by the timing of the transition. Therefore, these experiments do not tell us how neural stem cells terminate their life, which is the central question of the study. The discussion should be written accordingly and the title and the model in Fig 6 modified to reflect the importance of the end of life of the stem cells, the main theme of the paper.

We agree that our said experiments did not elucidate how NBs terminate at the end of neurogenesis. Nevertheless, our aim is to show that the timing of NB termination in the medulla is dependent on the timing of the NE-NB transition.

In Supplementary Figure 1, we showed that factors previously shown to be involved in NB termination in other lineages did not play similar roles in the medulla NBs. Thus, we think that NB termination in the medulla is likely regulated at the levels of the NE, but not the NBs themselves. Although we have briefly mentioned this in our manuscript, we hope by conducting the experiments suggested by the reviewer (see below), we can subsequently modify our model in Figure 6 and our discussion.

• The authors talk about Pros-dependent symmetric division and gliogenic switch as two separate processes, but these may be two sides of the same phenomenon. Tll+ gcm+ neural stem cells undergo Pros-dependent cell cycle exit, generating glial progeny. If the authors agree with this, could they update their model (and discussion) to reflect the fact that gliogenic switch occurs via a Pros-dependent symmetric division, and these are not two separate processes independently contributing to the depletion of the neural stem cell pool? Ideally, a triple staining between Dpn, Pros, and gcm would show that the symmetrically dividing cells seen by the authors are committed to the glial fate.

We will further test how gliogenesis is affected in pros RNAi clones. The results may shed light on whether Pros-mediated symmetric division is required for Gcm-mediated gliogenesis in the medulla. Regarding the model, we have summarized our findings and suggestions in Figure 5K, however, we will integrate this information into our final model.

In Figure 5C, we showed that at 12h APF, there are Dpn+ NBs in the medulla that expressed both Pros and Gcm, suggesting that it is very likely that Pros is upstream of Gcm to induce the glial cell fate switch of the medulla NBs.

• Why were Notch RNAi experiments assessed for the presence of neural stem cells at P12 and gcm RNAi experiments at P24? Given that most optic lobe neural stem cells disappear between P12-18, a subtle effect of gcm RNAi may have been missed. Do the authors have data for gcm RNAi at P12?

We hypothesized that the timing of NE-NB transition affects the timing of NB termination in the medulla. Because Notch KD was previously shown to induce precocious NE-NB transition in the OL, meaning that medulla NBs are born prematurely, we expected that this manipulation will lead to a corresponding premature elimination of the NBs. In contrast, gcm RNAi which inhibits the switch into the glial cell fate of the NBs, is expected to prolong the neurogenic phase of the NBs, and thereby, their persistence by 24h APF when WT NBs are eliminated.

• The authors should acknowledge that the inhibition of either apoptosis or autophagy alone may not be fully sufficient to prevent the death of NBs. In mushroom body neural stem cells, both processes must be inhibited simultaneously to produce a strong effect on their survival (Pahl et al. 2019, PMID 30773368).

We will add this information in our discussions.

• There is an important missing point that should be addressed: is there a specific point in time when all neural stem cells must stop their lineage wherever they are in the temporal series and either die or divide symmetrically? One possibility that is not discussed is that most neural stem cells end their life through a gliogenic symmetric division while those that were generated late must stop en route and die by apoptosis and/or autophagy. This would solve the strange diversity of end-of-life, which could be easily addressed by identifying the temporal stage of the neural stem cells that undergo apoptosis

We agree that it would be of interest to understand how there are diverse mechanisms by which medulla NBs terminate during pupal development. To address if temporal progression is involved in apoptosis of the medulla NBs, we will first characterize the expression of some temporal TFs (e.g., Ey, Slp, Tll) at 12h APF when we found a subset of medulla NBs undergo apoptosis in the wildtype animals.

Minor suggestions:

We agree with these minor modifications.

• Line 46: Specify that there are 8 type II neural stem cells in each hemisphere*.

• The statement in lines 181-182 that "cell death, and not autophagy, makes a minor contribution to..." should be replaced with "apoptosis, and not autophagy," as autophagy is also a type of cell death.

• The authors should adjust the logic of the section "Medulla neuroblasts terminate during early pupal development": Describe the wild-type pattern first (the decrease in the number of neural stem cells and their size with age) and then describe the perturbations aimed at disrupting the number and the size of neural stem cells

• Line 151 should refer to Fig. 2I-K, not Fig. 2J-K.

**Referees cross-commenting**

How can NBs die by different mechanisms?? This might only happen is they are in a different states, an issue that is not addressed.

it has been shown that optic lobe NBs end their life by a symmetric, gliogenic last division at the end of the last temporal window, and not by PCD.

It is likely, and the authors do hint at it, that NBs only die by PCD when they prematurely interrupt the temporal series in early pupation when neurons synchronously start undergoing maturation.

I believe that the authors should explain this, if this is indeed their model, and show that NBs die while still in early temporal windows.

Reviewer #2 (Significance (Required)):

Lineages of neural stem cells are of great interest to understand how many neural types are generated. They produce very diverse neurons, often in a highly stereotyped series. However, they must terminate their life when the animal becomes functional or if neurons need time to become mature before birth.

In the Drosophila optic lobes, neural stem cells are produced over a period of several days by a wave of neurogenesis that transforms a neuroepithelium into neural stem cells that undergo a series of temporal patterning steps. It has been reported that they finish their life when a symmetric division generates glial cells. The authors however analyze the end of a particular lineage, that of the latest born neural stem cells of the medulla.

The paper shows that neural stem cells stop being produced when the neuroepithelium is consumed. But how do the latest born neural stem cells stop their lineage?

The results show that they do so by several means, which is quite unexpected: they may die from apoptosis, or autophagy, by becoming glioblasts or by a terminal symmetric division.

There are no major issues affecting the conclusions

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

Summary

In this manuscript, the authors address the timing and mechanisms responsible for the termination of medulla neuroblasts in Drosophila visual processing centres, also known as optic lobes. Through time course experiments the authors demonstrate the medulla NBs are completely eliminated by 30h APF during early pupal development. By manipulating the Notch signalling pathway as well as proneural genes such as lethal of scute, the authors show that altering the NE-NB transition is sufficient to change the timing of NB termination. In contrast, ecdysone signalling and components of the mediator complex, known to terminate proliferation of central brain NBs, are not required for the termination of medulla NBs. Medulla NBs sequentially express a variety of temporal transcription factors to promote cellular diversity, however, the authors demonstrate that altering temporal factors such as Ey, Sco or Hth, does not affect the timing of the medulla NBs termination. Interestingly however overexpression of the transcription factor tailless can cease medulla NB termination via the conversion of type I to type II NB fate. They further go on to show the importance of the differentiation factor, Prospero, in promoting the differentiation of medulla NBs as well as terminating medulla neurogenesis during pupal development. Finally, in addition to differentiation, the authors show another mechanism responsible for the cessation of neurogenesis which is the commencement of gliogenesis. Through manipulation of the neurogenic to gliogenic switch by knockdown or overexpressing the glial regulatory gene, gcm, the authors show that even though the downregulation of gcm is is not sufficient to induce NB persistence, gcm overexpression can cause premature termination of NBs.

Major comments:

• Are the key conclusions convincing?

Yes, the key conclusions are convincing with proper controls, quantifications and statistical analyses.

• Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

The conclusion that temporal transcription factors (TTF) do not affect the timing of medulla NB termination is somewhat preliminary. The authors investigated a simplified temporal series including Homothorax, Eyeless, Sloppy-paired, Dichaete and Tailless. However, there are additional temporal factors that have not been examined for their potential involvement in medullar NB termination. Previous reports have identified several other temporal factors that play a role in medulla TTF cascade, such as, SoxNeuro (SoxN) and doublesex-Mab related 99B (Dmrt99B) that start their expression in the NE similar to Hth, however, Dmrt99B is likely to be repressed much later than Hth (Li, Erclik et al. 2013, Zhu, Zhao et al. 2022). At this point, it remains challenging to completely rule out the possibility that other temporal factors play a role in medullar NB termination or have redundant functions in regulating the timing of medulla NB cessation. It is suggested to tone down this claim and provide a brief discussion on alternative possibilities, citing relevant papers on the functions of other temporal factors in medullar NBs.

We agree.

• Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

Loss of pros by RNAi caused the formation of ectopic NBs and the NBs persist even at 24h APF. Do these NBs persist at 30h or 48h APF? Does overexpression of Pros result in early termination of medulla NBs?

We will do these experiments in clones as suggested.

• Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

Yes, I believe the suggested experiments are realistic in terms of time and resources, with an estimation of 3 months to complete the experiments.

• Are the data and the methods presented in such a way that they can be reproduced?

Yes.

• Are the experiments adequately replicated and statistical analysis adequate?

The experiments are straight forward and were performed with proper controls, supported by quantifications and proper statistical analyses. However, there is no mention about how many replicates were used.

We will add this information in our Material and Methods section.

Minor comments:

1. The authors use the eyR6F10-Gal4 driver in certain experiments. The eyR6F10-Gal4 driver is however expressed only in a subset of medulla NBs. Can the authors comment on what percentage of medulla NBs is the driver expressed in? We will characterize this.

Does the EGFR signalling pathway or JAK/STAT pathway affect the timing of termination of medulla NBs? Experiments are not necessary. The author can speculate on their roles.

We will modify our discussion accordingly.

Figure 1C has a p value of only 0.03 (*) but shows a strong reduction in the number of Dpn+ cells from 12h to 18h, etc. Is this correct? Also, is the p value the same for the comparison between 12h and 24h as well as 12h and 30h APF?

Yes. P-values showed no significant differences between 28-24h and 24-30h APF.

The controls in figure 2B and to some extent figure 2H show one major outlier (much higher than the other brain lobes in the control). Will the removal of this outlier affect the significance/ p-value of the experiment?

No, removing the outliers do not change the statical results.

In figure 2B what is the p-value between 12h and 18h APF? Is it *** as well?

No, it’s not significant.

Line 84 of the introduction introduces Tll, Gcm and Pros for the first time in the manuscript and should be written out in full.

We will change this.

• Are prior studies referenced appropriately?

Yes.

• Are the text and figures clear and accurate?

Yes.

• Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

Quite a few of data mentioned in the manuscript have been described as data not shown. I think it would be nice to show quantifications or representative images in the supplementary figures.

We will add the data which was previously not shown.

Reviewer #3 (Significance (Required)):

Since the mechanisms by which medulla NBs are terminated are currently unknow, this is an important and interesting study to understand how medulla neuroblasts in the optic lobe are terminated. The balance between stem cell maintenance and differentiation is critical for proper brain development and the results presented in this paper are impactful. Furthermore, Drosophila melanogaster is an excellent model to study stem cell niches and neuroblast temporal patterning. The authors provide key mechanisms namely cell death, Pros-mediated differentiation and the gliogenic switch that contribute to a better understanding of how the NB progenitor pool can be terminated in the Drosophila OL, which is largely supported by the data.

• Place the work in the context of the existing literature (provide references, where appropriate).

So far, most work in this field has focused on the regulation of the temporal factors to promote the progression of the TTF transcriptional cascade and thereby diversity of the neural progenitors (Li, Erclik et al. 2013, Naidu, Zhang et al. 2020, Ray and Li 2022, Zhu, Zhao et al. 2022). Furthermore, work on pathways such as EGFR and Notch signalling that allows the proneural wave to progress and subsequently induce neuroblast formation in a precise and orderly manner have also been studied (Yasugi, Umetsu et al. 2008, Yasugi, Sugie et al. 2010). Here, considering previous literature, the authors move one step forward to determine how and when these neuroblast progenitors cease proliferation during development thus providing mechanisms for the regulation of the neuroepithelial stem cell pool, its timely conversion into NSCs and the switch from neurogenesis to gliogenesis thus providing important implications for brain size determination and function.

• State what audience might be interested in and influenced by the reported findings.

Stem cell research, neurobiologists and developmental biologists.

• Define your field of expertise

Stem cells, developmental biology

Although thee are so many different styles of writing, especially on the internet, I think we can all agree in one way or another that knowing how to write professionally, whether it may be for a professional media outlet or even for a company website, is very important. Consider journalism. If a journalist working for the New York Times were to publish a story that was written in the style of something that came from Reddit, they would be fired immediately. Therefore, knowing how to differentiate when to use different writing styles is extremely important as a writer.

Author Response

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

eLife assessment:

This valuable study, of interest for students of the biology of genomes, uses simulations in combination with published data to examine how many TADs remain after cohesin depletion. The authors suggest that a significant subset of chromosome conformations do not require cohesin, and that knowledge of specific epigenetic states can be used to identify regions of the genome that still interact in the absence of cohesin. The theoretical approaches and quantitative analysis are state-of-the-art, and the data quality and strength of the conclusions are solid. However, because "physical boundaries (of domains?)" in the model appear to be a consequence of preserved TADs, rather than the other way around, the functional insights are limited.

Summary of the reviewer discussion for the authors:

While the simulations are state of the art and the reviewers appreciated that the approaches used here might help to resolve apparent discrepancies between conclusions from single-cell and bulk/ensemble techniques to study chromosome conformation, the work would benefit from clarification of what precisely is meant with "physical boundaries" and from a comparison of CCM and HIPPS models to understand commonalities and differences between them. In addition, more discussion of the relation of the current work to previous studies, such as Schwarzer et al., 2017, and Nuebler et al., 2018, would elevate the work and make the key claims more compelling. Please see also the detailed comments from the expert reviewers.

We thank the editor for the assessment and the reviewers for the incisive comments. We will address these comments one by one. In particular, we attempt to clarify the concept of “physical boundaries” and its relevance in our study. We hope our responses are satisfactory. We believe that our manuscript has benefitted substantially by revising the manuscript following the comments by the reviewers.

Below is our point-by-point response to the comments:

Reviewer #1 (Public Review):

Summary:

In this paper, Jeong et al. investigate the prevalence and cause of TADs that are preserved in eukaryotic cells after cohesin depletion. The authors perform an extensive analysis of previously published Hi-C data, and find that roughly 15% of TADs are preserved in both mouse liver cells and in HCT-116 cells. They confirm previous findings that epigenetic mismatches across the boundaries of TADs can cause TAD preservation. However, the authors also find that not all preserved TADs can be explained this way. Jeong et al. provide an argument based on polymer simulations that "physical boundaries" in 3D structures provide an additional mechanism that can lead to TAD preservation. However, in its current form, we do not find the argumentation and evidence that leads to this claim to be fully compelling.

Strengths:

We appreciate the extensive statistical analysis performed by the authors on the extent to which TAD's are preserved; this does seem like a novel and valuable contribution to the field.

We thank the reviewer for a succinct and comprehensive summary of our work and for appreciating value of our work.

Weaknesses:

1) As the authors briefly note, the fact that compartmentalization due to epigenetic mismatches can cause TAD-like structures upon cohesin depletion has already been discussed in the literature; see for example Extended Data Figure 8 in (Schwarzer et al., 2017) or the simulation study (Nuebler et al., 2018). We are hence left with the impression that the novelty of this finding is somewhat overstated in this manuscript.

It is unclear to us by studying the results in the Extended Data Figure 8 that the authors have shown that epigenetic mismatches cause TAD-like structures. As far as we can discern, the data, without a quantitative analysis, shows that may be new TAD-like structures that are not in the wild type appear when cohesin is deleted.

The studies by Schwarzer et al 2017 and Nuebler et al 2018 are relevant to our own investigation, which we undertook after scrutinizing the experiments in Schwarzer et al 2017 and the related work by Rao et. al in 2017 on a different cell line. In the summary of the Reviewer discussion, it is suggested we discuss the relation to the experimental study by Schwarzer et al 2017 and the computational work by Nuebler et al 2018.

(1) The results and the corresponding discussion in these two studies are different (may be complimentary) from our results. When referring to the Extended Data Figure 8 Schwarzer and co-authors state in the main text, “The finer compartmentalization explains most of the remaining or new domains and boundaries seen in Nipbl Hi-C maps”. We are not 100% sure what “remaining” means in this context. The Extended Data Fig. 8(a) shows the “common boundaries” is correlated with the eigenvectors of compartmentalization. If this indeed is what the reviewer is referring to, we believe that our study differs from theirs in two important ways: First, Extended Data Fig.8 (a) is a statistical analysis at the “ensemble” level. In our study, we examined the preservation of TADs at both individual and ensemble level with more detailed analysis. Second, in Extended Data Fig. 8(a), the “common boundaries” (incidentally we are uncertain how that was calculated) are compared to the eigenvectors of PCA analysis of the compartments (larger length scales). In contrast, in our study, we examined the correlation between TAD boundaries and the epigenetic profiles. We believe that this is an important distinction. The PCA analysis of compartments and “common boundaries” defined using (presumably) the insulation score are both derived from the Hi-C contact map. Epigenetic profile, on the other hand, is independent of Hi-C data. We believe our contribution, is to build the connection between epigenetic profiles with the preservation of TADs, and link it to 3D structures. For these reasons, we assert that our results are novel, and are not contained (or even implied) in the Schwarzer et al 2017 study.

The simulations in Neubler et al 2018, which were undertaken to rationalize the experimenrs, revealed that compartmentalization of small segments is enhanced after cohesin depletion, while TADs disappear, which support the broad claims that are made in the experiments. They assert that the structures generated are non-equilibrium. They do not address the emergence of preserved nor the observation of TAD-like structures at the single cell level. However, our goal was to elucidate the reasons for of preservation of TADs. By that we mean, the reasons why certain TADs are present in both the wild and cohesin depleted cells? Through a detailed analyses of two cells, polymer simulations, we have provided a structural basis to answer the question. Finally, we have provided a plausible between TAD preservation and maintenance of enhancer-promoter interactions by analyzing the Micro-C data. For all these reasons, we believe that our study is different from the results in the Extended Figure 8 or the simulations described by Neubler.

Let us summarize the new results in our study that are not contained in the studies referred to by this Reviewer. (1) We showed by analyzing the Hi-C data for mouse liver and HCT-16 that a non-negligible fraction of TAPs is preserved, which set in motion our detailed investigation. (2) Then, using polymer simulations on a different cell type, we generated quantitative insights (epigenetic mismatches as well as structural basis) for the preservation of TADs. Although not emphasized, we showed that deletion of cohesin in the GM12878 cells also give rise to P-TADs a prediction that suggests that the observations might be “universal”. (3) Rather than perform, time consuming polymer simulations, we calculated 3D structures directly from Hi-C data for the mouse liver and HCT-16 cells, which provided a structural basis for TAP preservation. (4) The 3D structures also showed how TAD-like features appear at the single cell level, which is in accord with imaging experiments. (5) Finally, we suggest that P-TADs may be linked to the maintenance of enhancer-promoter and promoter-promoter interactions by calculating the 3D structures using the recent Micro-C data.

For the reasons given above, we assert that our results are novel, and bring new perspectives that are not in the aforementioned insightful studies cited by the Reviewer.

2) It is not quite clear what the authors conceptually mean by "physical boundaries" and how this could offer additional insight into preserved TADs. First, the authors use the CCM model to show that TAD boundaries correlate with peaks in the single cell boundary probability distribution of the model. This finding is consistent with previous reports that TAD-like structures are present in single cells, and that specific TAD boundaries only arise as a population average.

The finding based on the CCM simulations hence seems to be that preserved TADs also arise as a population average in cohesin-depleted cells, but we do not follow what the term "physical boundaries" refers to in this context. The authors then use the Hi-C data to infer a maximumentropy-based HIPPS model. They find that preserved TADs often have boundaries that correspond to peaks in the single cell boundary probabilities of the inferred model. The authors seem to imply that these peaks in the boundary probability correspond to "physical boundaries" that cause the preservation of TADs. This argument seems circular; the model is based on inferring interaction strengths between monomers, such that the model recreates the input Hi-C map. This means that the ensemble average of the model should have a TAD boundary where one is present in the input Hi-C data. A TAD boundary in the Hi-C data would then seem to imply a peak in the model's single-cell boundary probability. (The authors do display two examples where this is not the case in Fig.3h, but looking at these cases by eye, they do not seem to correspond to strong TAD boundaries.) "Physical boundaries" in the model are hence a consequence of the preserved TADs, rather than the other way around, as the authors seem to suggest. At the very least the boundary probability in the HIPPS model is not an independent statistic from the Hi-C map (on which their model is constrained), so we have concerns about using the physical boundaries idea to understand where some of the preserved TADs come from.

There are many statements in this long comment that require us to unpack separately. First, using both the CCM simulations, and even more importantly using data-driven approach, we established that TAD-like structures are present in single cells with and without cohesin. The latter finding is fully consistent with imaging experiments. We are unaware of other computational efforts, before our work, demonstrating that this is the case. Perhaps, the Reviewer can point to the papers in the literature.

Regarding the statement that our arguments are circular, and lack of clarity of the meaning of physical boundary, please allow us to explain. First, we apologize for the confusion. Let us clarify our approach. We first used CCM to understand the potential origin of substantial fraction of P-TADs in the GM. The simulations, allowed us to generate the plausible mechanisms, for the origin of P-TADs. Because the CCM does reproduce the Hi-C data, we surmised that the general mechanisms inferred from these simulations could be profitably used to analyze the experiments. The simulations also showed that knowledge of 3D structures produces a muchneeded structural basis of P-TADs , and potentially for emergence of new TADs upon cohesin depletion.

Because 3D coordinates are needed to obtain structural insights into the role of cohesin, we use a novel method to obtain them without the need for simulations. In particular, we used the HIPPS method to obtain 3D coordinates with the Hi-C data as the sole input, which allowed us to calculate directly the boundary probabilities. The excellent agreement between the predicted 3D structures and imaging experiments suggests that meaningful information, not available in Hi-C, may be gleaned from the ensemble of calculated 3D structures.

Although “physical boundary”, a notion introduced by Zhuang, is defined in in the method section, it is apparently unclear for which we apologize. Because this is an important technical tool, we have added a summary in the main text in the revision. We did not mean to imply that the physical boundaries cause the preservation of TADs, although we found that maintenance of the enhancer-promoter contacts (see Fig. 8 in the revision) could be one of the potential reasons for the emergence of physical boundaries. We agree with the reviewer that physical boundaries are structural evidence of preserved TADs (not the cause), that is when a TAD is preserved, we can detect it by prominent physical boundary. The purpose and benefit of physical boundary analysis and using HIPPS in general is to obtain three-dimensional structures of chromosomes. Although both CCM simulations and HIPPS use Hi-C contact maps, three-dimensional structures provide additional information that is not present in the Hi-C data.

The arguments that the authors use to justify their claims could be clarified and strengthened. Here are some suggestions: -Explain the concept of "physical boundaries" more clearly in the main text.

As explained above, we have revised the text to clarify the concept and purpose of physical boundaries analysis. See Page 7.

• Justify why the boundary probabilities and the physical boundaries concept can be used to offer novel insight into where preserved TADs may come from.

Boundary probabilities and physical boundaries provide previously unavailable 3D structural information on the TADs structures both at the single-cell and population level. This provides a direct structural basis for determining which TADs are preserved. But in order to understand where P-TADs may come from, physical boundaries analysis alone is not sufficient. As we have shown in the analysis of enhancer-promoter contact, using physical boundary analysis from 3D structures, we can conclude that conservation of enhancer-promoter contact could be one of the reasons for the P-TAD.

• Explain more clearly what the additional value of using the HIPPS model to study TAD preservation is.

Our goal, as announced in the title is to elucidate the structural basis for the emergence of PTADs. The HIPPS method, which avoids doing simulations (like CCM and other polymer models used in the literature) provides an ensemble of 3D conformations using averaged contact map generated in Hi-C experiments. Even more importantly, HIPPS produce an ensemble of structures, which can be the basis for predicting the outcomes at the single-cell level. The accuracy of the generated structures has been shown in our previous work (Shi and ThirumalaiPRX 2021). In ensemble-averaged Hi-C experiments, TADs appear to be relatively stable. However, imaging experiments (Bintu et. al, 2018) have revealed that TADs are not fixed structures present in every single cell, but instead exhibit variability at the single-cell level. TADlike structures with distinct boundaries are observed in individual cells, and the location of these boundaries varies from cell to cell. However, these TAD-like structures still show a preferential positioning in 3D structures. Interestingly, the preferential positioning often corresponds to TAD boundaries observed in population-averaged Hi-C data. This suggests that while cohesin is involved in establishing the overall organization of TADs, other factors and mechanisms could also contribute to TAD formation at the individual cell level. In this study, we showed some boundaries of P-TADs upon cohesin loss in the Hi-C maps, align with preferential boundaries in individual 3D structures of chromosomes. The makes the finding that a subset of TADs is preserved upon cohesin is robust.

From a technical perspective, the use of HIPPS avoids time-consuming polymer simulations. The HIPPS is rapid and can be used to generate arbitrarily large ensemble of structures, allowing us calculate properties both at the single cell and ensemble level.

In addition, we'd like to offer the following feedback to the authors.

3) The discussion of enhancer-promoter loops as a cause of TAD preservation is interesting, but it would be interesting to know fraction of preserved TADs enhancer-promoter loops might explain.

We thank the reviewer for the excellent suggestion. We have done the suggested calculation. The results are shown in a new Figure.8 in the main text. We also moved the results on enhancer-promoter to the main results section from the Discussion section.

4) The last paragraph of the introduction seems to state that only the HIPPS model was used to find single-cell 3D structures and boundary probabilities. However, the main text suggests that the CCM model was also used for these purposes.

We have revised the text to clarify this point on pages 3-4. Also please see the discussion on the utility of HIPPS above.

5) When referring to the boundary probability, it would be useful if the authors always specified whether they refer to the boundary probability before or after cohesin depletion (or loop depletion in the CCM model). Statements such as "This implies that peaks in the boundary probabilities should correspond to P-TADs" are ambiguous; it is unclear if the authors mean that boundary probabilities before cohesin depletion predict that the boundary will be preserved, rather than that preserved TAD boundaries correlate with peaks in the boundary probability after cohesin depletion.

We thank the reviewer for the suggestion. Indeed, it may be confusing. Hence, we have revised the text in numerous places to clarify this point.

6) It would be interesting to analyze all TAD boundaries that are present after cohesin depletion, rather than just those that overlap with TAD boundaries in WT cells. This would give better statistics for answering the question what causes TAD-like structures in cells without cohesin.

We thank the reviewer for this excellent suggestion. First, this would we believe this deviate from the primary goal of this study: what leads to TAD preservation after cohesin deletion? Second, this has to be done very systematically, as we did here for P-TADs, in order draw meaningful conclusions. This is a very useful study for another occasion.

7) The use of a plethora of acronyms (P-TAD, CM, DM, CCM, HLM...) makes the paper difficult to read.

We have revised the text to change CM to “contact map” and “DM” to “distance map”. For PTADs, CCM, and WLM, we would argue that P-TAD is rather a clear and intuitive abbreviation and CCM/WLM refers to specific methods/models and replacing them with full names would make text more difficult to read. We hope the reviewer is okay with us keeping these acronyms.

Reviewer #2 (Public Review):

Summary:

Here Jeong et al., use a combination of theoretical and experimental approaches to define molecular contexts that support specific chromatin conformations. They seek to define features that are associated with TADs that are retained after cohesin depletion (the authors refer to these TADs as P-TADs). They were motivated by differences between single cell data, which suggest that some TADs can be maintained in the absence of cohesin, whereas ensemble HiC data suggest complete loss of TADs. By reananalyzing a number of HiC datasets from different cell types, the authors observe that in ensemble methods, a significant subset of TADs are retained. They observe that P-TADs are associated with mismatches in epigenetic state across TAD boundaries. They further observe that "physical boundaries" are associated with P-TAD maintenance. Their structure/simulation based approach appears to be a powerful means to generate 3D structures from ensemble HiC data, and provide chromosome conformations that mimic the data from single-cell based experiments. Their results also challenge current dogma in the field about epigenetic state being more related to compartment formation rather than TAD boundaries. Their analysis is particularly important because limited amounts of imaging data are presently available for defining chromosome structure at the single-molecule level, however, vast amounts of HiC and ChIP-seq data are available. By using HiC data to generate high quality simulated structural data, they overcome this limitation. Overall, this manuscript is important for understanding chromosome organization, particularly for contacts that do not require cohesin for their maintenance, and for understanding how different levels of chromosome organization may be interconnected. I cannot comment on the validity of the provided simulation methods and hope that another reviewer is qualified to do this.

We appreciate the reviewer for a comprehensive summary of our work, and we are happy that the reviewer finds our work important, which provides valuable insights to the field.

Specific comments

• It is unclear what defines a physical barrier. From reading the text and the methods, it is not entirely clear to me how the authors have designated sites of physical barriers. It may help to define this on pg 7, second to last paragraph, when the authors first describe instances of PTAD maintenance in the absence of epigenetic mismatch.

We thank the reviewer for the suggestions. The details of physical boundary designation are provided in the appendix data analysis. To make the concept and idea of physical boundary easy to understand, we have revised the text on page 7 in the revised main text.

• Figure 7 adds an interesting take to their approach. Here the authors use microC data to analyze promoter-enhancer/promoter-promoter contacts. These data are included as part of the discussion. I think this data could be incorporated into the main text, particularly because it provides a biological context where P-TADs would have a rather critical role.

We thank the reviewers for the suggestion. We also agree that results in Figure 7 provide novel insights on TAD formation and its possible preservation upon perturbation. We have followed the reviewer’s suggestion to move it to an independent section in the main results section as the last subsection.

• Figure 3a- the numbers here do not match the text (page 6, second to last paragraph). The numbers have been flipped for either chromosome 10 or chromosome 13 in the text or the figures.

We thank the reviewer for pointing out this error. In the revised main text, it has been corrected.

Reviewer #3 (Public Review):

This manuscript presents a comprehensive investigation into the mechanisms that explain the presence of TADs (P-TADs) in cells where cohesin has been removed. In particular, to study TADs in wildtype and cohesin depleted cells, the authors use a combination of polymer simulations to predict whole chromosome structures de novo and from Hi-C data. Interestingly, they find that those TADs that survive cohesin removal contain a switch in epigenetic marks (from compartment A to B or B to A) at the boundary. Additionally, they find that the P-TADs are needed to retain enhancer-promoter and promoter-promoter interactions.

Overall, the study is well-executed, and the evidence found provides interesting insights into genome folding and interpretations of conflicting results on the role of cohesin on TAD formation.

We are pleased with the reviewer’s positive assessment of our work.

To strengthen their claims, the authors should compare their de-novo prediction approach to their data-driven predictions at the single cell level.

We thank the reviewer for the very good suggestion. We are assuming that the Reviewer is asking us to compare the CCM simulations with HIPPS generated structures at the single cell level. We have shown, using the GM12878 cell data, that the polymer simulations reproduce the Hi-C contact maps (an average quantity) well (see Appendix Fig. 2 and Fig. 3). In addition, we show in Appendix Fig. 8 the comparison with ensemble averaged distance maps as well as at the single cell level for Chr 13 from the GM12878 cell. There are TAD-like structures at the single cell level just as we find for HCT-116 cell (Fig. 5 in the main text). Thus, the conclusions from de-novo prediction and data-driven predictions are consistent. In addition, in our previous publication introducing HIPPS in Phys Rev X 11: 011051 (2021), we showed that the method is quantitatively accurate in reproducing experimental data for all the interphase chromosomes.

Having demonstrated this consistency, we used computationally simple data-driven predictions to analyze HCT-116 and mouse liver cell lines for which Hi-C data with and without cohesin rather than perform multiple laborious polymer simulations.

Please see below for our response to specific comments.

1) It is confusing that the authors change continuously their label for describing B-A and A-B switches. They should choose one expression. I think that the label "switch" between A and B is more precise than "mismatch".

We have revised the text to make it consistent. Now it all reads “A-B”. Yes, the suggestion that we use switch is good but we think that mismatch is more concise. We trust that this Reviewer will indulge us on this point.

2) In the Abstract, the authors mention HCT-116 cells but do not specify which cells are these.

We have changed “HCT-116” in the abstract to “human colorectal carcinoma cell line”.

3) In the Abstract, it is unclear what the authors mean by "without any parameters"

In the theoretically based HIPPS method, there is no “free” parameter. In other words, the only parameter is uniquely determined. To avoid confusion, we have removed “without any parameters” from abstract.

4) In Results, what do the authors mean by 16% (26%)?

This refers the percentage of how many TADs are preserved after Nipbl and RAD21 removal in mouse and HCT-116 cells, respectively. Using TopDom method, we identified TAD boundaries in Wild and cohesin-depleted cells. There are 16% (959 out of 4176 – Fig. 1a) and 26% (1266 out of 4733 – Fig. 1b) of TADs are preserved after Nipbl and RAD21 removal in mouse and HCT-116 cells, respectively. We removed the percentages in the revised version.

5) In Results, the authors mention "more importantly, we did tune the value of any parameter to fit the experimental CMs". Did they mean that instead they didn't tune any parameter?

We apologize for the confusion. In the CCM, there is a single controlled parameter. We have changed the sentence to reflect this correctly.

6) In Results, section "CCM simulations reproduce wild-type Hi-C maps", Kullback-Leibler (KL) divergence is used to assess the correlation between two loci, but it is unclear what the value 0.04 stands for; is it a good or a bad correlation?

The value for Kullback-Leibler divergence can vary from 0 to infinity with 0 give the perfect correlation. Thus, 0.04 means that the correlation is excellent.

7) The authors use two techniques to obtain 3D structures, one is CCM, which takes the cohesin as constraints, and another is HIPPS, which reconstructs from Hi-C maps. Both seem to have good agreement with the Hi-C contact maps. However, did the authors compare the CCM with the HIPPS 3D structures?

This is detailed in response at the start of the reply to this Reviewer. As detailed in this response as well in the main text we used the CCM to generate hypotheses for the origin of P-TADs. In the process, we established the accuracy of CCM, which gives us confidence about the hypotheses. As explained above and emphasized in the revised version, CCM simulations are time consuming whereas generating 3D structures using HIPPS is computationally simple. Because HIPPS is also accurate, we used it to analyze the Hi-C data on mouse liver, HCT-116 as well as Micro-Data on mESC.

In our paper in Phys Rev X 11: 011051 (2021) we showed that HIPPS reproduces Hi-C data. In the current manuscript, we showed in Appendix Fig. 2 and Fig. 3 as well as in a study in 2018 (Shi and Thirumalai, Nat Comm.) that CCM is accurate as well. Thus, there is little doubt about the accuracies of the methods that we have developed.

8) In Results, section "P-TADs have prominent spatial domain boundaries", the authors constructed individual spatial distance matrices (DMs) using 10,000 simulated 3D structures. What are the differences among these 10,000 simulations? Do they start them with different initial structures?

The structures are generated using HIPPS which is data-driven method that uses Hi-C contact map as constraints. The method, which uses the maximum entropy theory, samples from a distribution that describe the structural ensemble of chromosome. The 10,000 structures are randomly sampled and are independent from each other. The HIPPS method is not a simulation, and hence the issue of initial structures does not arise.

9) In Methods, when the authors mention the "unknown parameter", do they use one parameter for all simulations (+/- cohesin) or is this parameter different for each system? Would this change the results?

We apologize for the confusion. The “unknown parameter” is the energy scale 𝜖 that describes the interaction strength between chromosome loci. We have revised the text in the method (page 27) to clarify it. The same value of 𝜖 is used for all CCM simulation with or without cohesin.

10) In Methods, when the authors perform DBSCAN clustering, they mention that they optimize the clustering parameters for each system. However, if they want to compare between different systems, the clustering parameters should be the same.

The purpose of DBSCAN is to capture the spatial clustering topology of chromosome loci. However, different cell types and chromosomes may have different overall density, which will impact the average distance between loci. If using the same parameters, such global changes will impact the result of clustering most and the intended spatial clustering topology can be distorted. Hence, we tune the clustering parameter for each system in order to ignore the global effect but only capture the local and topology of clustering of chromosome loci.

Grammar comments:

1) "structures, with sharp boundaries are present, at.."

We thank the reviewer for pointing out the error. We have fixed it.

2) "Three headlines emerge from these studies are:"

We have fixed it.

3) "both the cell lines"

We have fixed it.

Author Response

Reviewer #1 (Public Review):

This manuscript presents the first evidence for a plastic enhancement in the response of pial cortical arterioles to external stimulation. Specifically, they show (p8; Figure 3A-C) that repeated application of a visual stimulus at 0.25 Hz, at the upper edge of the vasomotor response, leads to a greater change in the diameter of pial arterioles at that frequency. This adds to the earlier, referenced work of Mateo et al (2017) that showed locking - or entrainment of pial arteriole vasomotion - by stimuli at different (0.0 to 0.3 Hz) frequencies.

We thank the reviewer for positively identifying the value of our manuscript.

The manuscript has a major flaw. Much as there is plasticity that leads to an increase in the amplitude of vasomotion at the drive frequency, the authors need to show reversibility. This could possibly be accomplished by driving the visual system at a different frequency, say 0.15 Hz, and observing if the 0.25 Hz response is then diminished. The authors could then test if their observation is repeatable by again driving at 0.25 Hz. Unless I missed the presentation on this point, there is no evidence for reversibility.

The reviewer has raised a very important point of view. In our experiments, the visually induced vasomotion (or visual stimulus-triggered vasomotion) was always entrained by repeated trials of the 0.25 Hz temporal frequency stimuli. When the visual stimulation stops, the vasomotion frequency lock to 0.25 Hz quickly dissipates. After saturated training with this stimulus, the parameters of the visual stimulus were switched, for example to 0.15 Hz. The animal quickly adapted to this new stimulus paradigm and the vasomotion was frequency-locked to 0.15 Hz. The adaptation to this new paradigm occurred well within 5 minutes. In Fig. 5, various paradigms were randomly tested. In some of the trials, 0.25 Hz stimulus was tested after 0.15 Hz. The vasomotion also quickly adapted back to the 0.25 Hz. We agree with the reviewer that this reversibility could have been explicitly documented in the manuscript.

Drew, P. J., A. Y. Shih, J. D. Driscoll, P. M. Knutsen, D. Davalos, P. Blinder, K. Akassoglou, P. S. Tsai, and D. Kleinfeld. 2010. 'Chronic optical access through a polished and reinforced thinned skull', Nature Methods, 7: 981-84.

Morii, S., A. C. Ngai, and H. R. Winn. 1986. 'Reactivity of rat pial arterioles and venules to adenosine and carbon dioxide: With detailed description of the closed cranial window technique in rats', Journal of Cerebral Blood Flow & Metabolism, 6: 34-41.

Reviewer #2 (Public Review):

Sasaki et al. investigated methods to entrain vasomotion in awake wild-type mice across multiple regions of the brain using a horizontally oscillating visual pattern which induces an optokinetic response (HOKR) eye movement. They found that spontaneous vasomotion could be detected in individual vessels of their wild-type mice through either a thinned cranial window or intact skull preparation using a widefield macro-zoom microscope. They showed that low-resolution autofluorescence signals coming from the brain parenchyma could be used to capture vasomotion activity using a macro-zoom microscope or optical fibre, as this signal correlates well with the intensity profile of fluorescently-labelled single vessels. They show that vasomotion can also be entrained across the cortical surface using an oscillating visual stimulus with a range of parameters (with varying temporal frequencies, amplitudes, or spatial cycles), and that the amplitude spectrum of the detected vasomotion frequency increases with repeated training sessions. The authors include some control experiments to rule out fluorescence fluctuations being due to artifacts of eye movement or screen luminance and attempt to demonstrate some functional benefit of vasomotion entraining as HOKR performance improves after repeat training. These data add in an interesting way to the current knowledge base on vasomotion, as the authors demonstrate the ability to entrain vasomotion across multiple brain areas and show some functional significance to vasomotion with regards to information processing as HOKR task performance correlates well with vascular oscillation amplitudes.

We thank the reviewer for summarizing the value of our study and recognizing its significance.

The aims of the paper are mostly well supported by the data, but some streamlining of the data presentation would improve overall clarity. The third aim to establish the functional significance of vasomotion in relation to plasticity in information processing could be better supported by the inclusion of some additional control experiments.

We thank the reviewer for recognizing our vast amount of data supporting our findings. We agree that better data presentation could have improved the clarity of the manuscript.

Specifically:

1) The clarity and comprehensibility of the paper could be significantly enhanced by incorporating additional details in both the introduction and discussion sections. In the introduction, a succinct definition of the frequency range of vasomotion should be provided, as well as a better description of the horizontal optokinetic response (i.e. as they have in the results section in the first paragraph below the 'Entrainment of vasomotion with visual stimuli presentation' sub-heading). The discussion would benefit from the inclusion of a clear summary of the results presented at the start, and the inclusion of stronger justification (i.e. more citations) with regards to the speculation about vasomotion and neuronal plasticity (e.g. paragraph 5 includes no citations).

We agree that a better description of vasomotion and horizontal optokinetic response could have been provided in the introduction. As the reviewer suggests, the discussion could also have started with the following summary of the results.

“We show that visually induced vasomotion can be frequency-locked to the visual stimulus and can be entrained with repeated trials. The initial drive for the vasomotion, or the sensory-evoked hyperemia, must be coming from the neuronal activity in the visual system. The vasomotion is likely triggered by activation of the neurovascular interaction (Kayser, 2004; van Veluw et al., 2020). Surprisingly, the entrained vasomotion was observed not only in the visual cortex but also widely throughout the surface of the brain and deep in the cerebellar flocculus. The global entrainment could be realized through separate mechanisms from the local neurovascular coupling. What is also unknown is where the plasticity occurs. The neuronal visual response in the primary visual cortex could potentially decrease with repeated visual stimulation presentation as the adaptive movement of the eye should decrease the retinal slip. With repeated training sessions, a more static projection of the presented image will likely be shown to the retina. The neurovascular coupling could be enhanced with increased responsiveness of the vascules and vascular-to-vascular coupling could also be potentiated.”

2) The novel methods for detecting vasomotion using low-resolution imaging techniques are discussed across the first four figures, but this gets a little bit confusing to follow as the authors jump back and forth between the different imaging and analysis techniques they have employed to capture vasomotion. The data presentation could be better streamlined - for instance by presenting only the methods most relevant for the functional dataset (in Figures 5-7), with the additional information regarding the various controls to establish the use of autofluorescence intensity imaging as a valid method for capturing vasomotion reduced to fewer figure panels, or moved to supplementary figures so as to not detract from the main novel findings contributed in this study.

We apologize for the confusing presentation of the data. Many of the initial figures were technical; however, we feel that following these steps was necessary to logically conclude that shadow imaging of the autofluorescence could be used as an indicator of vasomotion. We do agree with the reviewer that going back and forth between different techniques can be confusing. We could have added separate supplementary figures to introduce the various methods used upfront before going into the findings.

3) The authors heavily rely on representative traces from individual vessels to illustrate their findings, particularly evident in Figures 1-4. While these traces offer a valuable visualization, augmenting their approach by presenting individual data points across the entire dataset, encompassing all animals and vessels, would significantly enhance the robustness of their claims. For instance, in Figures 1 and 2, where average basal and dilated traces are depicted for a representative vessel, supplementing these with graphs showcasing peak values across all measured vessels would enable the authors to convey a more holistic representation of their data. Or in Figure 3, where the amplitude spectrum is presented for individual Texas red fluorescence intensity changes in V1 across novice, trained, and expert mice, incorporating a summary graph featuring the amplitude spectrum value at 0.25Hz for each individual trace (across animals/imaging sessions), followed by statistical analysis, would fortify the strength of their assertions. Moreover, providing explicit details on sample sizes for each individual figure panel (where not a representative trace), including the number of animals or vessels/imaging sessions, would contribute to transparency and aid readers in assessing the generalisability of the findings.

We agree with the reviewer that summarization of the data across a number of vessels/imaging sessions would lead to more generalization of the findings. However, contrary to what the reviewer described, we did summarize the vessel diameter expansion events across multiple vessel observations in Fig. 1F, G. The vasomotion parameters were not summarized for observation in intact skull shown in Fig. 2. However, this figure was intended just to show that vessel boundary cannot be well defined in intact skull imaging and Texas Red intensity or autofluorescence intensity fluctuation would give a better indication of vessel diameter fluctuation. In Fig. 3G, the peak ratio of 0.25 Hz was calculated for individual animals at Novice, Trained, and Expert levels and summarized for n = 5 animals. Statistical analysis was also done. The variability between imaging sessions within individual animals was not analyzed; thus, this could have been indicated.

4) In the experiments where mice are classed as "novice", "trained" or "expert", the inclusion of the specific range of the number of training sessions for each category would improve replicability.

We agree with the reviewer that classification on the level of training should have been explicitly indicated. Mice experiencing the first visual training session were defined as “Novice”. The mice that have experienced 3 training sessions are the “Trained” mice and the performance of the “Trained” mice during the 4th training session was evaluated. Mice that experienced 8 to 11 rounds of visual training sessions are the “Expert” mice.

5) The authors don't state whether mice were habituated to the imaging set-up prior to the first data collection, as head-fixation and restraint can be stress-inducing for animals, especially upon first exposure, which could impact their neurovascular coupling responses differentially in "novice" versus "trained" imaging sessions (e.g. see Han et al., 2020, DOI: https://doi.org/10.1523/JNEUROSCI.1553-20.2020). The stress associated with a tail vein injection prior to imaging could also partially explain why mice didn't learn very well if Texas Red was injected before the training session. If no habituation was conducted in these experiments, the study would benefit from the inclusion of some control experiments where "novice" responses were compared between habituated and non-habituated animals.

We agree with the reviewer that stress could well affect spontaneous vasomotion as well as visually induced vasomotion (or visual stimulus-triggered vasomotion). As the reviewer suggested, we could have compared the habituated and non-habituated mice to the initial visually induced vasomotion response. In addition, whether the experimentally induced increase in stress would interfere with the vasomotion or not could also be studied. With the Texas Red experiments, we observed that tail-vein injection stress appeared to interfere with the HOKR learning process. In the experiments presented in Fig. 3, Texas Red was injected before session 1. Vasomotion entrainment likely progressed with sessions 2 and 3 training. Before session 4, Texas Red was injected again to visualize the vasomotion. The vasomotion was clearly observed in session 4, indicating that the stress induced by tail-vein injection could not interfere with the generation of visually induced vasomotion.

6) The experiments regarding the brain-wide vasomotion entrainment across the cortical surface would benefit from some additional information about how brain regions were identified (e.g. particularly how V1 and V2 were distinguished given how close together they are).

The brain regions were identified by referring to the Mouse Brain Atlas. As the skull was intact, the location of bregma, lambda, and midline was clearly visible. We agree with the reviewer that strict separation of V1 and V2 could be difficult if we rely on the brain atlas alone. However, what we wanted to emphasize was that there was no specific localization of the vasomotion entrainment effect.

7) Whilst the authors show that HOKR task performance and vasomotion amplitude are increased with repeated training to provide some support to their aim of investigating the functional significance of vasomotion with regards to information processing plasticity, the inclusion of some additional control experiments would provide stronger evidence to address this aim. For instance, if vasomotion signalling is blocked or reduced (e.g. using optogenetics or in an AD mouse model where arteriole amyloid load restricts vasomotion capacity), does flocculus-dependent task performance (e.g. HOKR eye movements) still improve with repeated exposure to the external stimulus.

We agree that experimental intervention to vasomotion is ideal to test the functional significance of vasomotion. As pharmacological intervention lacks specificity, we are currently exploring the optogenetic approach. We have never thought of using the AD mouse as a model of restricted vasomotion by amyloid, and we agree this would be an interesting model to study. However, the AD mouse model would also have deficits other than the restricted vasomotion. On the other hand, we could test whether the repeated presentation of slowly oscillating visual stimuli can have beneficial effects in improving the cognitive abilities of AD model mice.

Reviewer #3 (Public Review):

Summary:

Here the authors show global synchronization of cerebral blood flow (CBF) induced by oscillating visual stimuli in the mouse brain. The study validates the use of endogenous autofluorescence to quantify the vessel "shadow" to assess the magnitude of frequency-locked cerebral blood flow changes. This approach enables straightforward estimation of artery diameter fluctuations in wild-type mice, employing either low magnification wide-field microscopy or deep-brain fibre photometry. For the visual stimuli, awake mice were exposed to vertically oscillating stripes at a low temporal frequency (0.25 Hz), resulting in oscillatory changes in artery diameter synchronized to the visual stimulation frequency. This phenomenon occurred not only in the primary visual cortex but also across a broad cortical and cerebellar surface. The induced CBF changes adapted to various stimulation parameters, and interestingly, repeated trials led to plastic entrainment. The authors control for different artefacts that may have confounded the measurements such as light contamination and eye movements but found no influence of these variables. The study also tested horizontally oscillating visual stimuli, which induce the horizontal optokinetic response (HOKR). The amplitude of eye movement, known to increase with repeated training sessions, showed a strong correlation with CBF entrainment magnitude in the cerebellar flocculus. The authors suggest that parallel plasticity in CBF and neuronal circuits is occurring. Overall, the study proposes that entrained "vasomotion" contributes to meeting the increased energy demand associated with coordinated neuronal activity and subsequent neuronal circuit reorganization.

We thank the reviewer for providing a thorough summarization of our manuscript.

Strengths:

• The paper describes a simple and useful method for tracking vasomotion in awake mice through an intact skull.

• The work controls for artefacts in their primary measurements.

• There are some interesting observations, including the nearly brain-wide synchronization of cerebral blood flow oscillations to visual stimuli and that this process only occurs after mice are trained in a visual task.

• This topic is interesting to many in the CBF, functional imaging, and dementia fields.

We thank the reviewer for positively recognizing the strength of the paper.

Weaknesses:

• I have concerns with the main concepts put forward, regarding whether the authors are actually studying vasomotion as they state, as opposed to functional hyperemia which is sensory-induced changes in blood flow, which is what they are actually doing. I recommend several additional experiments/analyses for them to explore. This is mostly further characterizing their effect which will benefit the interpretations.

We recognized that the terminology used in our paper was not explicitly explained. Traditionally, “vasomotion” is defined as the dilation and constriction of the blood vessels that occurs spontaneously at low frequencies in the 0.1 Hz range without any apparent external stimuli. Sensory-induced changes in the blood flow are usually called “hyperemia”. However, in our paper, we used the term, vasomotion, literally, to indicate both forms of “vascular” “motion”. Therefore, the traditional vasomotion was called “spontaneous vasomotion” and the hyperemia induced with slow oscillating visual stimuli was called “visually induced vasomotion”.

Using our newly devised methods, we show the presence of “spontaneous vasomotion”. However, this spontaneous vasomotion was often fragmented and did not last long at a specific frequency. With visual stimuli that slowly oscillated at temporal frequencies close to the frequency of spontaneous vasomotion, oscillating hyperemia, or “visually induced vasomotion” was observed.

• Neuronal calcium imaging would also benefit the study and improve the interpretations.

In our paper, we mainly studied the visually induced vasomotion (or visual stimulus-triggered vasomotion). Therefore, visual stimulation must first activate the neurons and, through neurovascular coupling, the initial drive for vasomotion is likely triggered. However, visually induced vasomotion is not observed in novice animals. Therefore, the visually induced vasomotion is not a simple sensory reaction of the vascular in response to neuronal activity in the primary visual cortex. We also do not know how the synchronized vasomotion can spread throughout the whole brain. Where the plasticity for vasomotion entrainment occurs is also unknown. To identify the extent of the neuronal contribution to the vasomotion triggering, whole brain synchronization, and vasomotion entrainment, simultaneous neuronal calcium imaging would be ideal. However, due to the fact that fluorescent Ca2+ indicators expressed in neurons would also be distorted by the “shadow” effect from the vasomotion, exquisite imaging techniques would be required.

• The plastic effects in vasomotion synchronization that occur with training are interesting but they could use an additional control for stress. Is this really a plastic effect, or is it caused by progressively decreasing stress as trials and progress? I recommend a habituation control experiment.

As also pointed out by reviewer #2, we agree that, whether stress would affect visually induced vasomotion or not could be studied. Studying the visually induced vasomotion in mice well-habituated to the experimental apparatus would give an idea of whether stress could truly be a profounding factor affecting vasomotion. On the other hand, whether acutely induced stress can interfere with the already entrained vasomotion could also be studied. In the experiments presented in Fig. 3, Texas Red was injected via the tail vein, which would be quite stressful for the mouse. However, in the trained mouse, visually induced vasomotion could be observed regardless of the stress. It is likely that stress can interfere with the acquisition of vasomotion entrainment, but the already acquired entrainment will not be canceled with acute stress induced by tail-vein injection. We agree that further relationship between stress and vasomotion and plasticity related to vasomotion entrainment could be investigated.

Appraisal

I think the authors have an interesting effect that requires further characterization and controls. Their interpretations are likely sound and additional experiments will continue to support the main hypothesis. If brain-wide synchrony of blood flow can be trained and entrained by external stimuli, this may have interesting therapeutic potential to help clear out toxic proteins from the brain as seen in several neurodegenerative diseases.

We thank the reviewer for the positive evaluation of our manuscript. Strong entrainment of visually induced vasomotion was observed with a simple presentation of slowly oscillating visual stimuli for several days. This is a totally non-invasive method to train the vasomotion capacity. As the reviewer recognizes, potential benefits for the treatment of dementia and neurodegenerative diseases could be evaluated with further studies.

The differentiation between visible and invisible disabilities in this text serves as a crucial reminder of the broad spectrum of disabilities and the varied experiences of those living with them. It underscores the need for greater awareness and sensitivity towards individuals whose disabilities may not be immediately apparent. The mention of the unjust stigma faced by individuals with invisible disabilities raises important questions about societal attitudes and the need for a shift towards more compassionate and informed perspectives. This section also implicitly advocates for a more inclusive definition of disability, one that acknowledges the complexity and diversity of individual experiences, thereby fostering a more accommodating and supportive community.

This explains different ways to help people with disabilities, like using special tools or making places easier for everyone to use. An example could be the tiktok update allowing for autogenerated subtitles so people didn't have to consciously write out every line. It shows that helping disabled people often means changing our surroundings or technology to fit their needs better.

Цікаво, наскільки змінилася ситуація з часу написання цієї статті. Звичайно, "базові" соціологічні твори були написані переважно білими чоловіками (переважно, бо були Рут Бенедикт та Маргарет Мід, наприклад), проте з огляду на авторство джерел, що вивчаються принаймні в Могилянці на соціології, кількість впливових соціологинь значно зросла

I think it is important to have a foundation for us teachers to refer to in order to help lead us in the right direction to make the "right" solution. This is important because sometimes teachers make connections with their students or have certain feelings towards an individual due to there behavior or whatever it may be which can hinder there decisions making. So it is important that we do not let our feelings get in the way of our decision making but rather rely on the ethics of the situation.

Author Response

We thank the editors and the reviewers for their assessment of our revised manuscript. Please see bellow, our answers to the recommendations by reviewer #2.

Figure S2F - Seems like a very narrow range of parameters. Is there some fine tuning here?

The range of values of tau_P that yields previous-trial biases is bounded by below and above for the following reasons: above a certain value of tau_P (therefore large integration time), the bump that had formed in the previous trial is not strong enough to remain stable for a long time, and therefore dissipates by the time the current trial starts (especially when adaptation is fast, towards the left of the third panel). Below a certain value, instead, this integration timescale is small enough to quickly form a representation of the current trial, hence the bump from the previous trial quickly dissipates (due to mutual inhibition). This interplay between the integration and the adaptation timescale as well as considering a phenomenon which is bounded in time (how close the activity bump is to the second stimulus of the previous trial which is presented between -22.4 and -5.6 seconds from the moment we are considering) yields a region for tau_P which is bounded. This region, however, appears narrow due to the limited number of points we have considered for the simulation grid.

Regarding my comment on lapse at the boundaries (old line 221). Lapse parameters in psychometric curves correspond to errors on the "easy" trials. But the mechanistic explanation for lapse trials is that there is a non-zero probability for the subject to respond in a manner that is random and independent of the stimulus. In the case of extreme stimuli, this is the only reason for errors, and thus looking at the edges of the psychometric curves allows to calculate lapse rate. But - the usual assumption for underlying mechanism is that the subject lapses in all trials, regardless of stimulus. If I understand correctly, this is different than the mechanistic reason for lapses in the network model, which was described as something that happens more in the edges than in the center. Or more generally, to be a stimulus-dependent effect.

We thank the reviewer for this clarification. The reviewer is right that in our mechanistic model, lapses (as defined by errors on easy trials) are more likely to occur for extreme stimuli, due to the vicinity to the boundary of the attractor. Such errors also occur for non-extreme stimuli, when delay intervals are long enough for the bump in PPC to drift to the boundaries. In experiments, lapse trials as described by the reviewer occur due to multiple different reasons; for lapse that is independent of the stimuli, mechanisms such as attention have been thought to play a role, this however is not included in our model.

What are the parameters for the distributions (skewed, bimodal, ...)?

These parameters are reported in the legend of Fig.6, where the distributions appear.

Bump with adaptation. Sorry for the draft-like comment. I don't think the existing studies are in the form you describe. I do think it might be useful to point readers to these studies. If an interested reader wishes to understand network dynamics in this and similar scenarios, it might be useful to have the pointers. The reference I had in mind was Romani, S., & Tsodyks, M. (2015). Short‐term plasticity based network model of place cells dynamics. Hippocampus, 25(1), 94-105.

We thank the reviewer for the clarification, and we will include this reference in the Version of Record.

---

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

eLife assessment

This is an important study about the mechanisms underlying our capacity to represent and hold recent events in our memory and how they are influenced by past experiences. A key aspect of the model put forward here is the presence of discrete jumps in neural activity with the posterior parietal region of the cortex. The strength of evidence is largely solid, with some weaknesses noted in the methodology. Both reviewers suggested ways in which this aspect of the model can to be tested further and resolve conflicts with previously published experimental results, in particular the study by Papadimitriou et al 2014 in Journal of Neurophysiology.

We thank the editors for their assessment. As mentioned in the cover letter, we have addressed all the reviewers’ concerns and would like to request and update of the assessment to reflect the revisions we have made.

Public Reviews:

We thank both reviewers for their careful reading and feedback that helped clarify many aspects of the model. Below, we address their comments.

Reviewer #1 (Public Review):

This paper aims to explain recent experimental results that showed deactivating the PPC in rats reduced both the contraction bias and the recent history bias during working memory tasks. The authors propose a twocomponent attractor model, with a slow PPC area and a faster WM area (perhaps mPFC, but unspecified). Crucially, the PPC memory has slow adaptation that causes it to eventually decay and then suddenly jump to the value of the last stimulus. These discrete jumps lead to an effective sampling of the distribution of stimuli, as opposed to a gradual drift towards the mean that was proposed by other models. Because these jumps are single-trial events, and behavior on single events is binary, various statistical measures are proposed to support this model. To facilitate this comparison, the authors derive a simple probabilistic model that is consistent with both the mechanistic model and behavioral data from humans and rats. The authors show data consistent with model predictions: longer interstimulus intervals (ISIs) increase biases due to a longer effect over the WM, while longer intertrial intervals (ITIs) reduce biases. Finally, they perform new experiments using skewed or bimodal stimulus distributions, in which the new model better fits the data compared to Bayesian models.

The mechanistic proposed model is simple and elegant, and it captures both biases that were previously observed in behavior, and how these are affected by the ISI and ITI (as explained above). Their findings help rethink whether our understanding of contraction bias is correct.

On the other hand, the main proposal - discrete jumps in PPC - is only indirectly verified.

We agree with the reviewer that the evidence for discrete jumps in PPC has been provided in behavioural results (short-term, n-back trial biases), and not from neural data. However, we believe electrophysiological investigations are out of the scope of the current manuscript and future works are needed to further verify the results.

The model predicts a systematic change in bias with inter-trial-interval. Unless I missed it, this is not shown in the experimental data. Perhaps the self-paced nature of the experiments allows to test this?

We thank the reviewer for this great suggestion.

We had not previously looked at this in the data for the reason that in the simulations, the ITI is set to either 2.2, 6 or 11 seconds, whereas the experiment is self-paced. Therefore, any comparison with the simulation should be made carefully.

However, after the reviewer’s suggestion, we did look at the change in the bias with the inter-trial interval, by dividing trials according to ITIs lower than 3 seconds (“short” ITI), and higher than 3 seconds (“long” ITI). This choice was motivated by the shape of the distribution of ITIs, which is bimodal, with a peak around 1 second, and another after 3 seconds (new Fig 8F). Hence, we chose 3 seconds as it seemed a natural division. However, 3 seconds also happens to be approximately the 75th percentile of the distribution, and this means that there is much more data in the “short” ITI than the “long” ITI set. In order to have sufficient data in the “long” ITI for clearer effects we used all of our dataset – the negatively skewed, and also two bimodal distributions (of which only one was shown in the manuscript, for succinctness). This larger dataset allows us to clearly see not only a decreasing contraction bias with increasing ITI (Fig 8G), but also a decreasing onetrial-back attractive bias with increasing ITI (Fig 8H). We have uploaded all the datasets as well as scripts used to analyze them to this repository: https://github.com/vboboeva/ParametricWorkingMemory_Data.

The data in some of the figures in the paper are hard to read. For instance, Figure 3B might be easier to understand if only the first 20 trials or so are shown with larger spacing. Likewise, Figure 5C contains many overlapping curves that are hard to make out.

We have limited the dynamics in Fig 3B to the first 50 trials for better visibility. Likewise, as suggested, we report the standard error of the mean instead of the standard deviation in old Fig 5C (new Fig 6C) – this allows for the different curves to be better discernible.

There is a gap between the values of tau_PPC and tau_WM. First - is this consistent with reports of slower timescales in PFC compared to other areas?

Recent studies by Xiao-Jing Wang and colleagues (Refs. 1-3 below) suggest that may be the case. In Wang et al 2023, Ref 1 below), the authors use a generative model to study the concept of bifurcation in space in working memory, that is accompanied by an inverted-V shape of the time constants as a function of cortical hierarchy.

Briefly, they propose a generative model of the cortex with modularity, incorporating repeats of a canonical local circuit connected via long-range connections. In particular, the authors define a hierarchy for each local circuit. At a critical point in this hierarchy axis, there is a phase transition from monostability to bistability in the firing rate. This means that a local circuit situated below the critical point will only display a low activity steady state, while those above the critical point additionally display a persistent activity steady state.

The model predicts a critical slowing down of the neural fluctuations at the critical point, resulting in an inverted-V shape of the time constants as a function of the hierarchy. They test the predictions of their model – the bifurcation in space and that inverted-V-shaped time constants as a function of the hierarchy - on connectome-based models of the macaque and mouse cortex. Interestingly both datasets show similar behavior. In particular, during working memory, frontal areas (higher in the hierarchy, e.g. area 24c in macaques) has a smaller time constant relative to posterior parietal areas (lower in the hierarchy, like LIP or f7). We have now cited this new work.

[1] https://www.biorxiv.org/content/10.1101/2023.06.04.543639v1

[2] https://elifesciences.org/articles/72136

[3] https://www.biorxiv.org/content/10.1101/2022.12.05.519094v3.abstract

Second - is it important for the model, or is it mostly the adaptation timescale in PPC that matters?

We have run simulations producing a phase diagram with tau_theta^P on the x-axis, tau^P on the y-axis, and in color, the fraction of trials in which the bump is in the vicinity of a target (Fig S2 F), before the network is presented with the second stimulus. This target can be the first stimulus s_1 (left), mean over stimuli (middle) and previous trial’s stimulus (right)). White point corresponds to parameters of the default network.

In this phase diagram, the lowest value that tau_P takes is tau_WM=0.01. When tau_P=tau_WM, the bump is rarely in the vicinity of 1-trial-back stimulus, and we can see that tau_PPC should be greater than tau_WM in order for the model to yield 1-trial back effects. We conclude that it is indeed important for tau_PPC > tau_WM.

We have included this in Fig S2 F of the manuscript.

Regarding the relation to other models, the model by Hachen et al (Ref 45) also has two interacting memory systems. It could be useful to better state the connection, if it exists.

The model proposed by Hachen et al is conceptually different in that one module stores the mean of the sensory stimulus; it could be related to a variant of our model where adaptation is turned off in the PPC network (Fig S2 A). However, the task they model is also different: subjects have to learn the location of a boundary according to which the stimulus is classified as ‘weak’ or ‘strong’, set by the experimenter. Hence, it is a task where learning is needed - this contrasts with the task we are modelling, where only working memory is required. How task demands reconfigure existing circuits via dynamics and/or learning to perform different computations is a fascinating area of research that is outside the scope of this work.

Reviewer #2 (Public Review):

Working memory is not error free. Behavioral reports of items held in working memory display several types of bias, including contraction bias and serial dependence. Recent work from Akrami and colleagues demonstrates that inactivating rodent PPC reduces both forms of bias, raising the possibility of a common cause.

In the present study, Boboeva, Pezzotta, Clopath, and Akrami introduce circuit and descriptive variants of a model in which the contents of working memory can be replaced by previously remembered items. This volatility manifests as contraction bias and serial dependence in simulated behavior, parsimoniously explaining both sources of bias. The authors validate their model by showing that it can recapitulate previously published and novel behavioral results in rodents and neurotypical and atypical humans.

Both the modeling and the experimental work is rigorous, providing compelling evidence that a model of working memory in which reports sometimes sample past experience can produce both contraction bias and serial dependence, and that this model is consistent with behavioral observations across rodents and humans in the parametric working memory (PWM) task.

Evidence for the model advanced by the authors, however, remains incomplete. The model makes several bold predictions about behavior and neural activity, untested here, that either conflict with previous findings or have yet to be reported but are necessary to appropriately constrain the model.

First, in the most general (descriptive) formulation of the Boboeva et al. model, on a fraction of trials items in working memory are replaced by items observed on previous trials. In delayed estimation paradigms, which allow a more direct behavioral readout of memory items on a trial-by-trial basis than the PWM task considered here, reports should therefore be locked to previous items on a fraction of trials rather than display a small but consistent bias towards previous items. However, the latter has been reported (e.g., in primate spatial working memory, Papadimitriou et al., J Neurophysiol 2014). The ready availability of delayed estimation datasets online (e.g., from Rademaker and colleagues, https://osf.io/jmkc9/) will facilitate in-depth investigation and reconciliation of this issue.

As pointed out by the reviewer, in the PWM task that we are modelling here, the activity in the network is used to make a binary decision. However, it is possible to directly analyse the network activity before the onset of the second stimulus.

In their manuscript, Papadimitriou et al. study a memory-guided saccade task in nonhuman primates and argue that the animals display a small but consistent bias towards previous items (Fig 2). In that figure, the authors compute the error as the difference between the saccade direction and target direction in each trial. They compute this error for all trials in which the preceding trial’s target direction is between 35° and 85° relative to the current trial (counterclockwise with respect to the current trial’s target). They discover that the residual error distribution is unimodal with a mode at 1.29° and a mean at 2.21° (positive, so towards the preceding target’s direction), from which they deduce a small but systematic bias towards previous trial targets.

We have computed a similar measure for our network with default parameters (Table 1), by subtracting the location of the bump at the end of the delay interval (s_hat(t), ‘saccade’) from the initial location of the first stimulus in the current trial (s1(t) or the ‘target’). We have done this for all trials where s1(t)=0.2, and where s2(t-1) takes specific values. These distributions are characterized by two modes. The first corresponds to those trials where the bump is not displaced in WM (i.e. mean of zero). We can also see the appearance of a second mode at the location of s1(t) - s2(t-1), corresponding to the displacements towards the preceding trial’s stimulus described in the main text. If, instead, we limit the analysis to a small range of previous trials close to s1(t) (similar to Papadimitriou et al) then the distribution of residual errors will appear unimodal, as the two modes merge. Importantly, note that there is a large variability around the second mode, expressing a more complex dynamics in the network. As can be seen in Fig 3B, the location of the bump is not always slaved to the one in the PPC in a straightforward way -- due to the adaptation in the PPC, the global inhibition in the connectivity kernel, as well as interleaved design for various delay intervals, the WM bump can be displaced in nontrivial ways (see also Recommendation no 4), yielding the dispersion around the second peak. It remains to be seen whether such patterns can be observed in the data from previous works on continuous working memory recall (including Papadimitriou et al). However, to our knowledge, such detailed and full analysis of errors at the level of individual trials has not been done.

In summary, this analysis shows that the type of dynamics in our network is not one of the two cases: 1) small and systematic bias in each and every trial or 2) large error that occurs only rarely; rather, the dispersion around both modes suggests that the dynamics in our model are a mixture of these two limit cases.

We have also performed another typical analysis, reported in several continuous recall tasks (e.g. Jazayeri and Shadlen 2010) where contraction bias has been reported. We plot WM bump locations after the delay period for every trial (s_hat(t)), and their averages, against the nominal value of s1(t). We see that the mean WM location deviates from the identity line toward the mean values of s1(t), again showing contraction bias as an average effect, while individual trials follow the dynamics explained above.

We have now included a new section on continuous recall (Sect. 1.5 and a new figure (Fig 5)), which details the two above-mentioned analyses. The analysis of freely available datasets of delayed estimation tasks, unfortunately, is out of the scope of this work, and we leave such analyses to future studies.

Second, the bulk of the modeling efforts presented here are devoted to a circuit-level description of how putative posterior parietal cortex (PPC) and working-memory (WM) related networks may interact to produce such volatility and biases in memory. This effort is extremely useful because it allows the model to be constrained by neural observations and manipulations in addition to behavior, and the authors begin this line of inquiry here (by showing that the circuit model can account for effects of optogenetic inactivation of rodent PPC).

Further experiments, particularly electrophysiology in PPC and WM-related areas, will allow further validation of the circuit model. For example, the model makes the strong prediction that WM-related activity should display 'jumps' to states reflecting previously presented items on some trials. This hypothesis is readily testable using modern high-density recording techniques and single-trial analyses.

As mentioned in response to the previous comment, we note again that in the WM network, the bump ‘displacement’ has a complex dynamics -- the examples we have provided in Fig 1A and 2B mainly show the cases in which jumps occur in the WM network, but this is not the only type of dynamics we observe in the model. We do have instances in which the continuity of the model causes drift across values, and we have now replaced the right panel in Fig 2B with one such instance, in order to emphasize that this displacement towards the previous trial’s stimulus (s2(t-1)) can occur in various ways. For a more thorough analysis, we have analyzed the distance between s1(t) and the position of the bump in the WM network at the end of the delay period s_hat(t), conditioned on specific values of s1(t) and s2(t-1) (Fig 5C). In this figure, we can see the appearance of two modes: one centered around 0, corresponding to the correct trials where the stimulus is kept in WM (s1(t) = s_hat(t)), and another mode centered around s2(t-1), the location of the second stimulus of the previous trial, where the bump is displaced. Note, as we explain in Sect. 1.5, the large dispersion around this second mode, which suggests that the bump is not always displaced to that specific location and may undergo drift.

We agree with the reviewer that future electrophysiological experiments (or analysis of existing datasets) are necessary for validation of these results.

Finally, while there has been a refreshing movement away from an overreliance on p-values in recent years (e.g., Amrhein et al., PeerJ 2017), hypothesis testing, when used appropriately, provides the reader with useful information about the amount of variability in experimental datasets. While the excellent visualizations and apparently strong effect sizes in the paper mitigate the need for p-values to an extent, the paucity of statistical analysis does impede interpretation of a number of panels in the paper (e.g., the results for the negatively skewed distribution in 5D, the reliability of the attractive effects in 6a/b for 2- and 3- trials back).

We share the reviewer’s criticism towards the misuse of p-values – in order for a clearer interpretation of old Fig 5D (new Fig 7E), we have looked at the 2 and 3 trials-back biases by using all of our dataset – the negatively skewed, and also two bimodal distributions (of which only one was shown in the manuscript). This larger dataset of 43 subjects (approximately 17,200 trials) allows us to clearly see the 2 and 3 trial back attractive biases, and the effect that the delay interval exerts on them.

Reviewer #1 (Recommendations For The Authors):

Fig 5 A&C - It might be beneficial to separate the distribution of stimuli from the performance. It is hard to read the details of the performance, especially with error bars.

Following the next recommendation, we have exchanged the standard deviation to standard errors of the mean, hopefully this allows to better read the performance.

Fig 5C. The number of participants should be written. Perhaps standard errors instead of standard deviation?

We have now changed the standard deviation to standard errors of the mean and included the number of participants in the figure.

Fig 2B - hard to understand, because there is no marking of where "perfect" memory of s1 would be.

The perfect memory of s1 is shown in the upper panel as black bars.

Fig 3B. dot number 9 (blue, around 0.7) - why is WM higher than stimulus?

This trial has a long ISI (blue means 10s). During this delay, the bump in the PPC, under the influence of adaptation, drifts far below the first stimulus (note that the previous trial also had its first stimulus in the same location, as a result of which the adaptative thresholds have built up significantly, causing the bump to move away from that location). During this delay period, neurons in the WM network receive inputs from the PPC network: if this input is strong enough, it can disrupt an existing bump; if not, this input still exerts inhibiting influence on the existing bump via the global inhibition in the connectivity. This can cause an existing bump to slowly drift in a random direction, and finally dissipate. Note that the lines in Fig 2B represent the neuron with the maximal activity, this activity may be a stable bump, or an unstable bump that may soon dissipate.

Other examples with similar dynamics include trials 43 and 54.

L167 fewer -> smaller

We have now corrected this.

Fig 3C - bump can also be in between. Is this binned?

We have not binned the length of the attractor; to produce that figure, we check whether the position of the neuron with the maximal firing rate is within a distance of ±5% of the length of the whole line attractor from the target location.

L221 Lapse at the boundary of attractor. This seems very different from behavior. Specifically, if it is in the boundaries, it should be stimulus dependent.

Very sorry, we did not manage to understand the reviewer’s comment.

L236 are -> is

We have now corrected this.

Fig S4 - should be mostly in main text.

Part of this figure is in Fig 6A, but given the amount of detail, we think Supplementary Material is better suited.

L253-254. Differences across all distributions - very minor except the bimodal case.

That is correct, this is why we conducted the experiment with the bimodal distribution, to better differentiate the predictions of the two models.

L273 extra comma after "This probability"

We have now corrected this.

ITI was only introduced in section 1.5.2. Perhaps worth mentioning the default 5s value earlier in the paper.

We have now mentioned this in line 97-98.

Fig S6B title: perhaps "previous stimuli"?

We have now corrected this.

L364 i"n A given trial"

Equation 2 - no decay term?

Thank you for pointing out this error, we have now corrected this.

Equation 5,6 are j^W and j^P indices of neurons in those populations?

Yes, j^W indexes neurons in the WM network, and j^P those in the PPC. We have now added this in the text for clarity.

Bump with adaptation - other REFs? Sandro?

We are aware of continuous bump attractors implementing short-term synaptic plasticity in various studies (including by Sandro Romani), but not in the form we have described. May the reviewer kindly point us towards the relevant literature.

Free boundary - what is the connectivity for neurons 1 and N? Is it weaker than others? Is the integral still 1? Does this induce some bias on the extreme values?

The connectivity of the network is all-to-all. However, as expressed by Eq. (3), the distance-dependent contribution to the weights, K, decreases exponentially as we move from neuron 1 onwards, and from neuron N down. The sum (or integral, in the large-N limit) of the K_ij for j on either side of neuron i is unity only when i is sufficiently far from 1 or N. We have rephrased the paragraph starting in line 516 to make this clearer.

The presence of a boundary could introduce a bias in theory, but in practice, it affects the dynamics only when the bump drifts sufficiently close to it. The smallest stimulus in the simulated task has amplitude 0.2, with width 0.05, which implies the activation of 50 neurons on either side of neuron 400. If one compares this with the width of the kernel K in stimulus space (d_0 = 0.02), which spans ~10 neurons, we can see that the bump of activity stays mostly far from the boundary. It is possible, though it is observed rarely, when several consecutive long delay intervals happen to occur, that the bump in PPC drifts beyond the location corresponding to either the minimum or maximum stimulus.

Code availability?

Code simulating the dynamics of the network as well as analysing the resulting data can be found in the following repository: https://github.com/vboboeva/ParametricWorkingMemory Code used to analyse human behavioural data and fit them with our statistical model can be found in this repository: https://github.com/vboboeva/ParametricWorkingMemory_Data Code used to run the auditory PWM experiments with human subjects (adapted from Akrami et al 2018) can be found here: https://github.com/vboboeva/Auditory_PWM_human

L547 stimuli

We have now corrected this.

Equation 14 uses both stimuli. Was this the same for the rest of analysis in the paper (first figures for instance)?

This equation was used for all GLM analyses (Figs 9 and S6).

D0 is very small (0.02). Does this mean that activity is essentially discrete in the model? Fig 1A & 2B - the two examples of model activity suggest this is the case. In other words - are there cases where the continuity of the model causes drift across values? Can you show an example (similar to Fig 1A)?

Since this point has been raised beforehand, we refer to the first comment, Fig 2B and Sect. 1.5 for the response to this question.

Table 1 - inter trial interval 6. Text says 5

We have now corrected this in the text.

Reviewer #2 (Recommendations For The Authors):

In addition to my review above, I just have a few minor comments:

• If I understood correctly, the squares inside the purple rectangle in Figure 1B are meant to show a gradation from red to blue, but this was hard to make out in the pdf.

Actually the squares are all on one side or the other of the diagonal, therefore they do not have any gradation.

• line 164: "The resulting dynamics... [are]?"

We have corrected this in the text.

• Fig 7B legend: "The network performance is on average worse for longer ITIs" – correct?

This was a mistake, we have replaced worse with better.

Other comments

We realized that the colorbar reported the incorrect fraction classified in Figs 1B, 2C, 7B (new 8B), S2C, S3A, S5B. We have corrected this in the new version of the manuscript.

We also found a minor mistake in one of our analysis codes that computed the n-trial back biases for different delay intervals. This did not change our results, actually made the effects clearer. The figures concerned are Fig 3F and new Fig 7E.

Author Response

eLife assessment

This study presents important findings for understanding cortical processing of color, binocular disparity, and naturalistic textures in the human visual cortex at the spatial scale of cortical layers and columns using state-of-the-art high-resolution fMRI methods at ultra-high magnetic field strength (7 T). Solid evidence supports an interesting layer-specific informational connectivity analysis to infer information flow across early visual areas for processing disparity and color signals. While the question of how the modularity of representation relates to cortical hierarchical processing is interesting and fundamental, the findings that texture does not map onto previously established columnar architecture in V2 is suggestive but would benefit from further controls. The successful application of high-resolution fMRI methods to study the functional organization along cortical columns and layers is relevant to a broad readership interested in general neuroscience.

Thank you for your assessment of our manuscript "Mesoscale functional organization and connectivity of color, disparity, and naturalistic texture in human second visual area ". We have carefully considered the public reviews and have outlined our plans of revision by providing point-by-point responses to the reviewers’ comments.

Reviewer #1 (Public Review):

To support the finding that texture is not represented in a modular fashion, additional possibilities must be considered. These include the effectiveness and specificity of the texture stimulus and control stimuli, (b) further analysis of possible structure in images that may have been missed, and (c) limitations of imaging resolution.

Thank you for your suggestions. We will provide evidence and additional analyses to show that there was indeed a large difference in high-order statistical information between the texture and control stimuli in our study, and thus the contrast between the two stimuli should be effective in localizing the processing of high-order texture information. Compared to the previous studies, another reason for the weaker texture selectivity in the current study could be the smaller number of images used and the slower rate of image presentation. Although our fMRI result at 1-mm isotropic resolution did not show a modular processing of naturalistic texture in CO-stripe columns, this does not exclude the possibility that smaller modules exist beyond the current fMRI resolution. We will discuss these limitations in the revised manuscript.

More in-depth analysis of subject data is needed. The apparent structure in the texture images in peripheral fields of some subjects calls for more detailed analysis. e.g. Relationship to eccentricity and the need for a 'modularity index' to quantify the degree of modularity. A possible relationship to eccentricity should also be considered.

We will perform further analysis based on your suggestion, especially regarding the relationship between eccentricity and modulation index. We will discuss this possibility in the revised manuscript.

Given what is known as a modular organization in V4 and V3 (e.g. for color, orientation, curvature), did images reveal these organizations? If so, connectivity analysis would be improved based on such ROIs. This would further strengthen the hierarchical scheme.

Thank you for your suggestion. The informational connectivity analyses used highly informative voxels by feature selection, which may already represent information from the modular organizations in these higher visual areas. We will examine the functional maps for possible modular organizations.

Reviewer #2 (Public Review):

In lines 162-163, it is stated that no clear columnar organization exists for naturalistic texture processing in V2. In my opinion, this should be rephrased. As far as I understand, Figure 2B refers to the analysis used to support the conclusion. The left and middle bar plots only show a circular analysis since ROIs were based on the color and disparity contrast used to define thin and thick stripes. The interesting graph is the right plot, which shows no statistically significant overlap of texture processing with thin, thick, and pale stripe ROIs. It should be pointed out that this analysis does not dismiss a columnar organization per se but instead only supports the conclusion of no coincidence with the CO-stripe architecture.

Reviewer #1 also raised a similar concern. We agree that there may be a smaller functional module of textures in area V2 at a finer spatial scale than our fMRI resolution. We will rephrase our conclusions to be more precise.

In Figure 3, cortical depth-dependent analyses are presented for color, disparity, and texture processing. I acknowledge that the authors took care of venous effects by excluding outlier voxels. However, the GE-BOLD signal at high magnetic fields is still biased to extravascular contributions from around larger veins. Therefore, the highest color selectivity in superficial layers might also result from the bias to draining veins and might not be of neuronal origin. Furthermore, it is interesting that cortical profiles with the highest selectivity in superficial layers show overall higher selectivity across cortical depth. Could the missing increase toward the pial surface in other profiles result from the ROI definition or overall smaller signal changes (effect size) of selected voxels? At least, a more careful interpretation and discussion would be helpful for the reader.

We will discuss the limitations of cortical depth-dependent analysis using GE-BOLD fMRI. All our stimuli produced robust activations in these visual areas, thus the flat laminar profiles of modulatory indices are unlikely to be caused by smaller signal changes. We will show the original BOLD responses in addition to the modulation index.

I was slightly surprised that no retinotopy data was acquired. The ROI definition in the manuscript was based on a retinotopy atlas plus manual stripe segmentation of single columns. Both steps have disadvantages because they neglect individual differences and are based on subjective assessment. A few points might be worth discussing: (1) In lines 467-468, the authors state that V2 was defined based on the extent of stripes. This classical definition of area V2 was questioned by a recent publication (Nasr et al., 2016, J Neurosci, 36, 1841-1857), which showed that stripes might extend into V3. Could this have been a problem in the present analysis, e.g., in the connectivity analysis? (2) The manual segmentation depends on the chosen threshold value, which is inevitably arbitrary. Which value was used?

The retinotopic atlas on the standard surface is usually quite accurate in defining the boundaries of early visual areas. Although some stripes may extend into V3, these patterns should be more robust in V2. In our analysis, we selected only those with clear organizations within the retinotopic atlas. Thus, the signal contribution from V3 is likely to be small and would not affect the pattern of results. In addition, the results between V3 and V2 could be very different, we will compare the pattern of results from these areas in additional analyses. The threshold for segmentation is abs(T)>2, we will clarify this in the method.

The use of 1-mm isotropic voxels is relatively coarse for cortical depth-dependent analyses, especially in the early visual cortex, which is highly convoluted and has a small cortical thickness. For example, most layer-fMRI studies use a voxel size of around isotropic 0.8 mm, which has half the voxel volume of 1 mm isotropic voxels. With increasing voxel volume, partial volume effects become more pronounced. For example, partial volume with CSF might confound the analysis by introducing pulsatility effects.

We agree that the 1-mm isotropic voxel is much smaller in volume than the 0.8-mm isotropic voxel, but the resolution along the cortical depth is not a large difference. In addition to our study, there are also other studies showing that fMRI at 1-mm isotropic resolution is capable of resolving cortical depth-dependent signals. Also, our fMRI slices were oriented perpendicular to the calcarine sulcus, the higher in-plane resolution will also benefit in resolving depth-dependent signals. We will discuss these issues about fMRI resolution in the revised manuscript.

The SVM analysis included a feature selection step stated in lines 531-533. Although this step is reasonable for the training of a machine learning classifier, it would be interesting to know if the authors think this step could have reintroduced some bias to remaining draining vein contributions.

Several precautions have been taken in the ROI definition to reduce the influence of large draining veins. The same number of voxels were selected from each cortical depth for the SVM analysis, thus there was no bias from the superficial layers susceptible to draining veins. Also, since both feedforward and feedback connections involved the superficial voxels, the remaining influence of large draining veins should be comparable between the two connections.

Reviewer #3 (Public Review):

The authors tend to overclaim their results.

Thank you for your comments. We will add more control analyses to strengthen our findings, and have appropriate discussion of results.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary: The authors study the appearance of oscillations in motifs of linear threshold systems, coupled in specific topologies. They derive analytical conditions for the appearance of oscillations, in the context of excitatory and inhibitory links. They also emphasize the higher importance of the topology, compared to the strength of the links. Finally, the results are confirmed with WC oscillators, which are also linear. The findings are to some extent confirmed with spiking neurons, though here results are less clear, and they are not even mentioned in the Discussion.

Overall, the results are sound from a theoretical perspective, but I still find it hard to believe that they are of significant relevance for biological networks, or in particular for the oscillations of BG-thalamus-cortex loop in PD. I find motifs in general to be too simplistic for multiscale and generally large networks as is the case in the brain. Moreover, the division of regions is more or less arbitrary by definition, and having such a strong dependence on an odd/even number of inhibitory links is far from reality. Another limitation is the fact that the cortex is considered a single node. Similarly, decomposing even such a coarse network in all possible (238 in this case) motifs doesn't seem of much relevance, when I assume that the emergence of pathological rhythms is more of an emergent phenomenon.

Strengths:

From the point of view of nonlinear dynamics, the results are solid, and the intuition behind the proofs of the theorems is well explained.

Weaknesses:

As stated in the summary, I find the work to be too theoretical without a real application in biological systems or the brain, where the networks are generally very large.

We respectfully disagree with the reviewer here. The second half of the paper is all about explaining a biological problem. We have shown the validity of our theoretical results (which indeed were obtained in idealized settings) to explain emergence of oscillations in the basal ganglia. We clearly show that our theoretical results hold both in a rate-based model and in a network model with spiking neurons. The model with spiking neurons is one of the most complete network models of the basal ganglia available in the literature. So we emphasize that we have provided a clear application of our results for the brain networks.

It is not the problem in the simplicity of the model or of the topology, it is often the case that the phenomena are explained by very reduced systems, but the problem is that the applicability of the finding cannot be extended. E.g. the Kuramoto model uses all-to-all coupling, or similar with QIF neurons which also need to follow a Lorentzian distribution in order to derive a mean field.

We do not understand this comment. There is no need to extend these results to a network of Kuramoto models because in that setting we already assume that individual nodes/populations are oscillating – there is no problem of emergence of oscillations. Here, we are specifically considering a setting in which nodes themselves are not oscillators. We agree that we, at this point, have no insight into how to extend our analytical proof to a situation where individual nodes are spiking.

But in those cases, relaxing the strict conditions that were necessary for the derivations, still conserves the main findings of the analysis, which I don't see being the case here. The odd/even number rule is too strict, and talking about a fixed and definite number of cycles in the actual brain seems too simplistic.

We have clearly relaxed most of our assumptions when we considered a network model of basal ganglia in which each subpopulation is a collection of spiking neurons. And as we have shown our results still hold (see Figure 5). Again our model is about oscillations in a network of networks i.e. network of brain regions.

At meso-scale it is not unreasonable to find such cycles and even-odd number rules. We have shown this for the case of a cortico-basal ganglia model. We can also extend this to cortico-thalamic networks and so on. We have already emphasized this point in the introduction to avoid any confusion: see lines 62-66 – “We prove this conjecture for the threshold-linear network (TLN) model without delays which can closely capture the dynamics of neural populations. Therefore, it is implicit that our results do not hold at the neuronal level but rather at the level of neuron populations/brain regions e.g. the basal ganglia (BG) network which can be described a network of different nuclei.” and lines 69-70 – ’Within the framework of the odd-cycle theory, distinct nuclei are associated with either excitatory or inhibitory nodes.’

Being linear is another strong assumption, and it is not clear how much of the results are preserved for spiking neurons, even though there is such an analysis, or maybe for other nonlinear types of neuronal masses.

Clearly our results hold in a network of spiking neurons (see Figure 5). It is of course interesting to ask whether our results hold in a network where individual spiking neurons have more complex spiking behavior like AdEx or Quadratic IF. But that kind of analysis deserves a full manuscript on its own.

Delays are also mentioned, and their impact on the oscillatory networks is as expected: it reduces the amplitude, but there is no link to the literature, where this is an established phenomenon during synchronization. Finally, the authors should also discuss the time-delays as a known phenomenon to cause or amplify oscillations at different frequencies in a network of coupled oscillators, e.g Petkoski & Jirsa Network Neuroscience 2022, Tewarie et al. NeuroImage 2019, Davis et al. Nat Commun 2021.

This is indeed a weakness of our model. But as the reviewer already knows, dynamical systems with delays are very difficult to analyze analytically. We have mentioned this in the limitations of the model and the analysis. In our simulations we have considered delays and when the delays are within reasonable limits our results hold.

Reviewer #2 (Public Review):

Summary:

The authors present here a mathematical and computational study of the topological/graph theory requirements to obtain sustained oscillations in neural network models. A first approach mathematically demonstrates that a given network of interconnected neural populations (understood in the sense of dynamical systems) requires an odd number of inhibitory populations to sustain oscillations. The authors extend this result via numerical simulations of (i) a simplified set of Wilson-Cowan networks, (ii) a simplified circuit of the cortico-basal ganglia network, and (iii) a more complex, spike-based neural network of basal ganglia network, which provides insight on experimental findings regarding abnormal synchrony levels in Parkinson's Disease (PD).

Strengths:

The work elegantly and effectively combines solid mathematical proof with careful numerical simulations at different levels of description, which is uncommon and provides additional layers of confidence to the study. Furthermore, the authors included detailed sections to provide intuition about the mathematical proof, which will be helpful for readers less inclined to the perusal of mathematical derivations. Its insightful and well-informed connection with a practical neuroscience problem, the presence of strong beta rhythms in PD, elevates the potential influence of the study and provides testable predictions.

Weaknesses:

In its current form, the study lacks a more careful consideration of the role of delays in the emergence of oscillations. Although they are addressed at certain points during the second part of the study, there are sections in which this could have been done more carefully, perhaps with additional simulations to solidify the authors' claims. Furthermore, there are several results reported in the main figures which are not explained in the main text. From what I can infer, these are interesting and relevant results and should be covered. Finally, the text would significantly benefit from a revision of the grammar, to improve the general readability at certain sections. I consider that all these issues are solvable and this would make the study more complete.

This point has been made by the first reviewer as well. So we repeat our answer:

This is indeed a weakness of our model. But as the reviewer already knows, dynamical systems with delays are very difficult to analyze analytically. We have mentioned this in the limitations of the model and the analysis. In our simulations we have considered delays and when the delays are within reasonable limits our results hold.

Reviewer #2 (Recommendations For The Authors):

As mentioned in my comments above, I think that the work is already quite solid and relevant but would significantly improve if some issues were addressed:

We would like to thank the reviewer for valuable comments and constructive feedback which has helped us greatly improve the manuscript.

1) While the authors acknowledge early on the limitations of this study in terms of not considering plasticity or neuron biophysics (line 72), I think that the absence of propagation delays should be explicitly included here. This absence leads to inaccuracies --for example, the sentence "Consider a small network of two nodes. If we connect them mutually with excitatory synapses, intuitively we can say that the two-population network will not oscillate" (line 74) is only correct if the delays (or signal latencies) are zero. With a proper delay, two excitatory neurons can engage in oscillations with a period given by two times the value of the delay.

A similar situation happens for inhibitory neurons, where the winner-take-all dynamics described in line 77 is only valid for zero delay. It is known that a homogeneous population of inhibitory spiking neurons with delayed synapses can lead to fast oscillations (Brunel and Hakim 1999), something which is also valid for the equivalent inhibitory single node with delayed self-inhibition. Indeed, a circuit of two inhibitory populations with delayed self- and cross-inhibition can generate oscillations, contradicting the main conclusion of the odd number of inhibitory nodes needed for oscillations.

Because of these considerations, I think the authors should be more careful when explaining the effects of delays, and state that their main results on the link between oscillations and having an odd number of inhibitory nodes are not valid when delays are considered. They could modify the sentences in lines 72-77 above and include a supplementary figure right after their simulation study for the Wilson-Cowan (to explain the examples above, and also the one in the next point).

The reviewer has brought up a critical point regarding the impact of propagation delays, and we completely concur with your assessment. In our study, we indeed did not comprehensively consider the effects of propagation delays in cycles with even inhibition, which may introduce inaccuracies in our conclusions.

We note that in the Wilson-Cowan model with delays, certain cycles with even number of inhibitory links can also generate oscillations with a period equal to twice the delay value. However, in our hand such oscillations were transient and dissipated quickly.

To better reflect the limitations of our research, we have made significant modifications to the relevant sections in our manuscript.

In line 100, we've added text to explicitly state that we considered delays in our simulations and acknowledged their potential to generate oscillations ("Given the importance of delays in biological network such as BG, we will consider them in the simulations.").

In line 102, we've clarified that our conclusions are based on a scenario without delays ("In this following, we give simple examples of the possibility of oscillation (or not) based on the connectivity characteristics of small networks without delays. Let us start with a network of two nodes.").

Additionally, in line 230, we've included a reference figure supplement 3-2 to highlight the outcomes in terms of oscillations ("EII network only resulted in transient oscillations (Fig. 3, figure supplement 3-1, figure supplement 3-2)").

In lines 234-237, we've added a sentence discussing the role of synaptic delays in generating transient oscillations in cycles with an even number of inhibitory components, referring to figure supplement 3-2 ("In networks with even number of inhibitory connections (e.g. EII, EEE, II), synaptic delays are the sole mechanism for initiating oscillations, however, unless delays are precisely tuned such oscillations will remain transient (see Supplementary figure supplement 3-2)").

Moreover, in response to the reviewer’s suggestion, we have included an additional figure supplement 3-2 to illustrate how cycles with even inhibitory components generate transient oscillations when propagation delays are taken into account. This figure provides a visual representation of the phenomenon and enhances the clarity of our findings.

2) In Figure 3, two motifs (III and EII) are explored to demonstrate the validity of the results across different parameters. Delays don't seem to play a disruptive role in these two cases, but the results seem to be different for other motifs not considered here. Aside from the examples mentioned above, I can imagine how a motif of EEE (i.e. a circle of three excitatory Wilson-Cowan neurons) would display oscillations when delays are included, as the activation would 'circulate' along the ring. However, this EEE motif has an even number of inhibitory units (or perhaps zero is considered an exception, but if so it's not mentioned in the text).

We thank the reviewer for this observation regarding Figure 3. Indeed, the impact of delays may differ for other motifs not considered in our study. For example, as the reviewer has correctly anticipated, a motif of EEE (a circular network of three excitatory Wilson-Cowan neurons) would exhibit oscillations when delays are included, as activation could 'circulate' along the ring.

To address this concern,we have performed new simulations (added as a new supplementary figure supplement 3-2). As illustrated in figure supplement 3-2, oscillations may indeed arise in the EEE motif when delays are introduced. However, these oscillations will eventually dissipate – at least with our settings.

3) Figures 1b, 1c, and 4e display interesting results, but these are absent from the main text. Please include the description of those results. Particularly the case of Figs 1b and 1c seems very relevant to understanding the main results in the context of more complex networks, in which multiple loops with odd and even numbers of inhibitory units would coexist in the network. Does the number of odd-inhibitory loops in a given network affect somehow the power or frequency of the resulting network oscillations? It would be interesting to show this.

Indeed, we did not explain Figs 1b,c and 4e properly. Now we have revised the manuscript in the following way to incorporate these results:

In lines 124-128, we added the following text to introduce the concept: "We can generalize these results to cycles of any size, categorizing them into two types based on the count of their inhibitory connections in one direction (referred to as the odd cycle rule, as illustrated in Fig. 1b). More complex networks can also be decomposed into cycles of size 2…N (where N is number of nodes), and predict the ability of the network to oscillate (as shown in Fig. 1c)" In line 298, we included the following text to highlight the relevant result: "Next, we removed the STN output (equivalent to inhibition of STN), the Proto-D2-Arky subnetwork generated oscillations for weak positive inputs to the D2-SPNs (Fig.4e, bottom)."

How the number of odd/even loops affect the frequency is an interesting question. Intuitively there should be a relation between the two. However, a complete treatment of this question is beyond the scope of the manuscript but we think that in a network with identical node properties, more odd cycles should imply higher oscillation power.

4) The cortico-BG model is focused on how inactivating STN could suppress (or not) beta oscillations, following experimental observations. However, besides mechanisms for extinguishing oscillations, it would be interesting to see if the progressive emergence of pathological beta oscillations could be explained by the modification of some of the nodes in the model (for example, explicitly mimicking the loss of dopaminergic neurons in the substantia nigra). This could be a very interesting additional figure in the main text.

This is an interesting suggestion. Something similar has been already done – e.g. Kumar et al. (2010) showed that progressive increase of inhibition of GPe can lead to oscillations. Similarly Holgado et al. (2008) showed how progressive change in the mutual connectivity between STN and GPe can cause oscillations. More recently, Ortone et al. (PloS Comp. Biol 2023) and Azizpour et al. (2023 Bioarxiv) have also shown the effect of progressive change in individual node properties on oscillations in basal ganglia using numerical simulations. Our work in a way provides the theoretical backing to their work. Therefore, we think it is not necessary to again show these results in our model. Instead we have cited these papers. Lines 392-396

5) I observed some grammatical inconsistencies in the text, some of them are indicated below. I would suggest carefully going through the text to correct those issues or seeking help with editing.

-line 32 "...which can closely capture the neural population dynamics". Which population dynamics? Do the authors refer to general neural dynamics?

-line 33 "long term behavior" -> long-term behavior

-line 68 "given the ionic channel composition" -> "given its ionic channel composition"

We apologize for the grammatical inconsistencies in our manuscript. We have made the necessary corrections to improve the clarity and accuracy of our text.

Reviewer #3 (Recommendations For The Authors):

This manuscript is useful for analytically showing that a cyclic network of threshold-linear neural populations can only oscillate if it has an odd number of inhibitory nodes with strong enough connections. Establishing this result, which holds under rather narrow assumptions, relies on standard tools from dynamical system theory. I find the strength of support for this result to be incomplete for the reasons detailed below:

Although the mathematical arguments used appear to be correct, the manuscript lacks in rigor and clarity. For instance, the main result presented in theorem 2 is stated in a very unclear fashion: aside from the oddity of the number of inhibitory nodes, there are two conditions to check, which determines four cases. This can be explained in a much more straightforward way without introducing four relations in equations 4-7.

We acknowledge the reviewer’s concern regarding the presentation of the main result in Theorem 2.

We would like to emphasize that the introduction of four relations in equations 4-7 was intended to provide a detailed and transparent exposition of the conditions for the main result. While we understand that this approach may appear less straightforward, it allows for a more comprehensive understanding of the underlying logic and the multiple factors influencing the outcomes.

However, we are open to suggestions for more concise and clear ways to express these conditions if the reviewer has specific recommendations or if there are alternative approaches that the reviewer believes would be more effective in conveying the information.

Moreover, equation 3 in that same theorem is clearly wrong.

We sincerely apologize for the typographical error in equation 3 within the same theorem. We thank the reviewer for noticing this. We have revised the text to rectify this mistake. The equation has now been corrected to ensure its accuracy.

The proof of theorem 2 relies on standard linear algebra and can be improved as well: there are typos, approximations, and missing words (see line 664). The rigor of the exposition is also unsatisfactory. For instance, the proof of Lemma 1 ends with the sentence: "Similarly as before, the convergence of the dynamics driven by the left and right terms ends the proof". I don't know what this means.

We thank the reviewer for the comments and suggestions. We have made the necessary adjustments to enhance the rigor and clarity of our mathematical reasoning in the revised manuscript.

In line 644, we have provided clarification for the sentence you found unclear. The revised version now offers a more precise explanation that should help in understanding the proof.

At the same time, the intuitive arguments presented in the main text are vague at best and do not really help grasping the possible generalizability of the results. For instance, I do not understand the message of panel B in Figure 2 and there seems to be no explanation about it in the main text.

The main purpose of Figure 2B is to offer a visual representation of the concept and to serve as an aid for readers who may prefer a graphical illustration over extensive equations. While we understand that the figure may not provide a complete explanation on its own, it is intended to complement the text and mathematical content presented in the main text. In the revised version we have added the explanation of Figure 2B.

Aside from the analytical result, most of the paper consists in simulating networks with distinct inhibitory cyclic structure to validate the theoretical argument. I do not find this approach particularly convincing due to the qualitative nature of the numerical results presented. There is little quantitative analysis of the network structure in relation to the emergence of oscillations. It is also hard to judge whether the examples discussed are cherry picked or truly representative of a large class of dynamics.

The reviewer has a valid concern about numerical simulations and qualitative nature of the results. We would like to provide some perspective on our approach.

In our paper, the primary focus is on the mathematical proof, which rigorously establishes the existence of our results. However, we understand that numerical simulations are valuable for illustrating the applicability of the theoretical framework and providing insights into the practical implications.

If we get into the quantitative description of all the results, the manuscript will become prohibitively long. We acknowledge that there is a balance to be struck between theory and numerical examples in a research paper. We believe that, in conjunction with the mathematical proof, the numerical simulations serve the purpose of illustrating the existence of our results in specific examples. While we cannot provide an exhaustive exploration of all possible network structures, we have chosen representative cases to demonstrate the applicability of our findings. Some of these are already provided in figure supplements S3-1 and S3-3. In the absence of specific suggestions from the reviewer we would like to leave it as is.

Moreover, the authors apply their cycle analysis to real-world networks by considering cycles of inhibitory nodes independently, whereas the same nodes can belong to several cycles. I find it hard to believe that considering these cycles independently should be enough to make predictions about the emergence of oscillations, as these cycles must interact with one another via shared nodes. I do not understand the color coding used to mark distinct cycles in supplementary figures. There is also not enough information to understand figures in the main text. For instance, I do not understand what the grids are representing in panel B, Figure 4.

We have clarified the color coding and added more information to understand the figures. We appreciate the reviewer’s concern about our application of cycle analysis to real-world networks and the clarity of our figures. It is not a matter of belief – we have provided a mathematical proof and complemented that with illustrative examples from real-world networks i.e. cortico-basal ganglia network with both rate-based and spiking neurons. Clearly our results hold.

Regarding the color coding in supplementary figures, we have revised the color scheme to make it more intuitive and informative in caption of figure 4: we use different colors to mark potential oscillators in each motif in BG, and each color means an oscillator from panel a. For more details, see figure supplements 4-1–4-6. The colors now represent distinct cycles more clearly, helping readers better interpret the figures.

Reviewer #1 (Public Review):

Summary & Assessment:

The catalytic core of the eukaryotic decapping complex consists of the decapping enzyme DCP2 and its key activator DCP1. In humans, there are two paralogs of DCP1, DCP1a, and DCP1b, that are known to interact with DCP2 and recruit additional cofactors or coactivators to the decapping complex; however, the mechanisms by which DCP1 activates decapping and the specific roles of DCP1a versus DCP1b, remain poorly defined. In this manuscript, the authors used CRISPR/Cas9-generated DCP1a/b knockout cells to begin to unravel some of the differential roles of human DCP1a and DCP1b in mRNA decapping, gene regulation, and cellular metabolism. While this manuscript presents some new and interesting observations on human DCP1 (e.g. human DCP1a/b KO cells are viable and can be used to investigate DCP1 function; only the EVH1 domain, and not its disordered C-terminal region which recruits many decapping cofactors, is apparently required for efficient decapping in cells; DCP1a and b target different subsets of mRNAs for decay and may regulate different aspects of metabolism), there are several major issues that undercut some of the main conclusions of the paper, and some key claims that are incompletely or inconsistently supported by the presented data.

Strengths & well-supported claims:

• Through in vivo tethering assays in CRISPR/Cas9-generated DCP1a/b knockout cells, the authors show that DCP1 depletion leads to significant defects in decapping and the accumulation of capped, deadenylated mRNA decay intermediates.

• DCP1 truncation experiments reveal that only the EVH1 domain of DCP1 is necessary to rescue decapping defects in DCP1a/b KO cells.

• RNA and protein immunoprecipitation experiments suggest that DCP1 acts as a scaffold to help recruit multiple decapping cofactors to the decapping complex (e.g. EDC3, DDX6, PATL1 PNRC1, and PNRC2), but that none of these cofactors are essential for DCP2-mediated decapping in cells.

• The authors investigated the differential roles of DCP1a and DCP1b in gene regulation through transcriptomic and metabolomic analysis and found that these DCP1 paralogs target different mRNA transcripts for decapping and have different roles in cellular metabolism and their apparent links to human cancers. (Although I will note that I can't comment on the experimental details and/or rigor of the transcriptomic and metabolomic analyses, as these are outside my expertise.)

Weaknesses & incompletely supported claims:

1) A central mechanistic claim of the paper is that "DCP1a can regulate DCP2's cellular decapping activity by enhancing DCP2's affinity to RNA, in addition to bridging the interactions of DCP2 with other decapping factors. This represents a pivotal molecular mechanism by which DCP1a exerts its regulatory control over the mRNA decapping process." Similar versions of this claim are repeated in the abstract and discussion sections. However, this appears to be entirely at odds with the observation from in vitro decapping assays with immunoprecipitated DCP2 that showed DCP1 knockout does not significantly affect the enzymatic activity of DCP2 (Figures 2B-D; I note that there may be a very small change in DCP2 activity shown in panel C, but this may be due to slightly different amounts of immunoprecipitated DCP2 used in the assay, as suggested by panel D). If DCP1 pivotally regulates decapping activity by enhancing RNA binding to DCP2, why is no difference in decapping activity observed in the absence of DCP1? Furthermore, the authors show only weak changes in relative RNA levels immunoprecipitated by DCP2 with versus without DCP1 (~2-3 fold change; consistent with the Valkov 2016 NSMB paper, which shows what looks like only modest changes in RNA binding affinity for yeast Dcp2 +/- Dcp1). Is the argument that only a 2-3 fold change in RNA binding affinity is responsible for the sizable decapping defects and significant accumulation of deadenylated intermediates observed in cells upon Dcp1 depletion? (and if so, why is this the case for in-cell data, but not the immunoprecipitated in vitro data?)

The authors acknowledge this apparent discrepancy between the in vitro DCP2 decapping assays and in-cell decapping data, writing: "this observation could be attributed to the inherent constraints of in vitro assays, which often fall short of faithfully replicating the complexity of the cellular environment where multiple factors and cofactors are at play. To determine the underlying cause, we postulated that the observed cellular decapping defect in DCP1a/b knockout cells might be attributed to DCP1 functioning as a scaffold." This is fair. They next show that DCP1 acts as a scaffold to recruit multiple factors to DCP2 in cells (EDC3, DDX6, PatL1, and PNRC1 and 2). However, while DCP1 is shown to recruit multiple cofactors to DCP2 (consistent with other studies in the decapping field, and primarily through motifs in the Dcp1 C-terminal tail), the authors ultimately show that *none* of these cofactors are actually essential for DCP2-mediated decapping in cells (Figures 3A-F). More specifically, the authors showed that the EVH1 domain was sufficient to rescue decapping defects in DCP1a/b knockout cells, that PNRC1 and PNRC2 were the only cofactors that interact with the EVH1 domain, and finally that shRNA-mediated PNRC1 or PNCR2 knockdown has no effect on in-cell decapping (Figures 3E and F). Therefore, based on the presented data, while DCP1 certainly does act as a scaffold, it doesn't seem to be the case that the major cellular decapping defect observed in DCP1a/b knockout is due to DCP1's ability to recruit specific cofactors to DCP2.

So as far as I can tell, the discrepancy between the in vitro (DCP1 not required) and in-cell (DCP1 required) decapping data, remains entirely unresolved. Therefore, I don't think that the conclusions that DCP1 regulates decapping by (a) changing RNA binding affinity (authors show this doesn't matter in vitro, and that the change in RNA binding affinity is very small) or (b) by bridging interactions of cofactors with DCP2 (authors show all tested cofactors are dispensable for robust in-cell decapping activity), are supported by the evidence presented in the paper (or convincingly supported by previous structural and functional studies of the decapping complex).

2) Related to the RNA binding claims mentioned above, are the differences shown in Figure 3H statistically significant? Why are there no error bars shown for the MBP control? (I understand this was normalized to 1, but presumably, there were 3 biological replicates here that have some spread of values?). The individual data points for each replicate should be displayed for each bar so that readers can better assess the spread of data and the significance of the observed differences. I've listed these points as major because of the key mechanistic claim that DCP1 enhances RNA binding to DCP2 hinges in large part on this data.

3) Also related to point (1) above, the kinetic analysis presented in Figure 2C shows that the large majority of transcript is mostly decapped at the first 5-minute timepoint; it may be that DCP2-mediated decapping activity is actually different in vitro with or without DCP1, but that this is being missed because the reaction is basically done in less than 5 minutes under the conditions being assayed (i.e. these are basically endpoint assays under these conditions). It may be that if kinetics were done under conditions to slow down the reaction somewhat (e.g. lower Dcp2 concentration, lower temperatures), so that more of the kinetic behavior is captured, the apparent discrepancy between in vitro and in-cell data would be much less. Indeed, previous studies have shown that in yeast, Dcp1 strongly activates the catalytic step (kcat) of decapping by ~10-fold, and reduces the KM by only ~2 fold (Floor et al, NSMB 2010). It might be beneficial to use purified proteins here (only a Western blot is used in Figure 2D to show the presence of DCP2 and/or DCP1, but do these complexes have other, and different, components immunoprecipitated along with them?), if possible, to better control reaction conditions.

This contradiction between the in vitro and in-cell decapping data undercuts one of the main mechanistic takeaways from the first half of the paper. This needs to be addressed/resolved with further experiments to better define the role of DCP1-mediated activation, or the mechanistic conclusions significantly changed or removed.

4) The second half of the paper compares the transcriptomic and metabolic profiles of DCP1a versus DCP1b knockouts to reveal that these target a different subset of mRNAs for degradation and have different levels of cellular metabolites. This is a great application of the DCP1a/b KO cells developed in this paper and provides new information about DCP1a vs b function in metazoans, which to my knowledge has not really been explored at all. However, the analysis of DCP1 function/expression levels in human cancer seems superficial and inconclusive: for example, the authors conclude that "...these findings indicate that DCP1a and DCP1b likely have distinct and non-redundant roles in the development and progression of cancer", but what is the evidence for this? I see that DCP1a and b levels vary in different cancer cell types, but is there any evidence that these changes are actually linked to cancer development, progression, or tumorigenesis? If not, these broader conclusions should be removed.

5) The authors used CRISPR-Cas9 to introduce frameshift mutations that result in premature termination codons in DCP1a/b knockout cells (verified by Sanger sequencing). They then use Western blotting with DCP1a or DCP1b antibodies to confirm the absence of DCP1 in the knockout cell lines. However, the DCP1a antibody used in this study (Sigma D5444) is targeted to the C-terminal end of DCP1a. Can the authors conclusively rule out that the CRISPR/Cas-generated mutations do not result in the production of truncated DCP1a that is just unable to be detected by the C-terminally targeted antibody? While it is likely the introduced premature termination codon in the DCP1a gene results in nonsense-mediated decay of the resulting transcript, this outcome is indeed supported by the knockout results showing large defects in cellular decapping which can be rescued by the addition of the EVH1 domain, it would be better to carefully validate the success of the DCP1a knockout and conclusively show no truncated DCP1a is produced by using N-terminally targeted DCP1a antibodies (as was the case for DCP1b).

Some additional minor comments:

• More information would be helpful on the choice of DCP1 truncation boundaries; why was 1-254 chosen as one of the truncations?<br /> • Figure S2D is a pretty important experiment because it suggests that the observed deadenylated intermediates are in fact still capped; can a positive control be added to these experiments to show that removal of cap results in rapid terminator-mediated degradation?

Author Response

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

eLife assessment

The work is a useful contribution towards understanding the role of archaeal and plant D-aminoacyl-tRNA deacylase 2 (DTD2) in deacylation and detoxification of D-Tyr-tRNATyr modified by various aldehydes produced as metabolic byproducts in plants. It integrates convincing results from both in vitro and in vivo experiments to address the long-standing puzzle of why plants outperform bacteria in handling reactive aldehydes and suggests a new strategy for stress-tolerant crops. The impact of the paper is limited by the fact that only one modified D-aminoacyl tRNA was examined, in lack of evidence that plant eEF1A mimics EF-Tu in protecting L-aminoacyl tRNAs from modification, and in failure to measure accumulation of toxic D-aminoacyl tRNAs or impairment of translation in plant cells lacking DTD2.

We have now addressed all the drawbacks as follows:

‘only one modified D-aminoacyl tRNA was examined’

We wish to clarify that only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009) and D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007) and it also recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure: 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

Author response image 1.

DTD2 recycles multiple D-aa-tRNAs with different side chain chemistry and size. Growth of wildtype (WT), dtd null strain (∆dtd), and Pyrococcus horikoshii DTD2 (PhoDTD2) complemented ∆dtd strains of E. coli K12 cells with 500 µM IPTG along with A) no D-amino acids, B) 2.5 mM D-tyrosine, C) 30 mM D-aspartate and D) 5 mM D-tryptophan.

‘lack of evidence that plant eEF1A mimics EF-Tu in protecting L-aminoacyl tRNAs from modification’

To understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). This suggests that eEF-1A will mimic EF-Tu in protecting L-aa-tRNAs from reactive aldehydes. Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

Author response image 2.

Elongation factor enantio-selects L-aa-tRNAs through D-chiral rejection mechanism. A) Surface representation showing the cocrystal structure of EF-Tu with L-Phe-tRNAPhe. Zoomed-in image showing the binding of L-phenylalanine with side chain projected outside of binding site of EF-Tu (PDB id: 1TTT). B) Zoomed-in image of amino acid binding site of EF-Tu bound with L-phenylalanine showing the selection of amino group of amino acid through main chain atoms (PDB id: 1TTT). C) Modelling of D-phenylalanine in the amino acid binding site of EF-Tu shows severe clashes with main chain atoms of EF-Tu. Modelling of smallest chiral amino acid, alanine, in the amino acid binding site of EF-Tu shows D) no clashes with L-alanine and E) clashes with D-alanine. F) Modelling of D-alanine in the amino acid binding site of eEF-1A shows clashes with main chain atoms. (*Represents modelled molecule). G) Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and animals) showing conserved amino acid binding site residues. (Key residues are marked with red star).

Author response image 3.

Elongation factor protects L-aa-tRNAs from aldehyde modification. A) Cartoon representation showing the cocrystal structure of EF-Tu with L-Phe-tRNAPhe (PDB id: 1TTT). B) Zoomed-in image of amino acid binding site of EF-Tu bound with L-phenylalanine (PDB id: 1TTT). C) Cartoon representation showing the cryoEM structure of eEF-1A with tRNAPhe (PDB id: 5LZS). D) Image showing the overlap of EF-Tu:L-Phe-tRNAPhe crystal structure and eEF-1A:tRNAPhe cryoEM structure (r.m.s.d. of 1.44 Å over 292 Cα atoms). E) Zoomed-in image of amino acid binding site of eEF-1A with modelled L-alanine (PDB id: 5ZLS). (*Modelled) F) Overlap showing the amino acid binding site residues of EF-Tu and eEF-1A. (EF-Tu residues are marked in black and eEF-1A residues are marked in red).

‘failure to measure accumulation of toxic D-aminoacyl tRNAs or impairment of translation in plant cells lacking DTD2’

We agree that measuring the accumulation of D-aa-tRNA adducts from plant cells lacking DTD2 is important. We tried to characterise the same with dtd2 mutant plants extensively through Northern blotting as well as mass spectrometry. However, due to the lack of information about the tissue getting affected (root or shoot), identity of aa-tRNA as well as location of aa-tRNA (cytosol or organellar), we are so far unsuccessful in identifying them from plants. Efforts are still underway to identify them from plant system lacking DTD2. However, we have used a bacterial surrogate system, E. coli, as used earlier in Mazeed M. et al., Science Advances, 2021 to show the accumulation of D-aa-tRNA adducts in the absence of dtd. We could identify the accumulation of both formaldehyde and MG modified D-aa-tRNA adducts via mass spectrometry (Figure: 4). These results are now included in the revised manuscript (in line no: 190-197 and Figure: S5).

Author response image 4.

Loss of DTD results in accumulation of modified D-aminoacyl adducts on tRNAs in E. coli. Mass spectrometry analysis showing the accumulation of aldehyde modified D-Tyr-tRNATyr in A) Δdtd E. coli, B) formaldehyde and D-tyrosine treated Δdtd E. coli, and C) MG and D-tyrosine treated Δdtd E. coli. ESI-MS based tandem fragmentation analysis for unmodified and aldehyde modified D-Tyr-tRNATyr in D) Δdtd E. coli, E) and F) formaldehyde and D-tyrosine treated Δdtd E. coli, G) and H) MG and D-tyrosine treated Δdtd E. coli.

Response to Public Reviews:

We are grateful for the reviewers’ positive feedback and their comments and suggestions on this manuscript. Reviewer 1 has indicated two weaknesses and Reviewer 2 has none. We have now addressed all the concerns of the Reviewers.

Reviewer #1 (Public Review):

Summary:

This work is an extension of the authors' earlier work published in Sci Adv in 2001, wherein the authors showed that DTD2 deacylates N-ethyl-D-aminoacyl-tRNAs arising from acetaldehyde toxicity. The authors in this study, investigate the role of archaeal/plant DTD2 in the deacylation/detoxification of D-Tyr-tRNATyr modified by multiple other aldehydes and methylglyoxal (produced by plants). Importantly, the authors take their biochemical observations to plants, to show that deletion of DTD2 gene from a model plant (Arabidopsis thaliana) makes them sensitive to the aldehyde supplementation in the media especially in the presence of D-Tyr. These conclusions are further supported by the observation that the model plant shows increased tolerance to the aldehyde stress when DTD2 is overproduced from the CaMV 35S promoter. The authors propose a model for the role of DTD2 in the evolution of land plants. Finally, the authors suggest that the transgenic crops carrying DTD2 may offer a strategy for stress-tolerant crop development. Overall, the authors present a convincing story, and the data are supportive of the central theme of the story.

We are happy that reviewer found our work convincing and would like to thank the reviewer for finding our data supportive to the central theme of the manuscript.

Strengths:

Data are novel and they provide a new perspective on the role of DTD2, and propose possible use of the DTD2 lines in crop improvement.

We are happy for this positive comment on the manuscript.

Weaknesses:

(a) Data obtained from a single aminoacyl-tRNA (D-Tyr-tRNATyr) have been generalized to imply that what is relevant to this model substrate is true for all other D-aa-tRNAs (term modified aa-tRNAs has been used synonymously with the modified Tyr-tRNATyr). This is not a risk-free extrapolation. For example, the authors see that DTD2 removes modified D-Tyr from tRNATyr in a chain-length dependent manner of the modifier. Why do the authors believe that the length of the amino acid side chain will not matter in the activity of DTD2?

We thank the reviewer for bringing up this important point. As mentioned above, we wish to clarify that only half of the aminoacyl-tRNA synthetases are known to charge D-amino acids and only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009). D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007). Moreover, we have previously shown that it recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

(b) While the use of EFTu supports that the ternary complex formation by the elongation factor can resist modifications of L-Tyr-tRNATyr by the aldehydes or other agents, in the context of the present work on the role of DTD2 in plants, one would want to see the data using eEF1alpha. This is particularly relevant because there are likely to be differences in the way EFTu and eEF1alpha may protect aminoacyl-tRNAs (for example see description in the latter half of the article by Wolfson and Knight 2005, FEBS Letters 579, 3467-3472).

We thank the reviewer for bringing up this important point. As mentioned above, to understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

Reviewer #2 (Public Review):

In bacteria and mammals, metabolically generated aldehydes become toxic at high concentrations because they irreversibly modify the free amino group of various essential biological macromolecules. However, these aldehydes can be present in extremely high amounts in archaea and plants without causing major toxic side effects. This fact suggests that archaea and plants have evolved specialized mechanisms to prevent the harmful effects of aldehyde accumulation.

In this study, the authors show that the plant enzyme DTD2, originating from archaea, functions as a D-aminoacyl-tRNA deacylase. This enzyme effectively removes stable D-aminoacyl adducts from tRNAs, enabling these molecules to be recycled for translation. Furthermore, they demonstrate that DTD2 serves as a broad detoxifier for various aldehydes in vivo, extending its function beyond acetaldehyde, as previously believed. Notably, the absence of DTD2 makes plants more susceptible to reactive aldehydes, while its overexpression offers protection against them. These findings underscore the physiological significance of this enzyme.

We thank the reviewer for the positive comments the manuscript.

Response to recommendation to authors:

Reviewer #1 (Recommendations For The Authors):

I enjoyed reading the manuscript entitled, "Archaeal origin translation proofreader imparts multi aldehyde stress tolerance to land plants" from the Sankaranarayanan lab. This work is an extension of their earlier work published in Sci Adv in 2001, wherein they showed that DTD2 deacylates N-ethyl-D-aminoacyl-tRNAs arising from acetaldehyde toxicity. Now, the authors of this study (Kumar et al.) investigate the role of archaeal/plant DTD2 in the deacylation/detoxification of D-Tyr-tRNATyr modified by multiple other aldehydes and methylglyoxal (which are produced during metabolic reactions in plants). Importantly, the authors take their biochemical observations to plants, to show that deletion of DTD2 gene from a model plant (Arabidopsis thaliana) makes them sensitive to the aldehyde supplementation in the media especially in the presence of D-Tyr. These conclusions are further supported by the observation that the model plant shows increased tolerance to the aldehyde stress when DTD2 is overproduced from the CaMV 35S promoter. The authors propose a model for the role of DTD2 in the evolution of land plants. Finally, the authors suggest that the transgenic crops carrying DTD2 may offer a strategy for stress-tolerant crop development. Overall, the authors present a convincing story, and the data are supportive of the central theme of the story.

We are happy that reviewer enjoyed our manuscript and found our work convincing. We would also like to thank reviewer for finding our data supportive to the central theme of the manuscript.

I have the following observations that require the authors' attention.

1) The title of the manuscript will be more appropriate if revised to, "Archaeal origin translation proofreader, DTD2, imparts multialdehyde stress tolerance to land plants".

Both the reviewer’s suggested to change the title. We have now changed the title based on reviewer 2 suggestion.

2) Abstract (line 19): change, "physiologically abundantly produced" to "physiologically produced".

As per the reviewer’s suggestion, we have now changed it to "physiologically produced".

3) Introduction (line 50): delete, 'extremely'.

We have removed the word 'extremely' from the Introduction.

4) Line 79: change, "can be utilized" to "may be explored".

We have changed "can be utilized" to "may be explored" as suggested by the reviewers.

5) Results in general:

(a) Data obtained from a single aminoacyl-tRNA (D-Tyr-tRNATyr) have been generalized to imply that what is relevant to this model substrate is true for all other D-aa-tRNAs (term modified aa-tRNAs has been used synonymously with the modified D-Tyr-tRNATyr). This is a risky extrapolation. For example, the authors see that DTD2 removes modified D-Tyr from tRNATyr in a chain-length dependent manner of the modifier. Why do the authors believe that the length of the amino acid side chain will not matter in the activity of DTD2?

We thank the reviewer for bringing up this important point. As mentioned above, we wish to clarify that only half of the aminoacyl-tRNA synthetases are known to charge D-amino acids and only D-Leu (Yeast), D-Asp (Bacteria, Yeast), D-Tyr (Bacteria, Cyanobacteria, Yeast) and D-Trp (Bacteria) show toxicity in vivo in the absence of known DTD (Soutourina J. et al., JBC, 2000; Soutourina O. et al., JBC, 2004; Wydau S. et al., JBC, 2009). D-Tyr-tRNATyr is used as a model substrate to test the DTD activity in the field because of the conserved toxicity of D-Tyr in various organisms. DTD2 has been shown to recycle D-Asp-tRNAAsp and D-Tyr-tRNATyr with the same efficiency both in vitro and in vivo (Wydau S. et al., NAR, 2007). Moreover, we have previously shown that it recycles acetaldehyde-modified D-Phe-tRNAPhe and D-Tyr-tRNATyr in vitro as shown in our earlier work (Mazeed M. et al., Science Advances, 2021). We have earlier shown that DTD1, another conserved chiral proofreader across bacteria and eukaryotes, acts via a side chain independent mechanism (Ahmad S. et al., eLife, 2013). To check the biochemical activity of DTD2 on D-Trp-tRNATrp, we have now done the D-Trp, D-Tyr and D-Asp toxicity rescue experiments by expressing the archaeal DTD2 in dtd null E. coli cells. We found that DTD2 could rescue the D-Trp toxicity with equal efficiency like D-Tyr and D-Asp (Figure 1). Considering the action on multiple side chains with different chemistry and size, it can be proposed with reasonable confidence that DTD2 also operates based on a side chain independent manner.

(b) Interestingly, the authors do suggest (in the Materials and Methods section) that the experiments were performed with Phe-tRNAPhe as well as Ala-tRNAAla. If what is stated in Materials and Methods is correct, these data should be included to generalize the observations.

We regret for the confusing statement. We wish to clarify that L- and D-Tyr-tRNATyr were used for checking the TLC-based aldehyde modification, EF-Tu based protection assays and deacylation assays, D-Phe-tRNAPhe was used to characterise aldehyde-based modification by mass spectrometry and L-Ala-tRNAAla was used to check the modification propensity of multiple aldehydes. We used multiple aa-tRNAs to emphasize that aldehyde-based modifications are aspecific towards the identity of aa-tRNAs. All the data obtained with respective aa-tRNAs are included in manuscript.

(c) While the use of EFTu supports that the ternary complex formation by the elongation factor can resist modifications of L-Tyr-tRNATyr by the aldehydes or other agents, in the context of the present work on the role of DTD2 in plants, one would want to see the data using eEF1alpha. This is particularly relevant because there are likely to be differences in the way EFTu and eEF1alpha may protect aminoacyl-tRNAs (for example see description in the latter half of the article by Wolfson and Knight 2005, FEBS Letters 579, 3467-3472).

We thank the reviewer for bringing up this important point. As mentioned above, to understand the role of plant eEF1A in protecting L-aa-tRNAs from aldehyde modification, we have done a thorough sequence and structural analysis. We analysed the aa-tRNA bound elongation factor structure from bacteria (PDB ids: 1TTT) and found that the side chain of amino acid in the amino acid binding site of EF-Tu is projected outside (Figure: 2A; 3A). In addition, the amino group of amino acid is tightly selected by the main chain atoms of elongation factor thereby lacking a space for aldehydes to enter and then modify the L-aa-tRNAs and Gly-tRNAs (Figure: 2B; 3B). Modelling of D-amino acid (D-phenylalanine and smallest chiral amino acid, D-alanine) in the same site shows serious clashes with main chain atoms of EF-Tu, indicating D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2C-E). Next, we superimposed the tRNA bound mammalian eEF-1A cryoEM structure (PDB id: 5LZS) with bacterial structure to understand the structural differences in terms of tRNA binding and found that elongation factor binds tRNA in a similar way (Figure: 3C-D). Modelling of D-alanine in the amino acid binding site of eEF-1A shows serious clashes with main chain atoms, indicating a general theme of D-chiral rejection during aa-tRNA binding by elongation factor (Figure: 2F; 3E). Structure-based sequence alignment of elongation factor from bacteria, archaea and eukaryotes (both plants and mammals) shows a strict conservation of amino acid binding site (Figure: 2G). Minor differences near the amino acid side chain binding site (as indicated in Wolfson and Knight, FEBS Letters, 2005) might induce the amino acid specific binding differences (Figure: 3F). However, those changes will have no influence when the D-chiral amino acid enters the pocket, as the whole side chain would clash with the active site. We have now included this sequence and structural conservation analysis in our revised manuscript (in text: line no 107-129; Figure: 2 and S2). Overall, our structural analysis suggests a conserved mode of aa-tRNA selection by elongation factor across life forms and therefore, our biochemical results with bacterial elongation factor Tu (EF-Tu) reflect the protective role of elongation factor in general across species.

6) Results (line 89): Figure: 1C-G (not B-G).

As correctly pointed out by the reviewer(s), we have changed it to Figure: 1C-G.

7) Results (line 91): Figure: S1B-G (not C-G).

We wish to clarify that this is correct.

8) Line 97: change, "propionaldehyde" to "propionaldehyde (Figure: 1H)".

As per the reviewer’s suggestion, we have now changed, "propionaldehyde" to "propionaldehyde (Figure: 1H)".

9) Line 124: The statement, "DTD2 cleaved all modified D-aa-tRNAs at 50 pM to 500 nM range (Figure: 2A_D)" is not consistent with the data presented. For example, Figure 2D does not show any significant cleavage. Figure S2A-B also does not show cleavage.

We thank the reviewers for pointing this out. We have changed the sentence to “DTD2 cleaved majority of aldehyde modified D-aa-tRNAs at 50 pM to 500 nM range".

10) Line 131: Cleavage observed in Fig. S2E is inconsistent with the generalized statement on DTD1.

We wish to clarify that the minimal activity seen in Fig. S2E is inconsistent with the general trend of DTD1’s biochemical activity seen on modified D-aa-tRNAs. In addition, we have earlier shown that D-aa-tRNA fits snugly in the active site of DTD1 (Ahmad S. et al., eLife, 2013) whereas the modified D-aa-tRNA cannot bind due to the space constrains in the active site of DTD1 (Mazeed M. et al., Science Advances, 2021). Therefore, this minimal activity could be a result of technical error during this biochemical experiment and could be considered as no activity.

11) Lines 129-133: Citations of many figure panels particularly in the supplementary figures are inconsistent with generalized statements. This section requires a major rewrite or rearrangement of the figure panels (in case the statements are correct).

We thank the reviewers for bringing forth this point and we have accordingly modified the statement into “DTD2 from archaea recycled short chain aldehyde-modified D-aa-tRNA adducts as expected (Figure: 3E-G) and, like DTD2 from plants, it did not act on aldehyde-modified D-aa-tRNAs longer than three chains (Figure: 3H; S3C-D; S4G-L)”.

12) Line 142: I don't believe one can call PTH a proofreader. Its job is to recycle tRNAs from peptidyl-tRNAs.

We thank the reviewers for pointing out this very important point. This is now corrected.

13). Line 145: change, "DTD2 can exert its protection for" to "DTD2 may exert protection from".

As per the reviewer’s suggestion, we have now changed"DTD2 can exert its protection for" to "DTD2 may exert protection from".

14) Line 148: change, "a homozygous line (Figure: 3A) and checked for" to "homozygous lines (Figure: 3A) and checked them for".

As per the reviewer’s suggestion, we have now changed, "a homozygous line (Figure: 3A) and checked for" to "homozygous lines (Figure: 3A) and checked them for".

15) Line 148: Change, the sentence beginning with dtd2 as follows. Similar to earlier results30-32, dtd2-/- (dtd2 hereafter) plants were susceptible to ethanol (Figure: S4A) confirming the non-functionality DTD2 gene in dtd2 plants.

As per the reviewer’s suggestion, we have now changed the sentence accordingly.

16) Line 161: change, "linked" to "associated".

As per the reviewer’s suggestion, we have now changed "linked" to "associated".

17) Lines 173-176: It would be interesting to know how well the DTD2 OE lines do in comparison to the other known transgenic lines developed with, for example, ADH, ALDH, or AOX lines. Any ideas would help appreciate the observation with DTD2 OE lines!

We greatly appreciate the reviewer’s suggestion. We have not done any comparison experiment with any transgenic lines so far. However, it can be potentially done in further studies with DTD2 OE lines.

18) Line 194: change, "necessary" with "present".

As per the reviewer’s suggestion, we have now changed "necessary" with "present".

19) Line 210: what is meant by 'huge'? Would 'significant' sound better?

As per the reviewer’s suggestion, we have now changed "huge" with "significant".

20) Lines 239-243: This needs to be rephrased. Isn't alpha carbonyl of the carboxyl group that makes ester bond with the -CCA end of the tRNA required for DTD2 activity as well? Are you referring to the carbonyl group in the moiety that modifies the alpha-amino group? Please clarify. The cited reference (no. 64) of Atherly does not talk about it.

We regret for the confusing statement. To clarify, we were referencing to the carbonyl carbon of the modification post amino group of the amino acid in aa-tRNAs (Figure: 5). We have now included a figure (Figure: S4Q of revised manuscript) to show the comparison of the carbonyl group for the better clarity. The cited reference Atherly A. G., Nature, 1978 shows the activity of PTH on peptidyl-tRNAs and peptidyl-tRNAs possess carbonyl carbon at alpha position post amino group of amino acid in L-aa-tRNAs.

Author response image 5.

Figure showing the difference in the position of carbonyl carbon in acetonyl and acetyl modification on aa-tRNAs.

21) Line 261: thrive (not thrives).

As per the reviewer’s suggestion, we have now changed it to thrive.

22) In Fig3A: second last lane, it should be dtd-/-:: AtDTDH150A (not dtd-/-:: AtDTDH150A).

We thank the reviewers for pointing out this, we have corrected it.

23). Materials and methods: Please clarify which experiments used tRNAPhe, tRNAAla, PheRS, etc. Also, please carefully check all other details provided in this section.

As per the reviewer’s suggestion, we would like to provide a table below explaining the use of different substrates as well as enzymes in our experiments.

Author response table 1.

24) Figure legends (many places): p values higher than 0.05 (not less than) are denoted as ns.

We thank the reviewers for pointing out this. We have corrected it.

Reviewer #2 (Recommendations For The Authors):

I have only minor comments for the authors:

Title: I would replace "Archeal origin translation proofreader" with " A translation proofreader of archeal origin"

As per the reviewer’s suggestion, we have now changed the title.

Abstract: This section could benefit from some rewriting. For instance, at the outset, the initial logical connection between the first and second sentences of the abstract is somewhat unclear. At the very least, I would suggest swapping their order to enhance the narrative flow. Later in the text, the term "chiral proofreading systems" is introduced; however, it is only in a subsequent sentence that these systems are explained to be responsible for removing stable D-aminoacyl adducts from tRNA. Providing an immediate explanation of these systems would enhance the reader's comprehension. The authors switch from the past participle tense to the present tense towards the end of the text. I would recommend that they choose one tense for consistency. In the final sentence, I would suggest toning down the statement and replacing "can be used" with "could be explored." (https://www.nature.com/articles/d41586-023-02895-w). The same comment applies to the introduction, line 79.

As per the reviewer’s suggestion, we have now changed the abstract appropriately.

General note: Conventionally, the use of italics is reserved for the specific species "Arabidopsis thaliana," while the broader genus "Arabidopsis" is not italicized.

We acknowledge the reviewer for this pertinent suggestion. This is now corrected in revised version of our manuscript.

General note: I would advise the authors against employing bold characters in conjunction with colors in the figures.

We thank the reviewer for this suggestion. We have now changed it appropriately in revised version of our manuscript.

Figure 1A: I recommend including the concentrations of the various aldehydes used in the experiment within the figure legend. While this information is available in the materials and methods section, it would be beneficial to have it readily accessible when analyzing the figure.

As per the reviewer’s suggestion, we have now included the concentrations in figure legend.

Figure 1I, J: some error bars are invisible.

We thank the reviewers for pointing out this, we have corrected it.

Figure 2M: The table could be simplified by removing aldehydes for which it was not feasible to demonstrate activity. The letter "M" within the cell labeled "aldehydes" appears to be a typographical error, presumably indicating the figure panel.

As per the reviewer’s suggestion, we have now changed this appropriately.

Figure 3: For consistency with the other panels in the figure, I recommend including an additional panel to display the graph depicting the impact of MG on germination.

As per the reviewer’s suggestion, we have now changed this appropriately.

Figure 4: Considering that only one plant is presented, it would be beneficial to visualize the data distribution for the other plants used in this experiment, similar to what the authors have done in panel A of the same figure.

We thank the reviewer for bringing up this point. We wish to clarify that we have done experiment with multiple plants. However, for the sake of clarity, we have included the representative images. Moreover, we have included the quantitative data for multiple plants in Figure 3C-G.

Figure 5E: The authors may consider presenting a chronological order of events as they believe they occurred during evolution.

We thank the reviewer for the suggestion. However, it is very difficult to pinpoint the chronology of the events. Aldehydes are lethal for systems due to their hyper reactivity and systems would require immediate solutions to survive. Therefore, we think that both problem (toxic aldehyde production) and its solution (expansion of aldehyde metabolising repertoire and recruitment of archaeal DTD2) might have appeared simultaneously.

Figure 6: The model appears somewhat crowded, which may affect its clarity and ease of interpretation. The authors might also consider dividing the legend sentence into two separate sentences for better readability.

As per the reviewer’s suggestion, we have now changed this appropriately.

Line 149: I recommend explicitly stating that ethanol metabolism produces acetaldehyde. This clarification will help the general reader immediately understand why DTD2 mutant plants are sensitive to ethanol.

As per the reviewer’s suggestion, we have now changed this appropriately.

Line 289: there is a typographical error, "promotor" instead of the correct term "promoter.".

We thank the referee for pointing out this, we have now corrected it.

Figure S5: The root morphology of DTD2 OE plants appears to exhibit some differences compared to the WT, even in the absence of a high concentration of aldehydes. It would be valuable if the authors could comment on these observed differences unless they have already done so, and I may have overlooked it.

We thank the referee for pointing out this. We do see minor differences in root morphology, but they are more pronounced with aldehyde treatments. The reason for this phenotype remains elusive and we are trying to understand the role of DTD2 in root development in detail in further studies.

Some Curiosity Questions (not mandatory for manuscript acceptance):

1) Do DTD2 OE plants display an earlier flowering phenotype than wild-type Col-0?

We have not done detailed phenotyping of DTD2 OE plants. However, our preliminary observations suggest no differences in flowering pattern as compared to wild-type Col-0.

2) What is the current understanding of the endogenous regulation of DTD2?

We have not done detailed analysis to understand the endogenous regulation of DTD2.

3) Could the protective phenotype of DTD2 OE plants in the presence of aldehydes be attributed to additional functions of this enzyme beyond the removal of stable D-aminoacyl adducts from tRNAs?

Based on the available evidence regarding the biochemical activity and in vivo phenotypes of DTD2, it appears that removal of stable D-aminoacyl adducts from tRNA is key for the protective phenotype of DTD2 OE.

A Suggestion for Future Research (not required for manuscript acceptance):

The authors could explore the possibility of overexpressing DTD2 in pyruvate decarboxylase transgenic plants and assess whether this strategy enhances flood tolerance without incurring a growth penalty under normal growth conditions.

We thank the referee for this interesting suggestion for future research. We will surely keep this in mind while exploring the flood tolerance potential of DTD2 OE plants.

Many academics and modern people may think this way, but it is far from a "bedrock value".

In many indigenous cultures knowledge was carefully sectioned and cordoned off.

And as we know that knowledge itself is power (ipsa scientia potestas est - Francis Bacon) many people have frequently cordoned off access to information.

Author Response

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

eLife assessment

This valuable study advances our understanding of the forces that shape the genomic landscape of transposable elements. By exploiting both long-read sequencing of mutation accumulation lines and in vivo transposition assays, the authors offer compelling evidence that structural variation rather than transposition largely shapes transposable element copy number evolution in budding yeast. The work will be of interest to the transposable element and genome evolution communities.

Public Reviews:

Reviewer #1 (Public Review):

Henault et al build on their own previous work investigating the longstanding hypothesis that hybridization between divergent populations can activate transposable element mobilization (transposition). Previously they created crosses of increasing sequence divergence, using both intra- and inter-species hybrids, and passaged them neutrally for hundreds of generations. Their previous work showed that neither hybrids isolated from natural environments nor hybrids from their mutation accumulation lines showed consistent evidence of increased transposable element content. Here, they sequence and assemble long-read genomes of 127 of their mutation-accumulation lines and annotate all existing and de novo transposable elements. They find only a handful of de novo transposition events, and instead demonstrate that structural variation (ploidy, aneuploidy, loss of heterozygosity) plays a much larger role in the transposable element load in a given strain. They then created transposable element reporter constructs using two different Ty1 elements from S. paradoxus lineages and measured the transposition rate in a number of intraspecific crosses. They demonstrate that the transposition rate is dependent on both the Ty1 sequence and the copy number of genomic transposable elements, the latter of which is consistent with what has been observed in the literature on transposable element copy number control in Saccharomyces. To my knowledge, others have not directly tested the effect of Ty1 sequence itself (have not created diverse Ty1 reporter constructs), and so this is an interesting advance. Finally, the authors show that mitotype has a moderate effect on transposition rate, which is an intriguing finding that will be interesting to explore in future work.

This study represents a large effort to investigate how genetic background can influence transposable element load and transposition rate. The long read sequencing, assembly, and annotation, and the creation of these reporter constructs are non-trivial. Their results are straightforward, well supported, and a nice addition to the literature.

The authors state that the results from their current work support results taken from their previous study using short-read sequencing data of the same lines. The argument that follows is whether the authors gained anything novel from long-read sequencing. I would like to see the authors make a stronger argument for why this new work was necessary, and a more detailed view of similarities or differences from their previous study (when should others choose to do long read vs. short read of evolved lines?).

We thank the reviewer for the suggestion. While we initially aimed to justify the relevance and novelty of the current in relation to our previous study, we understand that this justification may not have been strong enough.

In the second paragraph of the introduction, we explain how the multidimensional nature of TE load makes it more complex to characterize that simply reporting the abundance of a given TE family in a given genome. We added the following concluding sentence to further emphasize the importance of long reads in TE-focused genome inference:

“As such, ongoing technological and computational advances in genome inference, including long-read sequencing, will certainly be key to getting a detailed understanding of the dynamics of TEs and the underpinning evolutionary forces.”

In the penultimate introductory paragraph, we summarize our previous work from 2020 and highlight that the evolution of Ty contents in MA lines was inferred from aggregate measures of genomic abundance of TE families using short reads. We then make the point that combinations of multiple SVs could affect the landscape of TEs in ways that are not reflected by crude short-read measures. We added the following sentence to further emphasize this point and contrast it with the necessity of using more powerful methodologies for genome resolution:

“Under this scenario, measuring Ty family abundance would yield no significant net change, and the dissection of the underlying SVs using short reads could often be challenging.”

Relatedly, the authors should report the rates of structural variants that they observe. How are these results similar/different from other mutation-accumulation work in S. cerevisiae?

Since this work does not attempt to provide an exhaustive report of all the SVs in the MA lines, but rather focus on attributing an SV type to individual loci occupied by TEs, we cannot include these estimates, excepted for de novo transposition itself (see below). We added the following sentence to the Results section on the classification of Ty loci by SV types:

“We note that the current methodology does not aim at providing an exhaustive quantification of all SVs in the MA lines, as previously done for some SV types (Marsit et al., 2021), but focuses solely on loci containing Ty elements.”

We added estimates of the average retrotransposition rate in the MA experiment based on the number of de novo insertions detected in the MA lines genomes.

Figure 4:

“The average retrotransposition rates estimated from the counts of de novo insertions (per line per generation per element) are the following: CC1, 1.0✕10-5; CC2, 4.9✕10-6; CC3, 7.6✕10-6; BB1, 1.5✕10-5; BC2, 1.7✕10-5; BA1, 6.5✕10-6; BA2, 2.2✕10-5; BSc1, 3.6✕10-5.”

We added the following paragraph in the Discussion section to specifically discuss these estimates in relation to the in vivo measurements.

“We note that while the CC crosses tend to have the lowest retrotransposition rates as estimated from the de novo insertions (~1✕10-5 per line per generation per element; Figure 4), these values are several orders of magnitude higher than the in vivo measures in SpC backgrounds. The discrepancy between these estimates could be due to uncharacterized biases inherent to each method. They could also be linked to differences between the parental genotypes used to generate the MA crosses and the fluctuation assays. One major difference is the use of ade2 genotypes in the MA parents, a strategy that was initially adopted to provide a marker for the loss of mitochondrial respiration (Joseph and Hall, 2004; Lynch et al., 2008). It has been shown that the induction of adenine starvation through minimal adenine concentration in the medium and deletion of ADE2, which inactivates the adenine de novo biosynthesis pathway, increases Ty1 transcript levels (Todeschini et al., 2005), resulting in higher transposition rates. Rich complex medium like the one that was used for the MA experiment (YPD) can exhibit substantial variation in adenine concentration (VanDusen et al., 1997), and adenine can quickly become the limiting nutrient for ade2 strains (Kokina et al., 2014). Thus, we cannot exclude that the choice of initial ade2 genotypes could have inflated the transposition rates in the MA experiment.”

Since the authors show a small, but consistent influence of mitotype on transposition rates, adding further evidence for the role of mtDNA in regulating transposition, I'm curious what the transposition rate of a p0 strain is. I think including these results could make this observation more compelling.

We agree that measuring in vivo transposition rates in ρ0 backgrounds would be an interesting avenue. However, there is a large distinction between having non-functional mitochondrial respiration in ρ0 strains and inheriting diverse functional mtDNA haplotypes. The effects we show are all linked to the reciprocal inheritance of intact mtDNAs, producing ρ+ strains that are all respiration-competent, as shown by our growth confirmations on non-fermentable carbon sources for all the diploid backgrounds generated. While potentially interesting, adding transposition rates measures for the ρ0 backgrounds seems hard to justify in the context of our results.

Reviewer #2 (Public Review):

This is an interesting follow-up study that uses long-read sequencing to examine previously constructed mutation accumulation lines between wild populations of S. cerevisiae and S. paradoxus. They also complement this work with reporter assays in hybrid backgrounds. The authors are attempting to test the hypothesis that hybridization leads to genome shock and unrestrained transposition. The paper largely confirms previous results (suggesting hybridization does not increase transposition) that are well cited and discussed in the paper, both from this group and from the Smukowski Heil/Dunham group but extends them to a new set of species/hybrids and with some additional resolution via the long read sequencing. The paper is well written and clear and I have no serious complaints.

In the abstract, the authors make three primary claims:

Structural variation plays a strong role in TE load.

Transposition plays only a minor role in shaping the TE landscape in MA lines.

Transposition rates are not increased by hybridization but are affected by genotype-specific factors.

I found all three claims supported, albeit with some minor questions below:

Structural variation plays a strong role in TE load.

Convinced of this result. However:

Line 185-187/Figure 3C: I'm curious given that the changes in Ty count are so often linked to changes in gross DNA sequence whether the count per total DNA sequence is actually changing on average in these genomes. Ie., does hybridization tend to increase TE count via CNV or does hybridization tend to increase DNA content in the MA lines and TEs come along for the ride?

The Ty content definitely “rides along” with the rest of the genome that is affected by retrotransposition-unrelated SVs. To further highlight this point, we added a panel (E) to Figure 3 in which we correlate the net Ty copy number change (same as panel D, formerly C) to the corresponding genome size, which reflects the amount of DNA lost/gained by all SV types. We added the following to the results section:

“The distributions of net Ty CN change per MA line showed that most crosses had significant gains (Figure 3D), suggesting that Ty load can often increase as a result of random genetic drift. Some (but not all) of these crosses also exhibited significant increases in genome size after evolution (Supplemental Figure S7A). The net Ty CN changes per MA line subgenome were globally correlated to the corresponding changes in subgenome size (Figure 3E). Even after excluding polyploid lines (which have the largest changes in both Ty CN and genome size), we found a significant relationship between the two variables (mixed linear model with random intercepts and slopes for MA crosses, P-value=3.71✕10-9; Supplemental Figure S7B), indicating that SVs affecting large portions of the genome have a substantial impact on the Ty landscape.”

One question about ploidy (lines 175-177):

Both aneuploidy and triploidy seem easy to call from this data. A 3:1 tetraploidy as well. However, in Figure 2B there are tetraploids that are around the 1:1 line. How are the authors calling ploidy for these strains? This was not clear to me from the text.

This detail was indeed missing from the manuscript. The ploidy level of all MA lines was previously measured by DNA staining and flow cytometry, and the ploidy level of the subgenomes of each polyploid MA line was previously inferred from short-read sequencing. We modified the figure captions and the main text to include this along with the corresponding references:

Figure 2:

“The ploidy level of each line was previously determined by DNA staining and flow cytometry (Charron et al., 2019; Marsit et al., 2021).”

Main text:

“The ratio of classified bases per subgenome was consistent with the corresponding ploidy levels: triploid BC lines had two copies of the SpC subgenome, while tetraploid lines had both SpC subgenomes duplicated (Charron et al., 2019; Marsit et al., 2021) (Figure 2B).”

“Finally, we used the ploidy level of each MA line subgenome as previously measured by flow cytometry and short-read sequencing (Charron et al., 2019; Marsit et al., 2021).”

Reviewer #3 (Public Review):

Henault et al. address the important open question of whether hybridization could trigger TE mobilization. To do this they analysed MA lines derived from crosses of Saccharomyces paradoxus and Saccharomyces cerevisiae using long-read sequencing. These MA lines were already analysed in a previous publication using Illumina short-read data but the novelty of this work is the long-read sequencing data, which may reveal previously missed information. It is an interesting message of this study that hybridization between the two species did not lead to much TE activity. Due to this low activity, the authors performed an additional TE activity assay in vivo to measure transposition rates in hybrid backgrounds. The study is well written and I cannot spot any major problems. The study provides some important messages (like the influence of the genotype and mitochondrial DNA on transposition rates).

Major comments

• What I miss the most in this work is the perspective of the host defence against TEs in Saccharmoces. Based on such a mechanistic perspective, why do the authors think that hybridization could lead to a TE reactivation? For example, in Drosophila small RNAs important for the defence against a TE, are solely maternally transmitted. Hybrid offspring will thus solely have small-RNAs complementary to the TEs of the mother but not to the TEs of the father, therefore a reactivation of the paternal TEs may be expected. I was thus wondering, what is the situation in yeast. Why would we expect an upregulation of TEs? Without such a mechanistic explanation the hypothesis that TEs should be upregulated in hybrids is a bit vague, based on a hunch.

We agree with the reviewer that in the first version of the manuscript, the justification for the investigation of the reactivation hypothesis in the first place was not self-sufficient and relied too much on our previous work, upon which this article builds. We extensively remodeled the introduction to better justify the investigation of this hypothesis in the context of the current knowledge on the regulation of Ty elements in Saccharomyces.  

Reviewer #1 (Recommendations For The Authors):

It's interesting that the net change in transposable element copy number in mutation accumulation lines is either insignificant or gain, and never a significant loss. I think this could make a nice discussion point regarding the roles of drift and selection on TE load.

We thank the reviewer for the suggestion and agree that this is an interesting perspective that we did not explore in the first version of the manuscript. We thus included a short discussion point in the Results:

“The distributions of net Ty CN change per MA line showed that most crosses had significant gains (Figure 3D), suggesting that Ty load can often increase as a result of random genetic drift.”

We also added the following paragraph to the discussion section:

“Our experiments illustrate how under weakened natural selection efficiency, TE load can increase in hybrid genomes by the action of transposition-unrelated SVs. This offers a nuanced perspective on the classical interpretation of the transposition-selection balance model (Charlesworth et al., 1994; Charlesworth and Langley, 1989), in which increased TE load would be predominantly driven by the relaxation of purifying selection against TE insertions generated by de novo transposition. Our results suggest that SVs arising in the context of hybridization can act as a significant source of TE insertion polymorphisms which natural selection can purge more or less efficiently, depending on the population genetic context. This is closely related to the idea that sexual reproduction could favor the spread of TE families, contributing to their evolutionary success (Hickey, 1982; Zeyl et al., 1996). Since the insertion polymorphisms that contribute to increase TE load mostly originate from standing genetic variation, they could be less deleterious and thus harder for natural selection to purge efficiently.”

The point about the role of LOH in TE load is cool!

We thank the reviewer for their enthusiasm, it is one of our favorite results as well.

Figure 1: Add a figure component of the green box and label it Ty1 or TE.

We modified Figure 1 accordingly.

Figure 2C: what is the assembly size ratio?

We added the following sentence to the figure caption to clarify what we define as assembly size ratio:

“Assembly size ratio refers to the ratio of subgenome assembly size to the corresponding parental assembly size.”

Something cut off in the N50 plot axis

Unfortunately, we can’t seem to understand what the reviewer meant with this comment, nothing seems cut out of the figure panel 2C in any of our versions of the manuscript.

Reviewer #2 (Recommendations For The Authors):

These are all minor comments/suggestions that the authors can take or leave.

Line 42: "fuels" should be "fuel".

Since the verb refers to “source” and not “variants”, we believe it should be at the third person singular.

Line 43: unclear what the authors mean by "regroup".

We understand how this phrasing may sound strange. We modified the sentence accordingly:

“Structural variation is a term that encompasses a broad variety of large-scale sequence alterations”

Line 51-52: There are a couple of really nice papers that could be cited here from Anna Selmecki's group (Todd et al. 2020, Todd and Selmecki 2019, both in eLife).

We thank the reviewer for the suggestions, we included some of these references in the manuscript.

Figure 1: This is a nice cartoon! I'd suggest spelling out LOH here for a truly naive reader.

We modified the Figure 1 accordingly.

Figure 3A: One thing that is slightly lost here in the presentation is the relative frequency of the different events because of the changing scales across 3A. I can see why you want to do it this way, but would consider whether there may be a way to present this that makes it more obvious how much more frequent polyploidy is than excision for example.

We agree with the reviewer that the focus of this visualization is to compare crosses and individual MA lines within SV types, and fails to display the relative importance of each SV type. We solved this by including an additional panel (new 3A) that shows how the number of Ty loci affected by each SV type scales in comparison to others.

Figure 5: I'm not a fan of the gray bars highlighting the individual strains. This made the graph less intuitively readable for me.

We tend to agree with the reviewer and rolled back to a previous version of Figure 5 that was lighter on annotations.

One thing I would like to see in the future from this data (definitely not in this paper) is genome rearrangements within these hybrid MA lines. How often are there structural changes and how often are those changes mediated by repeats including TEs?

We completely agree with the reviewer that this would be a very interesting avenue, with a distinct (and likely higher) set of challenges at the analysis level compared to simply focusing on TE sequences like we did here. We hope to be able to tackle this goal in the future of this project.

Reviewer #3 (Recommendations For The Authors):

• I'm not from the yeast field. But why this focus on the Ty-load? Are Ty's the only active TEs in yeast? Provide some background on the TE landscape in yeast and a justification for focusing on Ty's.

We agree with the reviewer that this point was only implicit in the introduction. We modified the introductory segment on Saccharomyces yeasts to mention that Ty retrotransposons are the only TEs found in these genomes, thus explaining the exclusive focus on them. It now reads as follows:

“In the case of Saccharomyces cerevisiae, the only TEs found are five families of long terminal repeat (LTR) retrotransposons families named Ty1-Ty5 (Kim et al., 1998).”

• 56 I would argue that Petrov et al 2003 is not the best citation for arguing that TEs can lead to genomic rearrangement through ectopic recombination. Petrov solely showed that some long TE families are at lower population frequency than short TE families ones. This could be due to many reasons (e.g. recent activity of long TEs - mostly LTRs) but Petrov interpreted the data as being due to ectopic recombination. Petrov, therefore, did not demonstrate any direct evidence for the involvement of ectopic recombination.

We agree with the reviewer that this reference is not the best choice to simply support the role of TEs in generating ectopic recombination events and modified the references accordingly.

• For the assembly the authors used two steps 1) separate the reads based on similarity to a subgenome 2) and assembly the reads from the resulting two sets separately. This is probably the only viable approach, but I'm wondering if this step can lead to some biases (many reads may not be assigned to one sub-genome or assigned to the wrong sub-genome). An alternative, possibly less biased approach, would be to use one of the emerging assemblers that promise to assemble sub-genomes. Maybe discuss why this approach was not pursued.

We completely agree that our method has some level of bias. We adopted it because it seemed the most appropriate to answer our question, which required to resolve individual TE insertions at the level of single haplotype sequences. One specific challenge of this dataset is that we have a relatively wide range of nucleotide divergence between parental subgenomes in the different MA crosses, from <1% to ~15%. The efficiency of haplotype separation from tools that are not necessarily designed to be tunable with respect to the level of nucleotide divergence seemed uncertain, which is why we opted for a custom methodology. Although read non-classification remains a problem that is hard to solve (and would remain so using orthogonal strategies), we believe that read misclassification is minimized by our stringent criteria for read classification. The goal of this study was not to develop a tool nor to benchmark our approach against existing diploid assembly tools. It yielded phased genome representations that were of sufficient completeness and contiguity to confidently answer our questions, and we believe that pushing the discussion towards technical considerations would fall outside of our main objective.

• The authors used a decision tree to classify Ty loci. What were the training data? How were the trees validated? Decision tree is a technical term for a classifier in machine learning. I do not think the authors used machine learning in this work, but rather an "an ad-hoc set of rules". The term decision tree in this study is misleading.

We believe that the term “decision tree” can simply refer to a hierarchy of conditional rules implemented as a classification algorithm. As the reviewer pointed, it is clear from the manuscript that none of the analyses performed include any form of training or fitting of a machine learning classifier. However, we agree that its specific reference to the machine learning classifier can create unnecessary confusion. We thus agree to remove this term from the manuscript and replaced all its instances by “a hierarchy of binary rules”.

• 272: as it is the CNC explanation does not make a lot of sense to me; some information is missing, is p22 expression increasing with copy numbers?

Yes, p22 expression correlates positively with the CN of p22-expressing Ty1 elements.

Why are the two alternative downstream codons important?

We thought it would be useful to mention the two start codons at this point because later in the discussion, we bring the conservation of the first start codon as an observation consistent with the putative expression of p22 in S. paradoxus. We also thought that it helped clarify the mechanism by which the N-truncated version of the protein is expressed.

p22 interferes with assembly viral particles when in high copy numbers, but what happens when at low copy numbers, is it essential for retroviral activity? Is it even necessary for the virus or just some garbage product (they mention N-truncated).

To our knowledge, these questions regarding the potential molecular functions of p22 outside of a retrotransposition restriction factor are still open. We added details to the background on CNC in the Introduction and Results section to help clarify some the points raised:

Introduction:

“The best known regulation mechanism in yeast is termed copy number control (CNC) and was characterized in the Ty1 family of S. cerevisiae. This mechanism is a potent copy-number dependent negative feedback loop by which increasing the CN of Ty1 elements strengthens their repression (Czaja et al., 2020; Garfinkel et al., 2003; Saha et al., 2015).”

Results:

“The mechanism of negative copy-number dependent self-regulation of retrotransposition (CNC) was characterized in the Ty1 family of S. cerevisiae (Garfinkel et al., 2016). This mechanism relies on the expression of an N-truncated variant of the Ty1 capsid/nucleocapsid Gag protein (p22) from two downstream alternative start codons (Nishida et al., 2015; Saha et al., 2015). p22 expression scales up with the CN of Ty1 elements that encode it (Tucker et al., 2015), which gradually interferes with the assembly of the viral-like particles essential for Ty1 replication (Cottee et al., 2021; Saha et al., 2015). Thus, CNC yields a steep negative relationship between the retrotransposition rate measured with a tester element and the number of Ty1 copies in the genome (Garfinkel et al., 2003; Tucker et al., 2015).”

• mtDNA influences transposition, is anything known about the mechanism?

When presenting this result, we make it clear that this finding is not new and was previously observed in S. cerevisiae x S. uvarum hybrids by Smukowski-Heil et al. (2021). In this reference, the authors discuss multiple mechanisms by which mitochondrial biology and mito-nuclear interplay may affect transposition rate, although their data cannot support one specific hypothesis. Our data does not to allow to further dissect the mechanistic basis of the mtDNA effect, not more than the effect of distinct Ty1 natural variants. Since we simply provide new independent evidence for the mtDNA effect, it seems to us that repeating the discussion on putative mechanisms while bringing no support to any given hypothesis would be of limited relevance.

• During the first reading, I got quite confused about what CN means (copy number as it turned out). I suggest using abbreviations only if absolutely necessary, and I'm not entirely convinced it is necessary here. But I leave this to the discretion of the authors.

We agree that the excessive use of abbreviations in manuscripts is annoying. However, in this case, “copy number” is used so extensively that its abbreviation seemed to improve the reading experience. Thus, we would prefer to keep it unchanged.

• Fig 3D: Wilcoxon Rank sum test. It is not clear to me what was tested here? Which data were used?

We confirm that the statistical test employed is the Wilcoxon signed-rank test, and not the Wilcoxon rank-sum test (also known as Mann-Whitney U-test). The Wilcoxon signed-rank test is used here as a non-parametric one-sample test against the null hypothesis that the distribution is centered around zero.

• de novo -> italics

We choose to follow the recommendation of the general style conventions of the ACS guide for scholarly communications not to italicize common Latin terms like “de novo”, “e.g.” and “i.e.”.

This part of the assignment confused me, but now I think I understand better. I realized that sometimes we don't have all the answers, and sometimes the "answers" are just a collaboration of opinions and it is up to you to decide.All the sources I am finding are not direct answers to my question, but they do relate and spark my thoughts and more questions.

Author Response

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

The reviewers make some suggestions aimed towards increasing the clarity of the manuscript, and I suggest that the authors examine those carefully. In particular, the figure is difficult to read and could contain additional information to help the reader's interpretation. For example, Reviewer 1 suggests including sample age estimates alongside depth, while Reviewer 3 also notes that there is missing information in the figure. Apart from the figure, Reviewer 1 suggests two additional analysis to help explain the amount of mammoth DNA recovered, which they observe is much higher than previous similar investigations. This would seem to be an important issue to address, given the surprising nature of the findings. In addition to this larger issue, the Reviewer makes a few important suggestions for supplementary material that may be needed to support the authors' statements.

Some additional recommended edits -- in particular to the text and included references to related studies -- are suggested by Reviewers 2 and 3, and both commented on the lack of a publicly-available data repository. The authors may also wish to comment on or revisit their differential treatment of wooly mammoth vs. wooly rhinoceros samples, though I suspect this has more to do with low read numbers for the rhinos.

Thank you very much for the positive assessment of our manuscript and clear suggestions for revision. We address these points below.

Reviewer #1 (Recommendations For The Authors):

I have a few suggestions that might further improve the manuscript:

It is difficult for the reader to follow which core slices exactly have been sampled and sequenced. The authors mention 23 samples were taken from core LK-001 and 16 samples from core LK-007. From the text it remains unclear to me what the exact age of each of these samples is. Figure 1 shows the depth at which the LK-001 core was sampled, maybe sample age estimates could be included here.

Thanks for pointing this out. We have added approximate ages to Figure 1, added the depth range to the text (“from 1.5 to 80 cm”; l. 73-74, caption Figure 1), and reworked the table of the sampling depths in the supplement.

Line 84-87. The authors mention the retrieval of DNA from several expected Arctic taxa, however no further data regarding these findings is given in the manuscript. It would be useful to report the same numbers for these species as the ones given for the Mammuthus and woolly rhinoceros, which would allow for a comparison of the relative abundance of the DNA between these species. Are the expected Arctic species for instance at much higher (DNA) abundance in the samples? It would also be interesting to know if the authors discovered DNA from extant species that are unlikely to have occurred in the geographic region. A (supplementary)table listing the number of mapped reads to each of the respective mitogenomes for each sequence library would be useful for the reader.

We added a supplementary table (S8) indicating the numbers of reads assigned to mammals.

Line 90: I am somewhat amazed by the amount of mammoth DNA the authors recovered from these cores. A total depth of over 400X of the mitogenome is quite extraordinary and I am not aware of any ancient sediment study to date that has retrieved a similar amount of data. For instance, the Wang et al. 2021 paper, which the authors cite, sequenced over 400 samples and did not find any mammoth DNA in 70% of those. For the 30% of samples showing signs of mammoth DNA they retrieved on average 530 sequence reads. In this study the authors find on average ~20.000 reads, in 22 out of the 23 sequence libraries. This makes me wonder if the way the mapping was performed has been too lenient, resulting in possible spurious mappings? To really confirm the authenticity of the mammoth (and woolly rhino data) I would suggest two additional analysis:

1) Mapping all the sequence libraries to a reference consisting of the complete Asian-elephant genome (for instance https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_024166365.1/), the complete human genome (+mitogenome) and the Asian elephant mitogenome. This could possibly reduce spurious mappings as conserved regions between the genomes are filtered out and could also reduce the possible mapping of NUMTS. If the authors could show that after such a mapping approach a significant number of reads are still assigned to the Asian elephant part (including the mitogenome) of the reference, the reported findings would be strengthened.

2) I also suggest to construct a mitochondrial haplotype network from the obtained DNA, while also including previously published Asian and African elephants as well as previously published mammoth mitogenomes. If the obtained haplotypes indeed show that they cluster within the known haplotype diversity of mammoth, that would be strong support for the authenticity of the data

The same analysis could be considered for the woolly rhino data, although the lower read numbers might make this analysis challenging.

We agree that the amount of mammoth DNA is surprising, which is why we opted for further laboratory experiments for confirmation of the hybridization capture results of the first core, i.e., 1) DNA extraction from a second core of a different lake, 2) a quantitative PCR approach (ddPCR), and 3) metabarcoding. Our results of the highly specific ddPCR and metabarcoding assays confirmed considerable amounts of mammoth DNA in two sediment cores of different lakes, thus we have no doubts regarding the authenticity of the data. Considering the large amount of mammoth DNA, the high number of reads, and particularly the high mitogenome coverage, we argue that the effect of some spurious mapping is negligible and does not affect the main outcome and conclusions of our study. Although we agree that a haplotype network would be interesting, such analyses would stretch beyond the focus of this publication.

Line 91: The authors mention negative controls (extraction and library blanks) did not produce any reads assigned to mammals. This is quite remarkable, as in my experience low levels of (human)contamination are almost always present in the blanks. Could the authors comment on why they think the blanks did not show any signal of mammalian DNA?

The hybridization capture enrichment and the filtration and mapping procedures likely eliminated human contamination. Also, the data were mapped against Arctic mammal mitogenomes, which did not include human reference sequences. However, six of the sediment samples contained human sequences (now shown in supplementary table S8), albeit at low read counts (mean = 65)

Line 97: "mapping suggested that the sequences throughout the core originated from multiple individuals" The authors do not provide any supporting data showing this. I think that an analysis (for instance based on allele frequencies) has to be included in manuscript to support this claim.

We agree that his claim was not sufficiently supported. We performed further analyses including genomic data of previously retrieved mammoth remains and assigned our data to these haplogroups; the results were added to the main text and are shown as a figure (Fig. 2).

Line 98: "Signatures of post-mortem DNA decay were comparably minor."

Do the authors know if the used hybridisation enrichment method can distort the measurement of post-mortem damage? Are for instance reads with C-T substitutions less likely to be captured by the baits?

To our knowledge, there is no study suggesting that damaged sites are less likely to be captured. In general, the hybridization capture procedure is not overly specific, and studies report that DNA is readily and preferentially captured as long as the difference between baits and DNA is not above 10%.

Line 100: "The proportions of bases did not suggest a substantial deviation from those in the reference genomes or in the closest extant relative of Mammuthus, the Asian elephant (Elephas maximus)."

It is not clear to me what the authors mean by this. Could the authors explain how this was measured and what their interpretation of this result is?

We realize that the sentence was unclear. We meant that the nucleotide composition was similar to that of the reference genomes or the closest extant relative. However, as we do not consider this important for the argument, we have removed this sentence from the manuscript.

Given the high number of recovered mammoth reads in the samples, it would be interesting to know how much mammoth reads are present in the sample before enrichment capture with the baits. Shotgun sequencing the raw extract of one of the samples with the highest number of mammoth reads might allow for a rough estimate of mammoth DNA abundance compared to the other extant species (e.g. reindeer, Arctic lemming and hare) found in the sample(s). This could give further clarification about the extent of stratigraphy disturbance and its overall effect on the DNA based community reconstruction. However, this is just a suggested additional analysis and not something I believe crucial for supporting the overall findings in this manuscript.

We fully agree that this would be a highly interesting and informative additional analysis to perform. It was, however, not possible to perform this additional analyses in the course of the current experiments.

Finally, I could not find a public link to the (sequence)data produced in this study. I strongly encourage the authors to make their data publicly available.

Thank you for pointing this out. We have added a Data Availability paragraph, including the respective reference.

Reviewer #2 (Recommendations For The Authors):

In the Discussion it is mentioned that the reasons for Mammoth extinction are not entirely clear but are largely attributed to sudden climate warming (and add some relevant citations). However, there is also abundant literature that suggest humans also played a role in their extinction (for instance, a recent one, Damien et al. (2022) at Ecology Letters 25: 127-137).

We agree with the reviewer and have added some the recent citation highlighting the possible influence of humans.

One possibility to add further interest to this paper would be to conduct a phylogenetic tree with the Mammoth mitogenome(s) retrieved and a reference dataset; it could be interesting to know where do they fall in the phylogeny -already abundant with tens of individuals- and maybe it could be even possible to roughly estimate their date. There are some papers that report many Mammoth mitogenomes, including of course some from Siberia; for instance Chang et al. (2017) at Sci Reports and also Fellow Yates et al. (2017) also at Sci Reports (the latter mainly from Central Europe).

We are well aware of the amount of mt genomes available for mammoth, and such an analyses would be an interesting addition, potentially also offering the possibility to date the DNA. However, the analyses was hampered and would be less secure for this dataset, as our sequences display quite some variation among each other, suggesting that we have a mix of multiple mt genomes, which we cannot readily distinguish. We thus refrain from this, also because we instead provide multiple lines of evidence for the existence of the mammoth DNA in the surface sediment core (metabarcoding, ddPCR).

Minor points:

-Correct wooly to woolly

Revised.

-In the sampling description it is not totally clear if the samples were taken at 1 cm each (it is mentioned that core LK-001 is sliced in the field at 1-cm steps for radiometric dating and later it is explained that 23 samples were analyzed from this core, but it is unclear if they represent 23 cm of core)

-Maybe the authors could briefly define some terms such as "talik"

Revised.

Reviewer #3 (Recommendations For The Authors):

Maybe I missed this but I could not find a data availability statement or the location of the repository

We have added a Data Availability paragraph, including the respective reference.

It would be good to see some additional analysis on the distribution of the woolly rhinoceros DNA through the sediment core - like the figure for the mammoth i.e read numbers vs depth.

We have added to the supplements a table showing the numbers of assigned mammal reads over the core depths (Table S8). However, as rhinoceros reads are considerable rarer in our results, we did not produce a figure.

Would it be possible to be more explicit about the multiple mammoth individuals, could you calculate a minimum number or haplotypes for example.

We agree that his claim was not sufficiently supported and added results from additional analyses (incl. Fig. 2). Please see our response above.

Based on the aim stated in the introduction, the analysis of the Arctic biodiversity of this area is missing, it would be nice to see these result added or maybe the focus needs to be changed for clarity.

We now explicitly state that this objective pertains to a different study, which is currently still in preparation for publication.

The single main figure needs a bit more consideration. For example in panel A - there was no information on the transformation performed or what the general trend line refers to. Do the results in panel B refer to all 22 libraries? What is the x-axis in Panel C and what do the coloured lines refer to? Additionally, I think the figure needs to be in higher resolution with increased text size on all axes.

We revised the figure and the caption for clarity and readability.

Finally this might be an accidental typo - but when referring to the sample aged at around 8,677 years in text it states this the 36.5 cm sample (line 130 and 192), but the supplementary says this is the 51cm sample (Table S6). This would maybe impact potential conclusions. Would you be able to clarify this.

Thank you for noting this error, we revised it.

Author Response

Reviewer #1 (Public Review):

Summary:

The manuscript by Dubicka and co-workers on calcification in miliolid foraminifera presents an interesting piece of work. The study uses confocal and electron microscopy to show that the traditional picture of calcification in porcelaneous foraminifera is incorrect.

Strengths:

The authors present high-quality images and an original approach to a relatively solid (so I thought) model of calcification.

Weaknesses:

There are several major shortcomings. Despite the interesting subject and the wonderful images, the conclusions of this manuscript are simply not supported at all by the results. The fluorescent images may not have any relation to the process of calcification and should therefore not be part of this manuscript. The SEM images, however, do point to an outdated idea of miliolid calcification. I think the manuscript would be much stronger with the focus on the SEM images and with the speculation of the physiological processes greatly reduced.

Reply: We would like to give thanks for all of the highly valuable comments. Prior to our study, we were also convinced that the calcification model of Miliolid (porcelaneous) foraminifera was relatively solid. Nevertheless, our SEM imaging results surprisingly contradicted the old model. The main difference is the in situ biomineralization of calcitic needles that precipitate within the chamber wall after deposition of ACC-bearing vesicles. We agree that our fluorescence studies presented in the paper are not conclusive evidence for the calcification model used by the studied Miliolid species. However, our fluorescent results show that “the old model” (sensu Hemleben et al., 1986) is not completely outdated. Most of the fluorescent imaging data show a vesicular transport of substrates necessary for calcification. This transport is presented by Calcein labelling experiments (Movie 1 that show a high number of dynamic endocytic vesicles of sea water circulation within the cytoplasm. These very fine Calcein-labelled vesicles are most likely responsible for transport and deposition of Ca2+ ions. This is partly consistent with the model presented by Hemleben et al. (1986). We may speculate that calcite nucleation is already occurring within the transported vesicles, but at this stage of research we have no evidence for this phenomenon.

Further live imaging fluorescence data show autofluorescence of vesicles upon excitation at 405 nm (emission 420–480 nm) associated with acidic vesicles marked by pH-sensitive LysoGlow84, may be a hint indicating association of ACC-bearing vesicles with acidic vesicles. Such spatial association of these vesicles may indicate a mechanism of pH elevation in the vesicles transporting Ca2+-rich gel to the calcifying wall of the new chamber.

We will do our best to limit the physiological interpretation presented based on fluorescence studies in the revised version of the manuscript. We are convinced that our fluorescent live imaging experiments provide important observations in biomineralizing Miliolid foraminifera, which are still missing in the existing literature. It should be stressed that all the fluorescent experiments and SEM observations were based on specimens constructing and biomineralizing new chambers. All of them belong to the same species and come from the same culture. Due to the aforementioned reasons, it is worthwhile presenting these complimentary results of our study. In the future they may be helpful in further exploration and understanding of all aspects of calcification in foraminifera.

Reviewer #2 (Public Review):

Summary:

Dubicka et al. in their paper entitled " Biocalcification in porcelaneous foraminifera" suggest that in contrast to the traditionally claimed two different modes of test calcification by rotallid and porcelaneous miliolid formaminifera, both groups produce calcareous tests via the intravesicular mineral precursors (Mg-rich amorphous calcium carbonate). These precursors are proposed to be supplied by endocytosed seawater and deposited in situ as mesocrystals formed at the site of new wall formation within the organic matrix. The authors did not observe the calcification of the needles within the transported vesicles, which challenges the previous model of miliolid mineralization. Although the authors argue that these two groups of foraminifera utilize the same calcification mechanism, they also suggest that these calcification pathways evolved independently in the Paleozoic.

Reply: We would like to acknowledge the review and all valuable comments. We do not argue that Miliolida and Rotallida utilise an identical calcification mechanism, but both groups utilize less divergent crystallization pathways, where mesocrystalline chamber walls are created by accumulating and assembling particles of pre-formed liquid amorphous mineral phase.

Strengths:

The authors document various unknown aspects of calcification of Pseudolachlanella eburnea and elucidate some poorly explained phenomena (e.g., translucent properties of the freshly formed test) however there are several problematic observations/interpretations which in my opinion should be carefully addressed.

Weaknesses:

1) The authors (line 122) suggest that "characteristic autofluorescence indicates the carbonate content of the vesicles (Fig. S2), which are considered to be Mg-ACCs (amorphous MgCaCO3) (Fig. 2, Movies S4 and S5)". Figure S2 which the authors refer to shows only broken sections of organic sheath at different stages of mineralization. Movie S4 shows that only in a few regions some vesicles exhibit red autofluorescence interpreted as Mg-ACC (S5 is missing but probably the authors were referring to S3). In their previous paper (Dubicka et al 2023: Heliyon), the authors used exactly the same methodology to suggest that these are intracellularly formed Mg-rich amorphous calcium carbonate particles that transform into a stable mineral phase in rotaliid Aphistegina lessonii. However, in Figure 1D (Dubicka et al 2023) the apparently carbonate-loaded vesicles show the same red autofluorescence as the test, whereas in their current paper, no evidence of autofluorescence of Mg-ACC grains accumulated within the "gel-like" organic matrix is given. The S3 and S4 movies show circulation of various fluorescing components, but no initial phase of test formation is observable (numerous mineral grains embedded within the organic matrix - Figures 3A and B - should be clearly observed also as autofluorescence of the whole layer). Thus the crucial argument supporting the calcification model (Figure 5) is missing. There is no support for the following interpretation (lines 199-203) "The existence of intracellular, vesicular intermediate amorphous phase (Mg-ACC pools), which supply successive doses of carbonate material to shell production, was supported by autofluorescence (excitation at 405 nm; Fig. 2; Movies S3 and S4; see Dubicka et al., 2023) and a high content of Ca and Mg quantified from the area of cytoplasm by SEM-EDS analysis (Fig. S6)."

Reply: We used laser line 405nm and multiphoton excitation to detect ACCs. These wavelengths (partly) permeate the shell to excite ACCs autofluorescence. The autofluorescence of the shells is present as well, but it is not clearly visible in movieS4 as the fluorescence of ACCs is stronger. This may be related to the plane/section of the cell which is shown. The laser permeates the shell above the ACCs (short distance), but to excite the shell CaCO3 around foraminifera in the same three-dimensional section where ACCs are shown, the light must pass a thick CaCO3 area due to the three-dimensional structure of the foraminifera shell. Therefore, the laser light intensity is reduced. In a revised version a movie/image with reduced threshold will be shown.

2) The authors suggest that "no organic matter was detected between the needles of the porcelain structures (Figures 3E; 3E; S4C, and S5A)". Such a suggestion, which is highly unusual considering that biogenic minerals almost by definition contain various organic components, was made based only on FE-SEM observation. The authors should either provide clearcut evidence of the lack of organic matter (unlikely) or may suggest that intense calcium carbonate precipitation within organic matrix gel ultimately results in a decrease of the amount of the organic phase (but not its complete elimination), alike the pure calcium carbonate crystals are separated from the remaining liquid with impurities ("mother liquor"). On the other hand, if (249-250) "organic matrix involved in the biomineralization of foraminiferal shells may contain collagen-like networks", such "laminar" organization of the organic matrix may partly explain the arrangement of carbonate fibers parallel to the surface as observed in Fig. 3E1.

Reply: We agree with the reviewer that biogenic minerals should, by definition, contain some organic components. We wrote that "no organic matter was detected between the needles of the porcelain structures” as we did not detect any organic structures based only on our FE-SEM observations. We are convinced that the shell incorporates a limited amount of organic matrix. We will rephrase this part of the text to avoid further confusion.

3) The author's observations indeed do not show the formation of individual skeletal crystallites within intracellular vesicles, however, do not explain either what is the structure of individual skeletal crystallites and how they are formed. Especially, what are the structures observed in polarized light (and interpreted as calcite crystallites) by De Nooijer et al. 2009? The author's explanation of the process (lines 213-216) is not particularly convincing "we suspect that the OM was removed from the test wall and recycled by the cell itself".

Reply: Thank you for this comment. We will do our best to supplement our explanations. We are aware of the structures observed in polarized light by De Nooijer et al. (2009). However, Goleń et al. (2022, Protist, https://doi.org/10.1016/j.protis.2022.125886) showed that organic polymers may also exhibit light polarization. Additional experimental studies are needed to distinguish these types of polarization. We will aim to investigate this issue in our future research.

4) The following passage (lines 296-304) which deals with the concept of mesocrystals is not supported by the authors' methodology or observations. The authors state that miliolid needles "assembled with calcite nanoparticles, are unique examples of biogenic mesocrystals (see Cölfen and Antonietti, 2005), forming distinct geometric shapes limited by planar crystalline faces" (later in the same passage the authors say that "mesocrystals are common biogenic components in the skeletons of marine organisms" (are they thus unique or are they common)? It is my suggestion to completely eliminate this concept here until various crystallographic details of the miliolid test formation are well documented.

Reply: Our intention was to express that mesocrystals are common biogenic components in the skeletons of marine organisms, however Miliolid needles that form distinct geometric shapes limited by planar crystalline faces are unique type of mesocrystals.

Author Response

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

To the Senior Editor and the Reviewing Editor:

We sincerely appreciate the valuable comments provided by the reviewers, the reviewing editor, and the senior editor. After carefully reviewing and considering the comments, we have addressed the key concerns raised by the reviewers and made appropriate modifications to the article in the revised manuscript.

The main revisions made to the manuscript are as follows:

1) We have added comparison experiments with TNDM (see Fig. 2 and Fig. S2).

2) We conducted new synthetic experiments to demonstrate that our conclusions are not a by-product of d-VAE (see Fig. S2 and Fig. S11).

3) We have provided a detailed explanation of how our proposed criteria, especially the second criterion, can effectively exclude the selection of unsuitable signals.

4) We have included a semantic overview figure of d-VAE (Fig. S1) and a visualization plot of latent variables (Fig. S13).

5) We have elaborated on the model details of d-VAE, as well as the hyperparameter selection and experimental settings of other comparison models.

We believe these revisions have significantly improved the clarity and comprehensibility of the manuscript. Thank you for the opportunity to address these important points.

Reviewer #1

Q1: “First, the model in the paper is almost identical to an existing VAE model (TNDM) that makes use of weak supervision with behaviour in the same way [1]. This paper should at least be referenced. If the authors wish they could compare their model to TNDM, which combines a state space model with smoothing similar to LFADS. Given that TNDM achieves very good behaviour reconstructions, it may be on par with this model without the need for a Kalman filter (and hence may achieve better separation of behaviour-related and unrelated dynamics).”

Our model significantly differs from TNDM in several aspects. While TNDM also constrains latent variables to decode behavioral information, it does not impose constraints to maximize behavioral information in the generated relevant signals. The trade-off between the decoding and reconstruction capabilities of generated relevant signals is the most significant contribution of our approach, which is not reflected in TNDM. In addition, the backbone network of signal extraction and the prior distribution of the two models are also different.

It's worth noting that our method does not require a Kalman filter. Kalman filter is used for post hoc assessment of the linear decoding ability of the generated signals. Please note that extracting and evaluating relevant signals are two distinct stages.

Heeding your suggestion, we have incorporated comparison experiments involving TNDM into the revised manuscript. Detailed information on model hyperparameters and training settings can be found in the Methods section in the revised manuscripts.

Thank you for your valuable feedback.

Q2: “Second, in my opinion, the claims regarding identifiability are overstated - this matters as the results depend on this to some extent. Recent work shows that VAEs generally suffer from identifiability problems due to the Gaussian latent space [2]. This paper also hints that weak supervision may help to resolve such issues, so this model as well as TNDM and CEBRA may indeed benefit from this. In addition however, it appears that the relative weight of the KL Divergence in the VAE objective is chosen very small compared to the likelihood (0.1%), so the influence of the prior is weak and the model may essentially learn the average neural trajectories while underestimating the noise in the latent variables. This, in turn, could mean that the model will not autoencode neural activity as well as it should, note that an average R2 in this case will still be high (I could not see how this is actually computed). At the same time, the behaviour R2 will be large simply because the different movement trajectories are very distinct. Since the paper makes claims about the roles of different neurons, it would be important to understand how well their single trial activities are reconstructed, which can perhaps best be investigated by comparing the Poisson likelihood (LFADS is a good baseline model). Taken together, while it certainly makes sense that well-tuned neurons contribute more to behaviour decoding, I worry that the very interesting claim that neurons with weak tuning contain behavioural signals is not well supported.”

We don’t think our distilled signals are average neural trajectories without variability. The quality of reconstructing single trial activities can be observed in Figure 3i and Figure S4. Neural trajectories in Fig. 3i and Fig. S4 show that our distilled signals are not average neural trajectories. Furthermore, if each trial activity closely matched the average neural trajectory, the Fano Factor (FF) should theoretically approach 0. However, our distilled signals exhibit a notable departure from this expectation, as evident in Figure 3c, d, g, and f. Regarding the diminished influence of the KL Divergence: Given that the ground truth of latent variable distribution is unknown, even a learned prior distribution might not accurately reflect the true distribution. We found the pronounced impact of the KL divergence would prove detrimental to the decoding and reconstruction performance. As a result, we opt to reduce the weight of the KL divergence term. Even so, KL divergence can still effectively align the distribution of latent variables with the distribution of prior latent variables, as illustrated in Fig. S13. Notably, our goal is extracting behaviorally-relevant signals from given raw signals rather than generating diverse samples from the prior distribution. When aim to separating relevant signals, we recommend reducing the influence of KL divergence. Regarding comparing the Poisson likelihood: We compared Poisson log-likelihood among different methods (except PSID since their obtained signals have negative values), and the results show that d-VAE outperforms other methods.

Author response image 1.

Regarding how R2 is computed: , where and denote ith sample of raw signals, ith sample of distilled relevant signals, and the mean of raw signals. If the distilled signals exactly match the raw signals, the sum of squared error is zero, thus R2=1. If the distilled signals always are equal to R2=0. If the distilled signals are worse than the mean estimation, R2 is negative, negative R2 is set to zero.

Thank you for your valuable feedback.

Q3: “Third, and relating to this issue, I could not entirely follow the reasoning in the section arguing that behavioural information can be inferred from neurons with weak selectivity, but that it is not linearly decodable. It is right to test if weak supervision signals bleed into the irrelevant subspace, but I could not follow the explanations. Why, for instance, is the ANN decoder on raw data (I assume this is a decoder trained fully supervised) not equal in performance to the revenant distilled signals? Should a well-trained non-linear decoder not simply yield a performance ceiling? Next, if I understand correctly, distilled signals were obtained from the full model. How does a model perform trained only on the weakly tuned neurons? Is it possible that the subspaces obtained with the model are just not optimally aligned for decoding? This could be a result of limited identifiability or model specifics that bias reconstruction to averages (a well-known problem of VAEs). I, therefore, think this analysis should be complemented with tests that do not depend on the model.”

Regarding “Why, for instance, is the ANN decoder on raw data (I assume this is a decoder trained fully supervised) not equal in performance to the relevant distilled signals? Should a well-trained non-linear decoder not simply yield a performance ceiling?”: In fact, the decoding performance of raw signals with ANN is quite close to the ceiling. However, due to the presence of significant irrelevant signals in raw signals, decoding models like deep neural networks are more prone to overfitting when trained on noisy raw signals compared to behaviorally-relevant signals. Consequently, we anticipate that the distilled signals will demonstrate superior decoding generalization. This phenomenon is evident in Fig. 2 and Fig. S1, where the decoding performance of the distilled signals surpasses that of the raw signals, albeit not by a substantial margin.

Regarding “Next, if I understand correctly, distilled signals were obtained from the full model. How does a model perform trained only on the weakly tuned neurons? Is it possible that the subspaces obtained with the model are just not optimally aligned for decoding?”：Distilled signals (involving all neurons) are obtained by d-VAE. Subsequently, we use ANN to evaluate the performance of smaller and larger R2 neurons. Please note that separating and evaluating relevant signals are two distinct stages.

Regarding the reasoning in the section arguing that smaller R2 neurons encode rich information, we would like to provide a detailed explanation:

1) After extracting relevant signals through d-VAE, we specifically selected neurons characterized by smaller R2 values (Here, R2 signifies the proportion of neuronal activity variance explained by the linear encoding model, calculated using raw signals). Subsequently, we employed both KF and ANN to assess the decoding performance of these neurons. Remarkably, our findings revealed that smaller R2 neurons, previously believed to carry limited behavioral information, indeed encode rich information.

2) In a subsequent step, we employed d-VAE to exclusively distill the raw signals of these smaller R2 neurons (distinct from the earlier experiment where d-VAE processed signals from all neurons). We then employed KF and ANN to evaluate the distilled smaller R2 neurons. Interestingly, we observed that we could not attain the same richness of information solely through the use of these smaller R2 neurons.

3) Consequently, we put forth and tested two hypotheses: First, that larger R2 neurons introduce additional signals into the smaller R2 neurons that do not exist in the real smaller R2 neurons. Second, that larger R2 neurons aid in restoring the original appearance of impaired smaller R2 neurons. Our proposed criteria and synthetic experiments substantiate the latter scenario.

Thank you for your valuable feedback.

Q4: “Finally, a more technical issue to note is related to the choice to learn a non-parametric prior instead of using a conventional Gaussian prior. How is this implemented? Is just a single sample taken during a forward pass? I worry this may be insufficient as this would not sample the prior well, and some other strategy such as importance sampling may be required (unless the prior is not relevant as it weakly contributed to the ELBO, in which case this choice seems not very relevant). Generally, it would be useful to see visualisations of the latent variables to see how information about behaviour is represented by the model.”

Regarding "how to implement the prior?": Please refer to Equation 7 in the revised manuscript; we have added detailed descriptions in the revised manuscript.

Regarding "Generally, it would be useful to see visualizations of the latent variables to see how information about behavior is represented by the model.": Note that our focus is not on latent variables but on distilled relevant signals. Nonetheless, at your request, we have added the visualization of latent variables in the revised manuscript. Please see Fig. S13 for details.

Thank you for your valuable feedback.

Recommendations: “A minor point: the word 'distill' in the name of the model may be a little misleading - in machine learning the term refers to the construction of smaller models with the same capabilities.

It should be useful to add a schematic picture of the model to ease comparison with related approaches.”

In the context of our model's functions, it operates as a distillation process, eliminating irrelevant signals and retaining the relevant ones. Although the name of our model may be a little misleading, it faithfully reflects what our model does.

I have added a schematic picture of d-VAE in the revised manuscript. Please see Fig. S1 for details.

Thank you for your valuable feedback.

Reviewer #2

Q1: “Is the apparently increased complexity of encoding vs decoding so unexpected given the entropy, sparseness, and high dimensionality of neural signals (the "encoding") compared to the smoothness and low dimensionality of typical behavioural signals (the "decoding") recorded in neuroscience experiments? This is the title of the paper so it seems to be the main result on which the authors expect readers to focus. ”

We use the term "unexpected" due to the disparity between our findings and the prior understanding concerning neural encoding and decoding. For neural encoding, as we said in the Introduction, in previous studies, weakly-tuned neurons are considered useless, and smaller variance PCs are considered noise, but we found they encode rich behavioral information. For neural decoding, the nonlinear decoding performance of raw signals is significantly superior to linear decoding. However, after eliminating the interference of irrelevant signals, we found the linear decoding performance is comparable to nonlinear decoding. Rooted in these findings, which counter previous thought, we employ the term "unexpected" to characterize our observations.

Thank you for your valuable feedback.

Q2: “I take issue with the premise that signals in the brain are "irrelevant" simply because they do not correlate with a fixed temporal lag with a particular behavioural feature hand-chosen by the experimenter. As an example, the presence of a reward signal in motor cortex [1] after the movement is likely to be of little use from the perspective of predicting kinematics from time-bin to time-bin using a fixed model across trials (the apparent definition of "relevant" for behaviour here), but an entire sub-field of neuroscience is dedicated to understanding the impact of these reward-related signals on future behaviour. Is there method sophisticated enough to see the behavioural "relevance" of this brief, transient, post-movement signal? This may just be an issue of semantics, and perhaps I read too much into the choice of words here. Perhaps the authors truly treat "irrelevant" and "without a fixed temporal correlation" as synonymous phrases and the issue is easily resolved with a clarifying parenthetical the first time the word "irrelevant" is used. But I remain troubled by some claims in the paper which lead me to believe that they read more deeply into the "irrelevancy" of these components.”

In this paper, we employ terms like ‘behaviorally-relevant’ and ‘behaviorally-irrelevant’ only regarding behavioral variables of interest measured within a given task, such as arm kinematics during a motor control task. A similar definition can be found in the PSID[1].

Thank you for your valuable feedback.

[1] Sani, Omid G., et al. "Modeling behaviorally relevant neural dynamics enabled by preferential subspace identification." Nature Neuroscience 24.1 (2021): 140-149.

Q3: “The authors claim the "irrelevant" responses underpin an unprecedented neuronal redundancy and reveal that movement behaviors are distributed in a higher-dimensional neural space than previously thought." Perhaps I just missed the logic, but I fail to see the evidence for this. The neural space is a fixed dimensionality based on the number of neurons. A more sparse and nonlinear distribution across this set of neurons may mean that linear methods such as PCA are not effective ways to approximate the dimensionality. But ultimately the behaviourally relevant signals seem quite low-dimensional in this paper even if they show some nonlinearity may help.”

The evidence for the “useless” responses underpin an unprecedented neuronal redundancy is shown in Fig. 5a, d and Fig. S9a. Specifically, the sum of the decoding performance of smaller R2 neurons and larger R2 neurons is significantly greater than that of all neurons for relevant signals (red bar), demonstrating that movement parameters are encoded very redundantly in neuronal population. In contrast, we can not find this degree of neural redundancy in raw signals (purple bar).

The evidence for the “useless” responses reveal that movement behaviors are distributed in a higher-dimensional neural space than previously thought is shown in the left plot (involving KF decoding) of Fig. 6c, f and Fig. S9f. Specifically, the improvement of KF using secondary signals is significantly higher than using raw signals composed of the same number of dimensions as the secondary signals. These results demonstrate that these dimensions, spanning roughly from ten to thirty, encode much information, suggesting that behavioral information exists in a higher-dimensional subspace than anticipated from raw signals.

Thank you for your valuable feedback.

Q5: “there is an apparent logical fallacy that begins in the abstract and persists in the paper: "Surprisingly, when incorporating often-ignored neural dimensions, behavioral information can be decoded linearly as accurately as nonlinear decoding, suggesting linear readout is performed in motor cortex." Don't get me wrong: the equivalency of linear and nonlinear decoding approaches on this dataset is interesting, and useful for neuroscientists in a practical sense. However, the paper expends much effort trying to make fundamental scientific claims that do not feel very strongly supported. This reviewer fails to see what we can learn about a set of neurons in the brain which are presumed to "read out" from motor cortex. These neurons will not have access to the data analyzed here. That a linear model can be conceived by an experimenter does not imply that the brain must use a linear model. The claim may be true, and it may well be that a linear readout is implemented in the brain. Other work [2,3] has shown that linear readouts of nonlinear neural activity patterns can explain some behavioural features. The claim in this paper, however, is not given enough”

Due to the limitations of current observational methods and our incomplete understanding of brain mechanisms, it is indeed challenging to ascertain the specific data the brain acquires to generate behavior and whether it employs a linear readout. Conventionally, the neural data recorded in the motor cortex do encode movement behaviors and can be used to analyze neural encoding and decoding. Based on these data, we found that the linear decoder KF achieves comparable performance to that of the nonlinear decoder ANN on distilled relevant signals. This finding has undergone validation across three widely used datasets, providing substantial evidence. Furthermore, we conducted experiments on synthetic data to show that this conclusion is not a by-product of our model. In the revised manuscript, we added a more detailed description of this conclusion.

Thank you for your valuable feedback.

Q6: “Relatedly, I would like to note that the exercise of arbitrarily dividing a continuous distribution of a statistic (the "R2") based on an arbitrary threshold is a conceptually flawed exercise. The authors read too much into the fact that neurons which have a low R2 w.r.t. PDs have behavioural information w.r.t. other methods. To this reviewer, it speaks more about the irrelevance, so to speak, of the preferred direction metric than anything fundamental about the brain.”

We chose the R2 threshold in accordance with the guidelines provided in reference [1]. It's worth mentioning that this threshold does not exert any significant influence on the overall conclusions.

Thank you for your valuable feedback.

[1] Inoue, Y., Mao, H., Suway, S.B., Orellana, J. and Schwartz, A.B., 2018. Decoding arm speed during reaching. Nature communications, 9(1), p.5243.

Q7: “I am afraid I may be missing something, as I did not understand the fano factor analysis of Figure 3. In a sense the behaviourally relevant signals must have lower FF given they are in effect tied to the temporally smooth (and consistent on average across trials) behavioural covariates. The point of the original Churchland paper was to show that producing a behaviour squelches the variance; naturally these must appear in the behaviourally relevant components. A control distribution or reference of some type would possibly help here.”

We agree that including reference signals could provide more context. The Churchland paper said stimulus onset can lead to a reduction in neural variability. However, our experiment focuses specifically on the reaching process, and thus, we don't have comparative experiments involving different types of signals.

Thank you for your valuable feedback.

Q8: “The authors compare the method to LFADS. While this is a reasonable benchmark as a prominent method in the field, LFADS does not attempt to solve the same problem as d-VAE. A better and much more fair comparison would be TNDM [4], an extension of LFADS which is designed to identify behaviourally relevant dimensions.”

We have added the comparison experiments with TNDM in the revised manuscript (see Fig. 2 and Fig. S2). The details of model hyperparameters and training settings can be found in the Methods section in the revised manuscripts.

Thank you for your valuable feedback.

Reviewer #3

Q1.1: “TNDM: LFADS is not the best baseline for comparison. The authors should have compared with TNDM (Hurwitz et al. 2021), which is an extension of LFADS that (unlike LFADS) actually attempts to extract behaviorally relevant factors by adding a behavior term to the loss. The code for TNDM is also available on Github. LFADS is not even supervised by behavior and does not aim to address the problem that d-VAE aims to address, so it is not the most appropriate comparison. ”

We have added the comparison experiments with TNDM in the revised manuscript (see Fig. 2 and Fig. S2). The details of model hyperparameters and training settings can be found in the Methods section in the revised manuscripts.

Thank you for your valuable feedback.

Q1.2: “LFADS: LFADS is a sequential autoencoder that processes sections of data (e.g. trials). No explanation is given in Methods for how the data was passed to LFADS. Was the moving averaged smoothed data passed to LFADS or the raw spiking data (at what bin size)? Was a gaussian loss used or a poisson loss? What are the trial lengths used in each dataset, from which part of trials? For dataset C that has back-to-back reaches, was data chopped into segments? How long were these segments? Were the edges of segments overlapped and averaged as in (Keshtkaran et al. 2022) to avoid noisy segment edges or not? These are all critical details that are not explained. The same details would also be needed for a TNDM comparison (comment 1.1) since it has largely the same architecture as LFADS.

It is also critical to briefly discuss these fundamental differences between the inputs of methods in the main text. LFADS uses a segment of data whereas VAE methods just use one sample at a time. What does this imply in the results? I guess as long as VAEs outperform LFADS it is ok, but if LFADS outperforms VAEs in a given metric, could it be because it received more data as input (a whole segment)? Why was the factor dimension set to 50? I presume it was to match the latent dimension of the VAE methods, but is the LFADS factor dimension the correct match for that to make things comparable?

I am also surprised by the results. How do the authors justify LFADS having lower neural similarity (fig 2d) than VAE methods that operate on single time steps? LFADS is not supervised by behavior, so of course I don't expect it to necessarily outperform methods on behavior decoding. But all LFADS aims to do is to reconstruct the neural data so at least in this metric it should be able to outperform VAEs that just operate on single time steps? Is it because LFADS smooths the data too much? This is important to discuss and show examples of. These are all critical nuances that need to be discussed to validate the results and interpret them.”

Regarding “Was the moving averaged smoothed data passed to LFADS or the raw spiking data (at what bin size)? Was a gaussian loss used or a poisson loss?”: The data used by all models was applied to the same preprocessing procedure. That is, using moving averaged smoothed data with three bins, where the bin size is 100ms. For all models except PSID, we used a Poisson loss.

Regrading “What are the trial lengths used in each dataset, from which part of trials? For dataset C that has back-to-back reaches, was data chopped into segments? How long were these segments? Were the edges of segments overlapped and averaged as in (Keshtkaran et al. 2022) to avoid noisy segment edges or not?”:

For datasets A and B, a trial length of eighteen is set. Trials with lengths below the threshold are zero-padded, while trials exceeding the threshold are truncated to the threshold length from their starting point. In dataset A, there are several trials with lengths considerably longer than that of most trials. We found that padding all trials with zeros to reach the maximum length (32) led to poor performance. Consequently, we chose a trial length of eighteen, effectively encompassing the durations of most trials and leading to the removal of approximately 9% of samples. For dataset B (center-out), the trial lengths are relatively consistent with small variation, and the maximum length across all trials is eighteen. For dataset C, we set the trial length as ten because we observed the video of this paradigm and found that the time for completing a single trial was approximately one second. The segments are not overlapped.

Regarding “Why was the factor dimension set to 50? I presume it was to match the latent dimension of the VAE methods, but is the LFADS factor dimension the correct match for that to make things comparable?”: We performed a grid search for latent dimensions in {10,20,50} and found 50 is the best.

Regarding “I am also surprised by the results. How do the authors justify LFADS having lower neural similarity (fig 2d) than VAE methods that operate on single time steps? LFADS is not supervised by behavior, so of course I don't expect it to necessarily outperform methods on behavior decoding. But all LFADS aims to do is to reconstruct the neural data so at least in this metric it should be able to outperform VAEs that just operate on single time steps? Is it because LFADS smooths the data too much?”: As you pointed out, we found that LFADS tends to produce excessively smooth and consistent data, which can lead to a reduction in neural similarity.

Thank you for your valuable feedback.

Q1.3: “PSID: PSID is linear and uses past input samples to predict the next sample in the output. Again, some setup choices are not well justified, and some details are left out in the 1-line explanation given in Methods.

Why was a latent dimension of 6 chosen? Is this the behaviorally relevant latent dimension or the total latent dimension (for the use case here it would make sense to set all latent states to be behaviorally relevant)? Why was a horizon hyperparameter of 3 chosen? First, it is important to mention fundamental parameters such as latent dimension for each method in the main text (not just in methods) to make the results interpretable. Second, these hyperparameters should be chosen with a grid search in each dataset (within the training data, based on performance on the validation part of the training data), just as the authors do for their method (line 779). Given that PSID isn't a deep learning method, doing a thorough grid search in each fold should be quite feasible. It is important that high values for latent dimension and a wider range of other hyperparmeters are included in the search, because based on how well the residuals (x_i) for this method are shown predict behavior in Fig 2, the method seems to not have been used appropriately. I would expect ANN to improve decoding for PSID versus its KF decoding since PSID is fully linear, but I don't expect KF to be able to decode so well using the residuals of PSID if the method is used correctly to extract all behaviorally relevant information from neural data. The low neural reconstruction in Fid 2d could also partly be due to using too small of a latent dimension.

Again, another import nuance is the input to this method and how differs with the input to VAE methods. The learned PSID model is a filter that operates on all past samples of input to predict the output in the "next" time step. To enable a fair comparison with VAE methods, the authors should make sure that the last sample "seen" by PSID is the same as then input sample seen by VAE methods. This is absolutely critical given how large the time steps are, otherwise PSID might underperform simply because it stopped receiving input 300ms earlier than the input received by VAE methods. To fix this, I think the authors can just shift the training and testing neural time series of PSID by 1 sample into the past (relative to the behavior), so that PSID's input would include the input of VAE methods. Otherwise, VAEs outperforming PSID is confounded by PSID's input not including the time step that was provided to VAE.”

Thanks for your suggestions for letting PSID see the current neural observations. We did it per your suggestions and then performed a grid search for the hyperparameters for PSID. Specifically, we performed a grid search for the horizon hyperparameter in {2,3,4,5,6,7}. Since the relevant latent dimension should be lower than the horizon times the dimension of behavior variables (two-dimensional velocity in this paper) and increasing the dimension will reach performance saturation, we directly set the relevant latent dimensions as the maximum. The horizon number of datasets A, B, C, and synthetic datasets is 7, 6, 6 and 5, respectively.

And thus the latent dimension of datasets A, B, and C and the synthetic dataset is 14, 12, 12 and 10, respectively.

Our experiments show that KF can decode information from irrelevant signals obtained by PSID. Although PSID extracts the linear part of raw signals, KF can still use the linear part of the residuals for decoding. The low reconstruction performance of PSID may be because the relationship between latent variables and neural signals is linear, and the relationship between latent variables and behaviors is also linear; this is equivalent to the linear relationship between behaviors and neural signals, and linear models can only explain a small fraction of neural signals.

Thank you for your valuable feedback.

Q1.4: “CEBRA: results for CEBRA are incomplete. Similarity to raw signals is not shown. Decoding of behaviorally irrelevant residuals for CEBRA is not shown. Per Fig. S2, CEBRA does better or similar ANN decoding in datasets A and C, is only slightly worse in Dataset B, so it is important to show the other key metrics otherwise it is unclear whether d-VAE has some tangible advantage over CEBRA in those 2 datasets or if they are similar in every metric. Finally, it would be better if the authors show the results for CEBRA on Fig. 2, just as is done for other methods because otherwise it is hard to compare all methods.”

CEBRA is a non-generative model, this model cannot generate behaviorally-relevant signals. Therefore, we only compared the decoding performance of latent embeddings of CEBRA and signals of d-VAE.

Thank you for your valuable feedback.

Q2: “Given the fact that d-VAE infers the latent (z) based on the population activity (x), claims about properties of the inferred behaviorally relevant signals (x_r) that attribute properties to individual neurons are confounded.

The authors contrast their approach to population level approaches in that it infers behaviorally relevant signals for individual neurons. However, d-VAE is also a population method as it aggregates population information to infer the latent (z), from which behaviorally relevant part of the activity of each neuron (x_r) is inferred. The authors note this population level aggregation of information as a benefit of d-VAE, but only acknowledge it as a confound briefly in the context of one of their analyses (line 340): "The first is that the larger R2 neurons leak their information to the smaller R2 neurons, causing them contain too much behavioral information". They go on to dismiss this confounding possibility by showing that the inferred behaviorally relevant signal of each neuron is often most similar to its own raw signals (line 348-352) compared with all other neurons. They also provide another argument specific to that result section (i.e., residuals are not very behavior predictive), which is not general so I won't discuss it in depth here. These arguments however do not change the basic fact that d-VAE aggregates information from other neurons when extracting the behaviorally relevant activity of any given neuron, something that the authors note as a benefit of d-VAE in many instances. The fact that d-VAE aggregates population level info to give the inferred behaviorally relevant signal for each neuron confounds several key conclusions. For example, because information is aggregated across neurons, when trial to trial variability looks smoother after applying d-VAE (Fig 3i), or reveals better cosine tuning (Fig 3b), or when neurons that were not very predictive of behavior become more predictive of behavior (Fig 5), one cannot really attribute the new smoother single trial activity or the improved decoding to the same single neurons; rather these new signals/performances include information from other neurons. Unless the connections of the encoder network (z=f(x)) is zero for all other neurons, one cannot claim that the inferred rates for the neuron are truly solely associated with that neuron. I believe this a fundamental property of a population level VAE, and simply makes the architecture unsuitable for claims regarding inherent properties of single neurons. This confound is partly why the first claim in the abstract are not supported by data: observing that neurons that don't predict behavior very well would predict it much better after applying d-VAE does not prove that these neurons themselves "encode rich[er] behavioral information in complex nonlinear ways" (i.e., the first conclusion highlighted in the abstract) because information was also aggregated from other neurons. The other reason why this claim is not supported by data is the characterization of the encoding for smaller R2 neurons as "complex nonlinear", which the method is not well equipped to tease apart from linear mappings as I explain in my comment 3.”

We acknowledge that we cannot obtain the exact single neuronal activity that does not contain any information from other neurons. However, we believe our model can extract accurate approximation signals of the ground truth relevant signals. These signals preserve the inherent properties of single neuronal activity to some extent and can be used for analysis at the single-neuron level.

We believe d-VAE is a reasonable approach to extract effective relevant signals that preserve inherent properties of single neuronal activity for four key reasons:

1) d-VAE is a latent variable model that adheres to the neural population doctrine. The neural population doctrine posits that information is encoded within interconnected groups of neurons, with the existence of latent variables (neural modes) responsible for generating observable neuronal activity [1, 2]. If we can perfectly obtain the true generative model from latent variables to neuronal activity, then we can generate the activity of each neuron from hidden variables without containing any information from other neurons. However, without a complete understanding of the brain’s encoding strategies (or generative model), we can only get the approximation signals of the ground truth signals.

2) After the generative model is established, we need to infer the parameters of the generative model and the distribution of latent variables. During the inference process, inference algorithms such as variational inference or EM algorithms will be used. Generally, the obtained latent variables are also approximations of the real latent variables. When inferring the latent variables, it is inevitable to aggregation the information of the neural population, and latent variables are derived through weighted combinations of neuronal populations [3].

This inference process is consistent with that of d-VAE (or VAE-based models).

3) Latent variables are derived from raw neural signals and used to explain raw neural signals. Considering the unknown ground truth of latent variables and behaviorally-relevant signals, it becomes evident that the only reliable reference at the signal level is the raw signals. A crucial criterion for evaluating the reliability of latent variable models (including latent variables and generated relevant signals) is their capability to effectively explain the raw signals [3]. Consequently, we firmly maintain the belief that if the generated signals closely resemble the raw signals to the greatest extent possible, in accordance with an equivalence principle, we can claim that these obtained signals faithfully retain the inherent properties of single neurons. d-VAE explicitly constrains the generated signal to closely resemble the raw signals. These results demonstrate that d-VAE can extract effective relevant signals that preserve inherent properties of single neuronal activity.

Based on the above reasons, we hold that generating single neuronal activities with the VAE framework is a reasonable approach. The remaining question is whether our model can obtain accurate relevant signals in the absence of ground truth. To our knowledge, in cases where the ground truth of relevant signals is unknown, there are typically two approaches to verifying the reliability of extracted signals:

1) Conducting synthetic experiments where the ground truth is known.

2) Validation based on expert knowledge (Three criteria were proposed in this paper). Both our extracted signals and key conclusions have been validated using these two approaches.

Next, we will provide a detailed response to the concerns regarding our first key conclusion that smaller R2 neurons encode rich information.

We acknowledge that larger R2 neurons play a role in aiding the reconstruction of signals in smaller R2 neurons through their neural activity. However, considering that neurons are correlated rather than independent entities, we maintain the belief that larger R2 neurons assist damaged smaller R2 neurons in restoring their original appearance. Taking image denoising as an example, when restoring noisy pixels to their original appearance, relying solely on the noisy pixels themselves is often impractical. Assistance from their correlated, clean neighboring pixels becomes necessary.

The case we need to be cautious of is that the larger R2 neurons introduce additional signals (m) that contain substantial information to smaller R2 neurons, which they do not inherently possess. We believe this case does not hold for two reasons. Firstly, logically, adding extra signals decreases the reconstruction performance, and the information carried by these additional signals is redundant for larger R2 neurons, thus they do not introduce new information that can enhance the decoding performance of the neural population. Therefore, it seems unlikely and unnecessary for neural networks to engage in such counterproductive actions. Secondly, even if this occurs, our second criterion can effectively exclude the selection of these signals. To clarify, if we assume that x, y, and z denote the raw, relevant, and irrelevant signals of smaller R2 neurons, with x=y+z, and the extracted relevant signals become y+m, the irrelevant signals become z-m in this case. Consequently, the irrelevant signals contain a significant amount of information. It's essential to emphasize that this criterion holds significant importance in excluding undesirable signals.

Furthermore, we conducted a synthetic experiment to show that d-VAE can indeed restore the damaged information of smaller R2 neurons with the help of larger R2 neurons, and the restored neuronal activities are more similar to ground truth compared to damaged raw signals. Please see Fig. S11a,b for details.

Thank you for your valuable feedback.

[1] Saxena, S. and Cunningham, J.P., 2019. Towards the neural population doctrine. Current opinion in neurobiology, 55, pp.103-111.

[2] Gallego, J.A., Perich, M.G., Miller, L.E. and Solla, S.A., 2017. Neural manifolds for the control of movement. Neuron, 94(5), pp.978-984.

[3] Cunningham, J.P. and Yu, B.M., 2014. Dimensionality reduction for large-scale neural recordings. Nature neuroscience, 17(11), pp.1500-1509.

Q3: “Given the nonlinear architecture of the VAE, claims about the linearity or nonlinearity of cortical readout are confounded and not supported by the results.

The inference of behaviorally relevant signals from raw signals is a nonlinear operation, that is x_r=g(f(x)) is nonlinear function of x. So even when a linear KF is used to decode behavior from the inferred behaviorally relevant signals, the overall decoding from raw signals to predicted behavior (i.e., KF applied to g(f(x))) is nonlinear. Thus, the result that decoding of behavior from inferred behaviorally relevant signals (x_r) using a linear KF and a nonlinear ANN reaches similar accuracy (Fig 2), does not suggest that a "linear readout is performed in the motor cortex", as the authors claim (line 471). The authors acknowledge this confound (line 472) but fail to address it adequately. They perform a simulation analysis where the decoding gap between KF and ANN remains unchanged even when d-VAE is used to infer behaviorally relevant signals in the simulation. However, this analysis is not enough for "eliminating the doubt" regarding the confound. I'm sure the authors can also design simulations where the opposite happens and just like in the data, d-VAE can improve linear decoding to match ANN decoding. An adequate way to address this concern would be to use a fully linear version of the autoencoder where the f(.) and g(.) mappings are fully linear. They can simply replace these two networks in their model with affine mappings, redo the modeling and see if the model still helps the KF decoding accuracy reach that of the ANN decoding. In such a scenario, because the overall KF decoding from original raw signals to predicted behavior (linear d-VAE + KF) is linear, then they could move toward the claim that the readout is linear. Even though such a conclusion would still be impaired by the nonlinear reference (d-VAE + ANN decoding) because the achieved nonlinear decoding performance could always be limited by network design and fitting issues. Overall, the third conclusion highlighted in the abstract is a very difficult claim to prove and is unfortunately not supported by the results.”

We aim to explore the readout mechanism of behaviorally-relevant signals, rather than raw signals. Theoretically, the process of removing irrelevant signals should not be considered part of the inherent decoding mechanisms of the relevant signals. Assuming that the relevant signals we extracted are accurate, the conclusion of linear readout is established. On the synthetic data where the ground truth is known, our distilled signals show a significant improvement in neural similarity to the ground truth when compared to raw signals (refer to Fig. S2l). This observation demonstrates that our distilled signals are accurate approximations of the ground truth. Furthermore, on the three widely-used real datasets, our distilled signals meet the stringent criteria we have proposed (see Fig. 2), also providing strong evidence for their accuracy.

Regarding the assertion that we could create simulations in which d-VAE can make signals that are inherently nonlinearly decodable into linearly decodable ones: In reality, we cannot achieve this, as the second criterion can rule out the selection of such signals. Specifically,z=x+y=n^2+y, where z, x, y, and n denote raw signals, relevant signals, irrelevant signals and latent variables. If the relevant signals obtained by d-VAE are n, then these signals can be linear decoded accurately. However, the corresponding irrelevant signals are n^2-n+z; thus, irrelevant signals will have much information, and these extracted relevant signals will not be selected. Furthermore, our synthetic experiments offer additional evidence supporting the conclusion that d-VAE does not make inherently nonlinearly decodable signals become linearly decodable ones. As depicted in Fig. S11c, there exists a significant performance gap between KF and ANN when decoding the ground truth signals of smaller R2 neurons. KF exhibits notably low performance, leaving substantial room for compensation by d-VAE. However, following processing by d-VAE, KF's performance of distilled signals fails to surpass its already low ground truth performance and remains significantly inferior to ANN's performance. These results collectively confirm that our approach does not convert signals that are inherently nonlinearly decodable into linearly decodable ones, and the conclusion of linear readout is not a by-product by d-VAE.

Regarding the suggestion of using linear d-VAE + KF, as discussed in the Discussion section, removing the irrelevant signals requires a nonlinear operation, and linear d-VAE can not effectively separate relevant and irrelevant signals.

Thank you for your valuable feedback.

Q4: “The authors interpret several results as indications that "behavioral information is distributed in a higher-dimensional subspace than expected from raw signals", which is the second main conclusion highlighted in the abstract. However, several of these arguments do not convincingly support that conclusion.

4.1) The authors observe that behaviorally relevant signals for neurons with small principal components (referred to as secondary) have worse decoding with KF but better decoding with ANN (Fig. 6b,e), which also outperforms ANN decoding from raw signals. This observation is taken to suggest that these secondary behaviorally relevant signals encode behavior information in highly nonlinear ways and in a higher dimensions neural space than expected (lines 424 and 428). These conclusions however are confounded by the fact that A) d-VAE uses nonlinear encoding, so one cannot conclude from ANN outperforming KF that behavior is encoded nonlinearly in the motor cortex (see comment 3 above), and B) d-VAE aggregates information across the population so one cannot conclude that these secondary neurons themselves had as much behavior information (see comment 2 above).

4.2) The authors observe that the addition of the inferred behaviorally relevant signals for neurons with small principal components (referred to as secondary) improves the decoding of KF more than it improves the decoding of ANN (red curves in Fig 6c,f). This again is interpreted similarly as in 4.1, and is confounded for similar reasons (line 439): "These results demonstrate that irrelevant signals conceal the smaller variance PC signals, making their encoded information difficult to be linearly decoded, suggesting that behavioral information exists in a higher-dimensional subspace than anticipated from raw signals". This is confounded by because of the two reasons explained in 4.1. To conclude nonlinear encoding based on the difference in KF and ANN decoding, the authors would need to make the encoding/decoding in their VAE linear to have a fully linear decoder on one hand (with linear d-VAE + KF) and a nonlinear decoder on the other hand (with linear d-VAE + ANN), as explained in comment 3.

4.3) From S Fig 8, where the authors compare cumulative variance of PCs for raw and inferred behaviorally relevant signals, the authors conclude that (line 554): "behaviorally-irrelevant signals can cause an overestimation of the neural dimensionality of behaviorally-relevant responses (Supplementary Fig. S8)." However, this analysis does not really say anything about overestimation of "behaviorally relevant" neural dimensionality since the comparison is done with the dimensionality of "raw" signals. The next sentence is ok though: "These findings highlight the need to filter out relevant signals when estimating the neural dimensionality.", because they use the phrase "neural dimensionality" not "neural dimensionality of behaviorally-relevant responses".”

Questions 4.1 and 4.2 are a combination of Q2 and Q3. Please refer to our responses to Q2 and Q3.

Regarding question 4.3 about “behaviorally-irrelevant signals can cause an overestimation of the neural dimensionality of behaviorally-relevant responses”: Previous studies usually used raw signals to estimate the neural dimensionality of specific behaviors. We mean that using raw signals, which include many irrelevant signals, will cause an overestimation of the neural dimensionality. We have modified this sentence in the revised manuscripts.

Thank you for your valuable feedback.

Q5: “Imprecise use of language in many places leads to inaccurate statements. I will list some of these statements”

5.1) In the abstract: "One solution is to accurately separate behaviorally-relevant and irrelevant signals, but this approach remains elusive due to the unknown ground truth of behaviorally-relevant signals". This statement is not accurate because it implies no prior work does this. The authors should make their statement more specific and also refer to some goal that existing linear (e.g., PSID) and nonlinear (e.g., TNDM) methods for extracting behaviorally relevant signals fail to achieve.

5.2) In the abstract: "we found neural responses previously considered useless encode rich behavioral information" => what does "useless" mean operationally? Low behavior tuning? More precise use of language would be better.

5.3) "... recent studies (Glaser 58 et al., 2020; Willsey et al., 2022) demonstrate nonlinear readout outperforms linear readout." => do these studies show that nonlinear "readout" outperforms linear "readout", or just that nonlinear models outperform linear models?

5.4) Line 144: "The first criterion is that the decoding performance of the behaviorally-relevant signals (red bar, Fig.1) should surpass that of raw signals (the red dotted line, Fig.1).". Do the authors mean linear decoding here or decoding in general? If the latter, how can something extracted from neural surpass decoding of neural data, when the extraction itself can be thought of as part of decoding? The operational definition for this "decoding performance" should be clarified.

5.5) Line 311: "we found that the dimensionality of primary subspace of raw signals (26, 64, and 45 for datasets A, B, and C) is significantly higher than that of behaviorally-relevant signals (7, 13, and 9), indicating that behaviorally-irrelevant signals lead to an overestimation of the neural dimensionality of behaviorally-relevant signals." => here the dimensionality of the total PC space (i.e., primary subspace of raw signals) is being compared with that of inferred behaviorally-relevant signals, so the former being higher does not indicate that neural dimensionality of behaviorally-relevant signals was overestimated. The former is simply not behavioral so this conclusion is not accurate.

5.6) Section "Distilled behaviorally-relevant signals uncover that smaller R2 neurons encode rich behavioral information in complex nonlinear ways". Based on what kind of R2 are the neurons grouped? Behavior decoding R2 from raw signals? Using what mapping? Using KF? If KF is used, the result that small R2 neurons benefit a lot from d-VAE could be somewhat expected, given the nonlinearity of d-VAE: because only ANN would have the capacity to unwrap the nonlinear encoding of d-VAE as needed. If decoding performance that is used to group neurons is based on data, regression to the mean could also partially explain the result: the neurons with worst raw decoding are most likely to benefit from a change in decoder, than neurons that already had good decoding. In any case, the R2 used to partition and sort neurons should be more clearly stated and reminded throughout the text and I Fig 3.

5.7) Line 346 "...it is impossible for our model to add the activity of larger R2 neurons to that of smaller R2 neurons" => Is it really impossible? The optimization can definitely add small-scale copies of behaviorally relevant information to all neurons with minimal increase in the overall optimization loss, so this statement seems inaccurate.

5.8) Line 490: "we found that linear decoders can achieve comparable performance to that of nonlinear decoders, providing compelling evidence for the presence of linear readout in the motor cortex." => inaccurate because no d-VAE decoding is really linear, as explained in comment 3 above.

5.9) Line 578: ". However, our results challenge this idea by showing that signals composed of smaller variance PCs nonlinearly encode a significant amount of behavioral information." => inaccurate as results are confounded by nonlinearity of d-VAE as explained in comment 3 above.

5.10) Line 592: "By filtering out behaviorally-irrelevant signals, our study found that accurate decoding performance can be achieved through linear readout, suggesting that the motor cortex may perform linear readout to generate movement behaviors." => inaccurate because it us confounded by the nonlinearity of d-VAE as explained in comment 3 above.”

Regarding “5.1) In the abstract: "One solution is to accurately separate behaviorally-relevant and irrelevant signals, but this approach remains elusive due to the unknown ground truth of behaviorally-relevant signals". This statement is not accurate because it implies no prior work does this. The authors should make their statement more specific and also refer to some goal that existing linear (e.g., PSID) and nonlinear (e.g., TNDM) methods for extracting behaviorally relevant signals fail to achieve”:

We believe our statement is accurate. Our primary objective is to extract accurate behaviorally-relevant signals that closely approximate the ground truth relevant signals. To achieve this, we strike a balance between the reconstruction and decoding performance of the generated signals, aiming to effectively capture the relevant signals. This crucial aspect of our approach sets it apart from other methods. In contrast, other methods tend to emphasize the extraction of valuable latent neural dynamics. We have provided elaboration on the distinctions between d-VAE and other approaches in the Introduction and Discussion sections.

Thank you for your valuable feedback.

Regarding “5.2) In the abstract: "we found neural responses previously considered useless encode rich behavioral information" => what does "useless" mean operationally? Low behavior tuning? More precise use of language would be better.”:

In the analysis of neural signals, smaller variance PC signals are typically seen as noise and are often discarded. Similarly, smaller R2 neurons are commonly thought to be dominated by noise and are not further analyzed. Given these considerations, we believe that the term "considered useless" is appropriate in this context. Thank you for your valuable feedback.

Regarding “5.3) "... recent studies (Glaser 58 et al., 2020; Willsey et al., 2022) demonstrate nonlinear readout outperforms linear readout." => do these studies show that nonlinear "readout" outperforms linear "readout", or just that nonlinear models outperform linear models?”:

In this paper, we consider the two statements to be equivalent. Thank you for your valuable feedback.

Regarding “5.4) Line 144: "The first criterion is that the decoding performance of the behaviorally-relevant signals (red bar, Fig.1) should surpass that of raw signals (the red dotted line, Fig.1).". Do the authors mean linear decoding here or decoding in general? If the latter, how can something extracted from neural surpass decoding of neural data, when the extraction itself can be thought of as part of decoding? The operational definition for this "decoding performance" should be clarified.”:

We mean the latter, as we said in the section “Framework for defining, extracting, and separating behaviorally-relevant signals”, since raw signals contain too many behaviorally-irrelevant signals, deep neural networks are more prone to overfit raw signals than relevant signals. Therefore the decoding performance of relevant signals should surpass that of raw signals. Thank you for your valuable feedback.

Regarding “5.5) Line 311: "we found that the dimensionality of primary subspace of raw signals (26, 64, and 45 for datasets A, B, and C) is significantly higher than that of behaviorally-relevant signals (7, 13, and 9), indicating that behaviorally-irrelevant signals lead to an overestimation of the neural dimensionality of behaviorally-relevant signals." => here the dimensionality of the total PC space (i.e., primary subspace of raw signals) is being compared with that of inferred behaviorally-relevant signals, so the former being higher does not indicate that neural dimensionality of behaviorally-relevant signals was overestimated. The former is simply not behavioral so this conclusion is not accurate.”: In practice, researchers usually used raw signals to estimate the neural dimensionality. We mean that using raw signals to do this would overestimate the neural dimensionality. Thank you for your valuable feedback.

Regarding “5.6) Section "Distilled behaviorally-relevant signals uncover that smaller R2 neurons encode rich behavioral information in complex nonlinear ways". Based on what kind of R2 are the neurons grouped? Behavior decoding R2 from raw signals? Using what mapping? Using KF? If KF is used, the result that small R2 neurons benefit a lot from d-VAE could be somewhat expected, given the nonlinearity of d-VAE: because only ANN would have the capacity to unwrap the nonlinear encoding of d-VAE as needed. If decoding performance that is used to group neurons is based on data, regression to the mean could also partially explain the result: the neurons with worst raw decoding are most likely to benefit from a change in decoder, than neurons that already had good decoding. In any case, the R2 used to partition and sort neurons should be more clearly stated and reminded throughout the text and I Fig 3.”:

When employing R2 to characterize neurons, it indicates the extent to which neuronal activity is explained by the linear encoding model [1-3]. Smaller R2 neurons have a lower capacity for linearly tuning (encoding) behaviors, while larger R2 neurons have a higher capacity for linearly tuning (encoding) behaviors. Specifically, the approach involves first establishing an encoding relationship from velocity to neural signal using a linear model, i.e., y=f(x), where f represents a linear regression model, x denotes velocity, and y denotes the neural signal. Subsequently, R2 is utilized to quantify the effectiveness of the linear encoding model in explaining neural activity. We have provided a comprehensive explanation in the revised manuscript. Thank you for your valuable feedback.

[1] Collinger, J.L., Wodlinger, B., Downey, J.E., Wang, W., Tyler-Kabara, E.C., Weber, D.J., McMorland, A.J., Velliste, M., Boninger, M.L. and Schwartz, A.B., 2013. High-performance neuroprosthetic control by an individual with tetraplegia. The Lancet, 381(9866), pp.557-564.

[2] Wodlinger, B., et al. "Ten-dimensional anthropomorphic arm control in a human brain− machine interface: difficulties, solutions, and limitations." Journal of neural engineering 12.1 (2014): 016011.

[3] Inoue, Y., Mao, H., Suway, S.B., Orellana, J. and Schwartz, A.B., 2018. Decoding arm speed during reaching. Nature communications, 9(1), p.5243.

Regarding Questions 5.7, 5.8, 5.9, and 5.10:

We believe our conclusions are solid. The reasons can be found in our replies in Q2 and Q3. Thank you for your valuable feedback.

Q6: “Imprecise use of language also sometimes is not inaccurate but just makes the text hard to follow.

6.1) Line 41: "about neural encoding and decoding mechanisms" => what is the definition of encoding/decoding and how do these differ? The definitions given much later in line 77-79 is also not clear.

6.2) Line 323: remind the reader about what R2 is being discussed, e.g., R2 of decoding behavior using KF. It is critical to know if linear or nonlinear decoding is being discussed.

6.3) Line 488: "we found that neural responses previously considered trivial encode rich behavioral information in complex nonlinear ways" => "trivial" in what sense? These phrases would benefit from more precision, for example: "neurons that may seem to have little or no behavior information encoded". The same imprecise word ("trivial") is also used in many other places, for example in the caption of Fig S9.

6.4) Line 611: "The same should be true for the brain." => Too strong of a statement for an unsupported claim suggesting the brain does something along the lines of nonlin VAE + linear readout.

6.5) In Fig 1, legend: what is the operational definition of "generating performance"? Generating what? Neural reconstruction?”

Regarding “6.1) Line 41: "about neural encoding and decoding mechanisms" => what is the definition of encoding/decoding and how do these differ? The definitions given much later in line 77-79 is also not clear.”:

We would like to provide a detailed explanation of neural encoding and decoding. Neural encoding means how neuronal activity encodes the behaviors, that is, y=f(x), where y denotes neural activity and, x denotes behaviors, f is the encoding model. Neural decoding means how the brain decodes behaviors from neural activity, that is, x=g(y), where g is the decoding model. For further elaboration, please refer to [1]. We have included references that discuss the concepts of encoding and decoding in the revised manuscript. Thank you for your valuable feedback.

[1] Kriegeskorte, Nikolaus, and Pamela K. Douglas. "Interpreting encoding and decoding models." Current opinion in neurobiology 55 (2019): 167-179.

Regarding “6.2) Line 323: remind the reader about what R2 is being discussed, e.g., R2 of decoding behavior using KF. It is critical to know if linear or nonlinear decoding is being discussed.”:

This question is the same as Q5.6. Please refer to the response to Q5.6. Thank you for your valuable feedback.

Regarding “6.3) Line 488: "we found that neural responses previously considered trivial encode rich behavioral information in complex nonlinear ways" => "trivial" in what sense? These phrases would benefit from more precision, for example: "neurons that may seem to have little or no behavior information encoded". The same imprecise word ("trivial") is also used in many other places, for example in the caption of Fig S9.”:

We have revised this statement in the revised manuscript. Thanks for your recommendation.

Regarding “6.4) Line 611: "The same should be true for the brain." => Too strong of a statement for an unsupported claim suggesting the brain does something along the lines of nonlin VAE + linear readout.”

We mean that removing the interference of irrelevant signals and decoding the relevant signals should logically be two stages. We have revised this statement in the revised manuscript. Thank you for your valuable feedback.

Regarding “6.5) In Fig 1, legend: what is the operational definition of "generating performance"? Generating what? Neural reconstruction?””:

We have replaced “generating performance” with “reconstruction performance” in the revised manuscript. Thanks for your recommendation.

Q7: “In the analysis presented starting in line 449, the authors compare improvement gained for decoding various speed ranges by adding secondary (small PC) neurons to the KF decoder (Fig S11). Why is this done using the KF decoder, when earlier results suggest an ANN decoder is needed for accurate decoding from these small PC neurons? It makes sense to use the more accurate nonlinear ANN decoder to support the fundamental claim made here, that smaller variance PCs are involved in regulating precise control”

Because when the secondary signal is superimposed on the primary signal, the enhancement in KF performance is substantial. We wanted to explore in which aspect of the behavior the KF performance improvement is mainly reflected. In comparison, the improvement of ANN by the secondary signal is very small, rendering the exploration of the aforementioned questions inconsequential. Thank you for your valuable feedback.

Q8: “A key limitation of the VAE architecture is that it doesn't aggregate information over multiple time samples. This may be why the authors decided to use a very large bin size of 100ms and beyond that smooth the data with a moving average. This limitation should be clearly stated somewhere in contrast with methods that can aggregate information over time (e.g., TNDM, LFADS, PSID) ”

We have added this limitation in the Discussion in the revised manuscript. Thanks for your recommendation.

Q9: “Fig 5c and parts of the text explore the decoding when some neurons are dropped. These results should come with a reminder that dropping neurons from behaviorally relevant signals is not technically possible since the extraction of behaviorally relevant signals with d-VAE is a population level aggregation that requires the raw signal from all neurons as an input. This is also important to remind in some places in the text for example:

• Line 498: "...when one of the neurons is destroyed."

• Line 572: "In contrast, our results show that decoders maintain high performance on distilled signals even when many neurons drop out."”

We want to explore the robustness of real relevant signals in the face of neuron drop-out. The signals our model extracted are an approximation of the ground truth relevant signals and thus serve as a substitute for ground truth to study this problem. Thank you for your valuable feedback.

Q10: “Besides the confounded conclusions regarding the readout being linear (see comment 3 and items related to it in comment 5), the authors also don't adequately discuss prior works that suggest nonlinearity helps decoding of behavior from the motor cortex. Around line 594, a few works are discussed as support for the idea of a linear readout. This should be accompanied by a discussion of works that support a nonlinear encoding of behavior in the motor cortex, for example (Naufel et al. 2019; Glaser et al. 2020), some of which the authors cite elsewhere but don't discuss here.”

We have added this discussion in the revised manuscript. Thanks for your recommendation.

Q11: “Selection of hyperparameters is not clearly explained. Starting line 791, the authors give some explanation for one hyperparameter, but not others. How are the other hyperparameters determined? What is the search space for the grid search of each hyperparameter? Importantly, if hyperparameters are determined only based on the training data of each fold, why is only one value given for the hyperparameter selected in each dataset (line 814)? Did all 5 folds for each dataset happen to select exactly the same hyperparameter based on their 5 different training/validation data splits? That seems unlikely.”

We perform a grid search in {0.001, 0.01,0.1,1} for hyperparameter beta. And we found that 0.001 is the best for all datasets. As for the model parameters, such as hidden neuron numbers, this model capacity has reached saturation decoding performance and does not influence the results.

Regarding “Importantly, if hyperparameters are determined only based on the training data of each fold, why is only one value given for the hyperparameter selected in each dataset (line 814)? Did all 5 folds for each dataset happen to select exactly the same hyperparameter based on their 5 different training/validation data splits”: We selected the hyperparameter based on the average performance of 5 folds data on validation sets. The selected value denotes the one that yields the highest average performance across the 5 folds data.

Thank you for your valuable feedback.

Q12: “d-VAE itself should also be explained more clearly in the main text. Currently, only the high-level idea of the objective is explained. The explanation should be more precise and include the idea of encoding to latent state, explain the relation to pip-VAE, explain inputs and outputs, linearity/nonlinearity of various mappings, etc. Also see comment 1 above, where I suggest adding more details about other methods in the main text.”

Our primary objective is to delve into the encoding and decoding mechanisms using the separated relevant signals. Therefore, providing an excessive amount of model details could potentially distract from the main focus of the paper. In response to your suggestion, we have included a visual representation of d-VAE's structure, input, and output (see Fig. S1) in the revised manuscript, which offers a comprehensive and intuitive overview. Additionally, we have expanded on the details of d-VAE and other methods in the Methods section.

Thank you for your valuable feedback.

Q13: “In Fig 1f and g, shouldn't the performance plots be swapped? The current plots seem counterintuitive. If there is bias toward decoding (panel g), why is the irrelevant residual so good at decoding?”

The placement of the performance plots in Fig. 1f and 1g is accurate. When the model exhibits a bias toward decoding, it prioritizes extracting the most relevant features (latent variables) for decoding purposes. As a consequence, the model predominantly generates signals that are closely associated with these extracted features. This selective signal extraction and generation process may result in the exclusion of other potentially useful information, which will be left in the residuals. To illustrate this concept, consider the example of face recognition: if a model can accurately identify an individual using only the person's eyes (assuming these are the most useful features), other valuable information, such as details of the nose or mouth, will be left in the residuals, which could also be used to identify the individual.

Thank you for your valuable feedback.

Author Response

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

Summary:

In this interesting work, the authors investigated an important topical question: when we see travelling waves in cortical activity, is this due to true wave-like spread, or due to sequentially activated sources? In simulations, it is shown that sequential brain module activation can show up as a travelling wave - even in improved methods such as phase delay maps - and a variety of parameters is investigated. Then, in ex-vivo turtle eye-brain preparations, the authors show that visual cortex waves observable in local field potentials are in fact often better explained as areas D1 and D2 being sequentially activated. This has implications for how we think about travelling wave methodology and relevant analytical tools.

Strengths:

I enjoyed reading the discussion. The authors are careful in their claims, and point out that some phenomena may still indeed be genuine travelling waves, but we should have a higher evidence bar to claim this for a particular process in light of this paper and Zhigalov & Jensen (2023) (ref 44). Given this careful discussion, the claims made are well-supported by the experimental results. The discussion also gives a nice overview of potential options in light of this and future directions.

The illustration of different gaussian covariances leading to very different latency maps was interesting to see.

Furthermore, the methods are detailed and clearly structured and the Supplementary Figures, particularly single trial results, are useful and convincing.

We are glad the reviewer found our manuscript “interesting”, the questions we raise “important”, our claims “well-supported by the experimental results”, and our methods “detailed and clearly structured”.

The details of the sequentially activated Gaussian simulations give some useful results, but the fundamental idea still appears to be "sequential activation is often indistinguishable from a travelling wave", an idea advanced e.g. by Zhigalov & Jensen (2023). It takes a while until the (in my opinion) more intriguing experimental results.

To emphasize the experimental results, we switched between the analytical results and the experimental results. Correspondingly, figure 2 now illustrates the more intriguing experimental results and figure 3 the analytical results. In addition, we added subtitles to the different sections of the results to ease the navigation through the paper and to enable the readers to access the different sections more easily.

One of the key claims is that the spikes are more consistent with two sequentially activated modules rather than a continuous wave (with Fig 3k and 3l key to support this). Whilst this is more consistent, it is worth mentioning that there seems to be stochasticity to this and between-trial variability, especially for spikes.

In the revised manuscript we added the reviewer’s comment about stochasticity, and we discuss its possible origins:

"The transition was also not clear when examining spiking responses in some of the trials (as indicated by high DIP scores, Figure 2K). However, the observation that temporal grouping became more pronounced when using ALSA (a more robust estimate of local excitability) (Figure 2L,N), suggests that high DIP values may result from variability in the spike times of single neurons, and not necessarily from the lack of modular activation. Such issues can be resolved by denser sampling of spiking activity in the tissue."

Recommendations For The Authors:

The eye-cortex turtle preparation is not the most common. I would add more context about how specific the results are to this preparation vs how comparable it is to human data.

We added a sentence explaining the relevance of our preparation: “Finally, while the layered organization of turtle cortex is different than that of mammalian cortex, the basic excitability features of both tissues are similar (Connors and Kriegstein, 1986; Hemberger et al., 2019; Kriegstein and Connors, 1986; Larkum et al., 2008; Shein-Idelson et al., 2017b), and substantial differences in the manner by which field potentials and spikes spread through the tissue are not to be expected.”

Philosophical question: when does a 'module' become small enough for it to count as a travelling wave? More on this could be added to the discussion. I think we are in the very early days for a true understanding of travelling waves, and I wonder if these sequentially activated modules will functionally correspond to the known cortical segregation, or if it varies by area/task.

We agree with the reviewer that macroscopic waves could be composed of smaller modules (or single neurons at the smallest scale). Our results suggest that modular patterns can be classified as wave patterns both at large scales (of brain areas) and smaller scales of local neural circuits. Therefore, we believe it is necessary to make this distinction across different scales. We sharpened this point in the first paragraph of the discussion:

"…We showed that LFP measurements indicative of waves propagating across turtle cortex are underlined by discrete and consecutively activated neuronal populations, and not by a continuously propagating wavefront of spikes (Figure 2). Similarly, activation profiles that resemble continuous travelling waves in EEG simulations can be underlined by consecutive activation of two discrete cortical regions (Figure 1). We replicated these results using an analytical model and demonstrated that a simple scenario of sequentially activated Gaussians can exhibit WLPs with a rich diversity of spatiotemporal profiles (Figure 3). Our results offer insight into the scenarios and conditions for WLP detection by identifying failure points that should be considered when identifying travelling waves and therefore suggest caution when interpreting continuous phase latency maps as microscopically propagating wave patterns. Such failure points may exist both when examining activity at the scale of brain regions (Figure 1) and smaller neural circuits (Figure 2). Therefore, our results suggest that the discrepancy between modular and wave activation should be examined across spatial scales. Specifically, it is not necessarily the case that at the fine grained (single neuron) scale activation patterns are modular, but, following coarse graining, smooth wave patterns emerge. Rather, modular activation may hierarchically exist across scales (Kaiser and Hilgetag, 2010; Meunier et al., 2010) and may be masked by smeared spatial supra-threshold excitability boundaries. Below we discuss these limitations across techniques and their implications.”

I would advise the authors to focus on the experimental data, perhaps by putting the simulations second, and by putting some of the equation details that are in Methods into the Supplementary Information. Whilst the simulation parameter space is well-explored, the fundamental idea of spreading Gaussians is relatively simple, and the current manuscript organization detracted from the main message for me a little bit.”

Following the referee’s suggestion, we switched between the section with experimental data and the one with the analytic model (see response to comment 1). In addition, to ease the reading of the methods, we moved the mathematical derivation and related equations to appendix 1.

Things I thought about that you may also enjoy thinking about: Could we tell something about sequential sources vs travelling waves by the nature of the wave - e.g. shape or dispersion? If some wave properties are conserved whilst travelling, this could be evidence for travelling vs two sources.

This is a wonderful suggestion. We are currently working on a follow up publication with a new approach to do exactly that! We think that this new body of work is outside the scope of this paper.

Could synaptic potentials spread like waves, but spikes more in modular bursts? This would also explain the LFP vs spikes difference - maybe travelling waves of EPSPs are there priming the network, 'looking' for suitable modules to activate, which then activate sequentially. The current discussion is quite spike-focused - could some information be in synaptic potentials after all?

This is an interesting idea with intriguing functional implications. We added this idea to our discussion (see paragraph below). In addition, to emphasize our discussion on synaptic potentials, we reorganized the paragraphs in the discussion to separate between our discussion on sub-threshold excitability (which is mostly synaptic) and supra-threshold excitability which is the focus of the second part of the discussion.

“Variability in responses may also be explained by differences in propagation mechanisms (Ermentrout and Kleinfeld, 2001; Muller et al., 2018; Wu et al., 2008). Several reports suggest that waves are underlined by propagation along axonal collaterals (Muller et al., 2018, 2014). Both the transmembrane voltage-gated currents excited during action potentials as well as the post-synaptic currents along axonal boutons can potentially contribute to measured signals. However, such waves travel at high propagation speeds and are not compatible with the wide diversity of wave velocities and mechanisms of local neuronal interactions (Ermentrout and Kleinfeld, 2001; Feller et al., 1996). An intriguing possibility is that such axonal waves prime neuronal excitability by sub-threshold inputs that later result in modular supra-threshold activation. The ability to experimentally discriminate between axonal inputs and local spiking excitability (e.g. by reporters with different wavelengths) can potentially resolve such discrepancies.

Our turtle cortex results (Figure 2) exemplify how contrasting sub-threshold LFP measurements with supra-threshold spiking measurements can yield different conclusions about the nature of activity spread….”

Joint Public Review:

Summary:

In this interesting work, the authors investigated an important topical question: when we see travelling waves in cortical activity, is this due to true wave-like spread, or due to sequentially activated sources? In simulations, it is shown that sequential brain module activation can show up as a travelling wave - even in improved methods such as phase delay maps - and a variety of parameters is investigated. Then, in ex-vivo turtle eye-brain preparations, the authors show that visual cortex waves observable in local field potentials are in fact often better explained as areas D1 and D2 being sequentially activated. This has implications for how we think about travelling wave methodology and relevant analytical tools.

Strengths:

I enjoyed reading the discussion. The authors are careful in their claims, and point out that some phenomena may still indeed be genuine travelling waves, but we should have a higher evidence bar to claim this for a particular process in light of this paper and Zhigalov & Jensen (2023) (ref 44). Given this careful discussion, the claims made are well-supported by the experimental results. The discussion also gives a nice overview of potential options in light of this and future directions.

The illustration of different gaussian covariances leading to very different latency maps was interesting to see.

Furthermore, the methods are detailed and clearly structured and the Supplementary Figures, particularly single trial results, are useful and convincing.

Author Response:

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

Joint Public Review:

[…] While this does not rule out criticality in the brain, it decidedly weakens the evidence for it, which was based on the following logic: critical systems give rise to power law behavior; power law behavior is observed in cortical networks; therefore, cortical networks operate near a critical point. Given, as shown in this paper, that power laws can arise from noncritical processes, the logic breaks. Moreover, the authors show that criticality does not imply optimal information transmission (one of its proposed functions). This highlights the necessity for more rigorous analyses to affirm criticality in the brain. In particular, it suggests that attention should be focused on the question "does the brain implement a dynamical latent variable model?".

These authors are not the first to show that slowly varying firing rates can give rise to power law behavior (see, for example, Touboul and Destexhe, 2017; Priesemann and Shriki, 2018). However, to our knowledge they are the first to show crackling, and to compute information transmission in the critical state.

We thank the reviewers for their thoughtful assessment of our paper.

We would push back on the assessment that our model ‘has nothing to do with criticality,’ and that we observed ‘signatures of criticality [that] emerge through fundamentally non-critical mechanisms.’ This assessment partially stems from the definition of criticality provided in the Public Comment, that ‘criticality is a very specific set of phenomena in physics in which fundamentally local interactions produce unexpected long-range behavior.’

Our disagreement is largely focused on this definition, which we do not think is a standard definition. Taking the favorite textbook example, the Ising model, criticality is characterized by a set of power-law divergences in thermodynamic quantities (e.g., susceptibility, specific heat, magnetization) at the critical temperature, with exponents of these power laws governed by scaling laws. It is not defined by local interactions. All-to-all Ising model is generally viewed as showing a critical behavior at a certain temperature, even though interactions there are manifestly non-local. It is possible that, by “local” in the definition, the Public Comment meant that interactions are “collective” and among microscopic degrees of freedom. However, that same all-to-all Ising model is mathematically equivalent to the mean-field model, where criticality is achieved through large fluctuations of the mean field, but not through microscopic interactions.

More commonly, criticality is defined by power laws and scaling relationships that emerge at a critical value of a parameter(s) of the system. That is, criticality is defined by its signatures. What is crucial in all such definitions is that this atypical, critical state requires fine tuning. For example, in the textbook example of the Ising model, a parameter (the temperature) must be tuned to a critical value for critical behavior to appear. In the branching process model that generates avalanche criticality, criticality requires tuning m=1. The key result of our paper is that all signatures expected for avalanche criticality (power laws, crackling, and, as shown below, estimates of the branching rate m), and hence the criticality itself, appear without fine-tuning.

As we discussed in our introduction, there are a few other instances of signatures of criticality (and hence of criticality itself) emerging without fine-tuning. The first we are aware of was the demonstration of Zipf’s Law (by Schwab, et al. 2014, and Aitchison et al. 2016), a power-law relationship between rank and frequency of states, which was shown to emerge generically in systems driven by a broadly distributed latent variable. A second example, arising from applications of coarse-graining analysis to neural data (cf., Meshulam et al. 2019; also, Morales et al., 2023), was demonstrated in our earlier paper (Morrell et al. 2021). Thus, here we have a third example: the model in this paper generates signatures of criticality in the statistics of avalanches of activity, and it does so without fine-tuning (cf., Fig. 2-3).

The rate at which these ‘criticality without fine-tuning' examples are piling up may inspire revisiting the requirement of fine-tuning in the definition of criticality, and our ongoing work (Ngampruetikorn et al. 2023) suggests that criticality may be more accurately defined through large fluctuations (variance > 1/N) rather than through fine-tuning or scaling relations.

References:

• Schwab DJ, Nemenman I, Mehta P. “Zipf’s Law and Criticality in Multivariate Data without FineTuning.” Phys Rev Lett. 2014 Aug; doi::101103/PhysRevLett.113.068102,

• Aitchison L, Corradi N, Latham PE. “Zipf’s Law Arising Naturally When There Are Underlying, Unobserved Variables.” PLOS Computational biology. 2016 12; 12(12):1-32. doi:10.1371/journal.pcbi.1005110

• Meshulam L, Gauthier JL, Brody CD, Tank DW, Bialek W. “Coarse Graining, Fixed Points, and Scaling in a Large Population of Neurons.” Phys Rev Lett. 2019 Oct; doi: 10.1103/PhysRevLett.123.178103.

• Morales GB, di Santo S, Muñoz MA. “Quasiuniversal scaling in mouse-brain neuronal activity stems from edge-of-instability critical dynamics.” Proceedings of the National Academy of Sciences. 2023; 120(9):e2208998120.

• Morrell MC, Sederberg AJ, Nemenman I. “Latent Dynamical Variables Produce Signatures of Spatiotemporal Criticality in Large Biological Systems.” Phys Rev Lett. 2021 Mar; doi: 10.1103/PhysRevLett.126.118302.

• Ngampruetikorn, V., Nemenman, I., Schwab, D., “Extrinsic vs Intrinsic Criticality in Systems with Many Components.” arXiv: arXiv:2309.13898 [physics.bio-ph]

Major comments:

1) For many readers, the essential messages of the paper may not be immediately clear. For example, is the paper criticizing the criticality hypothesis of cortical networks, or does the criticism extend deeper, to the theoretical predictions of "crackling" relationships in physical systems as they can emerge without criticality? Statements like "We show that a system coupled to one or many dynamical latent variables can generate avalanche criticality ..." could be misinterpreted as affirming criticality. A more accurate language is needed; for instance, the paper could state that the model generates relationships observed in critical systems. The paper should provide a clearer conclusion and interpretation of the findings in the context of the criticality hypothesis of cortical dynamics.

Please see the response to the Public Review, above. To clarify the essential message that the dynamical latent variable model produces avalanche criticality without fine-tuning, we have made revisions to the abstract and introduction. This point was already made in the discussion (first sentence).

Key sentences changed in the abstract:

"… We find that populations coupled to multiple latent variables produce critical behavior across a broader parameter range than those coupled to a single, quasi-static latent variable, but in both cases, avalanche criticality is observed without fine-tuning of model parameters. … Our results suggest that avalanche criticality arises in neural systems in which activity is effectively modeled as a population driven by a few dynamical variables and these variables can be inferred from the population activity."

In the introduction, we changed the final sentence to read:

"These results demonstrate how criticality in neural recordings can arise from latent dynamics in neural activity, without need for fine-tuning of network parameters."

2) On lines 97-99, the authors state that "We are agnostic as to the origin of these inputs: they may be externally driven from other brain areas, or they may arise from recurrent dynamics locally". This idea is also repeated at the beginning of the Summary section. Perhaps being agnostic isn't such a good idea: it's possible that the recurrent dynamics is in a critical regime, which would just push the problem upstream. Presumably you're thinking of recurrent dynamics with slow timescales that's not critical? Or are you happy if it's in the critical regime? This should be clarified.

We have amended this sentence to clarify that any latent dynamics with large fluctuations would suffice:

”We are agnostic as to the origin of these inputs: they may be externally driven from other brain areas, or they may arise from large fluctuations in local recurrent dynamics.”

3) Even though the model in Equation 2 has been described in a previous publication and the Methods section, more details regarding the origin and justification of this model in the context of cortical networks would be helpful in the Results section. Was it chosen just for simplicity, or was there a deeper reason?

This model was chosen for its simplicity: there are no direct interactions between neurons, coupling between neurons and latent variables is random, and simulation is straightforward. More complex latent dynamics or non-random structure in the coupling matrices could have been used, but our aim was to explore this model in the simplest setting possible.

We have revised the Results (“Avalanche scaling in a dynamical latent variable model,” first paragraph) to justify the choice of the model:

"We study a model of a population of neurons that are not coupled to each other directly but are driven by a small number of dynamical latent variables -- that is, slowly changing inputs that are not themselves measured (Fig.~\ref{fig:fig1}A). We are agnostic as to the origin of these inputs: they may be externally driven from other brain areas, or they may arise from large fluctuations in local recurrent dynamics. The model was chosen for its simplicity, and because we have previously shown that this model with at least about five latent variables can produce power laws under the coarse-graining analysis \citep{Morrell2021}."

We have added the following to the beginning of the Methods section expanding on the reasons for this choice:

"We study a model from Morrell 2021, originally constructed as a model of large populations of neurons in mouse hippocampus. Neurons are non-interacting, receiving inputs reflective of place-field selectivity as well as input current arising from a random projection from a small number of dynamical latent variables, representing inputs shared across the population of neurons that are not directly measured or controlled. In the current paper, we incorporate only the latent variables (no place variables), and we assume that every cell is coupled to every latent variable with some randomly drawn coupling strength."

4) The Methods section (paragraph starting on line 340) connects the time scale to actual time scales in neuronal systems, stating that "The timescales of latent variables examined range from about 3 seconds to 3000 seconds, assuming 3-ms bins". While bins of 3 ms are relevant for electrophysiological data from LFPs or high-density EEG/MEG, time scales above 10 seconds are difficult to generate through biophysically clear processes like ionic channels and synaptic transmission. The paper suggests that slow time scales of the latent variables are crucial for obtaining power law behavior resembling criticality. Yet, one way to generate such slow time scales is via critical slowing down, implying that some brain areas providing input to the network under study may operate near criticality. This pushes the problem toward explaining the criticality of those external networks. Hence, discussing potential sources for slow time scales in latent variables is crucial. One possibility you might want to consider is sources external to the organism, which could easily have time scales in the 1-24 hour range.

As the reviewers note, it is a possibility that slow timescales arise from some other brain area in which dynamics are slow due to critical dynamics, but many other plausible sources exist. These include slowly varying sensory stimuli or external sources, as suggested by the reviewers. It is also possible to generate “effective” slow dynamics from non-critical internal sources. One example, from recordings in awake mice, is the slow change in the level of arousal that occurs on the scale of many seconds to minutes. These changes arise from release of neuromodulators that have broad effects on neural populations and correlations in activity (for a focused review, see Poulet and Crochet, 2019).

We have added the following sentence to the Methods section where timescales of latent variables was discussed:

"The timescales of latent variables examined range from about $3$ seconds to $3000$ seconds, assuming $3$-ms bins. Inputs with such timescales may arise from external sources, such as sensory stimuli, or from internal sources, such as changes in physiological state."

5) It is common in neuronal avalanche analysis to calculate the branching parameter using the ratio of events in consecutive bins. Near-critical systems should display values close to 1, especially in simulations without subsampling. Including the estimated values of the branching parameter for the different cases investigated in this study could provide more comprehensive data. While the paper acknowledges that the obtained exponents in the model differ from those in a critical branching process, it would still be beneficial to offer the branching parameter of the observed avalanches for comparison.

The reviewers requested that the branching parameter be computed in our model. We point out that, for the quasi-stationary latent variables (as in Fig. 3), a branching parameter of 1 is expected because the summed activity at time t+k is, on average, equal to the summed activity at time t, regardless of k. Numerics are consistent with this expectation. Following the methodology for an unbiased estimate of the branching parameter from Wilting and Priesemann (2018), we checked an example set of parameters (epsilon = 8, eta = 3) for quasi-stationary latent fields. We found that the naïve (biased) estimate of the branching parameter was 0.94, and that the unbiased estimator was exp(−1.4⋅10−8) ≈ 0.999999986.

For faster time scales, it is no longer true that summed activity is constant over time, as the temporal correlations in activity decay exponentially. Using the five-field simulation from Figure 2, we calculated the branching parameter for several values of tau. The biased estimates of m are 0.76 (𝜏=50), 0.79 (𝜏=500), and 0.79 (𝜏=5000). The corrected estimates are 0.98 (𝜏=50), 0.998 (𝜏=500), and 0.9998 (𝜏=5000).

6) In the Discussion (l 269), the paper suggests potential differences between networks cultured in vitro and in vivo. While significant differences indeed exist, it's worth noting that exponents consistent with a critical branching process have also been observed in vivo (Petermann et al 2009; Hahn et al. 2010), as well as in large-scale human data.

We thank the reviewers for pointing out these studies, and we have added the missing one (Hahn et al. 2010) to our reference list. The following was added to the discussion, in the section “Explaining Experimental Exponents:”

"A subset of the in vivo recordings analyzed from anesthetized cat (Hahn et al. 2010) and macaque monkeys (Petermann et al. 2009) exhibited a size distribution exponent close to 1.5."

Along these lines, we noted two additional studies of high relevance that have been published since our initial submission (Capek et al. 2023, Lombardi et al. 2023), and we have added these references to the discussion of experimental exponents.

Minor comments:

1) The term 'latent variable' should be rigorously explained, as it is likely to be unfamiliar to some readers.

Sentences and clauses have been added to the Introduction, Results and the Methods to clarify the term:

Intro: “Numerous studies have reported relatively low-dimensional structure in the activity of large populations of neurons [refs], which can be modeled by a population of neurons that are broadly and heterogeneously coupled to multiple dynamical latent (i.e., unobserved) variables.”

Results: “We studied a population of neurons that are not coupled to each other directly but are driven by a small number of dynamical latent variables -- that is, slowly changing inputs that are not themselves measured.”

Methods: “Neurons are non-interacting, receiving inputs reflective of place-field selectivity as well as input current reflecting a random projection from a small number of dynamical latent variables, representing inputs shared across the population of neurons that are not directly measured.”

2) There's a relatively important typo in the equations: Eq. 2 and Eq. 6 differ by a minus sign in the exponent. Eqs. 3 and 4 use the plus sign, but epsilon_0 on line 198 uses the minus sign. All very confusing until we figured out what was going on. But easy to fix.

Thank you for catching this. We have made the following corrections:

1) Figures adopted the sign convention that epsilon > 0, with larger values of epsilon decreasing the activity level. Signs in Eqs. 3 and 4 have been corrected to match.

2) Equation 5 was missing a minus sign in front of the Hamiltonian. Restoring this minus sign fixed the discrepancy between 2 and 6.

3) In Eq. 7, the left hand side is zeta'/zeta', which is equal to 1. Maybe it should be zeta'/zeta? Fixed, thank you.

Additional comments:

The authors are free to ignore these; they are meant to improve the paper.

We are extremely grateful for the close reading of our paper and note the actions taken below.

1) We personally would not use the abbreviation DLV; we find abbreviations extremely hard to remember. And DLV is not used that often.

Done, thank you for the suggestion.

2) l 198: epsilon_0 = -log(2^{1/N}-1) was kind of hard to picture -- we had to do a little algebra to make sense of it. Why not write e^{-epsilon_0} = 2^{1/N}-1 \approx log(2)/N, which in turn implies that epsilon_0 ~ log(N)?

Thank you, good point. We have added a sentence now to better explain:

"...which is maximized at $\epsilon_0 = - \log (2^{1/N} - 1)$, independent of $J_i$ and $\eta$. After some algebra, we find that $\epsilon_0 \sim \log N$ for large $N$."

3) Typo on l 202: "We plot P_ava as a function of epsilon in Fig. 4B". 4B --> 4D.

Done

4) It would be easier on the reader if the tables were all in one place. It would be even nicer to put the parameters in the figure captions. Or at least N; that one is kind of important.

Table placement was a Latex issue, which we have now fixed. We also have included links between tables and relevant figures and indicated network size.

5) What's x_i in Eqs. 7 and 8?

We added a sentence of explanation. These are the individual observations of avalanche sizes or durations, depending on what is being fit.

6) The latent variables evolve according to an Ornstein-Uhlenbeck process. But we might equally expect oscillations or non-normal behavior coupling dynamical modes, and these are likely to give different behavior with respect to avalanches. It might be worth commenting on this.

7) The model assumes a normal distribution of the coupling strengths between the latent variables and the binary units. Discussing the potential effects of different types of random coupling could provide interesting insights.

Both 6 and 7 are interesting questions. At this point, we could speculate that the main results would be qualitatively unchanged, provided dynamics are sufficiently slow and that the distribution of coupling strengths is sufficiently broad (that is, there is variance in the coupling matrix across individual neurons). Further studies would be needed to make these statements more precise.

8) In Fig 1, tau_f = 1E4 whereas in Fig 2 tau_f = 5E3. Why the difference?

For Figure 1, we chose a set of parameters that gave clear scaling. In Figure 2, we saw some value in showing more than one example of scaling, hence different parameters for the examples in Fig 2 than Fig 1. Note that the Fig 1 simulations are represented in Fig. 2 G-J, as the 5-field simulation with tau_F = 1e4.

Reviewer #2 (Public Review):

Summary:<br /> The dominant paradigm in the past decade for modeling the ventral visual stream's response to images has been to train deep neural networks on object classification tasks and regress neural responses from units of these networks. While object classification performance is correlated to the variance explained in the neural data, this approach has recently hit a plateau of variance explained, beyond which increases in classification performance do not yield improvements in neural predictivity. This suggests that classification performance may not be a sufficient objective for building better models of the ventral stream. Lindsey & Issa study the role of factorization in predicting neural responses to images, where factorization is the degree to which variables such as object pose and lighting are represented independently in orthogonal subspaces. They propose factorization as a candidate objective for breaking through the plateau suffered by models trained only on object classification. They claim that (i) maintaining these non-class variables in a factorized manner yields better neural predictivity than ignoring non-class information entirely, and (ii) factorization may be a representational strategy used by the brain.

The first of these claims is supported by their data. The second claim does not seem well-supported, and the usefulness of their observations is not entirely clear.

Strengths:<br /> This paper challenges the dominant approach to modeling neural responses in the ventral stream, which itself is valuable for diversifying the space of ideas.

This paper uses a wide variety of datasets, spanning multiple brain areas and species. The results are consistent across the datasets, which is a great sign of robustness.

The paper uses a large set of models from many prior works. This is impressively thorough and rigorous.

The authors are very transparent, particularly in the supplementary material, showing results on all datasets. This is excellent practice.

Weaknesses:<br /> 1. The primary weakness of this paper is a lack of clarity about what exactly is the contribution. I see two main interpretations: (1-A) As introducing a heuristic for predicting neural responses that improve over-classification accuracy, and (1-B) as a model of the brain's representational strategy. These two interpretations are distinct goals, each of which is valuable. However, I don't think the paper in its current form supports either of them very well:

(1-A) Heuristic for neural predictivity. The claim here is that by optimizing for factorization, we could improve models' neural predictivity to break through the current predictivity plateau. To frame the paper in this way, the key contribution should be a new heuristic that correlates with neural predictivity better than classification accuracy. The paper currently does not do this. The main piece of evidence that factorization may yield a more useful heuristic than classification accuracy alone comes from Figure 5. However, in Figure 5 it seems that factorization along some factors is more useful than others, and different linear combinations of factorization and classification may be best for different data. There is no single heuristic presented and defended. If the authors want to frame this paper as a new heuristic for neural predictivity, I recommend the authors present and defend a specific heuristic that others can use, e.g. [K * factorization_of_pose + classification] for some constant K, and show that (i) this correlates with neural predictivity better than classification alone, and (ii) this can be used to build models with higher neural predictivity. For (ii), they could fine-tune a state-of-the-art model to improve this heuristic and show that doing so achieves a new state-of-the-art neural predictivity. That would be convincing evidence that their contribution is useful.

(1-B) Model of representation in the brain. The claim here is that factorization is a general principle of representation in the brain. However, neural predictivity is not a suitable metric for this, because (i) neural predictivity allows arbitrary linear decoders, hence is invariant to the orthogonality requirement of factorization, and (ii) neural predictivity does not match the network representation to the brain representation. A better metric is representational dissimilarity matrices. However, the RDM results in Figure S4 actually seem to show that factorization does not do a very good job of predicting neural similarity (though the comparison to classification accuracy is not shown), which suggests that factorization may not be a general principle of the brain. If the authors want to frame the paper in terms of discovering a general principle of the brain, I suggest they use a metric (or suite of metrics) of brain similarity that is sensitive to the desiderata of factorization, e.g. doesn't apply arbitrary linear transformations, and compare to classification accuracy in addition to invariance.

Overall, I suggest the authors clarify exactly what their claim is, then focus on that claim and present results to justify it. If neither of the claims above can be supported by evidence, then this paper still has value as an idea that they spent effort trying to test, but they should not suggest these claims in the paper. In that case, it may also be possible to increase the value of the contribution by characterizing how the structure of class-free variable representations impacts correlation with neural fit, instead of just comparing existence vs absence (invariance) of this information. For example, evaluate the degree to which local or global orthogonality matters, or the degree to which curvature of the embedding matters.

2. I think the comparison to invariance, which is pervasive throughout the paper, is not very informative. First, it is not surprising that invariance is more weakly correlated with neural predictivity than factorization, because invariant representations lose information compared to factorized representations. Second, there has long been extensive evidence that responses throughout the ventral stream are not invariant to the factors the authors consider, so we already knew that invariance is not a good characterization of ventral stream data.

3. The formalization of the factorization metric is not particularly elegant, because it relies on computing top K principal components for the other-parameter space, where K is arbitrarily chosen as 10. While the authors do show that in their datasets the results are not very sensitive to K (Figure S5), that is not guaranteed to be the case in general. I suggest the authors try to come up with a formalization that doesn't have arbitrary constants. For example, one possibility that comes to mind is E[delta_a x delta_b], where 'x' is the normalized cross product, delta_a, and delta_b are deltas in representation space induced by perturbations of factors a and b, and the expectation is taken over all base points and deltas. This is just the first thing that comes to mind, and I'm sure the authors can come up with something better. The literature on disentangling metrics in machine learning may be useful for ideas on measuring factorization.

4. The authors defined the term "factorization" according to their metric. I think introducing this new term is not necessary and can be confusing because the term "factorization" is vague and used by different researchers in different ways. Perhaps a better term is "orthogonality", because that is clear and seems to be what the authors' metric is measuring.

5. One general weakness of the factorization paradigm is the reliance on a choice of factors. This is a subjective choice and becomes an issue as you scale to more complex images where the choice of factors is not obvious. While this choice of factors cannot be avoided, I suggest the authors add two things: First, an analysis of how sensitive the results are to the choice of factors (e.g. transform the basis set of factors and re-run the metric); second, include some discussion about how factors may be chosen in general (e.g. based on temporal statistics of the world, independent components analysis, or something else).

We did an exercise in my "Models of Effective Helping" class yesterday where we were given a list of about 12 people and the list had their ages and a brief description of who they are. We had to choose as a group eight people from the list to go on a life raft and the rest of them would die. You could also choose to save yourself as one of the eight. There was a heated debate over whether its ethical or proper to choose to save yourself over somebody else. I couldn't fathom the thought of me choosing for somebody to die over myself, but the majority of the class agreed to save themselves and kill somebody else. What I thought was even more shocking was that a few of the people on the list were teenagers and people were trying to justify killing them over themselves. What I learned is that there are many people who truly believe that they are the center of the universe, and it may just be human nature to think so.

This seems like the thinking of someone who does not believe in equality. I don't think that we need such severe levels of poverty to achieve a healthy social order. It should not be impossible to get out of poverty. Poverty should not be a life sentence.

This paragraph of the article tells how technological developments have the potential to redefine disability. For example, the development of cochlear implants and hearing aids has changed the way society views hearing loss. Similarly, Braille displays and screen reading technology have revolutionized the way blind people access information. I think these are all very significant things, and increasingly people are seeing disability as a spectrum rather than a binary state (disabled vs. non-disabled). Individuals may have different levels of skills in different areas. For example, a person may have a mild visual impairment that has no effect on most activities, but can be a major handicap in specific situations, such as dimly lit areas. I've known people with this problem so I think the point is important.

Certainly, considering disability in the context of design involves creating products, environments, and systems that are inclusive and accessible to individuals with a diverse range of abilities. Engaging individuals with disabilities in the design process helps to identify specific needs and challenges they may face. This approach ensures that the final design reflects the diverse perspectives and requirements of the user community.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Nitrogen metabolism is of fundamental importance to biology. However, the metabolism and biochemistry of guanidine and guanidine containing compounds, including arginine and homoarginine, have been understudied over the last few decades. Very few guanidine forming enzymes have been identified. Funck et al define a new type of guanidine forming enzyme. It was previously known that 2-oxogluturate oxygenase catalysis in bacteria can produce guanidine via oxidation of arginine. Interestingly, the same enzyme that produces guanidine from arginine also oxidises 2-oxogluturate to give the plant signalling molecule ethylene. Funck et al show that a mechanistically related oxygenase enzyme from plants can also produce guanidine, but instead of using arginine as a substrate, it uses homoarginine. The work will stimulate interest in the cellular roles of homoarginine, a metabolite present in plants and other organisms including humans and, more generally, in the biochemistry and metabolism of guanidines.

1) Significance

Studies on the metabolism and biochemistry of the small nitrogen rich molecule guanidine and related compounds including arginine have been largely ignored over the last few decades. Very few guanidine forming enzymes have been identified. Funck et al define a new guanidine forming enzyme that works by oxidation of homoarginine, a metabolite present in organisms ranging from plants to humans. The new enzyme requires oxygen and 2oxogluturate as cosubstrates and is related, but distinct from a known enzyme that oxidises arginine to produce guanidine, but which can also oxidise 2-oxogluturate to produce the plant signalling molecule ethylene.

Overall, I thought this was an exceptionally well written and interesting manuscript. Although a 2-oxogluturate dependent guanidine forming enzyme is known (EFE), the discovery that a related enzyme oxidises homoarginine is really interesting, especially given the presence of homoarginine in plant seeds. There is more work to be done in terms of functional assignment, but this can be the subject of future studies. I also fully endorse the authors' view that guanidine and related compounds have been massively understudied in recent times. I would like to see the possibility that the new enzyme makes ethylene explored. Congratulations to the authors on a very nice study.

Response: We thank the reviewer for the positive evaluation of our manuscript. In the revised version, we have emphasized more clearly that we found no evidence for ethylene production by the recombinant enzymes. The other suggestions of the reviewer are also considered in the revised version as detailed below.

Reviewer #2 (Public Review):

In this study, Dietmar Funck and colleagues have made a significant breakthrough by identifying three isoforms of plant 2-oxoglutarate-dependent dioxygenases (2-ODD-C23) as homo/arginine-6-hydroxylases, catalyzing the degradation of 6-hydroxyhomoarginine into 2aminoadipate-6-semialdehyde (AASA) and guanidine. This discovery marks the very first confirmation of plant or eukaryotic enzymes capable of guanidine production.

The authors selected three plant 2-ODD-C23 enzymes with the highest sequence similarity to bacterial guanidine-producing (EFE) enzymes. They proceeded to clone and express the recombinant enzymes in E coli, demonstrating capacity of all three Arabidopsis isoforms to produce guanidine. Additionally, by precise biochemical experiments, the authors established these three 2-ODD-C23 enzymes as homoarginine-6-hydroxylases (and arginine-hydroxylase for one of them). Furthermore, the authors utilized transgenic plants expressing GFP fusion proteins to show the cytoplasmic localization of all three 2-ODD-C23 enzymes. Most notably, using T-DNA mutant lines and CRISPR/Cas9-generated lines, along with combinations of them, they demonstrate the guanidine-producing capacity of each enzyme isoform in planta. These results provide robust evidence that these three 2-ODD-C23 Arabidopsis isoforms are indeed homoarginine-6-hydroxylases responsible for guanidine generation.

The findings presented in this manuscript are a significant contribution for our understanding of plant biology, particularly given that this work is the first demonstration of enzymatic guanidine production in eukaryotic cells. However, there are a couple of concerns and potential ways for further investigation that the authors should (consider) incorporate.

Firstly, the observation of cytoplasmic and nuclear GFP signals in the transgenic plants may also indicate cleaved GFP from the fusion proteins. Thus, the authors should perform Western blot analysis to confirm the correct size of the 2-ODD-C23 fusion proteins in the transgenic protoplasts.

Secondly, it may be worth measuring pipecolate (and proline?) levels under biotic stress conditions (particularly those that induce transcript changes of these enzymes, Fig S8). Given the results suggesting a potential regulation of the pathway by biotic stress conditions (eg. meJA), these experiments could provide valuable insights into the physiological role of guanidine-producing enzymes in plants. This additional analysis may give a significance of these enzymes in plant defense mechanisms.

Response: We thank also reviewer 2 for the positive evaluation and useful suggestions. We performed the proposed GFP Western blot, which indeed indicated the presences of both, fulllength fusion proteins and free GFP, which can explain the partial nuclear localization. We fully agree that further experiments with biotic and abiotic stress will be required to determine the physiological function of the 2-ODD-C23 enzymes. However, the list of potential experiments is long and they are beyond the scope of the present manuscript.

Reviewer #1 (Recommendations For The Authors):

Specific points

Overall, I thought this was a very interesting study, comprising biochemical, cellular, and in vivo studies. Of course more could be done on each of these, and likely will be, but I think the assignment of biochemical function is very strong, across all three approaches. The one new experiment I would like to see is a clear demonstration of whether ethylene is produced - unlikely but should be tested.

We had mentioned our failure to detect ethylene production by the plant enzymes in the previous version and have made it more prominent and reliable by including ethylene production as positive control in the new supplementary figure S5.

Abstract

Delete 'hitherto overlooked' - this is implicit 'but is more likely' to 'is likely'?

Agreed and modified

Introduction

Second sentence - what about relevant small molecule primary metabolites including precursors of proteins/nucleic acids.

We modified the sentence accordingly.

Paragraph 2 - maybe also note EFE produces glutamate semi aldehyde, via arginine C-5 oxidation.

Paragraph 2 has been re-phrased according to your suggestion.

Overall, I thought the introduction was exceptionally well written.

Perhaps either in the introduction, or later, note there are other 2OG oxygenases that oxidise arginine/arginine derivatives in various ways, e.g. clavaminate synthase/arginine hydroxylases/desaturases.

We added a sentence mentioning the arginine hydroxylases VioC and OrfP to the introduction and included VioC into the sequence comparison in supplementary figure 2 to show that these enzymes, as well as NapI, are very different from EFE and the plant hydroxylases.

Results

Paragraph 1 - qualify similarity and refer to/give a structurally informed sequence alignment, including EFE

A new supplemental figure S2 was added with sequence identity values and a structurally informed alignment. The text has been modified accordingly.

Paragraph 2 - briefly state method of guanidine analysis

We included a reference to the M&M section and mentioned LC-MS in paragraph 2.

Figure 1 - trivial point - proteins are not expressed/genes are

We have modified the legend to figure 1. However, we would like to point out that terms like “recombinant protein expression” are widely used in the field. A quick search with google Ngram viewer shows that “protein expression” started to appear in the mid-80ies and its use stayed constantly at 1/8th of “gene expression”.

Define errors clearly in all figure legends, clearly defining biological/technical repeats<br /> Page 6 - was the His-tag cleared to ensure no issues with Ni contamination?

We treat individual plants or independent bacterial cultures as biological replicates. Only in the case of enzyme activity assays with NAD(P)H, technical replicates were used and this has been indicated in the legend of figure 6.

Lower case 'p' in pentafluorobenzyl corrected

In Figure 2 make clear the hydroxylated intermediates are not observed

We now use grey color for the intermediates and have put them in brackets. Additionally we state in the figure legend that these intermediates were not detected.

Pages 6-7 - I may have missed this but it's important to investigate what happens to the 2OG. Is succinate the only product or is ethylene also produced? This possibility should also be considered in the plant studies, i.e. is there any evidence for responses related to perturbed ethylene metabolism. The authors consider a signalling role relating to AASA/P6C, but seem to ignore a potential ethylene connection.

As stated above, we checked for ethylene production with negative result. EFE produced 6 times more guanidine than the plant enzymes under the same condition, but even 100-fold lower ethylene production would have been clearly detected.

Page 12 - 'plants have been shown to....' Perhaps note how hydroxy guanidine is made?

We now mention the canavanine-γ-lyase that cleaves canavanine into hydroxyguanidine and homoserine.

Overall, I thought the discussion was good, but perhaps a bit long/too speculative on pages 12/13 and this detracted from the biochemical assignment of the enzyme. I'd suggest shortening the discussion somewhat - the precise roles of the enzyme can be the subject of future work. As indicated above, some discussion on potential links to ethylene would be appreciated.

Since reviewer 2 wanted more (speculative) discussion on the role of the 2-ODD-C23 enzymes and there was no detectable ethylene production, we took the liberty to leave the discussion largely unaltered.

I'd also like to see some more consideration/metabolic analyses of guanidine related metabolism in the genetically modified plants.

Such analyses will certainly be included in future experiments once we get an idea about the physiological role of the 2-ODD-C23 enzymes.

Page 16 - mass spectrometry

Corrected.

Please add a structurally informed sequence alignment with EFE and other 2OG oxygenases acting on arginine/derivatives.

An excerpt of the alignment is now presented in supplementary figure S2.

Reviewer #2 (Recommendations For The Authors):

I would like to see more discussion in the manuscript about the possible interconnection/roles between 2-ODD-C23 guanidine-producing, lysine- ALD1-Pipecolate producing, and proline metabolism pathways during both biotic and abiotic stresses.

Since we were unable to detect pipecolate in any of our plant samples and also our preliminary results with biotic stress did not produce any evidence for a function of the 2ODD-C23 enzymes in the tested defense responses, we would like to postpone such extended discussion until we find a condition where the physiological function of these enzymes is evident.

Fig. 4: Authors should change colors for Col-0, 0.2 HoArg and ctrl? They look too similar in my pdf file.

We changed the colors in figure 4 and hope that the enhanced contrast is maintained during the production of the final version of our article.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Sender et al describe a model to estimate what fraction of DNA becomes cell-free DNA in plasma. This is of great interest to the community, as the amount of DNA from a certain tissue (for example, a tumor) that becomes available for detection in the blood has important implications for disease detection.

However, the authors' methods do not consider important variables related to cell-free DNA shedding and storage, and their results may thus be inaccurate. At this stage of the paper, the methods section lacks important detail. Thus, it is difficult to fully assess the manuscript and its results.

Strengths:

The question asked by the authors has potentially important implications for disease diagnosis. Understanding how genomic DNA degrades in the human circulation can guide towards ways to enrich for DNA of interest or may lead to unexpected methods of conserving cell-free DNA. Thus, the question "how much genomic DNA becomes cfDNA" is of great interest to the scientific and medical community. Once the weaknesses of the manuscript are addressed, I believe this manuscript has the potential to be a widely used resource.

Weaknesses:

There are two major weaknesses in how the analysis is presented. First, the methods lack detail. Second, the analysis does not consider key variables in their model.

Issues pertaining to the methods section.

The current manuscript builds a flux model, mostly taking values and results from three previous studies: 1) The amount of cellular turnover by cell type, taken from Sender & Milo, 2021

2) The fractions of various tissues that contribute DNA to the plasma, taken from Moss et al, 2018 and Loyfer et al, 2023

My expertise lies in cell-free DNA, and so I will limit my comments to the manuscripts in (2). Paper by Loyfer et al (additional context):

Loyfer et al is a recent landmark paper that presents a computational method for deconvoluting tissues of origin based on methylation profiles of flow-sorted cell types. Thus, the manuscript provides a well-curated methylation dataset of sorted cell-types. The majority of this manuscript describes the methylation patterns and features of the reference methylomes (bulk, sorted cell types), with a smaller portion devoted to cell-free DNA tissue of origin deconvolution.

I believe the data the authors are retrieving from the Loyfer study are from the 23 healthy plasma cfDNA methylomes analyzed in the study, and not the re-analysis of the 52 COVID-19 samples from Cheng et al (MED 2021).

Paper by Moss et al (additional context):

Moss et al is another landmark paper that predates the Loyfer et al manuscript. The technology used in this study (methylation arrays) is outdated but is an incredible resource for the community. This paper evaluates cfDNA tissues of origin in health and different disease scenarios. Again, I assume the current manuscript only pulled data from healthy patients, although I cannot be sure as it is not described in the methods section.

This manuscript:

The current manuscript takes (I think) the total cfDNA concentration from males and females from the Moss et al manuscript (pooled cfDNA; 2 young male groups, 2 old male groups, 2 young female groups, 2 old female groups, Supplementary Dataset; "total_cfDNA_conc" tab). I believe this is the data used as total cfDNA concentration. It would be beneficial for all readers if the authors clarified this point.

The tissues of origin, in the supplemental dataset ("fraction" tab), presents the data from 8 cell types (erythrocytes, monocytes/macrophages, megakaryocytes, granulocytes, hepatocytes, endothelial cells, lymphocytes, other). The fractions in the spreadsheet do not match the Loyfer or Moss manuscripts for healthy individuals. Thus, I do not know what values the supplementary dataset represents. I also don't know what the deconvolution values are used for the flux model.

The integration of these two methods lack detail. Are the authors here using yields (ie, cfDNA concentrations) from Moss et al, and tissue fractions from Loyfer et al? If so, why? There are more samples in the Loyfer manuscript, so why are the samples from Moss et al. being used? The authors are also selectively ignoring cell-types that are present in healthy individuals (Neurons from Moss et al, 2018). Why?

Appraisal:

At this stage of the manuscript, I think additional evidence and analysis is required to confirm the results in the manuscript.

Impact:

Once the authors present additional analysis to substantiate their results, this manuscript will be highly impactful on the community. The field of liquid biopsies (non-invasive diagnostics) has the potential to revolutionize the medical field (and has already in certain areas, such as prenatal diagnostics). Yet, there is a lack of basic science questions in the field. This manuscript is an important step forward in asking more "basic science" questions that seek to answer a fundamental biological question.

We thank the reviewer for the valuable comments on our analysis. In response to the feedback, we have updated the analysis to address all critical points as described below and revised the text to enhance the clarity of our methodology. One notable improvement to our analysis involved ensuring better alignment between the cohort data for cfDNA plasma concentration and cell turnover estimates. To achieve this, we utilized the total plasma concentration of cfDNA from a study conducted by Meddeb et al. 2019, taking into account the influence of age and sex on these concentrations and specifically focusing on a cohort of relatively young and healthy individuals. Additionally, we considered expected variations related to sex, age, and other pertinent factors, as outlined in the studies by Meddeb et al. 2019 and Madsen et al. 2019.

In addition, we have addressed concerns regarding the technical aspects of cfDNA analysis, providing detailed explanations of their limited impact on our analysis and the resulting conclusions.

Reviewer #2 (Public Review):

Summary:

Cell-free DNA (cfDNA) are short DNA fragments released into the circulation when cells die. Plasma cfDNA level is thought to reflect the degree of cell-death or tissue injury. Indeed, plasma cfDNA is a reliable diagnostic biomarker for multiple diseases, providing insights into disease severity and outcomes. In this manuscript, Dr. Sender and colleagues address a fundamental question: What fraction of DNA released from cell death is detectable as plasma cfDNA? The authors use public data to estimate the amount of DNA produced from dying cells. They also utilize public data to estimate plasma cfDNA levels. Their calculations showed that <10% of DNA released is detectable as plasma cfDNA, the fraction of detectable cfDNA varying by tissue sources. The study demonstrates new and fundamental principles that could improve disease diagnosis and treatment via cfDNA.

Strengths:

1) The experimental approach is resource-mindful taking advantage of publicly available data to estimate the fraction of detectable cfDNA in physiological states. The authors did not assess if the fraction of detectable cfDNA changes in disease conditions. Nonetheless, their pioneering study lays the foundation and provides the methods needed for a similar assessment in disease states.

2) The findings of this study potentially explain discrepancies in measured versus expected tissue-specific cfDNA from some tissues. For example, the gastrointestinal tract is subject to high cell turnover and release of DNA. Yet, only a small fraction of that DNA ends up in plasma as gastrointestinal cfDNA.

3) The study proposes potential mechanisms that could account for the low fraction of detectable cfDNA in plasma relative to DNA released. This includes intracellular or tissue machinery that could "chew up" DNA released from dying cells, allowing only a small fraction to escape into plasma as cfDNA. Could this explain why the gastrointestinal track with an elaborate phagosome machinery contributes a small fraction of plasma cfDNA? Given the role of cfDNA as damage-associated molecular pattern in some diseases, targeting such a machinery may provide novel therapeutic opportunities.

Weaknesses:

In vitro and in vivo studies are needed to validate these findings and define tissue machinery that contribute to cfDNA production. The validation studies should address the following limitations of the study design: -

1) Align the cohorts to estimate DNA production and plasma cfDNA levels. Cellular turnover rate and plasma cfDNA levels vary with age, sex, circadian clock, and other factors (Madsen AT et al, EBioMedicine, 2019). This study estimated DNA production using data abstracted from a homogenous group of healthy control males (Sender & Milo, Nat Med 2021). On the other hand, plasma cfDNA levels were obtained from datasets of more diverse cohort of healthy males and females with a wide range of ages (Loyfer et al. Nature, 2023 and Moss et al., Nat Commun, 2018).

2) "cfDNA fragments are not created equal". Recent studies demonstrate that cfDNA composition vary with disease state. For example, cfDNA GC content, fraction of short fragments, and composition of some genomic elements increase in heart transplant rejection compared to no-rejection state (Agbor-Enoh, Circulation, 2021). The genomic location and disease state may therefore be important factors to consider in these analyses.

3) Alternative sources of DNA production should be considered. Aside from cell death, DNA can be released from cells via active secretion. This and other additional sources of DNA should be considered in future studies. The distinct characteristics of mitochondrial DNA to genomic DNA should also be considered.

We appreciate the reviewer's comments on our analysis. In response to the feedback, we have updated to address key points and revised the text accordingly.

1) We have incorporated several enhancements to improve the coherence of our analysis. In our revised examination, we drew upon the total plasma concentration of cfDNA, as documented in a study conducted by (Meddeb et al. 2019), while considering the influence of age and sex on these concentrations. To ensure the cohort's alignment, we focus on relatively young and healthy individuals, specifically those below the age of 47. This approach allowed for a more meaningful comparison with the estimated DNA flux from a reference male human aged between 20 and 30 years.

There was no specific estimate for a cohort of young males in both Meddeb et al. and Loyfer et al.; however, we factored in the expected variations stemming from sex, age, and other relevant factors, as elucidated in literature (Meddeb et al. 2019; Madsen et al. 2019). Thus, we demonstrate that sex and age have a small effect on the cfDNA concentrations and thus are unlikely to alter our conclusions substantially when considering a healthy population. We summarize the changes in the first paragraph, replacing the “Tissue-specific cfDNA concentration” subsection of the method, and the fourth paragraph added to the discussion.

2) In this study, we addressed the total amount of cfDNA in healthy individuals without regard to GC content, representation of different genomic regions, or fragment length, as the goal was to understand if cell death rates are fully accounted for by cfDNA concentration. We agree that it will be interesting to study the relative representation of the genome in cfDNA and the processes that determine cfDNA concentration in pathologies beyond the rate of cell death. These topics for future research fall beyond this study's scope.

3) We know only a few specific cases whereby DNA is released from cells that are not dying. These include the release of DNA from erythroblasts and megakaryocytes to generate anucleated erythrocytes and platelets (Moss et al. 2022, cited in our paper) and the release of NETs from neutrophils.

The presence of cfDNA fragments originating from megakaryocytes and erythroblasts indicates the elimination of megakaryocytes and erythroblasts and the birth of erythrocytes and platelets. However, the considerations in the rest of the paper still apply: the concentration of cfDNA from these sources is far lower than expected from the cell turnover rate.

Concerning NETosis: the presence of cfDNA originating in neutrophils that have not died would reduce the concentration of cfDNA from dying neutrophils and thus further increase the discrepancy, which is the topic of our study (under-representation of DNA from dying cells in plasma).

We neglected mitochondrial DNA, as it is not measured in methylation cell-of-origin analysis. Similarly to the argument above, if some of the total DNA measured in plasma is in fact, mitochondrial, this would mean that genomic cfDNA concentration is actually lower than the estimates, meaning that an even smaller fraction of DNA from dying cells is measured in plasma.

Recommendations For The Authors

Reviewer #1 (Recommendations For The Authors):

I think readers would appreciate the authors commenting or addressing the following points, in addition to addressing the concerns I raised about the methods section in the public review:

What variables and considerations did the authors omit in this study?

1) Cell-free DNA is found in virtually every biofluid.

Thus, the fact that cell-free DNA is not present in the plasma does not mean it cannot be detected elsewhere. This also implies that phagocytosis may not be the only factor related to cfDNA not being present in the blood. One example (of many, many others) is neutrophil-derived cell-free DNA, which is present in the urine.

Indeed, dying cells and their DNA can be consumed locally, released into the blood, or shed outside the body. The latter is a function of tissue topology. For example, intestinal epithelial cell turnover releases material to the lumen of the gut (i.e., stool); kidney and bladder cell turnover releases material to urine; and lung epithelium releases material to the air spaces. In these cases, the absence of cfDNA in plasma is expected. However, in cases where tissue topology dictates release to blood, low representation in cfDNA indicates local consumption or a related mechanism. In Figure 1 of the manuscript, we distinguish between tissues according to their topology, labeling organs that shed material to the outside denoted by open circles.

Neutrophil-derived DNA in urine likely represents a local process in the kidney (neutrophils that penetrate the epithelium and fall into the urine). Neutrophils that die elsewhere in the body must release cfDNA to the blood before it can reach the urine. Hence, quantifying plasma cfDNA is a legitimate approach for assessing the relationship between cell death and cfDNA. The revised text clarifies this point. We made revisions to the initial paragraph in the results section and a paragraph within the discussion to provide clarity on this topic:

“Based on atlases of human cell type-specific methylation signatures, Moss et al. and Loyfer et al. analyzed the main cell types contributing to plasma cfDNA. They found the primary sources of plasma cfDNA to be blood cells: granulocytes, megakaryocytes, macrophages, and/or monocytes (the signature could not differentiate between the last two), lymphocytes, and erythrocyte progenitors. Other cells that had detectable contributions are endothelial cells and hepatocytes. Qualitatively, these cells represent most of the leading cell types in cellular turnover, as shown in Sender & Milo 2021 (Sender and Milo 2021). Epithelial cells of the gastrointestinal tract, lung, kidney, bladder, and skin are other cell types that significantly contribute to cellular turnover. Dying cells in these tissues are shed into the gut lumen, the air spaces, the urine, or out of the skin (note that while DNA from gut, lung, and kidney epithelial cells can be found in stool, bronchoalveolar lavage, and urine, the fate of DNA from skin cells is not known). This arrangement may explain why DNA from these cell types is not represented in plasma cfDNA in healthy conditions. Therefore, it appears that cells with high cfDNA plasma levels are those with relatively high turnover that are not being shed out of the body.”

“A comparison between the different types of cells shows a trend in which less DNA flux from cells with higher turnover gets to the bloodstream. In particular, a tiny fraction (1 in 3x104) of DNA from erythroid progenitors arrives at the plasma, indicating an extreme efficiency of the DNA recovery mechanism. Erythroid progenitors are arranged in erythroblastic islands. Up to a few tens of erythroid progenitors surround a single macrophage that collects the nuclei extruded during the erythrocyte maturation process (pyrenocytes) (Chasis and Mohandas 2008). The amount of DNA discarded through the maturation of over 200 billion erythrocytes per day (Sender and Milo 2021) exceeds all other sources of homeostatic discarded DNA. Our findings indicate that the organization of dedicated erythroblastic islands functions highly efficiently regarding DNA utilization. Neutrophils are another high-turnover cell type with a low level of cfDNA. When contemplating the process of NETosis (Vorobjeva and Chernyak 2020), the existence of cfDNA originating from live neutrophils would potentially diminish the concentration of cfDNA released by dying neutrophils, thereby amplifying the observed ratio for this particular cell type. The overall trend of higher turnover resulting in a lower cfDNA to DNA flux ratio may indicate similar design principles, in which the utilization of DNA is better in tissues with higher turnover. However, our analysis is limited to only several cell types (due to cfDNA test and deconvolution sensitivities), and extrapolation to cells with lower cell turnover is problematic.”

2) Effect of biofluid storage.

Cell-free DNA continues to degrade after it is extracted via blood draw. This is not expected to change tissue of origin predictions (although that remains to be shown in the literature), but definitely affects extraction yield. This is not accounted for (or even discussed) in the manuscript. It would be important to understand how this was done for the data presented here.

The paper integrates data from multiple recent studies that adhered to state-of-the-art procedures requiring rapid processing of blood samples. In fact, earlier studies that were not careful to isolate plasma quickly typically reported very high concentrations due to the lysis of leukocytes and artifactual release of genomic DNA. Rapid plasma isolation and DNA extraction typically yield 5ng/ml in healthy donors, as stated in the paper (last paragraph of Results).

3) Batch effects

Batch effects are not discussed here and can affect cfDNA yields.

Our analysis relies on data reported by multiple studies from different groups, which independently results in similar key findings (total concentration of cfDNA and the relative contribution of different tissues). Thus, batch effects are unlikely to affect the calculations markedly.

4) Cell-free DNA extraction kits

Different kits and methods extract cell-free DNA at different quantities. Importantly, much research has been done recently that most kits are not sensitive for ultrashort cell-free DNA (of lengths ~50bp). This may represent most of the DNA present in plasma. This raises an important question: are the yields that are being used in Moss et al (where I presume the total concentration is taken from) accurate? Is there more cell-free DNA that was missed? While the importance of this ultrashort cfDNA has yet to be shown, it is in the blood. Thus, the authors' model may underestimate ratios by not accounting for this. This is mentioned in the discussion, but it is not evident why it was not added into the model.

The Qiagen cfDNA extraction kit can detect 50bp fragments. As shown in the specification sheets of the kit (https://www.qiagen.com/us/products/diagnostics-and-clinical-research/solutions-for -laboratory-developed-tests/qiasymphony-dsp-circulating-dna-kit), urine DNA contains abundant DNA fragments that peak at 50bp. In contrast, plasma cfDNA does not contain such fragments at appreciable concentrations. This suggests that small fragments, 50-150bp long, are not a major component of cfDNA, and thus, our measurements of the total concentration of cfDNA are not dramatically underestimated.

The convention regarding the size distribution of cfDNA fragments is based on extensive evidence using multiple approaches. For example, a study that profiled the DNA released by multiple cell lines in vitro (Aucamp et al. 2017) used another kit for DNA isolation – the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Düren, Germany). This kit does extract fragments that are 50bp long (nucleospin-gel-and-pcr-clean-up-mini). Indeed, the DNA released from cultured cells did contain a peak at 50bp, but it was minor compared with the nucleosome-size peak.

More recently, several studies did suggest the presence of ultra-short cfDNA fragments, 50 bp long on average, and concluded that such fragments might be present at a molar concentration that is comparable to that of nucleosome-protected DNA (for example, (Hisano et al. 2021)).

Thus, our model estimates can be off by up to 2-fold (that is, actual cfDNA concentration measured in most studies overlooks the small fragments and thus underestimates the actual concentration of cfDNA by 2-fold). This is incorporated into the revised manuscript.

We note that we cannot exclude the presence of abundant ultra-short DNA fragments (e.g., 10bp long). However, such fragments are not measurable in cfDNA analysis. Thus, we can refine our conclusion and state that only a small fraction of DNA of dying cells appears as measured cfDNA. We included a section in the methods detailing the integration of a potential factor for the short fragments and revised the discussion:

“The overall plasma cfDNA concentration was multiplied by a factor of 1.5 to accommodate for the presence of small fragments of approximately 50 base pairs of cfDNA in the plasma. These fragments are suggested to contribute comparable molar concentrations (Hisano, Ito, and Miura 2021). Despite having approximately one-third of the mass, it is reasonable to presume that these fragments represent a similar number of genomes. This assumption is based on the idea that their source is a broken nucleosome unit, and the fragments represent the portion that was not degraded. Given the restricted data and its interpretation, we consider factors spanning the range of 1 (negligible effect) and 2 (doubling of the amount). The chosen factor, 1.5, is selected as the midpoint within this range of uncertainty.”

“In this study, we report a surprising, dramatic discrepancy between the measured levels of cfDNA in the plasma and the potential DNA flux from dying cells. One hypothetical explanation for that discrepancy is the limited sensitivity of typical cfDNA assays to short DNA fragments, which may contribute a significant fraction of the overall cfDNA mass. Regular cfDNA analysis shows a size distribution concentrated around a length of 165 base pairs (bp). The sizes in ctDNA vary more, but most are longer than 100 bp (Alcaide et al. 2020; Udomruk et al. 2021). Recent studies suggested a significant fraction of single-strand ultrashort fragments (length of 25-60 bp) (Cheng et al. 2022; Hisano, Ito, and Miura 2021). However, the total amount of DNA contained in these fragments is less than or comparable to that of the longer “regular” nucleosome-protected cfDNA fragments (Cheng et al. 2022; Hisano, Ito, and Miura 2021), arguing against ultrashort fragments as a dominant explanation for the “missing” cfDNA material. We integrated the estimate provided by Hisano et al. into our analysis as a modifying factor for both the total concentration and uncertainty of plasma cfDNA. Importantly, this incorporation did not alter the overall conclusions, as the discrepancy between the cfDNA plasma concentration and potential DNA flux remains on the same order of magnitude. We note that we cannot exclude the presence of abundant DNA fragments that are even shorter (e.g., 10bp long) and are not measurable in cfDNA analysis. Thus, our formal conclusion is that only a small fraction of the DNA of dying cells appears as measurable cfDNA.”

5) Health status of samples analyzed.

Health, sex and physical activity affects cfDNA yields. This is not accounted for or discussed in the manuscript.

We incorporated several enhancements to improve our analysis in response to the provided feedback. In our revised examination, we drew upon the total plasma concentration of cfDNA, as documented in a study conducted by (Meddeb et al. 2019), while considering the influence of age and sex on these concentrations. To ensure the cohort's alignment, we focus on relatively young and healthy individuals, specifically those below the age of 47. This approach allowed for a more meaningful comparison with the estimated DNA flux from a reference male human aged between 20 and 30 years.

Furthermore, we factored in the expected variations stemming from sex, age, and other relevant factors, as elucidated in the works of (Meddeb et al. 2019; Madsen et al. 2019). Our intent in doing so was to demonstrate that these factors are unlikely to alter our conclusions substantially when considering a healthy population. We summarize the changes in the first paragraph, replacing the “Tissue-specific cfDNA concentration” subsection of the method, and the fourth paragraph added to the discussion:

“Our estimates for total plasma cfDNA concentration were derived from the median concentration observed in individuals below 47 years of age (n=52), as reported by (Meddeb et al. 2019). To complement this, we integrated our total concentration estimates with data on the proportion of cfDNA originating from specific cell types, leveraging a plasma methylome deconvolution method described by (Loyfer et al. 2023), which did not provide absolute quantities of cfDNA). To quantify the uncertainty associated with our cfDNA concentration estimates, we employed a methodology that considered several sources of variation. First, we incorporated the confidence interval of the median concentration reported by Meddeb et al. as a measure of uncertainty. Additionally, we accounted for individual-specific and analytic variations based on the study by (Madsen et al. 2019), encompassing factors such as the precise timing of measurements and assay precision. These sources of uncertainty were combined using the approach outlined below.”

“Our current analysis focused on estimating plasma cfDNA concentration and cellular turnover in a cohort of healthy, relatively young individuals. The total plasma cfDNA concentrations were sourced from healthy individuals below 47 years, as reported by (Meddeb et al. 2019). We use data analyzed based on plasma samples from healthy individuals to estimate the proportion of cfDNA originating from specific cell types (Loyfer et al. 2023). These values were then compared to the potential DNA flux resulting from homeostatic cellular turnover, estimated for reference healthy males aged between 20 and 30 (Sender and Milo 2021). In our analysis, we considered various sources of uncertainty, including inter-individual variation, variability in the timing of sample collection, and analytical precision (Madsen et al. 2019; Meddeb et al. 2019). These factors collectively contributed to an uncertainty factor of less than 3. Importantly, this level of uncertainty does not alter our conclusion regarding the relatively small fraction of DNA present in plasma as cfDNA. Furthermore, we acknowledge that age and sex can impact total cfDNA concentration, as demonstrated by (Meddeb et al. 2019), with potential variations of up to 30%. However, as the results of our analysis present a much larger difference, these effects do not change the conclusions drawn from our analysis. Nevertheless, age and health status may influence the proportion of cfDNA originating from specific cell types and their corresponding cellular turnover rates. Consequently, the ratios themselves may vary in the elderly population or individuals with underlying health conditions.”

Reviewer #2 (Recommendations For The Authors):

1) Align the cohorts to estimate DNA production and plasma cfDNA levels. Cellular turnover rate and plasma cfDNA levels vary with age, sex, circadian clock, and other factors (Madsen AT et al, EBioMedicine, 2019). This study estimated DNA production using data abstracted from a homogenous group of healthy control males (Sender & Milo, Nat Med 2021). On the other hand, plasma cfDNA levels were obtained from datasets of more diverse cohort of healthy males and females with a wide range of ages (Loyfer et al. Nature, 2023 and Moss et al., Nat Commun, 2018).

We have incorporated several enhancements to improve the coherence of our analysis. In our revised examination, we drew upon the total plasma concentration of cfDNA, as documented in a study conducted by (Meddeb et al. 2019), while considering the influence of age and sex on these concentrations. To ensure the cohort's alignment, we focus on relatively young and healthy individuals, specifically those below the age of 47. This approach allowed for a more meaningful comparison with the estimated DNA flux from a reference male human aged between 20 and 30 years.

There was no specific estimate for a cohort of young males in both Meddeb et al. and Loyfer et al.; however, we factored in the expected variations stemming from sex, age, and other relevant factors, as elucidated in literature (Meddeb et al. 2019; Madsen et al. 2019). Thus, we demonstrate that sex and age have a small effect on the cfDNA concentrations and thus are unlikely to alter our conclusions substantially when considering a healthy population.

We summarize the changes in the first paragraph, replacing the “Tissue-specific cfDNA concentration” subsection of the method, and the fourth paragraph added to the discussion.

“Our estimates for total plasma cfDNA concentration were derived from the median concentration observed in individuals below 47 years of age (n=52), as reported by (Meddeb et al. 2019). To complement this, we integrated our total concentration estimates with data on the proportion of cfDNA originating from specific cell types, leveraging a plasma methylome deconvolution method described by (Loyfer et al. 2023), which did not provide absolute quantities of cfDNA). To quantify the uncertainty associated with our cfDNA concentration estimates, we employed a methodology that considered several sources of variation. First, we incorporated the confidence interval of the median concentration reported by Meddeb et al. as a measure of uncertainty. Additionally, we accounted for individual-specific and analytic variations based on the study by (Madsen et al. 2019), encompassing factors such as the precise timing of measurements and assay precision. These sources of uncertainty were combined using the approach outlined below.”

“Our current analysis focused on estimating plasma cfDNA concentration and cellular turnover in a cohort of healthy, relatively young individuals. The total plasma cfDNA concentrations were sourced from healthy individuals below 47 years, as reported by (Meddeb et al. 2019). We use data analyzed based on plasma samples from healthy individuals to estimate the proportion of cfDNA originating from specific cell types (Loyfer et al. 2023). These values were then compared to the potential DNA flux resulting from homeostatic cellular turnover, estimated for reference healthy males aged between 20 and 30 (Sender and Milo 2021). In our analysis, we considered various sources of uncertainty, including inter-individual variation, variability in the timing of sample collection, and analytical precision (Madsen et al. 2019; Meddeb et al. 2019). These factors collectively contributed to an uncertainty factor of less than 3. Importantly, this level of uncertainty does not alter our conclusion regarding the relatively small fraction of DNA present in plasma as cfDNA. Furthermore, we acknowledge that age and sex can impact total cfDNA concentration, as demonstrated by (Meddeb et al. 2019), with potential variations of up to 30%. However, as the results of our analysis present a much larger difference, these effects do not change the conclusions drawn from our analysis. Nevertheless, age and health status may influence the proportion of cfDNA originating from specific cell types and their corresponding cellular turnover rates. Consequently, the ratios themselves may vary in the elderly population or individuals with underlying health conditions.”

2) "cfDNA fragments are not created equal". Recent studies demonstrate that cfDNA composition vary with disease state. For example, cfDNA GC content, fraction of short fragments, and composition of some genomic elements increase in heart transplant rejection compared to no-rejection state (Agbor-Enoh, Circulation, 2021). The genomic location and disease state may therefore be important factors to consider in these analyses.

In this study, we addressed the total amount of cfDNA in healthy individuals without regard to GC content, representation of different genomic regions, or fragment length, as the goal was to understand if cell death rates are fully accounted for by cfDNA concentration. We agree that it will be interesting to study the relative representation of the genome in cfDNA and the processes that determine cfDNA concentration in pathologies beyond the rate of cell death. These topics for future research fall beyond this study's scope.

3) Alternative sources of DNA production should be considered. Aside from cell death, DNA can be released from cells via active secretion. This and other additional sources of DNA should be considered in future studies. The distinct characteristics of mitochondrial DNA to genomic DNA should also be considered.

We know only a few specific cases whereby DNA is released from cells that are not dying. These include the release of DNA from erythroblasts and megakaryocytes to generate anucleated erythrocytes and platelets (Moss et al. 2022, cited in our paper) and the release of NETs from neutrophils.

The presence of cfDNA fragments originating from megakaryocytes and erythroblasts indicates the elimination of megakaryocytes and erythroblasts and the birth of erythrocytes and platelets. However, the considerations in the rest of the paper still apply: the concentration of cfDNA from these sources is far lower than expected from the cell turnover rate.

Concerning NETosis: the presence of cfDNA originating in neutrophils that have not died would reduce the concentration of cfDNA from dying neutrophils and thus further increase the discrepancy, which is the topic of our study (under-representation of DNA from dying cells in plasma).

We updated a paragraph in the discussion regarding this issue:

“A comparison between the different types of cells shows a trend in which less DNA flux from cells with higher turnover gets to the bloodstream. In particular, a tiny fraction (1 in 3x104) of DNA from erythroid progenitors arrives at the plasma, indicating an extreme efficiency of the DNA recovery mechanism. Erythroid progenitors are arranged in erythroblastic islands. Up to a few tens of erythroid progenitors surround a single macrophage that collects the nuclei extruded during the erythrocyte maturation process (pyrenocytes) (Chasis and Mohandas 2008). The amount of DNA discarded through the maturation of over 200 billion erythrocytes per day (Sender and Milo 2021) exceeds all other sources of homeostatic discarded DNA. Our findings indicate that the organization of dedicated erythroblastic islands functions highly efficiently regarding DNA utilization. Neutrophils are another high-turnover cell type with a low level of cfDNA. When contemplating the process of NETosis (Vorobjeva and Chernyak 2020), the existence of cfDNA originating from live neutrophils would potentially diminish the concentration of cfDNA released by dying neutrophils, thereby amplifying the observed ratio for this particular cell type. The overall trend of higher turnover resulting in a lower cfDNA to DNA flux ratio may indicate similar design principles, in which the utilization of DNA is better in tissues with higher turnover. However, our analysis is limited to only several cell types (due to cfDNA test and deconvolution sensitivities), and extrapolation to cells with lower cell turnover is problematic.”

We neglected mitochondrial DNA, as it is not measured in methylation cell-of-origin analysis. Similarly to the argument above, if some of the total DNA measured in plasma is in fact mitochondrial, this would mean that genomic cfDNA concentration is actually lower than the estimates, meaning that an even smaller fraction of DNA from dying cells is measured in plasma.

Given that we assume the effect of revision is conditional on the nature of that revision, it's not clear to me what "DAIR vs. Revision" means for late infection silo,

Just thinking through it for myself...

The options are:

• A = DAIR -> B = 12w
• A = Revision(one) -> B in {12w, 6w}
• A = Revision(two) -> B in {12w, 7d}
• C in (no rifampicin, rifampicin)

Currently, (just focusing on surgery/duration, and making 12w the reference for all groups rather than 7w/6d) cell parameters as specified in the model are:

$$ \begin{matrix} \text{DAIR} \ \text{Revision(one), 12w} \ \text{Revision(two), 12w} \ \text{Revision(one), 6w} \ \text{Revision(two), 7d} \end{matrix} \begin{pmatrix} \alpha \ \alpha + \beta_A \ \alpha + \beta_A \ \alpha + \beta_A + \beta_{B1} \ \alpha + \beta_A + \beta_{B2} \end{pmatrix} $$

So, effect of one-stage + 12 w assumed equal to the effect of two-stage + 12w, then revision type specific deviations from those. And "DAIR vs Revision" (beta_A) is really DAIR vs. weighted average of one-stage 12w and two-stage 12w, i.e. ignores the duration options.

I'm guessing this is the only randomised comparison we can make: a weighted average of one/two + 12w is the "default" revision.

As an alternative, I assume it's plausible that one-stage 12w and two-stage 12w differ due to clinician selection of one/two stage. So there may be preference (or maybe it just makes things messy) to have something like

$$ \begin{matrix} \text{DAIR} \ \text{Revision(one), 12w} \ \text{Revision(two), 12w} \ \text{Revision(one), 6w} \ \text{Revision(two), 7d} \end{matrix} \begin{pmatrix} \alpha \ \alpha + \beta_{A1} \ \alpha + \beta_{A2} \ \alpha + \beta_{A1} + \beta_{B1} \ \alpha + \beta_{A2} + \beta_{B2} \end{pmatrix} $$

Noting that \beta_{A1} and \beta_{A2} aren't "causal" in the sense that any differences could just be due to selection bias rather than differences in effectiveness of one/two stage.

The effect of revision versus DAIR will depend on what "revision" means. We can't just compare one-stage 12 weeks to DAIR vs 12 weeks because surgeon's choose who gets one-stage. Only comparison that seems to make sense is weighted combination of one/two stage, with weight as observed in the trial. I think that comparison makes sense, but maybe not.

Assume that $$p_{A1}$$ is the proportion randomised to revision who are selected to receive one-stage and $$1 - p_{A1}$$ the proportion selected to receive two-stage. Then the comparison for any revision versus DAIR might be taken to mean

$$ p_{A1}(0.5\beta_{A1} + 0.5(\beta_{A1} + \beta_{B1}))\ + (1 - p_{A1})(0.5\beta_{A2} + 0.5(\beta_{A2} + \beta_{B2})) $$

which just explicitly allows for the differences. Or some other combination of groups, where we assume that selection of one/two stage in the trial is the same in the population. Presumably though there are issues in interpreting such a comparison as "causal", unless also adjust for factors determining one/two stage selection.

The \beta_{A1} and \beta_{A2} are necessary for estimation of the duration effects, unless willing to assume no differences between one/two stage 12w.

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

We thank the reviewers for their reading of our manuscript, which we believe has led to substantial improvements.

To aid clarity, we have split Fig. 1 into three separate figures.

For convenience, we have put all major changes in the text in blue.

Reviewer #1

Evidence, reproducibility and clarity

Summary: Hui et al. tackle a crucial question in biology: what factors influence the preference for carbon sources in yeasts?

They reveal that the growth rate on palatinose exceeds that on glucose,

The above statement is incorrect --- we think the reviewer may have confused sugars.

despite palatinose utilization being repressed in the presence of glucose. Consequently, the favored carbon source does not necessarily align with the one supporting the fastest growth rate. The study also delves into potential regulatory mechanisms governing carbon source preference and dismisses certain existing theories, such as the general carbon flux sensing mechanism proposed by Okano et al. [25].

Major comments: None

Minor comments:

The authors suggest that a higher growth rate implies a higher glycolytic flux (l63), a crucial assumption underpinning their interpretation of the absence of a ``general carbon flux sensing mechanism' (l65). To substantiate this significant conclusion, they could calculate the extracellular uptake fluxes (based on the time-course concentrations of biomass and substrates).

This suggestion is a good one, but unfortunately the number of data points in the new Fig. 3 are insufficient to estimate the uptake flux reliably.

To address whether glycolytic flux increases, we have added a new paragraph to the introduction explaining how all the sugars we consider feed upper glycolysis, providing either its first or second metabolite. We therefore think it highly likely that any differences in growth rate are generated by differences in glycolytic flux. Indeed, Hackett et al., 2016, showed that the glycolytic flux increases with growth rate when they changed extracellular glucose concentrations. We now include this reference in the Discussion.

The accumulation of certain by-products is known to be toxic, reducing cellular growth rate (e.g., acetate DOI: 10.1038/srep42135, ethanol DOI: 10.1016/B978-0-12-040308-0.50006-9, etc.), while they can also enhance growth under specific conditions (e.g., acetate DOI: 10.15252/embj.2022113079). Considering this is crucial to rule out certain hypotheses, such as the possibility that a by-product produced during growth on the first carbon source would not modulate growth on the second carbon source, potentially influencing the growth rate differentially in each phase. Although the authors use mutant strains to eliminate the role of some C2 compounds (acetate and ethanol), alternative pathways could be implicated in the (co-)utilization of these by-products. This aspect should be discussed, and ideally, the authors could quantify the time-course concentrations of by-products to assess their potential role.

We agree with the reviewer that extracellular acetate and ethanol may inhibit growth, although budding yeast might be less sensitive than E. coli, the subject of most of the studies provided.

Nevertheless, we think it unlikely that these chemicals modify the decision-making we see. First, the icl1Δ mutant we tested is unable to consume ethanol (Fernandez et al., 1992) or acetate (Lee et al., 2011) --- we now include these references in the SI --- and yet has wild-type behaviour (Fig. S2D). Second, we observe that isomaltase expression strongly decreases in the presence of galactose when we grow cells in a microfluidic device (Fig. S4), just like it does in batch culture (Fig. 3A), even though the constant flow of medium through the device removes any chemicals the cells excrete.

The general flux-sensing regulatory mechanism proposed by Okano et al. [25], which has been dismissed by this study, has recently been questioned, as discussed in DOI: 10.15252/embj.2022113079. This aspect should be included in the discussion.

Okano et al. studied E. coli while we study budding yeast. We therefore have shown that the understanding for that organism does not transfer to our eukaryotic example. We suspect that control in budding yeast combines both flux-sensing and specific regulation, as we say in the discussion, and so we consider our results to build on those of Okano et al.

Significance

Strengths & limitations: The work is robust, and the experiments in the study have been appropriately designed and conducted. The primary question of this study has been tackled using a combination of experimental and computational methods to thoroughly assess various regulatory and functional aspects. However, there are gaps in the data that could enhance key conclusions, notably the absence of glycolytic flux measurements. Moreover, further evidence is needed to substantiate the assertion that by-products do not play a role in carbon source preference.

Advance: This study represents a significant step forward in comprehending the nutritional strategy of microbes. The authors demonstrate that the preferred carbon source may not necessarily be the one supporting the fastest growth rate. Furthermore, they dismiss certain theories that have been proposed to explain the growth strategy of microbes on mixed carbon sources.

Audience: By addressing a fundamental question in life science, this work is important in the field of biology in general and of particular interest in systems biology, biotechnology, synthetic biology, and health. Consequently, it will be of interest to a broad audience.

Reviewer #2

Evidence, reproducibility and clarity

Summary: The authors have used microtiter plates to produce growth profiles on combinations of different sugars. From this data they have evaluated whether the sugars are co-consumed or if there is a preference for either sugar, seen as a diauxic shift. They found diauxie between galactose and the disaccharide palatinose, but co-consumption between palatinose and fructose. They further used strains with perturbations in their GAL regulon to attempt to explain this discrepency.

Major comments:

I unfortunately found a large portion of the present manuscript unintelligable.

Firstly, figures were incorrect to the point I could not dechiffre them: Figure 2A-C have black solid and dashed lines in the legend that are not found in the graph, instead there are orange and blue dashed lines in the graph with no legends. Figure 4C has no description of the y-axis. The growth rates in Figure 1C are very hard to follow, and there are definitely local maxima in both the blue and green profiles that are not being discussed (at 15-20 h). I cannot evaluate the conclusions drawn from the data until these issues have been resolved.

We apologise for the difficulties experienced by this reviewer.

The black lines in the old Fig. 2's legend, now Fig. 4, explain the different styles used: dashed lines are for single sugars regardless of their concentration and full lines are for mixtures regardless of their concentration. We now explicitly say this in the caption.

We have fixed the missing label in what is now Fig. 6C and have moved the statement that we are showing two biological replicates for each set of concentrations earlier in Fig. 2's caption.

We now explore the meaning of the shoulder for the fructose-palatinose mixture in Fig. 2B in the Discussion. This point is not a local maximum, unlike the case for diauxie, because the growth rate always decreases. The shoulder for the glucose-palatinose mixture was likely an artefact generated by measurement noise at low ODs because it was not present when we repeated the experiment. We now use that data for Fig. 2A & B. We also include a new Fig. S5 showing that there are sucrose-palatinose concentrations too that have a similar shoulder.

Secondly, the language in the Results and Discussion sections is confusing. Alternating between present and imperfect tense as well as active and passive form makes it hard to distinguish the authors own results from literature findings (Results are usually written in passive, imperfect tense). Examples are found on lines 24, 29, 37-38, 59, 84, 131, and 165.

We have made both sections flow more smoothly with substantial re-writing. As before, we cite all results that are not our own.

The authors also do not consider the differences and similarities in catabolic pathways for assimilation of galactose, fructose and palatinose. Even if they do not see a reason to continue that as a possible explanation for the co-consumption between fructose and palatinose a discussion of why it is disregarded would not be out of place here.

A good point, and we now state in the Introduction that all the sugars we study feed upper glycolysis.

Significance

There is some novelty to the authors findings, but I would argue it is being overstated in the present manuscript. Some examples of studies looking at catabolite repression, the main cause of diauxie, of sugars other than glucose can be found in: Simpson-Lavy and Kupiec (2019), Gancedo (1998), Prasad and Venkatesh (2008) and Borgstrom et al (2022).

We strongly disagree with this statement. The papers cited do not address, as we do, the co-consumption between two sugars neither of which is glucose. Where they study two sugars, they always study glucose.

Simpson-Lavy and Kupiec, 2019, investigate the interaction between acetate and ethanol, neither of which are sugars. Further, they are not independent carbon sources because cells convert ethanol into acetate when catabolising ethanol.

Gancedo, 1998, is a review of glucose repression and describes how glucose represses the expression of genes for other sugars. Although Gancedo mentions ``galactose repression', this repression is of genes encoding enzymes for gluconeogenesis and the TCA and glyoxylate cycles, not of other sugar regulons, our subject.

Prasad and Venkatesh, 2008, also focus on glucose and the well studied diauxie between glucose and galactose.

Borgstrom et al., 2022, focus too on glucose and growth on glucose and xylose in recombinant strains. The standard laboratory strains we study have not be artificially engineered to consume xylose. They do mention that galactose causes repression of TPS1, which encodes an enzyme that synthesises the storage carbohydrate trehalose. This repression is again not of a sugar catabolic regulon, our subject.

I would not say that the field would be significantly advanced by the publication of this manuscript, and the authors have themselves not explained the application of futhering the understanding palatinose metabolism in yeast. As mentioned above, the catabolite repression potential of galactose is already known, it just hasn't been shown for palatinose specifically before.

We again strongly disagree. Our findings are novel. The reviewer did not provide any evidence for galactose repression of other sugar regulons, which is not widely recognised as we emphasised in the Discussion. We believe that the reviewer has confused the known "galactose repression' of gluconeogenic or TCA-cycle genes with our new report of repression of other sugar regulons in the presence of the sugar catabolised by the regulon.

I would recommend a complete rewrite of the manuscript as presented, with a lower stated novelty, clearer language and comprehensible figures.

Reviewer #3

Evidence, reproducibility and clarity

Summary: Microbes grow at different growth rates in different carbon sources. When more than one carbon sources are present in the media microbes often show a preference over certain carbon sources, and 'non-preferred' carbons sources are used only when the preferred carbon source is exhausted in the media, this process called diauxic shift.

Why microbes exhibit such utilization preference over certain carbon sources, is an interesting question in microbiology and evolutionary biology, and the molecular mechanisms that enable microbes to preferentially use one carbon over another is worth investigating. It is intuitive to think that microbes will prefer to use a carbon source that confers maximum growth rate, but when tested experimentally it has been often observed that a carbon source in which microbes grow at sub optimal growth rate is actually preferentially used.

Although the reviewer states that "it has been often observed that a carbon source in which microbes grow at sub optimal growth rate is actually preferentially used“, we are unaware of this work and would appreciate references, particularly for budding yeast. The most systematic study we know, in E. coli by Aidelberg et al., 2014 --- reference 13, concludes that "the faster the growth rate, the higher the sugar on the hierarchy“, the opposite behaviour.

In this study authors demonstrate that budding yeast prefer to use galactose over palatinose, but not over sucrose or fructose where all three sugars can support faster growth rate compared to palatinose. Authors presented data where preferential galactose use and diauxic shift is observed in the growth curve when galactose and palatinose or glucose and palatinose combinations were used.

No diauxic shift was observed in the growth curve when fructose-palatinose, or sucrose-palatinose combination were used. In fructose-palatinose and sucrose-palatinose combinations growth curves agree more with co-utilization strategies. Authors used transcriptomics and genetic perturbations to decipher the molecular mechanism of such preferential carbon use, and reports preference of galactose over palatinose is achieved by preventing positive feedback of MAL regulon, which encodes the genes for palatinose catabolism. We found this observation is interesting and the molecular mechanism of such preferential carbon use is nicely described in this paper. We also find claims authors made are well supported by experiments. Although catabolite repression and diauxic transitions are known in yeast, and authors also pointed out such previous references, but preferential use of a slower carbon source, i.e. galactose over at least one other fast-growing carbon is interesting enough for publication. We would like to support the publication of this article, but we have major concerns about the data analysis and data presentation. Authors must address our concerns which are mentioned below.

Major comments:

1. This study mainly hinges on growth rate measurements, but we found growth rates are not properly represented in the figures. Growth curves are always shown in linear scale, which makes it almost impossible to compare fast and slow growth when presented in same plot. All growth curves must be shown on log scale.

We have changed all growth curves to log2 scale, following New et al., 2014, rather than Monod's choice of linear scale that we had originally.

Our conclusions are unaffected.

1. Growth rates of the Yeast strain growing individual single carbon sources (galactose, palatinose, sucrose and fructose) should be shown as a figure panel and t-test should be performed to conclude if the individual growth rates are significantly different or not.

We already showed these growth rates in their own panel in Fig. 1B. Following the reviewer's suggestion, we have now added their statistical significance to the caption.

1. Growth phase, lag phase, diauxic shift and post shift growth should be clearly shown in figure 2 and 4, each phase should be clearly marked, carbons used in each phase should be mentioned on the plot. Also, the growth curve must be plotted using log scale.

Although we have changed all growth curves to log scale, we decided against include this additional labelling for two reasons. First, we are presenting evidence that some of the growth we observe is diauxic and labelling the curves as diauxic before we discuss this evidence undermines that discussion. Second, any further labels would clutter the figures, and we believe would hinder rather than help the reader.

Instead we changed the colour scheme and the boldness of the diauxic growth curves in Fig. 2, which we hope the reviewer agrees adds the clarity they felt was missing.

1. Authors has taken in account that MAL12 gene overexpression causes long lag when cells need to switch to maltose from glucose, and shown deletion of IMA1 decreases the lag with subsequent 2% growth rate increase in palatinose. How significant is this increase?

We have confirmed the statistical significance through a t-test and added the results to the caption of Fig. 6C.

1. Authors have an interesting observation that in sucrose-palatinose and fructose palatinose combinations, most probably co utilization of the carbons is taking place. Authors should discuss this in more details. In galactose-palatinose scenario intracellular galactose-based repression of gal80 and subsequent lack of feed forward of the Mal regulon is expected to stop co-utilization of palatinose. As authors have RNA seq data, can they make predictions for other carbon pairs, where sequential utilization can occur based on their model?

We agree and have added more discussion of the fructose- and sucrose-palatinose mixtures to the Discussion and a new figure, Fig. S5.

Our RNAseq data reveals the difference in gene expression caused by an active versus an inactive GAL regulon. In Fig. S11, we show that the hexose transporters HXT2 and HXT7 are down regulated in 0.1% fructose when the GAL regulon is active, perhaps implying that cells are able to prioritise galactose over other hexoses. Nevertheless, to predict if particular carbon sources are therefore favoured, we would need to know whether cells use specific hexose transporters to drive growth on different carbon sources, which has been little investigated.

Minor comments:

1. In figure 5, authors attempted to summarize the model, which is informative, but it will be more useful for non-specific reader if a cell-based cartoon, with transports on surface and catabolic enzymes inside is also added.

We have re-designed Fig. 5, now Fig. 7, following this suggestion and agree it improves clarity.

In this schematic diagram, switch from galactose (blue line) to red line (palatinose) shows a mixed color zone, it's a bit confusing, as this represents a bi-stable state. Authors should clearly comment on possibility of biostability while discussing their proposed mechanism.

In the new figure, this part has been removed.

1. The author may want to put their work in the context of other recent observations that bacteria do not try to maximize their growth rates in many conditions. Fast growth is often associated with expansive tradeoffs, and a carbon source which confers fast growth rate may confer selective disadvantage. Thus, there are evolutionary benefits of sub-optimal growth, which could be discussed in the manuscript. In this regard a recent study (bioRxiv (2023) doi:10.1101/2023.08.22.554312.) has established the link between resource allocation strategies, growth rates and tradeoffs, which may be taken in account while discussing. Are there any known tradeoffs, when galactose is used over palatinose and which is not the case sucrose or fructose?

This is an interesting reference looking at growth on a single carbon source. We are unaware of similar tradeoffs relevant to our study. For example, we see little evidence for a constraint on the proteome because in a strain with a constitutively active GAL regulon there is no change in phenotype if we delete the genes for the three highly expressed GAL enzymes (Fig. S6B). Nevertheless and as we state in the penultimate paragraph of the Discussion, we agree that such a constraint must exist, although perhaps this constraint is ecological.

Referees cross-commenting

As other reviewers pointed out, this study has merit and addressed interesting questions, but needed to be written well in a more understandable form, we agree with this assessment. Also figures must be made much clearer, as all of the reviewers pointed out. In summary, this is an interesting study, but needs some work before publication.

Significance

General assessment: Strength and limitations:

This study addressed an interesting question regarding resource preference and growth rate optimization in microbes. This is an important question in the field. Study is well designed and claims are backed up with experimental results. One of the limitations of the study is lack of predictability. Authors explained the mechanism for one pair of carbon sources, but how applicable that will be in general is not clear.

We would argue that one of our important findings is to demonstrate that the scientific community is missing the information needed to make such predictions. We provide a counter example to the generally accepted belief that accurate predictions can be made using growth rates. Our work poses the question: what then are the physiological variables required to predict how a cell will consume a pair of carbon sources?

Advance: This study helps to advance our knowledge. Their observation regarding preferential utilization of a carbon source which supports slower growth over a carbon source which can support faster growth, and the molecular mechanism provided will help researchers to understand resource allocation strategies better.

Audience: Microbiology, systems biology, evolutionary biology, fermentation and bio process engineering research.

Reviewer expertise: Biochemistry, systems biology, metabolic strategies and tradeoffs in microbes, microbial ecology.

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

Review of the paper by Yu Huo et al.

Summary:

Microbes grow at different growth rates in different carbon sources. When more than one carbon sources are present in the media microbes often show a preference over certain carbon sources, and 'non-preferred' carbons sources are used only when the preferred carbon source is exhausted in the media, this process called diauxic shift. Why microbes exhibit such utilization preference over certain carbon sources, is an interesting question in microbiology and evolutionary biology, and the molecular mechanisms that enable microbes to preferentially use one carbon over another is worth investigating. It is intuitive to think that microbes will prefer to use a carbon source that confers maximum growth rate, but when tested experimentally it has been often observed that a carbon source in which microbes grow at sub optimal growth rate is actually preferentially used. In this study authors demonstrate that budding yeast prefer to use galactose over palatinose, but not over sucrose or fructose where all three sugars can support faster growth rate compared to palatinose. Authors presented data where preferential galactose use and diauxic shift is observed in the growth curve when galactose and palatinose or glucose and palatinose combinations were used.

No diauxic shift was observed in the growth curve when fructose-palatinose, or sucrose-palatinose combination were used. In fructose-palatinose and sucrose-palatinose combinations growth curves agree more with co-utilization strategies. Authors used transcriptomics and genetic perturbations to decipher the molecular mechanism of such preferential carbon use, and reports preference of galactose over palatinose is achieved by preventing positive feedback of MAL regulon, which encodes the genes for palatinose catabolism. We found this observation is interesting and the molecular mechanism of such preferential carbon use is nicely described in this paper. We also find claims authors made are well supported by experiments. Although catabolite repression and diauxic transitions are known in yeast, and authors also pointed out such previous references, but preferential use of a slower carbon source, i.e. galactose over at least one other fast-growing carbon is interesting enough for publication. We would like to support the publication of this article, but we have major concerns about the data analysis and data presentation. Authors must address our concerns which are mentioned below.

Major comments:

1. This study mainly hinges on growth rate measurements, but we found growth rates are not properly represented in the figures. Growth curves are always shown in linear scale, which makes it almost impossible to compare fast and slow growth when presented in same plot. All growth curves must be shown on log scale.
2. Growth rates of the Yeast strain growing individual single carbon sources (galactose, palatinose, sucrose and fructose) should be shown as a figure panel and t-test should be performed to conclude if the individual growth rates are significantly different or not.
3. Growth phase, lag phase, diauxic shift and post shift growth should be clearly shown in figure 2 and 4, each phase should be clearly marked, carbons used in each phase should be mentioned on the plot. Also, the growth curve must be plotted using log scale.
4. Authors has taken in account that MAL12 gene overexpression causes long lag when cells need to switch to maltose from glucose, and shown deletion of IMA1 decreases the lag with subsequent 2% growth rate increase in palatinose. How significant is this increase?
5. Authors have an interesting observation that in sucrose-palatinose and fructose palatinose combinations, most probably co utilization of the carbons is taking place. Authors should discuss this in more details. In galactose-palatinose scenario intracellular galactose-based repression of gal80 and subsequent lack of feed forward of the Mal regulon is expected to stop co-utilization of palatinose. As authors have RNA seq data, can they make predictions for other carbon pairs, where sequential utilization can occur based on their model?

Minor comments

1. In figure 5, authors attempted to summarize the model, which is informative, but it will be more useful for non-specific reader if a cell-based cartoon, with transports on surface and catabolic enzymes inside is also added.

In this schematic diagram, switch from galactose (blue line) to red line (palatinose) shows a mixed color zone, it's a bit confusing, as this represents a bi-stable state. Authors should clearly comment on possibility of biostability while discussing their proposed mechanism. 2. The author may want to put their work in the context of other recent observations that bacteria do not try to maximize their growth rates in many conditions. Fast growth is often associated with expansive tradeoffs, and a carbon source which confers fast growth rate may confer selective disadvantage. Thus, there are evolutionary benefits of sub-optimal growth, which could be discussed in the manuscript. In this regard a recent study (bioRxiv (2023) doi:10.1101/2023.08.22.554312.) has established the link between resource allocation strategies, growth rates and tradeoffs, which may be taken in account while discussing. Are there any known tradeoffs, when galactose is used over palatinose and which is not the case sucrose or fructose?

Referees cross-commenting

As other reviewers pointed out, this study has merit and addressed interesting questions, but needed to be written well in a more understandable form, we agree with this assessment. Also figures must be made much clearer, as all of the reviewers pointed out. In summary, this is an interesting study, but needs some work before publication.

Significance

General assessment: Strength and limitations: This study addressed an interesting question regarding resource preference and growth rate optimization in microbes. This is an important question in the field. Study is well designed and claims are backed up with experimental results. One of the limitations of the study is lack of predictability. Authors explained the mechanism for one pair of carbon sources, but how applicable that will be in general is not clear.

Advance: This study helps to advance our knowledge. Their observation regarding preferential utilization of a carbon source which supports slower growth over a carbon source which can support faster growth, and the molecular mechanism provided will help researchers to understand resource allocation strategies better.

Audience: Microbiology, systems biology, evolutionary biology, fermentation and bio process engineering research.

Reviewer expertise: Biochemistry, systems biology, metabolic strategies and tradeoffs in microbes, microbial ecology.

Author Response

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

We would firstly like to thank all reviewers for their comments and support of this manuscript.

Reviewer #1 (Recommendations For The Authors):

No further recommendations.

Reviewer #2 (Recommendations For The Authors):

All of my comments have been sufficiently addressed.

Reviewer #3 (Recommendations For The Authors):

Thanks for responding to my former recommendations constructively. I believe these points have been fully addressed in this new version.

However, I have not seen any comments on the points I raised in my former public review concerning the I-2 dependence of the FonSIX4 cell death. Do you know whether FonSIX4 would trigger cell death in tissues not expressing any I-2?

We are a little confused concerning this comment. I-2 is a different class of resistance protein (NLR) that recognises Avr2 and this is likely to be intracellular. From the previous public review, we believe reviewer 3 may have been asking us to clarify the dependence of I (MM or M82) on FonSIX4 cell death. We have performed these controls by expressing FonSIX4 and associated FonSIX4/Avr1 chimeras in N. benthamiana (with the PR-1 signal peptide for efficient secretion of effectors) and it does not cause cell death in the absence of the I receptor – see S11F Fig. This was not explicitly conveyed in text so we have included the following in text: “Using the N. benthamiana assay we show FonSIX4 is recognised by I receptors from both cultivars (IM82 and iMoneymaker) and cell death is dependent on the presence of IM82 or iMoneymaker (Fig 5B, S11 Fig).”

I still recommend discussing whether the Avr1 residues crucial for Avr activity are in the same structural regions of the C-terminal domain where previous work has identified residues under diversifying selection in symbiotic fungal FOLD proteins.

The region important for recognition does encompass some residues within the structural region identified to be under diversifying selection in FOLD effectors from Rhizophagus irregularis previously reported (two residues within one beta-strand). However, we also see residues that don’t overlap to this area. We also note that the mycFOLD proteins analysed in symbiotic fungi are heavily skewed towards strong structurally similarity with FolSIX6 (similar cysteine spacing within both N and C-domains and structural orientation of the N and C-domains) rather than Avr1. We are under the impression that Avr1 was not included in the analysis of diversifying selection in symbiotic fungal FOLD proteins, it also is unclear to us if close Avr1 homologues are present. With this in mind, and considering our already lengthy discussion (as previously highlighted during reviewer), we have decided not to include further discussion concerning this point.

---

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

We would like to thank the editor(s) and reviewers for their work concerning our manuscript. Most of the suggested changes were related to text changes which we have incorporated into the revised version. Please find our response to reviewers below.

Reviewer #1 (Recommendations For The Authors):

I only have very minor suggestions for the authors. The first one comes from reading the manuscript and finding it very dense with so many acronyms. This will limit the audience that will read the study and appreciate its impact. This is more noticeable in the Results, with many passages that I would suggest moving to Methodology.

We thank reviewer 1 for their very positive review. We understand that due to the nature of this study, which includes many protein alleles/mutations that were expressed with different boundaries etc., it is difficult to achieve this. Reviewer 2 asked for more details to be provided. We hope we have achieved a nice balance in the revised manuscript.

Something else that would facilitate the reading of the manuscript is the effectors name. The authors use the SIX name or the Avr name for some effectors and it makes it difficult to follow up.

We have tried to make this consistent for Avr1 (SIX4), Avr2 (SIX3) and Avr3 (SIX1). Other SIX effectors are not known Avrs so the SIX names were used.

Reading the manuscript and seeing how in most of the sections the authors used a computational approach followed by an experimental approach, I wonder why Alphafold2-multimer was not used to investigate the interaction between the effector and the receptor?

This is a great suggestion, we have certainly investigated this, however to date there is no experimental evidence to directly support the direct interaction between I and Avr1. Post review, we spent some time trying to capture an interaction using a co-immunoprecipitation approach however to date we have not been able to obtain robust data that support this. We are currently looking to study this utilising protein biophysics/biochemistry but this work will take some time.

Reviewer #2 (Recommendations For The Authors):

We thank reviewer 2 for the very thorough editing and recommendations. We have incorporated all minor text edits below into the manuscript.

Line 43: perhaps "Effector recognition" instead of "Effector detection", to be consistent with line 51?

Line 60: Change to "leads".

Line 79: Italicise Avr2.

Line 94: Add the acronym ETI in parentheses after "effector-triggered immunity".

Line 106: "(Leptosphaeria Avirulence-Supressing)" should be "(Leptosphaeria Avirulence and Supressing)".

Line 112: Change "defined" to "define".

Line 119: Spell out the species name on first use.

Line 205: Glomeromycota is a division rather than a genus. Consistent with Fig 2, it also does not need to italicized.

Line 207: Change "basidiomycete" to "Division Basidiomycota", consistent with Fig 2.

Line 214: Change "alignment of Avr1, Avr3, SIX6 and SIX13" to "alignment of the mature Avr1, Avr3, SIX6 and SIX13 sequences".

Line 324: Change "solved structures" to "solved protein structures".

Line 335: Spell out acronyms like "MS" on first use in figure legends. Also dpi in other figure legends.

Line 341: replace "effector-triggered immunity (ETI)" with "(ETI)" - see comment on Line 94.

Line 370: Change "domains" to "domain".

Line 374: In the title, change "C-terminus" to C-domain", consistent with the rest of the figure legend.

Line 404: Change "(basidiomycetes and ascomycetes)" to "(Basidiomycota and Ascomycota fungi)", consistent with Fig 2C.

Line 416: Change "in" to "by".

Line 427: un-italicize the parentheses.

Line 519: First mention of NLR. Spell out the acronym on first use in main text. S5 and S11 figure titles should be bolded.

Line 852: Replace "@" with "at".

S4 Table: Gene names should be italicised.

S5 Table: Needs to be indicated that the primer sequences are in the 5´-3´ orientation.

With regards to the Agrobacterium tumefaciens-mediated transient expression assays involving co-expression of the Avr1 effector and I immune receptor, the authors need to make clear how many biological replicates were performed as this information is only provided for the ion leakage assay.

We have added these data to the figure legend

Line 57: For me, the text "Fol secretes a limited number of structurally related effectors" reads as Fol secretes structurally related effectors, but very few of them are structurally related. Perhaps it would be better to say that the effector repertoire of Fol is made up of proteins that adopt a limited number of structural folds, or that the effector repertoire can be classified into a reduced set of structural families?

This edit has been incorporated.

Lines 66-67: Subtle re-wording required for "The best-characterized pathosystem is F. oxysporum f. sp. lycopersici (Fol)", as a pathosystem is made up of a pathogen and its host. Perhaps "The best-characterized pathosystem involves F. oxysporum f. sp. lycopersici (Fol) and tomato".

Sentence has been reworded.

Line 113 and throughout: Stick with one of "resistance protein", "receptor", "immune receptor" and "immunity receptor" throughout the manuscript.

We have decided to use both receptor and immunity receptor as not all receptors investigated in the manuscript provide immunity.

Lines 149-150: The title does not fully represent what is shown in the figure. The text "that is unique among fungal effectors" can be deleted as there is nothing in Fig 1 that shows that the fold is unique to fungal effectors.

Figure title has been changed.

Line 173: The RMSD of Avr3 is stated as being 3.7 Å, but in S3 Fig it is stated as being 3.6 Å.

This was a mistake in the main text and has been corrected.

Lines 202-204: This sentence needs to be reworded, as the way that it is written implies that the Diversispora and Rhizophagus genera are in the Ascomycota division. Also, "Ascomycetes" should be changed to "Ascomycota fungi", consistent with Fig 2.

Sentence has been reworded.

Line 233: "Scores above 8". What type of scores? Z-scores?

These are Z-scores. This has been added in text.

Lines 242-246: It is stated that SIX9 and SIX11 share structural similarity to various RNA-binding proteins, but no scores used to make these assessments is given. The scores should be provided in the text.

Z-scores have been added.

Fig 4A: SIX3 should be Avr2, consistent with line 292. The gene names should be italicised in Fig 4A.

SIX3 was changed to Avr2. Gene names have been italicised.

Line 356: Subtle rewording required, as "co-infiltrated with both IM82 and iMoneymaker" implies that you infiltrated with protein rather than Agrobacterium strains.

Sentence has been reworded.

Fig 5A, Fig 5C and Line 380: Light blue is used, but this looks grey. Perhaps change colour, as grey is already used to show the pro-domain in Fig 5A (or simply change the colour used to highlight the pro-domain)?

Colour depicting the C-domain was changed.

Lines 530-531: This text is no longer correct. Rlm4 and Rlm3 are now known to be alleles of Rlm9. See: Haddadi, P., Larkan, N. J., Van deWouw, A., Zhang, Y., Neik, T. X., Beynon, E., ... & Borhan, M. H. (2022). Brassica napus genes Rlm4 and Rlm7, conferring resistance to Leptosphaeria maculans, are alleles of the Rlm9 wall‐associated kinase‐like resistance locus. Plant Biotechnology Journal, 20(7), 1229.

We thank the reviewer for picking this up. This text has been updated.

Line 553: Provide more information on what the PR1 signal peptide is.

More information about the PR1 signal peptide has been added.

Lines 767-781: Descriptions and naming conventions of proteins throughout the figure legend need to be consistent and better reflect their makeup. For example, I think it would be best to put the sequence range after each protein mentioned - e.g. Avr118-242 or Avr159-242 instead of Avr1, PSL1_C37S18-111 instead of PSL1_C37S, etc. Furthermore, it is often stated that a protein is full-length when it lacks a signal peptide - my thought is that if a proteins lack its signal peptide, it is not full-length. The acronym "PD" also needs to be spelled out as "pro-domain (PD)" in the figure legend.

We have incorporated sequence range for proteins that were produced upon first use. Sequence ranges that were modelled in AlphaFold2 were not added in text because they can be found in Supplementary Table 3.

Lines 853-845: It is stated the sizes of proteins are indicated above the chromatogram in S10 Fig, but this is not the case. It is also not clear from S10B Fig that the faint peaks correspond to the peaks in the Fig 4B chromatogram. In S10D Fig, the stick of C58S is difficult to see. Perhaps change the colour or use an arrow/asterisk?

Protein size estimates have been added above the chromatogram. Added text to indicate that the faint peaks correspond to peaks in Fig 4B. Added an asterisk in S10D Fig to identify the location of C58.

S14 Fig is not mentioned/referenced in the main text of the manuscript.

This was a mistake and has been added.

The reference list needs to be updated to accommodate those referenced bioRxiv preprints that have now been published in peer-reviewed journals.

The reference list has been updated.

Reviewer #3 (Recommendations For The Authors):

It would be good to discuss whether the pro-domains affecting virulence or avirulence activity.

Kex2, the protease that cleaves the pro-domain functions in the golgi. We therefore suspect that the pro-domain is removed prior to secretion. For recombinant protein production in E. coli we find that these pro-domains are necessary to obtain soluble protein (doi: 10.1111/nph.17516). As we require the pro-domain for protein production and can not completely removing them from our preps, we cannot perform experiments to test this and subsequently comment further. In a paper that identified SIX effectors in tomato utilising proteomics approach (https://bsppjournals.onlinelibrary.wiley.com/doi/10.1111/j.1364-3703.2007.00384.x), it appears that the pro-domains were not captured in this analysis. This supports the conclusion that they are not associated with the mature/secreted protein.

The authors stated that the C-terminal domain of SIX6 has a single disulfide bond unique to SIX6. Please clarify in which context is it unique: in Fusarium or across all FOLD proteins?

This is in direct comparison to Avr1 and Avr3. The disulfide in the C-domain of SIX6 is unique compared to Avr1 and Avr3. This has been made clear in text.

The structural similarity of FOLD proteins to other known structures have been discussed (lines 460ff), but it is not clear whether all structures and models identified in this work would yield cysteine inhibitor and tumor necrosis factors as best structural matches in the database or whether this is specific to a single FOLD protein. Please consider discussing recently published findings by others (Teulet et al. 2023, New Phytologist) on this aspect.

This analysis was performed for Avr1, we obtained relatively low similarity hits for Avr3/Six6. We have updated this text accordingly… “Unfortunately, the FOLD effectors share little overall structural similarity with known structures in the PDB outside of the similarity with each other. At a domain level, the N-domain of the FOLD effector Avr1 has some structural similarities with cystatin cysteine protease inhibitors (PDB code: 4N6V, PDB code: 5ZC1) [60, 61], and the C-domain with tumour necrosis factors (PDB code: 6X83) [62] and carbohydrate-binding lectins (PDB code: 2WQ4) [63]. Relatively weak hits were observed for Avr3/Six6.”

It might be useful to clearly point out that the ToxA fold and the C-terminus of the FOLD fold are different.

We have secondary structural topology maps of the FOLD and ToxA-like families in S8 Fig which highlight the differences in topology between these two families.

Please add information to Fig.S8 listing the approach to generate the secondary structure topology maps.

We have added this information in the figure caption.

Author Response

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

eLife assessment

This work presents H3-OPT, a deep learning method that effectively combines existing techniques for the prediction of antibody structure. This work is important because the method can aid the design of antibodies, which are key tools in many research and industrial applications. The experiments for validation are solid.

Comments to Author:

Several points remain partially unclear, such as:

1). Which examples constitute proper validation;

Thank you for your kind reminder. We have modified the text of the experiments for validation to identify which examples constitute proper validation. We have corrected the “Finally, H3-OPT also shows lower Cα-RMSDs compared to AF2 or tFold-Ab for the majority of targets in an expanded benchmark dataset, including all antibody structures from CAMEO 2022” into “Finally, H3-OPT also shows lower Cα-RMSDs compared to AF2 or tFold-Ab for the majority (six of seven) of targets in an expanded benchmark dataset, including all antibody structures from CAMEO 2022” and added the following sentence in the experimental validation section of our revised manuscript to clarify which examples constitute proper validation: “AlphaFold2 outperformed IgFold on these targets”.

2) What the relevance of the molecular dynamics calculations as performed is;

Thank you for your comment, and I apologize for any confusion. The goal of our molecular dynamics calculations is to compare the differences in binding affinities, an important issue of antibody engineering, between AlphaFold2-predicted complexes and H3-OPT-predicted complexes. Molecular dynamics simulations enable the investigation of the dynamic behaviors and interactions of these complexes over time. Unlike other tools for predicting binding free energy, MM/PBSA or MM/GBSA calculations provide dynamic properties of complexes by sampling conformational space, which helps in obtaining more accurate estimates of binding free energy. In summary, our molecular dynamics calculations demonstrated that the binding free energies of H3-OPT-predicted complexes are closer to those of native complexes. We have included the following sentence in our manuscript to provide an explanation of the molecular dynamics calculations: “Since affinity prediction plays a crucial role in antibody therapeutics engineering, we performed MD simulations to compare the differences in binding affinities between AF2-predicted complexes and H3-OPT-predicted complexes.”.

3) The statistics for some of the comparisons;

Thank you for the comment. We have incorporated statistics for some of the comparisons in the revised version of our manuscript and added the following sentence in the Methods section: “We conducted two-sided t-test analyses to assess the statistical significance of differences between the various groups. Statistical significance was considered when the p-values were less than 0.05. These statistical analyses were carried out using Python 3.10 with the Scipy library (version 1.10.1).”.

4) The lack of comparison with other existing methods.

We appreciate your valuable comments and suggestions. Conducting comparisons with a broader set of existing methods can further facilitate discussions on the strengths and weaknesses of each method, as well as the accuracy of our method. In our study, we conducted a comparison of H3-OPT with many existing methods, including AlphaFold2, HelixFold-Single, ESMFold, and IgFold. We demonstrated that several protein structure prediction methods, such as ESMFold and HelixFold-Single, do not match the accuracy of AlphaFold2 in CDR-H3 prediction. Additionally, we performed a detailed comparison between H3-OPT, AlphaFold2, and IgFold (the latest antibody structure prediction method) for each target.

We sincerely thank the comment and have introduced a comparison with OmegaFold. The results have been incorporated into the relevant sections (Fig 4a-b) of the revised manuscript.

Author response image 1.

Public Reviews

Comments to Author:

Reviewer #1 (Public Review):

Summary:

The authors developed a deep learning method called H3-OPT, which combines the strength of AF2 and PLM to reach better prediction accuracy of antibody CDR-H3 loops than AF2 and IgFold. These improvements will have an impact on antibody structure prediction and design.

Strengths:

The training data are carefully selected and clustered, the network design is simple and effective.

The improvements include smaller average Ca RMSD, backbone RMSD, side chain RMSD, more accurate surface residues and/or SASA, and more accurate H3 loop-antigen contacts.

The performance is validated from multiple angles.

Weaknesses:

1) There are very limited prediction-then-validation cases, basically just one case.

Thanks for pointing out this issue. The number of prediction-then-validation cases is helpful to show the generalization ability of our model. However, obtaining experimental structures is both costly and labor-intensive. Furthermore, experimental validation cases only capture a limited portion of the sequence space in comparison to the broader diversity of antibody sequences.

To address this challenge, we have collected different datasets to serve as benchmarks for evaluating the performance of H3-OPT, including our non-redundant test set and the CAMEO dataset. The introduction of these datasets allows for effective assessments of H3-OPT’s performance without biases and tackles the obstacle of limited prediction-then-validation cases.

Reviewer #2 (Public Review):

This work provides a new tool (H3-Opt) for the prediction of antibody and nanobody structures, based on the combination of AlphaFold2 and a pre-trained protein language model, with a focus on predicting the challenging CDR-H3 loops with enhanced accuracy than previously developed approaches. This task is of high value for the development of new therapeutic antibodies. The paper provides an external validation consisting of 131 sequences, with further analysis of the results by segregating the test sets into three subsets of varying difficulty and comparison with other available methods. Furthermore, the approach was validated by comparing three experimentally solved 3D structures of anti-VEGF nanobodies with the H3-Opt predictions

Strengths:

The experimental design to train and validate the new approach has been clearly described, including the dataset compilation and its representative sampling into training, validation and test sets, and structure preparation. The results of the in-silico validation are quite convincing and support the authors' conclusions.

The datasets used to train and validate the tool and the code are made available by the authors, which ensures transparency and reproducibility, and allows future benchmarking exercises with incoming new tools.

Compared to AlphaFold2, the authors' optimization seems to produce better results for the most challenging subsets of the test set.

Weaknesses:

1) The scope of the binding affinity prediction using molecular dynamics is not that clearly justified in the paper.

We sincerely appreciate your valuable comment. We have added the following sentence in our manuscript to justify the scope of the molecular dynamics calculations: “Since affinity prediction plays a crucial role in antibody therapeutics engineering, we performed MD simulations to compare the differences in binding affinities between AF2-predicted complexes and H3-OPT-predicted complexes.”.

2) Some parts of the manuscript should be clarified, particularly the ones that relate to the experimental validation of the predictions made by the reported method. It is not absolutely clear whether the experimental validation is truly a prospective validation. Since the methodological aspects of the experimental determination are not provided here, it seems that this may not be the case. This is a key aspect of the manuscript that should be described more clearly.

Thank you for the reminder about experimental validation of our predictions. The sequence identities of the wild-type nanobody VH domain and H3 loop, when compared with the best template, are 0.816 and 0.647, respectively. As a result, these mutants exhibited low sequence similarity to our dataset, indicating the absence of prediction bias for these targets. Thus, H3-OPT outperformed IgFold on these mutants, demonstrating our model's strong generalization ability. In summary, the experimental validation actually serves as a prospective validation.

Thanks for your comments, we have added the following sentence to provide the methodological aspects of the experimental determination: “The protein expression, purification and crystallization experiments were described previously. The proteins used in the crystallization experiments were unlabeled. Upon thawing the frozen protein on ice, we performed a centrifugation step to eliminate any potential crystal nucleus and precipitants. Subsequently, we mixed the protein at a 1:1 ratio with commercial crystal condition kits using the sitting-drop vapor diffusion method facilitated by the Protein Crystallization Screening System (TTP LabTech, mosquito). After several days of optimization, single crystals were successfully cultivated at 21°C and promptly flash-frozen in liquid nitrogen. The diffraction data from various crystals were collected at the Shanghai Synchrotron Research Facility and subsequently processed using the aquarium pipeline.”

3) Some Figures would benefit from a clearer presentation.

We sincerely thanks for your careful reading. According to your comments, we have made extensive modifications to make our presentation more convincing and clearer (Fig 2c-f).

Author response image 2.

Reviewer #3 (Public Review):

Summary:

The manuscript introduces a new computational framework for choosing 'the best method' according to the case for getting the best possible structural prediction for the CDR-H3 loop. The authors show their strategy improves on average the accuracy of the predictions on datasets of increasing difficulty in comparison to several state-of-the-art methods. They also show the benefits of improving the structural predictions of the CDR-H3 in the evaluation of different properties that may be relevant for drug discovery and therapeutic design.

Strengths:

The authors introduce a novel framework, which can be easily adapted and improved. The authors use a well-defined dataset to test their new method. A modest average accuracy gain is obtained in comparison to other state-of-the art methods for the same task while avoiding testing different prediction approaches.

Weaknesses:

1) The accuracy gain is mainly ascribed to easy cases, while the accuracy and precision for moderate to challenging cases are comparable to other PLM methods (see Fig. 4b and Extended Data Fig. 2). That raises the question: how likely is it to be in a moderate or challenging scenario? For example, it is not clear whether the comparison to the solved X-ray structures of anti-VEGF nanobodies represents an easy or challenging case for H3-OPT. The mutant nanobodies seem not to provide any further validation as the single mutations are very far away from the CDR-H3 loop and they do not disrupt the structure in any way. Indeed, RMSD values follow the same trend in H3-OPT and IgFold predictions (Fig. 4c). A more challenging test and interesting application could be solving the structure of a designed or mutated CDR-H3 loop.

Thank you for your rigorous consideration. When the experimental structure is unavailable, it is difficult to directly determinate whether the target is easy-to-predict or challenging. We have conducted our non-redundant test set in which the number of easy-to-predict targets is comparable to the other two groups. Due to the limited availability of experimental antibody structures, especially nanobody structures, accurately predicting CDR-H3 remains a challenge. In our manuscript, we discuss the strengths and weakness of AlphaFold2 and other PLM-based methods, and we introduce H3-OPT as a comprehensive solution for antibody CDR3 modeling.

We also appreciate your comment on experimental structures. We fully agree with your opinion and made attempts to solve the experimental structures of seven mutants, including two mutants (Y95F and Q118N) which are close to CDR-H3 loop. Unfortunately, we tried seven different reagent kits with a total of 672 crystallization conditions, but were unable to obtain crystals for these mutants. Despite the mutants we successfully solved may not have significantly disrupted the structures of CDR-H3 loops, they have still provided valuable insights into the differences between MSA-based methods and MSA-free methods (such as IgFold) for antibody structure modeling.

We have further conducted a benchmarking study using two examples, PDBID 5U15 and 5U0R, both consisting of 18 residues in CDR-H3, to evaluate H3-OPT's performance in predicting mutated H3 loops. In the first case (target 5U15), AlphaFold2 failed to provide an accurate prediction of the extended orientation of the H3 loop, resulting in a less accurate prediction (Cα-RMSD = 10.25 Å) compared to H3-OPT (Cα-RMSD = 5.56 Å). In the second case (target 5U0R, a mutant of 5U15 in CDR3 loop), AlphaFold2 and H3-OPT achieved Cα-RMSDs of 6.10 Å and 4.25 Å, respectively. Additionally, the Cα-RMSDs of OmegaFold predictions were 8.05 Å and 9.84 Å, respectively. These findings suggest that both AlphaFold2 and OmegaFold effectively captured the mutation effects on conformations but achieved lower accuracy in predicting long CDR3 loops when compared to H3-OPT.

2) The proposed method lacks a confidence score or a warning to help guide the users in moderate to challenging cases.

We appreciate your suggestions and we have trained a separate module to predict confidence scores. We used the MSE loss for confidence prediction, where the label error was calculated as the Cα deviation of each residue after alignment. The inputs of this module are the same as those used for H3-OPT, and it generates a confidence score ranging from 0 to 100.

3) The fact that AF2 outperforms H3-OPT in some particular cases (e.g. Fig. 2c and Extended Data Fig. 3) raises the question: is there still room for improvements? It is not clear how sensible is H3-OPT to the defined parameters. In the same line, bench-marking against other available prediction algorithms, such as OmegaFold, could shed light on the actual accuracy limit. We totally understand your concern. Many papers have suggested that PLM-based models are computationally efficient but may have unsatisfactory accuracy when high-resolution templates and MSA are available (Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies, Ruffolo, J. A. et al, 2023). However, the accuracy of AF2 decreased substantially when the MSA information is limited. Therefore, we directly retained high-confidence structures of AF2 and introduced a PSPM to improve the accuracy of the targets with long CDR-H3 loops and few sequence homologs. The improvement in mean Cα-RMSD demonstrated the room for accurately predicting CDR-H3 loops.

We also appreciate your kind comment on defined parameters. In fact, once a benchmark dataset is established, determining an optimal cutoff value through parameter searching can indeed further improve the performance of H3-OPT in CDR3 structure prediction. However, it is important to note that this optimal cutoff value heavily depends on the testing dataset being used. Therefore, we provide a recommended cutoff value and offer a program interface for users who wish to manually define the cutoff value based on their specific requirements. Here, we showed the average Cα-RMSDs of our test set under different confidence cutoffs and the results have been added in the text accordingly.

Author response table 1.

We also appreciate your reminder, and we have conducted a benchmark against OmegaFold. The results have been included in the manuscript (Fig 4a-b).

Author response image 3.

Reviewer #1 (Recommendations For The Authors):

1) In Fig 3a, please also compare IgFold and H3-OPT (merge Fig. S2 into Fig 3a)

In Fig 3b, please separate Sub2 and Sub3, and add IgFold's performance.

Thank you very much for your professional advice. We have made revisions to the figures based on your suggestions.

Author response image 4.

2) For the three experimentally solved structures of anti-VEGF nanobodies, what are the sequence identities of the VH domain and H3 loop, compared to the best available template? What is the length of the H3 loop? Which category (Sub1/2/3) do the targets belong to? What is the performance of AF2 or AF2-Multimer on the three targets?

We feel sorry for these confusions. The sequence identities of the VH domain and H3 loop are 0.816 and 0.647, respectively, comparing with the best template. The CDR-H3 lengths of these nanobodies are both 17. According to our classification strategy, these nanobodies belong to Sub1. The confidence scores of these AlphaFold2 predicted loops were all higher than 0.8, and these loops were accepted as the outputs of H3-OPT by CBM.

3) Is AF2-Multimer better than AF2, when using the sequences of antibody VH and antigen as input?

Thanks for your suggestions. Many papers have benchmarked AlphaFold2-Multimer for protein complex modeling and demonstrated the accuracy of AlphaFold2-Multimer on predicting the protein complex is far from satisfactory (Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants, Rui Yin, et al., 2022). Additionally, there is no significantly difference between AlphaFold2 and AlphaFold2-Multimer on antibody modeling (Structural Modeling of Nanobodies: A Benchmark of State-of-the-Art Artificial Intelligence Programs, Mario S. Valdés-Tresanco, et al., 2023)

From the data perspective, we employed a non-redundant dataset for training and validation. Since these structures are valuable, considering the antigen sequence would reduce the size of our dataset, potentially leading to underfitting.

4) For H3 loop grafting, I noticed that only identical target and template H3 sequences can trigger grafting (lines 348-349). How many such cases are in the test set?

We appreciate your comment from this perspective. There are thirty targets in our database with identical CDR-H3 templates.

Reviewer #2 (Recommendations For The Authors):

• It is not clear to me whether the three structures apparently used as experimental confirmation of the predictions have been determined previously in this study or not. This is a key aspect, as a retrospective validation does not have the same conceptual value as a prospective, a posteriori validation. Please note that different parts of the text suggest different things in this regard "The model was validated by experimentally solving three structures of anti-VEGF nanobodies predicted by H3-OPT" is not exactly the same as "we then sought to validate H3-OPT using three experimentally determined structures of anti-VEGF nanobodies, including a wild-type (WT) and two mutant (Mut1 and Mut2) structures, that were recently deposited in protein data bank". The authors are kindly advised to make this point clear. By the way, "protein data bank" should be in upper case letters.

We gratefully thank you for your feedback and fully understand your concerns. To validate the performance of H3-OPT, we initially solved the structures of both the wild-type and mutants of anti-VEGF nanobodies and submitted these structures to Protein Data Bank. We have corrected “that were recently deposited in protein data bank” into “that were recently deposited in Protein Data Bank” in our revised manuscript.

• It would be good to clarify the goal and importance of the binding affinity prediction, as it seems a bit disconnected from the rest of the paper. Also, it would be good to include the production MD runs as Sup, Mat.

Thanks for your valuable comment. We have added the following sentence in our manuscript to clarify the goal and importance of the molecular dynamics calculations: “Since affinity prediction plays a crucial role in antibody therapeutics engineering, we performed MD simulations to compare the differences in binding affinities between AF2-predicted complexes and H3-OPT-predicted complexes.”. The details of production runs have been described in Method section.

• Has any statistical test been performed to compare the mean Cα-RMSD values across the modeling approaches included in the benchmark exercise?

Thanks for this kind recommendation. We conducted a statistical test to assess the performance of different modeling approaches and demonstrated significant improvements with H3-OPT compared to other methods (p<0.001). Additionally, we have trained H3-OPT with five random seeds and compared mean Cα-RMSD values with all five models of AF2. Here, we showed the average Cα-RMSDs of H3-OPT and AlphaFold2.

Author response table 1.

• In Fig. 2c-f, I think it would be adequate to make the ordering criterion of the data points explicit in the caption or the graph itself.

We appreciate your comment and suggestion. We have revised the graph in the manuscript accordingly.

Author response image 5.

• Please revise Figure S2 caption and/or its content. It is not clear, in parts b and c, which is the performance of H3-OPT. Why weren´t some other antibody-specific tools such as IgFold included in this comparison?

Thanks for your comments. The performance of H3-OPT is not included in Figure S2. Prior to training H3-OPT, we conducted several preliminary studies, and the detailed results are available in the supplementary sections. We showed that AlphaFold2 outperformed other methods (including AI-based methods and TBM methods) and produced sub-angstrom predictions in framework regions. The comparison of IgFold with other methods was discussed in a previous work (Fast, accurate antibody structure prediction from deep learning on massive set of natural antibodies, Ruffolo, J. A. et al, 2023). In that study, we found that IgFold largely yielded results comparable to AlphaFold2 but with lower prediction cost. Additionally, we have also conducted a detailed comparison of CDR-H3 loops with IgFold in our main text.

• It is stated that "The relative binding affinities of the antigen-antibody complexes were evaluated using the Python script...". Which Python script?

Thank you for your comments, and I apologize for the confusion. This python script is a module of AMBER software, we have corrected “The relative binding affinities of the antigen-antibody complexes were evaluated using the python script” into “The relative binding affinities of the antigen-antibody complexes were evaluated using the MMPBSA module of AMBER software”.

Reviewer #3 (Recommendations For The Authors):

Does H3-OPT improve the AF2 score on the CDR-H3? It would be interesting to see whether grafted and PSPM loops improve the pLDDT score by using for example AF2Rank [https://doi.org/10.1103/PhysRevLett.129.238101]. That could also be a way to include a confidence score into H3-OPT.

We are so grateful for your kind question. H3-OPT could not provide a confidence score for output in current version, so we did not know whether H3-OPT improve the AF2 score or not.

We appreciate your kind recommendations and have calculated the pLDDT scores of all models predicted by H3-OPT and AF2 using AF2Rank. We showed that the average of pLDDT scores of different predicted models did not match the results of Cα-RMSD values.

Author response table 3.

Therefore, we have trained a separate module to predict the confidence score of the optimized CDR-H3 loops. We hope that this module can provide users with reliable guidance on whether to use predicted CDR-H3 loops.

The test case of Nb PDB id. 8CWU is an interesting example where AF2 outperforms H3-OPT and PLMs. The top AF2 model according to ColabFold (using default options and no template [https://doi.org/10.1038/s41592-022-01488-1]) shows a remarkably good model of the CDR-H3, explaining the low Ca-RMSD in the Extended Data Fig. 3. However, the pLDDT score of the 4 tip residues (out of 12), forming the hairpin of the CDR-H3 loop, pushes down the average value bellow the CBM cut-off of 80. I wonder if there is a lesson to learn from that test case. How sensible is H3-OPT to the CBM cut-off definition? Have the authors tried weighting the residue pLDDT score by some structural criteria before averaging? I guess AF2 may have less confidence in hydrophobic tip residues in exposed loops as the solvent context may not provide enough support for the pLDDT score.

Thanks for your valuable feedback. We showed the average Cα-RMSDs of our test set under different confidence cutoffs and the results have been added in the text accordingly.

Author response table 4.

We greatly appreciate your comment on this perspective. Inspired on your kind suggestions, we will explore the relationship between cutoff values and structural information in related work. Your feedback is highly valuable as it will contribute to the development of our approach.

A comparison against the new folding prediction method OmegaFold [https://doi.org/10.1101/2022.07.21.500999] is missed. OmegaFold seems to outperform AF2, ESM, and IgFold among others in predicting the CDR-H3 loop conformation (See [https://doi.org/10.3390/molecules28103991] and [https://doi.org/10.1101/2022.07.21.500999]). Indeed, prediction of anti-VEGF Nb structure (PDB WT_QF_0329, chain B in supplementary data) by OmegaFold as implemented in ColabFold [https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/beta/omegafold.ipynb] and setting 10 cycles, renders Ca-RMSD 1.472 Å for CDR-H3 (residues 98-115).

We appreciate your valuable suggestion. We have added the comparison against OmegaFold in our manuscript. The results have been included in the manuscript (Fig 4a-b).

Author response image 6.

In our test set, OmegaFold outperformed ESMFold in predicting the CDR-H3 loop conformation. However, it failed to match the accuracy of AF2, IgFold, and H3-OPT. We discussed the difference between MSA-based methods (such as AlphaFold2) and MSA-free methods (such as IgFold) in predicting CDR-H3 loops. Similarly, OmegaFold provided comparative results with HelixFold-Single and other MSA-free methods but still failed to match the accuracy of AlphaFold2 and H3-OPT on Sub1.

The time-consuming step in H3-OPT is the AF2 prediction. However, most of the time is spent in modeling the mAb and Nb scaffolds, which are already very well predicted by PLMs (See Fig. 4 in [https://doi.org/10.3390/molecules28103991]). Hence, why not use e.g. OmegaFold as the first step, whose score also correlates to the RMSD values [https://doi.org/10.3390/molecules28103991]? If that fails, then use AF2 or grafting. Alternatively, use a PLM model to generate a template, remove/mask the CDR loops (at least CDR-H3), and pass it as a template to AF2 to optimize the structure with or without MSA (e.g. using AF2Rank).

Thanks for your professional feedbacks. It is really true that the speed of MSA searching limited the application of high-throughput structure prediction. Previous studies have demonstrated that the deep learning methods performed well on framework residues. We once tried to directly predict the conformations of CDR-H3 loops using PLM-based methods, but this initial version of H3-OPT lacking the CBM could not replicate the accuracy of AF2 in Sub1. Similarly, we showed that IgFold and OmegaFold also provide lower accuracy in Sub1 (average Cα-RMSD is 1.71 Å and 1.83 Å, respectively, whereas AF2 predicted an average of 1.07 Å). Therefore, The predictions of AlphaFold2 not only produce scaffolds but also provide the highest quality of CDR-H3 loops when high-resolution templates and MSA are available.

Thank you once again for your kind recommendation. In the current version of H3-OPT, we have highlighted the strengths of H3-OPT in combining the AF2 and PLM models in various scenarios. AF2 can provide accurate predictions for short loops with fewer than 10 amino acids, and PLM-based models show little or no improvement in such cases. In the next version of H3-OPT, as the first step, we plan to replace the AF2 models with other methods if any accurate MSA-free method becomes available in the future.

Line 115: The statement "IgFold provided higher accuracy in Sub3" is not supported by Fig. 2a.

We are sorry for our carelessness. We have corrected “IgFold provided higher accuracy in Sub3” into “IgFold provided higher accuracy in Sub3 (Fig. 3a)”.

Lines 195-203: What is the statistical significance of results in Fig 5a and 5b?

Thank you for your kind comments. The surface residues of AF2 models are significantly higher than those of H3-OPT models (p < 0.005). In Fig. 5b, H3-OPT models predicted lower values than AF2 models in terms of various surface properties, including polarity (p <0.05) and hydrophilicity (p < 0.001).

Lines 212-213: It is not easy to compare and quantify the differences between electrostatic maps in Fig. 5d. Showing a Dmap (e.g. mapmodel - mapexperiment) would be a better option. Additionally, there is no methodological description of how the maps were generated nor the scale of the represented potential.

Thank you for pointing this out. We have modified the figure (Fig. 5d) according to your kind recommendation and added following sentences to clarify the methodological description on the surface electrostatic potential:

“Analysis of surface electrostatic potential

We generated two-dimensional projections of CDR-H3 loop’s surface electrostatic potential using SURFMAP v2.0.0 (based on GitHub from February 2023: commit: e0d51a10debc96775468912ccd8de01e239d1900) with default parameters. The 2D surface maps were calculated by subtracting the surface projection of H3-OPT or AF2 predicted H3 loops to their native structures.”

Author response image 7.

Lines 237-240 and Table 2: What is the meaning of comparing the average free energy of the whole set? Why free energies should be comparable among test cases? I think the correct way is to compare the mean pair-to-pair difference to the experimental structure. Similarly, reporting a precision in the order of 0.01 kcal/mol seems too precise for the used methodology, what is the statistical significance of the results? Were sampling issues accounted for by performing replicates or longer MDs?

Thanks for your rigorous advice and pointing out these issues. We have modified the comparisons of free energies of different predicted methods and corrected the precision of these results. The average binding free energies of H3-OPT complexes is lower than AF2 predicted complexes, but there is no significant difference between these energies (p >0.05).

Author response table 4.

Comparison of binding affinities obtained from MD simulations using AF2 and H3-OPT.

Thanks for your comments on this perspective. Longer MD simulations often achieve better convergence for the average behavior of the system, while replicates provide insights into the variability and robustness of the results. In our manuscript, each MD simulation had a length of 100 nanoseconds, with the initial 90 nanoseconds dedicated to achieving system equilibrium, which was verified by monitoring RMSD (Root Mean Square Deviation). The remaining 10 nanoseconds of each simulation were used for the calculation of free energy. This approach allowed us to balance the need for extensive sampling with the verification of system stability.

Regarding MD simulations for CDR-H3 refinement, its successful application highly depends on the starting conformation, the force field, and the sampling strategy [https://doi.org/10.1021/acs.jctc.1c00341]. In particular, the applied plan MD seems a very limited strategy (there is not much information about the simulated times in the supplementary material). Similarly, local structure optimizations with QM methods are not expected to improve a starting conformation that is far from the experimental conformation.

Thank you very much for your valuable feedback. We fully agree with your insights regarding the limitations of MD simulations. Before training H3-OPT, we showed the challenge of accurately predicting CDR-H3 structures. We then tried to optimize the CDR-H3 loops by computational tools, such as MD simulations and QM methods (detailed information of MD simulations is provided in the main text). Unfortunately, these methods were not expected to improve the accuracy of AF2 predicted CDR-H3 loops. These results showed that MD simulations and QM methods not only are time-consuming, but also failed to optimize the CDR-H3 loops. Therefore, we developed H3-OPT to tackle these issues and improve the accuracy of CDR3-H3 for the development of antibody therapeutics.

Text improvements

Relevant statistical and methodological parameters are presented in a dispersed manner throughout the text. For example, the number of structures in test, training, and validation datasets is first presented in the caption of Fig. 4. Similarly, the sequence identity % to define redundancy is defined in the caption of Fig. 1a instead of lines 87-88, where authors define "we constructed a non-redundant dataset with 1286 high-resolution (<2.5 Å)". Is the sequence redundancy for the CDR-H3 or the whole mAb/Nb?

Thank you for pointing out these issues. We have added the number of structures in each subgroup in the caption of Fig. 1a: “Clustering of the filtered, high-resolution structures yielded three datasets for training (n = 1021), validation (n = 134), and testing (n = 131).” and corrected “As data quality has large effects on prediction accuracy, we constructed a non-redundant dataset with 1286 high-resolution (<2.5 Å) antibody structures from SAbDab” into “As data quality has large effects on prediction accuracy, we constructed a non-redundant dataset (sequence identity < 0.8) with 1286 high-resolution (<2.5 Å) antibody structures from SAbDab” in the revised manuscript. The sequence redundancy applies to the whole mAb/Nb.

The description of ablation studies is not easy to follow. For example, what does removing TGM mean in practical terms (e.g. only AF2 is used, or PSPM is applied if AF2 score < 80)? Similarly, what does removing CBM mean in practical terms (e.g. all AF2 models are optimized by PSPM, and no grafting is done)? Thanks for your comments and suggestions. We have corrected “d, Differences in H3-OPT accuracy without the template module. e, Differences in H3-OPT accuracy without the CBM. f, Differences in H3-OPT accuracy without the TGM.” into “d, Differences in H3-OPT accuracy without the template module. This ablation study means only PSPM is used. e, Differences in H3-OPT accuracy without the CBM. This ablation study means input loop is optimized by TGM and PSPM. f, Differences in H3-OPT accuracy without the TGM. This ablation study means input loop is optimized by CBM and PSPM.”.

Authors should report the values in the text using the same statistical descriptor that is used in the figures to help the analysis by the reader. For example, in lines 223-224 a precision score of 0.75 for H3-OPT is reported in the text (I assume this is the average value), while the median of ~0.85 is shown in Fig. 6a.

Thank you for your careful checks. We have corrected “After identifying the contact residues of antigens by H3-OPT, we found that H3-OPT could substantially outperform AF2 (Fig. 6a), with a precision of 0.75 and accuracy of 0.94 compared to 0.66 precision and 0.92 accuracy of AF2.” into “After identifying the contact residues of antigens by H3-OPT, we found that H3-OPT could substantially outperform AF2 (Fig. 6a), with a median precision of 0.83 and accuracy of 0.97 compared to 0.64 precision and 0.95 accuracy of AF2.” in proper place of manuscript.

Minor corrections

Lines 91-94: What do length values mean? e.g. is 0-2 Å the RMSD from the experimental structure?

We appreciate your comment and apologize for any confusion. The RMSD value is actually from experimental structure. The RMSD value evaluates the deviation of predicted CDR-H3 loop from native structure and also represents the degree of prediction difficulty in AlphaFold2 predictions. We have added following sentence in the proper place of the revised manuscript: “(RMSD, a measure of the difference between the predicted structure and an experimental or reference structure)”.

Line 120: is the "AF2 confidence score" for the full-length or CDR-H3?

We gratefully appreciate for your valuable comment and have corrected “Interestingly, we observed that AF2 confidence score shared a strong negative correlation with Cα-RMSDs (Pearson correlation coefficient =-0.67 (Fig. 2b)” into “Interestingly, we observed that AF2 confidence score of CDR-H3 shared a strong negative correlation with Cα-RMSDs (Pearson correlation coefficient =-0.67 (Fig. 2b)” in the revised manuscript.

Line 166: Do authors mean "Taken" instead of "Token"?

We are really sorry for our careless mistakes. Thank you for your reminder.

Line 258: Reference to Fig. 1 seems wrong, do authors mean Fig. 4?

We sincerely thank the reviewer for careful reading. As suggested by the reviewer, we have corrected the “Fig. 1” into “Fig. 4”.

Author response image 7.

Point out which plot corresponds to AF2 and which one to H3-OPT

Thanks for pointing out this issue. We have added the legends of this figure in the proper positions in our manuscript.

Author Response

Reviewer #1 (Public Review):

“A sample size of 3 idiopathic seems underpowered relative to the many types of genetic changes that can occur in ASD. Since the authors carried out WGS, it would be useful to know what potential causative variants were found in these 3 individuals and even if not overlapping if they might expect to be in a similar biological pathway.

If the authors randomly selected 3 more idiopathic cell lines from individuals with autism, would these cell lines also have altered mTOR signaling? And could a line have the same cell biology defects without a change in mTOR signaling? The authors argue that the sample size could be the reason for lack of overlap of the proteomic changes (unlike the phosphor-proteomic overlaps), which makes the overlapping cell biology findings even more remarkable. Or is the phenotyping simply too crude to know if the phenotypes truly are the same?”

We appreciate these thoughtful comments and also agree that of several models, our studies indicate the possibility of mTOR alteration in multiple forms of ASD. As above, we are currently pursuing this hypothesis with newly acquired DOD support. With regard to the I-ASD population, we agree that there are a large variety of genetic changes that can occur in genetically undefined ASDs. Indeed, this is precisely why we expected to see “personalized” phenotypes in each I-ASD individual when we embarked on this study. At that time, several years ago, we had planned to expand the analyses to more I-ASD individuals to assess for additional personalized phenotypes. However, as our studies progressed, we were surprised to find convergence in our I-ASD population in terms of neurite outgrowth and migration and later proteomic results showing convergence in mTOR. We found it particularly remarkable that despite a sample size of 3 that this convergence was noted. When we had the opportunity to extend our studies to the 16p11.2 deletion population, we were thrilled to conduct the first comparison between I-ASD and a genetically defined ASD and, as such, the scope of the paper turned towards this comparison. We do agree that analyses of the other I-ASD individuals would be a beneficial endeavor, both to understand how pervasive NPC migration and neurite deficits are in autism and to assess the presence of mTOR dysregulation. Furthermore, it would be important to see whether alterations in other pathways could also lead to similar cell biological deficits, though we know that other studies of neurodevelopmental disorders have found such cellular dysregulations without reporting concurrent mTOR dysregulation. Given our current grant funding to extend these analyses, such experiments within this manuscript would not be feasible.

Regarding the phenotyping methods used, we decided to assess neurite outgrowth and migration as they are both cytoskeleton dependent processes that are critical for neurodevelopment and are often regulated by the same genes. Furthermore, similar analyses have been applied to Fragile-X Syndrome, 22q11.2 deletion syndrome, and schizophrenia NPCs (Shcheglovitov A. et al., 2013; Mor-Shaked H. et al., 2016; Urbach A. et al., 2010; Kelley D. J. et al., 2008; Doers M. E. et al., 2014; Brennand K. et al., 2015; Lee I. S. et al., 2015; Marchetto M. C. et al., 2011). As such, it seems that multiple underlying etiologies can lead to similar dysregulated cellular phenotypes that can contribute to a variety of neurodevelopmental disorders. On a more global level, there are only a few different cellular functions a developing neuron can undergo, and these include processes such as proliferation, survival, migration, and differentiation. Thus, to understand neurodevelopmental disorders, it is important to study the more “crude” or “global” cellular functions occurring during neurodevelopment to determine whether they are disrupted in disorders such as ASD. In our studies we find that there are indeed dysregulations in many of these basic developmental processes, indicating that the typical steps that occur for normal brain cytoarchitecture may be disrupted in ASD. To understand why, we then further utilized molecular studies to “zoom” in on potential mechanisms which implicated common dysregulation in mTOR signaling as one driver for these common cellular phenotypes. As suggested, we did complete WGS on all the I-ASD individuals and did not see any overlapping genetic variants between the three I-ASD individuals as mentioned in our manuscript. The genetic data was published in a larger manuscript incorporating the data (Zhou A. et al., 2023). However, there were variants that were unique to each I-ASD individual which were not seen in their unaffected family members, and it is possible these variants could be contributing to the I-ASD phenotypes. We also utilized IPA to conduct pathway analysis on the WGS data utilizing the same approach we did in analysis of p- proteome and proteome data. From WGS data, we selected high read-quality variants that were found only in I-ASD individuals and had a functional impact on protein (ie excluding synonymous variants). The enriched pathways obtained from this data were strikingly different from the pathways we found in the p-proteome analysis and are now included in supplemental Figure 6 in the manuscript. Briefly, the top 5 enriched pathways were: O-linked glycosylation, MHC class 1 signaling, Interleukin signaling, Antigen presentation, and regulation of transcription.

Reviewer #2 (Public Review):

1) I found that interpreting how differential EF sensitivity is connected to the rest of the story difficult at times. First, it is unclear why these extracellular factors were picked. These are seemingly different in nature (a neuropeptide, a growth factor and a neuromodulator) targeting largely different pathways. This limits the interpretation of the ASD subtype-specific rescue results. One way of reframing that could help is that these are pro-migratory factors instead of EFs broadly defined that fail to promote migration in I-ASD lines due to a shared malfunctioning of the intracellular migration machinery or cell-cell interactions (possibly through tight junction signaling, Fig S2A). Yet, this doesn't explain the migration/neurite phenotypes in 16p11 lines where EF sensitivity is not altered, overall implying that divergent EF sensitivity independent of underlying mTOR state. What is the proposed model that connects all three findings (divergent EF sensitivity based on ASD subtypes, 2 mTOR classes, convergent cellular phenotypes)?

We thank you for the kind assessment of our manuscript and for the thought-provoking questions posed. In terms of extracellular factors, for our study, we defined extracellular factor as any growth factor, amino acid, neurotransmitter, or neuropeptide found in the extracellular environment of the developing cells. The EFs utilized were selected due to their well-established role in regulation of early neurodevelopmental phenotypes, their expression during the “critical window” of mid-fetal development (as determined by Allan Brain Atlas), and in the case of 5-HT, its association with ASD (Abdulamir H. A. et al., 2018; Adamsen D. et al., 2014; Bonnin A. et al., 2011; Bonnin A. et al., 2007; Chen X. et al., 2015; El Marroun H. et al., 2014; Hammock E. et al., 2012; Yang C. J. et al., 2014; Dicicco-Bloom E. et al., 1998; Lu N. et al., 1998; Suh J. et al., 2001; Watanabe J. et al., 2016; Gilmore J. H. et al., 2003; Maisonpierre P. C. et al., 1990; Dincel N. et al., 2013; Levi- Montalcini R., 1987). Lastly, prior experiments in our lab with a mouse model of neurodevelopmental disorders, had shown atypical responses to EFs (IGF-1, FGF, PACAP). As such, when we first chose to use EFs in human NPCs we wanted to know 1) whether human NPCs even responded to these EFs, 2) whether EFs regulated neurite outgrowth and migration and 3) would there be a differential response in NPCs derived from those with ASD. Our studies were initiated on the I-ASD cohort and given the heterogeneity of ASD we had hypothesized we would get “personalized” neurite and migration phenotypes. Due to this reason, we also wanted to select multiple types of EFs that worked on different signaling pathways. Ultimately, instead of personalized phenotypes we found that all the I-ASD NPCs did not respond to any of the EFs tested whereas the 16p11.2 deletion NPCS did – this was therefore the only difference we found between these two “forms” of ASD. As noted, in I-ASD the lack of response to EFs can be ameliorated by modulating mTOR. However, in the 16p11.2 deletion, despite similar mTOR dysregulation as seen in I-ASD, there is no EF impairment. We do not have a cohesive model to explain why the 16pDel individuals differ from the I-ASD model other than to point to the p- proteomes which do show that the 16pDel NPCs are distinct from the I-ASD NPCs. It seems that mTOR alteration can contribute to impaired EF responsiveness in some NPCs but perhaps there is an additional defect that needs to be present in order for this defect to manifest, or that 16p11.2 deletion NPCs have specific compensatory features. For example, as noted in the thoughtful comment, the p-proteome canonical pathway analysis shows tight junction malfunction in I-ASD which is not present in the 16pDel NPCs and it could be the combination of mTOR dysregulation + dysregulated tight junction signaling that has led to lack of response to EFs in I-ASD. Regardless, we do not think the differences between two genetically distinct ASDs diminish the convergent mTOR results we have uncovered. That is, regardless of whatever defects are present in the ASD NPCs, we are able to rescue it with mTOR modulation which has fascinating implications for treatment and conceptualization for ASD. Lastly, we see our EF studies as an important inclusion as it shows that in some subtypes of ASD, lack of response to appropriate EFs could be contributing to neurodevelopmental abnormalities. Moreover, lack of response to these EFs could have implications for treatment of individuals with ASD (for example, SSRI are commonly used to treat co-morbid conditions in ASD but if an individual is unresponsive to 5- HT, perhaps this treatment is less effective). We have edited the manuscript to include an additional discussion section to address the EFs more thoroughly and have included a few extra sentences in the introduction as well!

2) A similar bidirectional migration phenotype has been described in hiSPC-derived human cortical interneurons generated from individuals with Timothy Syndrome (Birey et al 2022, Cell Stem Cell). Here, authors show that the intracellular calcium influx that is excessive in Timothy Syndrome or pharmacologically dampened in controls results in similar migration phenotypes. Authors can consider referring to this report in support of the idea that bimodal perturbations of cardinal signaling pathways can converge upon common cellular migration deficits.

We thank you for pointing out the similar migration phenotype in the Timothy Syndrome paper and have now cited it in our manuscript. We have also expanded on the concept of “too much or too little” of a particular signaling mechanism leading to common outcomes.

3) Given that authors have access to 8 I-ASD hiPSC lines, it'd very informative to assay the mTOR state (e.g. pS6 westerns) in NPCs derived from all 8 lines instead of the 3 presented, even without assessing any additional cellular phenotypes, which authors have shown to be robust and consistent. This can help the readers better get a sense of the proportion of high mTOR vs low- mTOR classes in a larger cohort.

We have already addressed this in response to reviewer 1 and the essential revisions section, providing our reasoning for not expanding the study to all 8 I-ASD individuals.

4) Does the mTOR modulation rescue EF-specific responses to migration as well (Figure 7)

We did not conduct sufficient replicates of the rescue EF specific responses to migration due to the time consuming and resource intensive nature of the neurosphere experiments. Unlike the neurite experiments, the neurosphere experiments require significantly more cells, more time, selection of neurospheres based on a size criterion, and then manual trace measurements. We did one experiment in Family-1 where we utilized MK-2206 to abolish the response of Sib NPCs to PACAP. Likewise, adding SC-79 to I-ASD-1 neurospheres allowed for response to PACAP.

Author response image 1.

Author response image 2.

Reviewer #3: Public Review

We appreciate the kind, detailed and very thorough review you provided for us!

The results on the mTOR signaling pathway as a point of convergence in these particular ASD subtypes is interesting, but the discussion should address that this has been demonstrated for other autism syndromes, and in the present manuscript, there should be some recognition that other signaling pathways are also implicated as common factors between the ASD subtypes.

With regards to the mTOR pathway, we had included the other ASD syndromes in which mTOR dysregulation has been seen including tuberous sclerosis, Cowden Syndrome, NF-1, as well as Fragile-X, Angelman, Rett and Phelan McDermid in the final paragraph of the discussion section “mTOR Signaling as a Point of Convergence in ASD”. We have now expanded our discussion to include that other signaling pathways such as MAPK, cyclins, WNT, and reelin which have also been implicated as common factors between the ASD subtypes.

The conclusions of this paper are mostly well supported by data, but for the cell migration assay, it is not clear if the authors control for initial differences in the inner cell mass area of the neurospheres in control vs ASD samples, which would affect the measurement of migration.

Thank you for this thoughtful comment! When we first started our migration data, inner cell mass size was indeed a major concern for which we controlled in our methods. First, when plating the neurospheres, we would only collect spheres when a majority of spheres were approximately a diameter of 100 um. Very large spheres often could not be imaged due to being out of focus and very small spheres would often disperse when plated. Thus, there were some constraints to the variability of inner cell mass size.

Furthermore, when we initially collected data, we conducted a proof of principal test to see if initial inner cell mass area (henceforth referred to as initial sphere size or ISS) influenced migration data. To do so, we obtained migration and ISS data from each diagnosis (Sib, NIH, I-ASD, 16pASD). Then we utilized R studio to see if there is a relationship between Migration and ISS in each diagnosis category using the equation (lm(Migration~ISS, data=bydiagnosis). In this equation, lm indicates linear modeling and (~) is a term used to ascertain the relationship between Migration and ISS and the term data=bydiagnosis allows the data to be organized by diagnosis

The results were expressed as R-squared values indicating the correlation between ISS and Migration for each diagnosis and the p-value showing statistical significance for each comparison. As shown in Author response table 1, for each data set, there is minimal correlation between Migration and ISS in each data set. Moreover, there are no statistically significant relationships between Migration and ISS indicating that initial sphere size DOES NOT influence migration data in any of our data-sets.

Author response table 1.

Lastly, utilizing R, we modeled what predicted migration would be like for Sib, NIH, I-ASD, and 16pASD if we accounted for ISS in each group. Raw migration data was then plotted against the predicted data as in Author response image 3.

Author response image 3.

As shown in the graph, there are no statistical differences between the raw migration data (the data that we actually measured in the dish) and the modeled data in which ISS is accounted for as a variable. As such, we chose not to normalize to or account for ISS in our other experiments. We have now included the above R studio analyses in our supplemental figures (Figure S1) as well.

Also, in Fig 5 and 6, panels I and J omit the effects of drug on mTOR phosphorylation as shown for other conditions.

Both SC-79 and MK2206 were selected in our experiments after thorough analysis of their effects on human epithelial cells and other cultured cells (citations in manuscript). However, initially, we did not know whether either of these drugs would modulate the mTOR pathway in human NPCs, thus, in Figures 5A,5D, 6A and 6D we chose to focus on two of our data-sets to establish the effect of these drugs in human NPCs. Our experiments in Family-1 and Family-2 showed us that SC-79 increases PS6 in human NPCs while MK-2206 downregulates it. Once this was established, we knew the drugs would have similar effects in the NPCs from the other families. Thus, we only conducted a proof of principle test to confirm the drug does indeed have the intended effect in I-ASD-3 and 16pDel. We have included these proof of principle westerns in Figure 5I, 5K, 6I and 6K to show that the effects of these drugs are reproducible across all our NPC lines. We did not include quantification since the data is only from our single proof of principle western.

Author response

eLife assessment

Using a genetically controlled experimental setting, the authors find that the lack of Polycomb-dependent epigenetic programming in the oocyte and early embryo influences the developmental trajectory through gestation in the mouse. By showing a two-phase outcome of early growth restriction followed by enhancement, the authors address previous inconsistencies in the field. However, the link with placenta function and gene misregulation is not yet fully supported.

We thank the Reviewers for their constructive comments. In response we have added significantly more data to the study and substantially rewritten the manuscript. New data include analyses of glucose, amino acid and metabolite levels in fetal and maternal blood samples, more highly resolved fetal growth analyses, a more detailed study of the hyperplastic placenta including IF analyses of labyrinth area, labyrinth to placenta and capillary to labyrinth ratios. We have also added analyses of placental DNA methylation state in offspring from oocytes lacking EED, which reveals a range of DNA methylation changes at imprinted and non-imprinted genes in HET-hom offspring compared to HET-het or WT-wt controls.

Reviewer #1 (Public Review):

Oberin, Petautschnig et. al investigated the developmental phenotypes that resulted from oocyte-specific loss of the EED (Embryonic Ectoderm Development) gene - a core component of the Polycomb repressive complex 2 (PRC2), which possess histone methyltransferase activity and catalyses trimethylation of histone H3 at lysine 27 (H3K27). The PRC2 complex plays essential roles in regulating chromatin structure, being an important regulator of cellular differentiation and development during embryogenesis. As novel findings, the authors find that PRC2-dependent programming in the oocyte, via loss of the core component EE2, causes placental hyperplasia and propose that the increase of placental transplacental flux of nutrients leads to fetal and postnatal overgrowth. At the mechanistic level, they show altered expression of genes previously implicated in placental hyperplasia phenotypes. They also establish interesting parallelism with the placental hyperplasia phenotype that is frequently observed in cloned mice.

Strengths:

The mouse breeding experiments are very well designed and are powerful to exclude potential confounding genetic effects on the developmental phenotypes that resulted from the loss of EED in oocytes. Another major strength is the developmental profiling across gestation, from pre-implantation to late gestation.

Weaknesses:

The evidence for 'oocyte' programming is restricted to phenotypic and gene expression analysis, without measurements of epigenetic dysregulation. It would be an added value if the authors could show evidence for altered H3K27me3 or DNA methylation in the placenta, for example.

In an earlier previous study we identified a large number of developmentally important genes that accumulated H3K27me3 in primary-secondary stage growing oocytes and were repressed by EED (Jarred et al., 2022 Clinical Epigenetics). However, H3K27me3 was removed from all from these genes during preimplantation development, indicating that maternal inheritance of H3K27me3 at a wide range of genes is unlikely (Jarred et al., 2022 Clinical Epigenetics). Consistent with this only a small number of genes, including Slc38a4 and C2MC, have been shown to be functionally important in H3K27me3-dependent imprinting (Matoba et al., 2022 Genes and Development). Moreover, a related study showed that deletion of Setd2 and consequent loss of H3K36me3 in oocytes led to spreading of H3K27me3 into regions that were otherwise marked by H3K36me3 and DNA methylation (Xu et al. 2019 Nature Genetics 51:844–56). Based on these studies, we proposed that loss of EED and H3K27me3 may result in the ectopic spreading of H3K36me3 and DNA methylation in oocytes and that altered DNA methylation may then be transmitted to offspring and affect developmental outcomes (Jarred et al., 2022 Clinical Epigenetics)

Given this hypothesis we analysed DNA methylation rather than H3K27me3 in the placenta of WT-wt, HET- het and HET-hom offspring. This revealed differentially methylated regions (DMRs) in HET-hom placentas at two H3K27me3 imprinted genes Sfmbt2 (C2MC) and Mbnl2, five classically imprinted genes and at 74 DMRs not associated with imprinted loci. Together, our data supports the hypothesis from Jarred et al., 2022 Clinical Epigenetics that loss of EED in oocytes results in altered DNA methylation patterning at both imprinted and non-imprinted genes in offspring and that this is likely to affect offspring growth and development. However, whether these changes result from direct alteration of DNA methylation in oocytes remains unclear.

These new data are now included in results (Lines 387-409), Figure 6I, Supplementary File H-J and Discussion Lines 569-581.

Reviewer Comment 1. The claim that placental hyperplasia drives offspring catch-up growth is not supported by current experimental data. The authors do not address if transplacental flux is increased in the hyperplastic placentae, measure amino acids and glucose in fetal/maternal plasma, or perform tetraploid rescue experiments to ascertain the contribution of the placenta to growth phenotypes. Furthermore, it is unclear, from the current data, if the surface area for nutrient transport is actually increased in the hyperplastic placenta and the extent to which other cell populations (i.e. spongiotrophoblasts) are affected in addition to glycogen cells. In addition, one of the supporting conclusions that the placenta is a key contributor to fetal overgrowth is based on a very crude measurement - placenta efficiency - which the authors claim is increased in the homozygous mutants compared to controls. After analysing the data carefully, I find evidence for decreased placental efficiency instead. I believe that the authors mistakenly present the data as placenta to fetal weight ratios, which led to the misinterpretation of the 'efficiency' concept.

We thank the reviewer for pointing out our error in the placental efficiency data and we have now corrected the placental efficiency graphs (fetal/placental weight ratios) and updated the text throughout the manuscript as required (Figure 3I-K). As requested and described below, we have also added significantly more data, which support the conclusion that placental function is not enhanced in HET-hom mice and is unlikely to support fetal growth recovery.

The new data and analyses we have added include:

1. Further analyses of glycogen-enriched and non-glycogen-enriched cell counts in the decidua and junctional zones (Figure 4F-J)

2. Total glycogen cell counts for male and female placentas (Figure 4 – figure supplement 1F)

3. New analyses of fetal blood glucose levels at E17.5 and E18.5 and matching data from the mothers of each litter (Figure 4M)

4. New analyses of the circulating amino acid levels and metabolites in fetal blood of E17.5 offspring and matching data from the mothers of each litter (Figure 8)

5. New IF analyses of CD31 (PECAM-1) and combined this with machine learning assisted quantitative analyses of labyrinth and capillary areas using HALO (Figure 5)

6. Separated male and female offspring and placental weights at E14.5 and E17.5 and total areas of the placenta, decidua, junctional zone and labyrinth (Figure 3 – figure supplement 1) which provide more insight into potential sex-specific differences in HET-hom offspring and placenta

We have significantly re-written the results and discussion to reflect our new data and interpretation.

While we did not assess transplacental flux, our new data revealed: 1. HET-hom fetuses had lower blood glucose levels at E18.5; 2. Circulating levels of amino acids and a wide range of metabolites did not differ between HET-hom and control offspring, or between the mothers of these offspring; 3. HET-hom placentas had lower total labyrinth area, labyrinth/placenta and capillary/labyrinth ratios based on analysis of total capillary and labyrinth areas, indicating that the surface area for nutrient transfer is not increased

Together these data strongly indicate that hyperplastic HET-hom placentas do not provide greater support to HET-hom fetuses than controls, and that increased placental function in HET-hom offspring is unlikely to explain the late gestation fetal growth recovery we observed in HET-hom offspring or how HET-hom offspring were able to attain normal weights by birth.

While we have not directly counted the spongiotrophoblast populations, we have now included analyses of both the glycogen-enriched and non-glycogen cell populations in the junctional zone and the decidua (Figure 4H-K). This revealed an increased area of both glycogen-enriched and non-glycogen cells in the junctional zone and in the decidua of HET-hom placentas, consistent with the greater junctional zone/placenta ratio observed in HET-hom placentas (Figure 4D). Together with data in Figure 4C-F and Supp. Fig. 3, our observations demonstrate that the overall decidua and junctional zone areas were increased in HET-hom offspring, but there was a disproportionate expansion of the junctional zone that was caused by increased areas of both glycogen and non-glycogen-enriched cells.

Tetraploid rescue experiments would require a very significant amount of time and investment and are technically very demanding. While creation of complementary tetraploid offspring would be informative, unfortunately these experiments are beyond the scope of this current study.

Reviewer Comment 1 cont. The authors do not mention alternative explanations for the observed fetal catch-up and postnatal overgrowth. Why would oocyte epigenetic programming effects be restricted to the placenta, and not include fetal organs?

Our intention was certainly not to convey a message that effects may be placenta specific. Indeed, our ongoing work beyond the scope of this study provides evidence for effects in other tissues (brain and bones) that will be published elsewhere. Our new data clearly show low placental efficiency, fetal blood glucose, low capillary/labyrinth ratio and no impact on circulating fetal amino acid or metabolite levels in HET-hom offspring. In light of these new data, we have reinterpreted the findings of this study and substantially updated the discussion.

Given our observations that fetal growth rate markedly increased during late gestation, but placental efficiency was reduced, our data strongly indicate that the effects of altered epigenetic oocyte programming due to loss of Eed affect both the placenta and the fetus. While our findings are significant, the precise mechanism underlying this growth response in HET-hom fetuses remains unknown. Understanding this mechanism will require substantially more work that will be the subject of future studies.

Reviewer #2 (Public Review):

Consistent fetal growth trajectories are vital for survival and later life health. The authors utilise an elegant and novel animal model to tease apart the role of Eed protein in the female germline from the role of somatic Eed. The authors were able to experimentally attribute placental overgrowth - particularly of the endocrine region of the placenta - to the function of Eed protein in the oocyte. Loss of Eed protein in the oocyte was also associated with dynamic changes in fetal growth and prolonged gestation. It was not determined whether the reported catch-up growth apparent on the day of birth was due to enhanced fetal growth very late in gestation, a longer gestational time ie the P0 pups are effectively one day "older" compared to the controls, or the pups catching up after birth when consuming maternal milk.

To understand if increased growth occurred in HET-hom fetuses prior to birth, we have now included analyses of offspring weight at E18.5 (Figure 2F), all pups collected with a verified E19.5 birth date (Figure 2J) and for pups from similar litter sizes (5-7 pups) at E19.5 (Figure 2K). Together with our existing data, these additional analyses provide average weights for fetuses at E14.5, E17.5, E18.5 and pups born on E19.5. This confirmed that HET-hom offspring undergo enhanced growth in the last few days of pregnancy, resulting in the progression of substantially growth and developmentally restricted HET-hom fetuses at E14.5, to pups with normal weight at birth within the 40% of pregnancies that were born on E19.5 in a normal gestational time.

However, in addition, gestational length was increased by one to two days in 60% of pregnancies from hom oocytes, but not in control pregnancies from het or wt oocytes. As average weights were significantly greater in all surviving HET-hom offspring at P0 (i.e. surviving pups born on E19.5-E21.5; Figure 2G), it appears that this additional gestational time contributed to the offspring overgrowth. This is logical, however it does not explain how growth and developmentally delayed fetuses at E14.5 attained normal weight and developmental stage by E19.5 (Figure 2J-K).

Together our data clearly show that HET-hom offspring undergo enhanced growth during the late stages of pregnancy, allowing them to resolve the developmental delay and growth insufficiency observed at E14.5 so that they were born at normal weight and stage at E19.5. In addition, increased gestational time contributes to weight of pups delivered on E20.5 or 21.5, partly explaining the overgrowth phenotype observed in this model.

The idea that increased milk consumption may explain the overgrowth of HET-hom offspring is interesting. It is possible that the increased growth rate of HET-hom offspring continues after birth and contributes to overgrowth. However, examining this outcome in a tightly controlled manner is complicated given that we cannot predict the day of birth of HET-hom litters, and that these litters are generally small and would need to be fostered on the day of birth alongside control litters. Given these challenges and that our primary observation is that HET-hom offspring underwent fetal growth recovery during pregnancies of normal length and via extension of gestational length, we have not examined the possibility of increased milk consumption after birth.

We have updated the results to reflect the new analyses and have provided relevant discussion to address these data. Our description of these data can be found in Results (lines 165-197) and in Figure 2.

Reviewer #3 (Public Review):

My understanding of the main claims of the paper, and how they are justified by the data are discussed below:

Overall, loss of PRC2 function in the developing oocyte and early embryo causes:

1) Growth restriction from at least the blastocyst stage with low cell counts and midgestational developmental delay.

Strengths:

• Live embryo imaging added an important dimension to this study. The authors were able to confirm an unquantified finding from a previous lab (reduced time to 2-cell stage in oocyte-deletion Eed offspring, Inoue 2018, PMID: 30463900) as well as identify developmental delay and mortality at the blastocyst- hatching transition.

• For the weight and morphological analysis the authors are careful to provide isogenic controls for most of the experiments presented. This means that any phenotypes can be attributed to the oocyte genotype rather than any confounding effects of maternal or paternal genotype.

• Overall, there is good evidence that oocyte deletion of Eed results in early embryonic growth restriction, consistent with previous observations (Inoue 2018, PMID: 30463900).

Reviewer 3, Comment 1: Weaknesses: Gaps in the reporting of specific features of the methodology make it difficult to interpret/understand some of the results.

While we are unsure exactly which methods Reviewer 3 would like expanded, we have updated parts that we thought required further detail and allow more informed interpretation of the results. These include methods for placental histology (Lines 650-669) and immuno- histochemistry (Lines 671-690), and new methods for CD31 immunofluorescence (Lines 692-714), glucose and metabolomics (Lines 752-769) and DNA methylation (RRBS; Lines 734-750) analyses.

To clarify the approach taken for histology, immunohistochemical and immunofluorescent staining, sections were cut in compound series from the centre of each placenta, ensuring that we collected representative data for each sample. QuPath was used to quantify the decidual and junctional zone areas in one complete, fully intact midline section for each placenta as close to the midline as possible. This provided data from 10 placentas for each genotype. In addition, glycogen-enriched and non-glycogen-enriched cells were identified and quantified using machine learning assisted QuPath analyses of the whole placenta, decidua and junctional zone regions. We have also added quantitative analyses of the labyrinth and labyrinth capillary network using immunofluorescent CD31 staining and machine learning assisted HALO software. This new analysis of placental morphology is included in the methods section.

Moreover, as there were no sex-specific differences in placental morphology or weight, we combined the samples from both sexes to provide greater numbers for analysis in each genotype. For example, as described for the analyses of labyrinth and capillaries using CD31 IF, 4 placentas of each sex were used for data collection. This provided data from a total of 8 placentas (4 male and 4 female) for each genotype from a total of 17 WT-wt (9 male and 8 female), 21 HET-het (9 male and 12 female) and 24 HET-hom (16 male and 8 female) sections (2-3 sections/placenta).

Reviewer 3, Comment 2: Placental hyperplasia with disproportionate overgrowth of the junctional trophoblast especially the glycogen trophoblast (GlyT) cells.

Strengths: • The authors provide a comprehensive description of how placental and embryo weight is affected by the oocyte-Eed deletion through mid-to-late gestation development. The case for placentomegaly is clear.

Weaknesses:

• The placental efficiency data presented in Figure 3G-I is incorrect. Placental efficiency is calculated as embryo mass/placental mass, and it increases over the late gestation period. For e14.5 for example (Fig3G), WT-wt embryo mass = ~0.3g, placenta mass = 0.11g (from Fig 3D) = placental efficiency 2.7; HET-hom = 0.25/0.12 = 2.1. The paper gives values: WT-wt 0.5, HET-hom 0.7. Have the authors perhaps divided placenta weight by embryo mass? This would explain why the E17.5 efficiencies are so low (WT-wt 0.11 rather than a more usual figure of 8.88. If this is the case then the authors' conclusion that placental efficiency is improved by oocyte deletion of Eed is wrong - in fact, placental efficiency is severely compromised.

The authors have performed cell type counting on histological sections obtained from placentas to discover which cells are contributing to the placentomegaly. This data is presented as %cell type area in the main figure, though the untransformed cross-sectional area for each cell type is shown in the supplementary data. This presentation of the data, as well as the description of it, is misleading because, while it emphasises the proportional increase in the endocrine compartment of the placenta it downplays the fact that the exchange area of the mutant placentas is vastly expanded. This is important for two reasons.

Firstly, the whole placenta is increased in size suggesting that the mechanism is not placental lineage- specific and instead acting on the whole organ. Secondly in relation to embryonic growth, generally speaking, genetic manipulations that modify labyrinthine volume tend to have a positive correlation with fetal mass whereas the relationship between junctional zone volume and embryonic mass is more complex (discussed in Watson PMID: 15888575, for example). The authors should reconsider how they present this data in light of the previous point.

We thank the reviewer for pointing out our error in the placental efficiency analysis and apologise for this error. We have corrected the presentation and interpretation of these data and have described this in detail in our response to Reviewer 1, Comment 1.

As discussed in our response to Reviewer 1, Comment 1, we have added a range of analyses to determine whether placental efficiency was enhanced in HET-hom offspring. These include measuring fetal and maternal circulating glucose levels (Figure 4K), individual amino acids and an extensive range of metabolites (Figure 8) and providing CD31 immunofluorescent analyses of labyrinth area, labyrinth/placental ratio and capillary/labyrinth ratio in HET-hom and control placentas (Figure 5).

We also added analyses of glycogen enriched and non-glycogen-enriched cell counts in the decidua and junctional zones. As suggested by Reviewer 3, both glycogen-enriched and non-enriched cell populations are significantly increased in HET-hom placentas.

Combined, these new analyses significantly expand the study and support the conclusion that placental efficiency in HET-hom offspring was either compromised or not different from controls, depending on the analysis. We find no evidence that placental efficiency was increased in HET-hom offspring and have reworked our results and discussion sections to reflect these new data and interpretation.

Reviewer 3, Comment 2 cont: Again, some of the methods are not clearly reported making interpretation difficult - especially how they have estimated their GlyT number.

As outlined in our response to Reviewer 3 Comment 1, in the methods section we have added further detail of how we counted glycogen-enriched and non-enriched cells in the decidua and junctional zone regions of sections for the middle of WT-wt, WT-het, HET-het and HET-hom placentas (Lines 650-669).

Reviewer 3, Comment 3: Perinatal embryonic/pup overgrowth.

Strengths:

• The overgrowth exhibited by the oocyte-Eed-deleted pups is striking and confirms the previous work by this group (Prokopuk, 2018). This is an important finding, especially in the context of understanding how PRC2-group gene mutations in humans cause overgrowth syndromes. It is also intriguing because it indicates that genetic/environmental insults in the mother that affect her gamete development can have long-term consequences on offspring physiology.

Weaknesses:

• Is the overgrowth intrauterine or is it caused by the increase in gestation length? The way the data is reported makes it impossible to work this out. The authors show that gestation time is consistently lengthened for mothers incubating oocyte-Eed-deleted pups by 1-2 days. In the supplementary material, the mutant embryos are not larger than WT at e19.5, the usual day of birth. Postnatal data is presented as day post-parturition. It would probably be clearer to present the embryonic and postnatal data as days post coitum. In this way, it will be obvious in which period the growth enhancement is taking place. This is information really important to determine whether the increased growth of the mutants is due to a direct effect of the intrauterine environment, or perhaps a more persistent hormonal change in the mother that can continue to promote growth beyond the gestation period.

We have used embryonic day (E) to denote embryo and fetal age throughout the study – this is the same as using DPC (i.e. E19.5 is equivalent to 19.5 DPC). As described in the Methods “Collection of post-implantation embryos, placenta and postnatal offspring”, mice were time mated for two-four nights, with females plug checked daily. Positive plugs were noted as day E0.5.

To make the data presentation clearer, we have shown the data for surviving HET-hom pups born on E19.5 (Figure 2J) separately from all HET-hom surviving pups born on E19.5-E21.5. (Figure 2G). As discussed in our response to Reviewer 2, we have also included growth data for pregnancies at E14.5, E17.5, E18.5 (Fig. 2C-F) and E19.5 (Figure 2J,K), as well as P0 (combined data for surviving pups born E19.5-E21.5), and P3 (combined data for surviving pups born E19.5-E21.5, Figure 2G,H).

These data clearly show that HET-hom fetuses are substantially growth and developmentally delayed at E14.5 (Figure 2D), but HET-hom pups born on E19.5 are the same weight as WT-wt, WT-het and HET-het control pups (Figure 2J). This demonstrates that weight of HET-hom fetuses is normalised in utero between E14.5 and day of birth on E19.5.

Importantly, as requested by Reviewer 3, we have separated average weight for all surviving pups with a day of birth of E19.5-21.5 (Figure 2G) from average weight of pups born on E19.5 only (Figure 2J). These analyses revealed that the average weight of surviving pups born between E19.5-21.5 was significantly higher than for controls (Figure 2G), but the average weight of pups born on E19.5 only was not. It is therefore clear that extended gestation also contributed to increased HET-hom pup birth weight. We have updated these additional analyses in Results (Lines 165-197) and Figure 2

As revealed in Figure 2H, it is also possible/likely that growth of HET-hom pups during the three days post- partum may have contributed to the offspring overgrowth we observed in this and our previous study (Prokopuk et al., 2018 Clinical Epigenetics). However, we cannot determine whether there is a contribution from a persistent maternal hormonal change that promotes post-natal offspring growth or whether there is an innate growth benefit in HET-hom pups. As this is very difficult to dissect, separating these possibilities is beyond the scope of our study.

Reviewer 3, Comment 4: "fetal growth restriction followed by placental hyperplasia, .. drives catch-up growth that ultimately results in perinatal offspring overgrowth".

Here the authors try to link their observations, suggesting that i) the increased perinatal growth rate is a consequence of placentomegaly, and ii) the placentomegaly/increased fetal growth is an adaptive consequence of the early growth restriction. This is an interesting idea and suggests that there is a degree of developmental plasticity that is operating to repair the early consequences of transient loss of Eed function.

Strengths:

• Discrepancies between earlier studies are reconciled. Here the authors show that in oocyte-Eed-deleted embryos growth is initially restricted and then the growth rate increases in late gestation with increased perinatal mass.

Weaknesses:

• Regarding the dependence of fetal growth increase on placental size increase, this link is far from clear since placental efficiency is in fact decreased in the mutants (see above).

• "Catch-up growth" suggests that a higher growth rate is driven by an earlier growth restriction in order to restore homeostasis. There is no direct evidence for such a mechanism here. The loss of Eed expression in the oocyte and early embryo could have an independent impact on more than one phase of development.

Firstly, there is growth restriction in the early phase of cell divisions. Potentially this could be due to depression of genes that restrain cell division on autosomes, or suppression of X-linked gene expression (as has been previously reported, Inoue, 2018 PMID: 30463900). The placentomegaly is explained by the misregulation of non-canonically imprinted genes, as the authors report (and in agreement with other studies, e.g. Inoue, 2020. PMID: 32358519).

• Explaining the perinatal phase of growth enhancement is more difficult. I think it is unlikely to be due to placentomegaly. Multiple studies have shown that placentomegaly following somatic cell nuclear transfer (SCNT) is caused by non-canonically imprinted genes, and can be rescued by reducing their expression dosage. However, SCNT causes placentomegaly with normal or reduced embryonic mass (for example -Xie 2022, PMID: 35196486), not growth enhancement. Moreover, since (to my knowledge) single loss of imprinting models of non-canonically imprinted genes do not exist, it is not possible to understand if their increased expression dosage can drive perinatal overgrowth, and if this is preceded by growth restriction and thus constitutes 'catch up growth'.

Reviewer 3 is correct in their assessment that placental efficiency was decreased in HET- hom offspring and we have corrected the placental efficiency analysis based on fetal/placental weight ratios (discussed in detail in our response to Reviewer 1 Comment 1). We have added substantially more data (glucose, amino acids, metabolites, labyrinth capillary area and density). These data support the conclusion that a placentally driven advantage for HET-hom fetal growth is unlikely, despite our observation that HET- hom fetuses are developmental delayed and underweight at E14.5, but are born at normal weight after a normal gestational length (19.5 days) (discussed in our responses to Reviewer 3, Comment 3 and Reviewer 2).

This demonstrates that HET-hom fetuses are able to attain normal birth weight despite being initially growth restricted state at E14.5, and that this occurs despite low placental function. Moreover, as we compared isogenic offspring with heterozygous loss of Eed (Het-het compared to HET-hom offspring) the outcomes we observed in HET-hom offspring originate from loss of EED in the growing oocyte or loss of maternal EED in the zygote strongly suggesting that a non-genetic mechanism is involved.

As pointed out by Reviewer 3, the initial developmental delay in HET-hom offspring may be due to increased expression of genes that regulate cell proliferation – this could clearly explain the lower number of cells we observed in the ICM and the growth delay at later stages of embryonic and fetal development. Another possibility is that maternal PRC2 provided by the oocyte promotes cell divisions in preimplantation embryos We have discussed these possibilities on Lines 467-476.

In addition, Matoba et al 2022 demonstrated that deletion of maternal Xist together with Eed was able to rescue male-biased lethality in offspring from oocytes lacking Eed, revealing a clear role for X-linked genes in this phenotype (Matoba et al 2022, Genes and Development). However, deletion of maternal Xist did not properly normalise survival offspring from Eed null oocytes (i.e. Eed/Xist double maternal null litters were smaller than litters derived from wild type oocytes) strongly suggesting other mechanisms provide the capacity for HET-hom offspring to attain normal weight at birth. We have added further discussion of the Matoba study in the context of our study on of the Discussion (Lines 544-555)

Finally, with respect to the outcomes for SCNT derived offspring, we extracted SCNT fetal growth and placental weight data from the supplementary data included in Matoba et al., 2018 Cell Stem Cell. 2018;23(3):343-54.e5 and compared it with data collected in our study (Figure 7). This analysis revealed that the weights of placentas and fetuses of offspring derived via SCNT were very similar to the HET-hom offpsring in our study and we have discussed the similarities and potential differences between HET-hom and SCNT offspring in the Discussion (Lines 478-500).

As pointed out by Reviewer 3, deletion of maternal non-canonically imprinted genes partially or fully rescued the placental hyperplasia phenotype in both SCNT derived and offspring from oocyte lacking EED. However, as we have discussed, the mechanisms underlying other aspects of the offspring phenotype, such as fetal growth recovery of HET-hom offspring observed in our study, remain unknown. Moreover, the comparison we provide in Figure 7 strongly indicates that HET-hom and SCNT fetuses are similarly delayed at E14.5 and undergo similar fetal growth recovery before birth, but the mechanism also remains unknown. Together, it appears that offspring derived from either Eed-null oocytes or by SCNT have an innate ability to remediate fetal growth restriction during the late stages of pregnancy without a requirement to correct maternally inherited impacts mediated by Xist or H3K27me3-dependent imprinting.

Author Response

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

We thank the editors and reviewers for their helpful comments, which have allowed us to improve the manuscript.

Response to reviewer 2

We thank the reviewer for this positive feedback, which requires no further revision.

Response to reviewer 3

We thank the reviewer for highlighting these additional points and provide further explanations on these below.

Firstly, we started the analysis from a baseline of year 2000 because the largest international donor (the Global Fund) uses baseline malaria levels in the period 2000-2004 as the basis of their current allocation calculations (The Global Fund, Description of the 2020-2022 Allocation Methodology, December 2019). In the paper we compare our optimal strategy to a simplified version of this method, represented by our “proportional allocation” strategy.

Even if our simulations started in the year 2015, a direct comparison with the Global Technical Strategy for Malaria 2016-2030 would not be possible due to the different approaches taken. The GTS was developed to progress towards malaria elimination globally and set ambitious targets of at least 90% reduction in malaria case incidence and mortality rates and malaria elimination in at least 35 countries by 2030 compared to 2015. Mathematical modelling at the time suggested that 90% coverage of WHO-recommended interventions (vector control, treatment and seasonal malaria chemoprevention) would be needed to approach this target (Griffin et al. 2016, Lancet Infectious Diseases). The global annual investment requirements to meet GTS targets were estimated at US$6.4 billion by 2020 and US$8.7 billion by 2030 (Patouillard et al. 2017, BMJ Global Health). This strategy therefore considers what resources would be required to achieve a specific global target, but not the optimized allocation of resources.

Investments into malaria control have consistently been below the estimated requirements for the GTS milestones (World Health Organization 2022, World Malaria Report 2022). In our study, we therefore take a different perspective on how limited budgets can be optimally allocated to a single intervention (insecticide-treated nets) across countries/settings to achieve the best possible outcome for two objectives that are different to the GTS milestones (either minimizing the global case burden, or minimizing both the global case burden and the number of settings not having yet reached a pre-elimination phase). As stated in the discussion, our estimate of allocating 76% of very low budgets to high-transmission settings was similar to the global investment targets estimated for the GTS, where the 20 countries with the highest burden in 2015 were estimated to require 88% of total investments (Patouillard et al. 2017, BMJ Global Health). Nevertheless, we also show that if higher budgets were available, allocating the majority to low-transmission settings co-endemic for P. falciparum and P. vivax would achieve the largest reduction in global case burden. We acknowledge the modelling of a single intervention as one of the key limitations of this analysis, but this simplification was necessary in order to perform the complex optimisation problem. Computationally it would not have been feasible to optimize across a multitude of intervention and coverage combinations.

A further limitation raised by the reviewer is the lack of cross-species immunity between P. falciparum and P. vivax in our model. While cross-reactivity between antibodies against these two species has been observed in previous studies and the potential implications of this would be important to explore in future work, we did not include it here as little is known to date about the epidemiological interactions between different malaria parasite species (Muh et al. 2020, PLoS Neglected Tropical Diseases).

Lastly, we did not assume that transmission was homogenous within the four transmission settings in our study (very low, low, moderate, high); transmission dynamics were simulated separately in each country, accounting for heterogeneous mosquito bite exposure. However, results were summarised for the broader transmission settings since many other country-specific factors were not accounted for (see discussion) and the findings should not be used to inform individual country allocation decisions.

---

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

Author response to peer review

We thank the reviewers for their insightful comments, which raise several important points regarding our study. As the reviewers have recognised, we introduced a number of simplifications in order to perform this complex optimisation problem, such as by restricting the analysis to a single intervention (insecticide-treated nets) and modelling countries at a national level. Despite their clear relevance to the study, computationally it would not have been feasible to run the multitude of scenarios suggested by reviewer 1, which we recognise as a limitation. As such we agree with the assessment that this study primarily represents a thought experiment, based on substantive modelling and aggregate scenario-based analysis, to assess whether current policies are aligned with an optimal allocation strategy or whether there might be a need to consider alternative strategies. The findings are relevant primarily to global funders and should not be used to inform individual country allocation decisions, and also point to avenues for further research. This perspective also underlies our decision to start the analysis from a baseline of year 2000 as opposed to modelling the current 2023 malaria situation: the largest international donor (the Global Fund) uses baseline malaria levels in the period 2000-2004 as the basis of their allocation calculations (The Global Fund, Description of the 2020-2022 Allocation Methodology, December 2019) (1). A simplified version of this method is represented by our “proportional allocation” strategy. We have made several revisions to the manuscript to address the points raised by the reviewers, as detailed below.

Reviewer #1 (Public Review):

1. The authors present a back-of-the-envelope exploration of various possible resource allocation strategies for ITNs. They identify two optimal strategies based on two slightly different objective functions and compare 3 simple strategies to the outcomes of the optimal strategies and to each other. The authors consider both P falciparum and P vivax and explore this question at the country level, using 2000 prevalence estimates to stratify countries into 4 burden categories. This is a relevant question from a global funder perspective, though somewhat less relevant for individual countries since countries are not making decisions at the global scale.

Thank you for this summary of the paper. We agree that our analysis is of relevance to global funders, but is not meant to inform individual country allocation decisions. In the discussion, we now state:

p. 12 L19: “Therefore, policy decisions should additionally be based on analysis of country-specific contexts, and our findings are not informative for individual country allocation decisions.”

1. The authors have made various simplifications to enable the identification of optimal strategies, so much so that I question what exactly was learned. It is not surprising that strategies that prioritize high-burden settings would avert more cases.

Thank you for raising this point. Indeed, several simplifying assumptions were necessary to ensure the computational feasibility of this complex optimization problem. As a result, our study primarily represents a thought experiment to assess whether current policies are aligned with an optimal allocation strategy or whether there might be a need to consider alternative strategies. As now further outlined in the introduction, approaches to this have differed over time and it remains a relevant debate for malaria policy.

p. 2 L22: “However, there remains a lack of consensus on how best to achieve this longer-term aspiration. Historically, large progress was made in eliminating malaria mainly in lower-transmission countries in temperate regions during the Global Malaria Eradication Program in the 1950s, with the global population at risk of malaria reducing from around 70% of the world population in 1950 to 50% in 2000 (2). Renewed commitment to malaria control in the early 2000s with the Roll Back Malaria initiative subsequently extended the focus to the highly endemic areas in sub-Saharan Africa (3).”

We believe our findings not only confirm an “expected” outcome – that prioritizing high-burden settings would avert more cases – but also clearly illustrate various consequences of different allocation strategies that are implemented or considered in reality, which may not be so obvious. For example, we found that initially allocating a larger share of the budget to high-transmission countries could be both almost optimal in terms of reducing clinical cases and maximising the number of countries reaching pre-elimination. We also observed a trade-off between reducing burden and reducing the global population at risk (“shrinking the map”) through a focus on near-elimination settings, and estimate the loss in burden reduction when following an elimination target.

1. Generally, I found much of the text confusing and some concepts were barely explained, such that the logic was difficult to follow.

Thank you for bringing this to our attention, and we regret to hear the manuscript was confusing to read. We believe that the revisions made as a result of the reviewer comments have now made the manuscript much easier to follow. We additionally passed the manuscript to a colleague to identify confusing passages, and have added a number of sentences to clarify key concepts and improve the structure.

1. I am not sure why the authors chose to stratify countries by 2000 PfPR estimates and in essence explore a counterfactual set of resource allocation strategies rather than begin with the present and compare strategies moving forward. I would think that beginning in 2020 and modeling forward would be far more relevant, as we can't change the past. Furthermore, there was no comparison with allocations and funding decisions that were actually made between 2000 and 2020ish so the decision to begin at 2000 is rather confusing.

Thank you for pointing this out. We have now made the rationale for this choice clearer in the manuscript. Our main reason for this was to allow comparison with the Global Fund funding allocation, which is largely based on malaria disease burden in 2000-2004. As stated in the paper, malaria prevalence estimates in the year 2000 are commonly considered to represent a “baseline” endemicity level, before large-scale implementation of interventions in the following decades. In the manuscript, the transmission-related element of the Global Fund allocation algorithm is represented in our “proportional allocation” strategy. Previously this was only mentioned in the methods, but we have now added the following in the results to address this comment of the reviewer:

p. 6 L12: “Strategies prioritizing high- or low-transmission settings involved sequential allocation of funding to groups of countries based on their transmission intensity (from highest to lowest EIR or vice versa). The proportional allocation strategy mimics the current allocation algorithm employed by the Global Fund: budget shares are mainly distributed according to malaria disease burden in the 2000-2004 period. To allow comparison with this existing funding model, we also started allocation decisions from the year 2000.”

The Global Fund framework additionally considers economic capacity and other specific factors, and we have now also included a direct comparison with the 2020-2022 Global Fund allocation in Supplementary Figure S12 (see Author response image 1).

We agree that looking at allocation decisions from 2020 onward would also constitute a very interesting question. However, the high dimensionality in scenarios to consider for this would currently make it computationally infeasible to run on the global level. Not only would it have to include all interventions currently implemented and available for malaria at different levels of coverage, but also the option of scaling down existing interventions. Instead, our priority in this paper was to conduct a thought experiment including both P. falciparum and P. vivax on a large geographical scale.

Author response image 1.

Impact of the proportional allocation strategy and the 2020-2022 Global Fund allocation on global malaria cases (panel A) and the total population at risk of malaria (panel B) at varying budgets. Both strategies use the same algorithm for budget share allocation based on malaria disease burden in 2000-2004, but the Global Fund allocation additionally involves an economic capacity component and specific strategic priorities.

1. I realize this is a back-of-the-envelope assessment (although it is presented to be less approximate than it is, and the title does not reveal that the only intervention strategy considered is ITNs) but the number and scope of modeling assumptions made are simply enormous. First, that modeling is done at the national scale, when transmission within countries is incredibly heterogeneous. The authors note a differential impact of ITNs at various transmission levels and I wonder how the assumption of an intermediate average PfPR vs modeling higher and lower PfPR areas separately might impact the effect of the ITNs.

Thank you for this comment. We agree the title could be more specific and have changed this to “Resource allocation strategies for insecticide-treated bednets to achieve malaria eradication”.

Regarding the scale of ITN allocation, it is true that allocation at a sub-national scale could affect the results. However, considering this at a national scale is most relevant for our analysis because this is the scale at which global funding allocation decisions are made in practice. A sentence explaining this has been added in the methods.

p. 15 L8: “The analysis was conducted on the national level, since this scale also applies to funding decisions made by international donors (1).”

Further considering different geographical scales would also require introducing other assumptions, for example about how different countries would distribute funding sub-nationally, whether specific countries would take cooperative or competitive approaches to tackle malaria within a region or in border areas, and about delays in the allocation of bednets in specific regions. These interesting questions were outside of the scope of this work, but certainly require further investigation.

1. Second, the effect of ITNs will differ across countries due to variations in vector and human behavior and variation in insecticide resistance and susceptibility to the ITNs. The authors note this as a limitation but it is a little mind-boggling that they chose not to account for either factor since estimates are available for the historical period over which they are modeling.

Thank you for pointing this out. We did consider this and mentioned it as a limitation. Nevertheless, the complexity of accounting for this should also be recognised; for example, there is substantial uncertainty about the precise relationship between insecticide resistance and the population-level effect of ITNs (Sherrard-Smith et al., 2022, Lancet Planetary Health) (4). Additionally, our simulations extend beyond the 2000-2023 period so further assumptions about future changes to these factors would also be required. Simplifying assumptions are inherent to all mathematical modelling studies and we consider these particular simplifications acceptable given the high-level nature of the analysis.

1. Third, the assumption that elimination is permanent and nothing is needed to prevent resurgence is, as the authors know, a vast oversimplification. Since resources will be needed to prevent resurgence, it appears this assumption may have a substantial impact on the authors' results.

Thank you for this comment. In the discussion, we have now expanded on this:

p. 13 L3: “While our analysis presents allocation strategies to progress towards eradication, the results do not provide insight into allocation of funding to maintain elimination. In practice, the threat of malaria resurgence has important implications for when to scale back interventions.”

We believe that from a global perspective, the questions of funding allocation to achieve elimination vs to maintain it can currently still be considered separately given the large time-scales involved. The cost of preventing resurgence is not known, and one major problem in accounting for this would also be to identify relevant timescales to quantify this over.

1. The decision to group all settings with EIR > 7 together as "high transmission" may perhaps be driven by WHO definitions but at a practical level this groups together countries with EIR 10 and EIR 500. Why not further subdivide this group, which makes sense from a technical perspective when thinking about optimal allocation strategies?

Thank you for pointing this out. The WHO categories used are better interpreted in terms of the corresponding prevalence, which places countries with a prevalence of over 35% in the high transmission categories (WHO Guidelines for malaria, 31 March 2022) (5). We felt this is appropriate given that we are looking at theoretical global allocation patterns and do not aim to make recommendations for specific groups of countries or individual countries within sub-Saharan Africa that would be distinguished through the use of higher cut-offs. In our analysis, all 25 countries in the high transmission category were located in sub-Saharan Africa.

1. The relevance of this analysis for elimination is a little questionable since no one eliminates with ITNs alone, to the best of my understanding.

Thank you for this comment. We indeed state in the paper that ITNs alone are not sufficient to eliminate malaria. However, we still think that our analysis is relevant for elimination by taking a more theoretical perspective on reducing transmission using interventions. Starting from the 2000 baseline (or current levels) globally, large-scale transmission reductions such as those achieved by mass ITN distribution still represent the first key step on the path to malaria eradication, as shown in previous modelling work (Griffin et al., 2016, Lancet Infectious Diseases) (6). In the final phase of elimination, the WHO also recommends the addition of more targeted and reactive interventions (WHO Guidelines for malaria, 31 March 2022) (5). Our changes to the title of the article (“Resource allocation strategies for insecticide-treated bednets to achieve malaria eradication”) should now better reflect that we consider ITNs as just one necessary component to achieve malaria eradication.

Reviewer #2 (Public Review):

1. Schmit et al. analyze and compare different strategies for the allocation of funding for insecticide-treated nets (ITNs) to reduce the global burden of malaria. They use previously published models of Plasmodium falciparum and Plasmodium vivax malaria transmission to quantify the effect of ITN distribution on clinical malaria numbers and the population at risk. The impact of different resource allocation strategies on the reduction of malaria cases or a combination of malaria cases and achieving pre-elimination is considered to determine the optimal strategy to allocate global resources to achieve malaria eradication.

Strengths:

Schmit et al. use previously published models and optimization for rigorous analysis and comparison of the global impact of different funding allocation strategies for ITN distribution. This provides evidence of the effect of three different approaches: the prioritization of high-transmission settings to reduce the disease burden, the prioritization of low-transmission settings to "shrink the malaria map", and a resource allocation proportional to the disease burden.

Thank you for providing this summary and outline of the strengths of the paper.

1. Weaknesses:

The analysis and optimization which provide the evidence for the conclusions and are thus the central part of this manuscript necessitate some simplifying assumptions which may have important practical implications for the allocation of resources to reduce the malaria burden. For example, seasonality, mosquito species-specific properties, stochasticity in low transmission settings, and changing population sizes were not included. Other challenges to the reduction or elimination of malaria such as resistance of parasites and mosquitoes or the spread of different mosquito species as well as other beneficial interventions such as indoor residual spraying, seasonal malaria chemoprevention, vaccinations, combinations of different interventions, or setting-specific interventions were also not included. Schmit et al. clearly state these limitations throughout their manuscript.

The focus of this work is on ITN distribution strategies, other interventions are not considered. It also provides a global perspective and analysis of the specific local setting (as also noted by Schmit et al.) and different interventions as well as combinations of interventions should also be taken into account for any decisions.

Thank you for raising these points. As outlined at the beginning of our response, for computational reasons we indeed had to introduce several simplifying assumptions to perform this complex optimisation problem. As a result of these factors you highlighted, our study should primarily be interpreted as a thought experiment to assess whether current policies are aligned with an optimal allocation strategy or whether there might be a need to consider alternative strategies. The findings are relevant primarily to global funders and should not be used to inform individual country allocation decisions, which we have further clarified in the manuscript.

1. Nonetheless, the rigorous analysis supports the authors' conclusions and provides evidence that supports the prioritization of funding of ITNs for settings with high Plasmodium falciparum transmission. Overall, this work may contribute to making evidence-based decisions regarding the optimal prioritization of funding and resources to achieve a reduction in the malaria burden.

Thank you for this positive assessment of our work.

Reviewer #1 (Recommendations For The Authors):

1. L144: last paragraph, the focus on endemic equilibrium: I did not really understand this, when 39 years is mentioned later is that a different analysis? How are cases averted calculated in a time-agnostic endemic equilibrium analysis? Perhaps a little more detail here would be helpful.

A further explanation of this has been added in the results and methods.

p. 8 L 22: “To evaluate the robustness of the results, we conducted a sensitivity analysis on our assumption on ITN distribution efficiency. Results remained similar when assuming a linear relationship between ITN usage and distribution costs (Figure S10). While the main analysis involves a single allocation decision to minimise long-term case burden (leading to a constant ITN usage over time in each setting irrespective of subsequent changes in burden), we additionally explored an optimal strategy with dynamic re-allocation of funding every 3 years to minimise cases in the short term.”

p. 17 L25: “To ensure computational feasibility, 39 years was used as it was the shortest time frame over which the effect of re-distribution of funding from countries having achieved elimination could be observed.”

p. 18 L 9: “Global malaria case burden and the population at risk were compared between baseline levels in 2000 and after reaching an endemic equilibrium under each scenario for a given budget.”

1. L148: what is proportional allocation by disease burden and how is that different from prioritizing high-transmission settings?

Further details have been added in the text.

p. 6 L12: “Strategies prioritizing high- or low-transmission settings involved sequential allocation of funding to groups of countries based on their transmission intensity (from highest to lowest EIR or vice versa). The proportional allocation strategy mimics the current allocation algorithm employed by the Global Fund: budget shares are mainly distributed according to malaria disease burden in the 2000-2004 period. To allow comparison with this existing funding model, we also started allocation decisions from the year 2000.”

1. L198-9: did low transmission settings get the majority of funding at intermediate and maximum budgets because they have the most population (I think so, based on Fig 1)?

Yes, this is correct. We state in the results: “the optimized distribution of funding to minimize clinical burden depended on the available global budget and was driven by the setting-specific transmission intensity and the population at risk”.

1. L206: what is ITN distribution efficiency? This is not explained. What is the 39-year period? Why this duration?

Further explanations have been added in the results section, which were previously only detailed in the methods:

p. 8 L 22: “To evaluate the robustness of the results, we conducted a sensitivity analysis on our assumption on ITN distribution efficiency. Results remained similar when assuming a linear relationship between ITN usage and distribution costs (Figure S10)."

p. 17 L25: “To ensure computational feasibility, 39 years was used as it was the shortest time frame over which the effect of re-distribution of funding from countries having achieved elimination could be observed.”

1. L218: what is "no intervention with a high budget"? is this a phrasing confusion?

Yes, this has been changed.

p. 9 L14: “We estimated that optimizing ITN allocation to minimize global clinical incidence could, at a high budget, avert 83% of clinical cases compared to no intervention.”

1. L235-7: on comparing these results to previous work on the 20 highest-burden countries: is the definition of "high" similar enough across these studies that this is a relevant comparison?

We believe this is reasonably comparable, as looking at the 20 highest-burden countries encompasses almost the entire high-transmission group in our work (25 countries in total), on which the comparison is made.

1. L267-70: I didn't understand this sentence at all.

Thanks for flagging this. The sentence referred to is: “Allocation proportional to disease burden did not achieve as great an impact as other strategies because the funding share assigned to settings was constant irrespective of the invested budget and its impact, and we did not reassign excess funding in high-transmission settings to other malaria interventions.”

The previously mentioned added details on the proportional allocation strategy in the manuscript should now make this clearer, together with this clarification:

p. 11 L17: “In modelling this strategy, we did not reassign excess funding in high-transmission settings to other malaria interventions, as would likely occur in practice.”

For proportional allocation, a fixed proportion of the budget is calculated for each country based on disease burden, as described in the Global Fund allocation documentation (see Methods). However, since ITNs are the only intervention considered, this leads to a higher budget being allocated than is needed in some countries (i.e. where more funding doesn’t translate into further health gains).

1. L339 EIR range: 80 is high at the country level but areas within countries probably went as high as 500 back in 2000. How does this affect the modeled estimates of ITN impact?

The question of sub-national differences in transmission has been addressed in the public review comments. Briefly, we consider the national scale to be most relevant for our analysis because this is the scale at which global funding allocation decisions are made in practice. Although, as you correctly point out, the EIR affects ITN impact, it is not possible to conclude what the average effect of this would be on the country level without considering the following factors and introducing further assumptions on these: how would different countries distribute funding sub-nationally? Which countries would take cooperative or competitive approaches to tackle malaria within a region or in border areas? Would there be delays in the allocation of bednets in specific regions? These interesting questions were outside of the scope of this work, but certainly require further investigation.

1. L347 population size constant: births and deaths are still present, is that right? Unclear from this sentence

Yes, this is correct. Full details on the model can be found in the Supplementary Materials.

1. L370 estimating ITN distribution required to achieve simulated population usage: is this a single relationship for all of Africa? Is it based on ITNs distributed 2:1 -> % access -> % usage? So it accounts for allocation inefficiency?

Yes, this is represented by a single relationship for all of Africa to account for allocation inefficiency and is based on observed patterns across the continent and methodology developed in a previous publication (Bertozzi-Villa et al., 2021, Nature Communications) (7). Full details can be found in the Supplementary Materials (“Relationship between distribution and usage of insecticide-treated nets (ITNs)”, p. 21).

1. L375: the ITN unit cost is assumed constant across countries and time (I think, it doesn't say explicitly), is this a good assumption?

Yes, this is correct. We consider this a reasonable assumption within the scope of the paper. While delivery costs likely vary across countries, international funders usually have pooled procurement mechanisms for ITNs (The Global Fund, 2023, Pooled Procurement Mechanism Reference Pricing: Insecticide-Treated Nets).

1. L399: "single allocation of a constant ITN usage" it is not explained what exactly this means

Further explanations have been added in the manuscript.

p. 8 L24: “While the main analysis involves a single allocation decision to minimise long-term case burden (leading to a constant ITN usage over time in each setting irrespective of subsequent changes in burden), we additionally explored an optimal strategy with dynamic re-allocation of funding every 3 years to minimise cases in the short term.”

Reviewer #2 (Recommendations For The Authors):

1. Additionally to the public comments, the only major comment is that in this reviewer's opinion, the focus on ITNs as the only intervention should be made clearer at different places in the manuscript (e.g. in the discussion lines 303-304). Otherwise, there are only some minor comments (see below).

We have now modified the following sentence and also included this suggestion in the title (“Resource allocation strategies for insecticide-treated bednets to achieve malaria eradication”).

p. 13 L8: “Our analysis demonstrates the most impactful allocation of a global funding portfolio for ITNs to reduce global malaria cases.”

1. Minor comments:
2. It may be of interest to compare the maximum budget obtained from the optimization with other estimates of required funding and actual available funding.

Thank you for this interesting suggestion. Our maximum budget estimates are similar to the required investments projected for the WHO Global Technical Strategy: US$3.7 billion for ITNs in our analysis compared to between US$6.8 and US$10.3 billion total annual resources between 2020 and 2030, of which an estimated 55% would be required for (all) vector control (US$3.7 - US$5.7 billion) (Patouillard et al., 2016, BMJ Global Health) (8). However, it is well known that current spending is far below these requirements: total investments in malaria were estimated to be about US$3.1 billion per year in the last 5 years (World Health Organization, 2022, World Malaria Report 2022) (9).

1. Line 177: should "Figure S7" be bold?

Yes, this has been corrected.

1. Line 218: what does "no intervention with high budget" mean? Should this simply be "no intervention"?

This has been changed.

p. 9 L14: “We estimated that optimizing ITN allocation to minimize global clinical incidence could, at a high budget, avert 83% of clinical cases compared to no intervention.”

1. In this reviewer's opinion it would be easier for the reader if the weighting term in the objective function would be added in the Materials and Methods section. The weighting could be added without extending the section substantially and the explanation in lines 390-393 may be easier to understand.

Thank you for this suggestion. We agree and have added this in the main manuscript.

References

1. The Global Fund. Description of the 2020-2022 Allocation Methodology 2019 [Available from: https://www.theglobalfund.org/media/9224/fundingmodel_2020-2022allocations_methodology_en.pdf.

2. Hay SI, Guerra CA, Tatem AJ, Noor AM, Snow RW. The global distribution and population at risk of malaria: past, present, and future. Lancet Infect Dis. 2004;4(6):327-36.

3. Feachem RGA, Phillips AA, Hwang J, Cotter C, Wielgosz B, Greenwood BM, et al. Shrinking the malaria map: progress and prospects. The Lancet. 2010;376(9752):1566-78.

4. Sherrard-Smith E, Winskill P, Hamlet A, Ngufor C, N'Guessan R, Guelbeogo MW, et al. Optimising the deployment of vector control tools against malaria: a data-informed modelling study. The Lancet Planetary Health. 2022;6(2):e100-e9.

5. World Health Organization. WHO Guidelines for malaria, 31 March 2022. Geneva: World Health Organization; 2022. Contract No.: Geneva WHO/UCN/GMP/ 2022.01 Rev.1.

6. Griffin JT, Bhatt S, Sinka ME, Gething PW, Lynch M, Patouillard E, et al. Potential for reduction of burden and local elimination of malaria by reducing Plasmodium falciparum malaria transmission: a mathematical modelling study. The Lancet Infectious Diseases. 2016;16(4):465-72.

7. Bertozzi-Villa A, Bever CA, Koenker H, Weiss DJ, Vargas-Ruiz C, Nandi AK, et al. Maps and metrics of insecticide-treated net access, use, and nets-per-capita in Africa from 2000-2020. Nature Communications. 2021;12(1):3589.

8. Patouillard E, Griffin J, Bhatt S, Ghani A, Cibulskis R. Global investment targets for malaria control and elimination between 2016 and 2030. BMJ global health. 2017;2(2):e000176.

9. World Health Organization. World malaria report 2022. Geneva: World Health Organization; 2022. Report No.: 9240064893.

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 well done and interesting paper examining the connection between TXNIP and GDF15. The main thrust is that TXNIP upregulation chemotherapies, such as Oxa, results in an a down regulation of GDF15 early in tumorigenesis. Later in tumorigenesis, TXNIP upregulation is less pronounced, elevating GFP15 resulting in a blockage of tumor suppressive immune responses. Generally the work is convincing. For example, it's clear that TXNIP is up regulated by Oxa in an ROS and MondoA-dependent manner. Likewise its quite clear TXNIP loss reads to an upregulation of GDF15. However, it's also quite clear that Oxa suppresses GDF15 in a manner that appears to be completely independent of TXNIP. The writing in the paper implies strongly that there is a mechanistic connection between TXNIP and GDF15, but no experiments investigate this possibility.

We feel this is very fair and is reflective of a) perhaps an overemphasis of the TXNIP knockout observation and supportive tissue data, which suggests a relationship but not a mechanistic understanding b) an underemphasis of the data in Figure 3 that shows a decrease in GDF15 after oxaliplatin treatment in TXNIP knockout lines.

We have addressed these concerns in several ways:

1. We have carried out knockdown experiments using siRNA for ARRDC4, which we felt, given its regulation by MondoA and ROS, and homology to TXNIP, may also regulate GDF15. This was found to be the case and may explain the data in Figure 3. At the very least it shows that other factors involved in oxidative stress management may have similar impacts – a form of functional redundancy. Lines 553-559 “Finally, given our previous data (Figure S4) we looked to assess the role of ARRDC4 on GDF15 expression. In the absence of oxaliplatin, knocking down ARRDC4 in DLD1 and HCT15 cells drove an increase in GDF15. When challenged with oxaliplatin, both ARRDC4 and TXNIP expression increased and GDF15 decreased. When the ARRDC4 knockdown was challenged TXNIP increased further and GDF15 decreased further (Figure S6G-J). Given the common regulatory pathways and homology between TXNIP and ARRDC4, and their similar functional roles, we suggest these data are evidence of redundancy within this system. “

We have included some context in the discussion:

Lines 930-933: “Further support for both TXNIP and ARRDC4’s role in regulating GDF15 after the induction of ROS comes from a pan cancer meta-analysis assessing the impact of metformin (which has been reported to inhibit ROS) on gene expression. Here the top two downregulated genes were TXNIP and ARRDC4 and the top four upregulated genes were DDIT4, CHD2, ERN1 and GDF1572

We have tempered the text:

Lines 522-524 “It is important to note however that here we saw clear evidence that TXNIP was not solely responsible for the downregulation of GDF15 post oxaliplatin treatment, with decreased levels seen in knockout lines (Figure 3C-G, S5E).”

Lines 926-929 “It must be stressed that these data do not place TXNIP as the sole regulator of GDF15, for example ARRDC4 can also be seen to regulate GDF15. We envisage TXNIP as one of a number of ROS-dependent GDF15 regulators, with this redundancy potential evidence of the importance of this regulatory framework.”

We have carried out additional analysis detailed in the discussion and in Figure S12 which suggests TXNIP impacts MYC function, as reported elsewhere (detailed below). For ease, the key paper can be accessed through this link https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001778

Lines 934-956: “The main shortcoming of this paper is the lack of mechanistic understanding linking TXNIP to GDF15. There are 650 transcription factors that have been shown, or are predicted, to bind to GDF15 promoter and/or enhancer regions. By assessing our list of differentially expressed genes (Suppl. Table 1-2) for the presence of these factors we identified 6 GDF15 binding TFs that show significantly decreased expression after oxaliplatin treatment in both cell lines (ATF4, MYC, SREBF1, PHB2, HBP1, KLF9). There was only one, MYC, that was downregulated by oxaliplatin treatment (validated; Figure S12A), and with this downregulation partially being rescued in a matched TXNIP knockout line (Figure S12B). We then observed that c-myc has been shown or is predicted to bind to promoter/enhancer regions of the top five transcriptomic and proteomic differentials in TXNIP knockout lines, including TXNIP itself (apart from C16orf90). Even with c-myc’s promiscuity (binds to 10-20% of all promoters/enhancers) this may be suggestive of a specific relationship. Finally, when looking at the correlations between these 6 TFs and TXNIP and GDF15 in the TCGA COAD dataset, MYC has the greatest and most significant negative correlation to TXNIP (r=-0.4631 p=1.42e-28) and the greatest and most significant positive correlation to GDF15 (r=0.4653 p=7.32e-29). ATF4 and PHB2 are the other TFs in the list, that show the same significant trends (Figure S12C), and therefore may play a role in the TXNIP-independent oxaliplatin-dependent regulation of GDF15. Further exploration of these additional TFs is outside the scope of the current manuscript.

MYC’s role in bridging from TXNIP to GDF15 is further supported by a recent paper which shows that TXNIP is “a broad repressor of MYC genomic binding” and that “TXNIP loss mimics MYC overexpression”73. Furthermore, the inter-dependent regulatory relationship between MondoA, TXNIP, and MYC has been seen in a variety of models74, whilst the impact of NAC on MYC-dependent pathways has been seen in lymphoma75. These studies lend credence to the idea that MYC is the most likely TXNIP-regulated TF that regulates GDF15 in our systems.”

It seems equally likely that TXNIP and GDF15 represent independent parallel pathways. Even if TXNIP is a direct regulator of GDF15, it's also clear that other "factors" up or down-regulated by Oxa also contribute to the regulation of GDF15. These are not explored and even though TXNIP is highly regulated genes shown Figure 2 that are not identified or discussed that may also be contributing to GDF15 regulation.

As mentioned above, the new data suggests that at least one other factor, ARRDC4, can regulate GDF15 (changes upon oxaliplatin treatment) and that MYC is a potential mechanistic bridge between TXNIP and GDF15. Whilst assessing for the transcription factor that may link TXNIP and GDF15 we found an additional 5 TXNIP-independent factors (ATF4, PHB2, SREBF1, HBP1, KLF9) that bind to GDF15 promoter/enhancer regions and are downregulated post-oxaliplatin treatment. When looking at correlations between these factors and GDF15 in the TCGA COAD dataset, ATF4 and PHB2 correlate most closely with GDF15 (when removing MYC) and so we would cautiously suggest that these may be the most pertinent. This data is now included.

Further, the experiments treating PBMCs with conditioned media contain other cytokines/factors, in addition to GDF15, that likely also contribute the observed effects on the different immune cells understudy. The conditioned media from GDF15 knock out cells are a good experiment, but the media is not rigorously tested to see what other cytokines/factors might have also been depleted.

The TXNIP knockout media is the same as that analysed by mass spec and the protein array, however as the reviewer states there is no analysis (excluding assessing for the presence or absence of GDF15) on the double knockout supernatant or over-expression supernatant. The text has been corrected as follows:

Lines 675-679. “In light of other secreted factors being seen to be regulated by TXNIP (Figure 3A-B), we included double knockouts (TXNIP and GDF15 knockout; GTKO) as well as an overexpression system (GDF15a) to test for GDF15 specific effects. However, we do not know the impact of knocking out or overexpressing GDF15 on the broader secretome.”

Perhaps a GDF15 complementation experiment would help here.

We felt that the association between GDF15 and Treg induction is reasonably well established in the literature, and so once we saw that the supernatant from our GDF15 overexpression system (+/- CD48 blockade) complemented what has already been demonstrated, we were encouraged. However we needed more – hence the TCGA data and IHC staining.

Finally, even if completely independent, a TXNIP/GDF15 ratio does seem to have utility in determining chemo-therapeutic response.

We agree – we feel that conceptually this may be the most interesting part of the project and is an example of what can be done with these tools.

Other major points: 1. Please label the other highly regulated genes shown in Fig 2A and B. Might they also explain some of the underlying biology. This could be on the current figures or in a supplement, though the former is preferred.

Many thanks – we have done this.

Please address why the TXNIP induction is so much less in patient-derived organoids vs. cell line spheroids (Fig S2). By the western blots, TXNIP inductions in the organoids looks quite modest. Further, the text is quite cryptic and implies that the "upregulation" is similar in both organoids and spheroids.

You are absolutely correct. Many apologies, the wording has changed:

Lines 320-323 “In both models we observed the upregulation of TXNIP mRNA (Figure S2E-H) and TXNIP protein (Figure S2I-L) after oxaliplatin treatment, with spheroids showing greater responsiveness. This difference is most likely due to culturing conditions or differences in the number and location of cycling cells.”

We have two possible explanations. Firstly the media in which the organoids are cultured contains a lower glucose concentration than that used for the spheroids. As per some of our new data (Figure S3 – later in the rebuttal), the upregulation of TXNIP after oxaliplatin is glucose dependant, with lower concentrations resulting in less of a differential. Secondly, while restricted to the periphery, the Ki67 signal in DLD1 spheroids is quite pronounced indicating that, within the outer zone, many cells (probably the majority) are in the S/G1/G2 phase of the cell cycle at any given point in time (figure below this text).

This is not the case for the organoids, where the Ki67 (and pCDK1) signal is quite weak, and only sporadic in the outer layer. So we believe that there are many more rapidly cycling cells in the most drug-exposed layer of spheroids when compared to the comparable region in organoids. As the spheroid cells are likely cycling more rapidly, they would also be expected to be more adversely affected by the drug within the finite drug treatment window. Indeed, these spheroids grow large, and quite quickly. If the organoid cells are cycling more slowly and if, within the cell layer most exposed to drug, these cycling cells are less abundant, then the TXNIP response may well be subdued in organoids when compared with spheroids.

We have decided to not include the above (full) explanation and figure within the new draft, as we feel it may distract from the central message. However do let ourselves and the editor know if you disagree.

What was the rationale of performing the MS experiment on control and TXNIP KO DLD1 cells in the absence of oxaliplatin? The other experiments in Fig 3 clearly show that Oxa can repress GDF15 even in the absence of TXNIP, which implicates other pathways. ARRDC4? Or something else? This needs to be addressed.

We adopted this approach because of the order in which the assays occurred and technical issues surrounding the use of post-oxaliplatin treated supernatant. By the time we moved to the proteomics we had already identified, and validated, GDF15 as our number one candidate (initially from the protein array), in terms of response to oxaliplatin and dependence on TXNIP. This led us to the next stage of the project – to assess the environmental impacts of this factor in vitro before validation in situ. To do this, aware of the issue of contaminated recombinant GDF15, we decided early on to use cell line supernatant. We carried out some pilot studies on immune cells using supernatant from oxaliplatin treated cell lines and we had several technical issues (difficulty in determining the correct controls, immune cell death…). This changed the emphasis to using supernatant from knockout models rather than knockout and treated models. Before we began these assays in earnest we wanted to assess exactly what was enriched in TXNIP knockout supernatant and so we turned to proteomics. When this further validated GDF15, we then generated GDF15 and TXNIP/GDF15 knockouts to further elucidate GDF15’s role specifically.

With regards the other pathways, as you correctly predicted, ARRDC4 also appears to regulate GDF15 – many thanks for helping with this line of enquiry. Please see earlier in the rebuttal for more details and the data.

The data in 3J with the MondoA knockdown is not convincing. The knockdown is weak and TXNIP goes down a smidge. Agree that GDF15 goes up

We agree. We have re-run this and pooled the densitometry data – see new figure below (Panel 3J).

Minor points 1. Line 79. The "loss" of TXNIP/GDF15 axis is confusing. It's really loss of TXNIP and upregulation of GDF15, right?

Absolutely - corrected to responsiveness.

Lines 144-147: “Intriguingly, multiple models including patient-derived tumor organoids demonstrate that the loss of TXNIP and GDF15 responsiveness to oxaliplatin is associated with advanced disease or chemotherapeutic resistance, with transcriptomic or proteomic GDF15/TXNIP ratios showing potential as a prognostic biomarker.”

Please provide an explanation for the different stages in tables 1 and 2. This will likely not be clear to non-clinicians.

Many thanks. The following has been added at the bottom of the second table.

Lines 304-309: “The TNM staging system stands for Tumor, Node, Metastasis. T describes the size of the primary tumor (T1-2; 5cm). N describes the presence of tumor cells in the lymph nodes (N0; no lymph nodes. N1-3 >0). M describes whether there are any observable metastases (M0; no metastases. M1; metastases). The clinical stage system is as follows: I/II; the tumor has remained stable or grown, but hasn’t spread. III/IV; the tumor has spread, either locally (III) or systemically (IV).”

Line 231 should probably read ...cysteine (NAC), a reactive oxygen species inhibitor,

Many thanks - corrected

Line 247, should be RT-qPCR I think.

Many thanks - corrected

Lines 343-345. I don't quite understand the wording. Does this mean to say that 675 soluble proteins were not changed between the condition media from both cell populations?

Yes, exactly this. We have removed as this is superfluous and confusing.

The data in FigS1 B and C don't seem to reach the standard p value of > 0.05

Very true – we have rewritten the text to make sure the reader knows there is no significance.

Lines 269-271. “High levels of both the protein (significantly) and the transcript (not significantly) were seen to be associated with favourable prognosis (Figure 1G,H and S1B,C).”

**Referee Cross-Commenting**

cross comment regarding referees 2 and 3 above. I'm am convinced that TXNIP is at least contemporaneously upregulated with GDF15 downregulation. However, the strong implication from the writing is that TXNIP regulates GDF15 directly. I agree with the comment above that exploring mechanisms may be open-ended especially as TXNIP has been implicated in gene regulation by several different mechanism. I'd be satisfied with a more open-minded discussion of potential mechanisms by which TXNIP may repress GDF15 and the possibility of other parallel pathways that likely contribute to GDF15 repression.

Many thanks, this is a generous and understanding approach. As described above we have carried out extra analysis and have found 6 differentially regulated transcription factors which have been shown to bind GDF15 promoter or enhancer regions with 1 of these, MYC, being significantly affected in the TXNIP knockout cell lines, which in combination with supportive literature suggests a degree of TXNIP dependence. We have also identified ARRDC4 as an additional regulator of GDF15 – again please see above.

Reviewer #1 (Significance (Required)):

This is an interesting contribution but the mechanistic connection between GDF15 and TXNIP is relatively weak. That said, even as independent variables they do seem to have utility in predicting therapeutic response.

Many thanks for the comment – we concur. We have reanalysed our data looking for relevant transcription factors (those that bind GDF15 promoter / enhancer regions) finding MYC as the most likely bridge. Please see above.

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

The manuscript by Deng et al. investigates a mechanistic link between TXNIP and GDF15 expression and oxaliplatin treatment and acquired resistance. They observe an upregulation in TXNIP expression in the tumors of patients who have previously received chemotherapy. They demonstrate oxaliplatin-driven MondoA transcriptional activity is what underlies the induction of TXNIP. They further demonstrate that TXNIP is a negative regulator of GDF15 expression. Together, oxaliplatin induces MondoA activity and TXNIP expression, resulting in a downregulation of GDF15 expression and consequently decreased Treg differentiation.

Major Comments

1. The authors suggest that TXNIP induction and GDF15 downregulation are a common effect of chemotherapies; however, the mechanistic studies were limited to oxaliplatin. The authors should clarify this point through further investigation using other commonly used CRC chemotherapies (5-FU, irinotecan, etc.),or through textual changes. To be clear, I think that the oxaliplatin results could potentially stand on their own but would require additional clarification. For example, regarding the patient samples analyzed in 1D and 4F, which patients received oxaliplatin? Could the analysis of publicly available molecular data be drilled down to just the patients who received oxaliplatin?

Many thanks – this is an excellent point. Firstly, all the patients in 1D and 4F received oxaliplatin. Secondly, we have included new data looking at the impact of other chemotherapies (FOLRIRI, FU-5 and SN-38) on aspects of the study, ultimately finding that these processes (especially an anti-correlation between GDF15 and TXNIP changes upon chemo treatment) appear to be specific to oxaliplatin. These data have been added (Figure S11) and throughout the emphasis has been switched from chemotherapeutic treatment to oxaliplatin treatment.

Lines 796-799: “To check if the pre-treatment GDF15/TXNIP ratio could be used for patients treated with FOLFIRI we performed the same analyses finding no significance (S11A-D). This oxaliplatin specificity was then confirmed by western blot analysis in DLD1 and HCT15 cells treated with 5-FU or SN38 (Figure S11E-F).

Example of change of emphasis from ‘chemotherapy’ to ‘oxaliplatin’ – lines 139-142: “Here, in colorectal adenocarcinoma (CRC) we identify oxaliplatin-induced Thioredoxin Interacting Protein (TXNIP), a MondoA-dependent tumor suppressor gene, as a negative regulator of Growth/Differentiation Factor 15 (GDF15).”

The data demonstrating the induction of MondoA transcriptional activity and TXNIP expression in response to oxaliplatin treatment is quite convincing. The data regarding ROS induction of TXNIP is interesting, especially in light of other studies arguing that ROS limits MondoA activity (PMID: 25332233). Given this apparent disparity, I think that this study could really be strengthened by also investigating other potential mechanisms of oxaliplatin induction of MondoA. In particular, given many studies arguing for direct nutrient-regulation of MondoA, the authors should address the potential for oxaliplatin regulation of glucose availability and a potential glucose dependence of oxaliplatin-induced TXNIP. 2

In line with the previous point, since MondoA activity and TXNIP expression are sensitive to glucose levels, the authors should investigate oxaliplatin-regulation of TXNIP under physiological glucose levels. No need to replicate everything, just key experiments.

We feel these are excellent point and really help the piece – many thanks. We have carried out assays around these points suggested and have included the findings in the new draft – see below.

Lines 332-339: “As such, we went back to first principles and assessed the impact of different concentrations of glucose on TXNIP induction +/- oxaliplatin treatment, finding a concentration dependent effect (Figure S3A). Intriguingly, high glucose alone was able to induce increased TXNIP expression. We then assessed if oxaliplatin treatment drove an increase in glucose uptake, with this seen at concentrations >10mM (Figure S3B). Next, to investigate the impact of glucose metabolism, and consequent ROS generation, on TXNIP induction we treated cells with Antimycin A, an inhibitor of oxidative phosphorylation, finding a complete block in oxaliplatin-induced TXNIP (Figure S3C).”

The authors did a good job of linking TXNIP and GDF15 in untreated conditions; however, the data arguing for oxaliplatin regulation of GDF15 through TXNIP is less clear. For example, in 3B-H, oxaliplatin treatment reduces GDF15 approximately to the same extent in the NTC and TKO cells, potentially in line with a mechanism of downregulation that doesn't involve TXNIP.

A very salient point and completely in line with the other reviewers. We have carried out a few additional analyses mentioned previously in this letter. The most pertinent for this specific point are the experiments around ARRDC4, where we found evidence to suggest that, like TXNIP, it regulates GDF15.

Minor Comments

1. The presentation of data in Figure 5 is confusing. A-B include raw cell numbers, whereas C-F show "normalized proliferation." What does this mean? And how was the normalization done?

Apologies for this. Legend test has been corrected to “Normalised proliferation (normalised to MFI from control: i.e. cells treated with supernatant from NTC cells) on gated CD3+CD8+ or CD3+CD4+ cells is shown. n=6. (G-H) Normalised IFNg concentrations (normalised to MFI from control: i.e. cells treated with supernatant from NTC cells) in the supernatant of cells from C-F.” (lines 727-729).

**Referee Cross-Commenting**

cross-comment regarding reviewer #1

I agree with the referee that the link between TXNIP and GDF15 is weak, though as I mentioned before, this is particularly true in the context of oxaliplatin-regulation of TXNIP. I agree that given all the presented data, it is likely that oxaliplatin-regulation of TXNIP and GDF15 are independent. In my opinion, the referee brought up all valid concerns, but this is by far the biggest concern that I share.

We agree that this is the weakest aspect of the paper, however our new analyses plus supportive literature, suggests that the relationship between TXNIP and GDF15 may be mediated by MYC (please see above)

cross-comment regarding reviewer #3

The major concern that this referee addresses is whether another transcription factor supersedes the proposed MondoA/TXNIP induction in regulating GDF15 expression in later stage CRC. In my opinion, this another other concerns of the referee are all valid, but still I remain unconvinced that TXNIP induction underlies the oxaliplatin-regulation of GDF15. I think fleshing out that aspect of the study would potentially help the authors tease apart how this potential MondoA-TXNIP-GDF15 axis is dysregulated later in CRC progression.

This is a great discussion. Interestingly enough, c-myc is seen at higher levels in late stage CRC (Hu X, Fatima S, Chen M, Huang T, Chen YW, Gong R, Wong HLX, Yu R, Song L, Kwan HY, Bian Z. Dihydroartemisinin is potential therapeutics for treating late-stage CRC by targeting the elevated c-Myc level. Cell Death Dis. 2021 Nov 5;12(11):1053. Doi: 10.1038/s41419-021-04247-w. PMID: 34741022; PMCID: PMC8571272.), is seen as an important factor in resistance, and as this review argues, is driven by stress (Saeed H, Leibowitz BJ, Zhang L, Yu J. Targeting Myc-driven stress addiction in colorectal cancer. Drug Resist Updat. 2023 Jul;69:100963. Doi: 10.1016/j.drup.2023.100963. Epub 2023 Apr 20. PMID: 37119690; PMCID: PMC10330748.). So it is very plausible that the partial TXNIP-mediated regulation of myc in early / sensitive CRCs that we may be observing, and has been reported recently (TXNIP loss expands Myc-dependent transcriptional programs by increasing Myc genomic binding Lim TY, Wilde BR, Thomas ML, Murphy KE, Vahrenkamp JM, et al. (2023) TXNIP loss expands Myc-dependent transcriptional programs by increasing Myc genomic binding. PLOS Biology 21(3): e3001778. https://doi.org/10.1371/journal.pbio.3001778) is lost in late stage / resistant CRCs. If this is the case, in effect what we would have observed is the loss of a stress-associated method (TXNIP) of controlling c-myc activity. What makes our collective lives difficult is that, as reported “this expansion of Myc-dependent transcription following TXNIP loss occurs without an apparent increase in Myc’s intrinsic capacity to activate transcription and without increasing Myc levels.” (TXNIP loss expands Myc-dependent transcriptional programs by increasing Myc genomic binding Lim TY, Wilde BR, Thomas ML, Murphy KE, Vahrenkamp JM, et al. (2023) TXNIP loss expands Myc-dependent transcriptional programs by increasing Myc genomic binding. PLOS Biology 21(3): e3001778. https://doi.org/10.1371/journal.pbio.3001778)

Reviewer #2 (Significance (Required)):

Generally speaking the experiments are well controlled and the findings are significant and novel. Though the link between MondoA activity and ROS could be strengthened, and the data could be validated under more physiological settings. Further, the authors should clarify their interpretations so as to not overstate the findings.

Many thanks for the comments. We have taken onboard the need for more physiological settings and have included varying levels of glucose to reflect concentrations in different environments. We have repeated the siMondoA work in 3J strengthening the conclusions wrt its impact on TXNIP and GDF15 expression (see above).

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

In this well-written manuscript, the authors show that chemotherapy increases a MondoA-dependent oxidative stress-associated protein, TXNIP, in chemotherapy-responsive colorectal cancer cells. They show that TXNIP negatively regulates GDF-15 expression. GDF-15, in turn, correlates with the presence of T cells (Treg), and inhibits CD4 and CD8 T cell stimulation. In advanced disease and chemo-resistant cancers, upregulation of TXNIP and downregulation of GDF-15 appear to get lost. Based on a somewhat smallish data set, the authors suggest that the pre-treatment GDF-15/TXNIP ratio can predict responses to oxaliplatin treatment. This is a very interesting, novel finding. In general, the quality of the experiments and the data are high and the conclusions appear sound. Still, there are a number of aspects that should still be improved:

The observed loss of the ROS - MondoA - TXNIP - GDF15 axis in chemoresistant and/or metastatic tumors implies that another transcription factor or pathway becomes dominant upon tumor progression. As this switch would be key to better understanding the mechanism underlying the prognostic role of the TXNIP/GDF15 ratio, the authors should at least do data mining followed by ChEA or Encode (or other) analysis to identify transcription factors or pathways that become activated in late-stage/metastatic CRC cells. There is a high likelihood that a transcription factor or pathway involved in GDF-15 upregulation in cancer (e.g. p53, HIF1alpha, Nrf2, NF-kB, MITF, C/EBPß, BRAF, PI3K/AKT, MAPK p38, EGR1) supersedes the inhibitory effect of the MondoA-TXNIP axis. As it stands, the proposed loss of function of the ROS - MondoA - TXNIP - GDF-15 axis is far less convincing than almost all other aspects of the study.

An extremely fair point. We adopted a similar approach to that suggested – as mentioned above, we looked at TFs that bind to GDF15 promoter/enhancer regions and then looked at the presence of these in our transcriptomic data – specifically any evidence of change post oxaliplatin treatment. We found 6 such TFs that were decreased post-oxaliplatin treatment. We then looked for any evidence of TXNIP dependence in these TFs by comparing post-oxaliplatin treatment across NTC and TXNIP knockout lines, when we did this we found only one GDF15 promoter/enhancer binding TF was significantly changed: MYC. We then looked at the relationship between MYC,TXNIP, and GDF15 against the other 5 ‘control’ TFs in the TCGA COAD dataset, we found that MYC showed the strongest correlations, in the ‘correct’ directions. This finding was further backed up in the literature where a TXNIP knockout in a breast cancer model drove c-myc-dependent transcription, whilst c-myc has been observed to increase in later stage CRC patients, is associated with cellular stress and resistance. The collective evidence therefore suggests that MYC is the factor that is initially at least partially regulated by TXNIP, before this regulation is lost in advanced / resistant disease. Continuing on this line, it is likely that the predictive GDF15/TXNIP ratio is at least in part, a measure of c-myc responsiveness to oxaliplatin. All the while we must bear in mind TXNIP-independent oxaliplatin-dependent regulation of GDF15, most likely ARRDC4, as described earlier in this document.

Using pathway analysis software to compare our transcriptomic data from cell lines treated with/without oxaliplatin, the most likely pathways upstream of MYC/c-myc that are negatively affected by chemotherapy are BAG2, Endothelin-1, telomerase, ErbB2-ErbB3 and Wnt/B-catenin. When looking at the comparison of UTC and resistant lines’ transcripts there is only one key component of these pathways which is upregulated in both lines - ERBB3 – which has already been shown to be important in CRC metastasis and resistance (Desai O, Wang R. HER3- A key survival pathway and an emerging therapeutic target in metastatic colorectal cancer and pancreatic ductal adenocarcinoma. Oncotarget. 2023 May 10;14:439-443. doi: 10.18632/oncotarget.28421. PMID: 37163206; PMCID: PMC10171365.). It is highly speculative, but our data suggests the most likely pathway to supersede TXNIP in its (partial) regulation of MYC is the ErbB2-ErbB3 pathway.

My further criticisms are mostly more technical:

Figure 2 I-L: What was the extent of MondoA downregulation achieved by siRNA treatment? Could the effects also be seen with the small molecule mondoA inhibitor SBI-477 (or a related substance)?

This experiment has been repeated. The pooled densiometric data is also now given (please see above).

How do you explain the different GDF-15 levels between untreated non-target control cells (NTC) and TXNIP knock-down cells (TKO) in Figures 3C-F?

The only way to interpret this is that there is a TXNIP-independent pathway regulating GDF15 expression after oxaliplatin treatment, as described this is most likely to be ARRDC4 - the text has been updated to:

Lines 522-524: “It is important to note, however, that we saw clear evidence that TXNIP was not solely responsible for the downregulation of GDF15 post oxaliplatin treatment (Figure 3C-G, S6E).”

In figures 3 E-G the dots for the individual measurements should be indicated. This would be more informative than just the bar graphs.

Completed.

Figure 4C,D and Table 3: Data on the role of GDF-15 in CRC are largely valedictory of previous work (e.g. Brown et al. Clin Cancer Res 2003, 9(7):2642-2650, Wallin et al., Br J Cancer. 2011 May, 10;104(10):1619-27). Therefore, the previous studies should be cited.

Apologies for the oversight and many thanks – this is an excellent addition.

Figure 5C-F: Please indicate in the figure legend how proliferation was assessed.

Many thanks. This was noticed by another reviewer also. We have changed the text to include how the data was normalised: “(C-F) Labelled PBMCs were stimulated with anti-CD3 and anti-CD28 for 4 days in the presence of fresh supernatant from indicated cell lines, before being stained with anti-CD3 and anti-CD8 (C-D) or anti-CD4 (E-F) antibodies and measured by flow cytometry. Normalised proliferation (normalised to MFI from control: i.e. cells treated with supernatant from NTC cells) on gated CD3+CD8+ or CD3+CD4+ cells is shown. n=6. (G-H) Normalised IFNg concentrations (normalised to MFI from control: i.e. cells treated with supernatant from NTC cells) in the supernatant of cells from C-F.” (lines 724-730)

Figure S8E-G: Please indicate the analysed parameters in the graphs. In Figure S8G, the legend just indicates that "aggression of tumour" is dichotomized and plotted. This clearly requires a better definition.

Many thanks, this has been changed as per the below.

Lines 862-868: “(E-G) Receiver operating characteristic (ROC) curves showing area under the curve and p values for the use of GDF15/TXNIP ratio in predicting origin of cell line (E; primary; DLD1, HCT15, HT29, SW48 [n=4] or secondary; DiFi, LIM1215 [n=2]), sensitivity to oxaliplatin (F; parental DLD1 (plus biological repeat), HCT15 [n=3] or resistant DLD1 (plus biological repeat), HCT15 [n=3]), aggression of tumor (G; non-aggressive; The authors propose a novel ROS - MondoA - TXNIP - GDF15 - Treg axis, where MondoA activation, TXNIP up- and GDF-15 downregulation enhance tumor immunogenicity. While this axis has been analyzed in some detail, GDF-15 is not only linked to induction of regulatory T cells. There has been a report showing that GDF-15/MIC-1 expression in colorectal cancer correlates with the absence of immune cell infiltration (Brown et al. Clin Cancer Res 2003, 9(7):2642-2650). The link between GDF-15 and immune cell exclusion has also been confirmed in other conditions, including different cancers (Kempf et al. Nat Med 2011, 17(5):581-588, Roth P et al. Clin Cancer Res 2010, 16(15):3851-3859, Haake et al. Nat Commun 2023, 14(1):4253). A key mechanism is the GDF-15 mediated inhibition of LFA-1 activation on immune cells. As the authors argue that the described pathways turns cold tumors hot in response to oxaliplatin-based chemotherapy, this GDF-15 dependent immune cell exclusion mechanism might be at least as relevant than induction of Treg. Likewise, inhibition of dendritic cell maturation by GDF-15 (Zhou et al. PLoS One 2013, 8(11):e78618) could explain why GDF-15high tumors are immunologically cold. Reviewed in 3

The authors propose that the pathways discovered by them contributed to the "heating up" of the tumor microenvironment after oxaliplatin-based chemotherapy. The authors should thus look in their data sets for the presence of cytotoxic T cells and their possible correlation with TXNIP and GDF-15 levels.

This is a wonderful explanation – many thanks. We have taken the opportunity to assess the impact of GDF15 expression on a variety of T cell markers (Figure S9). In this data a negative association between GDF15 and CD8 CTLs can clearly be seen, as predicted by the reviewer.

Lines 712-717: “To assess if the GDF15-dependent presence of Tregs may be associated with a decrease in activated cytotoxic CD8 T cells, we interrogated the TCGA COAD dataset. We found that low GDF15 tumors carried significantly higher levels of CD8, CD69, IL2RA, CD28, PRF1, GZMA, GZMK, TBX21, EOMES and IRF4 (Figure S9); transcripts indicative of activated cytotoxic CD8 T cells. High GDF15 tumors were enrichment for FOXP3 and, interestingly, RORC (Figure S9). These data support the hypothesis that GDF15 induces Foxp3+ve Tregs which inhibit CD8 T cell proliferation and activation in the TME.”

The paragraph on GDF-15 receptors needs to be corrected: The purported role of a type 2 transforming growth factor (TGF)-beta receptor in GDF-15 signalling had been due to a frequent contamination of recombinant GDF-15 with TGF-beta (Olsen et al. PLoS One 2017, 12(11):e0187349). There have been a number of screenings for GDF-15 receptors that have all failed to show an interaction between GDF-15 and TGF-beta receptors. Instead, only GFRAL was found in these large-scale screenings (Emmerson et al. Nat Med 2017, 23(10):1215-1219, Hsu et al. Nature 2017, 550(7675):255-259, Mullican et al. Nat Med 2017, 23(10):1150-1157, Yang et al. Nat Med 2017, 23(10):1158-1166). The one subsequent report that shows a link between GDF-15, engagement of CD48 on T cells and induction of a regulatory phenotype (Wang et al. J Immunother Cancer 2021, 9(9)) still awaits independent validation. Considering that CD48 lacks an intracellular signaling domain that would be critical for a classical receptor function, I recommend to be more cautious regarding the role of CD48 as GDF-15 receptor. Given the mechanism outlined by Wang et al. the word interaction partner might be more apt. Moreover, an anti-GDF-15 antibody would be a good control for the experiments involving an anti-CD48 antibody in Figure 5.

Thank you so much for this concise and highly informative paragraph. We have changed the text to read:

202-204: “As a soluble protein, GDF15 exerts its effects by binding to its cognate receptor, GDNF-family receptor a-like (GFRAL)44,45,46,47 or interaction partner, CD48 receptor (SLAMF2)43, with the latter still requiring additional verification.”

We would have ideally included an anti-GDF15 antibody in the CD48 assay at the time but didn’t have the foresight. We have included the additional text to temper any conclusions.

Lines 701-711: “Furthermore, when stimulating naïve CD4 T cells in the presence of GDF15 enriched supernatant we were able to both differentiate these cells into functional Tregs and also block the generation of this functionality using an anti-CD48 antibody (Figure 5M-N). However, it must be stressed that the binding and functional impacts of GDF15’s interaction with CD48 still require further verification.”

Cell surface externalization of annexin A1 has been described as a failsafe mechanism to prevent inflammatory responses during secondary necrosis (PMID: 20007579). Thus, I am surprised that the authors list annexin A1 among the immune-stimulatory molecules exposed or released in response to chemotherapy-induced cell death (line 103). Please clarify!

We agree – it shouldn’t be there!! Removed. Many thanks.

**Referee Cross-Commenting**

Regarding the cross-comment by referee 2: In my opinion, the data shown in Figure 3C-H clearly demonstrates that TXNIP can repress GDF-15 expression. I agree that there will likely be further regulators. The GDF-15 promoter is constantly regulated by a multitude of factors (which mostly induce transcription). As downregulation of GDF-15 in response to oxaliplatin is the opposite of the frequently described induction of GDF-15 upon chemotherapy, net effects may always be "smudged" by contributions from different pathways (e.g. by cell stress due to siRNA transfection). Therefore, I believe that the data are as good as it will get. Accordingly, I would not force the authors to further amplify the observed effect.

Many thanks for your understanding – yes, GDF15 has >650 TFs that bind its promoter/enhancer regions – a number we found rather daunting. Happily your comments and those of the other reviewers inspired us to dig and we now have data that is supportive of MYC’s and ARRDC4’s involvement – detailed throughout this reply.

cross comment regarding referee #1: I share the general assessment of the referee and recognize the very detailed mechanistic analysis. To further support the moderate effects of the MondoA knockdown, a small molecule inhibitor like SBI-477 might be useful. (I had already suggested using this inhibitor to support these data.)

Many thanks for the suggestion. We opted to increase the number of siRNA repeats instead – with the data included in Figure 3J (above).

Still, my view on the potential relevance of oxaliplatin-induced, TXNIP-independent downregulation of GDF-15 differs from that of referee 1. In the clinics, platinum-based chemotherapy is one of the strongest inducers of GDF-15 (compare Breen et al. GDF-15 Neutralization Alleviates Platinum-Based Chemotherapy-Induced Emesis, Anorexia, and Weight Loss in Mice and Nonhuman Primates. Cell Metabolism 32(6), P938-950, 2020.DOI:https://doi.org/10.1016/j.cmet.2020.10.023). I was thus surprised that the authors found a pathway, which leads to an outcome that an exactly opposite effect.

This is fascinating that oxaliplatin drives this increase in GDF15 – we were unaware of this paper. Looking at figure 2(H-K), GDF15 is being produced from multiple non-diseased tissues after systemic chemotherapy – even at day 19 post-treatment – this suggests that wrt this study, systemic GDF15 could not be used as a readout of success or otherwise – which is extremely helpful! Thank you.

Thus far, the only obvious reason for reduced GDF-15 secretion upon treatment with cytotoxic drugs was a reduction in tumor cell number due to cytotoxicity.

Please do not discount this. This study was focused on the cells which survived oxaliplatin treatment – the cells which did not were discarded. Our view, given your input, would be a complex picture where in early stages systemic GDF15 goes up, due to off-target effects, but locally levels drop owing to cell death and this, and other, stress-related pathways in the remaining tumor cells.

Still, the authors managed to convince me that the described pathway (ROS - MondoA - TXNIP - GDF-15) exists. (Here, I still largely concur with referee 1.) Moreover, as we have identified some factors required for GDF-15 biosynthesis that could easily interact with TXNIP, I find the proposed mechanism plausible.

Extremely encouraging for us to hear!

Nevertheless, as a downregulation of GDF-15 in response to chemotherapy is hardly ever observed in late-stage cancers, I believe that the observed switch in pathway activation between early- and late-stage cancers might be highly relevant - in particular, as there is so much evidence for platinum-based induction of GDF-15 in late-stage cancer patients. Emphasizing the divergent clinical observations (e.g. by Breen et al.) could thus help to put the finding into perspective.

Very much agree. We did see this phenomenon in LIM1215 cells (Figure 6B) and the resistant lines we generated continually produced higher levels.

Analysing TXNIP-independent mechanisms involved in the oxaliplatin-dependent repression of GDF-15, as suggested by referee #1, will require enormous efforts and resources, and may still turn out to be fruitless. Personally, I would thus be content if the authors just mentioned possible contributions from other pathways upon cancer progression. To me, the described pathway seems to be limited to early-stage cancers, and the actual finding that GDF-15 is downregulated is an interesting observation, irrespective of further involved pathways.

Many thanks – this is extremely fair. Happily we have managed to make some tentative steps forward in highlighting the potential role of MYC, and the suggestion of redundancy wrt ARRDC4, but as you say, much more work needs to be done to fully understand these processes.

cross comment regarding referee #2: I fully agree with the referee that activation of the pathway by further chemotherapeutic drugs could be a valuable addition. As Guido Kroemer´s lab has described oxaliplatin to induce a more immunogenic cell death compared to other platinum-based chemotherapies, even a rather limited comparison between oxaliplatin and cisplatin could be very interesting.

Absolutely agree – extra data on this has been included in Figure S11, which is included earlier in this letter. We also uncovered a meta-analysis using metformin, which has been seen to inhibit ROS, where TXNIP and ARRDC4 are the top two downregulated transcripts whilst GDF15 appears in the top four upregulated. This may suggest that chemotherapeutic immunogenicity, at least through the presence or absence of GDF15, may in part be driven by ROS.

Lines 930-933: “Further support for both TXNIP and ARRDC4’s role in regulating GDF15 after the induction of ROS comes from a pan cancer meta-analysis assessing the impact of metformin (which has been reported to inhibit ROS) on gene expression. Here the top two downregulated genes were TXNIP and ARRDC4 and the top four upregulated genes were DDIT4, CHD2, ERN1 and GDF1572 “

Reviewer #3 (Significance (Required)):

In general, this is a very interesting manuscript describing a cascade of events that may contribute to successful chemotherapy (which likely requires induction of an immune response against dying tumor cells.) The observation that this pathway is only active in early/non-metastatic cancer cells is striking. Unfortunately, the authors cannot explain inactivation of this pathway in later stage/ metastatic/ highly aggressive cancers. Understanding this switch could easily be the most important finding triggered by this report. Therefore, I highly recommend to make some effort in this direction. Strikingly, the authors find that disruption of TXNIP-mediated GDF-15 downregulation is strongly associated with worse prognosis. They also suggest that this ratio could indicate whether a patient will respond to oxaliplatin-based chemotherapy.

This is again very fair – we have posited a potential mechanism for the loss of this switch elsewhere in this reply– one which involves a change in TXNIP-mediated MYC regulation and/or increased HER2-HER3 signalling – but although reasonable for a rebuttal (and publication in that context) we do not feel we have the evidence to include this within the full manuscript.

Altogether, the findings described in manuscript are very novel and may have prognostic (or, in case of the presumed loss of the MondoA - TXNIP - GDF-15 pathway) therapeutic implications. Thus, the manuscript certainly fills various gaps and should be of major interest for cell biologists working on immunogenic cell death, or colorectal cancer, or MondoA, TXNIP or GDF-15. Still, due to its translational implications, it would also be worthwhile reading for a large number of researchers in the oncology field.

We are very grateful for your kind comments.

1 Sinclair, L. V., Barthelemy, C. & Cantrell, D. A. Single Cell Glucose Uptake Assays: A Cautionary Tale. Immunometabolism 2, e200029, doi:10.20900/immunometab20200029 (2020).

2 Yu, F. X., Chai, T. F., He, H., Hagen, T. & Luo, Y. Thioredoxin-interacting protein (Txnip) gene expression: sensing oxidative phosphorylation status and glycolytic rate. J Biol Chem 285, 25822-25830, doi:10.1074/jbc.M110.108290 (2010).

3 Wischhusen, J., Melero, I. & Fridman, W. H. Growth/Differentiation Factor-15 (GDF-15): From Biomarker to Novel Targetable Immune Checkpoint. Front Immunol 11, 951, doi:10.3389/fimmu.2020.00951 (2020).

This ominous realization did not occur and come together for me until just now:

Kate's influence did not start with Kate directly. It would have started with her son Liam. I've not recognized until now the likely significant role he plays in this. He is her son. He would have already been fully traumatized by Kate or by the situation with his dad, depending on if it existed, but if it did or didn't, the fear/abandonment/insecure attachment disorder would be entrenched in both Kate and Liam and they would be reinforcing it in each other. Rhyanna working with Liam at Subway would have been the first contact in which casual conversation would begin the subtle campaign by Liam via trauma reenactment (and also fueled by being a teenage boy meets girl savior/peacocking mentality) that at first innocuously and then overtly was showing (manipulating into false belief) that she is victimized. Liam then notifies Mom of "the recruit", probably a genuine felt statement like "Mom, there's this girl at work and it sounds like she's going through what we went through and we could help her". Then Mom [Kate], which we know this happened, took the initiative to contact her (or told Liam to bring her over to the house to hangout so she could then introduce herself and have 'a talk' with her). Phone numbers were shared, instructions to not let Dad know where they lived were given, taking out to dinners were done, sharing of "stories about my husband we don't tell other people so please don't share this" were given about "my dangerous psychotic husband that Liam and I had to flee from and go on the run because the system couldn't save us so we had to act outside it". This matches the dynamic and origination story of every cult/radical "church"/scientology/NXIVM story I know and it is the same dynamic whether it's the pathogenic parent or pathogenic adult influence which in this way has an extra component of evolution. Ie, the pathogenic adult has created/obtained a pathogenic "victimized" subordinate follower. The follower then acts as a relatable/ice-breaking recruiter that has the effect on the target of " they're my peer, they're like me, I can therefore trust the accuracy of what they're saying more and am more willing to listen". Then when the follower eventually introduces the pathogenic adult, the critical judgement defence of the target is suppressed/ignored because the target has made the naive judgment error that since I believe and feel trusting in this peer, I can put that trust into someone he is introducing me too. And because that person is "the adult in the room" this person instantly gets, erroneously, the elevated security clearance in the target's mind that this person is a "trusted"+"adult"+"who understands me"+"has my best interest"+"and knows what I need". Additionally, when speaking with this adult, should the target's defense mechanisms of critical judgement start turning on, the target then looks to a reference point to "reality test", and the follower, Liam, is immediately on hand and present almost daily to act as that reference point nodding reassuringly when the target glances over [literally or metaphorically]. ..... Combine this with a parent who is getting sicker and sicker, who's observably by the child who knows her father well can tell his fear, anxiety (particularly regarding his ability to provide for them both), and sadness because of his non-improving sickness from a mysterious unknown deadly pandemic disease, a parent who is the SOLE parent and there is no second parent to reality test against and get reassuring grounded perspective (ie you are not victimized, dad isn't going to kill himself, yes this is a tough situation but we and you are not a victim and this is not a Hallmark/teen drama, and tough situations like this have long been and are a prolific part of human life and we can more than handle this situation and frankly will serve to accelerate your empowered growth and deeper understanding, meaning, passion, joy of life and further shedding of vulnerability to irrational and mismanagement of uncontrollable fear as a general skill set in your personal quiver. This all is the loss of the second, of which there may only be 2, fundamental defense mechanisms to safeguard a child's sound critical analytical/judgement skills. It is easy to empathize with a child's daily living experience, especially an adolescent, how these are the 2 mechanisms which are functioning by which they are consuming and assembling all new knowledge and understanding. #1 They first use their incumbent developed analytical/judgement skills to self analyze a concept or problem or question. #2 They verify that determination with their trusted source of truth and protection, ie their parents (a reality test). Perhaps this at the root of the common report "teenagers think they know everything". It's probably the first time the first mechanism is developed strongly enough to feel like it can safely be used in its own. And in being the first time, many errors will be made and in many of those errors the use of verification of mechanism 2 will not be used. An ill unimproving parent will exacerbate the error to not use mechanism 2. Fear and anxiety will exacerbate errors in mechanism number 1. Severity of those insults would proportionally affect the rate of error. Malfunctions in both mechanisms would have a multiplicative effect on damaging erroneous conclusions the child arrives at and the damage further choices on those erroneous conclusions causes. Then when the "virus" of the narcissistic/BLP cross generational shared persecutory delusion boundary violation gains entry into this now much increased "analytically vulnerable" child, it has the critically added effect of disabling mechanism 2 since the patent now becomes "all bad [splitting]". ..... Then ..... add to this child a history that she is a survivor, albeit exceptionally so, of incurring the pain and largely successful battle for separation from a very narcissistic mother and the family that produced that narcissism in her mother. The entire repercussions of that I am not sure, but relevant here is I think that means my child's developmental reality has a biased understanding and emotional sensitivity to the fear that a parent "I thought was normal, changed into a monster" and second "I fully believed a truth about the 1 of 2 people I trusted and depend on the most, and I was wrong. How can I trust my own conclusions now if I can't trust my own analytical and emotional judgement abilities?". No doubt also a fear and anxiety upregulating mechanism in and if itself, as well as providing a data point which can add confusion to a child frantically looking for understanding and/or can be leveraged to falsely rationalize the false narrative is correct especially when the pain of the truth is building and she is looking for any tool to suppress confronting that pain.

Then, as Rhyanna further looks for, or rather it is imposed onto her, the naive drama thirsty peer group, whom many know Liam and Kate, and whom with very good intention but naivety of teenagers who in Boulder Colorado are conditioned to both be very helpful and that money and wealth (like them) combined with middle aged Caucasian combined with a "Boulderite" personality with an air of non-confrontational superiority and cancel-culture tendencies is the equivalent of "insightful, wise, holder of truth, and generally the definition of what is good, righteous, and hold the authority to declare whom is bad and further that it is expected that they will declare whom is good and bad and that action further validates that they are and have such authorities" in these teenagers minds reenforces this false truth as accurate.. Then the school, then CIRT team "mental health professionals", then the mental health hospital centennial peaks, then Boulder county child welfare via multiple staff, then the court and the judge personally all buy in and propagate this false truth and reinforce it overtly or indirectly overtly, and some propagate it by simply ignoring and not speaking out against or in questioning validity, all reinforcing this false truth. ..... And given all this, given all these goddamn ignorant spineless children of men in their lack of knowledge or past traumas, and under the weight of their ignorance and cowardice and laziness, and then under the unreal weight and fear and confusion of her and her dad, her one parent who's been her warrior defender of knowledge, self discovery, safety, character, food, and shelter, and whom no other family support exists is now very possibly dying and cannot speak for himself or to her (because her confusion and outside influence is not allowing it) to tell her the truth and reassurance of the situation ....... her heart and mind refuse to yield. The pain from her heart refusing to give way to the lie, they are trying to make her believe had caused her to want to kill herself. My daughter s unyielding heart and character brought tears to a police officer who'd not had the fortune of experiencing someone like my daughter. And still, after a year and a half, my daughter, MY daughter, still holds fast and is unwilling to tell the COURT that her resistance is because of me and is instead because of her. Yeah, that's who my daughter is. That is the caliber here. She is her father's daughter.

I see you kid. You hold fast. I'm comin' for you.

PS - Attention needs to be given to Liam. With consideration towards his possible and to what degree of trauma, and the validity of the story regarding his father.. It is now a real question, is his father above and well, normative, searching for his son and or fallen into decline, suicidality, doom? Is Liam about to lose a father and be irreversibly severely damaged because of the complete irreversible devastation, which will also include the self blame he incurs and will not be able to reconcile.

PSS - likely it is both important and the is the time to revisit with focus Rhyannas feelings and understanding of her mom. She possibly stands to gain 1, a self confidence and esteem and complete obliteration of any feeling/false rationalization that she is somehow "less", that she is at fault, or that she is somehow "less capable" of a person now and going forward, 2) stamp out reactions of hate, tolerance, splitting, and walls she might form that would prevent problem solving, truth finding, and understanding so crucial to both abilities and finding of joy, particularly in relationships of love and family, 3) she stands to gain a mother and an entire side of a family and which is attained by a fulfilling relationship of her own architecture and which she is fully empowered to control and manage and nurture at her pleasure.

I think this the reason why Karl Popper proposed an alternative approach based on falsification rather than confirmation. because when we saying that all swans are white, confirming this statement would involve finding as many white swans as possible However, with Popper's approach it shows that the statement is scientific only if there is a way to show it is false. In this case, finding a single black swan would falsify the statement). Therefore, we can also say that confirmation alone cannot provide certainty or proof of a theory's validity.

This underlies the ideas of songlines and oral mnemonic practices and is related to Vannevar Bush's "associative trails" in As We May Think.

Luhmann, Niklas. “ZK II Zettel 9/8b.” Niklas Luhmann-Archiv, undated. https://niklas-luhmann-archiv.de/bestand/zettelkasten/zettel/ZK_2_NB_9-8b_V.

I believe the focus of the reading is happiness can't be fully observed from the outside due to the lack self understanding we may have pertaining to happiness. It's hard for a man to be happy when they're so many different outlooks on what happiness should look like.

It seems lie Freud is trying to say that thirst for technological advancements have birthed new problems related to human existence. When you look at it from that perspective it does paint a rather bleak picture. However I do not think it's quite that simple. A dilemma such this may never have a clear cut answer.

I don't know why but this made me think about fate vs free as we may make all the right choices but somethings may not fall in our favor.

People put so much focus on social comparison. I think people get tunnel vision on trying to find a group to fit into, rather than find a group that fits them. Both are important in the right context. Just as it's important to step out of your comfort zone for new people, it's just as important to seek out people with shared interests and hobbies.

Though history may seem trivial at times, I think that the real purpose of learning history is how we learn what works, as well as what we should never do again. Our goal should be to learn from our mistakes as a society. It also tells us so much about culture and in that we can also learn about our future.

Author Response

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

eLife assessment

This important study advances our understanding of the ways in which different types of communication signals differentially affect mouse behaviors and amygdala cholinergic/dopaminergic neuromodulation. Researchers interested in the complex interaction between prior experience, sex, behavior, hormonal status, and neuromodulation should benefit from this study. Nevertheless, the data analysis is incomplete at this stage, requiring additional analysis and description, justification, and - potentially - power to support the conclusions fully. With the analytical part strengthened, this paper will be of interest to neuroscientists and ethologists.

GENERAL COMMENTS ON REVIEWS AND REVISIONS

Experimental design

Here we address questions from several reviewers regarding our periods of neuromodulator and behavioral analysis. First, we recognize that the text would benefit from an overview of the experimental structure different from the narrative we provide in the first paragraphs of the Results. We now include this near the beginning for the Materials and Methods (page 17). We further articulate that the 10-minute time periods were dictated by the sampling duration required to perform accurate neurochemical analyses (and to reserve half of the sample in the event of a catastrophic failure of batch-processing samples). Since neurochemical release may display multiple temporal components (e.g., ACh: Aitta-aho et al., 2018) during playback stimulation, and since these could differ across neurochemicals of interest, we decided to collect, analyze, and report in two stimulus periods as well as one Pre-Stim control. We now clarify this in additional text in the Material and Methods (p. 24, lines 20-22; p. 26, lines 17-19). We decided not to include analyses of the post-stimulus period because this is subject to wider individual and neuromodulator-specific effects and because it weakens statistical power in addressing the core question—the change in neuromodulator release DURING vocal playback.

We also sought to clarify the meaning of the periods “Stim 1” and “Stim 2”; they are two data collection periods, using the same examplar sequences in the same order. We have added statements in the Material and Methods (p. 18, lines 4-7; Fig. caption, p. 39, lines 11-13) to clarify these periods.

For behavioral analyses, observation periods were much shorter than 10 mins, but the main purpose of behavioral analyses in this report is to relate to the neurochemical data. As a result, we matched the temporal features of the behavioral and neurochemical analyses (p. 22, lines 17-22). We plan a separate report, focused exclusively on a broader set of behavioral responses to playback, that may examine behaviors at a more granular level.

Data and statistical analyses

Reviewers 1 and 3 expressed concerns about our normalization of neurochemical data, suggesting that it diminishes statistical power or is not transparent. We note that normalization is a very common form of data transformation that does not diminish statistical power. It is particularly useful for data forms in which the absolute value of the measurement across experiments may be uninformative. Normalization is routine in microdialysis studies, because data can be affected by probe placement and factors affecting neurochemical recovery and processing. Recent examples include:

Li, Chaoqun, Tianping Sun, Yimu Zhang, Yan Gao, Zhou Sun, Wei Li, Heping Cheng, Yu Gu, and Nashat Abumaria. "A neural circuit for regulating a behavioral switch in response to prolonged uncontrollability in mice." Neuron (2023).

Gálvez-Márquez, Donovan K., Mildred Salgado-Ménez, Perla Moreno-Castilla, Luis Rodríguez-Durán, Martha L. Escobar, Fatuel Tecuapetla, and Federico Bermudez-Rattoni. "Spatial contextual recognition memory updating is modulated by dopamine release in the dorsal hippocampus from the locus coeruleus." Proceedings of the National Academy of Sciences 119, no. 49 (2022): e2208254119.

Holly, Elizabeth N., Christopher O. Boyson, Sandra Montagud-Romero, Dirson J. Stein, Kyle L. Gobrogge, Joseph F. DeBold, and Klaus A. Miczek. "Episodic social stress-escalated cocaine self-administration: role of phasic and tonic corticotropin releasing factor in the anterior and posterior ventral tegmental area." Journal of Neuroscience 36, no. 14 (2016): 4093-4105.

Bagley, Elena E., Jennifer Hacker, Vladimir I. Chefer, Christophe Mallet, Gavan P. McNally, Billy CH Chieng, Julie Perroud, Toni S. Shippenberg, and MacDonald J. Christie. "Drug-induced GABA transporter currents enhance GABA release to induce opioid withdrawal behaviors." Nature neuroscience 14, no. 12 (2011): 1548-1554.

However, since all reviewers requested raw values of neurochemicals, we provide these in supplementary tables 1-3. The manuscript references these table early in the Results (p. 6, lines 18-19) and in the Material and Methods (p. 27, lines 3-4)

All reviewers commented on correlation analyses that we presented, with different perspectives. Reviewer 2 questioned the validity of such analyses, performed across experimental groups, while Reviewer 1 pointed out that the analyses were redundant with the GLM. We agree with these criticisms, and note the challenges associated with correlations involving behaviors for which there is a “floor” in the number of observations. As a result, we have removed most correlation analyses from the manuscript. The text and figures have been modified accordingly. Due these changes, we have to decline requests of Reviewer 3 to include many more such analyses. While correlation analyses could still be performed between neurochemicals and behaviors for each group, the relatively small size of each experimental group, the large number of groups, and the even larger numbers of pairings between neurochemicals and behavior, the statistical power is very low. The only correlations we utilize in the manuscript concern the interpretation of our increased acetylcholine levels.

As part of this revision, we re-ran our statistical analyses on neuromodulators because of a calculation error in 3 animals (regarding baseline values). In a few instances, a significance level changed, but none of these changed a conclusion regarding neuromodulator changes under our experimental conditions.

Other revisions

INTRODUCTION: We modified the Introduction to provide both a more general framework and specific gaps in our understanding relating neuromodulators with vocal communication.

DISCUSSION: We have added material in the first two pages of the Discussion to provide more framework to our conclusions, to address the issues of the temporal aspects of neurochemical release and behavioral observations, and to identify limitations that should be addressed in future studies.

FIGURES: All figures are now in the main part of the manuscript. We modified most figures in response to reviewer comments. We removed neuromodulator – behavior correlations from several figures. We modified all box plots to ensure that all data points are visible. The visible data points match the numbers reported in figure captions. We brought 5-HIAA data into the main figures reporting on neuromodulator results.

Public Reviews:

Reviewer #1 (Public Review):

The manuscript addresses a fundamental question about how different types of communication signals differentially affect brain states and neurochemistry. In addition, the manuscript highlights the various processes that modulate brain responses to communication signals, including prior experience, sex, and hormonal status. Overall, the manuscript is well-written and the research is appropriately contextualized. The authors are thoughtful about their quantitative approaches and interpretations of the data.

That being said, the authors need to work on justifying some of their analytical approaches (e.g., normalization of neurochemical data, dividing the experimental period into two periods (as opposed to just analyzing the entire experimental period as a whole)) and should provide a greater discussion of how their data also demonstrate dissociations between neurochemical release in the basolateral amygdala and behavior (e.g., neurochemical differences during both of the experimental periods but behavioral differences only during the first half of the experimental period). The normalization of neurochemical data seems unnecessary given the repeated-measures design of their analysis and could be problematic; by normalizing all data to the baseline data (p. 24), one artificially creates a baseline period with minimal variation (all are "0"; Figures 2, 3 & 5) that could inflate statistical power.

Please see our general responses to structure of observation periods and normalization of neuromodulator data. Normalization is a common and appropriate procedure in microdialysis studies that does not alter statistical power.

We have included a section in the Discussion concerning the temporal relationship between behavioral responses and neurochemical changes in response to vocal playback (p. 12, lines 3-17). We note where the linkage is particularly strong (e.g., ACh release and flinching). This points to a need to examine these phenomena with finer temporal resolution, but also with the recognition that the brain circuits driving a behavioral response may extend beyond the BLA.

The Introduction could benefit from a priori predictions about the differential release of specific neuromodulators based on previous literature.

We added some material to the Introduction to provide additional rationale for the study. However, we did not attempt to develop predictions for the range of neuromodulators that we sought to test. The literature can lead to opposite predictions for a given neuromodulator. For example, acetylcholine could be associated with both positive and negative valence. Instead, we note in the Introduction the association of both DA and ACh with vocalizations.

The manuscript would also benefit from a description of space use and locomotion in response to different valence vocalizations.

We have provided additional descriptions of space use and video tracking data in Material and Methods (p. 23, lines 1-6). We now report a few correlations based on these data in the Results to demonstrate that increased ACh in Restraint males and Mating estrus females was not related to the amount of locomotion (p. 9, lines 8-14).

Nevertheless, the current manuscript seems to provide some compelling support for how positive and negative valence vocalizations differentially affect behavior and the release of acetylcholine and dopamine in the basolateral amygdala. The research is relevant to broad fields of neuroscience and has implications for the neural circuits underlying social behavior.

Reviewer #2 (Public Review):

Ghasemahmad et al. report findings on the influence of salient vocalization playback, sex, and previous experience, on mice behaviors, and on cholinergic and dopaminergic neuromodulation within the basolateral amygdala (BLA). Specifically, the authors played back mice vocalizations recorded during two behaviors of opposite valence (mating and restraint) and measured the behaviors and release of acetylcholine (ACh), dopamine (DA), and serotonin in the BLA triggered in response to those sounds.

Strength: The authors identified that mating and restraint sounds have a differential impact on cholinergic and dopaminergic release. In male mice, these two distinct vocalizations exert an opposite effect on the release of ACh and DA. Mating sounds elicited a decrease of Ach release and an increase of DA release. Conversely, restraint sounds induced an increase in ACh release and a trend to decrease in DA. These neurotransmission changes were different in estrus females for whom the mating vocalization resulted in an increase of both DA and ACh release.

Weaknesses: The behavioral analysis and results remain elusive, and although addressing interesting questions, the study contains major flaws, and the interpretations are overstating the findings.

Although Reviewer 2 raises several valid issues that we have addressed in our response and revision, we believe that none represent “major flaws” in the study that challenge the validity of our central conclusions. In brief, we will:

--provide enhanced description of behaviors (pp. 22-23 and Table 1)

--clarify / modify box-plot representations of data (p 28. Lines 3-9)

--point to our methods that describe corrections for multiple comparisons (p. 27; lines 15-16)

--revise figures to clarify sample size (Figs. 3-6)

Reviewer #3 (Public Review):

Ghasemahmad et al. examined behavioral and neurochemical responses of male and female mice to vocalizations associated with mating and restraint. The authors made two significant and exciting discoveries. They revealed that the affective content of vocalizations modulated both behavioral responses and the release of acetylcholine (ACh) and dopamine (DA) but not serotonin (5-HIAA) in the basolateral amygdala (BLA) of male and female mice. Moreover, the results show sex-based differences in behavioral responses to vocalizations associated with mating. The authors conclude that behavior and neurochemical responses in male and female mice are experience-dependent and are altered by vocalizations associated with restraint and mating. The findings suggest that ACh and DA release may shape behavioral responses to context-dependent vocalizations. The study has the potential to significantly advance our understanding of how neuromodulators provide internal-state signals to the BLA while an animal listens to social vocalizations; however, multiple concerns must be addressed to substantiate their conclusions.

Major concerns:

1) The authors normalized all neurochemical data to the background level obtained from a single pre-stimulus sample immediately preceding playback. The percentage change from the background level was calculated based on a formula, and the underlying concentrations were not reported. The authors should report the sample and background concentrations to make the results and analyses more transparent. The authors stated that NE and 5-HT had low recovery from the mouse brain and hence could not be tracked in the experiment. The authors could be more specific here by relating the concentrations to ACh, DA, and 5-HIAA included in the analyses.

Please see our general statement regarding normalization of neurochemical data. We have added supplemental tables that shows concentrations of dopamine, acetylcholine, 5-HIAA. We do not report serotonin or noradrenalin since these were below the detection threshold.

2) For the EXP group, the authors stated that each animal underwent 90-min sessions on two consecutive days that provided mating and restraint experiences. Did the authors record mating or copulation during these experiments? If yes, what was the frequency of copulation? What other behaviors were recorded during these experiences? Did the experiment encompass other courtship behaviors along with mating experiences? Was the female mouse in estrus during the experience sessions?

In the mating experience, mounting or attempted mounting was required for the animal to be included in subsequent testing. Since the session lasted 90 minutes, more general courtship behavior was likely. However, we did not record detailed behaviors or track estrous stage for the mating experience. See p. 21, line 20-22.

3) For the mating playback, the authors stated that the mating stimulus blocks contained five exemplars of vocal sequences emitted during mating interactions. The authors should clarify whether the vocal sequences were emitted while animals were mating/copulating or when the male and female mice were inside the test box. If the latter was the case, it might be better to call the playback "courtship playback" instead of "mating playback".

We have modified the Results (p. 5, lines 18-20) and Materials and Methods (p. 21, lines 8-15) to clarify our meaning. We continue to use the term “mating” because this refers to a specific set of behaviors associated with mounting and copulation, rather than the more general term “courtship”. We also indicate that we based these behaviors on previous work (e.g., Gaub et al., 2016).

4) Since most differences that the authors reported in Figure 3 were observed in Stim 1 and not in Stim 2, it might be better to perform a temporal analysis - looking at behaviors and neurochemicals over time instead of dividing them into two 10-minute bins. The temporal analysis will provide a more accurate representation of changes in behavior and neurochemicals over time.

Please see our general response to the structuring of experimental periods. The 10-min periods are the minimum for the neurochemical analyses, and we adopted the same periods for behavioral analyses to match the two types of observations. Our repeated measures analysis is a form of temporal analysis, since it compares values in three observation periods.

5) In Figures 2 and 3, the authors show the correlation between Flinching behavior and ACh concentration. The authors should report correlations between concentrations of all neurochemicals (not just ACh) and all behaviors recorded (not just Flinching), even if they are insignificant. The analyses performed for the stim 1 data should also be performed on the stim 2 data. Reporting these findings would benefit the field.

Please see general comments regarding correlation analyses. We removed almost all such analyses and references to them from the manuscript based on concerns of the other reviewers.

6) The mice used in the study were between p90 - p180. The mice were old, and the range of ages was considerable. Are the findings correlated with age? The authors should also discuss how age might affect the experiment's results.

Our p90-p180 mice are not “old”. CBA/CaJ mice display normal hearing for at least 1 year (Ohlemiller, Dahl, and Gagnon, JARO 11: 605-623, 2010) and adult sexual and social behavior throughout our observation period. They are sexually mature adults, appropriate for this study. We decline to perform correlation analyses with age, both because this was not a question for this study and because the very large number of correlations, for each experimental group (as requested by reviewer #2), render this approach statistically problematic.

7) The authors reported neurochemical levels estimated as the animals listened to the sounds played back. What about the sustained effects of changes in neurochemicals? Are there any potential long-term effects of social vocalizations on behavior and neurochemical levels? The authors might consider discussing long-term effects.

We have not included discussion of long term effects of neuromodulatory release, both because our data analysis doesn’t address it (see response to Comment #10) and because we desired to keep the Discussion focused on topics more closely related to the results.

8) Histology from a single recording was shown in supplementary figure 1. It would benefit the readers if additional histology was shown for all the animals, not just the colored schematics summarizing the recording probe locations. Further explanation of the track location is also needed to help the readers. Make it clear for the readers which dextran-fluorescein labeling image is associated with which track in the schematic.

Based on the recent publications cited in our overall response to reviewer comments about statistical methods, our reporting of histological location of microdialysis exceeds the standard. We believe that the inclusion of all histology is unnecessary and not particularly helpful. Raw photomicrographs do not always illustrate boundaries, so interpretation is required. However, we added a second photomicrograph example and we identified which tracks correspond to these photomicrographs (see Figure 2; now in main body of manuscript).

9) The authors did not control for the sounds being played back with a speaker. This control may be necessary since the effects are more pronounced in Stim 1 than in Stim 2. Playing white noise rather than restraint or courtship vocalizations would be an excellent control. However, the authors could perform a permutation analysis and computationally break the relationship between what sound is playing and the neurochemical data. This control would allow the authors to show that the actual neurochemical levels are above or below chance.

We considered a potential “control” stimulus in our experimental design. We concluded, based on our previous work (e.g., Grimsley et al., 2013; Gadziola et al., 2016), that white noise is not or not necessarily a neutral stimulus and therefore the results would not clarify the responses to the two vocal stimuli. Instead, we opted to use experience as a type of control. This control shows very clearly that temporal patterns and across-group differences in neurochemical response to playback disappear in the absence of experience with the associated behavior.

10) The authors indicated that each animal's post-vocalization session was also recorded. No data in the manuscript related to the post-vocalization playback period was included. This omission was a missed opportunity to show that the neurochemical levels returned to baseline, and the results were not dependent on the normalization process described in major concern #1. The data should be included in the manuscript and analyzed. It would add further support for the model described in Figure 6.

We decided not to include analyses of the post-stimulus period because this period is subject to wider individual and neuromodulator-specific effects and because it weakens statistical power in addressing the core question—the change in neuromodulator release DURING vocal playback. We agree that the general question is of interest to the field, but we don’t think our study is best designed to answer that question.

11) The authors could use a predictive model, such as a binary classifier trained on the CSF sampling data, to predict the type of vocalizations played back. The predictive model could support the conclusions and provide additional support for the model in Figure 6.

We recognize that a binary classifier could provide an interesting approach to support conclusions. However, we do not believe that the sample size per group is sufficient to both create and test the classifier.

Reviewer #1 (Recommendations For The Authors):

Major comments:

• Introduction: It would be useful to set up an experimental framework before delving into the results. What are the predictions about specific neuromodulators based on previous literature?

Because this narrative is laid out in the first two paragraphs of the Results, which immediately follow the Introduction, we believe that additional text in the Introduction on the experimental framework is redundant. As stated above, detailing predictions for a range of neuromodulators would make for a long and not particularly illuminating Introduction. We instead have related our findings to more general understanding of DA and ACh in the Discussion.

• There really isn't a major difference in stimuli during the "Stim 1" and "Stim 2" phases, and it's not clear why the authors divided the experimental period into two phases. Therefore, the authors need to justify their experimental approach. For example, the authors could first anecdotally mention that behavioral responses to playbacks seem to be larger in the first half of the playbacks than during the second half, therefore they individually analyzed each half of the experimental period. Or adopt a different approach to justify their design. Overall, the analytical approach is reasonable but it is currently not justified.

See general comment for analysis periods. As noted, we clarified these issues in several locations with Materials and Methods (pp. 24, lines 20-22; p. 26, lines 17-19). We also sought to clarify the meaning of the periods “Stim 1” and “Stim 2”; they are two data collection periods, using the same examplar sequences in the same order. We have added statements in the Material and Methods (p. 18, lines 4-7; Fig. caption, p. 39, lines 11-13).

• The normalization of neurochemical data seems problematic and unnecessary. By normalizing all data to the baseline data (p. 24), one artificially creates a baseline period with minimal variation (all are "0"; Figures 2, 3 & 5) and this has implications for statistical power. Because the analysis is a within-subjects analysis, this normalization is not necessary for the analysis itself. It can be useful to normalize data for visualization purposes, but raw data should be analyzed. Indeed, behavioral data are qualitatively similar to the neurochemical data, and those data are not normalized to baseline values.

Please see our general comment on this issue. We believe normalization does not affect statistical power and is both the standard way and an appropriate way to analyze microdialysis results. We include concentrations of ACh, DA, and 5-HIAA in supplementary tables?

• The authors should include a discussion (in the Discussion section) of how behavior and neurochemical release are associated during the first half of the experimental session but not in the second half (e.g., differences in Ach and DA release between mating and restraint groups during stim 1 and 2, but behavioral differences only during stim 1).

We have included a section in the Discussion concerning the temporal relationship between behavioral responses and neurochemical changes in response to vocal playback. We note that the linkage is particularly strong in some cases (e.g., ACh release and flinching). This points to a need to examine these phenomena with finer temporal resolution, but also with the recognition that the brain circuits driving a behavioral response may extend beyond the BLA.

Minor comments:

• Keywords: add "serotonin" (even though there are no significant differences on 5-HIAA, people interested in serotonin would find this interesting).

Added to keywords list.

• Do the authors collect data on the vocalizations of mice in response to these playbacks?

We monitored vocalizations during playback, noting that vocalizations–especially “Noisy” vocalization–were common. However, we did not record vocalizations and are therefore unable quantify our observations.

• First line of page 7: readers do not know about "stim 1" and "stim 2". Therefore, the authors need to describe their approach to analyzing behavior and neurochemical release.

We first introduce these terms earlier, citing Figure 1D,E. We have added some additional wording for further clarification. page 7, lines 4-5.

• Make sure citations are uniformly formatted (e.g., Inconsistencies in: "As male and female mice emit different vocalizations during mating (Finton et al., 2017; J. M. S. Grimsley et al., 2013; Neunuebel et al., 2015; Sales (née Sewell), 1972)").

We have reviewed and corrected citations throughout the manuscript.

• Last paragraph of page 7: "attending behavior" has not been defined yet.

Table 1 contains our description of the behaviors analyzed in this study. We have now inserted a reference to Table 1 earlier in the Results (p. 6, line 12).

• Figure 2E and 3G: I find these correlations to be redundant with the GLMs. This is because the significant relationship is likely to be driven by group differences in behavior and in neurochemical release.

Please see general comments regarding correlation analyses. We removed such analyses and references to them from the manuscript.

• Page 2, 2nd paragraph, 2nd sentence: this paragraph seems to be rooted in comparing and contrasting experienced and inexperienced mice, so there should be explicit comparisons in each sentence. For example, the 2nd sentence should read: "Whereas EXP estrus females demonstrated increased flinching behaviors in response to mating vocalizations, INEXP ....". This paragraph overall could use some refining.

We believe this refers to page 9. We have revised the paragraph to clarify our findings (Beginning p. 9, line 23).

• Page 9: "Further, there were no significant differences across groups during Stim 1 or Stim 2 periods. These results contrast sharply with those from all EXP groups, in which both ACh and DA release changed significantly during playback (Figs. 2C, 2D, 3E, 3F)." While I understand their perspective, this is misleading because changes were only observed during the Stim 1 period.

We have slightly revised the wording in this paragraph, because the restraint males did not show significant ACh decreases. However, we do not believe our statements mislead readers just because some changes are observed in only one of the stimulation periods (p 10, lines 13-16).

• Last paragraph of page 14: it would be useful to mention the increase in flinching in experienced females in response to mating vocalizations.

We have added a sentence in this paragraph relating flinching in estrus females to increased ACh (p. 15, lines 18-20).

• Was there a full analysis of locomotion in response to playbacks? I see that locomotion was correlated with neurochemical release but was it different in response to different stimuli? Were there changes to the part of the arena that mice occupied in response to restraint vs. mating vocalizations? Given their methods section, it would be useful for the authors to mention the results of the analyses of these aspects of movement.

We have provided additional descriptions of space use and video tracking data in Material and Methods (p. 23, lines 1-6). We now report additional results associated with these analyses (p. 8, lines 13-15; p. 9, lines 8-14).

• I believe that each experimental mouse only heard one of the stimuli (given the analytical approach). Because it is plausible to measure neurochemical release in response to both types of stimuli, I encourage the authors to be more explicit about this aspect of the experimental design (e.g., mention in Results section).

Sentence modified to read: “Each mouse received playback of either the mating or restraint stimuli, but not both: same-day presentation of both stimuli would require excessively long playback sessions, the condition of the same probe would likely change on subsequent days, and quality of a second implanted probe on a subsequent day was uncertain.” (p. 7, lines 5-9).

• Figure 1A and 1B: add labels to the panels so readers don't have to read the legend to know what spectrogram is associated with what context.

We added these labels to Figure 1.

• Table 1: in the definition of "still and alert", should this mention "abrupt attending" instead of "abrupt freezing"? The latter isn't described.

Yes, we intended “abrupt attending”, and now indicated that in Table 1

Reviewer #2 (Recommendations For The Authors):

Major comments:

• The authors report they performed manual behavioral analysis, and provide a table defining the different behaviors. However, it remains unclear how some of these behaviors were detected (such as still-and-alert events). A thorough description of the criteria used to define these events needs to be provided.

We have modified some descriptions of manually analyzed behaviors in Table 1, and have added additional description of how we developed this set of behaviors for analysis in the study (pp. 22-23).

• The box plots do not appear to represent the "minimum, first quartile, median, third quartile, and maximum values." as specified on page 24 (Methods). Indeed, the individual data points sometimes do not reach the max or min of the bar plot, and sometimes are way beyond them.

We used the “inclusive median” function in Excel to generate final boxplots. These boxplots will sometimes result in a data point being placed outside of the whiskers. SPSS considers these to be “outliers”, but our GLM analysis includes these values. We describe this in Data Analysis section of Materials and Methods (p. 28, lines 3-9)

• Some of the data are replicated in different Figures: Figure 2A and Figure 3C. While this is acceptable, the authors did not correct for multiple comparisons (dividing the p value by the number of comparisons).

Our analysis included corrections for multiple comparisons, as we have indicated on p. 27, lines 15-16.

• Overall, the sample sizes are too small (for example in Figure 3, non-estrus females are at n=3), and are different in experiments where they should be equal (Figure 2B: mating stim 1 is at n=5 and mating stim 2 is at n=3).

We apologize that sample sizes were not properly displayed in figures. Please note that sample sizes are identified in the figure captions. For neuromodulator data, all sample sizes are at least 7. For behavioral data, the minimum sample size is 5. We have revised Figures 3-6 to ensure that all data points are visible.

• It remains unclear why the impact of mating vocalizations has been tested only in males.

We assume the reviewer meant that only males were tested in restraint. We now indicate that our preliminary evidence indicated no difference in behavioral responses to restraint vocalization between males and females, so we opted to perform the neurochemical analysis for restraint only in males (page 22 lines 4-5). If there were no limitations to time and cost, we would have preferred to test responses to restraint in females as well. We note that such inclusion would have added up to 4 experimental groups (estrus and non-estrus groups in both EXP and INEXP groups).

• The correlation between the number of flinching and ACh release changes (Figure 2E) visually appears to be opposite between mating and restraint playbacks. The authors should perform independent correlations for these 2 playbacks.

Please see general comments regarding correlation analyses. We removed such analyses and references to them from the manuscript.

• The authors state that their findings "indicate that behavioral responses to salient vocalizations result from interactions between sex of the listener or context of vocal stimuli with the previous behavioral experience associated with these vocalizations.". However, in male mice, they do not report any difference in previous experience on flinching for both restraint and mating sounds, as well as no difference in rearing for the restrain sounds (Figure 4A-B). Thus, the discussion of these results should be completely revisited.

We revised the paragraph in question (p. 9, line 22 through p. 10, line 9). For instance, we note that significant differences between EXP male-mating and male-restraint flinching do not exist between the INEXP groups. We believe that the last sentence correctly summarizes findings described in this paragraph.

• For serotonin experiments in Figure S2 there are strong outliers (150% increase in 5HIAA release). Did the authors correlate these levels with the behavior of the animals?

Outliers are identified by the Excel function that generated the boxplots, but we have no reason to consider these as outliers and exclude them. As noted above, we have clarified that these “outliers” are the result of the Excel function in the Materials and Methods (p. 28, lines 3-9) and we have revised the plotting of data points

Minor comments:

• Mating vocalization playback is mainly emitted by males, thus, instead of a positive valence signal, this could also be interpreted as a competitive signal to other males.

There is support in the literature for viewing our mating stimulus as having positive valence. Gaub et al., 2016 describe the emission of stepped calls, lower frequency harmonics, and increased sound level as indicators of “positive emotion”. We have shown (Grimsley et al, 2013) that the female LFH vocalization can be highly attractive to male mice, under the right conditions, indicating something like “sex is happening”. The inclusion of both the male and female vocalizations in our stimuli was a key piece of our experimental design, based on our understanding of the contributions of both vocalizations to the meaning of the overall acoustic experience.

• Figure 1 should include panel titles.

No change. This information is available in the Figure caption.

• n=31 should be indicated in the EXP group.

We’re not sure where the reviewer is referring to this value.

• The color legend of Figure 1E is absent, making the Figure not understandable.

We added text in the Figure 1 caption to indicate that each color represents a different exemplar. We don’t think a legend provides additional useful information.

• The point of making two blocks (stim 1 and stim2) should be stated more clearly.

Please see general statement regarding experimental blocks. We have modified our description of these in an Experimental overview section in the Material and Methods.

• Including raw data of micro-dialysis in the supplementary figures would allow assessment of the variability and quality of the measurements.

We have added concentrations of neurochemicals in supplemental tables 1-3.

• Baseline (prestimulus) number of flinch and rearing should systematically be indicated (missing in Figure 4).

The focus in this figure is on the differences that occur in Stim 1 values. There are no differences between EXP and INEXP animals of any group during the Pre-Stim period. We now state that in the Figure 4 caption.

• Discussion: "increase in AMPA/NMDA currents". We believe the authors are referring to the ratio of AMPA to NMDA currents. This sentence should be reformulated.

These are modified to refer to “… the AMPA/NMDA current ratio…” in two locations in the Discussion (p. 14, lines 8-9; p. 15, line 4)

• Overall the discussion is very speculative and should rely more on the data.

We believe that the Discussion provides appropriate speculation that is based on our experimental data and previous literature. We have added a paragraph to identify limitations of our findings and recommendations of future experiments to resolve some issues (p. 12, lines 3-17)

Reviewer #3 (Recommendations For The Authors):

Minor concerns:

1) The authors stated that USVs are most likely to be emitted by males, and LFH are likely to be emitted by females. However, Oliveira-Stahl et al. 2023, Matsumoto et al. 2022, Warren et al. 2018, Heckman et al. 2017, Neunuebel et al., 2015 showed that females also emit USVs. The authors should mention that USVs are emitted by both males and females and discuss how the sex of the vocalizing animal (both males and females) can influence neuromodulator release.

The reviewer slightly mis-stated the wording of our text, changing the meaning significantly. Our wording is “These sequences included ultrasonic vocalizations (USVs) with harmonics, steps, and complex structure, mostly emitted by males, and low frequency harmonic calls (LFHs) emitted by females (Fig. 1A,C)…” This phrasing is correct and carefully chosen. The Discussion in Oliveira-Stahl et al 2023 (p. 10-11) supports our statement: “The exact fraction of USVs emitted by females as concluded in all previous studies on dyadic courtship has varied, ranging from 18%, 17.5%, and 16% to 10.5% in the present study…”.

2) The authors should explain why ECF from BLA was collected unilaterally from the left hemisphere.

p. 23, lines 9-11: We inserted a sentence to explain why we targeted the BLA unilaterally. “Since both left and right amygdala are responsive to vocal stimuli in human and experimental animal studies (Wenstrup et al., 2020), we implanted microdialysis probes into the left amygdala to maintain consistency with other studies in our laboratory..” Beyond that, the choice was arbitrary.

3) The authors said each animal recovered in its home cage for four days before the playback experiment. A 4-day period may not be sufficient for every animal to recover from surgery, so the authors should describe how a mouse's recovery was assessed.

p. 23, lines 20-23: We provide more description about the recovery and how it was assessed. Except for a few animals that were not included in the experiments, all animals recovered within 4 days.

4) The authors stated that each animal was exposed to 90-min sessions with mating and restraint behaviors in a counterbalanced design. This description for Figure 1D should also include the duration of the mating and restraint experience.

The Results that immediately precede citation to this figure include this information.

5) The authors stated, "Data are reported only from mice with more than 75% of the microdialysis probe implanted within the BLA". What are the implications of having 25% of the probe outside the BLA? The authors should shed more light on this by discussing this issue as it relates to the findings and commenting on where the other 25% of the probe was located.

We inserted a sentence to explain the rationale for this inclusion criterion. “We verified placement of microdialysis probes to minimize variability that could arise because regions surrounding BLA receive neurochemical inputs from different sources (e.g., cholinergic inputs to putamen and central amygdala).” (p. 25, lines 21-23).

All brain regions that surround BLA, dorsal, medial, ventral, or lateral, could have been sampled by the “other” 25%. Some of these, e.g., the central amygdala or caudate-putamen, have different sources of cholinergic input that may not have the same release pattern. We do not think it is worthy of further speculation in the Discussion. Due to the high cost of the neurochemical analysis, we often did not process the neurochemistry data if histology indicated that a probe missed the BLA target.

6) The authors confirmed that the estrus stage did not change during the experiment day by evaluating and comparing estrus prior to and after data collection. This strategy was a fantastic experimental approach, but the authors should have discussed the results. How did the results the authors included change when the females were in estrus before but not after data collection? What percentage of females started in estrus but ended in metestrus? Assuming that some females changed estrus state, were these animals excluded from the analyses?

All animals were in the same estrus state at the beginning and end of the playback session.

7). Authors cite Neunuebel et al., 2015 for the sentence "As male and female mice emit different vocalizations during mating". However, Neunuebel et al., 2015 showed vocalizations emitted during chasing--not mating. If mating is a general term for courtship, then this reference is appropriate, but see major concern #3.

In the Results (p. 8, line 5), we changed the phrasing to “courtship and mating” to include the Neunubel et al study.

As we indicate in our response to Public Comment #3, we have modified the Results (p. 5, lines 18-20) and Materials and Methods (p. 21, lines 8-15) to clarify our meaning. We continue to use the term “mating” because this refers to a specific set of behaviors associated with mounting and copulation, rather than the more general term “courtship”. We also indicate that we based these behaviors on previous work (e.g., Gaub et al., 2016).

8) Authors interpret Figure 3F as DA release showed a "consistent" increase during mating playback across all three experimental groups. However, the increase in the estrus female group is inconsistent, as seen in the graph. This verbiage should be reworded to describe the data more accurately.

p. 8, line 23 “consistent” was deleted.

9) In all the box plots, multiple data points overlay each other. A more transparent way of showing the data would be adding some jitter to the x value to make each data point visible. The mean (X's) in Figure 3D (pre-stim mating and mating estrus) are difficult to see, as are all the data points in mating non-estrus. Adding all the symbols to the figure legend or a key in the figure instead of the method section would aid the reader and make the plots easier to interpret

We have revised the boxplots to ensure that all data points are visible.

10) Some verbiage used in the discussion should be toned down. For example, "intense" experiences and "emotionally charged" vocalizations should be removed.

We have not changed these terms, which we believe are appropriate to describe these experiences and vocalizations.

11) The authors include "Emotional Vocalizations" in the title. It would be beneficial if the authors included more detail and references in the introduction to help set up the emotional content of vocalizations. It may benefit a broader readership as typically targeted by eLife.

We now cite Darwin and some more recent publications that articulate the general understanding that social vocalizations carry emotional content.

Author Response

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

eLife assessment

This study presents potentially valuable results on glutamine-rich motifs in relation to protein expression and alternative genetic codes. The author's interpretation of the results is so far only supported by incomplete evidence, due to a lack of acknowledgment of alternative explanations, missing controls and statistical analysis and writing unclear to non experts in the field. These shortcomings could be at least partially overcome by additional experiments, thorough rewriting, or both.

We thank both the Reviewing Editor and Senior Editor for handling this manuscript.

Based on your suggestions, we have provided controls, performed statistical analysis, and rewrote our manuscript. The revised manuscript is significantly improved and more accessible to non-experts in the field.

Reviewer #1 (Public Review):

Summary

This work contains 3 sections. The first section describes how protein domains with SQ motifs can increase the abundance of a lacZ reporter in yeast. The authors call this phenomenon autonomous protein expression-enhancing activity, and this finding is well supported. The authors show evidence that this increase in protein abundance and enzymatic activity is not due to changes in plasmid copy number or mRNA abundance, and that this phenomenon is not affected by mutants in translational quality control. It was not completely clear whether the increased protein abundance is due to increased translation or to increased protein stability.

In section 2, the authors performed mutagenesis of three N-terminal domains to study how protein sequence changes protein stability and enzymatic activity of the fusions. These data are very interesting, but this section needs more interpretation. It is not clear if the effect is due to the number of S/T/Q/N amino acids or due to the number of phosphorylation sites.

In section 3, the authors undertake an extensive computational analysis of amino acid runs in 27 species. Many aspects of this section are fascinating to an expert reader. They identify regions with poly-X tracks. These data were not normalized correctly: I think that a null expectation for how often poly-X track occur should be built for each species based on the underlying prevalence of amino acids in that species. As a result, I believe that the claim is not well supported by the data.

Strengths

This work is about an interesting topic and contains stimulating bioinformatics analysis. The first two sections, where the authors investigate how S/T/Q/N abundance modulates protein expression level, is well supported by the data. The bioinformatics analysis of Q abundance in ciliate proteomes is fascinating. There are some ciliates that have repurposed stop codons to code for Q. The authors find that in these proteomes, Q-runs are greatly expanded. They offer interesting speculations on how this expansion might impact protein function.

Weakness

At this time, the manuscript is disorganized and difficult to read. An expert in the field, who will not be distracted by the disorganization, will find some very interesting results included. In particular, the order of the introduction does not match the rest of the paper.

In the first and second sections, where the authors investigate how S/T/Q/N abundance modulates protein expression levels, it is unclear if the effect is due to the number of phosphorylation sites or the number of S/T/Q/N residues.

There are three reasons why the number of phosphorylation sites in the Q-rich motifs is not relevant to their autonomous protein expression-enhancing (PEE) activities:

First, we have reported previously that phosphorylation-defective Rad51-NTD (Rad51-3SA) and wild-type Rad51-NTD exhibit similar autonomous PEE activity. Mec1/Tel1-dependent phosphorylation of Rad51-NTD antagonizes the proteasomal degradation pathway, increasing the half-life of Rad51 from ∼30 min to ≥180 min (1). (page 1, lines 11-14)

Second, in our preprint manuscript, we have already shown that phosphorylation-defective Rad53-SCD1 (Rad51-SCD1-5STA) also exhibits autonomous PEE activity similar to that of wild-type Rad53-SCD (Figure 2D, Figure 4A and Figure 4C). We have highlighted this point in our revised manuscript (page 9, lines 19-21).

Third, as revealed by the results of Figure 4, it is the percentages, and not the numbers, of S/T/Q/N residues that are correlated with the PEE activities of Q-rich motifs.

The authors also do not discuss if the N-end rule for protein stability applies to the lacZ reporter or the fusion proteins.

The autonomous PEE function of S/T/Q-rich NTDs is unlikely to be relevant to the N-end rule. The N-end rule links the in vivo half-life of a protein to the identity of its N-terminal residues. In S. cerevisiae, the N-end rule operates as part of the ubiquitin system and comprises two pathways. First, the Arg/N-end rule pathway, involving a single N-terminal amidohydrolase Nta1, mediates deamidation of N-terminal asparagine (N) and glutamine (Q) into aspartate (D) and glutamate (E), which in turn are arginylated by a single Ate1 R-transferase, generating the Arg/N degron. N-terminal R and other primary degrons are recognized by a single N-recognin Ubr1 in concert with ubiquitin-conjugating Ubc2/Rad6. Ubr1 can also recognize several other N-terminal residues, including lysine (K), histidine (H), phenylalanine (F), tryptophan (W), leucine (L) and isoleucine (I) (68-70). Second, the Ac/N-end rule pathway targets proteins containing N-terminally acetylated (Ac) residues. Prior to acetylation, the first amino acid methionine (M) is catalytically removed by Met-aminopeptidases (MetAPs), unless a residue at position 2 is non-permissive (too large) for MetAPs. If a retained N-terminal M or otherwise a valine (V), cysteine (C), alanine (A), serine (S) or threonine (T) residue is followed by residues that allow N-terminal acetylation, the proteins containing these AcN degrons are targeted for ubiquitylation and proteasome-mediated degradation by the Doa10 E3 ligase (71).

The PEE activities of these S/T/Q-rich domains are unlikely to arise from counteracting the N-end rule for two reasons. First, the first two amino acid residues of Rad51-NTD, Hop1-SCD, Rad53-SCD1, Sup35-PND, Rad51-ΔN, and LacZ-NVH are MS, ME, ME, MS, ME, and MI, respectively, where M is methionine, S is serine, E is glutamic acid and I is isoleucine. Second, Sml1-NTD behaves similarly to these N-terminal fusion tags, despite its methionine and glutamine (MQ) amino acid signature at the N-terminus. (Page 12, line 3 to page 13, line 2)

The most interesting part of the paper is an exploration of S/T/Q/N-rich regions and other repetitive AA runs in 27 proteomes, particularly ciliates. However, this analysis is missing a critical control that makes it nearly impossible to evaluate the importance of the findings. The authors find the abundance of different amino acid runs in various proteomes. They also report the background abundance of each amino acid. They do not use this background abundance to normalize the runs of amino acids to create a null expectation from each proteome. For example, it has been clear for some time (Ruff, 2017; Ruff et al., 2016) that Drosophila contains a very high background of Q's in the proteome and it is necessary to control for this background abundance when finding runs of Q's.

We apologize for not explaining sufficiently well the topic eliciting this reviewer’s concern in our preprint manuscript. In the second paragraph of page 14, we cite six references to highlight that SCDs are overrepresented in yeast and human proteins involved in several biological processes (5, 43) and that polyX prevalence differs among species (79-82).

We will cite a reference by Kiersten M. Ruff in our revised manuscript (38).

K. M. Ruff, J. B. Warner, A. Posey and P. S. Tan (2017) Polyglutamine length dependent structural properties and phase behavior of huntingtin exon1. Biophysical Journal 112, 511a.

The authors could easily address this problem with the data and analysis they have already collected. However, at this time, without this normalization, I am hesitant to trust the lists of proteins with long runs of amino acid and the ensuing GO enrichment analysis. Ruff KM. 2017. Washington University in St.

Ruff KM, Holehouse AS, Richardson MGO, Pappu RV. 2016. Proteomic and Biophysical Analysis of Polar Tracts. Biophys J 110:556a.

We thank Reviewer #1 for this helpful suggestion and now address this issue by means of a different approach described below.

Based on a previous study (43), we applied seven different thresholds to seek both short and long, as well as pure and impure, polyX strings in 20 different representative near-complete proteomes, including 4X (4/4), 5X (4/5-5/5), 6X (4/6-6/6), 7X (4/7-7/7), 8-10X (≥50%X), 11-10X (≥50%X) and ≥21X (≥50%X).

To normalize the runs of amino acids and create a null expectation from each proteome, we determined the ratios of the overall number of X residues for each of the seven polyX motifs relative to those in the entire proteome of each species, respectively. The results of four different polyX motifs are shown in our revised manuscript, i.e., polyQ (Figure 7), polyN (Figure 8), polyS (Figure 9) and polyT (Figure 10). Thus, polyX prevalence differs among species and the overall X contents of polyX motifs often but not always correlate with the X usage frequency in entire proteomes (43).

Most importantly, our results reveal that, compared to Stentor coeruleus or several non-ciliate eukaryotic organisms (e.g., Plasmodium falciparum, Caenorhabditis elegans, Danio rerio, Mus musculus and Homo sapiens), the five ciliates with reassigned TAAQ and TAGQ codons not only have higher Q usage frequencies, but also more polyQ motifs in their proteomes (Figure 7). In contrast, polyQ motifs prevail in Candida albicans, Candida tropicalis, Dictyostelium discoideum, Chlamydomonas reinhardtii, Drosophila melanogaster and Aedes aegypti, though the Q usage frequencies in their entire proteomes are not significantly higher than those of other eukaryotes (Figure 1). Due to their higher N usage frequencies, Dictyostelium discoideum, Plasmodium falciparum and Pseudocohnilembus persalinus have more polyN motifs than the other 23 eukaryotes we examined here (Figure 8). Generally speaking, all 26 eukaryotes we assessed have similar S usage frequencies and percentages of S contents in polyS motifs (Figure 9). Among these 26 eukaryotes, Dictyostelium discoideum possesses many more polyT motifs, though its T usage frequency is similar to that of the other 25 eukaryotes (Figure 10).

In conclusion, these new normalized results confirm that the reassignment of stop codons to Q indeed results in both higher Q usage frequencies and more polyQ motifs in ciliates.  

Reviewer #2 (Public Review):

Summary:

This study seeks to understand the connection between protein sequence and function in disordered regions enriched in polar amino acids (specifically Q, N, S and T). While the authors suggest that specific motifs facilitate protein-enhancing activities, their findings are correlative, and the evidence is incomplete. Similarly, the authors propose that the re-assignment of stop codons to glutamine-encoding codons underlies the greater user of glutamine in a subset of ciliates, but again, the conclusions here are, at best, correlative. The authors perform extensive bioinformatic analysis, with detailed (albeit somewhat ad hoc) discussion on a number of proteins. Overall, the results presented here are interesting, but are unable to exclude competing hypotheses.

Strengths:

Following up on previous work, the authors wish to uncover a mechanism associated with poly-Q and SCD motifs explaining proposed protein expression-enhancing activities. They note that these motifs often occur IDRs and hypothesize that structural plasticity could be capitalized upon as a mechanism of diversification in evolution. To investigate this further, they employ bioinformatics to investigate the sequence features of proteomes of 27 eukaryotes. They deepen their sequence space exploration uncovering sub-phylum-specific features associated with species in which a stop-codon substitution has occurred. The authors propose this stop-codon substitution underlies an expansion of ploy-Q repeats and increased glutamine distribution.

Weaknesses:

The preprint provides extensive, detailed, and entirely unnecessary background information throughout, hampering reading and making it difficult to understand the ideas being proposed.

The introduction provides a large amount of detailed background that appears entirely irrelevant for the paper. Many places detailed discussions on specific proteins that are likely of interest to the authors occur, yet without context, this does not enhance the paper for the reader.

The paper uses many unnecessary, new, or redefined acronyms which makes reading difficult. As examples:

1) Prion forming domains (PFDs). Do the authors mean prion-like domains (PLDs), an established term with an empirical definition from the PLAAC algorithm? If yes, they should say this. If not, they must define what a prion-forming domain is formally.

The N-terminal domain (1-123 amino acids) of S. cerevisiae Sup35 was already referred to as a “prion forming domain (PFD)” in 2006 (48). Since then, PFD has also been employed as an acronym in other yeast prion papers (Cox, B.S. et al. 2007; Toombs, T. et al. 2011).

B. S. Cox, L. Byrne, M. F., Tuite, Protein Stability. Prion 1, 170-178 (2007). J. A. Toombs, N. M. Liss, K. R. Cobble, Z. Ben-Musa, E. D. Ross, [PSI+] maintenance is dependent on the composition, not primary sequence, of the oligopeptide repeat domain. PLoS One 6, e21953 (2011).

2) SCD is already an acronym in the IDP field (meaning sequence charge decoration) - the authors should avoid this as their chosen acronym for Serine(S) / threonine (T)-glutamine (Q) cluster domains. Moreover, do we really need another acronym here (we do not).

SCD was first used in 2005 as an acronym for the Serine (S)/threonine (T)-glutamine (Q) cluster domain in the DNA damage checkpoint field (4). Almost a decade later, SCD became an acronym for “sequence charge decoration” (Sawle, L. et al. 2015; Firman, T. et al. 2018).

L. Sawle and K, Ghosh, A theoretical method to compute sequence dependent configurational properties in charged polymers and proteins. J. Chem Phys. 143, 085101(2015).

T. Firman and Ghosh, K. Sequence charge decoration dictates coil-globule transition in intrinsically disordered proteins. J. Chem Phys. 148, 123305 (2018).

3) Protein expression-enhancing (PEE) - just say expression-enhancing, there is no need for an acronym here.

Thank you. Since we have shown that the addition of Q-rich motifs to LacZ affects protein expression rather than transcription, we think it is better to use the “PEE” acronym.

The results suggest autonomous protein expression-enhancing activities of regions of multiple proteins containing Q-rich and SCD motifs. Their definition of expression-enhancing activities is vague and the evidence they provide to support the claim is weak. While their previous work may support their claim with more evidence, it should be explained in more detail. The assay they choose is a fusion reporter measuring beta-galactosidase activity and tracking expression levels. Given the presented data they have shown that they can drive the expression of their reporters and that beta gal remains active, in addition to the increase in expression of fusion reporter during the stress response. They have not detailed what their control and mock treatment is, which makes complete understanding of their experimental approach difficult. Furthermore, their nuclear localization signal on the tag could be influencing the degradation kinetics or sequestering the reporter, leading to its accumulation and the appearance of enhanced expression. Their evidence refuting ubiquitin-mediated degradation does not have a convincing control.

Although this reviewer’s concern regarding our use of a nuclear localization signal on the tag is understandable, we are confident that this signal does not bias our findings for two reasons. First, the negative control LacZ-NV also possesses the same nuclear localization signal (Figure 1A, lane 2). Second, another fusion target, Rad51-ΔN, does not harbor the NVH tag (Figure 1D, lanes 3-4). Compared to wild-type Rad51, Rad51-ΔN is highly labile. In our previous study, removal of the NTD from Rad51 reduced by ~97% the protein levels of corresponding Rad51-ΔN proteins relative to wild-type (1).

Based on the experimental results, the authors then go on to perform bioinformatic analysis of SCD proteins and polyX proteins. Unfortunately, there is no clear hypothesis for what is being tested; there is a vague sense of investigating polyX/SCD regions, but I did not find the connection between the first and section compelling (especially given polar-rich regions have been shown to engage in many different functions). As such, this bioinformatic analysis largely presents as many lists of percentages without any meaningful interpretation. The bioinformatics analysis lacks any kind of rigorous statistical tests, making it difficult to evaluate the conclusions drawn. The methods section is severely lacking. Specifically, many of the methods require the reader to read many other papers. While referencing prior work is of course, important, the authors should ensure the methods in this paper provide the details needed to allow a reader to evaluate the work being presented. As it stands, this is not the case.

Thank you. As described in detail below, we have now performed rigorous statistical testing using the GofuncR package (Figure 11, Figure 12 and DS7-DS32).

Overall, my major concern with this work is that the authors make two central claims in this paper (as per the Discussion). The authors claim that Q-rich motifs enhance protein expression. The implication here is that Q-rich motif IDRs are special, but this is not tested. As such, they cannot exclude the competing hypothesis ("N-terminal disordered regions enhance expression").

In fact, “N-terminal disordered regions enhance expression” exactly summarizes our hypothesis.

On pages 12-13 and Figure 4 of our preprint manuscript, we explained our hypothesis in the paragraph entitled “The relationship between PEE function, amino acid contents, and structural flexibility”.

The authors also do not explore the possibility that this effect is in part/entirely driven by mRNA-level effects (see Verma Na Comms 2019).

As pointed out by the first reviewer, we present evidence that the increase in protein abundance and enzymatic activity is not due to changes in plasmid copy number or mRNA abundance (Figure 2), and that this phenomenon is not affected in translational quality control mutants (Figure 3).

As such, while these observations are interesting, they feel preliminary and, in my opinion, cannot be used to draw hard conclusions on how N-terminal IDR sequence features influence protein expression. This does not mean the authors are necessarily wrong, but from the data presented here, I do not believe strong conclusions can be drawn. That re-assignment of stop codons to Q increases proteome-wide Q usage. I was unable to understand what result led the authors to this conclusion.

My reading of the results is that a subset of ciliates has re-assigned UAA and UAG from the stop codon to Q. Those ciliates have more polyQ-containing proteins. However, they also have more polyN-containing proteins and proteins enriched in S/T-Q clusters. Surely if this were a stop-codon-dependent effect, we'd ONLY see an enhancement in Q-richness, not a corresponding enhancement in all polar-rich IDR frequencies? It seems the better working hypothesis is that free-floating climate proteomes are enriched in polar amino acids compared to sessile ciliates.

We thank this reviewer for raising this point, however her/his comments are not supported by the results in Figure 7.

Regardless, the absence of any kind of statistical analysis makes it hard to draw strong conclusions here.

We apologize for not explaining more clearly the results of Tables 5-7 in our preprint manuscript.

To address the concerns about our GO enrichment analysis by both reviewers, we have now performed rigorous statistical testing for SCD and polyQ protein overrepresentation using the GOfuncR package (https://bioconductor.org/packages/release/bioc/html/GOfuncR.html). GOfuncR is an R package program that conducts standard candidate vs. background enrichment analysis by means of the hypergeometric test. We then adjusted the raw p-values according to the Family-wise error rate (FWER). The same method had been applied to GO enrichment analysis of human genomes (89).

The results presented in Figure 11 and Figure 12 (DS7-DS32) support our hypothesis that Q-rich motifs prevail in proteins involved in specialized biological processes, including Saccharomyces cerevisiae RNA-mediated transposition, Candida albicans filamentous growth, peptidyl-glutamic acid modification in ciliates with reassigned stop codons (TAAQ and TAGQ), Tetrahymena thermophila xylan catabolism, Dictyostelium discoideum sexual reproduction, Plasmodium falciparum infection, as well as the nervous systems of Drosophila melanogaster, Mus musculus, and Homo sapiens (78). In contrast, peptidyl-glutamic acid modification and microtubule-based movement are not overrepresented with Q-rich proteins in Stentor coeruleus, a ciliate with standard stop codons.

Recommendations for the authors:

Please note that you control which revisions to undertake from the public reviews and recommendations for the authors.

Reviewer #1 (Recommendations For The Authors):

The order of paragraphs in the introduction was very difficult to follow. Each paragraph was clear and easy to understand, but the order of paragraphs did not make sense to this reader. The order of events in the abstract matches the order of events in the results section. However, the order of paragraphs in the introduction is completely different and this was very confusing. This disordered list of facts might make sense to an expert reader but makes it hard for a non-expert reader to understand.

Apologies. We endeavored to improve the flow of our revised manuscript to make it more readable.

The section beginning on pg 12 focused on figures 4 and 5 was very interesting and highly promising. However, it was initially hard for me to tell from the main text what the experiment was. Please add to the text an explanation of the experiment, because it is hard to figure out what was going on from the figures alone. Figure 4 is fantastic, but would be improved by adding error bars and scaling the x-axis to be the same in panels B,C,D.

Thank you for this recommendation. We have now scaled both the x-axis and y-axis equivalently in panels B, C and D of Figure 4. Error bars are too small to be included.

It is hard to tell if the key variable is the number of S/T/Q/N residues or the number of phosphosites. I think a good control would be to add a regression against the number of putative phosphosites. The sequences are well designed. I loved this part but as a reader, I need more interpretation about why it matters and how it explains the PEE.

As described above, we have shown that the number of phosphorylation sites in the Q-rich motifs is not relevant to their autonomous protein expression-enhancing (PEE) activities.

I believe that the prevalence of polyX runs is not meaningful without normalizing for the background abundance of each amino acid. The proteome-wide abundance and the assumption that amino acids occur independently can be used to form a baseline expectation for which runs are longer than expected by chance. I think Figures 6 and 7 should go into the supplement and be replaced in the main text with a figure where Figure 6 is normalized by Figure 7. For example in P. falciparum, there are many N-runs (Figure 6), but the proteome has the highest fraction of N’s (Figure 7).

Thank you for these suggestions. The three figures in our preprint manuscript (Figures 6-8) have been moved into the supplementary information (Figures S1-S3). For normalization, we have provided four new figures (Figures 7-10) in our revised manuscript.

The analysis of ciliate proteomes was fascinating. I am particularly interested in the GO enrichment for “peptidyl-glutamic acid modification” (pg 20) because these enzymes might be modifying some of Q’s in the Q-runs. I might be wrong about this idea or confused about the chemistry. Do these ciliates live in Q-rich environments? Or nitrogen rich environments?

Polymeric modifications (polymodifications) are a hallmark of C-terminal tubulin tails, whereas secondary peptide chains of glutamic acids (polyglutamylation) and glycines (polyglycylation) are catalyzed from the γ-carboxyl group of primary chain glutamic acids. It is not clear if these enzymes can modify some of the Q’s in the Q-runs.

To our knowledge, ciliates are abundant in almost every liquid water environment, i.e., oceans/seas, marine sediments, lakes, ponds, and rivers, and even soils.

I think you should include more discussion about how the codons that code for Q’s are prone to slippage during DNA replication, and thus many Q-runs are unstable and expand (e.g. Huntington’s Disease). The end of pg 24 or pg 25 would be good places.

We thank the reviewer for these comments.

PolyQ motifs have a particular length-dependent codon usage that relates to strand slippage in CAG/CTG trinucleotide repeat regions during DNA replication. In most organisms having standard genetic codons, Q is encoded by CAGQ and CAAQ. Here, we have determined and compared proteome-wide Q contents, as well as the CAGQ usage frequencies (i.e., the ratio between CAGQ and the sum of CAGQ, CAGQ, TAAQ, and TAGQ).

Our results reveal that the likelihood of forming long CAG/CTG trinucleotide repeats are higher in five eukaryotes due to their higher CAGQ usage frequencies, including Drosophila melanogaster (86.6% Q), Danio rerio (74.0% Q), Mus musculus (74.0% Q), Homo sapiens (73.5% Q), and Chlamydomonas reinhardtii (87.3% Q) (orange background, Table 2). In contrast, another five eukaryotes that possess high numbers of polyQ motifs (i.e., Dictyostelium discoideum, Candida albicans, Candida tropicalis, Plasmodium falciparum and Stentor coeruleus) (Figure 1) utilize more CAAQ (96.2%, 84.6%, 84.5%, 86.7% and 75.7%) than CAAQ (3.8%, 15.4%, 15.5%, 13.3% and 24.3%), respectively, to avoid the formation of long CAG/CTG trinucleotide repeats (green background, Table 2). Similarly, all five ciliates with reassigned stop codons (TAAQ and TAGQ) have low CAGQ usage frequencies (i.e., from 3.8% Q in Pseudocohnilembus persalinus to 12.6% Q in Oxytricha trifallax) (red font, Table 2). Accordingly, the CAG-slippage mechanism might operate more frequently in Chlamydomonas reinhardtii, Drosophila melanogaster, Danio rerio, Mus musculus and Homo sapiens than in Dictyostelium discoideum, Candida albicans, Candida tropicalis, Plasmodium falciparum, Stentor coeruleus and the five ciliates with reassigned stop codons (TAAQ and TAGQ).

Author response table 1.

Usage frequencies of TAA, TAG, TAAQ, TAGQ, CAAQ and CAGQ codons in the entire proteomes of 20 different organisms.

Pg 7, paragraph 2 has no direction. Please add the conclusion of the paragraph to the first sentence.

This paragraph has been moved to the “Introduction” section” of the revised manuscript.

Pg 8, I suggest only mentioning the PFDs used in the experiments. The rest are distracting.

We have addressed this concern above.

Pg 12. Please revise the "The relationship...." text to explain the experiment.

We apologize for not explaining this topic sufficiently well in our preprint manuscript.

SCDs are often structurally flexible sequences (4) or even IDRs. Using IUPred2A (https://iupred2a.elte.hu/plot_new), a web-server for identifying disordered protein regions (88), we found that Rad51-NTD (1-66 a.a.) (1), Rad53-SCD1 (1-29 a.a.) and Sup35-NPD (1-39 a.a.) are highly structurally flexible. Since a high content of serine (S), threonine (T), glutamine (Q), asparanine (N) is a common feature of IDRs (17-20), we applied alanine scanning mutagenesis approach to reduce the percentages of S, T, Q or N in Rad51-NTD, Rad53-SCD1 or Sup35-NPD, respectively. As shown in Figure 4 and Figure 5, there is a very strong positive relationship between STQ and STQN amino acid percentages and β-galactosidase activities. (Page 13, lines 5-10)

Pg 13, first full paragraph, "Futionally, IDRs..." I think this paragraph belongs in the Discussion.

This paragraph is now in the “Introduction” section (Page 5, Lines 11-15).

Pg. 15, I think the order of paragraphs should be swapped.

These paragraphs have been removed or rewritten in the “Introduction section” of our revised manuscript.

Pg 17 (and other parts) I found the lists of numbers and percentages hard to read and I think you should refer readers to the tables.

Thank you. In the revised manuscript, we have avoided using lists of numbers and percentages, unless we feel they are absolutely essential.

Pg. 19 please add more interpretation to the last paragraph. It is very cool but I need help understanding the result. Are these proteins diverging rapidly? Perhaps this is a place to include the idea of codon slippage during DNA replication.

Thank you. The new results in Table 2 indicate that the CAG-slippage mechanism is unlikely to operate in ciliates with reassigned stop codons (TAAQ and TAGQ).

Pg 24. "Based on our findings from this study, we suggest that Q-rich motifs are useful toolkits for generating novel diversity during protein evolution, including by enabling greater protein expression, protein-protein interactions, posttranslational modifications, increased solubility, and tunable stability, among other important traits." This idea needs to be cited. Keith Dunker has written extensively about this idea as have others. Perhaps also discuss why Poly Q rich regions are different from other IDRs and different from other IDRs that phase-separate.

Agreed, we have cited two of Keith Dunker’s papers in our revised manuscript (73, 74).

Minor notes:

Please define Borg genomes (pg 25).

Borgs are long extrachromosomal DNA sequences in methane-oxidizing Methanoperedens archaea, which display the potential to augment methane oxidation (101). They are now described in our revised manuscript. (Page 15, lines 12-14)

Reviewer #2 (Recommendations For The Authors):

The authors dance around disorder but never really quantify or show data. This seems like a strange blindspot.

We apologize for not explaining this topic sufficiently well in our preprint manuscript. We have endeavored to do so in our revised manuscript.

The authors claim the expression enhancement is "autonomous," but they have not ruled things out that would make it not autonomous.

Evidence of the “autonomous” nature of expression enhancement is presented in Figure 1, Figure 4, and Figure 5 of the preprint manuscript.

Recommendations for improving the writing and presentation.

The title does not recapitulate the entire body of work. The first 5 figures are not represented by the title in any way, and indeed, I have serious misgivings as to whether the conclusion stated in the title is supported by the work. I would strongly suggest the authors change the title.

Figure 2 could be supplemental.

Thank you. We think it is important to keep Figure 2 in the text.

Figures 4 and 5 are not discussed much or particularly well.

This reviewer’s opinion of Figure 4 and Figure 5 is in stark contrast to those of the first reviewer.

The introduction, while very thorough, takes away from the main findings of the paper. It is more suited to a review and not a tailored set of minimal information necessary to set up the question and findings of the paper. The question that the authors are after is also not very clear.

Thank you. The entire “Introduction” section has been extensively rewritten in the revised manuscript.

Schematics of their fusion constructs and changes to the sequence would be nice, even if supplemental.

Schematics of the fusion constructs are provided in Figure 1A.

The methods section should be substantially expanded.

The method section in the revised manuscript has been rewritten and expanded. The six Javascript programs used in this work are listed in Table S4.

The text is not always suited to the general audience and readership of eLife.

We have now rewritten parts of our manuscript to make it more accessible to the broad readership of eLife.

In some cases, section headers really don't match what is presented, or there is no evidence to back the claim.

The section headers in the revised manuscript have been corrected.

A lot of the listed results in the back half of the paper could be a supplemental table, listing %s in a paragraph (several of them in a row) is never nice

Acknowledged. In the revised manuscript, we have removed almost all sentences listing %s.

Minor corrections to the text and figures.

There is a reference to table 1 multiple times, and it seems that there is a missing table. The current table 1 does not seem to be the same table referred to in some places throughout the text.

Apologies for this mistake, which we have now corrected in our revised manuscript.

In some places its not clear where new work is and where previous work is mentioned. It would help if the authors clearly stated "In previous work...."

Acknowledged. We have corrected this oversight in our revised manuscript.

Not all strains are listed in the strain table (KO's in figure 3 are not included)

Apologies, we have now corrected Table S2, as suggested by this reviewer.

Author response table 2.

S. cerevisiae strains used in this study

The section provides a profound exploration of the complexities involved in understanding and evaluating disruptive behaviors in social media contexts. It compellingly illustrates how the formation of groups, the use of stereotypes, and the enforcement of norms are all deeply intertwined with our cognitive processes and societal structures. The examination of trolling as a form of disruption that can be deployed for both just and unjust ends invites readers to reflect on the multifaceted nature of these actions and their ethical implications.

Reviewer #1 (Public Review):

This valuable study demonstrates a novel mechanism by which implicit motor adaptation saturates for large visual errors in a principled normative Bayesian manner. Additionally, the study revealed two notable empirical findings: visual uncertainty increases for larger visual errors in the periphery, and proprioceptive shifts/implicit motor adaptation are non-monotonic, rather than ramp-like. This study is highly relevant for researchers in sensory cue integration and motor learning. However, I find some areas where statistical quantification is incomplete, and the contextualization of previous studies to be puzzling.

Issue #1: Contextualization of past studies.

While I agree that previous studies have focused on how sensory errors drive motor adaptation (e.g., Burge et al., 2008; Wei and Kording, 2009), I don't think the PReMo model was contextualized properly. Indeed, while PReMo should have adopted clearer language - given that proprioception (sensory) and kinaesthesia (perception) have been used interchangeably, something we now make clear in our new study (Tsay, Chandy, et al. 2023) - PReMo's central contribution is that a perceptual error drives implicit adaptation (see Abstract): the mismatch between the felt (perceived) and desired hand position. The current paper overlooks this contribution. I encourage the authors to contextualize PReMo's contribution more clearly throughout. Not mentioned in the current study, for example, PReMo accounts for the continuous changes in perceived hand position in Figure 4 (Figure 7 in the PReMo study).

There is no doubt that the current study provides important additional constraints on what determines perceived hand position: Firstly, it offers a normative Bayesian perspective in determining perceived hand position. PReMo suggests that perceived hand position is determined by integrating motor predictions with proprioception, then adding a proprioceptive shift; PEA formulates this as the optimal integration of these three inputs. Secondly, PReMo assumed visual uncertainty to remain constant for different visual errors; PEA suggests that visual uncertainty ought to increase (but see Issue #2).

Issue #2: Failed replication of previous results on the effect of visual uncertainty.

2a. A key finding of this paper is that visual uncertainty linearly increases in the periphery; a constraint crucial for explaining the non-monotonicity in implicit adaptation. One notable methodological deviation from previous studies is the requirement to fixate on the target: Notably, in the current experiments, participants were asked to fixate on the target, a constraint not imposed in previous studies. In a free-viewing environment, visual uncertainty may not attenuate as fast, and hence, implicit adaptation does not attenuate as quickly as that revealed in the current design with larger visual errors. Seems like this current fixation design, while important, needs to be properly contextualized considering how it may not represent most implicit adaptation experiments.

2b. Moreover, the current results - visual uncertainty attenuates implicit adaptation in response to large, but not small, visual errors - deviates from several past studies that have shown that visual uncertainty attenuates implicit adaptation to small, but not large, visual errors (Tsay, Avraham, et al. 2021; Makino, Hayashi, and Nozaki, n.d.; Shyr and Joshi 2023). What do the authors attribute this empirical difference to? Would this free-viewing environment also result in the opposite pattern in the effect of visual uncertainty on implicit adaptation for small and large visual errors?

2c. In the current study, the measure of visual uncertainty might be inflated by brief presentation times of comparison and referent visual stimuli (only 150 ms; our previous study allowed for a 500 ms viewing time to make sure participants see the comparison stimuli). Relatedly, there are some individuals whose visual uncertainty is greater than 20 degrees standard deviation. This seems very large, and less likely in a free-viewing environment.

2d. One important confound between clear and uncertain (blurred) visual conditions is the number of cursors on the screen. The number of cursors may have an attenuating effect on implicit adaptation simply due to task-irrelevant attentional demands (Parvin et al. 2022), rather than that of visual uncertainty. Could the authors provide a figure showing these blurred stimuli (gaussian clouds) in the context of the experimental paradigm? Note that we addressed this confound in the past by comparing participants with and without low vision, where only one visual cursor is provided for both groups (Tsay, Tan, et al. 2023).

Issue #3: More methodological details are needed.

3a. It's unclear why, in Figure 4, PEA predicts an overshoot in terms of perceived hand position from the target. In PReMo, we specified a visual shift in the perceived target position, shifted towards the adapted hand position, which may result in overshooting of the perceived hand position with this target position. This visual shift phenomenon has been discovered in previous studies (e.g., (Simani, McGuire, and Sabes 2007)).

3b. The extent of implicit adaptation in Experiment 2, especially with smaller errors, is unclear. The implicit adaptation function seems to be still increasing, at least by visual inspection. Can the authors comment on this trend, and relatedly, show individual data points that help the reader appreciate the variability inherent to these data?

3c. The same participants were asked to return for multiple days/experiments. Given that the authors acknowledge potential session effects, with attenuation upon re-exposure to the same rotation (Avraham et al. 2021), how does re-exposure affect the current results? Could the authors provide clarity, perhaps a table, to show shared participants between experiments and provide evidence showing how session order may not be impacting results?

3d. The number of trials per experiment should be detailed more clearly in the Methods section (e.g., Exp 4). Moreover, could the authors please provide relevant code on how they implemented their computational models? This would aid in future implementation of these models in future work. I, for one, am enthusiastic to build on PEA.

3f. In addition to predicting a correlation between proprioceptive shift and implicit adaptation on a group level, both PReMo and PEA (but not causal inference) predict a correlation between individual differences in proprioceptive shift and proprioceptive uncertainty with the extent of implicit adaptation (Tsay, Kim, et al. 2021). Interestingly, shift and uncertainty are independent (see Figures 4F and 6C in Tsay et al, 2021). Does PEA also predict independence between shift and uncertainty? It seems like PEA does predict a correlation.

References:

Avraham, Guy, Ryan Morehead, Hyosub E. Kim, and Richard B. Ivry. 2021. "Reexposure to a Sensorimotor Perturbation Produces Opposite Effects on Explicit and Implicit Learning Processes." PLoS Biology 19 (3): e3001147.<br /> Makino, Yuto, Takuji Hayashi, and Daichi Nozaki. n.d. "Divisively Normalized Neuronal Processing of Uncertain Visual Feedback for Visuomotor Learning."<br /> Parvin, Darius E., Kristy V. Dang, Alissa R. Stover, Richard B. Ivry, and J. Ryan Morehead. 2022. "Implicit Adaptation Is Modulated by the Relevance of Feedback." BioRxiv. https://doi.org/10.1101/2022.01.19.476924.<br /> Shyr, Megan C., and Sanjay S. Joshi. 2023. "A Case Study of the Validity of Web-Based Visuomotor Rotation Experiments." Journal of Cognitive Neuroscience, October, 1-24.<br /> Simani, M. C., L. M. M. McGuire, and P. N. Sabes. 2007. "Visual-Shift Adaptation Is Composed of Separable Sensory and Task-Dependent Effects." Journal of Neurophysiology 98 (5): 2827-41.<br /> Tsay, Jonathan S., Guy Avraham, Hyosub E. Kim, Darius E. Parvin, Zixuan Wang, and Richard B. Ivry. 2021. "The Effect of Visual Uncertainty on Implicit Motor Adaptation." Journal of Neurophysiology 125 (1): 12-22.<br /> Tsay, Jonathan S., Anisha M. Chandy, Romeo Chua, R. Chris Miall, Jonathan Cole, Alessandro Farnè, Richard B. Ivry, and Fabrice R. Sarlegna. 2023. "Implicit Motor Adaptation and Perceived Hand Position without Proprioception: A Kinesthetic Error May Be Derived from Efferent Signals." BioRxiv. https://doi.org/10.1101/2023.01.19.524726.<br /> Tsay, Jonathan S., Hyosub E. Kim, Darius E. Parvin, Alissa R. Stover, and Richard B. Ivry. 2021. "Individual Differences in Proprioception Predict the Extent of Implicit Sensorimotor Adaptation." Journal of Neurophysiology, March. https://doi.org/10.1152/jn.00585.2020.<br /> Tsay, Jonathan S., Steven Tan, Marlena Chu, Richard B. Ivry, and Emily A. Cooper. 2023. "Low Vision Impairs Implicit Sensorimotor Adaptation in Response to Small Errors, but Not Large Errors." Journal of Cognitive Neuroscience, January, 1-13.

Reviewer #3 (Public Review):

Summary<br /> In this paper, the authors model motor adaptation as a Bayesian process that combines visual uncertainty about the error feedback, uncertainty about proprioceptive sense of hand position, and uncertainty of predicted (=planned) hand movement with a learning and retention rate as used in state space models. The model is built with results from several experiments presented in the paper and is compared with the PReMo model (Tsay, Kim, et al., 2022) as well as a cue combination model (Wei & Körding, 2009). The model and experiments demonstrate the role of visual uncertainty about error feedback in implicit adaptation.

In the introduction, the authors notice that implicit adaptation (as measured in error-clamp-based paradigms) does not saturate at larger perturbations, but decreases again (e.g. Moorehead et al., 2017 shows no adaptation at 135{degree sign} and 175{degree sign} perturbations). They hypothesized that visual uncertainty about cursor position increases with larger perturbations since the cursor is further from the fixated target. This could decrease the importance assigned to visual feedback which could explain lower asymptotes.

The authors characterize visual uncertainty for 3 rotation sizes in the first experiment, and while this experiment could be improved, it is probably sufficient for the current purposes. Then the authors present a second experiment where adaptation to 7 clamped errors is tested in different groups of participants. The models' visual uncertainty is set using a linear fit to the results from experiment 1, and the remaining 4 parameters are then fit to this second data set. The 4 parameters are 1) proprioceptive uncertainty, 2) uncertainty about the predicted hand position, 3) a learning rate, and 4) a retention rate. The authors' Perceptual Error Adaptation model ("PEA") predicts asymptotic levels of implicit adaptation much better than both the PReMo model (Tsay, Kim et al., 2022), which predicts saturated asymptotes, or a causal inference model (Wei & Körding, 2007) which predicts no adaptation for larger rotations. In a third experiment, the authors test their model's predictions about proprioceptive recalibration, but unfortunately, compare their data with an unsuitable other data set. Finally, the authors conduct a fourth experiment where they put their model to the test. They measure implicit adaptation with increased visual uncertainty, by adding blur to the cursor, and the results are again better in line with their model (predicting overall lower adaptation) than with the PReMo model (predicting equal saturation but at larger perturbations) or a causal inference model (predicting equal peak adaptation, but shifted to larger rotations). In particular, the model fits experiment 2 and the results from experiment 4 show that the core idea of the model has merit: increased visual uncertainty about errors dampens implicit adaptation.

Strengths<br /> In this study, the authors propose a Perceptual Error Adaptation model ("PEA") and the work combines various ideas from the field of cue combination, Bayesian methods, and new data sets, collected in four experiments using various techniques that test very different components of the model. The central component of visual uncertainty is assessed in the first experiment. The model uses 4 other parameters to explain implicit adaptation. These parameters are 1) learning and 2) retention rate, as used in popular state space models, and the uncertainty (variance) of 3) predicted and 4) proprioceptive hand position. In particular, the authors observe that asymptotes for implicit learning do not saturate, as claimed before, but decrease again when rotations are very large and that this may have to do with visual uncertainty (e.g. Tsay et al., 2021, J Neurophysiol 125, 12-22). The final experiment confirms predictions of the fitted model about what happens when visual uncertainty is increased (overall decrease of adaptation). By incorporating visual uncertainty depending on retinal eccentricity, the predictions of the PEA model for very large perturbations are notably different from and better than, the predictions of the two other models it is compared to. That is, the paper provides strong support for the idea that visual uncertainty of errors matters for implicit adaptation.

Weaknesses<br /> Although the authors don't say this, the "concave" function that shows that adaptation does not saturate for larger rotations has been shown before, including in papers cited in this manuscript.

The first experiment, measuring visual uncertainty for several rotation sizes in error-clamped paradigms has several shortcomings, but these might not be so large as to invalidate the model or the findings in the rest of the manuscript. There are two main issues we highlight here. First, the data is not presented in units that allow comparison with vision science literature. Second, the 1 second delay between the movement endpoint and the disappearance of the cursor, and the presentation of the reference marker, may have led to substantial degradation of the visual memory of the cursor endpoint. That is, the experiment could be overestimating the visual uncertainty during implicit adaptation.

The paper's third experiment relies to a large degree on reproducing patterns found in one particular paper, where the reported hand positions - as a measure of proprioceptive sense of hand position - are given and plotted relative to an ever-present visual target, rather than relative to the actual hand position. That is, 1) since participants actively move to a visual target, the reported hand positions do not reflect proprioception, but mostly the remembered position of the target participants were trying to move to, and 2) if the reports are converted to a difference between the real and reported hand position (rather than the difference between the target and the report), those would be on the order of ~20{degree sign} which is roughly two times larger than any previously reported proprioceptive recalibration, and an order of magnitude larger than what the authors themselves find (1-2{degree sign}) and what their model predicts. Experiment 3 is perhaps not crucial to the paper, but it nicely provides support for the idea that proprioceptive recalibration can occur with error-clamped feedback.

Perhaps the largest caveat to the study is that it assumes that people do not look at the only error feedback available to them (and can explicitly suppress learning from it). This was probably true in the experiments used in the manuscript, but unlikely to be the case in most of the cited literature. Ignoring errors and suppressing adaptation would also be a disastrous strategy to use in the real world, such that our brains may not be very good at this. So the question remains to what degree - if any - the ideas behind the model generalize to experiments without fixation control, and more importantly, to real-life situations.

Specific comments:<br /> A small part of the manuscript relies on replicating or modeling the proprioceptive recalibration in a study we think does NOT measure proprioceptive recalibration (Tsay, Parvin & Ivry, JNP, 2020). In this study, participants reached for a visual target with a clamped cursor, and at the end of the reach were asked to indicate where they thought their hand was. The responses fell very close to the visual target both before and after the perturbation was introduced. This means that the difference between the actual hand position, and the reported/felt hand position gets very large as soon as the perturbation is introduced. That is, proprioceptive recalibration would necessarily have roughly the same magnitude as the adaptation displayed by participants. That would be several times larger than those found in studies where proprioceptive recalibration is measured without a visual anchor. The data is plotted in a way that makes it seem like the proprioceptive recalibration is very small, as they plot the responses relative to the visual target, and not the discrepancy between the actual and reported hand position. It seems to us that this study mostly measures short-term visual memory (of the target location). What is astounding about this study is that the responses change over time to begin with, even if only by a tiny amount. Perhaps this indicates some malleability of the visual system, but it is hard to say for sure.

Regardless, the results of that study do not form a solid basis for the current work and they should be removed. We would recommend making use of the dataset from the same authors, who improved their methods for measuring proprioception shifts just a year later (Tsay, Kim, Parvin, Stover, and Ivry, JNP, 2021). Although here the proprioceptive shifts during error-clamp adaptation (Exp 2) were tiny, and not quite significant (p<0.08), the reports are relative to the actual location of the passively placed unseen hand, measured in trials separate from those with reach adaptation and therefore there is no visual target to anchor their estimates to.

Experiment 1 measures visual uncertainty with increased rotation size. The authors cite relevant work on this topic (Levi & Klein etc) which has found a linear increase in uncertainty of the position of more and more eccentrically displayed stimuli.

First, this is a question where the reported stimuli and effects could greatly benefit from comparisons with the literature in vision science, and the results might even inform it. In order for that to happen, the units for the reported stimuli and effects should (also) be degrees of visual angle (dva).

As far as we know, all previous work has investigated static stimuli, where with moving stimuli, position information from several parts of the visual field are likely integrated over time in a final estimate of position at the end of the trajectory (a Kalman filter type process perhaps). As far as we know, there are no studies in vision science on the uncertainty of the endpoint of moving stimuli. So we think that the experiment is necessary for this study, but there are some areas where it could be improved.

Then, the linear fit is done in the space of the rotation size, but not in the space of eccentricity relative to fixation, and these do not necessarily map onto each other linearly. If we assume that the eye-tracker and the screen were at the closest distance the manufacturer reports it to work accurately at (45 cm), we would get the largest distances the endpoints are away from fixation in dva. Based on that assumed distance between the participant and monitor, we converted the rotation angles to distances between fixation and the cursor endpoint in degrees visual angle: 0.88, 3.5, and 13.25 dva (ignoring screen curvature, or the absence of it). The ratio between the perturbation angle and retinal distance to the endpoint is roughly 0.221, 0.221, and 0.207 if the minimum distance is indeed used - which is probably fine in this case. But still, it would be better to do fit in the relevant perceptual coordinate system.

The first distance (4 deg rotation; 0.88 dva offset between fixation and stimulus) is so close to fixation (even at the assumed shortest distance between eye and screen) that it can be considered foveal and falls within the range of noise of eye-trackers + that of the eye for fixating. There should be no uncertainty on or that close to the fovea. The variability in the data is likely just measurement noise. This also means that a linear fit will almost always go through this point, somewhat skewing the results toward linearity. The advantage is that the estimate of the intercept (measurement noise) is going to be very good. Unfortunately, there are only 2 other points measured, which (if used without the closest point) will always support a linear fit. Therefore, the experiment does not seem suitable to test linearity, only to characterize it, which might be sufficient for the current purposes. We'd understand if the effort to do a test of linearity using many more rotations requires too much effort. But then it should be made much clearer that the experiment assumes linearity and only serves to characterize the assumed linearity.

Final comment after the consultation session:<br /> There were a lot of discussions about the actual interpretation of the behavioral data from this paper with regards to past papers (Tsay et al. 2020 or 2021), and how it matches the different variables of the model. The data from Tsay 2020 combined both proprioceptive information (Xp) and prediction about hand position (Xu) because it involves active movements. On the other hand, Tsay et al. 2021 is based on passive movements and could provide a better measure of Xp alone. We would encourage you to clarify how each of the variables used in the model is mapped onto the outcomes of the cited behavioral experiments.

The reviewers discussed this point extensively during the consultation process. The results reported in the Tsay 2020 study reflect both proprioception and prediction. However, having a visual target contributes more than just prediction, it is likely an anchor in the workspace that draws the response to it. Such that the report is dominated by short-term visual memory of the target (which is not part of the model). However, in the current Exp 3, as in most other work investigating proprioception, this is calculated relative to the actual direction.

The solution is fairly simple. In Experiment 3 in the current study, Xp is measured relative to the hand without any visual anchors drawing responses, and this is also consistent with the reference used in the Tsay et al 2021 study and from many studies in the lab of D. Henriques (none of which also have any visual reach target when measuring proprioceptive estimates). So we suggest using a different data set that also measures Xp without any other influences, such as the data from Tsay et al 2021 instead.

These issues with the data are not superficial and can not be solved within the model. Data with correctly measured biases (relative to the hand) that are not dominated by irrelevant visual attractors would actually be informative about the validity of the PEA model. Dr. Tsay has so much other that we recommend using a more to-the-point data set that could actually validate the PEA model.

This section particularly the exploration of utilitarianism in the context of social media, provides a thought-provoking perspective on ethical decision-making. The concept of the utility calculus as a method of predicting the outcomes and moral implications of our actions highlights the importance of comprehensive data collection and the potential pitfalls of biased or incomplete data. The discussion cleverly highlighted the challenges of navigating social media in an ethical manner, which must consider

Author Response

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

eLife assessment

This important study reports jAspSnFR3, a biosensor that enables high spatiotemporal resolution of aspartate levels in living cells. To develop this sensor, the authors used a structurally guided amino acid substitution in a glutamate/aspartate periplasmic binding protein to switch its specificity towards aspartate. The in vitro and in cellulo functional characterization of the biosensor is convincing, but evidence of the sensor's effectiveness in detecting small perturbations of aspartate levels and information on its behavior in response to acute aspartate elevations in the cytosol are still lacking.

We thank the reviewers and editors for the detailed assessment of our work and for their constructive feedback. Most comments have now been experimentally addressed in the revised manuscript, which we feel is substantially improved from the initial draft.

Public Reviews:

Reviewer #1 (Public Review):

In this manuscript, Davidsen and coworkers describe the development of a novel aspartate biosensor jAspSNFR3. This collaborative work supports and complements what was reported in a recent preprint by Hellweg et al., (bioRxiv; doi: 10.1101/2023.05.04.537313). In both studies, the newly engineered aspartate sensor was developed from the same glutamate biosensor previously developed by the authors of this manuscript. This coincidence is not casual but is the result of the need to find tools capable of measuring aspartate levels in vivo. Therefore, it is undoubtedly a relevant and timely work carried out by groups experienced in aspartate metabolism and in the generation of metabolite biosensors.

Reviewer #2 (Public Review):

In this work the IGluSnFR3 sensor, recently developed by Marvin et al (2023) is mutated position S72, which was previously reported to switch the specificity from Glu to Asp. They made 3 mutations at this position, selected a S72P mutant, then made a second mutation at S27 to generate an Asp-specific version of the sensor. This was then characterized thoroughly and used on some test experiments, where it was shown to detect and allow visualization of aspartate concentration changes over time. It is an incremental advance on the iGluSnFR3 study, where 2 predictable mutations are used to generate a sensor that works on a close analog of Glu, Asp. It is shown to have utility and will be useful in the field of Asp-mediated biological effects.

Reviewer #3 (Public Review):

In this manuscript, Davidsen and collaborators introduce jAspSnFR3, a new version of aspartate biosensor derived from iGluSnFR3, that allows monitoring in real-time aspartate levels in cultured cells. A selective amino acids substitution was applied in a key region of the template to switch its specificity from glutamate to aspartate. The jAspSnFR3 does not respond to other tested metabolites and performs well, is not toxic for cultured cells, and is not affected by temperature ensuring the possibility of using this tool in tissues physiologically more relevant. The high affinity for aspartate (KD=50 uM) allowed the authors to measure fluctuations of this amino acid in the physiological range. Different strategies were used to bring aspartate to the minimal level. Finally, the authors used jAspSnFR3 to estimate the intracellular aspartate concentration. One of the highlights of the manuscript was a treatment with asparagine during glutamine starvation. Although didn't corroborate the essentiality of asparagine in glutamine depletion, the measurement of aspartate during this supplementation is a glimpse of how useful this sensor can be.

Reviewer #1 (Recommendations For The Authors):

The authors should evaluate the effectiveness of the sensor in detecting small perturbations of aspartate levels and its behavior in response to acute aspartate elevations in the cytosol. In vivo aspartate determinations were performed exclusively in conditions that cause aspartate depletion. By means the use of mitochondrial respiratory inhibitors or aspartate withdrawal, it was determined the reliability of the sensor performing readings during relatively long periods, until reaching a steady-state of aspartate-depletion 12-60 hours later. Although in Hellweg and coworkers, it has been demonstrated that a related aspartate sensor could detect increases in aspartate in cell overexpressing the aspartate-glutamate GLAST transporter, the differences reported here between both sensors advise testing whether this aspect is also improved, or not, using jAspSNFR3.

Similarly, Davidsen et al. did not test if the sensor can be able to detect transient variations in cytosolic aspartate levels. In proliferative cells aspartate synthesis is linked to NAD+ regeneration by ETC (Sullivan et al., 2015, Cell), indeed the authors deplete aspartate using CI or CIII inhibitors but do not analyze if those are recovered, and increased, after its removal. Furthermore, the sequential addition of oligomycin and uncouplers could generate measurable fluctuations of aspartate in the cytosol.

We agree with the reviewer that only including situations of aspartate depletion in our cell culture experiments provided an incomplete evaluation of the utility of this biosensor. In the revised manuscript we provide three additional experiments using secondary treatments that restore aspartate synthesis to conditions that initially caused aspartate depletion. First, we conducted experiments where cells expressing jAspSnFR3/NucRFP were changed into media without glutamine, inducing aspartate depletion, with glutamine being replenished at various time points to observe if GFP/RFP measurements recover. As expected, glutamine withdrawal caused a decay in the GFP/RFP signal and we found that restoring glutamine caused a subsequent restoration of the GFP/RFP signal at all time points, with each fully recovering the GFP/RFP signal over time (Revised Manuscript Figure 2E). Next, we conducted the experiment suggested by the reviewer, testing whether the published finding, that oligomycin induced aspartate limitation can be remedied by co-treatment with electron transport chain uncouplers, could be visualized using jAspSnFR3 measurements of GFP/RFP. Indeed, after 24 hours of oligomycin induced aspartate depletion, treatment with the ETC uncoupler BAM15 dose dependently restored GFP/RFP signal (Revised Manuscript Figure 2G). Finally, we also measured whether the ability of pyruvate to mitigate the decrease in aspartate upon co-treated with rotenone (Figure 2B) could also be detected in a sequential treatment protocol after aspartate depletion. Indeed, after 24 hours of aspartate depletion by rotenone treatment, the GFP/RFP signal was rapidly restored by additional treatment with pyruvate (Revised Manuscript Figure 2, figure supplement 1C). Collectively, these results provide support for the utility of jAspSnFR3 to measure transient changes in aspartate levels in diverse metabolic situations, including conditions that restore aspartate to cells that had been experiencing aspartate depletion.

Reviewer #2 (Recommendations For The Authors):

Weaknesses: Sensor basically identical to iGluSnFR3, but nevertheless useful and specific. The results support the conclusions, and the paper is very straightforward. I think the work will be useful to people working on the effects of free aspartate in biology and given it is basically iGluSnFR3, which is widely used, should be very reproducible and reliable.

We appreciate the reviewer’s comment that sensor is useful for specific detection of aspartate. We agree that the advance of the paper is primarily in demonstrating its utility to measure aspartate, rather than any fundamental innovation on the biosensor approach. We hope the fact that jAspSnFR3 derives from a well validated biosensor (iGluSnFR3) will support its adoption.

Reviewer #3 (Recommendations For The Authors):

Although this is a well-performed study, I have some comments for the authors to address:

1) A red tag version of the sensor (jAspSnFR3-mRuby3) was generated for normalization purposes, with this the authors plan to correct GFP signal from expression and movement artifacts. I naturally interpret "movement artifacts" as those generated by variations in cell volume and focal plane during time-lapse experiments. However, it was mentioned that jAspSnFR3-mRuby3 included a histidine tag that may induce a non-specific effect (responses to the treatment with some amino acids). This suggests that a version without the tag needs to be generated and that an alternative design needs to be set for normalization purposes. A nuclear-localized RFP was expressed in a second attempt to incorporate RFP as a normalization signal. Here the cell lines that express both signals (sensor and RFP) were generated by independent lentiviral transductions (insertions). Unless the number of insertions for each construct is known, this approach will not ensure an equimolar expression of both proteins (sensor and RFP). In this scenario is not clear how the nuclear expression of RFP will help the correction by expression or monitor changes in cell volume. The authors may be interested in attempting a bicistronic system to express both the sensor and RFP.

The reviewer noted several potential issues concerning the use of RFP for normalization, which will be separated into sections below:

Movement artifacts:

We are glad the reviewer raised this issue since we see how it was confusingly worded. We have deleted the text “and movement artefacts” from the sentence.

His-tag and non-specific responses to some amino acids:

We also found it concerning that non-specific responses to amino acids could potentially contribute to our RFP normalization signal, and so we conducted additional experiments to address whether this was likely to be an issue in intracellular measurements. We first tested whether the non-specific signal was related to the histidine tag, or was intrinsic to the mRuby3 protein itself, by comparing the fluorescence response to a titration of histidine (which showed the largest effect of red fluorescence), aspartate, and GABA (structurally related to glutamate and aspartate, but lacking a carboxylate group) across a group of mRuby containing variants, with or without histidine tags. We replicated the non-specific signal originally observed in jAspSnFR3-mRuby3-His and found that another biosensor with a histidine tagged on the C terminus of mRuby3 had a similar response (iGlucoSnFR2.mRuby3-His), as did mRuby3-His alone, indicating that the aspect of being fused with jAspSnFR3 or another binding protein was not required for this effect. Additionally, we also compared the fluorescence response of lysates expressing mRuby2 and mRuby3 without histidine tags and found that the non-specific signal was essentially absent (Revised Manuscript Figure 1, figure supplement 4B-D). Collectively. These data support our original hypothesis that the histidine tag was responsible for the non-specific signal, alleviating concerns about more substantial protein design issues or with using nuc-RFP for normalization. Since we also found that measuring aspartate signal using GFP/RFP ratios from cells with linked the jAspSnFR3-Ruby3-His agreed with measurements from cells separately expressing jAspSnFR3 and nucRFP (without a His tag), and the amino acid concentrations needed to significantly alter His tagged Ruby3 signal are above those typically found in cells, we conclude that this is unlikely to be a significant factor in cells. Nonetheless, we have added all the relevant data to the manuscript to allow readers to make their own decision about which construct would be best for their purposes.

Original text:

"Surprisingly, the mRuby3 component responds to some amino acids at high millimolar concentrations, indicating a non-specific effect, potentially interactions with the C-terminal histidine tag (Figure 1—figure Supplement 2, panel B). Notably, this increase in fluorescence is still an order of magnitude lower than the green fluorescence response and it occurs at amino acid concentrations that are unlikely to be achieved in most cell types."

Revised text:

"Surprisingly, the mRuby3 fluorescence of affinity-purified jAspSnFR3.mRuby3 responds to some amino acids at high millimolar concentrations, indicating a non-specific effect (Figure 1—figure Supplement 4, panel A). This was determined to be due to an unexpected interaction with the C-terminal histidine tag and could be reproduced with other proteins containing mRuby3 and purified via the same C-terminal histidine tag (Figure 1—figure Supplement 4, panel B and C). Interestingly, a structurally related, non-amino acid compound, GABA, does not elicit a change in red fluorescence; indicating, that only amino acids are interacting with the histidine tag (Figure 1—figure Supplement 4, panel D). Nevertheless, most of our cell culture experiments were performed with nuclear localized mRuby2, which lacks a C-terminal histidine tag, and these measurements correlated with those using the histidine tagged jAspSnFR3-mRuby3 construct (Figure 1—figure Supplement 1 panel D)."

Lentiviral transductions

We agree that splitting the two fluorescent proteins across two expression constructs and infections effectively guarantees that there will not be equimolar expression of jAspSnFR3 and RFP, however we do not think equimolar expression is necessary in this context. The primary goal of RFP measurements in these experiments (and in experiments using the jAspSnFR3-mRuby3 fused construct) is to control for global alterations in protein expression that might confound the interpretation that a change in GFP fluorescence corresponds to a change in aspartate levels. While a bicistronic system is arguably a better approach to improve the similarity of expression of jAspSnFR3 and nuc-RFP in a cell, we only require that the cells have consistent expression of both proteins across all cells in the population, not that the expression of one necessarily be a similar molarity to the other. We accomplish consistent expression of proteins by single cell cloning after expression of jAspSnFR3 and nucRFP (or jAspSnFR3-mRuby3), and screening for clones that have high enough expression of both proteins such that they are well detected by standard Incucyte conditions. Given that our data do not identify an obvious downside to separate expression of jASPSnFR3 and nuc-RFP compared to the fused jAspSnFR3-mRuby3 construct (where the fluorescent proteins are truly equimolar) (Figure 2, Figure Supplement 1C), we elected to prioritize the separate jAspSnFR3 and nuc-RFP combination, which provides additional opportunities to measure cell number in the same experiment (see below).

2) The authors were interested in establishing the temporal dynamics of aspartate depletion by genetics and pharmaceutical means. For the inhibition of mitochondrial complex I rotenone and metformin were used. Although the assays are clearly showing aspartate depletion the report of cell viability is missing. Considering that glutamine deprivation induces arrest in cell proliferation, I think will be important to know the conditions of the cell cultures after 60 hours of treatment with such inhibitors.

We agree that ensuring that cells are still viable in conditions where aspartate is depleted, as determined by GFP/RFP in jAspSnFR3 expressing cells, is an important goal. To this end, we added a new experiment investigating the restoration of glutamine on the GFP/RFP signal at different time points after glutamine depletion (Revised Manuscript Figure 2E, see response to reviewer 1). One advantage of using the nuclear RFP as a normalization marker is that it also enables measurements of nuclei counts, a surrogate measurement for cell number. In the same glutamine depletion experiment we therefore measured cell counts using nuclear RFP incidences and confluency as measurements of cell proliferation/growth. In both cases, the arrest in cell proliferation upon glutamine withdrawal was obvious, as was the restoration of cell proliferation following glutamine replenishment, with the amount of growth delay corresponding to the length of glutamine withdrawal (Revised Manuscript Figure 2, Figure Supplement 2A-B). Nonetheless, there was no obvious lasting defects in restarting cell proliferation even after 12 hours of glutamine withdrawal, indicating that cell viability is preserved. In the case of mitochondrial inhibitors, we also observe even that after 24 hours of treatment with oligomycin or rotenone, restoration of aspartate synthesis from BAM15 or pyruvate, respectively, can also restore GFP/RFP signal, supporting the conclusion that cellular metabolism is still active in these conditions (Revised Manuscript Figure 2G; Revised Manuscript Figure 2, figure supplement 1C).

3) The pH sensitivity was checked in vitro with jAspSnFR3-mRuby3 and the sensor reported suitable for measurements at physiological pH. It would be an opportunity to revisit the analysis for pH sensitivity in cultured cells using an untagged version of jAspSnFR3 coupled, for example, to a sensor for pH.

We thank the reviewer for the suggestion and agree that pH effects on sensor signal could be a confounding factor in some conditions. Unfortunately, measuring intracellular pH is not trivial and using multiple fluorescent sensors that change simultaneously would be complex to interpret, particularly in the absence of controls to unambiguously control intracellular pH and aspartate concentrations. Thus, we believe that proper investigation of the variable of pH is beyond the scope of this study. Nonetheless, we agree that measuring the contribution of pH to sensor signal is an important goal for future work, particularly if deploying it in conditions likely to cause substantial pH differences, such as comparing compartmentalized signal of jAspSnFR3 in the cytosol and mitochondria. We have added the following italicized text to the conclusions section to underscore this point:

“Another potential use for this sensor would be to dissect compartmentalized metabolism, with mitochondria being a critical target, although incorporating the influence of pH on sensor fluorescence will be an important consideration in this context.”

4) While the authors take an interesting approach to measuring intracellular aspartate concentration, it will be highly desirable if a calibration protocol can be designed for this sensor. Clearly, glutamine depletion grants a minimal ("zero") aspartate concentration. However, having a more dynamic way for calibration will facilitate the introduction of this tool for metabolism studies. This may be achieved by incorporating a cultured cell that already expresses the transporter or by ectopic expression in the cells that have already been used.

We appreciate the suggestion and would similarly desire a calibration protocol to serve as a quantitative readout of aspartate levels from fluorescence signal, if possible. While we do calibrate jAspSnFR3 fluorescence in purified settings, conducting an analogous experiment intracellularly is currently difficult, if not impossible. While we have several methods to constrain the production rate of aspartate (glutamine withdrawal, mitochondrial inhibitors, and genetic knockouts of GOT1 and GOT2), we cannot prevent cells from decreasing aspartate consumption and so cannot get a true intracellular zero to aid in calibration. Additionally, the impermeability of aspartate to cell membranes makes it challenging to specifically control intracellular concentrations using environmental aspartate, and the best-known aspartate transporter (SLC1A3) is concentrative and so has the reciprocal problem. Considering these issues, we are wary of implying to readers that any specific fluorescence measurement can be used to directly interpret aspartate concentration given the many variables that can impact its signal, both related to the biosensor system itself (expression of jAspSnFR3, expression of Nuc-RFP, sensitivity and settings of the fluorescence detector) and based on cell intrinsic variability (differences in basal ASP levels, different sensitivity to treatments, influence of pH, etc.). We maintain that jAspSnFR3 has utility to measure relative changes in aspartate within a cell line across treatment conditions and over time, but absolute quantitation of aspartate still will require complementary approaches, like mass spectrometry, enzymatic assays, or NMR.

5) jAspSnFR3 seems to have the potential to be incorporated easily for several research groups as a main tool. In general, a minor correction to replace F/F with ΔF/F in the text.

Thank you for catching this error, the text has been edited accordingly.

Author Response

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this work, the authors provide evidence to show that an increase in Kv7 channels in hilar mossy cells of Fmr1 knock out mice results in a marked decrease in their excitability. The reduction in excitatory drive onto local hilar interneurons produces an increased excitation/inhibition ratio in granule cells. Inhibiting Kv7 channels can help normalize the excitatory drive in this circuit, suggesting that they may represent a viable target for targeted therapeutics for fragile-x syndrome.

Strengths:

The work is supported by a compelling and thorough set of electrophysiological studies. The authors do an excellent job of analysing their data and present a very complete data set.

We thank the Reviewer for the positive comments.

Weaknesses:

There are no significant weaknesses in the experimental work, however the complexity of the data presentation and the lack of a schematic showing the organizational framework of this circuit make the data less accessible to non-experts in the field. I highly encourage a graphical abstract and network diagram to help individuals understand the implications of this work.

We thank the Reviewer for the suggestion, and added a schematic of the dentate network organization (Figure 1A).

The work is important as it identifies a unique regional and cell-specific abnormality in Fmr1 KO mice, showing how the loss of one gene can result in region-specific changes in brain circuits.

Reviewer #2 (Public Review):

Summary:

Deng et al. investigate, for the first time to my knowledge, the role that hippocampal dentate gyrus mossy cells play in Fragile X Syndrome. They provide strong evidence that, in slice preparations from Fmr1 knockout mice, mossy cells are hypoactive due to increased Kv7 function whereas granule cells are hyperactive compared to slices from wild-type mice. They provide indirect evidence that the weakness of mossy cell-interneuron connections contributes to granule cell hyperexcitability, despite converse adaptations to mossy cell inputs. The authors show that application of the Kv7 inhibitor XE991 is able to rescue granule cell hyperexcitability back to wild-type baseline, supporting the overall conclusion that inhibition of Kv7 in the dentate may be a potential therapeutic approach for Fragile X Syndrome. However, any claims regarding specific circuit-based intervention or analysis are limited by the exclusively pharmacological approach of the manipulations.

Strengths:

Thorough electrophysiological characterization of mossy cells in Fmr1 knockout mice, a novel finding.

Their electrophysiological approach is quite rigorous: patched different neuron types (GC, MC, INs) one at a time within the dentate gyrus in FMR1 KO and WT, with and without 'circuit blockade' by pharmacologically inhibiting neurotransmission. This allows the most detailed characterization possible of passive membrane/intrinsic cell differences in the dentate gyrus of Fmr1 knockout mice.

Provide several examples showing the use of Kv7 inhibitor XE991 is able to rescue excitability of granule cell circuit in Fmr1 knockout mice (AP firing in the intact circuit, postsynaptic current recordings, theta-gamma coupling stimulation).

We thank the Reviewer for the positive comments.

Weaknesses:

The implications for these findings and the applicability of the potential treatment for the disorder in a whole animal are limited due to the fact that all experiments were done in slices.

We appreciate the Reviewer’s point and agree. To address this concern, we have revised the Discussion to state that “the applicability of a circuit-wide approach as a potential treatment in vivo will require extensive future behavioral analyses, which are beyond the scope of the current study”. We also now emphasize in Discussion that “these findings provide a proof-of-principle demonstration that a circuit-based intervention can normalize dynamic E/I balance and restore dentate circuit output in vitro”.

The authors' interpretation of the word 'circuit-based' is problematic - there are no truly circuit-specific manipulations in this study due to the reliance on pharmacology for their manipulations. While the application of the Kv7 inhibitor may have a predominant effect on the circuit through changes to mossy cell excitability, this manipulation would affect many other cells within the dentate and adjacent brain regions that connect to the dentate that express Kv7 as well.

We appreciate the reviewer’s point but would like to clarify that by using a term “circuit-based” we did not intend to imply that it is a “’circuit-specific” intervention. Our intended interpretation of the term ‘circuit-based’ stems from the following reasoning: the dentate circuit has two types of excitatory neurons which show opposite excitability defects in FXS mice, thus presenting an irreconcilable conflict to correct pharmacologically for each cell type individually. Instead, we sought an approach to correct the overall dentate circuit output, rather than to restore excitability defects of individual cell types. Notably, when we pharmacologically isolated granule cells from the circuit, inhibition of Kv7 failed to restore their excitability, suggesting that normalization of the dentate output depends on the circuit activity. Since we focused on correcting dentate output using such a circuit-dependent approach, we used the term ‘circuit-based intervention’ to emphasize this notion.

Reviewer #3 (Public Review):

The paper by Deng, Kumar, Cavalli, Klyachko describes that, unlike in other cell types, loss of Fmr1 decreases the excitability of hippocampal mossy cells due to up-regulation of Kv7 currents. They also show evidence that while muting mossy cells appears to be a compensatory mechanism, it contributes to the higher activity of the dentate gyrus, because the removal of mossy cell output alleviates the inhibition of dentate principal cells. This may be important for the patho-mechanism in Fragile X syndrome caused by the loss of Fmr1.

These experiments were carefully designed, and the results are presented ‎in a very logical, insightful, and self-explanatory way. Therefore, this paper represents strong evidence for the claims of the authors. In the current state of the manuscript, there are only a few points that need additional explanation.

We thank the Reviewer for the positive comments.

One of the results, which is shown in the supplementary dataset, does not fit the main conclusions. Changes in the mEPSC frequency suggest that in addition to the proposed network effects, there are additional changes in the synaptic machinery or synapse number that are independent of the actual activity of the neurons. Since the differences of the mEPSC and sEPSC frequencies are similar and because only the latter can signal network effects, while the former is typically interpreted as a presynaptic change, it cannot be claimed that sEPSC frequency changes are due to the hypo-excitability of mossy cells.

We thank the Reviewer for this important point and agree. To address this concern, we now state in Results that “We note that changes in the excitatory drive onto interneurons include both mEPSC and sEPSC frequencies, which reflect not only potential deficits in excitability of their input cells, such as MCs, but also changes in synaptic connectivity/function, that may arise from homeostatic circuit reorganization/compensation (see Discussion)”.

We also now emphasize this point in Discussion by stating that “alterations in excitatory drives, including both mEPSC and sEPSC frequencies onto interneurons, suggest changes in the excitatory synapse number and/or function. Together with alterations in inhibitory drives these changes may reflect compensatory circuit reorganization of both excitatory and inhibitory connections, including mossy cell synapses”.

We also note in Discussion that “Such circuit reorganization can explain the balanced E/I drive onto granule cells in Fmr1 KO mice we observed in the basal state, which can result from reorganization of excitatory and inhibitory axonal terminals”.

Notably, our findings that Kv7 blocker acting by increasing MC excitability is sufficient to correct dentate output, supports the notion that hypo-excitability of mossy cells is a major factor contributing to dentate circuit E/I imbalance. This does not exclude the presence of additional mechanisms contributing to E/I imbalance, such as changes of synaptic connectivity or release machinery. To reflect this point, we revised the Results to temper the initial claim that “this analysis supports the notion that the hypo-excitability of MCs in Fmr1 KO mice caused (now replaced with “is a major factor contributing to”) the reduction of excitatory drive onto hilar interneurons, which ultimately results in reduced local inhibition”.

An apparent technical issue may imply a second weak point in the interpretation of the results. Because the IPSCs in the PP stimulation experiments (Fig 8) start within a few milliseconds, it is unlikely that its first ‎components originate from the PP-GC-MC-IN feedforward inhibitory circuit. The involvement of this circuit and MCs in the Kv7-dependent excitability changes is the main implication of the results of this paper. But this feedforward inhibition requires three consecutive synaptic steps and EPSP-AP couplings, each of them lasting for at least 1ms + 2-5ms. Therefore, the inhibition via the PP-GC-MC-IN circuit can be only seen from 10-20ms after PP stimulation. The earlier components of the cPSCs should originate from other circuit elements that are not related to the rest of the paper. Therefore, more isolated measurements on the cPSC recordings are needed ‎which consider only the later phase of the IPSCs. This can be either a measurement of the decay phase or a pharmacological manipulation that selectively enhances/inhibits a specific component of the proposed circuit.

We appreciate the Reviewer’s point. As we mentioned in Results: “The EPSP measured in granule cells in response to the PP stimulation integrates both excitatory and inhibitory synaptic inputs onto granule cells, including the direct synaptic input from the PP and all the PP stimulation-associated feedforward and feedback synaptic inputs. In other words, the EPSP in granule cells integrates all dentate circuit ‘operations’.” As the Reviewer pointed out, this is also the case in the measurements of cPSCs, which comprise all of PP stimulation-associated feedforward and feedback inhibition. We thank the Reviewer for the suggestion to isolate specific components of IPSC. However, we did not attempt to do it in this study for three reasons. First, activity of all of these circuit components likely overlaps extensively in time and it is difficult to identify the specific time point that can separate contributions from earlier canonical feed-forward and feed-back components from the contribution of the later MC-dependent PP-GC-MC-IN feed-forward component. Notably the tri-synapse PP-GC-MC-IN component differs temporarily from the canonical di-synaptic (PP-GC-IN) feed-back inhibition only by a single synaptic activation step, resulting in only a few milliseconds difference. Moreover, the temporal differences in the contributions of these components vary widely among different recordings making a uniform analysis very difficult. Second, we used three different metrics to assess E/I changes in cPSC measurements, which capture a wide range of temporal processes and their integration, including peak-to-peak measurements, the charge transfer, and the excitation window metrics. Third, the principal readout in our study was the overall dentate output (i.e., granule cell firing), which reflects the integration of all dentate circuit ‘operations’ thus making the overall cPSC measurements appropriate, in our view, for this readout.

I suggest refraining from the conclusions saying "‎MCs provide at least ~51% of the excitatory drive onto interneurons in WT and ~41% in KO mice", because too many factors (eg. IN cell types, slice condition, synaptic reliability) are not accounted for in these actual numbers, and these values are not necessary for the general observation of the paper.

We thank the reviewer for this suggestion, and have revised the manuscript accordingly.

There are additional minor issues about the presentation of the results.

We have carefully checked and corrected the minor errors that reviewer pointed out.

Recommendations for the authors:

Revisions that are considered essential for improved assessment regarding the strengths of support of the claims:

• Temper claims regarding circuit-based effects

• Temper claims regarding very specific quantitative assessments of synaptic drives

• Differentiate between monosynaptic inputs and inputs arriving through multiple synaptic contacts with proper analytical techniques.

We appreciate these suggestions and have revised the manuscript to address the concerns raised by the reviewers.

Reviewer #1 (Recommendations For The Authors):

The authors do an outstanding job of reviewing and presenting all of their data. This is a paper I will recommend all of my trainees read, as it is an excellent example of a complete research project. While I am impressed with the effort involved, I also wondered if the complexity and thoroughness of their presentations could make the story less accessible to non-expert readers. My comments are simply intended to help them present a more coherent and succinct story to a wider audience, though I am not sure I really provide any meaningful changes. This is simply a very thorough and complete body of work that the authors should be commended for. After reading it I felt they had gone above and beyond what most authors would provide in terms of data to support their story, and thus I had no doubt that a change in Kv7 plays a role in changing the excitability of the network.

We thank the Reviewer for the positive comments and great suggestions. We have made numerous changes to present our work in a more coherent and succinct way, in part by re-plotting some of the figures, as well as by adding a schematic of the dentate circuit in Figure 1.

Figure 1. A visual of mossy cells and the local circuit they are studying would be a useful addition to Figure. 1. I also feel this is important for conveying the story of how hypo-excitability can impact the E/I of the network. I think it has to be more of a cell structure/circuit-based figure than is presented in Supplementary Figure 8.

We thank the reviewer for this suggestion. We have added a schematic of the dentate circuit with all major cell types involved in Figure 1A.

Figure 1. A, B, and C tell a coherent story and are easy to understand. The interpretation of the phase plot in D is harder to access. Perhaps having this as a separate figure and providing a clearer presentation of the way the phaseplot was created (see Figure 3 Bove et al., 2019, Neuroscience 418; DOI: 10.1016/j.neuroscience.2019.08.048)

We appreciate the Reviewer’s point and agree. In order to keep Figure 1 more concise and readable, we removed the phase plot in the revised version. This change did not negatively impact the result presentation because the primary aim of this plot was to visualize changes in voltage threshold in an alternative way, but it was already clearly shown by the ramp-evoked AP traces (revised Figure 1D, insert), and thus was not essential to show.

Figure 1 E-N might be better situated in a supplementary graph as the characteristics of the AP aren't changing.

We understand the Reviewer’s point, but we feel it would be better to keep all action potential metrics together in one figure, to show that only a specific subset of parameters was affected in Fmr1 KO mice.

Figure 2: (A-D) I am not sure having so many figures is required given the focus is on having a small change in Ir at one membrane potential. I do worry that the significance appears to be due to 2 cells with an IR of over 100 in the WT group and 2 with an IR of around 62 in the KO group. All other cells are between 75-100 in both groups. I also worry a bit bc in the literature IRs between 55 and 125 seem to be commonly reported by groups that do this work normally (Buzsacki, Westbrook, etc.). I would be cautious about making too much out of this result.

We thank the Reviewer for these comments. We have performed additional analyses of these data, as also suggested by Reviewer 3 (Point #1), and improved presentation of the data in Figure 2D-F by showing the effect of XE991 on increasing input resistance in WT vs KO. We also plotted other panels in a similar way to show the comparisons between WT and KO, as well as comparisons within genotype +/- XE991, which makes the results easy to follow. For more details, please also see the response to Reviewer 3, Point 1.

Figure 2D-E: As in the text, this result is really pointing towards there being a Kv7 issue. Worries about the data in D aside, I think these two figures alone tell a clearer story. Figure 3 on the other hand tells a story of the effects of blocking Kv7 on membrane potential. Is this central to the story the others are trying to tell?

We thank the reviewer for this point. We believe that Figure 2, Figure 3 and Figure 4—figure supplement 1 together provide strong and multifaceted evidence to support changes in Kv7 function in Fmr1 KO mossy cells.

Figure 3. This is an interesting finding that shows how detailed their analysis was. Showing that the change in holding current in KO animals is greater than in WT is the first solid piece of evidence that there is a change in Kv7 in these cells that affects their excitability.

We appreciate the reviewer’s comment. As mentioned above, we believe that Figure 2, Figure 3 and Figure 4—figure supplement 1 together provide strong and multifaceted evidence to support changes in Kv7 function in Fmr1 KO mossy cells.

Figures 4 and 5 provide additional detail to support the idea that Kv& changes by showing how the E/I ratio and spontaneous minis are shifted in KO animals.

We thank the Reviewer for the comments.

Figures 6-8 build a compelling story for the reduction in excitatory drive in mossy cells affecting the network dynamics in excitatory/inhibitory interactions in DG cells.

We appreciate the Reviewer’s comment.

Reviewer #2 (Recommendations For The Authors):

1) Other than location and characteristic morphology, the other parameters that were used to identify mossy cells and granule cells were also parameters used to find differences in cellular properties between wild-type and Fmr1 KO mice (RMP, sEPSC frequency, etc.), which would confound the results shown. The use of available transgenic mouse lines would provide for a more unbiased screen of these cells. Afterhyperpolarization was also used as a parameter while screening cells, yet none of the data on this measurement is shown.

We thank the reviewer for this point and agree that transgenic mouse lines provide a more unbiased way to identify various types of neurons. However, since the present study involves analyses of at least three different types of neurons, establishing multiple transgenic lines labeling different types of dentate neurons in the Fmr1 KO mouse model would be very time consuming and beyond the current resources of the lab. We would also like to clarify that the three types of dentate neurons are easily distinguished according to the large differences in location, morphology and basal electrophysiological properties, none of which were essential in defining differences between genotypes. Specifically, granule cells are located in the granule cell layer, have a small cell body (<10 m), RMP around -80mV, capacitance ~20 pF, and infrequent sEPSCs (<20 events/min); mossy cells are located in the hilus, have a large cell body (>15 m), RMP around -65 mV, capacitance >100 pF, and fast afterhyperpolarization less than -10 mV (WT –5.1 ± 0.7 mV, KO -5.8 ± 0.5 mV); interneurons are located in the hilus or border of granule cell layer, have a relative smaller cell body (10-15 m), RMP around -55 mV, capacitance <60 pF, and afterhyperpolarization larger than -15 mV (WT -20.4 ± 1.3 mV, KO -19.8 ±1.4 mV). We note that the cells that could not be definitively classified into the three categories were not included in analyses, and we have now clarified this further in the Methods. To address the reviewer’s second concern regarding AHP, we now provided the corresponding values in the Methods.

2) A definitive way to test the cell-autonomous nature of the Kv7 changes would be to use female mice, who will have a mosaic of cells affected by the fragile X chromosome, and the Fmr1 KO cells could be engineered to express GFP to help identify them from wild-type cells.

We agree and appreciate this suggestion. This could be an interesting follow up study to further verify the cell-autonomous nature of Kv7 changes.

3) The authors heavily rely on XE991 as a selective Kv7 blocker. Is it blocking all Kv7 channels at the concentration used? If so, given the significant expression of Kv7 in the dentate as shown by Western blot, is it surprising that there is no effect of this inhibitor on wild-type slices in most cases?

We thank the reviewer for this important point. We used 10x of IC50 concentration in the present study, suggesting that more than 80% of Kv7 should be blocked. Notably, we observed several effects of XE991 in WT mice: it significantly increased input resistance (new Figure 2D-F), and strongly enhanced AP firing evoked by step depolarization (Figure 7E-H), although we did not observe effect of XE991 in WT in the analyses of spiking evoked by theta-gamma stimulation in Figure 8. However, this is not surprising. If a parameter we measured is predominately cell-autonomous (for example, input resistance), the effects of XE991 are easy to observe. However, if a parameter reflects integration of all dentate circuit operations (for example, AP probability in response to theta-gamma stimulation), it is difficult to detect the effect of XE991 in WT mice because the dentate circuit of WT mice has larger capability to maintain E/I balance in response to XE991.

4) E/I ratio is a helpful concept, and it is heavily relied upon in the results text, but statistically shaky, especially for sEPSC:sIPSCs since you are combining uncertainty in the sEPSC and sIPSC to make one very uncertain ratio that doesn't undergo any subsequent statistical confirmation (such as in Fig 4I).

We appreciate the reviewer’s point and apologize for the confusion in presentation of Fig 4I (and 5I), due to lack of detailed explanation. The E/I ratio shown in Figs. 4I (and 5I) is a single data-point estimate calculated from the mean values of independent sEPSC and sIPSC measurements (Figs. 4G-H and 5G-H, respectively). This ratio was used only as an estimate/illustration of the changes, rather than a precise determination of the shift in E/I balance. Because there is only one data-point for this ratio, statistical analysis is not possible. For this reason we performed extensive additional analyses in Figures 7 and 8, in which the EPSC and IPSC were measured from the same cells and at the same time to define the actual E/I ratio with the corresponding statistical analyses (i.e., a real matched and dynamic E/I ratio).

5) Is this mGlur2/CB1 specificity to PP/granule and MC axons, respectively, true in the Fmr1 KO mice? It is possible that mGluR2 and CB1 expression patterns are altered in FMR1 KO, thus the assumption used to isolate these distinct inputs may not hold true.

This is a very good point. We do assume that the specificity of Group II mGluR and CB1 is similar between Fmr1 KO and WT mice, but this is an assumption that we have not directly verified. However, our results in Figures 7 and 8 strongly support this assumption, because if it were not true, then our intervention would be unlikely to correct the excessive dentate output.

6) XE991 only normalized GC firing when other cells were not pharmacologically blocked. The authors suggest this means blockage of MC Kv7 reduces GC excitability back to normal...presumably by increasing MC --> IN --> GC firing. This is a conclusion from many indirect comparisons (comparing XE991 effect on GC with/without GABA and glutamate blockers; comparing MC firing rates with/without XE991, and using CB1 agonist versus mGluR2 agonist to say it is mossy cells that are mostly controlling INs) - a clincher experiment would be to acutely knockdown Kv7 in mossy cells specifically and measure GC and IN firing.

Thank you, this is a great suggestion. Indeed, as an expansion of this project, in the future studies we are planning to manipulate excitability of mossy cells through manipulating Kv7, or using chemogenetic or optogenetic approaches.

7) The reasoning behind the FMRP-Kv7 connection is quite weak, citing the paper Darnell 2011 as "translational target", but FMRP has myriad translational targets.

We agree, and attempted to define the mechanism of increased Kv7 function using co-immunoprecipitation approach, as well as immunostaining to look at cell-type specific expression changes. However, both of these approaches were difficult to interpret due to technical limitations of the available antibodies. We also note that “We did not further investigate the precise mechanisms underlying enhancement of Kv7 function in the absence of FMRP, since the present study primarily focuses on the functional consequences of abnormal cellular and circuit excitability”. To address this concern, we extensively discussed the potential mechanisms of FMRP-Kv7 connection, acknowledged in Discussion that “further studies will be needed to elucidate the precise mechanism responsible for the increased Kv7 function in Fmr1 KO mice”, and will continue to investigate it in the future studies.

8) The authors attempt to look for changes in Kv7 expression with Western blot, but since they hypothesize that Kv7 changes are mainly in the mossy cells, it is perhaps not surprising that they would not be able to see any changes when they look at dentate as a whole. Staining for Kv7 subunits to look at expression on a cellular level would be beneficial.

We appreciate the reviewer’s suggestion. We attempted to perform the suggested experiments using immunostaining for KCNQ2, KCNQ3 and KCNQ5 in different subtypes of dentate neurons. However, these experiments failed to produce interpretable results due to technical limitations of the available antibodies.

9) Is Kv7 localization or splice/composition different in FMR1 KO mice?

This is a very good point. As we mentioned in Point 8 above, we were not able to perform these experiments and do not have the answer at this point.

10) Regarding the 3 subtypes of interneurons in the dentate, the authors are pooling data based on similar intrinsic properties, but this conclusion may be affected by the low number of recorded neurons for the regular-spiking type. In addition, it is unclear whether these different interneuron types have differential circuit connectivity (most likely) which would make it imperative to keep circuit analysis for interneurons segregated into these cell types.

We appreciate the reviewer’s point. Indeed, these different interneuron types may have distinct circuit connectivity and contributions to circuit activity. However, identification of these 3 types of interneurons and determination of their respective functions is in itself a very extensive set of experiments which is beyond the scope of the current manuscript. We also note that the functional readout of circuit activity in our measurements was the AP firing and EPSPs evoked in granule cells by PP stimulation, which integrate all dentate circuit operations, including all of the feedforward and feedback loops which are mediated by all of these different types of interneurons. For simplicity, we thus pooled all interneuron data for the purposes of this study. But we fully agree that extensive future work is required to elucidate interneuron-type specific changes in Fmr1 KO mice and their contributions to the dentate circuit dysfunction.

11) To do statistics treating each cell individually, and therefore assuming each cell is independent of one another, is not correct. Two cells from the same mouse will be more similar than two cells from different mice, therefore they are not independent data points. Nested statistical methods (n cells from o slices from p mice) will be important in future work, as discussed by (Aarts et al., Nat. Neurosci. 2014).

We agree with the Reviewer’s point and appreciate this suggestion. In the present study, the cells tested in electrophysiological experiments were from at least 3 different mice for each condition, which help minimize this kind of errors.

Reviewer #3 (Recommendations For The Authors):

Is there a difference in the Rin at -45mV of the control cell after the application of XE991? This is important to appreciate whether the XE991-sensitive conductances contribute to the basal excitability of MCs. Furthermore, the statistical comparison of the Rin at -45mV of the FXS animals in the control solution and in the presence of XE991 would be also important‎. Actually, the most accurate measurement would be to show a difference in the acute Kv7-blockade between control and FXS animals, if that is possible with this blocker. Additionally, it would be also informative if the bar graphs in Fig.2 D & E were merged for this purpose, similarly as in the later figures.

We thank the Reviewer for this suggestion and agree. Following this suggestion, we have re-plotted the data in Figure 2 accordingly. Specifically, we now show that XE991 significantly increased input resistance in both WT and KO mossy cells, and the effect of XE991 on increasing input resistance was markedly larger in KO than WT mossy cells. For other figures, we have plotted data in a similar way to show the comparisons between WT and KO, as well as comparisons within genotype +/- XE991.

Because of the cell-to-cell variability of the voltage responses, it would be more informative and representative if the average of traces from all cells were shown in Fig.2 D & E.

We agree with the Reviewer’s point. For clarity of presentation, we presented the cell-to-cell variability of the data as scatter points of input resistance values in the bar graph (Figure 2E), together with the representative traces (Figure 2D). Plotting the average traces from all cells would result in a total of 30 traces for all the WT and KO mice, which is difficult to visually assess clearly.

On page 7, please clarify the recorded cell type in this sentence: "In ‎contrast, WIN markedly reduced the number of sEPSCs in both WT and KO mice...".

We thank the Reviewer for pointing out this omission and have clarified it in the revised version.

In Figures 6 C, F, and I, the title of the Y-axis should be normalized frequency. Please also correct the figure legend accordingly because the current sentence can be also interpreted as the absolute or total number of events that were compared, irrespective of the duration of the recordings.

We thank the Reviewer for this point and have corrected the revised version accordingly.

Author Response

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

eLife assessment

This study presents a valuable finding on the immunophenotypes of cancer treatment-related pneumonitis. The evidence supporting the claims of the authors is solid, although the inclusion of controls, as suggested by one of the reviewers, strengthened the study. The work will be of interest to cancer immunologists.

Response: We are thankful for the editor's recognition of the contribution our study makes to understanding the immunophenotypes associated with cancer treatment-related pneumonitis. We agree that the inclusion of control data is pivotal for benchmarking biomarkers. While our initial study design was constrained by the availability of BALF from healthy individuals within clinical settings, we addressed this limitation by incorporating scRNA-seq data from healthy control and COVID-19 BALF cells sourced from the GSE145926 dataset. This additional analysis has provided a baseline for comparison, revealing that CD16 is expressed in a minority of T cells in healthy BALF, specifically 1.0% of CD4+ T cells and 1.6% of CD8+ T cells. The inclusion of this data as Figures 6H and 6I in our manuscript offers a robust context for the significant increase in CD16-expressing T cells observed in patients with PCP, thus enhancing the robustness of our study's conclusions.

Author response image 1.

Reviewer #1 (Recommendations For The Authors):

Many thanks for giving me the opportunity to review your paper. I really enjoyed the way you carried out this work - for example, your use of a wide panel of markers and the use of two analytical methods - you have clearly given great thought to bias avoidance. I also greatly appreciated your paragraph on the limitations, as there are several, but you do not 'over-sell' your conclusions so there is no issue here for me.

To improve the piece, there are a few typos (eg 318 - specific to alpha-myosin) and I was briefly confused about the highlighted clusters in Figure 4. Perhaps mention why they are highlighted when they first appear in 4D instead of E?

Response: We have corrected the typos, and we have rearranged the sequence of Figures 3E and 3F, as well as 4D and 4E, to ensure a logical flow. Citrus-generated violin plots are now presented prior to the heatmap of the clusters, which better illustrates the progression of our analysis and the derivation of the clusters.

In terms of improvements to the data, obviously it would have been ideal if you had had some sort of healthy control as a point of reference for all cohorts, but working in the field I understand the difficulties in getting healthy BAL. It would be worth your while however trying to find more supportive data in the literature in general. There are studies which assess various immune markers in healthy BAL eg https://journal-inflammation.biomedcentral.com/articles/10.1186/1476-9255-11-9. and so I think it is worth looking wrt the main findings. For example, are CD16+ T cells seen in healthy BAL or any other conditions (at present the COVID study is being over-relied on)? Could these cells be gamma deltas? (gamma deltas frequently express CD8 and CD16, and can switch to APC like phenotypes).

Response: We are grateful for the reviewer's consideration of the practical challenges associated with collecting BALF from healthy individuals. Alternatively, we have supplemented our analysis with single-cell RNA sequencing data from BALF cells of healthy controls, as found in existing literature (Nature Medicine 2020; 26: 842-844). We have accessed to GSE145926 and downloaded data of BALF cells from healthy control (n=3) and severe COVID19 (n=6). The filtered gene-barcode matrix was first normalized using ‘NormalizeData’ methods in Seurat v.4 with default parameters. The top 2,000 variable genes were then identified using the ‘vst’ method in Seurat FindVariableFeatures function. Then PCA and UMAP was performed. T cells were identified as CD2 >1 and CD3E >1, and FCGR3A expression was explored using an expression threshold of 0.5. Violin plots and bar plots were generated by ggplot function.

Regarding the pivotal finding of increased CD16-expressing T cells in patients with PCP, the scRNA-seq data mining indicates that CD16 is expressed by a minority of T cells in healthy BALF—1.0% of CD4+ T cells and 1.6% of CD8+ T cells. These figures, now incorporated into our revised manuscript as Figures 6H and 6I, substantiate our findings. These cells could be gamma delta T cells, but we could not confirm it with the limited data. We will investigate in the future study. The main text has been updated to reflect these findings.

Author response image 2.

I would agree with your approach of not going down the transcript route, so just focus on protein expression.

I think you need to mention more about the impact of ICI on PD1 expression - in the methods you lose one approach owing to low T cell expression (132) but in the discussion you mention ICI induced high expression (311) as previously reported. This apparent contradiction needs an explanation.

Response: We acknowledge the need for clarification regarding the impact of ICIs on PD-1 expression. In the methods section, the low detection of PD-1 expression on T cells in patients treated with nivolumab was indeed noted; this was due to the competitive nature of the PD-1 detection antibody EH12.2 with nivolumab. As reported by Suzuki et al. (International Immunology 2020; 32: 547-557), T cells from patients with ICI-induced ILD, including those treated with nivolumab, exhibit upregulated PD-1 expression, where the PD-1 detection antibody (clone: MIH4). Conversely, as outlined by Yanagihara et al. (BBRC 2020; 527: 213-217), the PD-1 detection antibody clone EH12.2 conjugated with 155Gd (#3155009B) used in our study is unable to detect PD-1 when patients are under nivolumab treatment due to competitive inhibition. The absence of a metal-conjugated PD-1 antibody with the MIH4 clone presented a limitation in our study. Ideally, we would have conjugated the MIH4 antibody with 155Gd for our analysis, which is a refinement we aim to incorporate in future research. We have now included this discussion in our manuscript to clarify the contradiction between the methodological limitations and the high PD-1 expression induced by ICIs, as reported in the literature. This addition will guide readers through the nuances of antibody selection and its implications for detecting PD-1 expression in the context of ICI treatment.

Finally, since you have the severity data, it would be good to assess all the significantly different clusters against this metric, as you have done for CD16+ T cells. Not only may this reveal more wrt the impact of other immune populations, but it'll also give a point of reference for the CD16+ T cell data.

Response: Thank you for the suggestion to assess all significantly different clusters against the disease severity metric. We have expanded our analysis to include a thorough correlation study between the disease severity and intensity of various T-cell markers. Notably, we observed that intensity of CCR7 expression correlates with the disease severity. Although the precise biological significance of this correlation remains to be elucidated, it may suggest a role for CCR7+ T cells in the pathogenesis or progression of the disease. We have considered the potential implications of this finding and included it as Supplementary Figure 5. We have also discussed this observation in the discussion section.

Author response image 3.

Overall though I think this is a really nice study, with a potentially very significant finding in linking CD16+ T cells with severity. Congratulations.

Response: We would like to thank the reviewer’s heartful comments on our manuscript.

Reviewer #2 (Recommendations For The Authors):

General:

1) The fact that this is a retrospective study should be indicated earlier in the paper.

Response: Now we have mentioned the retrospective nature of the study in the method section as follows: In this retrospective study, patients who were newly diagnosed with PCP, DI-ILD, and ICI-ILD and had undergone BALF collection at Kyushu University Hospital from January 2017 to April 2022 were included. The retrospective study was approved by the Ethics Committee of Kyushu University Hospital (reference number 22117-00).

2) tSNE and UMAP are dimensionality reduction techniques that don't cluster the cells, the authors should specify what clustering algorithm was used subsequently (e.g FlowSOM)

Response: The cluster was determined manually by their expression pattern.

3) With regards to the role of CD16 in a potential exacerbated cytotoxicity in the fatal PCP case, the authors could measure the levels of C3a related proteins in patient serum to link to a common immunopathogenic pathway with COVID.

Response: We did not collect serum from the patients in this study as our research protocol was approved by the Ethics committee for the use of BALF only. However, we agree with your assessment that the measurement of serum C3a levels would be informative. In future studies, we will incorporate the measurement of serum C3a levels to provide more comprehensive insights into the impact of C3a on immune function. Thank you for your valuable feedback and for helping us to improve the quality of our research.

Line-specific:

101 The authors should provide some information on how the cryopreservation of the BALF was carried out.

Response: Upon collection, BALF samples were immediately centrifuged at 300 g for 5 minutes to pellet the cells. The resultant cell pellets were then resuspended in Cellbanker 1 cryopreservation solution (Takara, catalog #210409). This suspension was aliquoted into cryovials and gradually frozen to –80ºC using a controlled rate freezing method to ensure cell viability. The samples were stored at –80ºC until required for experimental analysis. We have added the information in the method section.

Fig 3B: It would be very helpful if the authors could add a supplementary figure with marker expression on the UMAP projection.

Response: We have added Supplementary Figure 4 with marker expression on the UMAP projection in Figure 3B.

Fig 4A: Same as Fig 3B

Response: We have added Supplementary Figure 5 with marker expression on the UMAP projection in Figure 4A.

Fig 5B: Same as Fig 3B

Response: We have added Supplementary Figure 6 with marker expression on the tSNE projection in Figure 5B.

266 Authors should state if the data is not shown with regards to differences in myeloid cell fractions

430 Marker intensity is not shown in panel D

Re: Corrected as follows: “Citrus network tree visualizing the hierarchical relationship of each marker between identified T cell ~”

446 The legend says patients have IPF, CTD-ILD, sarcoidosis but the figure shows PCP, DI-ILD, ICI-ILD.

Re: Corrected.

451 What do the authors mean in "Graphical plots represent individual samples"? Panel B is a dot plot of all samples.

Response: Corrected as “Dot plots represent ~”.

472 What do the authors mean in "Graphical plots represent individual samples"? Panel C is a dot plot of all samples.

Response: Corrected as “Dot plots represent ~”.

Reviewer #3 (Recommendations For The Authors):

An important thing is to add comparisons against healthy donors, at least. A common baseline is needed to firmly establish any biomarkers.

Response: We acknowledge the reviewer's concern regarding the comparison with healthy donors. Although our study did not initially include BALF collection from healthy controls due to the constraints of clinical practice, we recognize the importance of a control baseline to validate biomarkers. To address this, we have integrated scRNA-seq data from healthy control BALF cells available in public datasets (Nature Medicine 2020; 26: 842-844), accessed from GSE145926. This dataset includes BALF cells from healthy controls (n=3) alongside severe COVID-19 patients (n=6). Data mining confirmed that CD16 expression is in a minority of T cells in healthy BALF—1.0% of CD4+ T cells and 1.6% of CD8+ T cells. We have included this comparative data in our manuscript as Figures 6H and 6I to provide context for the observed increase in CD16-expressing T cells in PCP patients, which substantiates our findings.

Author response image 4.

Data analysis needs to go deeper. There are several other tools on Cytobank alone that would allow a more quantitative analysis of the data. Fold changes in marker expressions would be very important as measurements of phenotypic changes.

Response: We thank the reviewer for their constructive feedback on the depth of our data analysis. We acknowledge the value of a more quantitative approach, including the use of fold change measurements to assess phenotypic alterations, and recognize the potential insights such tools on Cytobank could provide. Due to the scope and limited space of the current study, we have focused our analysis on the most pertinent findings relevant to our research questions. We believe the present analysis serves the immediate objectives of this study. However, we agree that further quantitative analysis would enhance the understanding of the data. We have expanded our analysis to include a thorough correlation study between the disease severity of PCP and intensity of various T-cell markers. Notably, we observed that intensity of CCR7 expression correlates with the disease severity of PCP. Although the precise biological significance of this correlation remains to be elucidated, it may suggest a role for CCR7+ T cells in the pathogenesis or progression of the disease. We have considered the potential implications of this finding and included it as Supplementary Figure 5. We have also discussed this observation in the discussion section. We aim to consider these approaches in future work to build upon the foundation laid by this study. Your suggestions are invaluable and will be kept at the forefront as we plan subsequent research phases.

Author response image 5.

Reviewer #1 (Public Review):

Cytotoxic agents and immune checkpoint inhibitors are the most commonly used and efficacious treatments for lung cancers. However their use brings two significant pulmonary side-effects; namely Pneumocystis jirovecii infection and resultant pneumonia (PCP), and interstitial lung disease (ILD). To observe the potential immunological drivers of these adverse events, Yanagihara et al. analysed and compared cells present in the bronchoalveolar lavage of three patient groups (PCP, cytotoxic drug-induced ILD [DI-ILD], and ICI-associated ILD [ICI-ILD]) using mass cytometry (64 markers). In PCP, they observed an expansion of the CD16+ T cell population, with the highest CD16+ T proportion (97.5%) in a fatal case, whilst in ICI-ILD, they found an increase in CD57+ CD8+ T cells expressing immune checkpoints (TIGIT+ LAG3+ TIM-3+ PD-1+), FCRL5+ B cells, and CCR2+ CCR5+ CD14+ monocytes. Given the fatal case, the authors also assessed for, and found, a correlation between CD16+ T cells and disease severity in PCP, postulating that this may be owing to endothelial destruction. Although n numbers are relatively small (n=7-9 in each cohort; common numbers for CyTOF papers), the authors use a wide panel (n=65) and two clustering methodologies giving greater strength to the conclusions. The differential populations discovered using one or two of the analytical methods are robust: whole population shifts with clear and significant clustering. These data are an excellent resource for clinical disease specialists and pan-disease immunologists, with a broad and engaging contextual discussion about what they could mean.

Strengths:

• The differences in immune cells in BAL in these specific patient subgroups is relatively unexplored.

• This is an observational study, with no starting hypothesis being tested.

• Two analytical methods are used to cluster the data.

• A relatively wide panel was used (64 markers), with particular strength in the alpha beta T cells and B cells.

• Relevant biomarkers, beta-D-glucan and KL-6 were also analysed

• Appropriate statistics were used throughout.

• Numbers are low (7 cases of PCP, 9 of DI-ILD, and 9 of ICI-ILD) but these are difficult samples to collect and so in relative terms, and considering the use of CyTOF, these are good numbers.

• Beta-D-glucan shows potential as a biomarker for PCP (as previously reported) whilst KL-6 shows potential as a biomarker for ICI-ILD (not reported before). Interestingly, KL-6 was not seen to be increased in DI-ILD patients.

• Despite the relatively low n numbers and lack of matching there are some clear differentials. The CD4/CD8+CD16+HLA-DR+CXCR3+CD14- T cell result is striking - up in PCP (with EM CD4s significantly down) - whilst the CD8 EMRA population is clear in ICI-ILD and 'non-exhausted' CD4s, with lower numbers of EMRA CD8s in DI-ILD.

• The authors identify 17/31 significantly differentiated clusters of myeloid cells, eg CD11bhi CD11chi CD64+ CD206+ alveolar macrophages with HLA-DRhi in PCP.

• With respect to B cells, the authors found that FCRL5+ B cells were more abundant in patients with ICI-ILD compared to those with PCP and DI-ILD, suggesting these FCRL5+ B cells may have a role in irAE.

• One patient's extreme CD16+ T cell (97.5% positive) and death, led the authors to consider CD16+ T cells as an indicator of disease severity in PCP. This was then tested and found to be correct.

• Authors discuss results in context of literature leading them to suggest that CD16+ T cells may target endothelial cells and wonder if anti-complement therapy may be efficacious in PCP.

• Great discussion on auto-reactive T cell clones where the authors suggest that in ICI-ILD CD8s may react against healthy lung, driving ILD.

• An observation of CXCR3 in different CD8 populations in ICI-ILD and PCP lead the authors to hypothesise on the chemoattractants in the microenvironment.

• Excellent point suggesting CD57 may not always be a marker of senescence on T cells - reflective of growing change within the community.

• Well considered suggestion that FCRL5+ B cells may be involved in ICI-ILD driven autoimmunity.

• The authors discuss the main weaknesses in the discussion and stress that the findings detailed in the paper "demonstrate a correlation rather than proof of causation".

• Figures and legends are clear and pleasing to the eye.

Weaknesses:

• This is an observational study, with no starting hypothesis being tested.

• Only patients who were able to have a lavage taken have been recruited.

• One set of analysis wasn't carried out for one subgroup (ICI-ILD) as PD1 expression was negative owing to the use of nivolumab.

• Some immune cell subsets wouldn't be picked up with the markers and gating strategies used; e.g. NK cells.

• Some immune cells would be disproportionately damaged by the storage, thawing and preparation of the samples; e.g. granulocytes.

• Numbers are low (7 cases of PCP, 9 of DI-ILD, and 9 of ICI-ILD), sex, age and adverse event matching wasn't performed, and treatment regimen are varied and 'suspected' (suggesting incomplete clinical data) - but these are difficult samples to collect. These numbers drop further for some analyses e.g. T cell clustering owing to factors such as low cell number.

• The disease comparisons are with each other, there is no healthy control.

• Samples are taken at one time point.

• The discussion on probably the stand out result - the CD16+ T cells in PCP - relies on two papers - leading to a slightly skewed emphasis on one paper on CD16+ cells in COVID. There are other papers out there that have observed CD16+ T cells in other conditions. It is also worth being in mind that given the markers used, these CD16+ T cell may be gamma deltas.

• The discussion on ICI patient consistently showing increased PD1, could have been greater, as given the ICI is targeting PD1, one would expect the opposite as commented on, and observed, in the methods section.

Reviewer #2 (Public Review):

Yanagihara and colleagues investigated the immune cell composition of bronchoalveolar lavage fluid (BALF) samples in a cohort of patients with malignancy undergoing chemotherapy and with with lung adverse reactions including Pneumocystis jirovecii pneumonia (PCP) and immune-checkpoint inhibitors (ICIs) or cytotoxic drug induced interstitial lung diseases (ILDs). Using mass cytometry, their aim was to characterize the cellular and molecular changes in BAL to improve our understanding of their pathogenesis and identify potential biomarkers and therapeutic targets. In this regard, the authors identify a correlation between CD16 expression in T cells and the severity of PCP and an increased infiltration of CD57+ CD8+ T cells expressing immune checkpoints and FCLR5+ B cells in ICI-ILD patients.

The conclusions of this paper are mostly well supported by data, but some aspects of the data analysis need to be clarified and extended.

1) The authors should elaborate on why different set of markers were selected for each analysis step. E.g., Different set of markers were used for UMAP, CITRUS and viSNE in the T cell and myeloid analysis.

2) The authors should state if a normality test for the distribution of the data was performed. If not, non-parametric tests should be used.

3) The authors should explore the correlation between CD16 intensity and the CTCAE grade in T cell subsets such as EMRA CD8 T cells, effector memory CD4, etc as identified in Figure 1B.

4) The authors could use CITRUS to better assess the B cell compartment.

Reviewer #3 (Public Review):

The authors collected BALF samples from lung cancer patients newly diagnosed with PCP, DI-ILD or ICI-ILD. CyTOF was performed on these samples, using two different panels (T-cell and B-cell/myeloid cell panels). Results were collected, cleaned-up, manually gated and pre-processed prior to visualisation with manifold learning approaches t-SNE (in the form of viSNE) or UMAP, and analysed by CITRUS (hierarchical clustering followed by feature selection and regression) for population identification - all using Cytobank implementation - in an attempt to identify possible biomarkers for these disease states. By comparing cell abundances from CITRUS results and qualitative inspection of a small number of marker expressions, the authors claimed to have identified an expansion of CD16+ T-cell population in PCP cases and an increase in CD57+ CD8+ T-cells, FCRL5+ B-cells and CCR2+ CCR5+ CD14+ monocytes in ICI-ILD cases.

By the authors' own admission, there is an absence of healthy donor samples and, perhaps as a result of retrospective experimental design, also an absence of pre-treatment samples. The entire analysis effectively compares three yet-established disease states with no common baseline - what really constitutes a "biomarker" in such cases? The introduction asserts that "y characterizing the cellular and molecular changes in BAL from patients with these complications, we aim to improve our understanding of their pathogenesis and identify potential therapeutic targets" (lines 82-84). Given these obvious omissions, no real "changes" have been studied in the paper. These are very limited comparisons among three, and only these three, states.

Even assuming more thorough experimental design, the data analysis is unfortunately too shallow and has not managed to explore the wealth of information that could potentially be extracted from the results. CITRUS is accessible and convenient, but also make a couple of big assumptions which could affect data analysis - 1) Is it justified to concatenate all FCS files to analyse the data in one batch / small batches? Could there be batch effects or otherwise other biological events that could confuse the algorithm? 2) With a relatively small number of samples, and after internal feature selection of CITRUS, is the regression model suitable for population identification or would it be too crude and miss out rare populations? There are plenty of other established methods that could be used instead. Have those methods been considered?

Colouring t-SNE or UMAP (e.g. Figure 6C) plots by marker expression is useful for quick identification of cell populations but it is not a quantitative analysis. In a CyTOF analysis like this, it is common to work out fold changes of marker expressions between conditions. It is inadequate to judge expression levels and infer differences simply by looking at colours.

The relatively small number of samples also mean that most results presented in the paper are not statistical significant. Whilst it is understandable that it is not always possible to collect a large number of patient samples for studies like this, having several entire major figures showing "n.s." (e.g. Figures 3A, 4B and 5C), together with limitations in the comparisons themselves and inadequate analysis, make the observations difficult to be convincing, and even less so for the single fatal PCP case where N = 1.

It would also be good scientific practice to show evidence of sample data quality control. Were individual FCS files examined? Did the staining work? Some indication of QC would also be great.

This dataset generated and studied by the authors have the potential to address the question they set out to answer and thus potentially be useful for the field. However, in the current state of presentation, more evidence and more thorough data analysis are needed to draw any conclusions, or correlations, as the authors would like to frame them.

I feel like when people think about perception, this is the aspect that usually comes to mind. So much so that other aspects are often disregarded. I can attest to this myself. When I think about how I perceive things, I don't usually think about selecting information, or even organizing it. This gives some interesting perspective going forward in life. I'm realizing I may have more things to consider as I take in the world.

Author Response

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

Reviewer #1:

Summary:

This paper performs fine-mapping of the silkworm mutants bd and its fertile allelic version, bdf, narrowing down the causal intervals to a small interval of a handful of genes. In this region, the gene orthologous to mamo is impaired by a large indel, and its function is later confirmed using expression profiling, RNAi, and CRISPR KO. All these experiments are convincingly showing that mamo is necessary for the suppression of melanic pigmentation in the silkworm larval integument. The authors also use in silico and in vitro assays to probe the potential effector genes that mamo may regulate. Strengths: The genotype-to-phenotype workflow, combining forward (mapping) and reverse genetics (RNAi and CRISPR loss-of-function assays) linking mamo to pigmentation are extremely convincing.

Response: Thank you very much for your affirmation of our work. The reviewer discussed the parts of our manuscript that involve evolution sentence by sentence. We have further refined the description in this regard and improved the logical flow. Thank you again for your help.

Weaknesses:

1) The last section of the results, entitled "Downstream target gene analysis" is primarily based on in silico genome-wide binding motif predictions.

While the authors identify a potential binding site using EMSA, it is unclear how much this general approach over-predicted potential targets. While I think this work is interesting, its potential caveats are not mentioned. In fact the Discussion section seems to trust the high number of target genes as a reliable result. Specifically, the authors correctly say: "even if there are some transcription factor-binding sites in a gene, the gene is not necessarily regulated by these factors in a specific tissue and period", but then propose a biological explanation that not all binding sites are relevant to expression control. This makes a radical short-cut that predicted binding sites are actual in vivo binding sites. This may not be true, as I'd expect that only a subset of binding motifs predicted by Positional Weight Matrices (PWM) are real in vivo binding sites with a ChIP-seq or Cut-and-Run signal. This is particularly problematic for PWM that feature only 5-nt signature motifs, as inferred here for mamo-S and mamo-L, simply because we can expect many predicted sites by chance.

Response: Thank you very much for your careful work. The analysis and identification of transcription factor-binding sites is an important issue in gene regulation research. Techniques such as ChIP-seq can be used to experimentally identify the binding sites of transcription factors (TFs). However, reports using these techniques often only detect specific cell types and developmental stages, resulting in a limited number of downstream target genes for some TFs. Interestingly, TFs may regulate different downstream target genes in different cell types and developmental stages.

Previous research has suggested that the ZF-DNA binding interface can be understood as a “canonical binding model”, in which each finger contacts DNA in an antiparallel manner. The binding sequence of the C2H2-ZF motif is determined by the amino acid residue sequence of its α-helical component. Considering the first amino acid residue in the α-helical region of the C2H2-ZF domain as position 1, positions -1, 2, 3, and 6 are key amino acids for recognizing and binding DNA. The residues at positions -1, 3, and 6 specifically interact with base 3, base 2, and base 1 of the DNA sense sequence, respectively, while the residue at position 2 interacts with the complementary DNA strand (Wolfe SA et al., 2000; Pabo CO et al., 2001). Based on this principle, the binding sites of C2H2-ZF have good reference value. For the 5-nt PWM sequence, we referred to the study of D. melanogaster, which was identified by EMSA (Shoichi Nakamura et al., 2019). In the new version, we have rewritten this section.

Pabo CO, Peisach E, Grant RA. Design and selection of novel Cys2His2 zinc finger proteins. Annu Rev Biochem. 2001;70:313-340.

Wolfe SA, Nekludova L, Pabo CO. DNA recognition by Cys2His2 zinc finger proteins. Annu Rev Biophys Biomol Struct. 2000;29:183-212.

Nakamura S, Hira S, Fujiwara M, et al. A truncated form of a transcription factor Mamo activates vasa in Drosophila embryos. Commun Biol. 2019;2:422. Published 2019 Nov 20.

2) The last part of the current discussion ("Notably, the industrial melanism event, in a short period of several decades ... a more advanced self-regulation program") is flawed with important logical shortcuts that assign "agency" to the evolutionary process. For instance, this section conveys the idea that phenotypically relevant mutations may not be random. I believe some of this is due to translation issues in English, as I understand that the authors want to express the idea that some parts of the genome are paths of least resistance for evolutionary change (e.g. the regulatory regions of developmental regulators are likely to articulate morphological change). But the language and tone is made worst by the mention that in another system, a mechanism involving photoreception drives adaptive plasticity, making it sound like the authors want to make a Lamarckian argument here (inheritance of acquired characteristics), or a point about orthogenesis (e.g. the idea that the environment may guide non-random mutations).

Because this last part of the current discussion suffers from confused statements on modes and tempo of regulatory evolution and is rather out of topic, I would suggest removing it.

In any case, it is important to highlight here that while this manuscript is an excellent genotype-to-phenotype study, it has very few comparative insights on the evolutionary process. The finding that mamo is a pattern or pigment regulatory factor is interesting and will deserve many more studies to decipher the full evolutionary study behind this Gene Regulatory Network.

Response: Thank you very much for your careful work. In this part of the manuscript, we introduced some assumptions that make the statement slightly unconventional. The color pattern of insects is an adaptive trait. The bd and bdf mutants used in the study are formed spontaneously. As a frequent variation and readily observable phenotype, color patterns have been used as models for evolutionary research (Wittkopp PJ et al., 2011). Darwin's theory of natural selection has epoch-making significance. I deeply believe in the theory that species strive to evolve through natural selection. However, with the development of molecular genetics, Darwinism’s theory of undirected random mutations and slow accumulation of micromutations resulting in phenotype evolution has been increasingly challenged.

The prerequisite for undirected random mutations and micromutations is excessive reproduction to generate a sufficiently large population. A sufficiently large population can contain sufficient genotypes to face various survival challenges. However, it is difficult to explain how some small groups and species with relatively low fertility rates have survived thus far. More importantly, the theory cannot explain the currently observed genomic mutation bias. In scientific research, every theory is constantly being modified to adapt to current discoveries. The most famous example is the debate over whether light is a particle or a wave, which has lasted for hundreds of years. However, in the 20th century, both sides seemed to compromise with each other, believing that light has a wave‒particle duality.

In summary, we have rewritten this section to reduce unnecessary assumptions.

Wittkopp PJ, Kalay G. Cis-regulatory elements: molecular mechanisms and evolutionary processes underlying divergence. Nat Rev Genet. 2011;13(1):59-69.

Minor Comment:

The gene models presented in Figure 1 are obsolete, as there are more recent annotations of the Bm-mamo gene that feature more complete intron-exon structures, including for the neighboring genes in the bd/bdf intervals. It remains true that the mamo locus encodes two protein isoforms.

An example of the Bm-mamo locus annotation, can be found at: https://www.ncbi.nlm.nih.gov/gene/101738295 RNAseq expression tracks (including from larval epidermis) can be displayed in the embedded genome browser from the link above using the "Configure Tracks" tool.

Based on these more recent annotations, I would say that most of the work on the two isoforms remains valid, but FigS2, and particularly Fig.S2C, need to be revised.

Response: Thank you very much for your careful work. In this study, we referred to the predicted genes of SilkDB, NCBI and Silkbase. In different databases, there are varying degrees of differences in the number of predicted genes and the length of gene mRNA. Because the SilkDB database is based on the first silkworm genome, it has been used for the longest time and has a relatively large number of users. In the revised manuscript, we have added the predicted genes of NCBI and Silkbase in Figure S1.

Author response image 1.

The predicted genes and qPCR analysis of candidate genes in the responsible genomic region for bd mutant. (A) The predicted genes in SilkDB;(B) the predicted genes in Genbak;(C) the predicted genes in Silkbase;(D) analysis of nucleotide differences in the responsible region of bd;(E) investigation of the expression level of candidate genes.

Reviewer #2 (Public Review):

Summary:

The authors tried to identify new genes involved in melanin metabolism and its spatial distribution in the silkworm Bombyx mori. They identified the gene Bm-mamo as playing a role in caterpillar pigmentation. By functional genetic and in silico approaches, they identified putative target genes of the Bm-mamo protein. They showed that numerous cuticular proteins are regulated by Bm-mamo during larval development.

Strengths:

• preliminary data about the role of cuticular proteins to pattern the localization of pigments

• timely question

• challenging question because it requires the development of future genetic and cell biology tools at the nanoscale

Response: Thank you very much for your affirmation of our work. The reviewer's familiarity with the color patterns of Lepidoptera is helpful, and the recommendation raised has provided us with very important assistance. This has allowed us to make significant progress with our manuscript.

Weaknesses:

• statistical sampling limited

• the discussion would gain in being shorter and refocused on a few points, especially the link between cuticular proteins and pigmentation. The article would be better if the last evolutionary-themed section of the discussion is removed.

A recent paper has been published on the same gene in Bombyx mori (https://www.sciencedirect.com/science/article/abs/pii/S0965174823000760) in August 2023. The authors must discuss and refer to this published paper through the present manuscript.

Response: Thank you very much for your careful work. First, we believe that competitive research is sometimes coincidental and sometimes intentional. Our research began in 2009, when we began to configure the recombinant population. In 2016, we published an article on comparative transcriptomics (Wu et al. 2016). The article mentioned above has a strong interest in our research and is based on our transcriptome analysis for further research, with the aim of making a preemptive publication. To discourage such behavior, we cannot cite it and do not want to discuss it in our paper.

Songyuan Wu et al. Comparative analysis of the integument transcriptomes of the black dilute mutant and the wild-type silkworm Bombyx mori. Sci Rep. 2016 May 19:6:26114. doi: 10.1038/srep26114.

Reviewer #1 (Recommendations For The Authors):

1) please consider using a more recent annotation model of the B. mori genome to revise your Result Section 1, Fig.1, and Fig. S2. https://www.ncbi.nlm.nih.gov/gene/101738295

Specifically, you used BGIM_ gene models, while the current annotation such as the one above featured in the NCBI database provides more accurate intron-exon structures without splitting mamo into tow genes. I believe this can be done with minor revisions of the figures, and you could keep the BGIM_ gene names for the text.

Response: Thank you very much for your careful work. The GenBank of NCBI (National Center for Biotechnology Information) is a very good database that we often use and refer to in this research process. Our research started in 2009, so we mainly referred to the SilkDB database (Jun Duan et al., 2010), although other databases also have references, such as NCBI and Silkbase (https://silkbase.ab.a.u-tokyo.ac.jp/cgi-bin/index.cgi). Because the SilkDB database was constructed based on the first published silkworm genome data, it has been used for the longest time and has a relatively large number of users. Recently, researchers are still using these data (Kejie Li et al., 2023).

The problem with predicting the mamo gene as two genes (BGIBMGA012517 and BGIBMGA012518) in SilkDB is mainly due to the presence of alternative splicing of the mamo gene. BGIBMGA012517 corresponds to the shorter transcript (mamo-s) of the mamo gene. Due to the differences in sequencing individuals, sequencing methods, and methods of gene prediction, there are differences in the number and sequence of predicted genes in different databases. We added the pattern diagram of predicted genes from NCBI and Silkbase, and the expression levels of new predicted genes are shown in Supplemental Figure S1.

Jun Duan et al., SilkDB v2.0: a platform for silkworm (Bombyx mori) genome biology. Nucleic Acids Res. 2010 Jan;38(Database issue): D453-6. doi: 10.1093/nar/gkp801. Kejie Li et al., Transcriptome analysis reveals that knocking out BmNPV iap2 induces apoptosis by inhibiting the oxidative phosphorylation pathway. Int J Biol Macromol. 2023 Apr 1;233:123482. doi: 10.1016/j.ijbiomac.2023.123482. Epub 2023 Jan 31.

Author response image 2.

The predicted genes and qPCR analysis of candidate genes in the responsible genomic region for bd mutant. (A) The predicted genes in SilkDB;(B) the predicted genes in Genbak;(C) the predicted genes in Silkbase;(D) analysis of nucleotide differences in the responsible region of bd;(E) investigation of the expression level of candidate genes.

2) As I mentioned in my public review, I strongly believe the interpretation of the PWM binding analyses require much more conservative statements taking into account the idea that short 5-nt motifs are expected by chance. The work in this section is interesting, but the manuscript would benefit from a quite significant rewrite of the corresponding Discussion section, making it that the in silico approach is prone to the identification of many sites in the genomes, and that very few of those sites are probably relevant for probabilistic reasons. I would recommend statements such as "Future experiments assessing the in vivo binding profile of Bm-mamo (eg. ChIP-seq or Cut&Run), will be required to further understand the GRNs controlled by mamo in various tissues".

Response: Thank you very much for your careful work. Previous research has suggested that the ZF-DNA binding interface can be understood as a “canonical binding model”, in which each finger contacts DNA in an antiparallel manner. The binding sequence of the C2H2-ZF motif is determined by the amino acid residue sequence of its α-helical component. Considering the first amino acid residue in the α-helical region of the C2H2-ZF domain as position 1, positions -1, 2, 3, and 6 are key amino acids for recognizing and binding DNA. The residues at positions -1, 3, and 6 specifically interact with base 3, base 2, and base 1 of the DNA sense sequence, respectively, while the residue at position 2 interacts with the complementary DNA strand (Wolfe SA et al., 2000; Pabo CO et al., 2001). Based on this principle, the prediction of DNA recognition motifs of C2H2-type zinc finger proteins currently has good accuracy.

The predicted DNA binding sequence (GTGCGTGGC) of the mamo protein in Drosophila melanogaster was highly consistent with that of silkworms. In addition, in D. melanogaster, the predicted DNA binding sequence of mamo, the bases at positions 1 to 7 (GTGCGTG), was highly similar to the DNA binding sequence obtained from EMSA experiments (Seiji Hira et al., 2013). Furthermore, in another study on the mamo protein of Drosophila melanogaster, five bases (TGCGT) were used as the DNA recognition core sequence of the mamo protein (Shoichi Nakamura et al., 2019). In the JASPAR database (https://jaspar.genereg.net), there are also some shorter (4-6 nt) DNA recognition sequences; for example, the DNA binding sequence of Ubx is TAAT (ID MA0094.1) in Drosophila melanogaster. However, we used longer DNA binding motifs (9 nt and 15 nt) of mamo to study the 2 kb genomic regions near the predicted gene. Over 70% of predicted genes were found to have these feature sequences near them. This analysis method is carried out with common software and processes. Due to sufficient target proteins, the accessibility of DNA, the absence of suppressors, the suitability of ion environments, etc., zinc finger protein transcription factors are more likely to bind to specific DNA sequences in vitro than in vivo. Using ChIP-seq or Cut&Run techniques to analyze various tissues and developmental stages in silkworms can yield one comprehensive DNA-binding map of mamo, and some false positives generated by predictions can be excluded. Thank you for your suggestion. We will conduct this work in the next research step. In addition, for brevity, we deleted the predicted data (Supplemental Tables S7 and S8) that used shorter motifs.

Pabo CO, Peisach E, Grant RA. Design and selection of novel Cys2His2 zinc finger proteins. Annu Rev Biochem. 2001;70:313-340.

Wolfe SA, Nekludova L, Pabo CO. DNA recognition by Cys2His2 zinc finger proteins. Annu Rev Biophys Biomol Struct. 2000;29:183-212.

Anton V Persikov et al., De novo prediction of DNA-binding specificities for Cys2His2 zinc finger proteins. Nucleic Acids Res. 2014 Jan;42(1):97-108. doi: 10.1093/nar/gkt890. Epub 2013 Oct 3.

Seiji Hira et al., Binding of Drosophila maternal Mamo protein to chromatin and specific DNA sequences. Biochem Biophys Res Commun. 2013 Aug 16;438(1):156-60. doi: 10.1016/j.bbrc.2013.07.045. Epub 2013 Jul 20.

Shoichi Nakamura et al., A truncated form of a transcription factor Mamo activates vasa in Drosophila embryos. Commun Biol. 2019 Nov 20;2: 422. doi: 10.1038/s42003-019-0663-4. eCollection 2019.

3) In my opinion, the last section of the Discussion needs to be completely removed ("Notably, the industrial melanism event, in a short period of several decades ... a more advanced self-regulation program"), as it is over-extending the data into evolutionary interpretations without any support. I would suggest instead writing a short paragraph asking whether the pigmentary role of mamo is a Lepidoptera novelty, or if it could have been lost in the fly lineage.

Below, I tried to comment point-by-point on the main issues I had.

Wu et al: Notably, the industrial melanism event, in a short period of several decades, resulted in significant changes in the body color of multiple Lepidoptera species(46). Industrial melanism events, such as changes in the body color of pepper moths, are heritable and caused by genomic mutations(47).

Yes, but the selective episode was brief, and the relevant "carbonaria" mutations may have existed for a long time at low-frequency in the population.

Response: Thank you very much for your careful work. Moth species often have melanic variants at low frequencies outside industrial regions. Recent molecular work on genetics has revealed that the melanic (carbonaria) allele of the peppered moth had a single origin in Britain. Further research indicated that the mutation event causing industrial melanism of peppered moth (Biston betularia) in the UK is the insertion of a transposon element into the first intron of the cortex gene. Interestingly, statistical inference based on the distribution of recombined carbonaria haplotypes indicates that this transposition event occurred in approximately 1819, a date highly consistent with a detectable frequency being achieved in the mid-1840s (Arjen E Van't Hof, et al., 2016). From molecular research, it is suggested that this single origin melanized mutant (carbonaria) was generated near the industrial development period, rather than the ancient genotype, in the UK. We have rewritten this part of the manuscript.

Arjen E Van't Hof, et al., The industrial melanism mutation in British peppered moths is a transposable element. Nature. 2016 Jun 2;534(7605):102-5. doi: 10.1038/nature17951.

Wu et al: If relying solely on random mutations in the genome, which have a time unit of millions of years, to explain the evolution of the phenotype is not enough.

What you imply here is problematic for several reasons.

First, as you point out later, some large-effect mutations (e.g. transpositions) can happen quickly.

Second, it's unclear what "the time units of million of years" means here... mutations occur, segregate in populations, and are selected. The speed of this process depends on the context and genetic architectures.

Third, I think I understand what you mean with "to explain the evolution of the phenotype is not enough", but this would probably need a reformulation and I don't think it's relevant to bring it here. After all, you used loss-of-function mutants to explain the evolution of artificially selected mutants. The evolutionary insights from these mutants are limited. Random mutations at the mamo locus are perfectly sufficient here to explain the bd and bdf phenotypes and larval traits.

Response: Thank you very much for your careful work. Charles Darwin himself, who argued that “natural selection can act only by taking advantage of slight successive variations; she can never take a leap, but must advance by the shortest and slowest steps” (Darwin, C. R. 1859). This ‘micromutational’ view of adaptation proved extraordinarily influential. However, the accumulation of micromutations is a lengthy process, which requires a very long time to evolve a significant phenotype. This may be only a proportion of the cases. Interestingly, recent molecular biology studies have shown that the evolution of some morphological traits involves a modest number of genetic changes (H Allen Orr. 2005).

One example is the genetic basis analysis of armor-plate reduction and pelvic reduction of the three-spined stickleback (Gasterosteus aculeatus) in postglacial lakes. Although the marine form of this species has thick armor, the lake population (which was recently derived from the marine form) does not. The repeated independent evolution of lake morphology has resulted in reduced armor plate and pelvic structures, and there is no doubt that these morphological changes are adaptive. Research has shown that pelvic loss in different natural populations of three-spined stickleback fish occurs by regulatory mutations deleting a tissue-specific enhancer (Pel) of the pituitary homeobox transcription factor 1 (Pitx1) gene. The researchers genotyped 13 pelvic-reduced populations of three-spined stickleback from disparate geographic locations. Nine of the 13 pelvic-reduced stickleback populations had sequence deletions of varying lengths, all of which were located at the Pel enhancer. Relying solely on random mutations in the genome cannot lead to such similar mutation forms among different populations. The author suggested that the Pitx1 locus of the stickleback genome may be prone to double-stranded DNA breaks that are subsequently repaired by NHEJ (Yingguang Frank Chan et al., 2010).

The bd and bdf mutants used in the study are formed spontaneously. Natural mutation is one of the driving forces of evolution. Nevertheless, we have rewritten the content of this section.

Darwin, C. R. The Origin of Species (J. Murray, London, 1859).

H Allen Orr. The genetic theory of adaptation: a brief history. Nat Rev Genet. 2005 Feb;6(2):119-27. doi: 10.1038/nrg1523.

Yingguang Frank Chan et al., Adaptive evolution of pelvic reduction in sticklebacks by recurrent deletion of a Pitx1 enhancer. Science. 2010 Jan 15;327(5963):302-5. doi: 10.1126/science.1182213. Epub 2009 Dec 10.

Wu et al: Interestingly, the larva of peppered moths has multiple visual factors encoded by visual genes, which are conserved in multiple Lepidoptera, in the skin. Even when its compound eyes are covered, it can rely on the skin to feel the color of the environment to change its body color and adapt to the environment(48). Therefore, caterpillars/insects can distinguish the light wave frequency of the background. We suppose that perceptual signals can stimulate the GRN, the GRN guides the expression of some transcription factors and epigenetic factors, and the interaction of epigenetic factors and transcription factors can open or close the chromatin of corresponding downstream genes, which can guide downstream target gene expression.

This is extremely confusing because you are bringing in a plastic trait here. It's possible there is a connection between the sensory stimulus and the regulation of mamo in peppered moths, but this is a mere hypothesis. Here, by mentioning a plastic trait, this paragraph sounds as if it was making a statement about directed evolution, especially after implying in the previous sentence that (paraphrasing) "random mutations are not enough". To be perfectly honest, the current writing could be misinterpreted and co-opted by defenders of the Intelligent Design doctrine. I believe and trust this is not your intention.

Response: Thank you very much for your careful work. The plasticity of the body color of peppered moth larvae is very interesting, but we mainly wanted to emphasize that their skin shows the products of visual genes that can sense the color of the environment by perceiving light. Moreover, these genes are conserved in many insects. Human skin can also perceive light by opsins, suggesting that they might initiate light–induced signaling pathways (Haltaufderhyde K et al., 2015). This indicates that the perception of environmental light by the skin of animals and the induction of feedback through signaling pathways is a common phenomenon. For clarity, we have rewritten this section of the manuscript.

Haltaufderhyde K, Ozdeslik RN, Wicks NL, Najera JA, Oancea E. Opsin expression in human epidermal skin. Photochem Photobiol. 2015;91(1):117-123.

Wu et al: In addition, during the opening of chromatin, the probability of mutation of exposed genomic DNA sequences will increase (49).

Here again, this is veering towards a strongly Lamarckian view with the environment guiding specific mutation. I simply cannot see how this would apply to mamo, nothing in the current article indicates this could be the case here. Among many issues with this, it's unclear how chromatin opening in the larval integument may result in heritable mutations in the germline.

Response: Thank you very much for your careful work. Previous studies have shown that there is a mutation bias in the genome; compared with the intergenic region, the mutation frequency is reduced by half inside gene bodies and by two-thirds in essential genes. In addition, they compared the mutation rates of genes with different functions. The mutation rate in the coding region of essential genes (such as translation) is the lowest, and the mutation rates in the coding region of specialized functional genes (such as environmental response) are the highest. These patterns are mainly affected by the traits of the epigenome (J Grey Monroe et al., 2022).

In eukaryotes, chromatin is organized as repeating units of nucleosomes, each consisting of a histone octamer and the surrounding DNA. This structure can protect DNA. When one gene is activated, the chromatin region of this gene is locally opened, becoming an accessible region. Research has found that DNA accessibility can lead to a higher mutation rate in the region (Radhakrishnan Sabarinathan et al., 2016; Schuster-Böckler B et al., 2012; Lawrence MS et al., 2013; Polak P et al., 2015). In addition, the BTB-ZF protein mamo belongs to this family and can recruit histone modification factors such as DNA methyltransferase 1 (DMNT1), cullin3 (CUL3), histone deacetylase 1 (HDAC1), and histone acetyltransferase 1 (HAT1) to perform chromatin remodeling at specific genomic sites. Although mutations can be predicted by the characteristics of apparent chromatin, the forms of mutations are diverse and random. Therefore, this does not violate randomness. For clarity, we have rewritten this section of the manuscript.

J Grey Monroe, Mutation bias reflects natural selection in Arabidopsis thaliana. Nature. 2022 Feb;602(7895):101-105.

Sabarinathan R, Mularoni L, Deu-Pons J, Gonzalez-Perez A, López-Bigas N. Nucleotide excision repair is impaired by binding of transcription factors to DNA. Nature. 2016;532(7598):264-267.

Schuster-Böckler B, Lehner B. Chromatin organization is a major influence on regional mutation rates in human cancer cells. Nature. 2012;488(7412):504-507.

Lawrence MS, Stojanov P, Polak P, et al. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature. 2013;499(7457):214-218.

Polak P, Karlić R, Koren A, et al. Cell-of-origin chromatin organization shapes the mutational landscape of cancer. Nature. 2015;518(7539):360-364.

Mathew R, Seiler MP, Scanlon ST, et al. BTB-ZF factors recruit the E3 ligase cullin 3 to regulate lymphoid effector programs. Nature. 2012;491(7425):618-621.

Wu et al: Transposon insertion occurs in a timely manner upstream of the cortex gene in melanic pepper moths (47), which may be caused by the similar binding of transcription factors and opening of chromatin.

No, we do not think that the peppered moth mutation is Lamarckian at all, as seems to be inferred here (notice that by mentioning the peppered moth twice, you are juxtaposing a larval plastic trait and then a purely genetic wing trait, making it even more confusing). Also, the "in a timely manner" is superfluous, because all the data are consistent with a chance mutation being eventually picked up by strong directional mutation. The mutation and selection did NOT occur at the same time.

Response: Thank you very much for your careful work. The insertion of one transposon into the first intron of the cortex gene of industrial melanism in peppered moth occurred in approximately 1819, which is similar to the time of industrial development in the UK (Arjen E Van't Hof, et al., 2016). In multiple species of Heliconius, the cortex gene is the shared genetic basis for the regulation of wing coloring patterns. Interestingly, the SNP of the cortex, associated with the wing color pattern, does not overlap among different Heliconius species, such as H. erato dephoon and H. erato favorinus, which suggests that the mutations of this cortex gene have different origins (Nadeau NJ et al., 2016). In addition, in Junonia coenia (van der Burg KRL et al., 2020) and Bombyx mori (Ito K et al., 2016), the cortex gene is a candidate for regulating changes in wing coloring patterns. Overall, the cortex gene is an evolutionary hotspot for the variation of multiple butterfly and moth wing coloring patterns. In addition, it was observed that the variations in the cortex are diverse in these species, including SNPs, indels, transposon insertions, inversions, etc. This indicates that although there are evolutionary hotspots in the insect genome, this variation is random. Therefore, this is not completely detached from randomness.

Arjen E Van't Hof, et al., The industrial melanism mutation in British peppered moths is a transposable element. Nature. 2016 Jun 2;534(7605):102-5. doi: 10.1038/nature17951.

Nadeau NJ, Pardo-Diaz C, Whibley A, et al. The gene cortex controls mimicry and crypsis in butterflies and moths. Nature. 2016;534(7605):106-110.

van der Burg KRL, Lewis JJ, Brack BJ, Fandino RA, Mazo-Vargas A, Reed RD. Genomic architecture of a genetically assimilated seasonal color pattern. Science. 2020;370(6517):721-725.

Ito K, Katsuma S, Kuwazaki S, et al. Mapping and recombination analysis of two moth colour mutations, Black moth and Wild wing spot, in the silkworm Bombyx mori. Heredity (Edinb). 2016;116(1):52-59.

Wu et al: Therefore, we proposed that the genetic basis of color pattern evolution may mainly be system-guided programmed events that induce mutations in specific genomic regions of key genes rather than just random mutations of the genome.

While the mutational target of pigment evolution may involve a handful of developmental regulator genes, you do not have the data to infer such a strong conclusion at the moment.

The current formulation is also quite strong and teleological: "system-guided programmed events" imply intentionality or agency, an idea generally assigned to the anti-scientific Intelligent Design movement. There are a few examples of guided mutations, such as the adaptation phase of gRNA motifs in bacterial CRISPR assays, where I could see the term ""system-guided programmed events" to be applicable. But it is irrelevant here.

Response: Thank you very much for your careful work. The CRISPR-CAS9 system is indeed very well known. In addition, recent studies have found the existence of a Cas9-like gene editing system in eukaryotes, such as Fanzor. Fanzor (Fz) was reported in 2013 as a eukaryotic TnpB-IS200/IS605 protein encoded by the transposon origin, and it was initially thought that the Fz protein (and prokaryotic TnpBs) might regulate transposon activity through methyltransferase activity (Saito M et al., 2023). Fz has recently been found to be a eukaryotic CRISPR‒Cas system. Although this system is found in fungi and mollusks, it raises hopes for scholars to find similar systems in other higher animals. However, before these gene-editing systems became popular, zinc finger nucleases (ZFNs) were already being studied as a gene-editing system in many species. The mechanism by which ZFN recognizes DNA depends on its zinc finger motif (Urnov FD et al., 2005). This is consistent with the mechanism by which transcription factors recognize DNA-binding sites.

Furthermore, a very important evolutionary event in sexual reproduction is chromosome recombination during meiosis, which helps to produce more abundant alleles. Current research has found that this recombination event is not random. In mice and humans, the PRDM9 transcription factors are able to plan the sites of double-stranded breaks (DSBs) in meiosis recombination. PRDM9 is a histone methyltransferase consisting of three main regions: an amino-terminal region resembling the family of synovial sarcoma X (SSX) breakpoint proteins, which contains a Krüppel-associated box (KRAB) domain and an SSX repression domain (SSXRD); a PR/SET domain (a subclass of SET domains), surrounded by a pre-SET zinc knuckle and a post-SET zinc finger; and a long carboxy-terminal C2H2 zinc finger array. In most mammalian species, during early meiotic prophase, PRDM9 can determine recombination hotspots by H3K4 and H3K36 trimethylation (H3K4me3 and H3K36me3) of nucleosomes near its DNA-binding site. Subsequently, meiotic DNA DSBs are formed at hotspots through the combined action of SPO11 and TOPOVIBL. In addition, some proteins (such as RAD51) are involved in repairing the break point. In summary, programmed events of induced and repaired DSBs are widely present in organisms (Bhattacharyya T et al., 2019).

These studies indicate that on the basis of randomness, the genome also exhibits programmability.

Saito M, Xu P, Faure G, et al. Fanzor is a eukaryotic programmable RNA-guided endonuclease. Nature. 2023;620(7974):660-668.

Urnov FD, Miller JC, Lee YL, et al. Highly efficient endogenous human gene correction using designed zinc-finger nucleases. Nature. 2005;435(7042):646-651.

Bhattacharyya T, Walker M, Powers NR, et al. Prdm9 and Meiotic Cohesin Proteins Cooperatively Promote DNA Double-Strand Break Formation in Mammalian Spermatocytes [published correction appears in Curr Biol. 2021 Mar 22;31(6):1351]. Curr Biol. 2019;29(6):1002-1018.e7.

Wu et al: Based on this assumption, animals can undergo phenotypic changes more quickly and more accurately to cope with environmental changes. Thus, seemingly complex phenotypes such as cryptic coloring and mimicry that are highly similar to the background may have formed in a short period. However, the binding sites of some transcription factors widely distributed in the genome may be reserved regulatory interfaces to cope with potential environmental changes. In summary, the regulation of genes is smarter than imagined, and they resemble a more advanced self-regulation program.

Here again, I can agree with the idea that certain genetic architectures can evolve quickly, but I cannot support the concept that the genetic changes are guided or accelerated by the environment. And again, none of this is relevant to the current findings about Bm-mamo.

Response: Thank you very much for your careful work. Darwin's theory of natural selection has epoch-making significance. I deeply believe in the theory that species strive to evolve through natural selection. However, with the development of molecular genetics, Darwinism’s theory of undirected random mutations and slow accumulation of micromutations resulting in phenotype evolution has been increasingly challenged.

The prerequisite for undirected random mutations and micromutations is excessive reproduction to generate a sufficiently large population. A sufficiently large population can contain sufficient genotypes to face various survival challenges. However, it is difficult to explain how some small groups and species with relatively low fertility rates have survived thus far. More importantly, the theory cannot explain the currently observed genomic mutation bias. In scientific research, every theory is constantly being modified to adapt to current discoveries. The most famous example is the debate over whether light is a particle or a wave, which has lasted for hundreds of years. However, in the 20th century, both sides seemed to compromise with each other, believing that light has a wave‒particle duality.

Epigenetics has developed rapidly since 1987. Epigenetics has been widely accepted, defined as stable inheritance caused by chromosomal conformational changes without altering the DNA sequence, which differs from genetic research on variations in gene sequences. However, an increasing number of studies have found that histone modifications can affect gene sequence variation. In addition, both histones and epigenetic factors are essentially encoded by genes in the genome. Therefore, genetics and epigenetics should be interactive rather than parallel. However, some transcription factors play an important role in epigenetic modifications. Meiotic recombination is a key process that ensures the correct separation of homologous chromosomes through DNA double-stranded break repair mechanisms. The transcription factor PRDM9 can determine recombination hotspots by H3K4 and H3K36 trimethylation (H3K4me3 and H3K36me3) of nucleosomes near its DNA-binding site (Bhattacharyya T et al., 2019). Interestingly, mamo has been identified as an important candidate factor for meiosis hotspot setting in Drosophila (Winbush A et al., 2021).

Bhattacharyya T, Walker M, Powers NR, et al. Prdm9 and Meiotic Cohesin Proteins Cooperatively Promote DNA Double-Strand Break Formation in Mammalian Spermatocytes [published correction appears in Curr Biol. 2021 Mar 22;31(6):1351]. Curr Biol. 2019;29(6):1002-1018.e7.

Winbush A, Singh ND. Genomics of Recombination Rate Variation in Temperature-Evolved Drosophila melanogaster Populations. Genome Biol Evol. 2021;13(1): evaa252.

Reviewer #2 (Recommendations For The Authors):

Major comments

• A recent paper has been published on the same gene in Bombyx mori (https://www.sciencedirect.com/science/article/abs/pii/S0965174823000760) in August 2023. The authors must discuss and refer to this published paper through the present manuscript.

Response: Thank you very much for your careful work. First, we believe that competitive research is sometimes coincidental and sometimes intentional. Our research began in 2009, when we began to configure the recombinant population. In 2016, we published an article on comparative transcriptomics (Wu et al. 2016). The article mentioned above has a strong interest in our research and is based on our transcriptome analysis for further research, with the aim of making a preemptive publication.

To discourage such behavior, we cannot cite it and do not want to discuss it in our paper.

Songyuan Wu et al. Comparative analysis of the integument transcriptomes of the black dilute mutant and the wild-type silkworm Bombyx mori. Sci Rep. 2016 May 19:6:26114. doi: 10.1038/srep26114.

• line 52-54. The numerous biological functions of insect coloration have been thoroughly investigated. It is reasonable to expect more references for each function.

Response: Thank you very much for your careful work. We have made the appropriate modifications.

Sword GA, Simpson SJ, El Hadi OT, Wilps H. Density-dependent aposematism in the desert locust. Proc Biol Sci. 2000;267(1438):63-68. … Behavior.

Barnes AI, Siva-Jothy MT. Density-dependent prophylaxis in the mealworm beetle Tenebrio molitor L. (Coleoptera: Tenebrionidae): cuticular melanization is an indicator of investment in immunity. Proc Biol Sci. 2000;267(1439):177-182. … Immunity.

N. F. Hadley, A. Savill, T. D. Schultz, Coloration and Its Thermal Consequences in the New-Zealand Tiger Beetle Neocicindela-Perhispida. J Therm Biol. 1992;17, 55-61…. Thermoregulation.

Y. G. Hu, Y. H. Shen, Z. Zhang, G. Q. Shi, Melanin and urate act to prevent ultraviolet damage in the integument of the silkworm, Bombyx mori. Arch Insect Biochem. 2013; 83, 41-55…. UV protection.

M. Stevens, G. D. Ruxton, Linking the evolution and form of warning coloration in nature. P Roy Soc B-Biol Sci. 2012; 279, 417-426…. Aposematism.

K. K. Dasmahapatra et al., Butterfly genome reveals promiscuous exchange of mimicry adaptations among species. Nature.2012; 487, 94-98…. Mimicry.

Gaitonde N, Joshi J, Kunte K. Evolution of ontogenic change in color defenses of swallowtail butterflies. Ecol Evol. 2018;8(19):9751-9763. Published 2018 Sep 3. …Crypsis.

B. S. Tullberg, S. Merilaita, C. Wiklund, Aposematism and crypsis combined as a result of distance dependence: functional versatility of the colour pattern in the swallowtail butterfly larva. P Roy Soc B-Biol Sci.2005; 272, 1315-1321…. Aposematism and crypsis combined.

• line 59-60. This general statement needs to be rephrased. I suggest remaining simple by indicating that insect coloration can be pigmentary, structural, or bioluminescent. About the structural coloration and associated nanostructures, the authors could cite recent reviews, such as: Seago et al., Interface 2009 + Lloyd and Nadeau, Current Opinion in Genetics & Development 2021 + "Light as matter: natural structural colour in art" by Finet C. 2023. I suggest doing the same for recent reviews that cover pigmentary and bioluminescent coloration in insects. The very recent paper by Nishida et al. in Cell Reports 2023 on butterfly wing color made of pigmented liquid is also unique and worth to consider.

Response: Thank you very much for your careful work. We have made the appropriate modifications.

Insect coloration can be pigmentary, structural, or bioluminescent. Pigments are mainly synthesized by the insects themselves and form solid particles that are deposited in the cuticle of the body surface and the scales of the wings (10, 11). Interestingly, recent studies have found that bile pigments and carotenoid pigments synthesized through biological synthesis are incorporated into body fluids and passed through the wing membranes of two butterflies (Siproeta stelenes and Philaethria diatonica) via hemolymph circulation, providing color in the form of liquid pigments (12). The pigments form colors by selective absorption and/or scattering of light depending on their physical properties (13). However, structural color refers to colors, such as metallic colors and iridescence, generated by optical interference and grating diffraction of the microstructure/nanostructure of the body surface or appendages (such as scales) (14, 15). Pigment color and structural color are widely distributed in insects and can only be observed by the naked eye in illuminated environments. However, some insects, such as fireflies, exhibit colors (green to orange) in the dark due to bioluminescence (16). Bioluminescence occurs when luciferase catalyzes the oxidation of small molecules of luciferin (17). In conclusion, the color patterns of insects have evolved to be highly sophisticated and are closely related to their living environments. For example, cryptic color can deceive animals via high similarity to the surrounding environment. However, the molecular mechanism by which insects form precise color patterns to match their living environment is still unknown.

• RNAi approach. I have no doubt that obtaining phenocopies by electroporation might be difficult. However, I find the final sampling a bit limited to draw conclusions from the RT-PCR (n=5 and n=3 for phenocopies and controls). Three control individuals is a very low number. Moreover, it would nice to see the variability on the plot, using for example violin plots.

Response: Thank you very much for your careful work. In the RNAi experiment, we injected more than 20 individuals in the experimental group and control group. We have added the RNAi data in Figure 4.

Author response table 1.

• Figure 6. Higher magnification images of Dazao and Bm-mamo knockout are needed, as shown in Figure 5 on RNAi.

Response: Thank you very much for your careful work. We have added enlarged images.

Author response image 3.

• Phylogenetic analysis/Figure S6. I am not sure to what extent the sampling is biased or not, but if not, it is noteworthy that mamo does not show duplicated copies (negative selection?). It might be interesting to discuss this point in the manuscript.

Response: Thank you very much for your careful work. mamo belongs to the BTB/POZ zinc finger family. The members of this family exhibit significant expansion in vertebrates. For example, there are 3 members in C. elegans, 13 in D. melanogaster, 16 in Bombyx mori, 58 in M. musculus and 63 in H. sapiens (Wu et al, 2019). These members contain conserved BTB/POZ domains but vary in number and amino acid residue compositions of the zinc finger motifs. Due to the zinc finger motifs that bind to different DNA recognition sequences, there may be differences in their downstream target genes. Therefore, when searching for orthologous genes from different species, we required high conservation of their zinc finger motif sequences. Due to these strict conditions, only one orthologous gene was found in these species.

• Differentially-expressed genes and CP candidate genes (line 189-191). The manuscript would gain in clarity if the authors explain more in details their procedure. For instance, they moved from a list of 191 genes to CP genes only. Can they say a little bit more about the non-CP genes that are differentially expressed? Maybe quantify the number of CPs among the total number of differentially-expressed genes to show that CPs are the main class?

Response: Thank you very much for your careful work. The nr (Nonredundant Protein Sequence Database) annotations for 191 differentially expressed genes in Supplemental Table S3 were added. Among them, there were 19 cuticular proteins, 17 antibacterial peptide genes, 6 transporter genes, 5 transcription factor genes, 5 cytochrome genes, 53 enzyme-encoding genes and others. Because CP genes were significantly enriched in differentially expressed genes (DEGs), previous studies have found that BmorCPH24 can affect pigmentation. Therefore, we first conducted an investigation into CP genes.

• Interaction between Bm-mamo. It is not clear why the authors chose to investigate the physical interaction of Bm-mamo protein with the putative binding site of yellow, and not with the sites upstream of tan and DDC. Do the authors test one interaction and assume the conclusion stands for the y, tan and DDC?

Response: Thank you very much for your careful work. In D. melanogaster, the yellow gene is the most studied pigment gene. The upstream and intron sequences of the yellow gene have been identified as containing multiple cis-regulatory elements. Due to the important pigmentation role of the yellow gene and its variable cis-regulatory sequence among different species, it has been considered a research model for cis-regulatory elements (Laurent Arnoult et al. 2013, Gizem Kalay et al. 2019, Yaqun Xin et al. 2020, Yann Le Poul et al. 2020). We use yellow as an example to illustrate the regulation of the mamo gene. We added this description to the discussion.

Laurent Arnoult et al. Emergence and diversification of fly pigmentation through evolution of a gene regulatory module. Science. 2013 Mar 22;339(6126):1423-6. doi: 10.1126/science.1233749.

Gizem Kalay et al. Redundant and Cryptic Enhancer Activities of the Drosophila yellow Gene. Genetics. 2019 May;212(1):343-360. doi: 10.1534/genetics.119.301985. Epub 2019 Mar 6.

Yaqun Xin et al. Enhancer evolutionary co-option through shared chromatin accessibility input. Proc Natl Acad Sci U S A. 2020 Aug 25;117(34):20636-20644. doi: 10.1073/pnas.2004003117. Epub 2020 Aug 10.

Yann Le Poul et al. Regulatory encoding of quantitative variation in spatial activity of a Drosophila enhancer. Sci Adv. 2020 Dec 2;6(49):eabe2955. doi: 10.1126/sciadv.abe2955. Print 2020 Dec.

• Please note that some controls are missing for the EMSA experiments. For instance, the putative binding-sites should be mutated and it should be shown that the interaction is lost.

Response: Thank you very much for your careful work. In this study, we found that the DNA recognition sequence of mamo is highly conserved across multiple species. In D. melanogaster, studies have found that mamo can directly bind to the intron of the vasa gene to activate its expression. The DNA recognition sequence they use is TGCGT (Shoichi Nakamura et al. 2019). We chose a longer sequence, GTGCGTGGC, to detect the binding of mamo. This binding mechanism is consistent across species.

• Figure 7 and supplementary data. How did the name of CPs attributed? According to automatic genome annotation of Bm genes and proteins? Based on Drosophila genome and associated gene names? Did the authors perform phylogenetic analyses to name the different CP genes?

Response: Thank you very much for your careful work. The naming of CPs is based on their conserved motif and their arrangement order on the chromosome. In previous reports, sequence identification and phylogenetic analysis of CPs have been carried out in silkworms (Zhengwen Yan et al. 2022, Ryo Futahashi et al. 2008). The members of the same family have sequence similarity between different species, and their functions may be similar. We have completed the names of these genes in the text, for example, changing CPR2 to BmorCPR2.

Zhengwen Yan et al. A Blueprint of Microstructures and Stage-Specific Transcriptome Dynamics of Cuticle Formation in Bombyx mori. Int J Mol Sci. 2022 May 5;23(9):5155.

Ningjia He et al. Proteomic analysis of cast cuticles from Anopheles gambiae by tandem mass spectrometry. Insect Biochem Mol Biol. 2007 Feb;37(2):135-46.

Maria V Karouzou et al. Drosophila cuticular proteins with the R&R Consensus: annotation and classification with a new tool for discriminating RR-1 and RR-2 sequences. Insect Biochem Mol Biol. 2007 Aug;37(8):754-60.

Ryo Futahashi et al. Genome-wide identification of cuticular protein genes in the silkworm, Bombyx mori. Insect Biochem Mol Biol. 2008 Dec;38(12):1138-46.

• Discussion. I think the discussion would gain in being shorter and refocused on the understudied role of CPs. Another non-canonical aspect of the discussion is the reference to additional experiments (e.g., parthogenesis line 290-302, figure S14). This is not the place to introduce more results, and it breaks the flow of the discussion. I encourage the authors to reshuffle the discussion: 1) summary of their findings on mamo and CPs, 2) link between pigmentation mutant phenotypes, pigmentation pattern and CPs, 3) general discussion about the (evo-)devo importance of CPs and link between pigment deposition and coloration. Three important papers should be mentioned here:

1) Matsuoka Y and A Monteiro (2018) Melanin pathway genes regulate color and morphology of butterfly wing scales. Cell Reports 24: 56-65... Yellow has a pleiotropic role in cuticle deposition and pigmentation.

2) https://arxiv.org/abs/2305.16628... Link between nanoscale cuticle density and pigmentation

3) https://www.cell.com/cell-reports/pdf/S2211-1247(23)00831-8.pdf... Variation in pigmentation and implication of endosomal maturation (gene red).

Response: Thank you very much for your careful work. We have rewritten the discussion section.

1) We have summarized our findings.

Bm-mamo may affect the synthesis of melanin in epidermis cells by regulating yellow, DDC, and tan; regulate the maturation of melanin granules in epidermis cells through BmMFS; and affect the deposition of melanin granules in the cuticle by regulating CP genes, thereby comprehensively regulating the color pattern in caterpillars.

2) We describe the relationship among the pigmentation mutation phenotype, pigmentation pattern, and CP.

Previous studies have shown that the lack of expression of BmorCPH24, which encodes important components of the endocuticle, can lead to dramatic changes in body shape and a significant reduction in the pigmentation of caterpillars (53). We crossed Bo (BmorCPH24 null mutation) and bd to obtain F1(Bo/+Bo, bd/+), then self-crossed F1 and observed the phenotype of F2. The lunar spots and star spots decreased, and light-colored stripes appeared on the body segments, but the other areas still had significant melanin pigmentation in double mutation (Bo, bd) individuals (Fig. S13). However, in previous studies, introduction of Bo into L (ectopic expression of wnt1 results in lunar stripes generated on each body segment) (24) and U (overexpression of SoxD results in excessive melanin pigmentation of the epidermis) (58) strains by genetic crosses can remarkably reduce the pigmentation of L and U (53). Interestingly, there was a more significant decrease in pigmentation in the double mutants (Bo, L) and (Bo, U) than in (Bo, bd). This suggests that Bm-mamo has a stronger ability than wnt1 and SoxD to regulate pigmentation. On the one hand, mamo may be a stronger regulator of the melanin metabolic pathway, and on the other hand, mamo may regulate other CP genes to reduce the impact of BmorCPH24 deficiency.

3) We discussed the importance of (evo-) devo in CPs and the relationship between pigment deposition and coloring.

CP genes usually account for over 1% of the total genes in an insect genome and can be categorized into several families, including CPR, CPG, CPH, CPAP1, CPAP3, CPT, CPF and CPFL (68). The CPR family is the largest group of CPs, containing a chitin-binding domain called the Rebers and Riddiford motif (R&R) (69). The variation in the R&R consensus sequence allows subdivision into three subfamilies (RR-1, RR-2, and RR-3) (70). Among the 28 CPs, 11 RR-1 genes, 6 RR-2 genes, 4 hypothetical cuticular protein (CPH) genes, 3 glycine-rich cuticular protein (CPG) genes, 3 cuticular protein Tweedle motif (CPT) genes, and 1 CPFL (like the CPFs in a conserved C-terminal region) gene were identified. The RR-1 consensus among species is usually more variable than RR-2, which suggests that RR-1 may have a species-specific function. RR-2 often clustered into several branches, which may be due to gene duplication events in co-orthologous groups and may result in conserved functions between species (71). The classification of CPH is due to their lack of known motifs. In the epidermis of Lepidoptera, the CPH genes often have high expression levels. For example, BmorCPH24 had a highest expression level, in silkworm larvae epidermis (72). The CPG protein is rich in glycine. The CPH and CPG genes are less commonly found in insects outside the order Lepidoptera (73). This suggests that they may provide species specific functions for the Lepidoptera. CPT contains a Tweedle motif, and the TweedleD1 mutation has a dramatic effect on body shape in D. melanogaster (74). The CPFL members are relatively conserved in species and may be involved in the synthesis of larval cuticles (75). CPT and CPFL may have relatively conserved functions among insects. The CP genes are a group of rapidly evolving genes, and their copy numbers may undergo significant changes in different species. In addition, RNAi experiments on 135 CP genes in brown planthopper (Nilaparvata lugens) showed that deficiency of 32 CP genes leads to significant defective phenotypes, such as lethal, developmental retardation, etc. It is suggested that the 32 CP genes are indispensable, and other CP genes may have redundant and complementary functions (76). In previous studies, it was found that the construction of the larval cuticle of silkworms requires the precise expression of over two hundred CP genes (22). The production, interaction, and deposition of CPs and pigments are complex and precise processes, and our research shows that Bm-mamo plays an important regulatory role in this process in silkworm caterpillars. For further understanding of the role of CPs, future work should aim to identify the function of important cuticular protein genes and the deposition mechanism in the cuticle.

Minor comments - Title. At this stage, there is no evidence that Bm-mamo regulates caterpillar pigmentation outside of Bombyx mori. I suggest to precise 'silkworm caterpillars' in the title.

Response: Thank you very much for your careful work. We have modified the title.

• Abstract, line 29. Because the knowledge on pigmentation pathway(s) is advanced, I would suggest writing 'color pattern is not fully understood' instead of 'color pattern is not clear'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 29. I suggest 'the transcription factor' rather than 'a transcription factor'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 30. If you want to mention the protein, the name 'Bm-mamo' should not be italicized.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 30. 'in the silkworm'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 31. 'mamo' should not be italicized.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 31. 'in Drosophila' rather 'of Drosophila'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 32. Bring detail if the gamete function is conserved in insects? In all animals?

Response: Thank you very much for your careful work. The sentence was changed to “This gene has a conserved function in gamete production in Drosophila and silkworms and evolved a pleiotropic function in the regulation of color patterns in caterpillars.”

• Introduction, line 51. I am not sure what the authors mean by 'under natural light'. Please rephrase.

Response: Thank you very much for your careful work. We have deleted “under natural light”.

• line 43. I find that the sentence 'In some studies, it has been proven that epidermal proteins can affect the body shape and appendage development of insects' is not necessary here. Furthermore, this sentence breaks the flow of the teaser.

Response: Thank you very much for your careful work. We have deleted this sentence.

• line 51-52. 'Greatly benefit them' should be rephrased in a more neutral way. For example, 'colours pattern have been shown to be involved in...'.

Response: Thank you very much for your careful work. We have modified to “and the color patterns have been shown to be involved in…”

• line 62. CPs are secreted by the epidermis, but I would say that CPs play their structural role in the cuticle, not directly in the epidermis. I suggest rephrasing this sentence and adding references.

Response: Thank you very much for your careful work. We have modified “epidermis” to “cuticle”.

• line 67. Please indicate that pathways have been identified/reported in Lepidoptera (11). Otherwise, the reader does not understand if you refer to previous biochemical in Drosophila for example.

Response: Thank you very much for your careful work. We have modified this sentence. “Moreover, the biochemical metabolic pathways of pigments used for color patterning in Lepidoptera…have been reported.”

• line 69. Missing examples of pleiotropic factors and associated references. For example, I suggest adding: engrailed (Dufour, Koshikawa and Finet, PNAS 2020) + antennapedia (Prakash et al., Cell Reports 2022) + optix (Reed et al., Science 2011), etc. Need to add references for clawless, abdominal-A.

Response: Thank you very much for your careful work. We have made modifications.

• line 76. The simpler term moth might be enough (instead of Lepidoptera).

Response: Thank you very much for your careful work. We have modified this to “insect”.

• line 96. I would simplify the text by writing "Then, quantitative RT-PCR was performed..."

Response: Thank you very much for your careful work. We have modified this sentence.

• line 112. 'Predict' instead of 'estimate'?

Response: Thank you very much for your careful work. We have modified this sentence.

• line 113. I would rather indicate the full name first, then indicate mamo between brackets.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 144. The Perl script needs to be made accessible on public repository.

Response: Thank you very much for your careful work.

• line 147-150. Too many technical details here. The details are already indicated in the material and methods section. Furthermore, the details break the flow of the paragraph.

Response: Thank you very much for your careful work. We have modified this section.

• line 152. Needs to make the link with the observed phenotypes in Figure 1. Just needs to state that RNAi phenocopies mimic the mutant alleles.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 153-157. Too many technical details here. The details are already indicated in the material and methods section. Furthermore, the details break the flow of the paragraph.

Response: Thank you very much for your careful work. We have simplified this paragraph.

• line 170. Please rephrase 'conserved in 30 species' because it might be understood as conserved in 30 species only, and not in other species.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 182. Maybe explain the rationale behind restricting the analysis to +/- 2kb. Can you cite a paper that shows that most of binding sites are within 2kb from the start codon?

Response: Thank you very much for your careful work. We have modified this sentence.

• line 182. '14,623 predicted genes'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 183. '10,622 genes'

Response: Thank you very much for your careful work. We have modified this sentence.

• line 183. Redundancy. Please remove 'silkworm' or 'B. mori'.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 187. '10,072 genes'

Response: Thank you very much for your careful work. We have modified this sentence.

• line 188. '9,853 genes'

Response: Thank you very much for your careful work. We have modified this sentence.

• line 200. "Therefore, the differential...in caterpillars" is a strong statement.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 204. Remove "The" in front of eight key genes. Also, needs a reference... maybe a recent review on the biochemical pathway of melanin in insects.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 220. This sentence is too general and vague. Please explicit what you mean by "in terms of evolution". Number of insect species? Diversity of niche occupancy? Morphological, physiological diversity?

Response: Thank you very much for your careful work. We have modified this sentence.

• line 285. The verb "believe" should be replaced by a more neutral one.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 354-355. This sentence needs to be rephrased in a more objective way.

Response: Thank you very much for your careful work. We have rewritten this sentence.

• line 378. Missing reference for MUSCLE.

Response: Thank you very much for your careful work. We have modified this sentence.

• line 379. Pearson model?

Response: Thank you very much for your careful work. We have modified this sentence.

• line 408. "The CRISPRdirect online software was used...".

Response: Thank you very much for your careful work. We have modified this sentence.

• Figure 1. In the title, I suggest indicating Dazao, bd, bdf as it appears in the figure. Needs to precise 'silkworm larval development'.

Response: Thank you very much for your careful work. We have modified this figure title.

• Figure 3. In the title, is the word 'pattern' really necessary? In the legend, please indicate the meaning of the acronyms AMSG and PSG.

Response: Thank you very much for your careful work. We have modified this figure legend.

• Figure S7A. Typo 'Znic finger 1', 'Znic finger 2', 'Znic finger 3',

Response: Thank you very much for your careful work. We have fixed these typos. .