Reviewer #3 (Public Review):

Summary:

The study claims to investigate trunk representations in elephant trigeminal nuclei located in the brainstem. The researchers identify large protrusions visible from the ventral surface of the brainstem, which they examined using a range of histological methods. However, this ventral location is usually where the inferior olivary complex is found, which challenges the author's assertions about the nucleus under analysis. They find that this brainstem nucleus of elephants contains repeating modules, with a focus on the anterior and largest unit which they define as the putative nucleus principalis trunk module of the trigeminal. The nucleus exhibits low neuron density, with glia outnumbering neurons significantly. The study also utilizes synchrotron X-ray phase contrast tomography to suggest that myelin-stripe-axons traverse this module. The analysis maps myelin-rich stripes in several specimens and concludes that based on their number and patterning that they likely correspond with trunk folds; however this conclusion is not well supported if the nucleus has been misidentified.

Strengths:

The strength of this research lies in its comprehensive use of various anatomical methods, including Nissl staining, myelin staining, Golgi staining, cytochrome oxidase labeling, and synchrotron X-ray phase contrast tomography. The inclusion of quantitative data on cell numbers and sizes, dendritic orientation and morphology, and blood vessel density across the nucleus adds a quantitative dimension. Furthermore, the research is commendable for its high-quality and abundant images and figures, effectively illustrating the anatomy under investigation.

Weaknesses:

While the research provides potentially valuable insights if revised to focus on the structure that appears to be inferior olivary nucleus, there are certain additional weaknesses that warrant further consideration. First, the suggestion that myelin stripes solely serve to separate sensory or motor modules rather than functioning as an "axonal supply system" lacks substantial support due to the absence of information about the neuronal origins and the termination targets of the axons. Postmortem fixed brain tissue limits the ability to trace full axon projections. While the study acknowledges these limitations, it is important to exercise caution in drawing conclusions about the precise role of myelin stripes without a more comprehensive understanding of their neural connections.

Second, the quantification presented in the study lacks comparison to other species or other relevant variables within the elephant specimens (i.e., whole brain or brainstem volume). The absence of comparative data to different species limits the ability to fully evaluate the significance of the findings. Comparative analyses could provide a broader context for understanding whether the observed features are unique to elephants or more common across species. This limitation in comparative data hinders a more comprehensive assessment of the implications of the research within the broader field of neuroanatomy. Furthermore, the quantitative comparisons between African and Asian elephant specimens should include some measure of overall brain size as a covariate in the analyses. Addressing these weaknesses would enable a richer interpretation of the study's findings.

Reviewer #4 (Public Review):

Summary:

The authors report a novel isomorphism in which the folds of the elephant trunk are recognizably mapped onto the principal sensory trigeminal nucleus in the brainstem. Further, they identifiy the enlarged nucleus as being situated in this species in an unusual ventral midline position.

Strengths:

The identity of the purported trigeminal nucleus and the isomorphic mapping with the trunk folds is supported by multiple lines of evidence: enhanced staining for cytochrome oxidase, an enzyme associated with high metabolic activity; dense vascularization, consistent with high metabolic activity; prominent myelinated bundles that partition the nucleus in a 1:1 mapping of the cutaneous folds in the trunk periphery; near absence of labeling for the anti-peripherin antibody, specific for climbing fibers, which can be seen as expected in the inferior olive; and a high density of glia.

Weaknesses:

Despite the supporting evidence listed above, the identification of the gross anatomical bumps, conspicuous in the ventral midline, is problematic. This would be the standard location of the inferior olive, with the principal trigeminal nucleus occupying a more dorsal position. This presents an apparent contradiction which at a minimum needs further discussion. Major species-specific specializations and positional shifts are well-documented for cortical areas, but nuclear layouts in the brainstem have been considered as less malleable.

Reviewer #5 (Public Review):

After reading the manuscript and the concerns raised by reviewer 2 I see both sides of the argument - the relative location of trigeminal nucleus versus the inferior olive is quite different in elephants (and different from previous studies in elephants), but when there is a large disproportionate magnification of a behaviorally relevant body part at most levels of the nervous system (certainly in the cortex and thalamus), you can get major shifting in location of different structures. In the case of the elephant, it looks like there may be a lot of shifting. Something that is compelling is that the number of modules separated but the myelin bands correspond to the number of trunk folds which is different in the different elephants. This sort of modular division based on body parts is a general principle of mammalian brain organization (demonstrated beautifully for the cuneate and gracile nucleus in primates, VP in most of species, S1 in a variety of mammals such as the star nosed mole and duck-billed platypus). I don't think these relative changes in the brainstem would require major genetic programming - although some surely exists. Rodents and elephants have been independently evolving for over 60 million years so there is a substantial amount of time for changes in each l lineage to occur.

I agree that the authors have identified the trigeminal nucleus correctly, although comparisons with more out groups would be needed to confirm this (although I'm not suggesting that the authors do this). I also think the new figure (which shows previous divisions of the brainstem versus their own) allows the reader to consider these issues for themselves. When reviewing this paper, I actually took the time to go through atlases of other species and even look at some of my own data from highly derived species. Establishing homology across groups based only on relative location is tough especially when there appears to be large shifts in relative location of structures. My thoughts are that the authors did an extraordinary amount of work on obtaining, processing and analyzing this extremely valuable tissue. They document their work with images of the tissue and their arguments for their divisions are solid. I feel that they have earned the right to speculate - with qualifications - which they provide.

Author response:

Thank you for organising the review and providing us with the reviewer's feedback. These comments are very useful, and we would like to express our gratitude to the reviewers for their efforts.

The reviewers all point out a number of related improvements, relating to: 1) describing various processing steps more clearly, in the online documentation but also in the manuscript itself (e.g. for particle picking), 2) describing more clearly what features Ais offers, how these compare to those of other programmes, and how they might be interfaced with in third-party programmes (e.g. the expected format of models), and 3) a degree of subjectivity in discussion of the results presented in the manuscript (e.g. our statement that Pix2pix performed better in some cases than did other architectures).

We will address these points, as well as the various other suggestions, in the upcoming revised manuscript and updates to Ais.

Reviewer #1 (Public Review):

This paper describes "Ais", a new software tool for machine-learning-based segmentation and particle picking of electron tomograms. The software can visualise tomograms as slices and allows manual annotation for the training of a provided set of various types of neural networks. New networks can be added, provided they adhere to a Python file with an (undescribed) format. Once networks have been trained on manually annotated tomograms, they can be used to segment new tomograms within the same software. The authors also set up an online repository to which users can upload their models, so they might be re-used by others with similar needs. By logically combining the results from different types of segmentations, they further improve the detection of distinct features. The authors demonstrate the usefulness of their software on various data sets. Thus, the software appears to be a valuable tool for the cryo-ET community that will lower the boundaries of using a variety of machine-learning methods to help interpret tomograms.

eLife assessment

This work describes a new software platform for machine-learning-based segmentation of and particle-picking in cryo-electron tomograms. The program and its corresponding online database of trained models will allow experimentalists to conveniently test different models and share their results with others. The paper provides solid evidence that the software will be valuable to the community.

Reviewer #2 (Public Review):

Summary:

Last et al. present Ais, a new deep learning-based software package for the segmentation of cryo-electron tomography data sets. The distinguishing factor of this package is its orientation to the joint use of different models, rather than the implementation of a given approach. Notably, the software is supported by an online repository of segmentation models, open to contributions from the community.

The usefulness of handling different models in one single environment is showcased with a comparative study on how different models perform on a given data set; then with an explanation of how the results of several models can be manually merged by the interactive tools inside Ais.

The manuscripts present two applications of Ais on real data sets; one is oriented to showcase its particle-picking capacities on a study previously completed by the authors; the second one refers to a complex segmentation problem on two different data sets (representing different geometries as bacterial cilia and mitochondria in a mouse neuron), both from public databases.

The software described in the paper is compactly documented on its website, additionally providing links to some YouTube videos (less than an hour in total) where the authors videocapture and comment on major workflows.

In short, the manuscript describes a valuable resource for the community of tomography practitioners.

Strengths:

A public repository of segmentation models; easiness of working with several models and comparing/merging the results.

Weaknesses:

A certain lack of concretion when describing the overall features of the software that differentiate it from others.

Reviewer #3 (Public Review):

Summary:

In this manuscript, Last and colleagues describe Ais, an open-source software package for the semi-automated segmentation of cryo-electron tomography (cryo-ET) maps. Specifically, Ais provides a graphical user interface (GUI) for the manual segmentation and annotation of specific features of interest. These manual annotations are then used as input ground-truth data for training a convolutional neural network (CNN) model, which can then be used for automatic segmentation. Ais provides the option of several CNNs so that users can compare their performance on their structures of interest in order to determine the CNN that best suits their needs. Additionally, pre-trained models can be uploaded and shared to an online database.

Algorithms are also provided to characterize "model interactions" which allows users to define heuristic rules on how the different segmentations interact. For instance, a membrane-adjacent protein can have rules where it must colocalize a certain distance away from a membrane segmentation. Such rules can help reduce false positives; as in the case above, false negatives predicted away from membranes are eliminated.

The authors then show how Ais can be used for particle picking and subsequent subtomogram averaging and for the segmentation of cellular tomograms for visual analysis. For subtomogram averaging, they used a previously published dataset and compared the averages of their automated picking with the published manual picking. Analysis of cellular tomogram segmentation was primarily visual.

Strengths:

CNN-based segmentation of cryo-ET data is a rapidly developing area of research, as it promises substantially faster results than manual segmentation as well as the possibility for higher accuracy. However, this field is still very much in the development and the overall performance of these approaches, even across different algorithms, still leaves much to be desired. In this context, I think Ais is an interesting package, as it aims to provide both new and experienced users with streamlined approaches for manual annotation, access to a number of CNNs, and methods to refine the outputs of CNN models against each other. I think this can be quite useful for users, particularly as these methods develop.

Weaknesses:

Whilst overall I am enthusiastic about this manuscript, I still have a number of comments:

On page 5, paragraph 1, there is a discussion on human judgement of these results. I think a more detailed discussion is required here, as from looking at the figures, I don't know that I agree with the authors' statement that Pix2pix is better. I acknowledge that this is extremely subjective, which is the problem. I think that a manual segmentation should also be shown in a figure so that the reader has a better way to gauge the performance of the automated segmentation.

On page 7, the authors mention terms such as "emit" and "absorb" but never properly define them, such that I feel like I'm guessing at their meaning. Precise definitions of these terms should be provided.

For Figure 3, it's unclear if the parent models shown (particularly the carbon model) are binary or not. The figure looks to be grey values, which would imply that it's the visualization of some prediction score. If so, how is this thresholded? This can also be made clearer in the text.

Figure 3D was produced in ChimeraX using the hide dust function. I think some discussion on the nature of this "dust" is in order, e.g. how much is there and how large does it need to be to be considered dust? Given that these segmentations can be used for particle picking, this seems like it may be a major contributor to false positives.

Page 9 contains the following sentence: "After selecting these values, we then launched a batch particle picking process to determine lists of particle coordinates based on the segmented volumes." Given how important this is, I feel like this requires significant description, e.g. how are densities thresholded, how are centers determined, and what if there are overlapping segmentations?

The FSC shown in Figure S6 for the auto-picked maps is concerning. First, a horizontal line at FSC = 0 should be added. It seems that starting at a frequency of ~0.045, the FSC of the autopicked map increases above zero and stays there. Since this is not present in the FSC of the manually picked averages, this suggests the automatic approach is also finding some sort of consistent features. This needs to be discussed.

Page 11 contains the statement "the segmented volumes found no immediately apparent false positive predictions of these pores". This is quite subjective and I don't know that I agree with this assessment. Unless the authors decide to quantify this through subtomogram classification, I don't think this statement is appropriate.

In the methods, the authors note that particle picking is explained in detail in the online documentation. Given that this is a key feature of this software, such an explanation should be in the manuscript.

eLife assessment

The authors further corroborated their model that Netrin signaling promotes survival and dissemination of non-proliferating ovarian cancer cells. These valuable results were found to be of significant potential interest to cancer biologists in as much as they address gaps in knowledge pertinent to the mechanisms underpinning ovarian cancer spread. In general, it was thought that solid experimental evidence was provided to support the role of Netrin signaling in fueling ovarian cancer progression.

Joint Public Review:

In this article, the authors employed modified CRISPR screens ["guide-only (GO)-CRISPR"] in the attempt to identify the genes which may mediate cancer cell dormancy in the high grade serous ovarian cancer (HGSOC) spheroid culture models. Using this approach, they observed that abrogation of several of the components of the netrin (e.g., DCC, UNC5Hs) and MAPK pathways compromise survival of non-proliferative ovarian cancer cells. This strategy was complemented by the RNAseq approach which revealed that number of the components of the netrin pathway are upregulated in non-proliferative ovarian cancer cells, and that their overexpression is lost upon disruption of DYRK1A kinase that has been previously demonstrated to play a major role in survival of these cells. Perampalam et al. then employed a battery of cell biology approaches to support the model whereby the Netrin signaling governs the MEK-ERK axis to support survival of non-proliferative ovarian cancer cells. Moreover, the authors show that overexpression of Netrins 1 and 3 bolsters dissemination of ovarian cancer cells in the xenograft mouse model, while also providing evidence that high levels of the aforementioned factors are associated with poor prognosis of HGSOC patients.

Strengths:

In this valuable study Perampalam et al. developed a CRISPR-based screening approach to identify key genes that are enriched in high grade serous ovarian cancer spheroids. This led to a discovery that Netrin signaling plays a prominent role in survival of ovarian cancer cells. During revision, the authors provide additional evidence to support their central claims and to this end, it was found that they now provide solid evidence to substantiate the proposed model. This work is anticipated to be of interest to cancer biologists specializing in ovarian cancer biology.

Author response:

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

Reviewer #1 (Public Review):

Summary:

Perampalam et al. describe novel methods for genome-wide CRISPR screening to identify and validate genes essential for HGSOC spheroid viability. In this study, they report that Netrin signaling is essential for maintaining disseminated cancer spheroid survival, wherein overexpression of Netrin pathway genes increases tumor burden in a xenograft model of ovarian cancer. They also show that high netrin expression correlates with poor survival outcomes in ovarian cancer patients. The study provides insights into the biology of netrin signaling in DTC cluster survival and warrants development of therapies to block netrin signaling for treating serous ovarian cancer.

Strengths:

- The study identifies Netrin signaling to be important in disseminated cancer spheroid survival

- A Novel GO-CRISPR methodology was used to find key genes and pathways essential for disseminated cancer cell survival

Thanks for the endorsement of our work and its importance to metastasis in ovarian cancer.

Weaknesses:

- The term dormancy is not fully validated and requires additional confirmation to claim the importance of Netrin signaling in "dormant" cancer survival.

- Findings shown in the study largely relate to cancer dissemination and DTS survival rather than cancer dormancy.

Much of the validation of dormancy and cell cycle arrest in HGSOC spheroids, as well as the culture model, have been published previously and hence was not repeated here.  I think this reviewer will appreciate the updated citations and explanations to better illustrate the state of knowledge.  We have also added new experiments that further emphasize the dormant state of spheroid cells in culture and xenografts, as well as patient derived spheroids used in this study.

Reviewer #1 (Recommendations for Authors):

(1) It is unclear what spheroid/adherent enrichment ratio is and how it ties into genes affecting cell viability. Why is an ER below 1 the criteria for selecting survival genes?

Our screen uses the ‘guide only’ comparison in each culture condition to establish a gene score under that specific condition.  A low adherent score captures genes that are essential under standard culture conditions where cells are proliferating and this can include genes needed for proliferation or other basic functions in cell physiology.  A low spheroid score identifies the genes that are most depleted in suspension when cells are growth arrested and this is an indication of cell death in this condition.  Since gene knock outs are first established in adherent proliferating conditions, essential genes under these conditions will already start to become depleted from the population before suspension culture.  By selecting genes with a ratio of <1 we can identify those that are most relevant to dormant suspension culture conditions.  Ultimately, the lowest enrichment ratio scores represent genes whose loss of function is dispensable in the initial adherent condition, but critical for survival in suspension and this is what we aimed to identify. We’ve updated Figure 1B to illustrate this and we’ve updated the explanation of the enrichment ratio on page 6, lines 144 to 147 of the results.

(2) The WB for phospho-p38 in figure 1A for OVCAR8 line does not show increased phosphorylation in the spheroid relative to the adherent. If anything, phospho-p38 appears to be reduced in the spheroid. Can the authors provide a better western blot?

We’ve updated this blot with a longer exposure, see Figure 1A.  Phosphorylation levels of p38 are essentially unchanged in OVCAR8 cells in suspension culture, although the overall levels of p38 may be slightly reduced in dormant culture conditions.

(3) How did the authors confirm dormancy apart from western blot for phospho-ERK vs phospho-p38? Authors should add EdU/BrdU staining and/or Ki67 staining to confirm dormancy.

Previous publications that appear as citations 7,10, and 33 in the reference list established the growth arrest state of these cells in suspension culture in the past.  This included measuring other known markers of dormancy and quiescence such as p27, p130, and reduced cyclin/cdk activity and 3H-thymidine incorporation. In addition, other associated characteristics of dormancy such as EMT and catabolic metabolism have been demonstrated in these culture conditions (see citation 11 and Rafehi et al. Endocr. Relat. Cancer 23;147-59).  We’ve added these additional citations to our descriptions of dormant spheroid culture to better clarify the status of these cells in our experiments (see page 6, lines 126-28).  To ensure that cells are growth arrested in the experiments shown in this paper, we have updated Figure 1A to include blots of p130 and Ki67 to further emphasize that spheroid cells are not proliferating as the quiescence marker (p130) is high and the proliferative marker (Ki67) is lost in suspension culture.

(4) Can the authors report spheroid volume over time in culture? How was viability measured?

We’ve updated the methods (see page 27, line 574) to better highlight the description of cell survival that answers both of these questions. At the ends of experimental time points in both the screen and viability assays we captured live cells by replating on adherent plasticware. We fixed and stained with crystal violet and photographed plates to illustrate the sizes of spheroids (shown in Fig. 2 Supplement 1E, Fig. 6C, and 7D). We subsequently extracted the dye and quantitated it spectrophotometrically to quantitatively compare biomass of viable cells between experiments irrespective of the relatively random shapes of spheroids. We found reattachment and staining in this manner to match traditional viability assays such as CellTiter-Glo in a previous paper (10). Furthermore, biomass never increases in culture and diminishes gradually over time in culture consistent with the non-proliferative state of these experiments. Double checks of this equivalency of viability and reattached biomass measurments, as well as demonstrating that biomass is lost over time, are shown in Fig. 2 Supplement 1E that compares reattached crystal violet staining measurements with CellTiter-Glo for DYRK1A knock out cells over time in culture. In addition, we include a comparison of crystal violet staining of reattached spheroids with trypan blue dye exclusion in Fig. 5G and H. In both cases reattachment and more direct viability assays demonstrate the same conclusion that Netrin signaling supports viability in dormant culture.

(5) Please show survival significance of Netrin signaling genes in recurrence/relapse free survival to claim importance in cancer dormancy.

See Fig. 7 Supplement 1C where we include the recurrence free survival data. Netrin-1, and -3 high expressors also have a numerically shorter progression free survival but it is not statistically significant. Netrin-1 overexpression alone is also shown and it shows shorter survival with a P-value of 0.0735. Elevated survival of dormant cells in a residual disease state is expected to increase the chance of relapse and shorten this interval. Thus, this data is consistent with our model, but lacks statistical significance. 

There are many alternative ways to interpret what shorter progression free survival, or overall survival, may mean biologically. Since survival of dormant cells is but one of them, we also added new data to experimentally investigate the role of endogenous Netrin signaling in dormant residual disease in Fig. 6 and described on page 12, lines 266-87.  We used xenograft experiments to show OVCAR8 spheroids form and withdraw from the cell cycle equivalently to suspension culture following intraperitoneal injection.  Furthermore, loss of Netrin signaling due to receptor deletions compromises survival during this early window before disseminated lesions form.  This argues that Netrin signaling contributes to survival during this window of dormancy.  In addition, mice engrafted with mutant cells experience prolonged survival when Netrin signaling is blocked.  Together, these experiments further argue that Netrin signaling supports survival in the dormant, non-proliferative phase, and leads to reduced survival of mice.

(6) The authors show IHC staining of patient ascites derived HGSOC spheroids. However, no marker for dormancy is shown in these spheroids. Adding Ki67 staining or phospho-ERK vs phospho-p38 would be necessary to confirm cancer dormancy.

We have added new staining for Ki67 and p130 that compares these markers in HGSOC tumors where Ki67 is high and p130 is low with ascites derived spheroids where staining is the opposite. Importantly, expression of p130 is linked to cellular quiescence and is not found to accumulate in the nucleus of cells that are just transiting through G1.  This confirms that the ascites derived spheroids are dormant.  See Fig. 4A-E and described on page 9, lines 201-7.

(7) Overall, the findings are interesting in the context of cancer dissemination. There is not enough evidence for cancer dormancy and the importance of Netrin signaling in the survival of cancer dormancy. Overexpression of Netrin increases phosphorylation of ERK, leading one to expect an increase in proliferation. This suggests that Netrin breaks cancer cells out of dormancy, into a proliferative state.

We have found that the discovery of Netrin activation of MEK-ERK in growth arrested cells is counterintuitive to many cancer researchers.  However, this axis exists in other paradigms of Netrin signaling in axon outgrowth that are not proliferation related (see citation 26, Forcet et al. Nature 417; 443-7 as an example).  We have added Fig. 5D and descriptions on page 11, lines 244-52 to better clarify that Netrins CAN’T induce cell proliferation through ERK.  Addition of recombinant Netrin-1 can only induce ERK phosphorylation in suspension culture conditions and not in quiescent adherent conditions.  The small magnitude of ERK phosphorylation induced by Netrin-1 in suspension compared to treating adherent, quiescent cells with the same concentration of mitogenic EGF further emphasizes that this is not a proliferative signal.  Lastly, the new xenograft experiment in Fig. 6A-D (described on page 12, lines 266-81 demonstrates the growth arrested context in which Netrin signaling in dormant spheroids leads supports viability.

(8) If authors wish to claim cancer dormancy as the premise of their study, additional confirmatory experiments are required to support their claims. Alternatively, based on the current findings of the study, it would be best to change the premise of the article to Netrin signaling in cancer dissemination and survival of disseminated cancer spheroids rather than cancer dormancy.

I expect that this reviewer will agree that we have added more than sufficient explanations of background work on HGSOC spheroid dormancy from the literature, as well as new experiments that address their questions about dormancy in our experiments.

Reviewer #2 (Public Review):

Summary:

In this article, the authors employed modified CRISPR screens ["guide-only (GO)-CRISPR"] in the attempt to identify the genes which may mediate cancer cell dormancy in the high grade serous ovarian cancer (HGSOC) spheroid culture models. Using this approach, they observed that abrogation of several of the components of the netrin (e.g., DCC, UNC5Hs) and MAPK pathways compromise the survival of non-proliferative ovarian cancer cells. This strategy was complemented by the RNAseq approach which revealed that a number of the components of the netrin pathway are upregulated in non-proliferative ovarian cancer cells and that their overexpression is lost upon disruption of DYRK1A kinase that has been previously demonstrated to play a major role in survival of these cells. Perampalam et al. then employed a battery of cell biology approaches to support the model whereby the Netrin signaling governs the MEK-ERK axis to support survival of non-proliferative ovarian cancer cells. Moreover, the authors show that overexpression of Netrins 1 and 3 bolsters dissemination of ovarian cancer cells in the xenograft mouse model, while also providing evidence that high levels of the aforementioned factors are associated with poor prognosis of HGSOC patients.

Strengths:

Overall it was thought that this study is of potentially broad interest in as much as it provides previously unappreciated insights into the potential molecular underpinnings of cancer cell dormancy, which has been associated with therapy resistance, disease dissemination, and relapse as well as poor prognosis. Notwithstanding the potential limitations of cellular models in mimicking cancer cell dormancy, it was thought that the authors provided sufficient support for their model that netrin signaling drives survival of non-proliferating ovarian cancer cells and their dissemination. Collectively, it was thought that these findings hold a promise to significantly contribute to the understanding of the molecular mechanisms of cancer cell dormancy and in the long term may provide a molecular basis to address this emerging major issue in the clinical practice.

Thanks for the kind words about the importance of our work in the broader challenges of cancer treatment.

Weaknesses:

Several issues were observed regarding methodology and data interpretation. The major concerns were related to the reliability of modelling cancer cell dormancy. To this end, it was relatively hard to appreciate how the employed spheroid model allows to distinguish between dormant and e.g., quiescent or even senescent cells. This was in contrast to solid evidence that netrin signaling stimulates abdominal dissemination of ovarian cancer cells in the mouse xenograft and their survival in organoid culture. Moreover, the role of ERK in mediating the effects of netrin signaling in the context of the survival of non-proliferative ovarian cancer cells was found to be somewhat underdeveloped.

Experiments previously published in citation 7 show that growth arrest in patient ascites derived spheroids is fully reversible and that argued against non-proliferative spheroids being a form of senescence and moved this work into the dormancy field.  We have added extensive new support for our model systems and data to address the counterintuitive aspects of MEK-ERK signaling in survival instead of proliferation. 

Reviewer #1 Recommendations for Authors

(1) A better characterization of the spheroid model may be warranted, including staining for the markers of quiescence and senescence (including combining these markers with staining for the components of the netrin pathway)

See Figure 1A and page 6, lines 126-36 where we have added blots for Ki67 and p130 to better emphasize the arrested proliferative state of cells in our screening conditions.  We have also added these same controls for patient ascites-derived spheroids in Figure 4 and described on page 9, lines 203-7.  One realization from this CRISPR screen, and others in our lab, is that it identifies functionally important aspects of cell physiology and not necessarily ones that are easily explored using commercially available antibodies.  Netrin-1 and -3 staining of patient derived spheroids in Fig. 4, as well as cell line spheroids stained in Fig. 4 Supplement 1 further support the relevance of this pathway in dormant cancer cells because Netrins are expressed in the right place at the right time.  The Netrin-1 stimulation experiments in Fig. 5C were originally carried out to probe HGSOC cells for functionality of Netrin receptors since we couldn’t reliably detected them by blotting or staining with available antibodies.  This demonstrates that this pathway is active in the various HGSOC cell lines we’ve used and specifically, using OVCAR8 cells, we show it is only active in suspension culture conditions.

(2) In figure 1A it appears that total p38 levels are reduced in some cell lines in spheroid vs. adherent culture. The authors should comment on this.

These blots have been updated to be more clear.  Overall p38 levels may be reduced in some cell lines and when compared with activation levels of phosphorylated p38 it suggests the fraction of activated p38 is higher. OVCAR8 cells may be an exception where the overall activity level remains approximately the same.

(3) The authors should perhaps provide a clearer rationale for choosing to focus on the netrin signaling vs. e.g., GPCR signaling, and consider more explicit defining of "primary" vs. "tertiary" categories in Reactome gene set analysis.

We’ve updated Fig. 1E and the text on page7, lines 161-5 to illustrate which gene categories identified in the screen belong to which tiers of Reactome categories. It better visualizes why we have investigated the Axon guidance pathway that includes Netrin because it is a highly specific signaling pathway that scores similarly to the broader and less specific categories at the very top of the list. As an aside, the GPCR signaling and GPCR downstream signaling have proven to be fairly intractable categories.  As best we can tell the GPCR downstream signaling category is full of MAPK family members and likely represents some redundancy with MAPK further down.  

(4) In figure 3A-C, including factors whose expression did not appear to change between adherent and suspension conditions may be warranted as the internal control. Figure 3D-F may benefit from some sort of quantification.

The mRNA expression levels are normalized to GAPDH as an internal control. We have updated this figure and re-plotted it as fold change relative to adherent culture cells with statistical comparisons to indicate which are significantly upregulated in suspension culture.

The IHC experiments are now in Fig. 4D-F and show positive staining for Netrin-1 and -3.  Netrin-3 is easiest to see, while Netrin-1 is trickier because the difference with the no primary antibody control isn’t intensity, but the tint of the DAB stain.  We had to counter stain the patient spheroids with Hematoxylin in order for the slide scanner to find the best focal plane and make image registration between sections possible.  This unfortunately makes the Netrin-1 staining rather subtle.  For cell line spheroids in the Fig. 4, Supplement 1 we didn’t need the slide scanner and show negative controls without counter stain that are much more convincing of Netrin-1 detection and reassure us that our staining detects the intended target.  We’ve updated the labels in Fig. 4 and Fig. 4, Supplement 1 for this to be more intuitive.  Unfortunately, relying on the tint of the DAB stain leaves this as a qualitative experiment.

- In figure 4C-E the authors show that Netrin-1 stimulation induces ERK phosphorylation whereby it is argued that this is a "low-level" stimulation of ERK signaling required for the survival of ovarian cells in the suspension. This is however hard to appreciate, and it was thought that having adherent cells in parallel would be helpful to wage whether this indeed is a "low level" ERK activity. Moreover, the authors should likely include downstream substrates of ERK (e.g., RSKs) as well as p38 in these experiments. The control experiments for the effects of PD184352 on ERK phosphorylation also appear to be warranted. Finally, performing the experiments with PD184352 in the presence of Netrin-1 stimulation would also be advantageous.

We have added a new Netrin-1 stimulation experiment in Fig. 4D (described on page 11, line 244-52) that shows that Netrins can only activate  very low levels of ERK phosphorylation in suspension when proliferation is arrested. Netrin-1 stimulation of quiescent adherent cells where stimulation of proliferation is possible shows that Netrins are unable to activate ERK phosphorylation in this condition.  In contrast, we also stimulate quiescent adherent OVCAR8 cells with an equal concentration of EGF (a known mitogen) to offer high level ERK phosphorylation as a side by side comparison.  I think that this offers clear evidence that Netrin signaling is inconsistent with inducing cell proliferation.  We’ve also updated citations in the introduction to include citation 26 that offers a previously reported paradigm of Netrin-ERK signaling in axon outgrowth that is a non-cancer, non-proliferative context to remind readers that Netrins utilize MEK-ERK differently. 

We highlight Netrin-MEK-ERK signaling as key to survival for a number of reasons.  First, Netrin signaling in this paradigm does not fit the dependence receptor paradigm where loss of Netrin receptors protect against cell death.  Fig. 5B rules this out as receptor loss never offers a survival advantage, but clearly receptor deletions compromise survival in suspension culture.  Second, positive Netrin signaling is known to support survival by inactivating phosphorylation of DAPK1.  We’ve added this experiment as Fig. 5 Supplement 1D and show that loss of Netrin receptors doesn’t reduce DAPK1 phosphorylation in a time course of suspension culture.  Consequently, we conclude this isn’t the survival signal either.  Since MEK and ERK family members scored in our screen, we investigated their role in survival.  We now show two different MEK inhibitors with different inhibitory mechanisms to confirm that MEK inhibition induces cell death. In addition to the previous PD184352 inhibitor in our first submission, we’ve added Trametinib as well and this is shown in Fig. 5G.  Since it is surprising the MEK inhibition can kill instead of just arrest proliferation, we’ve also added another cell death assay in which we show trypan blue dye exclusion as a second look at survival.  This is now Fig. 5H.  Lastly, we include Trametinib inhibition of ERK phosphorylation in these assays in Fig. 5I.  While we leave open what takes place downstream of ERK, our model in Fig. 5J offers a very detailed look at the components upstream.

- Does inhibition of ERK prevent the abdominal spread of ovarian cancer cells? The authors may feel that this is out of the scope of the study, which I would agree with, but then the claims regarding ERK being the major mediator of the effects of netrin signaling should be perhaps slightly toned down.

We agree that loss of function xenograft experiments will enhance our discovery of Netrin’s role in dormancy and metastasis.  We have added a new Fig. 6 that uses xenografts with Netrin receptor deficient OVCAR8 cells (UNC5 4KO).  It demonstrates that two weeks following IP engraftment we can isolate spheroids from abdominal washes and that cells have entered a state of reduced proliferation as determined by lowered Ki67 expression as well as other proliferation inducing genes.  In the case of UNC5 4KO cells, there is significant attrition of these cells as determined by recovering spheroids in adherent culture (Fig.6C) and by Alu PCR to detect human cells in abdominal washes (Fig. 6D).  Lastly, xenografts of UNC5 4KO cells cause much less aggressive disease and significantly extend survival of these mice (Fig. 6E,F).  Not exactly the experiment that the reviewer is asking for, but a clear indication that Netrin signaling supports survival in xenograft model of dormancy.

- Notwithstanding that this could be deduced from figures 6D and F, it would be helpful if the number of mice used in each experimental group is clearly annotated in the corresponding figure legends. Moreover, indicating the precise statistical tests that were used in the figures would be helpful (e.g., specifying whether anova is one-way, two-way, or?)

We have added labels to what is now Fig. 8B to indicate the number of animals used for each genotype of cells.  We have also updated figure legends to include more details of statistical tests used in each instance.

Reviewer #2 (Public Review):

Summary:

The manuscript by Gitanjali Roy et al. applies deep transfer learning (DEGAS) to assign patient-level disease attributes (metadata) to single cells of T2D and non-diabetic patients, including obese patients. This led to the identification of a singular cluster of T2D-associated β-cells; and two subpopulations of obese- β-cells derived from either non-diabetic or T2D donors. The objective was to identify novel and established genes implicated in T2D and obesity. Their final goal is to validate their findings at the protein level using immunohistochemistry of pancreas tissue from non-diabetic and T2D organ donors.

Strengths:

This paper is well-written, and the findings are relevant for β-cell heterogeneity in T2D and obesity.

Weaknesses:

The validation they provide is not sufficiently strong: no DLK1 immunohistochemistry is shown of obese patient-derived sections. Additional presumptive relevant candidates from this transcriptomic analysis should be screened for, at the protein level.

eLife assessment

This is a useful study that used DEGAS, a deep transfer learning tool, to identify distinct pancreatic beta cell subpopulations that could be associated with type 2 diabetes (T2D) and/or obesity status. The data supporting the authors' findings is solid and demonstrates that DEGAS will be a helpful tool for analyzing cell-specific transcriptomic phenotypes. This study will be of interest to researchers studying the genetics of T2D.

Reviewer #1 (Public Review):

In this manuscript, Roy et al. used the previously published deep transfer learning tool, DEGAS, to map disease associations onto single-cell RNA-seq data from bulk expression data. The authors performed independent runs of DEGAS using T2D or obesity status and identified distinct β-cell subpopulations. β-cells with high obese-DEGAS scores contained two subpopulations derived largely from either non-diabetic or T2D donors. Finally, immunostaining using human pancreas sections from healthy and T2D donors validated the heterogeneous expression and depletion of DLK1 in T2D islets.

Strengths:

(1) This meta-analysis of previously published scRNA-seq data using a deep transfer learning tool.

(2) Identification of novel beta cell subclusters.

(3) Identified a relatively innovative role of DLK1 in T2D disease progression.

Weaknesses:

(1) There is little overlap of the DE list of bulk RNA-seq analysis in Figure 1D and 1E overlap with the DE list of pseudo-bulk RNA-seq analysis of all cells in Figure S2C.

(2) The biological meaning of "beta cells had the lowest scores compared to other cell types" is not clear.

(3) The figures and supplemental figures were not cited following the sequence, which makes the manuscript very difficult to read. Some supplemental figures, such as Figures S1C-S1D, S2B-S2E, S3A-S3B, were not cited or mentioned in the text.

(4) In Figure 7, the current resolution is too low to determine the localization of DLK1.

Author response:

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Roy et al. used the previously published deep transfer learning tool, DEGAS, to map disease associations onto single-cell RNA-seq data from bulk expression data. The authors performed independent runs of DEGAS using T2D or obesity status and identified distinct β-cell subpopulations. β-cells with high obese-DEGAS scores contained two subpopulations derived largely from either non-diabetic or T2D donors. Finally, immunostaining using human pancreas sections from healthy and T2D donors validated the heterogeneous expression and depletion of DLK1 in T2D islets.

Strengths:

(1) This meta-analysis of previously published scRNA-seq data using a deep transfer learning tool.

(2) Identification of novel beta cell subclusters.

(3) Identified a relatively innovative role of DLK1 in T2D disease progression.

We thank the reviewer for their constructive critiques and positive feedback. We hope to further apply deep transfer learning tools in future scRNA-seq meta-analyses.

Weaknesses:

(1) There is little overlap of the DE list of bulk RNA-seq analysis in Figure 1D and 1E overlap with the DE list of pseudo-bulk RNA-seq analysis of all cells in Figure S2C.

We thank the reviewer for this insightful thought and plan to perform additional analyses and comparisons to address this comment.

(2) The biological meaning of "beta cells had the lowest scores compared to other cell types" is not clear.

We agree with the reviewer and will amend this statement to clarify in the revised manuscript. In summary, the relatively lower T2D-DEGAS scores for beta cells overall compared to all other cell types (alpha cells, acinar cells, etc) reflects the fact that in T2D, beta cell-specific genes can be downregulated. This is also possibly due to beta cell loss in T2D and would be reflected in bulk islet RNAseq data. This affects the DEGAS model which is reflected in the scores of all cells in the scRNA-seq data (Fig 3A). For this reason, subsetting the beta cells and replotting them on their own (Fig 4B) is an important step to identify relative differences in DEGAS scores between different subsets of beta cells.

(3) The figures and supplemental figures were not cited following the sequence, which makes the manuscript very difficult to read. Some supplemental figures, such as Figures S1C-S1D, S2B-S2E, S3A-S3B, were not cited or mentioned in the text.

We apologize and thank the reviewer for pointing out these errors. All of the annotated errors will be amended in the revised manuscript.

(4) In Figure 7, the current resolution is too low to determine the localization of DLK1.

We will include the original highest-resolution confocal images in our resubmission. We will also improve the color combination to improve visibility of colocalization of DLK1 with Insulin.

Reviewer #2 (Public Review):

Summary:

The manuscript by Gitanjali Roy et al. applies deep transfer learning (DEGAS) to assign patient-level disease attributes (metadata) to single cells of T2D and non-diabetic patients, including obese patients. This led to the identification of a singular cluster of T2D-associated β-cells; and two subpopulations of obese- β-cells derived from either non-diabetic or T2D donors. The objective was to identify novel and established genes implicated in T2D and obesity. Their final goal is to validate their findings at the protein level using immunohistochemistry of pancreas tissue from non-diabetic and T2D organ donors.

Strengths:

This paper is well-written, and the findings are relevant for β-cell heterogeneity in T2D and obesity.

We thank the reviewer for their constructive critiques and positive feedback. We believe this study can improve our understanding β-cell heterogeneity in the context of T2D and obesity.

Weaknesses:

The validation they provide is not sufficiently strong: no DLK1 immunohistochemistry is shown of obese patient-derived sections. Additional presumptive relevant candidates from this transcriptomic analysis should be screened for, at the protein level.

Thank the reviewer for this suggestion. We are planning to perform new immunostaining of DLK1 in human pancreas tissue sections from non-diabetic lean, non-diabetic obese, T2D lean, and T2D obese donors. We also note that Table S6 contains the patient metadata for the pancreas samples we show in the current manuscript. Two of the T2D donors have BMI > 30 (obese). However, the non-diabetic donors have BMI between 26-29. Our new planned studies should address the question of differential DLK1 expression / beta cell heterogeneity in the context of both diabetes and obesity.

eLife assessment

ProtSSN is a valuable approach that generates protein embeddings by integrating sequence and structural information, demonstrating improved prediction of mutation effects on thermostability compared to sequence-only models. The work is currently incomplete as it lacks a thorough comparison against other recent top-performing methods that also incorporate structural data, such as SaProt, EVE-based models, and GEMME. Providing a comprehensive analysis benchmarking ProtSSN against these state-of-the-art structure-based approaches would significantly strengthen the evidence supporting the utility of ProtSSN's joint sequence-structure representations.

Reviewer #1 (Public Review):

Summary:

The authors introduce a denoising-style model that incorporates both structure and primary-sequence embeddings to generate richer embeddings of peptides. My understanding is that the authors use ESM for the primary sequence embeddings, take resolved structures (or use structural predictions from AlphaFold when they're not available), and then develop an architecture to combine these two with a loss that seems reminiscent of diffusion models or masked language model approaches. The embeddings can be viewed as ensemble-style embedding of the two levels of sequence information, or with AlphaFold, an ensemble of two methods (ESM+AlphaFold). The authors also gather external datasets to evaluate their approach and compare it to previous approaches. The approach seems promising and appears to out-compete previous methods at several tasks. Nonetheless, I have strong concerns about a lack of verbosity as well as the exclusion of relevant methods and references.

Advances:

I appreciate the breadth of the analysis and comparisons to other methods. The authors separate tasks, models, and sizes of models in an intuitive, easy-to-read fashion that I find valuable for selecting a method for embedding peptides. Moreover, the authors gather two datasets for evaluating embeddings' utility for predicting thermostability. Overall, the work should be helpful for the field as more groups choose methods/pretraining strategies amenable to their goals, and can do so in an evidence-guided manner.

Considerations:

Primarily, a majority of the results and conclusions (e.g., Table 3) are reached using data and methods from ProteinGym, yet the best-performing methods on ProteinGym are excluded from the paper (e.g., EVE-based models and GEMME). In the ProteinGym database, these methods outperform ProtSSN models. Moreover, these models were published over a year---or even 4 years in the case of GEMME---before ProtSSN, and I do not see justification for their exclusion in the text.

Secondly, related to the comparison of other models, there is no section in the methods about how other models were used, or how their scores were computed. When comparing these models, I think it's crucial that there are explicit derivations or explanations for the exact task used for scoring each method. In other words, if the pre-training is indeed an important advance of the paper, the paper needs to show this more explicitly by explaining exactly which components of the model (and previous models) are used for evaluation. Are the authors extracting the final hidden layer representations of the model, treating these as features, and then using these features in a regression task to predict fitness/thermostability/DDG etc.? How are the model embeddings of other methods being used, since, for example, many of these methods output a k-dimensional embedding of a given sequence, rather than one single score that can be correlated with some fitness/functional metric? Summarily, I think the text lacks an explicit mention of how these embeddings are being summarized or used, as well as how this compares to the model presented.

I think the above issues can mainly be addressed by considering and incorporating points from Li et al. 2024[1] and potentially Tang & Koo 2024[2]. Li et al.[1] make extremely explicit the use of pretraining for downstream prediction tasks. Moreover, they benchmark pretraining strategies explicitly on thermostability (one of the main considerations in the submitted manuscript), yet there is no mention of this work nor the dataset used (FLIP (Dallago et al., 2021)) in this current work. I think a reference and discussion of [1] is critical, and I would also like to see comparisons in line with [1], as [1] is very clear about what features from pretraining are used, and how. If the comparisons with previous methods were done in this fashion, this level of detail needs to be included in the text.

To conclude, I think the manuscript would benefit substantially from a more thorough comparison of previous methods. Maybe one way of doing this is following [1] or [2], and using the final embeddings of each method for a variety of regression tasks---to really make clear where these methods are performing relative to one another. I think a more thorough methods section detailing how previous methods did their scoring is also important. Lastly, TranceptEVE (or a model comparable to it) and GEMME should also be mentioned in these results, or at the bare minimum, be given justification for their absence.

[1] Feature Reuse and Scaling: Understanding Transfer Learning with Protein Language Models<br /> Francesca-Zhoufan Li, Ava P. Amini, Yisong Yue, Kevin K. Yang, Alex X. Lu<br /> bioRxiv 2024.02.05.578959; doi: https://doi.org/10.1101/2024.02.05.578959

[2] Evaluating the representational power of pre-trained DNA language models for regulatory genomics<br /> Ziqi Tang, Peter K Koo<br /> bioRxiv 2024.02.29.582810; doi: https://doi.org/10.1101/2024.02.29.582810

Reviewer #2 (Public Review):

Summary:

To design proteins and predict disease, we want to predict the effects of mutations on the function of a protein. To make these predictions, biologists have long turned to statistical models that learn patterns that are conserved across evolution. There is potential to improve our predictions however by incorporating structure. In this paper, the authors build a denoising auto-encoder model that incorporates sequence and structure to predict mutation effects. The model is trained to predict the sequence of a protein given its perturbed sequence and structure. The authors demonstrate that this model is able to predict the effects of mutations better than sequence-only models.

As well, the authors curate a set of assays measuring the effect of mutations on thermostability. They demonstrate their model also predicts the effects of these mutations better than previous models and make this benchmark available for the community.

Strengths:

The authors describe a method that makes accurate mutation effect predictions by informing its predictions with structure.

Weaknesses:

It is unclear how this model compares to other methods of incorporating structure into models of biological sequences, most notably SaProt (https://www.biorxiv.org/content/10.1101/2023.10.01.560349v1.full.pdf).

ProteinGym is largely made of deep mutational scans, which measure the effect of every mutation on a protein. These new benchmarks contain on average measurements of less than a percent of all possible point mutations of their respective proteins. It is unclear what sorts of protein regions these mutations are more likely to lie in; therefore it is challenging to make conclusions about what a model has necessarily learned based on its score on this benchmark. For example, several assays in this new benchmark seem to be similar to each other, such as four assays on ubiquitin performed at pH 2.25 to pH 3.0.

Reviewer #1 (Public Review):

Summary:

The authors demonstrate that the immunosuppressive environment in pancreatic ductal adenocarcinoma (PDAC) can be mitigated by a combination of ionizing radiation (IR), CCR5 inhibition, and PD1 blockade. This combination therapy increases tissue-resident natural killer (trNK) cells that facilitate CD8 T cell activity, resulting in a reduction of E-cadherin positive tumor cells. They identify a specific "hypofunctional" NK cell population in both mouse and human PDAC that supports CD8 T cell involvement. A trNK signature is found to be associated with better survival outcomes in PDAC and other solid tumors.

Overall, I think this is an interesting study that combines testing of therapeutic concepts in mice with bioinformatics analysis of single cell transcriptome data in primary tumors and exploration of clinical outcomes using signature genes in TCGA data. The key finding is that immunoregulatory properties of tumor infiltrating/resident CD56-bright NK cells (assumed to be non-cytotoxic) are beneficial for outcome through cross-talk with DC and recruitment of CD8 T cells. The latter is specifically induced by irradiation combined with CCR5i and PD1 blockade.

These results support the notion that IR/CCR5i/αPD1 combination treatment alters immune infiltration by reducing Tregs and increasing NK and CD8 T cells, thereby resulting in greater local tumor control.

Although the language was slightly modified in the revised version I think it is important to point out that transcripts (not protein expression) of KLRC2 is common in CD56bright NK cells and does not really reflect "adaptive-like" NK cells.

eLife assessment

This valuable manuscript provides an interesting account documenting the role of resident CD56(br) NK cells in driving interaction with dendritic cells that attract CD8+ T cells to the pancreas cancer tumor microenvironment (TME). The work convincingly illustrates how irradiation combined with CCR5i and PD1 blockade leads to a reduction in pancreatic cancer growth that correlates with a reduction in Treg cells and enhancement of NK and CD8 T cells in the TME. The correlation of NKC1 signature with survival in pancreatic cancer patients is indeed of broader interest regarding potential relevance to other types of cancer.

Reviewer #2 (Public Review):

Summary:

This work elaborates on a combined therapeutic approach comprising ionizing radiation and CCR5i/αPD1 immunotherapy as a promising strategy in pancreatic cancer. Previous research has established that NK cell-derived CCL5 and XCL1 play a crucial role in recruiting cDC1 cells to the tumor microenvironment, contributing to tumor control. In this study, by using a murine pancreatic cancer model, the authors propose that the addition of radiation therapy to CCR5i and αPD1 immunotherapy could upregulate CD8+ T cells and a subgroup of NK cells within the tumor and result in better tumor control. They further analyzed human single-cell sequencing data from pancreatic cancer patients and identified one subgroup of NK cells (NK C1) with tissue-resident features. Subsequent cell-cell contact analysis reveals the NK-cDC1-CD8 cell axis in pancreatic cancer. By analyzing TCGA data, they found that high NK C1 signature levels were associated with better survival in pancreatic cancer patients. Thus, radiotherapy could benefit the outcome of patients bearing low NK C1 signatures. Importantly, the positive correlation between NK C1 score with survival extends beyond pancreatic cancer, showing potential applicability across various solid cancers.

Strengths:

This study could add new insight into the clinical practice by introducing such novel combined therapy and shed light on the underlying immune cell dynamics. These findings hold potential for more effective and targeted treatment in the future. Mouse experiments nicely confirmed that such combined therapy could significantly reduce tumor volume. The elegant use of single-cell sequencing analysis and human database examination enriches the narrative and strengthens the study's foundation. Additionally, the notion that NK C1 signature correlates with patient survival in various solid cancers is of high interest and relevance.

Weaknesses:

The authors have addressed some of my concerns. However, others remain and should be discussed.

(1) The role of CCR5i requires further clarification/ discussion. While the authors demonstrated its capacity to reduce Treg in murine tumors, its impact on other cell populations, including NK cells and CD8+ T cells, was not observed. Nevertheless, the effect of CCR5i on tumor growth in Figure 2B seems pathogenic. If the combination of radiotherapy and αPD1 already can achieve good outcomes as shown in Figure 3A, the necessity to include CCR5i is questioned. Overall, a more comprehensive elucidation of the roles of CCL5 and CCR5i in this context would be good. Alternatively, this limitation should be discussed.<br /> (2) In line with this, spatial plots in Figure 4 did not include the group with only radiotherapy and αPD1. This inclusion would facilitate a clearer comparison and better highlight the essential role of CCR5i.<br /> (3) Human database analysis showed a positive correlation between NK C1 score and CCL5 in pancreatic cancer. Furthermore, radiotherapy could benefit the outcome of patients bearing low NK C1 scores. It would be interesting to test, if radiotherapy could also benefit patients with low CCL5 levels in this cohort. This is a key question since the role of CCL5/CCR5i is not well verified. Alternatively, this point could be mentioned and discussed.

Reviewer #3 (Public Review):

Summary:

In the submitted manuscript by Go et al, the authors evaluated the tumor microenvironment in pancreatic ductal adenocarcinoma (PDAC) and made a number of interesting observations, including the following: 1) CCL5 expression within the tumor microenvironment negatively correlated with clinical outcomes in human patients with PDAC; 2) there were both positive and negative correlations between CCL5 expression and the expression of specific genes (e.g. those encoding CD56 and CD16, respectively) included among gene signature lists for Treg, MDSC, TAM, and NK cells; 3) CCR5 inhibition with the inhibitor, maraviroc, reduced Treg infiltration but not that of other immune cell types in an orthotopic murine model of PDAC; 4) CCR5 inhibition augmented anti-PD1 immunotherapy when combined with ionizing radiation (IR) therapy in the murine model; 5) the above therapy resulted in increased infiltration of CD8+ cytotoxic T cells as well as of a subset of NKG2D-negative, tissue-residency (tr) marker expressing NK cells (deemed Cluster 1 NK in their data sets) that inversely correlated with the number of E-cadherin+ cells (i.e. tumor cells) and showed predicted interactions with cDC1 dendritic cells (including XCL1/XCL2 expressed by the NK and XCR1 expressed by the cDC1); 6) the authors identified a number of putative signals stemming from the trNK (e.g. IL-16, TNFSF14, FASLG, CSF, MIF) as well as incoming from cDC1s to NK (e.g. BAG6-NKp30); 7) these trNK cells positively correlated with good outcomes and with CD8+ T cell infiltrations in human PDAC as well as in many other solid tumor types; and 8) importantly, the benefit of IR therapy was specific to the subset of PDAC patients (represented in the TCGA dataset) that were predicted to have low amounts of trNK cells. The authors used murine experimental models, multi-plexed imaging analyses, and a number of publicly available sequencing data sets from human tumor samples to perform their investigations. Based on their findings, the authors proposed that combining IR with CCR5 inhibition and anti-PD1 immunotherapy is a promising strategy to treat solid cancers.

Strengths:

Overall, the collective analyses and conclusions appear to be novel and could be of high and rapid impact on the field, particularly in terms of directing clinical trials to incorporate IR with CCR5 inhibition and immunotherapy. The manuscript is well written; the figures are for the most part clear; and the Discussion is very thoughtful.

Weaknesses:

In the revised manuscript, the authors addressed my original concerns. I have no new major concerns with the study. One of the limitations is that the authors did not perform functional in vivo or ex vivo assays to address some of the major hypotheses that arose from the descriptive, correlative data; but overall, this does not detract from the enthusiasm for the work or the potential significance and impact of the study.

Author response:

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

Reviewer #1 (Public Review): 

Summary:

The authors demonstrate that the immunosuppressive environment in pancreatic ductal adenocarcinoma (PDAC) can be mitigated by a combination of ionizing radiation (IR), CCR5 inhibition, and PD1 blockade. This combination therapy increases tissue-resident natural killer (trNK) cells that facilitate CD8 T cell activity, resulting in a reduction of E-cadherin positive tumor cells. They identify a specific "hypofunctional" NK cell population in both mouse and human PDAC that supports CD8 T cell involvement. A trNK signature is found to be associated with better survival outcomes in PDAC and other solid tumors.   

Strengths: 

Overall, I think this is an interesting study that combines testing of therapeutic concepts in mice with bioinformatics analysis of single-cell transcriptome data in primary tumors and exploration of clinical outcomes using signature genes in TCGA data. The key finding is that immunoregulatory properties of tumor-infiltrating/resident CD56-bright NK cells (assumed to be non-cytotoxic) are beneficial for outcome through cross-talk with DC and recruitment of CD8 T cells. The latter is specifically induced by irradiation combined with CCR5i and PD1 blockade. 

"These results collectively support the notion that IR/CCR5i/αPD1 combination treatment alters immune infiltration by reducing Tregs and increasing NK and CD8 T cells, thereby resulting in greater local tumor control." I agree with this conclusion.  

Weaknesses:  

There are a few points to discuss and that the authors may want to address. 

(1)   "Notably, CCR5i significantly reduced Treg infiltration but had no effect on the infiltration of other immune cells, indicating the active recruitment of CCR5+ Tregs in PDAC (Figure 2B)." 

CCR5i treatment seems to inhibit infiltration of CD8 T cells and NK cells to a greater extent, in relative terms, compared to Treg, albeit it is not statistically significant. If this visual inspection of the graph does not reflect reality, additional experiments may be needed to verify the selective targeting of Tregs or confirm the fact that also CD8 T cells and NK cells are affected by single agent CCR5i. The reduced recruitment of Treg, NK cells, and CD8T cells was completely reversed when combined with irradiation. In the data shown in Figure 3E it seems as if CCR5i induced infiltration of Tregs along with other immune cells. However, this said, I agree with the conclusion of the authors that this combined treatment leads to an altered immune composition and ratio between Tregs and effector cells (CD8T cells and NK cells). Could this altered composition be displayed more clearly? 

We would like to thank the reviewer for their comments and agree that there is a trend for reduced NK and T-cell infiltration during CCR5i standalone treatment (as seen in Figure 2B), although it does not reach significance. To reflect this more clearly, we have added n.s (non-significant) for the NK cells and CD8+ T-cells and adjusted the text to reflect a trend for decreased NK and CD8+ T-cell infiltration (See Lines 162-165). Moreover, to reflect the data accurately, we have taken the Treg data out of the original Figure 2B and present it separately as a percentage of CD45+CD3+ T-cells.

(2) The definition of active and hypofunctional NK cells based on solely NKG2D expression alone seems like an oversimplification. I realize it is not trivial to test tumor-infiltrating NK cells from these tumors functionally but perhaps scRNAseq of the tumors would allow for characterization of cytotoxicity scores using KEGG or GO analysis or reversed gene set enrichment in responders/non-responders.  

We agree that scRNA-seq of tumors would add to the overall characterization of the tumor-infiltrating NK cells and their characterization, however we are currently unfortunately not in the position to carry out this experiment. We did however immunophenotype the tumor infiltrating NK cell population in more depth by also looking at NKp46 and NKG2D surface expression. This newly added data demonstrates not only increased infiltration of “bona-fide” trNK cells (based on surface expression of CD103+CD49a+) under the triple treatment combination, but more importantly these trNK have reduced levels of CD69, NKp46, NKG2D and increased TIM-3 surface expression compared to conventional NK cells – suggesting that these trNKs could be more hypoactive compared to the conventional NK cells. These data have been added to the manuscript as Figure 4E, F; Figure supplement 4E-G and Lines 244-260 in the revised manuscript. To clarify this difference, we have replaced the word “hypofunctional” with “hypoactive” throughout the manuscript.

(3) It seems as if the abstract refers to this phenotype incorrectly since the "hyporesponsive" subset is described as NKG2C-negative. 

We apologize for the typographic confusion and have corrected our abstract and changed the subset to NKG2D-negative (as was intended).

(4) "The NK_C1 cluster correlates best with the hypofunction NK phenotype observed in mice as similarly displayed reduced activation (reduced NKG7, NKp80, GZMA, and PRF1) with additional expression of tissue residency markers CD103, CD49a and, surprisingly, the adaptive activating receptor NKG2C (KLRC2) (Figure 5B, C)." 

There is no doubt that NK_C1 represents tumor-infiltrating NK cells with a CD56bright gene signature with a strong tissue resident score. However, the transcriptional expression of KLRC2 on these is not surprising! It is well established that KLRC2 transcripts (but not protein) are highly expressed on conventional CD56bright NK cells. There are several published sources where the authors can find such data for confirmation. Thus, this is not to be confused with adaptive NK cells having an entirely different transcriptional signature and expressing high levels of NKG2C at the cell surface. I strongly recommend reinterpreting the results based on the fact that KLRC2 is expressed at high levels in conventional CD56bright NK cells. If not, it would be important to verify that these tissueresident NK cells express NKG2C and not NKG2A at the cell surface. 

We agree with the reviewer and have modified the text accordingly in the revised manuscript (Lines 279-283), including references to tissue-resident adaptive-like cells as described previously in literature. 

(5) NCAM1 transcript alone is not sufficient to deconvolute CD56bright NK cells in TCGA data (Figure 7A). As a single marker, it likely reflects NK cell infiltration without providing further evidence on the contribution of the bright/dim components. Therefore, the use of the bright Tr NK signature described in Table 1 is very important (Figure 7B). Table 1 is not provided. Nor Supplementary Table 1. There is only one supplementary figure in the ppt attached.

We agree that a high NCAM1/CD56 single gene signature could also represent NK cell infiltration. We have rephrased this in the text accordingly (Lines 354-357). We apologize for the missing tables and Supplementary figures. We have added these now to the manuscript as Supplementary table 1.

Reviewer #2 (Public Review)  

Summary: 

This work elaborates on a combined therapeutic approach comprising ionizing radiation and CCR5i/αPD1 immunotherapy as a promising strategy in pancreatic cancer. Previous research has established that NK cell-derived CCL5 and XCL1 play a crucial role in recruiting cDC1 cells to the tumor microenvironment, contributing to tumor control. In this study, by using a murine pancreatic cancer model, the authors propose that the addition of radiation therapy to CCR5i and αPD1 immunotherapy could upregulate CD8+ T cells and a subgroup of NK cells within the tumor and result in better tumor control. They further analyzed human single-cell sequencing data from pancreatic cancer patients and identified one subgroup of NK cells (NK C1) with tissue-resident features. Subsequent cell-cell contact analysis reveals the NK-cDC1-CD8 cell axis in pancreatic cancer. By analyzing TCGA data, they found that high NK C1 signature levels were associated with better survival in pancreatic cancer patients. Thus, radiotherapy could benefit the outcome of patients bearing low NK C1 signatures. Importantly, the positive correlation between NK C1 score with survival extends beyond pancreatic cancer, showing potential applicability across various solid cancers.  

Strengths: 

This study could add new insight into the clinical practice by introducing such novel combined therapy and shed light on the underlying immune cell dynamics. These findings hold potential for more effective and targeted treatment in the future. Mouse experiments nicely confirmed that such combined therapy could significantly reduce tumor volume. The elegant use of single-cell sequencing analysis and human database examination enriches the narrative and strengthens the study's foundation. Additionally, the notion that NK C1 signature correlates with patient survival in various solid cancers is of high interest and relevance.  

Weaknesses: 

The role of CCR5i requires further clarification. While the authors demonstrated its capacity to reduce Treg in murine tumors, its impact on other cell populations, including NK cells and CD8+ T cells, was not observed. Nevertheless, the effect of CCR5i on tumor growth in Figure 2B should be shown. If the combination of radiotherapy and αPD1 already can achieve good outcomes as shown in Figure 3A, the necessity to include CCR5i is questioned. Overall, a more comprehensive elucidation of the roles of CCL5 and CCR5i in this context would be good.  

We would like to thank the reviewer for their comments and agree that standalone CCR5i also shows a trend of reduced infiltrating NK cells and CD8+ T-cells, although this does not reach significance. We have mentioned this trend in the manuscript (see Lines 162-165) and added n.s to Figure 2B as well. In regards to adding CCR5i; although we observe volumetric control by radiotherapy and anti-PD1, we observe an increase in necrosis induction only in the triple combination compared to radiotherapy combined with anti-PD1 – suggesting that there is an additive effect of CCR5i in our model only as a combination modality. We therefore believe that addition of CCR5i to radiotherapy and anti-PD1 has a beneficial effect. The growth curves for CCR5i alone were already presented in Figure 3A, and we have modified our manuscript to refer to this (see Lines 165-167).

(1) In line with this, spatial plots in Figure 4 did not include the group with only radiotherapy and αPD1. This inclusion would facilitate a clearer comparison and better highlight the essential role of CCR5i. 

We agree with the reviewer that inclusion of radiotherapy and αPD1 would facilitate a clear comparison of our data and our experiments did include single controls for radiotherapy and αPD1; however, unfortunately, the tissue slides were of bad quality and therefore not suitable for quantification. In line with this, we have added references to other studies that investigated the effect of immune checkpoint inhibitors in combination with radiotherapy (see Lines 169-172).

(2) NK C1 cells should be also analyzed in the mouse model. The authors suggest that NKNKG2Dve could be the cell population. Staining of inhibitory markers should be considered, for example, TIGIT and TIM3 as presented in Figure 5B. 

As per the reviewer suggestion, we have now included some additional data on the surface expression of inhibitory markers/activating receptor on tumor-infiltrating NK cells in our model under the triple combination. These additional data demonstrate increased infiltration of trNK under the triple combination that seem to be more ‘hypoactive’ than conventional NK cells.  This data has been added as Figure 4E in the revised Figure.

(3) While the cell-cell contact analysis generated from single-cell sequencing data is insightful, extending this analysis to the mouse model under therapy would be highly informative. NK and CD8 cells in the tumor increased upon the combined therapy. However, cDC1 was not characterized. Analysis regarding cDC1 would provide more information on the NK/cDC1/CD8 axis. 

We agree that looking into cDC1 would be highly interesting in our treatment model and its characterization is currently under investigation. The importance about the interaction between cDC1-NK cells has been described before by various groups, and we have provided additional references for that in our manuscript (see Lines 449-455)

(4) Human database analysis showed a positive correlation between NK C1 score and CCL5 in pancreatic cancer. Furthermore, radiotherapy could benefit the outcome of patients bearing low NK C1 scores. It would be interesting to test if radiotherapy could also benefit patients with low CCL5 levels in this cohort. 

We would like to thank the reviewer for their suggestion and please see the figure below for the comparison. Patients with CCL5high are enriched for NK_C1 (Figure 7D) and CCL5high patients with NK_C1high have significantly increased overall and disease-free survival compared to NK_C1low (Figure 7E); where those with NK_C1low significantly benefit from radiotherapy (Figure 7B). Accordingly, patients with CCL5high have significantly decreased overall survival compared to CCL5low patients, again confirming CCL5 as a prognostic marker (Figure 1A, Figure R1). When we look at CCL5low patients however, there is no additional significant benefit for radiotherapy (see insert below) in the CCL5low group (not significant; only significant p-values are shown). These data collectively support the strong correlation between CCL5 levels and NK_C1 enrichment, and imply that radiotherapy alone is insufficient to drive NK_C1 cells in the absence of high CCL5 gradients to improve overall survival. However, given the increased overall survival of CCL5low compared to CCL5high it is likely that other factors are at play. Future studies will be required to further elucidate the role of CCL5 gradients on NK_C1 cells and the beneficial effect of radiotherapy.

Author response image 1.

Overall survival of CCL5high versus CCL5low patients stratified into groups with and without radiotherapy using TCGA-PAAD. Log-rank p-value indicates the significance level across all groups while individual significant comparisons are shown as indicated.

Reviewer #3 (Public Review):

Summary

In the submitted manuscript by Go et al, the authors evaluated the tumor microenvironment in pancreatic ductal adenocarcinoma (PDAC) and made a number of interesting observations, including the following: 1) CCL5 expression within the tumor microenvironment negatively correlated with clinical outcomes in human patients with PDAC; 2) there were both positive and negative correlations between CCL5 expression and the expression of specific genes (e.g. those encoding CD56 and CD16, respectively) included among gene signature lists for Treg, MDSC, TAM, and NK cells; 3) CCR5 inhibition with the inhibitor, maraviroc, reduced Treg infiltration but not that of other immune cell types in an orthotopic murine model of PDAC; 4) CCR5 inhibition augmented anti-PD1 immunotherapy when combined with ionizing radiation (IR) therapy in the murine model; 5) the above therapy resulted in increased infiltration of CD8+ cytotoxic T cells as well as of a subset of NKG2D-negative, tissueresidency (tr) marker expressing NK cells (deemed Cluster 1 NK in their data sets) that inversely correlated with the number of E-cadherin+ cells (i.e. tumor cells) and showed predicted interactions with cDC1 dendritic cells (including XCL1/XCL2 expressed by the NK and XCR1 expressed by the cDC1); 6) the authors identified a number of putative signals stemming from the trNK (e.g. IL-16, TNFSF14, FASLG, CSF, MIF) as well as incoming from cDC1s to NK (e.g. BAG6-NKp30); 7) these trNK cells positively correlated with good outcomes and with CD8+ T cell infiltrations in human PDAC as well as in many other solid tumor types; and 8) importantly, the benefit of IR therapy was specific to the subset of PDAC patients (represented in the TCGA dataset) that were predicted to have low amounts of trNK cells. The authors used murine experimental models, multiplexed imaging analyses, and a number of publicly available sequencing data sets from human tumor samples to perform their investigations. Based on their findings, the authors proposed that combining IR with CCR5 inhibition and anti-PD1 immunotherapy is a promising strategy to treat solid cancers.  

Strengths

Overall, the collective analyses and conclusions appear to be novel and could be of high and rapid impact on the field, particularly in terms of directing clinical trials to incorporate IR with CCR5 inhibition and immunotherapy. The manuscript is well written; the figures are for the most part clear; and the Discussion is very thoughtful.   

Weaknesses

There were a number of minor typographical errors, missing references, or minor issues with the figures. In general, while many of the observations provided strong suggestive evidence of relationships, phenotypes, and functions, the authors often used language to indicate that such things were confirmed, validated, or proven. In fact, there was a paucity of such functional/confirmatory experiments. This does not necessarily detract from the overall significance, excitement for, and potential impact of the study; but the language could likely be adjusted to be more in keeping with the true nature of the findings. The main title and running title are a bit different; consider making them more similar.

We apologize for the typographical errors, missing references and issues with the figures. We have revised our manuscript, with a major focus on adjusting our language to more carefully reflect our data, and hope to have addressed all the concerns of the reviewer. The slight discrepancy between the main title and running title are to be able to convey the contents of this manuscript in a comprehensive way. 

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):  

Please make sure all files are made available. Also please check available datasets describing KLRC2 transcripts in CD56brights. This is not to be confused with an adaptive-like signature. 

We have added the missing table to the supplementary figures and revised the manuscript text in regards to KLRC2 transcript in our NK_C1 cluster and its implications for an adaptive-like signature in the context of tissue-residency (see Lines 279-283; 465-474).

Reviewer #2 (Recommendations For The Authors): 

Additional experiments as mentioned in the 'weakness' section could help to further strengthen this study. Besides these points, I would recommend the following: 

(1) The description in the figure should be more precise and clear. Especially in Figure 3A, it seems the addition of IR into CCR5i or CCR5i/aPD1 leads to a bigger tumor volume.  

We have adjusted the figure descriptions to more clearly describe the figures. We apologise for the confusion in Figure 3A, this was a figure legend error and has been correctly rectified in the revised Figures (i.e. closed symbols represent +IR conditions).

(2) The definition of Tregs in figures should be described, e.g. it is not specified which population is shown in Figure S2c.  

We have added a definition of Tregs (i.e. Live/CD45+CD3+CD4+FOXP3+) in our revised manuscript (see Lines 162-165). To avoid confusion, we have removed the subsequent gating of CCR5 and PD-1 of Tregs in our revised Supplementary Figures.

(3) Please add a bar in all histology figures, for example, Figure 2A, S2A, S3E. It seems in Figure S3D, E, the green group is missing.  

We have added the scale bar to all the indicated figures. Unfortunately, indeed as correctly pointed out by the reviewer, we are missing the green group (i.e. IR+CCR5i) as we felt that the excessive growth seen with CCR5i alone may have given a false impression of the extent of infiltration, therefore we did not include this in the original analysis and do not have the data in the Figure.

(4) Please check through the manuscript, there are some grammar mistakes.  

We apologise for the grammar mistakes in our original manuscript and have carefully revised the current manuscript to avoid grammar mistakes

(5) Figure S7B, the left cell lacks a name.  

We have annotated the left cell accordingly in our revised supplementary figure.

Reviewer #3 (Recommendations For The Authors): 

(1) Abbreviations (e.g. PDAC) should be spelled out the first time introduced in the manuscript.

We have adjusted this in our revised manuscript.

(2) Referring to the tissue-resident NK cells as "hypofunctional" may not be useful...they seem to be functional, just not in the conventional sense. The authors may want to consider another term, such as non-cytotoxic (given the low expression of cytolytic granules, etc) or immunoregulatory (as they actually refer to them on line 310).

We agree with the reviewer and have revised the manuscript to refer to them as “immunoregulatory” or “hypoactive” when appropriate. The latter is supported by the additional experiments as shown in Figure 4E.

(3) Barry et al 2018 Nat Med demonstrated that NK cells in melanoma could support cDC1s and promote positive clinical outcomes in the setting of immunotherapy. It would likely be beneficial to also cite this paper (e.g. on line 425). 

Thank you for the suggestion, which would work in line with our hypothesis of crosstalk between NK_C1 and cDC1. We have looked for FLT3L in our NK_C1 cluster and did not find any enrichment for FLT3L transcript (see Figure 5E). Nevertheless, we have added the reference in the discussion of our manuscript to further support the importance of crosstalk between cDC1 and NK cells (see Lines 449455)

(4) Figure 2B: by eye, it looks like the difference between CD8+ T cells in the two conditions would be significantly different; is this not the case? Same thing for the NK cells...what are the pvalues? 

We have added n.s. to our revised Figure 2B. The p-values for CD8+ T-cells and NK cells were 0.14 and 0.19 {2-tailed students t-test), respectively.

(5) The murine data strongly suggest that the combination therapy promotes trNK cell infiltration into the tumors, in turn resulting in cDC1-mediated CD8+ T cell infiltration and/or activation. It could be highly valuable/useful to functionally determine (e.g. by depleting NK cells in this model) if NK cells are required for the effects seen. 

We agree that depletion of NK cells could really solidify the findings even more, and it is part of ongoing investigations for future projects. However, it would be imperative to first characterise these NK cells in more depth as conventional global ablation of NK cells is excepted to highly impact immunosurveillance as well. This is part of current ongoing work.

(6) Figure 7B: how were "high" and "low" defined (for the NK signature)?

An enrichment score of the NK_C1 gene signature (see Table supplement 1) was first calculated per patient sample in the TCGA RNA-seq dataset using the Gene Set Variation Analysis (GSVA) method. A cut-off value was then determined using the maximally selected rank statistics (max-stat R package) method to divide patients into “high” and “low”. 

(7) Lines 164-165 of the Results: it would be good to include a reference supporting the statement.

We have added rephrased the manuscript and added corresponding references (see Lines 170-173 in revised manuscript).

(8) There are many conclusions and very speculative language based only on sequencing results, and these have not been validated (e.g. in the Discussion, lines 447-453). As another example, it was concluded that a decrease in NKG2D+ NK cells implied a reduction in overall NK cell cytolytic activity and that NKG2D- NK cells were hypofunctional and did not kill well. This was not tested. Generally, it would be useful for the authors to use language that conveys that the data are primarily suggestive (rather than "confirmatory", line 447) of relationships, phenotypes, and functions at this point. 

We thank the reviewer for their concerns and have carefully adapted the manuscript text to more clearly clarify the findings in a careful manner.

(9) On lines 246-247 the authors refer to cluster 3 NK cells, which express CD16, as "immature". The rationale for this designation is not provided, and most human NK cell development models hold that CD16+ NK cells represent the most mature subset(s). 

We apologize for the typographic error – later on we refer to the NK_C3 cluster as cytotoxic NK cells and we have corrected this in our revised manuscript (see Lines 273-275).

(10) On line 351, the authors reference supplemental Figure 7C...but I don't see this figure in the accompanying powerpoint file. 

This should have been Supplementary Figure 7B, and we have corrected it in the revised manuscript (see Lines 374-377)

(11) On line 417, the authors reference NKp40; this is likely a typographical error. 

This has been corrected in the revised manuscript to NKp46 (see Lines 439-442).

eLife assessment

The authors investigated the requirement and function of Blimp1/Prdm1 in murine natural killer (NK) cells and the ILC1 lineage of innate lymphoid cells, using a conditional knockout model. The single-cell mRNA-seq data provided here represent a valuable resource for the community, but the lack of mechanistic investigations leaves the study partially incomplete. The work will be of interest to the fields of innate lymphoid cell biology and tissue immunology.

Reviewer #1 (Public Review):

He et al. investigate the requirement and function of Blimp1 (encoded by Prdm1) in murine NK cells and ILC1. Employing a conditional knockout mouse model (Prdm1flox x Ncr1cre), the authors describe impaired abundance and maturation of Prdm1-deficient NK cells and ILC1 in different tissues. Blimp1-deficient NK cells have reduced expression of cytotoxic molecules (Gzmb, Prf1) and, in some instances, Ifng production, and Prdm1flox x Ncr1cre mice show impaired tumor control in experimental metastasis models. Using single cell RNA sequencing analysis, the authors propose that Prdm1 regulates JunB expression and NK cell maturation. Based on in silico analyses, the authors suggest manifold intercellular communication between NK/ILC1 and macrophages. Without following up on any of these potentially interesting suggestions, the authors conclude their study reiterating that Prdm1 regulates IFNg-production of tumor-infiltrating NK cells and ILC1.

Many of the reported functions of Blimp1 in NK cells have previously been identified using a mixed-chimera strategy comparing Prdm1 WT and KO NK cells (Kallies et al., Blood 2011). Here, the authors expand on these findings using a conditional model to delete Prdm1 in NK/ILC1 and single cell sequencing, and provide a more refined analysis of the functions of Blimp1 in these cells. Cell-chat analysis suggests close interactions fo Blimp-dependent NK/ILC1 subsets with hepatic macrophages, but these suggestions are not followed up by experiments. Potentially interesting differences in the macrophage compartment of Ncr1-Cre x Prdm1-fl/fl mice are suggested by the scc-RNA-Seq data, but are not validated e.g. by FACS. The study falls short in providing new mechanistic insights. Nevertheless, it is an interesting confirmation of "old" suggestions in a more refined setting, and the provided single-cell mRNA-Seq data represents a potentially valuable resource for the community.

Reviewer #2 (Public Review):

He and colleagues aimed to elucidate the role of the transcription factor Prdm1 in liver Type 1 ILCs (innate lymphoid cells), focusing on its regulatory mechanisms and potential implications for developing innovative immune therapy strategies against liver cancer​.

Strengths:

The study effectively integrates omics analyses and cytometry to explore Prdm1's impact on the cellular composition and immune regulation within the liver, providing a comprehensive view of its biological role​. Employing a conditional knockout mouse model adds specificity to their experiments, allowing for precise manipulation of the Prdm1 gene​​.

Weaknesses:

The study predominantly relies on limited mouse models, which may not fully represent the complexity of Type 1 ILC behavior across different cancer types. Some experimental designs, such as the limited in vitro killing assessments, and additional human data could be expanded to strengthen the findings and their interpretation​​.

The authors have demonstrated that Prdm1 plays a critical role in the function of NK cells and ILC1s within the liver, particularly in the context of tumor resistance. However, due to the use of specific disease models and lack of direct human data, the application of these findings to clinical settings remains speculative​​. While the study advances our understanding of liver ILC biology, further research is necessary to validate these effects in human systems and across more diverse cancer models​.

​Discussion on impact and utility:

This study contributes significantly to the field of immunology and cancer therapy by revealing potential new targets for immunotherapy of liver cancer. The methods and data provided could serve as a valuable resource for further research aimed at enhancing immune-based cancer treatments​.

​Additional Context for Interpretation:

Understanding the role of Prdm1 in the broader context of immune cell regulation and its interaction with other cellular components in the tumor microenvironment could be crucial. Further studies should explore the dynamic between Prdm1 expression, NK cell functionality, and tumor resistance mechanisms to fully harness the therapeutic potential of targeting this pathway in liver cancer​.

Author response:

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

Reviewer 1 (Public Review):

He et al. investigate the requirement and function of Blimp1 (encoded by Prdm1) in murine NK cells and ILC1. Employing a conditional knockout mouse model (Prdm1flox x Ncr1cre), the authors describe impaired abundance and maturation of Prdm1-deficient NK cells and ILC1 in different tissues. Blimp1-deficient NK cells have reduced expression of cytotoxic molecules (Gzmb, Prf1) and, in some instances, Ifng production, and Prdm1flox x Ncr1cre mice show impaired tumor control in experimental metastasis models. Using single-cell RNA sequencing analysis, the authors propose that Prdm1 regulates JunB expression and NK cell maturation. Based on in silico analyses, the authors suggest manifold intercellular communication between NK/ILC1 and macrophages. Without following up on any of these potentially interesting suggestions, the authors conclude their study reiterating that Prdm1 regulates IFNg-production of tumor-infiltrating NK cells and ILC1. Many of the reported functions of Blimp1 in NK cells have previously been identified using a mixed-chimera strategy comparing Prdm1 WT and KO NK cells (Kallies et al., Blood 2011). Here, the authors expand on these findings using a conditional model to delete Prdm1 in NK/ILC1 and single-cell sequencing and provide a more refined analysis of the functions of Blimp1 in these cells. Cell-chat analysis suggests close interactions of Blimp-dependent NK/ILC1 subsets with hepatic macrophages, but these suggestions are not followed up by experiments. Potentially interesting differences in the macrophage compartment of Ncr1-Cre x Prdm1-fl/fl mice are suggested by the scRNA-Seq data but are not validated e.g. by FACS. The study falls short in providing new mechanistic insights. Nevertheless, it is an interesting confirmation of "old" suggestions in a more refined setting, and the provided single-cell mRNA-Seq data represents a potentially valuable resource for the community. There are some control analyses that are required to support the conclusions of the authors, and I have a few suggestions that would help to improve the manuscript.

We sincerely appreciate your careful review and insightful feedback on our manuscript. We have carefully considered your comments and present the results of new experiments conducted in response to your suggestions. Please find the detailed responses below.

Major comments

Comment 1: The authors do not control for the potential effects of Cre expression. Expression of Cre from within the Ncr1 locus (using the mouse model established by Narni-Mancinelli et al.) has significant effects on NK cells and especially ILC1s (reducing their frequency and absolute numbers and altering their functionality. The authors should characterize the Ncr1cre mice used here (developed by Shanghai Model Organism Center) in this regard and should use proper controls (Ncr1Cre+ Prdm1wt/wt as control for Ncr1Cre+ Prdm1fl/fl, instead of WT littermates) for all of their key data, e.g. those depicted in Fig 1FG, 2ADFH, 7D, S2,3,4.

Response 1: This is a very insightful question that has posed a challenge for many researchers, including us, engaged in conditional knockout studies. The expression of Cre and the insertion of loxP sequences both have the potential to influence gene expression. This is because the region where loxP is inserted may contain regulatory sequences for the gene of interest. Ncr1-Cre is a frequently used transgenic mouse model in our laboratory. In our prior research, we also had concerns about the possible impact of Cre on NKp46 expression, which could lead to a decline in NK cell function. Therefore, in our previous studies focused on Smad4 expression in NK cells, we conducted similar experiments. In Figure 6 of our published paper in the Journal of Clinical Investigation (Wang et al., J Clin Invest, 2018), we compared NKp46-iCreTgfbr2fl/flSmad4fl/WT with NKp46-iCreTgfbr2fl/flSmad4fl/fl. Although the primary purpose is to establish Smad4's independence from TGF-β, it also allows for a comparison between Smad4fl/fl and Smad4fl/WT in the presence of Cre. In the critical phenotype we assessed, NKp46-iCreTgfbr2fl/flSmad4fl/fl (compared with NKp46-iCreTgfbr2fl/flSmad4fl/WT) exhibited the same phenotype as NKp46-iCreSmad4fl/fl (compared with NKp46WTSmad4fl/fl). This suggests that Cre's influence on NK cells may be within a reasonable and controllable range. Furthermore, in contrast to the decrease in Ncr1 expression caused by Cre, the reduction in the expression levels of genes targeted by Loxp knockout, such as Prdm1 in this study (Figure 1 E), is more significant. Therefore, with the current techniques and research methods, we believe that the data provided in this study can support the role of Prdm1 in

NK cells.

Comment 2: Several of the phenotypic findings on NK cells have been described before by Kallies et al. in 2011 (Ref 29), although using a different genetic Prdm1-ablation model (Prdm1-GFP/GFP knockin/knockout model). This study reported impaired NK cell maturation, reduced Gzmb expression, impaired in vivo cytotoxicity against subcutaneous RMA-S cells, impaired in vitro proliferation, comparable in vitro killing, increase in BM NK cell numbers. The authors should discuss/mention this more prominently in their manuscript, and highlight where they confirm or refine these previous findings, and where they actually provide new information.

Response 2: We appreciate your valuable suggestions. The article you referred to, published in Blood, is indeed an excellent work. While we had cited this article, our discussion regarding its specific content was limited. Based on your advice, we have made revisions and included the following content in our discussion section (page 24; line 489-493):

“In a study involving systemic knockout combined with competitive transplantation, it was found that Prdm1 promotes NK cell maturation and the expression of Gzmb. On the contrary, the same study also found that NK cells with Prdm1 deficiency exhibit heightened proliferation, increased survival, enhanced migratory abilities towards tumors, and greater cytotoxicity against subcutaneously implanted RMAS tumors (31).”.

Comment 3: What is the reason to refer to the enriched cluster in Blimp1-deficient NK cells as "Junbhi"? There is no follow-up for a function of Junb, and there are many other genes upregulated in these cells. Most critically, these cells seem to represent exactly the c-Kithi cells that Kallies et al. already showed and discussed in their paper. The authors should stain for Kit, and also refer to this. Also, MacKay et al. performed Blimp1-Chip-Seq (in T cells), maybe it would be interesting to check to which of the identified DEGs Blimp1 can bind.

Response 3: We appreciate the suggestion from the reviewer. We think a gene that supports the development of lymphocytes doesn't necessarily positively regulate their function. For example, JunB is essential for T cell development but can also induce T cell exhaustion (Lynn et al., Nature. 2019). Therefore, while Prdm1 has been shown to promote NK cell development, it cannot be assumed that it always positively regulates NK cell function, especially for anti-cancer immune surveillance. In this respect, we try to find a driving-factor of the impaired anti-tumor ability of Prdm1_Δ_Ncr1 NK cells. Although there are many other genes upregulated in this cluster (e.g. Kit), JunB attracts more our interest of its potential for regulating NK cells functions in cancer, whereas c-Kit is more likely a marker of NK cells maturation, which has been well-demonstrated by Kallies et al. and other studies. Our previous studies also showed that the expression of c-kit was decreased in mature NK cells, compared immature NK cells (Wang et al., J Clin Invest, 2018). 

The lack of following experiments of Junb is because we cannot find valuable surface markers to investigate the follow-up function of _Junb_hi cNK cluster. If we use intracellular markers, it is more likely an analysis of gene expression pattern, which has been well-described in our RNA-seq data. As we describe above, our study did not aim to further investigate the role of prdm1 in NK cells maturation, as the c-Kit expression was upregulated in Prdm1-kncok NK cells and correlated with NK cell maturation, which has been validated by Kallies et al.. 

We also have discussed the potential DEGs that could be bound and regulated by Prdm1 in our revised manuscript (page 27-28; line 561-571):

“Prdm1 and Hobit directly bound and repressed Tcf7 (18), which encoded TCF-1, a TF binding and limiting the activity of Gzmb regulatory element (69). Gzmb has been demonstrated directly bound and activated by Junb in NK cells, which suggested Gzmb expression regulated by multiple Prdm1/Hobit downstream signals (26). In human T cells, binding motif of JUNB was enriched in the binding sites of PRDM1 (70), indicating the essential role of PRDM1-JUNB axis during NK cell and T cell development. In NK cells deficient in Prdm1 expression, we noted a decrease in Gzmb levels alongside with an elevation in Junb expression. This indicates that Prdm1 not only facilitates the expression of Gzmb in NK cells but also suppresses Junb expression. Given that Junb is recognized as a positive regulator of Gzmb (71), this presents a complex interplay that seems contradictory. Therefore, it is imperative to develop a theoretical framework to comprehensively understand and interpret this paradoxical relationship.”.

Comment 4: cNK cells are considered circulating cells, that transiently pass through the liver.

Previous studies have suggested almost identical gene expression patterns in hepatic and splenic NK cells. In functional tests, they often "perform" identically. I am therefore a bit surprised that the authors find a differential dependency of Blimp1 for the IFNg production of splenic (no role of Blimp1) versus hepatic (Blimp1 regulating IFNg production) NK cells (Fig S3). Do the authors have any suggestions on that? The analyses are performed by 12+4h stimulations with IL12/18, which could involve the effects of altered bystander cells (as suggested by Figure 6). Therefore, these analyses should be provided upon standard 4h stimulations with IL12/18 and also with PMA/I under BFA. Note: liver and splenic cNK cells look quite different in the chosen histograms in Figures 7 A, B, C, yet there is massive variability in these analyses - is there any systematic/technical problem?

Response 4: We appreciate the valuable suggestion from the reviewer. Studies have suggested that, at the gene expression or transcriptomic level, liver NK cells exhibit more similarity to splenic NK cells while displaying greater divergence from liver ILC1s. However, we do not think that splenic NK cells or peripheral blood NK cells (which are more abundant in circulation) are entirely indistinguishable from liver NK cells. Notably, there are substantial differences in their maturity levels, with liver NK cells being more mature. Since we are examining the protein levels, a 4-hour stimulation period may not fully capture these distinctions. Even when considering the potential impact of bystander cells, the experimental design specifically targets Prdm1 knockout within NK cells, ensuring that the study accurately elucidates the role of Prdm1 in NK cells. For each experiment, we have implemented control measures, and any variances observed in the figures may be attributed to individual variations among the animals. It is also possible that the MFI values measured by flow cytometry exhibit larger variations than a percentage.

Comment 5: Figure 4 H/I - In contrast to NK cells in Fig 4E, F, the KO and WT ILC1s seem to co-cluster largely. Authors should validate differentially expressed genes. How strong is the effect of Blimp1 in ILC1s? Or is Blimp1 a critical TF driving effector differentiation in NK cells, while it has only subtle effects in ILC1 (these may be regulated by Hobit?)? This seems an interesting finding that should at least be discussed. For these types of small differences in ILC1, FACS confirmation analyses should be performed and findings be reevaluated using Cre-expressing controls (see above).

Response 5: We appreciate the suggestion from the reviewer. As request, we analyze the DEGs in liver cNK cells and ILC1s from our scRNA-seq data (revised Supplemental Figure 8, A and B). There only a few valuable DEGs in ILC1s compared to cNK cells. It’s likely that Prdm1 have more essential effect of cNK cells transcriptional program, while it plays more important role in keep the homeostasis of ILC1s population. We have discussed these points to better inform the readers. (page 27; line 554-561): 

“Previous studies have identified Hobit and Prdm1 as central regulators instructing tissue-dependent programs and retention of diverse tissue-resident lymphocytes (18, 51, 53). Liver ILC1s required Hobit, but not necessary for cNK cells (6). Expression of Prdm1 was remarkably higher in cNK cells versus ILC1s (18). While in our study, cNK cells and liver ILC1s reduced simultaneously in Prdm1ΔNcr1 mice, and even more significant in ILC1s. This indicates that while Prdm1 is expressed at lower levels in ILC1s, its role in preserving the quantity of ILC1s may be more crucial. Thus, Prdm1 and Hobit may have parallel program in instructing ILC1s functional development and maturation.”. 

We cannot find valuable surface marker to evaluate the change in ILC1s, as most of changes are intracellular markers.

Comment 6: The authors describe and discuss some of Figure 1 and 2 data as if Blimp1 would be involved in alternative NK versus ILC1 fates, but there is no evidence for this.

Response 6: There is no evidence that Prdm1 could alter the fate decision of the progenitor towards liver cNK or ILC1s. Although some studies reported the conversion between cNK cells and ILC1s in special contexts, it was widely accepted that liver cNK cells and ILC1s originated from different progenitors. While we observed changes in the proportions of liver cNK cells and ILC1 in Prdm1 KO mice, we still lack sufficient evidence to support the relative independence of NK and ILC1 development, as well as evidence to indicate that Prdm1 is exclusively responsible for NK and ILC1.

Regarding the changes in NK and ILC1 proportions after Prdm1 KO, we believe that both NK and ILC1 cells require Prdm1 to maintain their populations, with ILC1 possibly requiring it to a greater extent. This is the reason for the altered balance between NK and ILC1 cells following Prdm1 KO. We wish to clarify this point to prevent any misconceptions among readers. To address this, we have added the following content to the discussion section (page 25; line 509-516):

“Furthermore, although both liver NK cells and liver ILC1s require Prdm1 to maintain their quantity, liver ILC1s demonstrate a more pronounced dependency on Prdm1. However, it is currently widely believed that liver NK cells and liver ILC1s originate from different progenitors. It is worth noting that while we observed changes in the NK and ILC1 proportions after Prdm1 knockout, our data does not support the hypothesis that Prdm1 affects progenitor differentiation decisions, thereby influencing the fate selection of NK and ILC1. Further research may be needed to elucidate how Prdm1 regulates the balance between NK cells and ILC1s.”.

Comment 7: There are several recent studies suggesting a role for Hobit, homologue of Blimp1, in NK cells and in ILC1, and in the control of liver metastases. The authors should discuss similar and unique functions of Hobit and Blimp1, also in the regulation of gene expression patterns, and should refer to these studies.

Response 7: We would like to express our gratitude to the reviewer for your insightful comments, which bring forth a critical perspective. In accordance with the reviewer's suggestion, we have updated our discussion to include the diverse functions guided by Hobit and Prdm1 in regulating the development and function of cNK cells and ILC1s (page 27; line 554-561):

“Previous studies have identified Hobit and Prdm1 as central regulators instructing tissue-dependent programs and retention of diverse tissue-resident lymphocytes (18, 51, 53). Liver ILC1s required Hobit, but not necessary for cNK cells (6). Expression of Prdm1 was remarkably higher in cNK cells versus ILC1s (18). While in our study, cNK cells and liver ILC1s reduced simultaneously in Prdm1ΔNcr1 mice, and even more significant in ILC1s. This indicates that while Prdm1 is expressed at lower levels in ILC1s, its role in preserving the quantity of ILC1s may be more crucial. Thus, Prdm1 and Hobit may have parallel program in instructing ILC1s functional development and maturation.”.

As shown in Supplemental Figure 8, we analyzed two published scRNA-seq data performed with Hobit_KO mice and integrated DEGs in cNK cells and ILC1s with our data. We observed overlaps of DEGs in _Prdm1_Δ_Ncr1 and Hobit_KO between cNK cells and ILC1s, such as _Junb, Tcf7, Gzmb, and Prf1 (Supplemental Figure 8), indicating the similar regulatory network of Prdm1 and Hobit. These data are now described on page 19; lines 386-395:   

“We also compared the gene expression patterns between Prdm1 and Hobit (homologue of Blimp1) with two published scRNA-seq data (51, 53). Following the knockout of Hobit, the DEGs were primarily identified within ILC1s. Conversely, after the knockout of Prdm1, a greater number of DEGs were observed in cNK cells. This indicates that Prdm1 likely possesses a broader range of target genes within cNK cells, whereas Hobit appears to have a more pronounced impact on gene expression within ILC1s (Supplemental Figure 8, C-F). There are some overlaps between the downstream transcriptional profile of Prdm1 and Hobit in liver cNK cells and ILC1s (Supplemental Figure 8, G and H), such as Junb, Fosb, Tcf7, Kit, Gzmb, Prf1, and Cxcr6 was simultaneously upregulated or downregulated in both Prdm1ΔNcr1 and _Hobit_KO liver cNK cells or ILC1s, indicating the similar regulatory networks of Prdm1 and Hobit.”.

Comment 8: Figure 4: The authors should discuss (and cross-validate) their liver gene expression analyses in the context of published datasets of NK and ILC1, such as the ones by Lopez et al, Friedrich et al, Ducimetiere et al and Yomogida et al.

Response 8: We thank the reviewer for raising this important point. To address this question, we have now analyzed the gene expression of liver cNK cells and ILC1 in two published data mentioned above, also in the context of Hobit-knock. We compared gene expression of different clusters and described in our revised manuscript (page 19; lines 386-395). 

“We also compared the gene expression patterns between Prdm1 and Hobit (homologue of Blimp1) with two published scRNA-seq data (51, 53). Following the knockout of Hobit, the DEGs were primarily identified within ILC1s. Conversely, after the knockout of Prdm1, a greater number of DEGs were observed in cNK cells. This indicates that Prdm1 likely possesses a broader range of target genes within cNK cells, whereas Hobit appears to have a more pronounced impact on gene expression within ILC1s (Supplemental Figure 8, C-F). There are some overlaps between the downstream transcriptional profile of Prdm1 and Hobit in liver cNK cells and ILC1s (Supplemental Figure 8, G and H), such as Junb, Fosb, Tcf7, Kit, Gzmb, Prf1, and Cxcr6 was simultaneously upregulated or downregulated in both Prdm1ΔNcr1 and _Hobit_KO liver cNK cells or ILC1s, indicating the similar regulatory networks of Prdm1 and Hobit.”.

Recommendations For The Authors:

Comment 9: The use of a paired t-test analysis when comparing cells/groups from different mice is not correct. Instead, the authors should consider using e.g. an unpaired t-test and re-test the indicated significance (e.g. Figure 1F, Figure 2H).

Response 9: We thank the reviewer’s comments. As we used littermates for the experiments and they are compared side by side, so the paired t-test analysis is acceptable. We reanalysis the significance in the results of Figure 1F, and Figure 2H using unpaired t-test. The statistics significance of Figure 1F using unpaired t-test was same as using t-test. However, in Figure 2H, the reduced IFN-γ production not reach statistics significance when used un-paired t-test (Supplemental Figure 12B). It may attribute to the variation between different littermates, but the trend is still under the scope of our conclusion. We believe that employing a paired t-test between littermates could be also meaningful. As such, we kept both statistical methodologies to ensure a thorough evaluation.

Comment 10: In several instances, it is unclear whether data are pooled or representative (and if so, of how many analyses). This information needs to be provided for all analyses. 

Response 10: We apologize for the lack of details and have now provided the sufficient information in our figure legends. 

For example, we delete the number in original histogram to avoid the misunderstanding of the unclear whether data are pooled or representative (e.g. original Figure7 A-C; revised Figure7 A-C). Furthermore, we added the “representative” in figure legends of all flow cytometric plots to better inform readers (e.g. original Figure2, D and F; revised Figure2, B and D).

Comment 11: In the title and abstract authors use "type 1 ILCs" for both NK cells and ILC1, and it is difficult to understand which phenotypes correspond to cNK cells versus ILC1. Most of the analyses clearly separate these two different cell types. I would appreciate a lot being more accurate in the abstract, and describing cNK and ILC1 phenotypes in a clear way.

Response 11: We are really sorry for our inaccurate descriptions. According to Spits et al., (Spits et al., Nature Reviews Immunology, 2013) and other related studies, we have now adopted a more appropriate nomenclature as “Conventional NK cells” correspond to “cNK cells”, “Type 1 innate lymphoid cells” to “ILC1s”, and “Group 1 ILC” as the collective name of cNK and ILC1s. 

The definition of these cells was described in the introduction (page 4, line 52-53; line58-62): 

“Group 1 ILCs consist of cNK cells and ILC1s (1, 2), with distinct developmental trajectories and effect molecules (3).”, “In a state of homeostasis, liver group 1 ILCs (CD45+CD3-NK1.1+NKp46+) can be discriminated into cNK cells and ILC1s by the differential expression of CD49a and CD49b (2): cNK cells are marked by the expression of CD49b, while liver ILC1s exhibit a distinctive positivity for CD49a. Tumor Necrosis Factor Related Apoptosis Inducing Ligand (TRAIL) is also expressed on liver ILC1s, but not on cNK cells (10, 11).”. 

We also describe cNK and ILC1 phenotypes in our scRNA-seq data, as shown in page 13; line 259-261: 

“cNK cells expressed high levels of Itga2 (CD49b) and Eomes, while ILC1s had high levels expression of Itga1 (CD49a) and Tnfsf10 (Supplemental Figure 5, F and G).”.

Comment 12: In the abstract authors state "The present study unveiled a novel regulatory mechanism of Prdm1 in liver Type 1 ILCs, showing promising potential for developing innovative immune therapy strategies against liver cancer." - maybe authors should discuss how their findings could be used for therapeutic approaches?

Response 12: We appreciate comments from the reviewer. As there hasn't been a clear consensus on the role of Prdm1 in NK cells prior to this, some studies have suggested that Prdm1 can inhibit cytokine secretion by NK cells. Particularly, Kallies et al. in their 2011 article in Blood found that Prdm1 might suppress NK cell anti-tumor activity. Hence, there hasn't been any immunotherapy targeting Prdm1 in NK cells for cancer treatment. Our research demonstrates the enhancing role of Prdm1 in NK cell anti-tumor activity, providing theoretical support for NK cell therapy targeting Prdm1. 

We added the following content to the discussion section (page 29; line 605-609): 

“Further research may provide deeper insight into the role of PRDM1 in the anti-tumor function of human NK cells, enabling a more direct investigation of its application in cancer therapies. Given its important role in preserving liver cNK cells and ILC1s functional heterogeneity, enhancing Prdm1 function in human NK cells could potentially be a strategy to promote NK cell-based immunotherapy for cancer.”.

Comment 13: The authors should explain or interpret their data a bit more (e.g. what is the consequence of GSEA enriched in negative regulation of Il6 production? (Fig. 3D)  do NK cells produce Il6 (Figure 3)? What's the impact of Il17 signaling in NK/ILC1 (Figure 5). Do the authors suggest JunB-driven metabolic reprogramming (Suppl. Fig 6D-F?).

Response 13: We appreciate comments from the reviewer. The question of IL-6 production in NK cell also raised by another reviewer. We have checked the GSEA results, and found no valuable genes in IL-6 production in NK cells. According to the suggestions of another reviewer (Response to Reviewer 2 Comment, Comment 14), it may be prudent to omit this figure.

IL-17 signaling indicated the plasticity of ILC1s, that may be originated from the differentiation of ILC3, we added more discussion of this part (page 17; line 341-344). 

“Several ILC3 signature genes, such as Rora, Tmem176a, and Tmem176b (45), highly expressed in this cluster (Supplemental Figure 7D). Considering the close relationship between IL-17 mediated immunity response and ILC3 (1, 46), it is plausible that _Il7r_hi ILC1 cluster may be attributed, at least in part, to potential plasticity between ILC1 and ILC3 subsets.”.

The decreased mitochondrial function may have more relevance to NK cell exhaustion in tumors. Our data suggest that the elevated expression of JunB in NK cells may predispose them to exhaustion. Currently, our hypothesis regarding the promotion of NK cell exhaustion by high JunB expression is based on the observed correlation between JunB expression levels and exhaustion phenotypes (at the gene expression and IFN-γ secretion levels) and the findings in reference 67 (Lynn et al., Nature, 2019), where JunB was found to promote T cell exhaustion. However, we have not demonstrated causation between high JunB expression and exhaustion in NK cells. We propose that in NK cells, especially mature NK cells, excessive JunB expression may make them more sensitive to exhaustion inducers. Nevertheless, further research is needed to confirm this. To clarify this, we added the following content in the discussion section (page 26; line 537-543): 

“While our current data is not sufficient to definitively classify these cells as exhausted NK cells, it supports that a subpopulation, referred to Junbhi cluster, demonstrates an exhaustion-like phenotype.

The significant increase in this cell population following Prdm1 knockout in NK cells may potentially be one of the reasons why Prdm1ΔNcr1 mice lose their tumor-killing capacity. Whether the excessive expression of JunB in NK cells is also a contributing factor to their exhaustion, similar to T cells(65), requires further investigation.”.

Comment 14: Ref 25 and Ref 57 are the same publication?

Response 14: We are really sorry for our careless mistakes. We have checked all the reference and corrected the wrong format.

Comment 15: Figure 1, E - The method description of RT-PCR is missing. I apologize if I have overlooked this information.

Response 15: We have now added the description of RT-PCR in our revised method section (page 31; line 638-644):

“RNA was extracted from FACS-sorted NK cells or splenocytes using RNASimple Total RNA Kit (TIANGEN Biotech, 4992858) and subsequently reverse transcribed to cDNA with SuperScript VILO Master Mix (Thermo Fisher Scientific, 11755050) according to manufacturer’s instructions. qPCR was performed with SYBR Green Mix (Thermo Fisher Scientific, A25742) and CFX Opus 96 Real-Time PCR System (Bio-Rad). The relative mRNA expression level was calculated using 2-ddCt method. Primer sequences:           Prdm1: 5’-CAGAAACACTACTTGGTACA-3’; 5’-GATTGCTTGTGCTGCTAA-3’.”

Comment 16: Figure 1, F - The NKp46+CD3- gate for the liver seems to cut the population, not all cells are included.

Response 16: We appreciate the review’s comment and apologize for our carelessness. We expend our data with more samples and reanalyzed them with a more convincing gating strategy. We now update our figures (revised Figure 1G; revised Supplemental Figure 2A). Several changes have occurred in the data and conclusions, and we have accordingly revised these contents in our manuscript.

The original text is:

“Proportion and absolute number of cNK cells in blood, bone marrow, lung, liver, spleen, and lymph nodes were analyzed by flow cytometry. Compared with Prdm1+/+ mice, the percentage of cNK cells (CD3-NK1.1+NKp46+) among lymphocytes was decreased in all of these tissues except bone marrow and lymph nodes (Figure 1F; Supplemental Figure 2A). However, no significant difference was observed in the percentage of cNK cells among bone marrow-derived lymphocytes between Prdm1ΔNcr1 and Prdm1+/+ mice. The absolute number of cNK cells in blood, lung, liver, and spleen also decreased in Prdm1ΔNcr1 mice (Figure 1F; Supplemental Figure 2A). Only a slight decrease in the number of cNK cells was observed in the lymph nodes of Prdm1ΔNcr1 mice, which did not reach statistical significance either (Supplemental Figure 2A). In contrast, the absolute number of cNK cells in Prdm1fl/fl mice bone marrow is moderately higher than Prdm1ΔNcr1 mice (Figure 1F).”

The revised text is (page 8; line 142-146):

“Proportion and absolute number of cNK cells in blood, bone marrow, lung, liver, spleen, and lymph nodes were analyzed by flow cytometry. Compared with Prdm1+/+ mice, the percentage and absolute number of NK cells (CD45+CD3-NK1.1+NKp46+) among lymphocytes was decreased in all of these tissues, whereas increased number of NK cells were observed in bone marrow (Figure 1G; Supplemental Figure 2A).”

Comment 17: Figure 1, The y-axis labeling of lung CD3-NKp46+ cells (x10^3) is not correct.

Response 17: We are really sorry for our carelessness. We now check the labels and make sure they are correct.

Comment 18: Figure 1, The statistical significance of absolute numbers of NKp46+ cells in the bone marrow should be reviewed.

Response 18: We expend our data with more samples and reanalyzed them with a more convincing gating strategy. We observed significant increase of bone marrow NK cells quantity in our updated data. These changes are now described in our revised manuscript.

The original text is: 

“However, no significant difference was observed in the percentage of cNK cells among bone marrow-derived lymphocytes between Prdm1ΔNcr1 and Prdm1+/+ mice”, “In contrast, the absolute number of cNK cells in Prdm1fl/fl mice bone marrow is moderately higher than Prdm1ΔNcr1 mice (Figure 1F).”

The revised text is (page 8; line 142-146):

“Proportion and absolute number of cNK cells in blood, bone marrow, lung, liver, spleen, and lymph nodes were analyzed by flow cytometry. Compared with Prdm1+/+ mice, the percentage and absolute number of NK cells (CD45+CD3-NK1.1+NKp46+) among lymphocytes was decreased in all of these tissues, whereas increased number of NK cells were observed in bone marrow (Figure 1G; Supplemental Figure 2A).”

Comment 19: Figure 1, G - CD27 and CD11b are used to define maturation stages within NK cells. Here the authors are analyzing group 1 ILC instead (containing both NK cells and ILC1, especially in the liver). It would be better to pre-gate on Eomes+ or CD49b+ NK cells for this analysis.

Response 19: We apologize for the lack of details in this analysis. We have pre-gate CD49b+ NK cells for the maturation stages analysis. We have now added this statement in our revised manuscript and figure legend (page 8; line 149-151)

“The maturation of cNK cells (gated by CD45+CD3-NK1.1+NKp46+CD49b+) from blood, bone marrow, lung, liver, spleen, and lymph nodes were assessed, based on the expression of CD11b and CD27.”.

Comment 20: Supplementary Figure 1, A - The NKp46+CD3- gate seems to cut the population, not all cells are included. y-axis labeling of spleen CD3-NKp46+ cells (%) is not correct.

Response 20: Thanks, we have corrected these errors and shown in our revised supplementary Figure 2A.

Comment 21: Figure 2, D-G - Did the authors analyse the ILC1/NK compartment of the tumor? What is the abundance and phenotype of these cells dependent on Prdm1 expression? Proper Crecontrols should be used (see above).

Response 21: We appreciate the suggestions from the reviewer. As request, we have now added the analysis of cNK/ILC1s population in the context of tumor. The proportion changes of cNK cells and ILC1s in Prdm1_Δ_Ncr1 mice was similar with the no tumor-burden condition, while the number of both cNK cells and ILC1s decreased in tumor bearing liver (revised Figure 7D). These contents have been updated in our revised manuscript (page 23; line 479-481):

“The proportion changes of cNK cells and ILC1s in Prdm1ΔNcr1 mice was similar with the no tumorburden condition, while the number of both cNK cells and ILC1s have significant decreased in tumor-bearing liver (Figure 7D).”.

The reason why we did not use Cre-controls was described in comment 1.

Comment 22: Figure 2, H - Prdm1-deficient NK and ILC1 produce less Ifng in response to in vitro stimulations with Il-12 and /or Il-18, and bulk Seq analysis (Fig 3F) shows reduced Il12rb2 expression. Does the expression of cytokine receptors correlate with the maturation of NK cells? This could be analyzed from the single-cell RNA-seq dataset. The statistical significance of %Ifng after Il12/Il18 stimulation should be revisited (see above).

Response 22: We thank the reviewer for the suggestions. To address this question, we explored the expression of IL-12 and IL-18 receptors in cNK and ILC1 clusters. Within cNK clusters, Il12rb2, Il18r1 and Il18rap was highly expressed in Prf1hi and Cxcr3hi cNK clusters (revised Supplemental Figure 6H), indicating the IL-18 receptor expression correlated with the NK cell maturation. While in ILC1, these receptors mostly expressed on Il7r_hi and _Gzmb_hi ILC1 clusters (revised Supplemental Figure 7C). Significant decreased of _Il18r1 expression in Prdm1_Δ_Ncr1 cNK cells and ILC1s may associated with the impaired ability to produce IFN-γ. We now added this analysis (page 18; line 364-368):

“Within cNK cells, Il12rb2, Il18r1 and Il18rap was highly expressed in Prf1hi and Cxcr3hi cNK clusters (Supplemental Figure 6I), indicating the IL-18 receptor expression correlated with the NK cell maturation. While in ILC1, these receptors mostly expressed on Il7r_hi and _Gzmb_hi ILC1 clusters (Supplemental Figure 7D). Significant decreased of _Il18r1 expression in Prdm1ΔNcr1 cNK cells and ILC1s may associated with the impaired ability to produce IFN-γ.”.

The un-paired t test of IFN-γ production was displayed in revised supplemental Figure 12 B. Difference in IFN-γ production was found to be not significant when analyzed using an unpaired ttest in original Figure 2 H. However, significance was observed in tumor-bearing liver cNK cells and ILC1s, specifically under the context of IL-12/IL-18 stimulation, as depicted in the original Figure 7E using an unpaired t-test. These variations may be attributed to differences among different littermates. Despite these variations, the trend remains consistent with our overall conclusions. We believe that employing a paired t-test between littermates could be also meaningful. As such, we kept both statistical methodologies to ensure a thorough evaluation.

Comment 23: Figure 3, A-E - For bulk sequencing analysis, splenic CD3-NK1.1+NKp46+ were isolated. This population also contains ILC1 in the spleen (e.g. Flommersfeld et al.), although much less abundant compared to NK cells, and compared to the liver compartment. However, have the authors tested the abundance of splenic ILC1 in Prdm1-deficient mice, which may impact the gene expression data? In line with this the detection of altered Cxcr6 expression in Figure F, which is usually expressed by ILC1 rather than NK cells, may indicate an alteration in ILC1 numbers. The authors should validate the altered expression of CXCR6, Itga1, and Cx3cr1 on NK cells by flow cytometry.

Response 23: We cited the work of Flommersfeld et al. into our manuscript and have expanded our Results section to include the following information (page 19; line 377-385):

“Previous research found that spleen NK cells could be divided into three distinct groups based on their expression levels of CD27, CD62L, CD49a, and CD49b (52). CD27+CD62L- NK cells have remarkable high expression of Batf3, while it was only barely expressed in CD27+CD62L+ and CD27-CD62L+ NK cells (52). Based the sequencing data published by Flommersfeld et al., (GSE180978), a notable negative correlation was observed between the expression levels of Prdm1 and Batf3 (Supplemental Figure 8I). On top of that, our findings unveiled the negative regulatory influence of Prdm1 on Batf3 within both spleen and liver NK cells. This discovery highlights a potential upstream mechanism that may influence the hemostasis of the spleen NK cell subpopulations through Batf3.”.

We validated the expression of CD49a (Itga1) and CX3CR1 in liver cNK cells and ILC1s in our revised manuscript, which is described in our revised manuscript (page 9; line 170-174, page 14; line 231-233):

“Increased CD49a expression was also observed in Prdm1ΔNcr1 liver ILC1s, while it showed decreased expression in NKp46+ cells in the liver, bone marrow, and lymph nodes (Supplemental Figure 2, F and G).”, “The percentage of CX3CR1+ cNK cells was significantly decreased in multiple tissues of Prdm1_Δ_Ncr1 mice, while the proportion of CX3CR1+ ILC1 was increased in the liver (Figure 3F).”

Comment 24: Figure 3, F - Tnfsf26: which gene is this? is this a typo? Is a function of this gene in NK cells reported? Altered Batf3 expression suggests an impact on ILC1-like NK cells (Flommersfeld et al).

Response 24: We are very sorry for our mistakes. We have removed Tnfrsf26 from the heatmap.

Comment 25: Figure 3, G-J refer to Kallies data?! 

Response 25: Kallies‘s data has mentioned the reduced GzmB expression in Blimp1gfp/gfp mice. However, compared with Kallies’s study, we further analyzed the GzmB and Perforin expression in different mature stages of NK cells. Reduced GzmB expression not only due to the less mature phenotype in Prdm1-deficient NK cells, highlighting the role of Prdm1 in regulating NK cell function. So, we added these contents in the revised manuscript (page 12; line 233-242):

“Lower GZMB and PRF1 production was observed in Prdm1-deficient splenic cNK cells, liver cNK cells and ILC1s (Figure 3, H-K; Supplemental Figure 4, A-I). Notably, the proportion of GZMB+ and PRF1+ cNK cells was decreased among almost all of the maturation stages of cNK cells (Figure 3, J and K). The relative mean fluorescent intensities (MFIs) of GZMB and PRF1 consistently show a reduction across all developmental stages in PrdmΔNcr1 NK cells (Supplemental Figure 4, H and I). Yet, no statistical difference of PRF1 was found within the CD11b-CD27+ and CD11b+CD27+ subsets, likely due to the relatively lower perforin levels in these populations (Supplemental Figure 4I). These findings suggest that Prdm1 may directly influence cytotoxic molecule in NK cells, rather than impacting their anti-tumor abilities solely by affecting the maturation phenotype of Prdm1-deficient NK cells.”

In Discussion section (Kallies’s work is cited here in revised manuscript) (page 24; line 500-502):

“Our results not only confirmed a decrease in cytotoxic molecules in Prdm1-deficient NK cells (31) but also showed that the reduction in Gzmb and perforin is not solely attributable to the diminished maturation of these cells.”

Comment 26: Figure 3, G, I - How do the authors explain the high variability of GzmB and Prf1 in Prdm1+/+ cells? 2 samples have comparable values to Prdm1-deficient cells.

Response 26: This may be due to the inherent differences in MFI among different samples. In the revised version, we have added data on percentages, which exhibit much less variability (Figure 3, H and I). The MFIs of GZMB and PRF1 are moved to supplemental Figure 4 E and F.

Comment 27: Did the authors test the mice for potential germline recombination of the floxed allele, which has been suggested as a potential problem of Ncr1cre?

Response 27: We appreciate the insightful comments provided by the reviewer, and this is a really good question. In Prdm1fl/fl mice, germline recombination typically results in a systemic knockout of Prdm1, which can lead to embryonic lethality. Given that mice were successfully born in the current study, it is almost unlikely that germline recombination of Prdm1 occurred due to leaky expression of Cre.

To confirm this issue, we isolated splenocytes and assessed Prdm1 expression using qPCR. We observed no significant difference in Prdm1 expression between splenocytes from Prdm1+/+ and Prdm1ΔNcr1 mice (revised Figure 1F). This also indicated that germline recombination issues are unlikely to be present in the Prdm1ΔNcr1 mice.

Comment 28: Histograms do not show MFI

Response 28: We appreciate the comments provided by the reviewer. The MFI value was omitted.

Comment 29: Supplementary Figure 4, B - FACS plot labelling: Typo, Histograms do not show MFI.

Response 29: We sincerely thank the reviewer for careful reading. The typo in this figure was corrected. The MFI is omitted.

Comment 30: Figure 4, A - What are the cells in the red cluster in the middle of the UMAP, do they belong to B cells? Why do they cluster so separately? It is interesting, but also surprising that NK and ILC1 cluster map so far apart from each other (rather with CD8 or B cells? or NKT cells) - do the authors have any comments?

Response 30: We sincerely apologize for the mistakes in labeling a group of cells in our previous analysis. Upon a thorough re-evaluation, we have corrected the labels of several cell clusters that were previously misidentified. The revised heatmap (revised Supplemental Figure 5C) represents the marker genes for each cluster. Additionally, in our updated analysis (revised Figure 4A), we have included clusters for Epithelial cells, CD4+ T cells, NKT cells, and Kupffer cells. Please note, the red cluster identified in the center of the original heatmap corresponds to the CD4+ T cells.

We checked the markers of cNK cell and ILC1 clusters and confirmed they are labeled correctly, as Ncr1 and Klrb1c (NK1.1) was highly expressed in these clusters compared to others (revised Supplemental Figures 5E).

Comment 31: Does Junb expression correlate with the maturation stages of NK cells?

Response 31: Our previous research indicated that during the maturation process of NK cells, there was a decrease in the expression levels of Junb (negative correlation), whereas there was an increase in the expression levels of Prdm1 (Wang et al., J Clin Invest, 2018; Supplemental Figure 5c and Supplemental Figure 11).

Comment 32: The authors may consider validating their scRNA-seq data (e.g. by FACS analysis for highlighted markers, eg. cKit, Tcf7, Gzma, Cxcr3).

Response 32: We appreciate the suggestion from the reviewer. We validated several marker genes, including Gzmb, Prf1, and Cx3cr1 by FACS, as shown in the revised Figure 3 F-K. Currently, FACS cannot distinguish liver NK cells into as many distinct clusters as can be achieved through scRNAseq analysis. However, we expect that as technology progresses, we will be able to enhance our validation of the scRNA-seq data.

Comment 33: It is a bit unclear to me why authors refer to Cxcr3hi NK cells as tissue-resident. This is based on Cxcr3 and Ccr2 expression. To make this statement, a much more detailed analysis would be required. How are CD69, CD49a, or CXCR6 expression of these cells?

Response 34: We appreciate the suggestion from the reviewer. The primary reason for classifying this specific cluster of NK cells as tissue-resident is derived from the differential expression genes (DEGs) and Gene Ontology (GO) analysis, which demonstrate significant chemokine receptor activity within this cluster.

To make this statement more clearly, we check the expression of the above markers, but only Cd69 had expression in cNK clusters, which was highly expressed in _Junb_hi and _Cxcr3_hi cNK cells (revised Supplemental Figure 6D). We also used top30 DEGs in ILC1s versus cNK to calculate the module score in all cNK clusters, as _Cxcr3_hi cNK had highest score among these clusters (revised Supplemental Figure 6D). This part has been updated in our manuscript (page 15; line 298-308):

“Expression of tissue-resident markers Cd69 was also highly expressed in this clusters (Supplemental Figure 6D). The enrichment of chemokine receptors in the genes upregulated in the Cxcr3_hi cluster implying a greater likelihood of this cluster being tissue-resident compared with other cNK cell clusters (Figure 4H). To further confirmed tissue-resident properties of this clusters, we calculated the module score based on top30 DEGs in ILC1 versus cNK clusters, including _Cxcr6, Itga1, Cd160, Cd226, etc. _Cxcr3_hi cNK clusters have the highest score among all cNK clusters (Supplemental Figure 6H), indicating the similarity with liver ILC1s. In the tumor microenvironment, reports indicated that NK cells could transform into ILC1s (25). If this conversion of cNK cells into ILC1s also occurred under normal physiological conditions, then _Cxcr3_hi cNK cell cluster might be the most susceptible to such transformation.”

Comment 35: The authors suggest that Prdm1 regulates chemokine receptor expression. An alternative explanation could be that this is an indirect effect of altering the abundance of NK cell subsets.

Response 35: We are sorry for lacking the details in these figures. The input cell number of each genotype has now been added in following figure legends. 

Figure 4F, “Proportions of cNK cells among total cNK cells (left; 211 cells in Prdm1+/+, and 141 cells in Prdm1ΔNcr1) and within clusters (right).”; Figure 5C, “Proportions of ILC1s among total ILC1s in different genotypes (left; 114 cells in Prdm1+/+, and 63 cells in Prdm1ΔNcr1) and within each cluster (right).”; Figure 6C, “Proportions of MDMs and KCs among total macrophages in different genotypes (510 cells in Prdm1+/+, and 624 cells in Prdm1ΔNcr1).”

To minimize the effects of discrepancies in input numbers between samples with different genotypes, we represented the relative proportions of each cluster within its specific genotype (e.g. Supplemental Figure 6B; Supplemental Figure 7B; Supplemental Figure 9B).

Comment 36: Supplementary Figures 6 and 7, A - The formatting of gene annotations does not fit the heat maps (the gene names on the last rows are missing).

Response 36: We apologize for our careless mistakes. We have now addressed these mistakes.

Comment 37: Supplementary Figures 6 and 7, What is the consequence of compromised mitochondrial function? Increase apoptosis?

Response 37: In our experiments, we did not find that Prdm1 has an effect on the apoptosis of NK cells. Conversely, previous studies have found that Prdm1 might inhibit the proliferation of NK cells (C. Kucuk, et. al., PNAS, 2011). We acknowledge that there is ongoing debate regarding the precise definition of NK cell exhaustion. In our experiments, no changes were detected in the expression levels of surface markers (TIGIT) associated with exhaustion on NK cells following the knockout of Prdm1. However, we did note a significant reduction in the cytokine secretion capacity and tumor control efficacy of NK cells after Prdm1 knockout. We prefer to say that the consequence of compromised mitochondrial function might be increased exhaustion. As we mentioned in discussion part (line 482-483), mitochondrial fragmentation has been confirmed to be closely associated with NK cell exhaustion in tumor (Zheng et al. Nature immunology, 2019). Although the evidence to define the exhausted NK cells in Prdm1_Δ_Ncr1 was not sufficient, our data may support the compromised mitochondrial functions, at least in part, associated with the exhausted phenotype of Prdm1_Δ_Ncr1 NK cells in cancer. 

We have discussed these points in our revised manuscript (page 26; line 529-543): 

“Mitochondria are pivotal organelles crucial for cellular metabolism. Disruptions in mitochondrial function have been linked to T Cell exhaustion, attributed to glycolytic reprogramming (66). Similarly, mitochondrial fragmentation has been closely associated with NK cell exhaustion (67).

However, the concept of NK cell exhaustion isn't as firmly established as it is for T cells. Exhausted NK cells should primarily exhibit diminished functions. This is characterized by a diminished ability to destroy tumor cells, a reduced capability to activate other components of the immune system, and compromised proliferation and survival rates. Additionally, this reduced functionality is associated with a decline in the expression of molecules responsible for cytotoxic activity, lower production of IFN-γ, and metabolic disturbances that may arise from mitochondrial dysfunction. While our current data is not sufficient to definitively classify these cells as exhausted NK cells, it supports that a subpopulation, referred to Junb_hi cluster, demonstrates an exhaustion-like phenotype. The significant increase in this cell population following _Prdm1 knockout in NK cells may potentially be one of the reasons why Prdm1ΔNcr1 mice lose their tumor-killing capacity. Whether the excessive expression of JunB in NK cells is also a contributing factor to their exhaustion, similar to T cells(65), requires further investigation.”.

Comment 38: Figure 5, Describing the scRNA Seq data, the authors are switching a lot between Figure 4 and Figure 5. Maybe a reorganization of the Figures (Figure 4: NK cell; Figure 5: ILC1) could help.

Response 38: We appreciate the reviewer’s suggestion. We have now reorganized the Figure 4 and Figure 5.

Comment 39: Figure 5, We suggest naming one of the ILC1 clusters "Gzmbhi" to keep it consistent with the FACS data.

Response 39: We agree with this excellent suggestion and have now renaming the “Gzmahi” ILC1 cluster as “Gzmbhi” ILC1 cluster.

Comment 40: Figure 5, C - How was the JunB score derived (which genes were used)?

Response 40: The JunB score was calculated based on the expression of marker genes in _Junb_hi cNK clusters (DEGs in _Junb_hi cNK cluster compared to other clusters, as shown in revised Supplemental figure 6A). The score was calculated using “AddModuleScore” R package.

Comment 41: Figure 5, G, I - The authors highlight Il17 signaling pathway, what is the impact of Il17 on NK/ILC1? Did the authors check for ILC3 (Rorc expression) within the ILC1 cluster?

Response 41: The enrichment of IL-17 signaling pathway in Il7r_hi ILC1 indicated that this cluster encompass ILC1s originate from the conversion of Rorγt+ ILC3s. Although the Rorc expression was undetectable in all ILC1 clusters, we found several ILC3 marker genes highly expressed in this clusters (e.g. Rora, Tmem176a, Tmem176b) according to the ILC3 transcriptomes (Robinette et al., _Nature Immunology, 2015). 

We have added these contents in our revised manuscript (page 17; line 341-344): 

“Several ILC3 signature genes, such as Rora, Tmem176a, and Tmem176b (45), highly expressed in this cluster (Supplemental Figure 7D). Considering the close relationship between IL-17 mediated immunity response and ILC3 (1, 46), it is plausible that _Il7r_hi ILC1 cluster may be attributed, at least in part, to potential plasticity between ILC1 and ILC3 subsets.”.

Comment 42: Figure 5, The authors detect more Ly49E+ cytotoxic ILC1 in Prdm1fl Ncr1cre mice.

How does this observation fit to the reduced cytotoxicity of NK cells?

Response 42: The proportion of _Klra_hi ILC1 was increased, while the _Gzmb_hi ILC1 was decreased in _Prdm1_ΔNcr1 mice. Moreover, total number of three ILC1 cluster was reduced in _Prdm1_ΔNcr1 mice.

Comment 43: Line 350/351: Citation required.

Response 43: We added the respective reference. (reference 55 and 56).

Comment 44: Figure 6, The Cell-chat analysis provides interesting suggestions, but none are experimentally addressed. It is also difficult to evaluate these analyses: are any of the Mac subsets altered in frequency or phenotype in either genotype? This could be analyzed from the single-cell data in Fig 4. At the very least, flow cytometric validation of predicted shifts in the Mac compartment should be confirmed.

Response 44: We gratefully thanks for these valuable suggestions. As requested, we analyzed macrophages and validated some of the scRNA-seq data by flow cytometry. We have re-written this part with the analysis of altered proportion of two macrophage clusters (Kupffer cells and Monocyte-derived macrophages) (page 20-21; line 399-436):

“The scRNA sequencing analysis identified two well-established subpopulations of liver macrophages: the resident Kupffer Cells (KCs) and the Monocyte-Derived Macrophages (MDMs) (Figure 6, A-C; Supplemental Figure 9A). When comparing the total proportion of macrophages within the immune cell population of the liver between WT and Prdm1ΔNcr1 mice, there is an increase in Prdm1ΔNcr1 mice (Figure 6C). To confirm these findings, we utilized flow cytometry to define macrophages, including both KCs and MDMs, gating by CD45+Ly6G-F4/80+CD11b+ (Figure 6D).

Our analysis showed that, following the deletion of Prdm1 in Group 1 ILCs, there is a significant increase in both the proportion and number of macrophages in the liver (Figure 6D).

According to the transcriptional profile, liver macrophages further clustered and were labeled as “Ly6c2_hi”; “_Cxcl2_hi”; “_Ear2_hi” MDMs, and “_Mrc1_hi”; “_C1q_hi” KCs (Figure 6A, Supplemental Figure 9, A-E). Increased proportion of MDMs and KCs was observed in _Prdm1ΔNcr1 cells (Supplemental Figure 9B). Within MDMs clusters, Ly6c2_hi MDMs mainly compose of _Prdm1+/+ cells, while Prdm1ΔNcr1 cells concentrated in Cxcl2_hi cluster (Figure 6C). The scRNA-seq data reveal that following Prdm1 knockout in NKp46+ cells, there is a decrease in the proportion of KCs within the macrophage population, while the proportion of MDMs increases (Figure 6D). CX3CR1, a chemokine receptor, is extensively utilized to distinguish KCs and MDMs within macrophages. Cells expressing CX3CR1 are identified as MDMs, whereas those without CX3CR1 expression are categorized as KCs (56). Employing flow cytometry and leveraging CX3CR1 expression, we assessed the ratios of KCs and MDMs. However, diverging from the scRNA-seq findings, flow cytometry indicates that post-Prdm1 knockout in group 1 ILCs, there is a minor increase in the proportion of KCs within the total liver macrophages, and a decrease in the proportion of MDMs (Figure 6D; Supplemental Figure 9B). This discrepancy could stem from the different bases of classification: scRNA-seq defines KCs based on gene expression profiles, whereas flow cytometry differentiates between KCs and MDMs using the single surface marker, CX3CR1. Analysis of the macrophage subsets identified by scRNA-seq reveals that, while MDM clusters generally show high CX3CR1 expression, there exists a subset within MDMs, labeled _Mrc1hi, that also exhibits high levels of CX3CR1 (Supplemental Figure 9C). Consequently, if flow cytometry solely employs CX3CR1 for differentiating between KCs and MDMs, it could result in disparities when compared to scRNA-seq outcomes. Both KCs and MDMs has significantly increased in Prdm1ΔNcr1 mice, which was consist with the scRNA-seq data (Supplemental Figure 9, B and F). Despite the decrease in the proportion of Ly6c2hi MDMs in Prdm1ΔNcr1 mice, the expression levels of Ly6c2 exhibited minimal variation between WT and Prdm1ΔNcr1 mice (Supplemental Figure 9D). Intriguingly, within certain cellular subsets, notably the Ear2hi cluster, the Ly6c2 expression levels in KO mice were found to be higher than those in WT mice. Additionally, we employed flow cytometry to examine Ly6C expression within the macrophages. Similar with the scRNA-seq findings, there were no notable differences in Ly6C expression levels between WT and KO mice (Figure 6E; Supplemental Figure 9G).”.

The changes of the macrophage compartment indicated the potential influence of functional NK cells to macrophages. We have revised these parts in our results and discussion (line 590-601). However, to address more analysis on macrophage is worthy but would go beyond the scope of this manuscript, which will be a direction of our further work.

Comment 45: Figure 6, C1qhi Mac only are few cells/events, and interactions (or cells?) seem to be gone in the Prdm1-floxed mice. Is that true? Does it make sense to perform cell-chat analysis on so few cells?

Response 45: We have now added KCs to the cell-chat analysis, and this cluster was belonged to C1qhi KCs. We have revised the analysis of corresponding parts in our manuscript (page 20-21; line 408-428):

“According to the transcriptional profile, liver macrophages further clustered and were labeled as “Ly6c2_hi”; “_Cxcl2_hi”; “_Ear2_hi” MDMs, and “_Mrc1_hi”; “_C1q_hi” KCs (Figure 6A, Supplemental Figure 9, A-E). Increased proportion of MDMs and KCs was observed in _Prdm1ΔNcr1 cells (Supplemental Figure 9B). Within MDMs clusters, Ly6c2_hi MDMs mainly compose of _Prdm1+/+ cells, while Prdm1ΔNcr1 cells concentrated in Cxcl2_hi cluster (Figure 6C). The scRNA-seq data reveal that following Prdm1 knockout in NKp46+ cells, there is a decrease in the proportion of KCs within the macrophage population, while the proportion of MDMs increases (Figure 6D). CX3CR1, a chemokine receptor, is extensively utilized to distinguish KCs and MDMs within macrophages. Cells expressing CX3CR1 are identified as MDMs, whereas those without CX3CR1 expression are categorized as KCs (56). Employing flow cytometry and leveraging CX3CR1 expression, we assessed the ratios of KCs and MDMs. However, diverging from the scRNA-seq findings, flow cytometry indicates that post-Prdm1 knockout in group 1 ILCs, there is a minor increase in the proportion of KCs within the total liver macrophages, and a decrease in the proportion of MDMs (Figure 6D; Supplemental Figure 9B). This discrepancy could stem from the different bases of classification: scRNA-seq defines KCs based on gene expression profiles, whereas flow cytometry differentiates between KCs and MDMs using the single surface marker, CX3CR1. Analysis of the macrophage subsets identified by scRNA-seq reveals that, while MDM clusters generally show high CX3CR1 expression, there exists a subset within MDMs, labeled _Mrc1hi, that also exhibits high levels of CX3CR1 (Supplemental Figure 9C). Consequently, if flow cytometry solely employs CX3CR1 for differentiating between KCs and MDMs, it could result in disparities when compared to scRNA-seq outcomes.”.

Comment 46: Figure 6, C - Here the interactions of both Mac+ILC1 and Mac+NK are shown together. It would be interesting to separate this analysis (also Suppl. Fig 9A-B) into comparisons of Mac+ILC1 vs Mac1+NK from WT or Prdm1fl Ncr1 mice.

Response 46: As request, we re-analyzed this part in each genotype, which was showed in the Supplemental Figure 10. These data have now been described in (page 22; line 445-447).

“The reduction of interaction mostly occurred in the cross-talk of ILC1-MDM and ILC1-KC, whereas no difference was observed in cNK-MDM and cNK-KC interaction (Supplemental Figure 10, A-H)”

Comment 47: Supplementary Figure 9, A, B - Is this analysis using WT and Prdm1fl Ncr1cre dataset together? 

Response 47: Yes, we used WT and Prdm1_Δ_Ncr1 data together. As the request above, we separate this analysis from WT or Prdm1_Δ_Ncr1 Ncr1 mice. These data have now been described in (page 22; line 445-460):

“The reduction of interaction mostly occurred in the cross-talk of ILC1-MDM and ILC1-KC, whereas no difference was observed in cNK-MDM and cNK-KC interaction (Supplemental Figure 10, A-H). A reduction in the interaction of ligand-receptor, such as Mif-CD74, Cxcl16-Cxcr6, and Cxcl10-Cxcr3 was observed in Prdm1ΔNcr1 mice compared to Prdm1+/+ mice (Supplemental Figure 11). Compared to Prdm1+/+ mice, the information flow of CXCL and MIF pathways significantly decreased in Prdm1ΔNcr1 mice (Figure 6, H and I; Supplemental Figure 10, B, D, F, and H). These pathways play a crucial role in facilitating macrophage migration. The CXCL signaling was sent from Ly6c2_hi _Cxcl2_hi MDMs and _C1q_hi KC, targeting all ILC1 clusters and _Cxcr3_hi cNK cell clusters (Figure 6J). Of note, although the population of _Cxcl2_hi macrophage primarily comprised cells from _Prdm1ΔNcr1 mice, the interaction within the CXCL pathway between macrophages and group 1 ILCs was obviously less than Prdm1+/+ sample (Figure 6J). These changes could be linked to a decreased population of ILC1s and Cxcr3_hi cNK cell cluster in _Prdm1ΔNcr1 mice, implying that the homeostasis of _Cxcl2_hi macrophages required sufficient signals from cNK cells and ILC1s. The impaired CXCLCXCR interactions might subsequently lead to reduced recruitment and activation of group 1 ILCs and macrophages within the tumor microenvironment.”.

Comment 48: Figure 7, A-C -What is the consequence/interpretation of reduced Mitotracker staining? Any metabolic assays performed? The definition of NK cell "exhaustion" is unclear, is reduced IFNg enough for that? Is the concept of NK cell exhaustion clearly established? Only shortly touched upon in the discussion, the rationale for suggesting an exhausted phenotype, should be explained.

Response 48: MitoTracker was used to assess the mitochondrial mass. The reduced staining indicated compromised mitochondria function, which associated with mitochondrial fragmentation.

We believe that the exhaustion of NK cells is not as well-established a concept as it is for T cells. The purpose of detecting mitochondria in this study is to provide evidence for the relationship between Prdm1 and the exhaustion of NK cells. In the discussion section, we have added the following content (page 26; line 529-543):

“Mitochondria are pivotal organelles crucial for cellular metabolism. Disruptions in mitochondrial function have been linked to T Cell exhaustion, attributed to glycolytic reprogramming (66). Similarly, mitochondrial fragmentation has been closely associated with NK cell exhaustion (67).

However, the concept of NK cell exhaustion isn't as firmly established as it is for T cells. Exhausted NK cells should primarily exhibit diminished functions. This is characterized by a diminished ability to destroy tumor cells, a reduced capability to activate other components of the immune system, and compromised proliferation and survival rates. Additionally, this reduced functionality is associated with a decline in the expression of molecules responsible for cytotoxic activity, lower production of IFN-γ, and metabolic disturbances that may arise from mitochondrial dysfunction. While our current data is not sufficient to definitively classify these cells as exhausted NK cells, it supports that a subpopulation, referred to Junb_hi cluster, demonstrates an exhaustion-like phenotype. The significant increase in this cell population following _Prdm1 knockout in NK cells may potentially be one of the reasons why Prdm1ΔNcr1 mice lose their tumor-killing capacity. Whether the excessive expression of JunB in NK cells is also a contributing factor to their exhaustion, similar to T cells(65), requires further investigation.”.

Comment 49: Figure 7, x-axis labelling (MFI) of histograms is not correct. Do bar graphs and FACS plots show the same data? Does the number in the FACS plots indicate the MFI? If so, the FACS plots do not show representative samples?

Response 48: We appreciate the valuable comments provided by the reviewer. In the revised Figure 7, the MFI values have been removed. Bar graphs now display summary data from FACS histograms.

A representative sample close to the group's mean value was chosen for display in the histograms.

Comment 50: Figure 7, D - How are these data different from Figure 2H? Why is it now called "exhaustion", but not in 2H? Is the detected IFNg only driven by ex vivo stimulation with Il12/Il18? As above, a "standard" 4h assay should also be provided to allow better interpretation of potential differences. In the introduction, the authors cite the Ducimetiere study (Ref 5) highlighting "the primary function of ILC1 in suppressing the seeding of metastatic tumor cells in liver tissue". Thus, it would be interesting to test Ifng production by liver ILC1 and NK cells ex vivo at early time points of tumor inoculation.

Response 50: Tumors grow and proliferate within tissues, constituting one of the major causes of lymphocyte exhaustion. This part of the current study aims to investigate whether Prdm1 aids NK cells or ILC1 in resisting the exhaustion induced by malignant tumors. Specifically, we seek to ascertain whether the absence of Prdm1 renders NK cells or ILC1 more susceptible to exhaustion within the tumor microenvironment. Therefore, we will consider the capacity to secrete IFN-γ upon IL-12/IL-18 stimulation as one indicative aspect of exhaustion. It's crucial to emphasize that this assessment serves as only one piece of evidence, not the sole determinant. Overnight stimulation is a conventional method for studying NK cells and has been widely used across different laboratories, including our lab (e.g. Bream et al., Blood, 2003; Yu et al., Immunity, 2006; Wang et al., J Clin Invest, 2018). It's essential to clarify that our approach does not involve stimulating with tumor cells to evaluate the secretion capacity of IFN-γ by NK cells or ILC1.

Reviewer 2 (Public Review):

Summary:

This study offers a significant advancement in understanding liver innate lymphoid cell (ILC) biology by elucidating the role of the transcription factor Prdm1. It shows that Prdm1 is crucial in maintaining the balance between conventional natural killer (cNK) cells and ILC1s in the liver, with knockout models revealing a vital role in cancer defense mechanisms. Despite not affecting direct cytotoxicity, Prdm1 deficiency leads to increased cancer metastasis and reduced secretion of key molecules like IFN-γ, pointing to its importance in immune regulation. The use of single-cell RNA sequencing further underscores Prdm1's role in cellular communication within the liver's immune milieu. This study is a robust contribution to the field, providing insights that could inform new immunotherapy approaches for liver cancer.

Strengths:

The study's strength lies in its comprehensive approach, combining the specificity of Prdm1 conditional deletion in Ncr1-cre mice with integrative omics analyses and cutting-edge cytometry to delineate Prdm1's role in liver Type 1 ILC biology and its functional implications in tumor immunity. This multifaceted strategy not only clarifies Prdm1's influence on ILC composition and maturation but also conveys potential therapeutic insights for liver cancer immunotherapy.

We sincerely appreciate your interest and critical assessment of our manuscript. We have carefully read your comments and suggestions, and I am truly grateful for your expert guidance. We have worked on addressing each of your concerns and comments, and below we provide a point-to-point response. Please find the detailed responses below:

Weakness

Comment 1: A notable weakness of the study is the limited scope of in vivo disease models, primarily relying on the B16F10 melanoma model, which may not fully capture the complex behavior of Type 1 ILCs across diverse cancer types. Furthermore, the absence of direct human data, such as the effects of PRDM1 deletion in human NK cells or stem cells during their differentiation into NK and ILC1, leaves a gap in translating these findings to clinical settings.

Response 1: We appreciate the reviewer for raising these important points, which we see as a unique opportunity for future work to transform our understanding of Prdm1 and its targets as opposed to a weakness of the present study. 

In our revised manuscript, we have discussed these limitations of our study (page 29; line 602-609):

“While our findings underscore the importance of Prdm1 in liver cNK cells and ILC1s tumor immune surveillance, it does not be validated in human NK cells, whereas previous studies have found that PRDM1 might inhibit the proliferation and function of human NK cells (33, 73). Furthermore, we not provided an in-depth evaluation in multiple tumor models. Further research may provide deeper insight into the role of PRDM1 in the anti-tumor function of human NK cells, enabling a more direct investigation of its application in cancer therapies. Given its important role in preserving liver cNK cells and ILC1s functional heterogeneity, enhancing Prdm1 function in human NK cells could potentially be a strategy to promote NK cell-based immunotherapy for cancer.”.

Recommendations For The Authors:

(Introduction) 

Comment 2: Reference 1 appears slightly misplaced. You might find the nomenclature discussion in Spits et al., Nature Reviews Immunology, 2013, more appropriate.

Response 2: We are really sorry for our inaccurate descriptions. According to Spits et al., (Spits et al., Nature Reviews Immunology, 2013) and other related studies, we have now adopted a more appropriate nomenclature as “Conventional NK cells” correspond to “cNK cells”, “Type 1 innate lymphoid cells” to “ILC1s”, and “Group 1 ILC” as the collective name of cNK and ILC1s. 

The definition of these cells was described in the introduction (page 4, line 52-53; line58-62): 

“Group 1 ILCs consist of cNK cells and ILC1s (1, 2), with distinct developmental trajectories and effect molecules (3).”, “In a state of homeostasis, liver group 1 ILCs (CD45+CD3-NK1.1+NKp46+) can be discriminated into cNK cells and ILC1s by the differential expression of CD49a and CD49b (2): cNK cells are marked by the expression of CD49b, while liver ILC1s exhibit a distinctive positivity for CD49a. Tumor Necrosis Factor Related Apoptosis Inducing Ligand (TRAIL) is also expressed on liver ILC1s, but not on cNK cells (10, 11).”. 

We also describe cNK and ILC1 phenotypes in our scRNA-seq data, as shown in page 13; line 259-261: 

“cNK cells expressed high levels of Itga2 (CD49b) and Eomes, while ILC1s had high levels expression of Itga1 (CD49a) and Tnfsf10 (Supplemental Figure 5, F and G).”.

Comment 3: It has come to my attention that Reference 9 has been retracted. I recommend removing this citation to maintain the integrity of your references (https://doi.org/10.1182/blood.2023022801).

Response 3: We thank the reviewer’s comment and we now have removed this citation.

Comment 4: For a more comprehensive context around reference 15, consider citing Thierry Walzer's work ([https://rupress.org/jem/article/211/3/563/41636/T-bet-and-Eomes-instruct-thedevelopment-of-two)]) which aligns closely with your discussion.

Response 4: We agree with the reviewer’s suggestion and have added this citation in our introduction (page 4; line 64-66):

“Liver environment facilitated T-bet expression in the early stage of NK cells development, which results in Eomes repression. The repression of T-bet is required for Eomes+ NK cells (17).”.

(Results) 

Comment 5: The NK cell signature referenced in 32 has been questioned for its reliability as discussed by Cursons et al., CRI 2019 (https://pubmed.ncbi.nlm.nih.gov/31088844/). Reanalysis of data in Figure 1 B/C and Supplementary Figure 1 with the refined NK cell signature from Curson's work would be advantageous.

Response 5: We thank the reviewer’s comment. As requested, we reanalyzed our data using the refined NK cell signature from Cursons et al. (revised Figure 1 A-C; revised Supplemental Figure 1). Of note, the overall survival of liver cancer (LIHC) patients only reached statistics significance when compared high and low expression of refined PRDM1-NK signature with a median cutoff (Figure 1, A-C). The overall survival performed with quartile high and low expression of refined PRDM1-NK signature was moved to supplemental figure 1, G-I. 

The original text is: “Examination of 363 liver hepatocellular carcinoma (LIHC) patient samples from The Cancer Genome Atlas (TCGA) revealed a positive correlation between the expression of NK cell-associated genes (NCR1, NCR3, KLRB1, CD160, and PRF1) (32) and PRDM1 expression (Figure 1A). Patients with top and bottom quartiles of NK-PRDM1 signature expression were chosen for survival analysis (Figure 1B). Notably, patients with the NK-PRDM1_hi signature had better overall survival compared to the these with NK-_PRDM1_lo signature (Figure 1C). Similar results were also found in skin cutaneous melanoma (SKCM, n=454) and lung adenocarcinoma (LUAD, n=497) patients (Supplemental Figure 1, A-F). These data suggested that _PRDM1 in NK cells might be essential for immune surveillance in some solid tumors, including liver cancer. These findings prompted us to investigate the impact and mechanism of PRDM1 in NK cells and ILC1 within the context of liver cancer.”

We have rewritten this part in our revised manuscript (page 7; line 119-132): 

“Examination of 363 liver hepatocellular carcinoma (LIHC) patient samples from The Cancer Genome Atlas (TCGA) revealed a positive correlation between the expression of NK cell-associated genes (34) (NCR1, KLRB1, CD160, PRF1, etc.) and PRDM1 expression (Figure 1A). The patients are ordered from highest to lowest based on the expression of NK-Prdm1 for survival analysis (Figure 1B). Notably, patients exhibiting higher levels of NK-PRDM1 expression (above the median) experienced better survival outcomes compared to those with lower levels of NK-PRDM1 expression (below the median) (Figure 1C). Similar results were also found in skin cutaneous melanoma (SKCM, n=454) and lung adenocarcinoma (LUAD, n=497) patients (Supplemental Figure 1, A-F). Patients within the highest quartile of NK-PRDM1 signature expression demonstrated enhanced overall survival, a result that achieved statistical significance in LUAD and SKCM patients (Supplemental Figure 1, G-I). These data suggested that PRDM1 in NK cells might be essential for immune surveillance in solid tumors, including liver cancer, and prompted us to investigate the function and mechanism of PRDM1 in NK cells and ILC1 within the context of liver cancer.”.

Comment 6: The origin of the Ncr1-cre mice utilised should be clarified; is this the line developed by Eric Vivier? (https://www.pnas.org/doi/10.1073/pnas.1112064108).

Response 6: We did not use the line developed by Eric Vivier, our Ncr1-cre mice was purchase from Shanghai Model Organism Center, Inc.. We described this in our method parts (page 29-30; line 612-614): 

“Prdm1fl/fl mice were purchased from The Jackson Laboratory. Ncr1-iCre and B2m-/- mice were purchased from Shanghai Model Organisms Center, Inc.. Six- to twelve-week-old littermates were used for the experiment.”

Comment 7: Considering the known reduction of Ncr1 expression in Ncr1-cre mice and its implications, it is recommended to repeat the B16F10 experiments with the correct control, Ncr1cre/+ Prdm1+/+.

Response 7: This is an excellent question, and it has been raised by another reviewer and comprehensively answered (Reviewer 1, Comment 1). The answer is below: 

The expression of Cre and the insertion of loxP sequences both have the potential to influence gene expression. This is because the region where loxP is inserted may contain regulatory sequences for the gene of interest. Ncr1-Cre is a frequently used transgenic mouse model in our laboratory. In our prior research, we also had concerns about the possible impact of Cre on NKp46 expression, which could lead to a decline in NK cell function. Therefore, in our previous studies focused on Smad4 expression in NK cells, we conducted similar experiments. In Figure 6 of our published paper in the Journal of Clinical Investigation (Wang et al., J Clin Invest, 2018), we compared NKp46iCreTgfbr2fl/flSmad4fl/WT with NKp46-iCreTgfbr2fl/flSmad4fl/fl. Although the primary purpose is to establish Smad4's independence from TGF-β, it also allows for a comparison between Smad4fl/fl and Smad4fl/WT in the presence of Cre. In the critical phenotype we assessed, NKp46iCreTgfbr2fl/flSmad4fl/fl (compared with NKp46-iCreTgfbr2fl/flSmad4fl/WT) exhibited the same phenotype as NKp46-iCreSmad4fl/fl (compared with NKp46WTSmad4fl/fl). This suggests that Cre's influence on NK cells may be within a reasonable and controllable range. Furthermore, in contrast to the decrease in Ncr1 expression caused by Cre, the reduction in the expression levels of genes targeted by Loxp knockout, such as Prdm1 in this study (Figure 1 E), is more significant. Therefore, with the current techniques and research methods, we believe that the data provided in this study can support the role of Prdm1 in NK cells.

Comment 8: The proportion of ILC1 in wild-type mouse livers is notably higher than standard references. Could you confirm whether liver perfusion was performed before analysis? This procedure was not clearly detailed in the methods section.

Response 8: We apologize that we did not provide enough detail regarding this point in our original method. We had performed the liver perfusion before analysis. This has now been clarified in the method section of the revised text (page 30-31; line 630-636): 

“Mice were perfused with 1◊ PBS by portal vein puncture before harvesting tissues. Liver and lung was digested with 0.05% collagenase II for 30 minutes and filtered through 70 µm cell strainers, and mononuclear cells were isolated after subjected to density gradient using 30% and 70% percoll. Spleen were also removed and pressed through 70 µm filterers to obtain splenocytes. Peripheral blood mononuclear cells were obtained from peripheral blood after lysis of red blood cells (Biolegend, 420301). Flushing femurs and mechanical disruption of inguinal lymph nodes were performed to obtain cells from bone marrow and lymph nodes.”.

The lymphocyte proportions in mice from different laboratories may exhibit slight variations, possibly due to genetic background disparities. To minimize the influence of genetic backgrounds, paired littermates were used in the current study, wherein one is Prdm1 WT and the other has the Prdm1 gene knocked out in NK cells.

Comment 9: There appears to be inconsistency in reference formatting; for instance, Ref 39 does not match the formatting of other references. A thorough review of your citation format is suggested.

Response 9: We apologize for the inadvertent errors and we reviewed the citation format.

Comment 10: The information in Figures 2B and C may be better suited to the supplementary section as it does not significantly contribute to the main text.

Response 10: We agree with the reviewer’s suggestion and these are now moved to supplementary figures (Supplemental Figure 2).

Comment 11: The citation of reference 40 could be strengthened by including Sathe et al., 2014, which directly pertains to your findings (https://www.nature.com/articles/ncomms5539).

Response 11: We added the suggested reference.

Comment 12: Can the findings presented in Figure 2D/F be replicated using alternative models?

This would substantiate the versatility of your results.

Response 12: The current predominant in vivo tumor model for NK cells is primarily based on the use of B16F10 melanoma cells. These melanoma cells, with their low expression of MHC-I molecules, evade T cell-mediated immune surveillance, rendering them ideal targets for NK cells. Typically, this experimental melanoma metastasis assay involves tail vein injection, followed by nodules' detection in the lungs. To align with our investigation of liver-resident cNK and ILC1, we've introduced splenic injection (via the portal vein) and evaluated melanoma metastasis in the liver to reflect the anti-tumor capabilities of liver group 1 ILCs. We also explored subcutaneous tumor models, but we believe they may not effectively support Prdm1's role in cNK cells, particularly liver-resident NK cells and ILC1. While we've experimented with models using mouse liver tumor cells like Hepa 1-6, we found them less stable than B16F10 and less conducive to quantification. Should more suitable models or cells line emerge, we remain open to exploring them in future research.

Comment 13: The absence of in vitro killing assessments against B16F10 and YAC-1 leaves a gap in the NK cell characterisation which would be valuable to address.

Response 13: Isolating NK cells for ex vivo cytotoxicity assays typically requires stimulation with high concentrations of IL-2. Under such high IL-2 stimulation, many intracellular differences that contribute to difference in cytotoxicity, such as changes in transcription factors, are often masked. Another issue is that current ex vivo NK cell cytotoxicity assays often only isolate NK cells from the spleen. Liver-resident NK cells, on the other hand, are often limited in quantity and isolation methods, making it challenging to conduct ex vivo cytotoxicity assays effectively. If more sensitive detection methods become available, we will also incorporate ex vivo data into our future research endeavors.

Comment 14: The suggestion that NK cells produce IL-6 is indeed a bold one, and without additional validation through intracellular cytokine detection or ELISA, it may be prudent to omit these claims.

Response 14: We have checked the GSEA results, and found no valuable genes in IL-6 production.

Therefore, we have removed this figure.

Comment 15: The lack of fluorescence minus one (FMO) controls in Figure 3 and Supplementary

Figure 4 is noted; including these would enhance the validity of your gating strategies.

Response 15: As requested, we add the FMO controls in aforementioned figures.

Comment 16: There seems to be a minor mix-up in referring to Figure 4A in the scRNAseq results section, perhaps it was intended to refer to Figure 3A?

Response 16: We have corrected this part (line 247). We also double checked corrected the inaccuracies in the references to the figures. we apologize for the inadvertent errors.

Comment 17: The rich datasets generated from bulk and scRNAseq are commendable. However, I urge you to make these datasets publicly accessible with a GEO accession number.

Response 17: We appreciate the suggestion from the reviewer. We plan to upload our datasets when in the last version of our manuscript, which is also the request of the eLife policy.

Comment 18: Figure 4K is insightful, yet a similar analysis of the ILC1 cluster could provide a more rounded understanding.

Response 18: We thank the reviewer for the comments. We provide the similar analysis of ILC1s, as showing in revised Figure 5H. 

Comment 19: The metabolic RNA signatures featured in Supplementary Figure 6 are intriguing and warrant further validation, perhaps through Seahorse analysis. Such validation could merit their inclusion in the main figures.

Response 19: This is a very good suggestion. Currently, our data offer only limited indications in this context. We have chosen to validate some aspects of Prmd1's influence on cytotoxicity molecules. As for Prdm1's impact on other aspects of NK cells, such as metabolic functions, we may explore further in future research. Additionally, we hope that by publishing our research findings, laboratories worldwide can draw insights for their own studies and conduct relevant research based on this data.

Comment 20: It is difficult to discern whether the cells depicted in Figure 7D are truly tumorinfiltrating ILC1 or NK cells that have adopted ILC1-like characteristics. Intravenous injection of CD45-PE could clarify this distinction, and if they are the latter, it may be more appropriate to refer to them as ILC1-like cells.

Response 20: We completely agree with the reviewer's suggestion that "tumor-infiltrating lymphocytes" may not be accurate for the current experiment. Therefore, in the revised manuscript, we have changed it to "liver cNK or ILC1 from tumor-bearing livers.

eLife assessment

This important study demonstrates a link between an acute high fat diet, microglial metabolism and improved higher cognitive function. The evidence supporting the proposed mechanism in vivo is incomplete at this stage due to non-trivial technical limitations but the authors provide convincing in vitro metabolic characterization of primary microglia cultures to support the model. This work will be of interest to a broad audience in the field of neuroscience, metabolism, and immunology.

Reviewer #1 (Public Review):

In this study, Drougard et al. examined the consequences of an acute high fat diet (HFD) on microglia in mice. 3-day HFD influenced the regulation of systemic glucose homeostasis in a microglia-dependent and independent manner, as determined using microglial depletion with PLX5622. 3-day HFD increased microglial membrane potential and the levels of palmitate and stearate in cerebrospinal fluid in vivo. Using confocal imaging, respirometry and stable isotope-assisted tracing in primary microglial cultures, the authors suggest an increase in mitochondrial fission and metabolic remodelling occurs when exposed to palmitate, which increases the release of glutamate, succinate and itaconate that may alter neuronal metabolism. This acute microglial metabolic response following acute HFD is subsequently linked to improved higher cognitive function (learning and memory) in a microglia and DRP1-dependent manner.

Strengths:

Overall, this study is interesting and novel in linking acute high fat diet to changes in microglia and improved learning and memory in mice. The role for microglia and DRP1 in regulating glucose homeostasis and memory in vivo appears to be supported by the data. Palmitate (which is elevated in the CSF following acute HFD) is clearly used as a fuel by primary microglia ex vivo as determined using U-13C-plamitate tracing and metabolomics.

Weaknesses:

The authors suggest that utilisation of palmitate by microglia following HFD is the driver of the acute metabolic changes and that the release of microglial-derived lactate, succinate, glutamate and itaconate are causally linked to improvements in learning and memory. A weakness is that the authors provide no mechanistic link between beta-oxidation of palmitate (or other fatty acids) in microglia in vivo and the observed systemic metabolic and memory phenotypes. However, this reviewer acknowledges the technical difficulties of providing this evidence and approaches, such as microglia-specific deletion of CPT1a, will be an exciting avenue of research to explore for a subsequent study.

Reviewer #2 (Public Review):

The study by Drougard et al. aimed to answer a critical question on how high-fat diets trigger metabolic issues like obesity and diabetes. Their study revealed that an acute response by microglial cells in the brain to high-fat intake surprisingly benefits metabolism and cognitive function by rapidly metabolizing harmful fatty acids into alternative energy substrates like lactate and itaconate. Thus, short-term HFD intake seems to prompt a distinct beneficial response, suggesting a need for further exploration into the transition from acute to chronic effects.

Reviewer #3 (Public Review):

Drougard et al. explore microglial detection of a switch to high-fat diet and a subsequent metabolic response that benefits memory. The findings are both surprising and novel in the context of acute high-fat intake, with convincing evidence of increased CSF palmitate after 3 days of HFD. While the authors demonstrate compelling signs of microglial activation in multiple brain regions and unique metabolite release in tracing studies, they should address the following areas.

Major Points:

(1) It appears that the authors perform key metabolic assays in vitro/ex vivo using primary microglia from either neonatal or adult mice, which should be more clearly delineated especially for the 13C-palmitate tracing. In the case of experiments using primary microglia derived from mixed glial cultures stimulated with M-CSF, this system relies on neonatal mice. This is understandable given the greater potential yield from neonatal mice, but the metabolic state and energetic demands of neonatal and adult microglia differ as their functional roles change across the lifespan. The authors should either show that the metabolic pathways they implicate in neonatal microglia are also representative of adult microglia or perform additional experiments using microglia pooled from adult mice, especially because they link metabolites derived from neonatal microglia (presumably not under the effects of acute HFD) to improved performance in behavioral assays that utilize adult mice.

(2) The authors demonstrate that 3 days of HFD increases circulating palmitate by CSF metabolomics and that microglia can readily metabolize palmitate, but the causal link between palmitate metabolism specifically by microglia and improved performance in behavioral paradigms remains unclear. A previous body of research, alluded to by the authors, suggests that astrocyte shuttling of lactate to neurons improves long-term and spatial memory. The authors should account for palmitate that also could be derived from astrocyte secretion into CSF, and the relative contribution compared to microglia-derived palmitate. Specifically, although microglia can metabolize the palmitate in circulation, there is no direct evidence that the palmitate from the HFD is directly shuttled to microglia and not, for example, to astrocytes (which also express CX3CR1). Thus, the Barnes Maze results could be attributed to multiple cell types. Furthermore, the evidence provided in Figure 5J is insufficient to claim a microglia-dependent mechanism without showing data from mice on HFD with and without microglia depletion (analogous to the third and fourth bars in panel K).

(3) Given the emphasis on improved cognitive function, there is minimal discussion of the actual behavioral outcomes in both the results and discussion sections. The data that HFD-treated animals outperform controls should be presented in more detail both in the figure and in the text. For example, data from all days/trials of the Barnes Maze should be shown, including the day(s) HFD mice outperform controls. Furthermore, the authors should either cite additional literature or provide experimental evidence supporting the notion that microglia release of TCA-associated substrates into the extracellular milieu after HFD specifically benefits neuronal function cellularly or regionally in the brain, which could translate to improved performance in classical behavioral paradigms. The single reference included is a bit obscure, given the study found that increased lactate enhances fear memory which is a neural circuit not studied in the current manuscript. Are there no additional studies on more relevant metabolites (e.g., itaconate, succinate)?

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

In this study, Drougard et al. examined the consequences of an acute high fat diet (HFD) on microglia in mice. 3-day HFD influenced the regulation of systemic glucose homeostasis in a microglia-dependent and independent manner, as determined using microglial depletion with PLX5622. 3-day HFD increased microglial membrane potential and the levels of palmitate and stearate in cerebrospinal fluid in vivo. Using confocal imaging, respirometry and stable isotope-assisted tracing in primary microglial cultures, the authors suggest an increase in mitochondrial fission and metabolic remodeling occurs when exposed to palmitate, which increases the release of glutamate, succinate and itaconate that may alter neuronal metabolism. This acute microglial metabolic response following acute HFD is subsequently linked to improved higher cognitive function (learning and memory) in a microglia and DRP1-dependent manner.

Strengths:

Overall, this study is interesting and novel in linking acute high fat diet to changes in microglia and improved learning and memory in mice. The role for microglia and DRP1 in regulating glucose homeostasis and memory in vivo appears to be supported by the data.

Weaknesses:

The authors suggest that utilization of palmitate by microglia following HFD is the driver of the acute metabolic changes and that the release of microglial-derived lactate, succinate, glutamate and itaconate are causally linked to improvements in learning and memory. A major weakness is that the authors provide no mechanistic link between beta-oxidation of palmitate (or other fatty acids) in microglia and the observed systemic metabolic and memory phenotypes in vivo. Pharmacological inhibition of CPT1a could be considered or CPT1a-deficient microglia.

We thank Reviewer #1 for their time, effort and the critique. Indeed, we suggest that palmitate drives the aMMR response and associated improvements in learning and memory. In response to acute HFD we observe 1) increased in palmitate in CSF; 2) impaired mitochondrial ETC activity in primary microglia (within 12 hours of HFD); and 3) improved learning and memory. The greatest barrier to proving how acute palmitate uptake in microglia improves learning and memory in vivo is the protracted methodology required for microglial isolation and purification. The timeframes and relatively harsh digestion protocols required are currently incompatible with metabolomic tracing and well beyond those required for most cell-types used for metabolomic investigation.  We have tested and failed to obtain reproducible data across numerous in vivo protocols and finally settled on in vitro 13C palmitate treated neonatal microglia as the best current option. Primary neonatal microglia are accepted as one of the current best culture models by the microglial community (Valdercaos cell report 2014, Kim Cell Metab 2019). Using neonatal microglia we demonstrate that 13Cpalmitate label is processed to palmitoylcarnitine (Fig 4C) and acetylcarnitine (Fig 4D) indicating that microglial fatty acid metabolism acts via the canonical CPT1/CPT2 pathway. These experiments highlight that microglia process palmitate via beta oxidation generating acetyl coA and engaging the TCA cycle (Fig 4G-I).

We now acknowledge these technical limitations more clearly and highlight their impact on any conclusions regarding adult microglia in vivo:

Results “Microglia take up and metabolize free fatty acids”; 

“Due in part to the long isolation times required to generate pure primary adult microglia, metabolite tracing experiments on primary adult microglia are not currently feasible. We therefore chose primary murine neonatal microglia as our model of choice for more mechanistic experiments (Valdercaos, Cell Report 2014)”

And,

Discussion:

“We propose that aMMR could result from direct uptake, processing, and release of fatty acid derived carbons, and demonstrate that microglia are capable of metabolizing fatty acids towards diverse intracellular and extracellular pools.”

While acute ICV injection a CPT1a blocker would be of potential interest, the caveats associated with CPT1a inhibition in other cell-types (neurons, astrocytes, etc) and with targeting the appropriate brain region (currently unknown) currently preclude the effective use of this approach for to generate clear additional mechanistic insights. Similarly, given the time and resources required to generate, validate, optimize and experiment on a clean model of in vivo adult microglia-specific CPT1a knockout, this approach was deemed beyond the scope of this study. That said, the critique is important, and it should comprise a follow-up project.

Comment: Another major weakness is that the authors also suggest that 3-day HFD microglial response (increase membrane potential) is likely driven by palmitate-induced increases in itaconate feedforward inhibition of complex II/SDH. Whilst this is an interesting hypothesis, the in vitro metabolic characterization is not entirely convincing.

The reviewer is correct, we suggest that our data is consistent with a model where a palmitate-induced increase in itaconate inhibits complex II/SDH. While our findings do not comprise mechanistic proof, the hypothesis is supported by our Seahorse studies (Fig 2E) highlighting that a combined Palmitate + Succinate stimulation does not increase OCR beyond that of Palmitate alone; by primary microglial cell experiments highlighting that 3d-HFD treated adult primary microglia are refractory to succinate-induced mitochondrial membrane depolarization (Fig 2F); and by the identification of increased palmitate induced itaconate production/release in cultured primary neonatal microglia (Fig 4H). The data are consistent with an inhibition of complex II/ SDH and with increased itaconate secretion. They are also consistent with literature on more easily accessible myeloid lineages (Lampropoulou V, Cell Metab 2016).  

Comment: The authors suggest that acute palmitate appears to rapidly compromise or saturate complex II activity. Succinate is a membrane impermeable dicarboxylate. It can enter cells via MCT transporters at acidic pH. It is not clear that I) Succinate is taken up into microglia, II) If the succinate used was pH neutral sodium succinate or succinic acid, and III) If the observed changes are due to succinate oxidation, changes in pH or activation of the succinate receptor SUCNR1 on microglia. In the absence of these succinate treatments, there are no alterations in mitochondrial respiration or membrane potential following palmitate treatment, which does not support this hypothesis.

We thank Reviewer #1 for highlighting a lack of information in the material and methods. We have updated them accordingly as follows:

“For the electron transport chain experiments (ETC), the experiment was based on the Salabei et al. The cell suspension was incubated with the mitochondrial probe Tetramethylrhodamine TMRM (10mM; Abcam, Cat# ab228569) and fluorescent glucose analog 2-NBDG (Abcam, Cat# 235976) for 30min at 37degrees before FACS acquisition. For challenging the ETC, the cell pellet was resuspended in 500ul of warm MAS buffer solution + 1nM Plasma Membrane Permeabilizer (Agilent Seahorse XF PMP) in order to permeabilize the cells. Microglial cells were gated from CD45low-CD11b+ cells followed by singlet after forward and side scatter pattern. They were incubated each 90 seconds by the following drugs: 0,5ul of 100uM Rotenone (Sigma), 2ul of 2.5M Succinate adjusted to ph 7.4 with NaOH (succinic acid, Sigma) and 0.5ul of 1mM Antimycin (Sigma). Cytometry was performed on Fortessa (BD Bioscience) and analyzed with FlowJo v10 (Treestar).”

Following the updated protocol, we hope we highlighted that the succinate (solution of succinic acid ph 7.4) is reaching directly the ETC since the microglial cells have been permeabilized by the Plasma Membrane Permeabilizer (Agilent Seahorse XF PMP).

Comment: Intracellular itaconate measurements and quantification are lacking and IRG1 expression is not assessed. There also appears to be more labelled itaconate in neuronal cultures from control (BSA) microglia conditioned media, which is not discussed. What is the total level of itaconate in neurons from these conditioned media experiments? No evidence is provided that the in vivo response is dependent on IRG1, the mitochondrial enzyme responsible for itaconate synthesis, or itaconate. To causally link IRG1/itaconate, IRG1-deficient mice could be used in future work. 

We appreciate the interest, the exciting question, and the suggested future experiment. Indeed, our results suggest a difference in metabolite release between the BSA treated-microglia and palmitate treated-microglia and their impact on neurons comprises a prime question for future work. We have highlighted this in the discussion as well as adding a comment regarding relative levels of labelled itaconate as follows:

Results; Acute HFD induces widespread MMR and rapid modulation (…) memory  

“As a control for the direct uptake of 13C-glucose, we treated parallel neuronal cultures with the same fresh 13C-glucose tracing media originally added to the microglia. Intriguingly, and consistent with literature documenting poor direct glucose utilization by neurons [29], we found substantial m+3 lactate (as well as other metabolites) in neurons treated with microglial conditioned media, and at levels that far exceeded labelling triggered by glucose tracer alone (Fig 5A, middle column vs left column)(Suppl Fig S5B). The data indicate higher uptake of citrate and itaconate from the control microglia-conditioned media, further supporting the hypothesis that neuronal metabolism is reproducibly impacted by palmitate-triggered changes in microglial products. These data demonstrate that palmitate metabolism by microglia modulates neuronal carbon substrate use in vitro, and, they highlight the relative importance of this process compared to uptake of pure glucose. The data identify a candidate mechanism by which aMMR may alter neuronal function in vivo.”

Comment: While microglial DRP1 is causally implicated the role of palmitate is not convincing. Mitochondrial morphology changes are subtle including TOMM20 and DRP1 staining and co-localization - additional supporting data should be provided. Electron microscopy of mitochondrial structure would provide more detailed insight to morphology changes. Western blot of fission-associated proteins Drp1, phospho-Drp1 (S616), MFF and MiD49/51. Higher magnification and quality confocal imaging of DRP1/TOMM20. Drp1 recruitment to mitochondrial membranes can be assessed using subcellular fractionation.

We appreciate the reviewer’s comment. Previous work by others, already cited elsewhere in our manuscript

(PMCID: PMC7251564), has clearly demonstrated increased mitochondrial fragmentation and

phosphorylated DRP1 in 3d HFD animals. This very specific result can therefore be considered confirmatory / validating of existing literature, and important for inclusion of DRP1 in our overall model. We have made sure to better highlight this important literature accordingly:

Results; A rapid Microglial Mitochondria response to high fat diet

“Consistent with the in vivo observations above, in vitro palmitate exposure decreased microglial mitochondrial length within 24 hours, indicating that fatty acid exposure itself is sufficient to trigger mitochondrial fission in a cell autonomous manner (Fig 2G upper panels). This result also confirms observations by Kim et al. who observed mitochondrial fission and DRP1 phosphorylation upon 3d-HFD treated mice [Kim JD et al, Microglial UCP2 mediates Inflammation and Obesity induced by High Fat feeding, Cell Metab 2019].”

Comment: No characterization of primary microglia from DRP1-knockout mice is performed with palmitate treatment. Authors demonstrate an increase in both stearate and palmitate in CSF following 3day HFD. Only palmitate was tested in the regulation of microglial responses, but it may be more informative to test stearate and palmitate combined.

Testing stearate and palmitate combined is an interesting experiment for mimicking the global effect of HFD which is highly enriched with these two satured fatty acids, and then, more informative. In vitro stimulation of microglia model cells has been previously published by Valdearcos and al. (Cell Reports 2014) who studied the effect of a mix of stearate and palmitate on the Mediobasal Hypothalamus inflammation. Here, we build on their important findings by demonstrating that these 2 compounds are actually found in the CSF of 3d-HFD mice. Studies from other labs have also shown the presence of stearate and palmitate in the CSF of chronically obese and diabetic patients which highlights the importance of these findings (Melo HM et al. cell report 2020). While a systematic dissection of the roles of HFD-regulated CSF metabolites (including direct (diet containing) and indirect (secondary) is beyond the scope of this study, this point is important, not least because it highlights less well-studied metabolites and the potential of possible combinatorial interactions. We have highlighted this idea in the results as follows:

Results; A rapid Microglial Mitochondria response to high fat diet

“To test whether these observed fatty acid changes in the CSF might directly trigger aMMR, we switched to an in vitro primary neonatal microglia model and examined the effects of the more abundant of these, palmitate (Fig S2A-B).”

and, in the discussion as follows:

“Studies have identified stearate and palmitate in the CSF of patients with chronic obesity and with diabetes, reports that highlight the importance of these findings (Melo HM et al. cell report 2020). While a systematic dissection of the roles of HFD-regulated CSF metabolites (including direct (diet containing) and indirect (secondary)) is beyond the scope of this study, they represent priority areas for future investigation, particularly given the wide-range of fatty-acid specific biological effects in the literature, and the potential for combinatorial interactions.” 

Reviewer #1 (Recommendations For The Authors):

Congratulations on this interesting and novel work. Please see public review for details on potential experiments. While I would not expect all the experiments to be performed for this current study, it’s important to not overstate what the data is showing. For example, there is no causal link between palmitate oxidation in microglia or released metabolites (itaconate etc) from microglia in the effect on systemic glucose metabolism or memory. To make such claims more supporting data would be required.

We thank Reviewer #1 for their highly constructive critique_._

Reviewer #2 (Public Review):

The study "A rapid microglial metabolic response controls metabolism and improves memory" by Drougard et al. provides evidence that short-term HFD has a beneficial effect on spatial and learning memory through microglial metabolic reprogramming. The manuscript is well-written and the statistics were properly performed with all the data. However, there are concerns regarding the interpretation of the data, particularly the gap between the in vivo observations and the in vitro mechanistic studies.

In the PLX-5622 microglial depletion study, it is unclear what happened to the body weight, food intake, and day-night behavior of these mice compared to the vehicle control mice. It is important to address the innate immunity-dependent physiology affected by a long period of microglial depletion in the brain (also macrophages in the periphery). Furthermore, it would be beneficial to validate the images presented in Fig.1F by providing iba1 staining in chow diet-fed mice with or without PLX-5622 for 7-10 days. Additionally, high-quality images, with equal DAPI staining and comparable anatomical level, should be provided in both chow diet-fed mice and HFD-fed mice with or without PLX-5622 in the same region of hypothalamus or hippocampus. These are critical evidences for this project, and it is suggested that the authors provide more data on the general physiology of these mice, at least regarding body weight and food intake.

We are grateful to Reviewer #2 for their constructive comments and for their time and effort; and for highlighting the lack of experimental details regarding the PLX-5622 microglial depletion study. We followed the protocol established in Feng et al JCI 2017. No adverse effects on body weight, food intake and day-night behavior have been described in this study as well as in other studies for longer treatment (Sonia George et al Molecular Neurodegeneration 2019). We didn’t observe any differences in body weight and the food intake within or between groups, upon PLX administration. These data have been included as new Supplementary Fig 6 A-B.

The material and method was updated as follows:

“Animals were administered PLX5622-containing diet for 7-9 days without observable impact on the body weight or food intake (Fig S6A-B), using protocols adopted from [Feng et al JCi 2017, Sonia George et al Molecular Neurodegeneration 2019].”

Comment: It is also unclear whether the microglia shown in Fig.3A were isolated from mice 4 weeks after Tamoxifen injection. It is suggested that the authors provide more evidence, such as additional images or primary microglia culture, to demonstrate that the mitochondria had more fusion upon drp1 KO. It is recommended to use mito-tracker green/red to stain live microglia and provide good resolution images.

We thank Reviewer #2 for pointing out the lack of detailed information about Fig 3A. Microglial cells were indeed isolated from mice after the tamoxifen injection for highlighting the deletion. We updated the Material and methods with the text below;

“For the colocalization experiment, microglia were isolated from 10 to 12-week old drp1ko mice and their littermate controls, immediately fixed in PFA and stained with DRP1 (diluted 1:50 Cell signaling; Cat#8570) and tomm20 antibodies (diluted 1:1000, SantaCruz; Cat#sc177615).”

This experiment was performed as an additional control of the drp1 deletion from our knockout-mice. For this experiment we used Tomm20 since the microglia cells weren’t live after the addition of PFA. 

Comment: Regarding the data presented in Fig.5A, it is suggested that the authors profile the metabolomics of the microglial conditioned media (and provide the methods on how this conditioned media was collected) to determine whether there was already abundant lactate in the media. Any glucose-derived metabolites, e.g. lactate, are probably more preferred by neurons as energy substrates than glucose, especially in embryonic neurons (which are ready to use lactate in newborn brain).

With regards to Fig 5A, metabolomics of microglia conditioned media are provided as Fig 5A, Supp Figure 5Band we provided a supplementary table 2.

We thank Reviewer #2 for noting the lapse of technical detail. We updated the Material and methods with the following:

“For conditioned media experiments, microglial cells were incubated with DMEM (Gibco) without lactate completed with BSA-conjugated palmitate or Control BSA. Conditioned media was collected after the incubation, centrifuged 15min at 300g (4oC) and the supernatant transferred and frozen in a fresh tube avoiding the cells and debris pellet. Sample were immediately snap frozen or use for the neurons incubation.”

Any glucose-derived metabolites, e.g. lactate, are more preferred by neurons as energy substrates than glucose as described first in the literature by Prof. Pellerin and Prof. Magistretti via the astrocyte-neuron cooperation (PNAS 1994). Since their discovery, lactate has been explored and is well known as a key signaling molecule (Magistretti PJ Nat Rev Neurosciences 2018). We explored the role of lactate released from the microglia, and we demonstrated that it is taken up by neurons independently of any microglial pretreatment. This experiment highlights microglia as another lactate provider for the neurons (Fig 4N and Fig 5A). 

Comment: Finally, it is important to address whether PLX-5622 affects learning and spatial memory in chow diet-fed animals. Following the findings shown in Fig 5J and 5K, the authors should confirm these by any morphological studies on synapse, e.g. by synaptophysin staining or ultrastructure EM study in the area shown in Fig 5I.

We appreciate the comment and question. We performed the controls and included them now as Fig 5J and Fig S5 E-F-G. We do not observe any adverse effects of PLX5622 on learning and spatial memory in normal chow-fed animals. 

While we were unable to study the synapses as requested, it is important to note that no changes are expected given publications from other labs using the same protocol (Feng x JCI 2017 ,Spangenberg E Nat Com 2019), or longer PLX5622 treatment (Niiyama T eNeuro 2023, Witcher KG J neurosciences 2021), all four of which did not find morphological differences at synapses. 

Reviewer #2 (Recommendations For The Authors):

The authors should provide more evidence that palmitate is derived from HFD to prove that it mediates the HFD effects on the microglial mitochondria response. This could be done by adding 13C-palmitate into the HFD and performing metabolomics in isolated microglia from control mice (and Drp1-MG-KO mice, if possible).

We thank the Reviewer #2 for the enthusiastic revision. Unfortunately, we were unable to attempt this final suggested experiment. We have adjusted our wording accordingly and appreciate the reviewer’s understanding.

Reviewer #3 (Public Review):

Drougard et al. explore microglial detection of a switch to high-fat diet and a subsequent metabolic response that benefits memory. The findings are both surprising and novel in the context of acute highfat intake, with convincing evidence of increased CSF palmitate after 3 days of HFD. While the authors demonstrate compelling signs of microglial activation in multiple brain regions and unique metabolite release in tracing studies, they should address the following areas prior to acceptance of this manuscript.

Major Points:

(1) It appears that the authors perform key metabolic assays in vitro/ex vivo using primary microglia from either neonatal or adult mice, which should be more clearly delineated especially for the 13C-palmitate tracing. In the case of experiments using primary microglia derived from mixed glial cultures stimulated with M-CSF, this system relies on neonatal mice. This is understandable given the greater potential yield from neonatal mice, but the metabolic state and energetic demands of neonatal and adult microglia differ as their functional roles change across the lifespan. The authors should either show that the metabolic pathways they implicate in neonatal microglia are also representative of adult microglia or perform additional experiments using microglia pooled from adult mice, especially because they link metabolites derived from neonatal microglia (presumably not under the effects of acute HFD) to improved performance in behavioral assays that utilize adult mice.

We thank Reviewer #3 for their constructive critique and encouraging words. As indicated, the 13C-palmitate experiments were performed with primary microglia derived from mixed glial cultures stimulated with M-CSF and we demonstrated our primary cultures were almost pure by the supplementary experiments (supp Fig2A and B). Additional minor details in these contexts have been added to the Material and Methods.

The experiments focusing on the mitochondrial ETC were performed on sorted microglia from adult mice and parallels demonstrated with the neonatal cultures (the primary model for metabolic tracing). Compromised complex II activity under conditions of acute HFD/palmitate stimulation for instance were shown in both systems. Unfortunately, despite best-efforts, attempts to run 13C-palmitate tracing experiments on primary adult microglia failed, attributable in large part to the long (~4 hour) and harsh microglial extraction and sorting process. These experiments will require substantial follow-up efforts including the establishment and validation ideally of an adult microglia-neuron co-culture model that faithfully recapitulates most aspects of in vivo metabolic cross-talk. This noble aim is beyond the scope of this study. We have made sure to temper the  conclusions made in the manuscript and to not overstate the impact and interpretation of the in vitro work including updating the following sentences.

Results “Microglia take up and metabolize free fatty acids”; 

“Due in part to the long isolation times required to generate pure primary adult microglia, metabolite tracing experiments on primary adult microglia are not currently feasible. We therefore chose primary murine neonatal microglia as our model of choice for more mechanistic experiments (Valdercaos cell Report 2014)”

and Discussion:

“We propose that aMMR could result from direct uptake, processing, and release of fatty acid derived carbons, and demonstrate that microglia are capable of metabolizing fatty acids towards diverse intracellular and extracellular pools.”

Comment: The authors demonstrate that 3 days of HFD increases circulating palmitate by CSF metabolomics and that microglia can readily metabolize palmitate, but the causal link between palmitate metabolism specifically by microglia and improved performance in behavioral paradigms remains unclear. A previous body of research, alluded to by the authors, suggests that astrocyte shuttling of lactate to neurons improves long-term and spatial memory. The authors should account for palmitate that also could be derived from astrocyte secretion into CSF, and the relative contribution compared to microglia-derived palmitate. Specifically, although microglia can metabolize the palmitate in circulation, there is no direct evidence that the palmitate from the HFD is directly shuttled to microglia and not, for example, to astrocytes (which also express CX3CR1). 

We appreciate the comment. Indeed, this issue highlights one of the greatest challenges for efforts aimed at tracing (beyond doubt) that a single minor cell population contributes towards metabolic cross-talk in vivo. Our experiments show: increased CSF palmitate levels within one feeding cycle of HFD; rapidly induced microglial metabolic activation (characterized by increased mitochondrial membrane potential and impaired complex II activity); and that microglia mount a comparable mitochondrial activation profile in vitro when exposed to palmitate. They show in vitro using neonatal microglia that microglia take up and metabolize palmitate; that they release metabolites with neuro-modulatory potential; that neurons take these metabolites up and modulate their function differentially when exposed to control vs palmitate-treated microglia-conditioned media (in the absence of astrocytes). The experiments show through acute PLX-induced elimination of microglia, however crude, that this compartment impacts the acute HFD response, and using conditional deletion, that full DRP1 expression is required CX3CR1-CreERT2 targeted cells (primarily microglia deleting; Zhao et al 2019).  While these experiments cannot rule out a contribution of astrocytes to the observations in vivo, comparable experiments rarely can and we cannot rationalize why microglia should not have equal access to CSF palmitate for uptake or to neurons for substrate provisioning. We now better highlight this important issue, and temper our conclusions accordingly:

“Tanycytes and astrocytes have both been documented to release select metabolites into the extracellular environment [33, 34]. While suggestive, the experiments highlighted here do not rule out a contribution of these or cell types in coupling acute HFD intake to memory and learning.”

Comment: Thus, the Barnes Maze results could be attributed to multiple cell types. Furthermore, the evidence provided in Figure 5J is insufficient to claim a microglia-dependent mechanism without showing data from mice on HFD with and without microglia depletion (analogous to the third and fourth bars in panel K).

Agreed. We appreciate the comment. We have now added the requested HFD condition to Figure 5J. The data support our previous interpretation of the data. 

Comment: Given the emphasis on improved cognitive function, there is minimal discussion of the actual behavioral outcomes in both the results and discussion sections. The data that HFD-treated animals outperform controls should be presented in more detail both in the figure and in the text. For example, data from all days/trials of the Barnes Maze should be shown, including the day(s) HFD mice outperform controls. Furthermore, the authors should either cite additional literature or provide experimental evidence supporting the notion that microglia release of TCA-associated substrates into the extracellular milieu after HFD specifically benefits neuronal function cellularly or regionally in the brain, which could translate to improved performance in classical behavioral paradigms. The single reference included is a bit obscure, given the study found that increased lactate enhances fear memory which is a neural circuit not studied in the current manuscript. Are there no additional studies on more relevant metabolites (e.g., itaconate, succinate)?

We agree. We have now re-plotted the behavioral test to better highlight that the HFD-treated animals outperform controls, as requested (Fig S7 and S8). We also added the requested literature. While we cannot be sure our search captured all relevant studies, we find a relative paucity of studies that characterize CSF metabolite changes in the context of acute high fat feeding or that demonstrate the ability of CSF substrates to convincingly improve memory and learning in vivo at physiological levels. Indeed, while simple, we feel the findings are of substantial novelty and highlight an area for significant future research. We have tempered our conclusions throughout and added to the discussion as follows:

“Such substrate release could mediate the learning and memory effects that accompany aMMR; they are consistent with the data of other studies that have examined metabolite associations with learning and memory (itaconate [Morgunov IG, microorganisms 2020; Xiong J, Neuromolecular med 2023], succinate [Serra FT neurosciences letter 2022; Cline BH, BMC neurosciences 2012].”

Minor Points:

(1) In Figure 5J the latency to find the hole was noticeably higher (mean around 150s) than the latency in panel K (mean around 100s for controls, and 60s for Drp1MGWT on HFD). This suggests high variability between experiments using this modified version of the Barnes Maze, despite the authors assertion that a standard Barnes Maze was employed and the results were reproducible at multiple institutions. Why do Drp1MGWT mice on control diet find the escape hole significantly faster than WT mice on control diet in panel J? Given the emphasis on cognitive improvement following acute HFD as a novel finding, the authors should explain this discrepancy.

We appreciate this question and comment. Indeed, as the reviewer knows, behavioral tests including the Barnes test show variation with genetic background, and with environment and context (eg. age, caging density, litter size, behavioral state and more (Inglis A, Physiol Behavior 2019; Loos M Mamm Genome 2015; and unpublished observations). We do not know the exact origin of the difference mentioned above but our best guess would be that it stems from either environmental differences  that are ever present in vivaria (seasonal, mouse house room, cage-changing cycles, etc) and/or, differences between the background genetics (eg. presence of Cre transgene and linked genome, genetic drift) or precise experimental differences between the cohorts (eg. repeated tamoxifen-injection paradigm for the deletion group). All of our experiments were performed in parallel, with all relevant animal groups equally represented in every run, and,and used age- and sex-matched individuals from congenic strains. Wherever possible, controls and test animals were littermates to minimize within strain variance attributable to litter effects (litter size, maternal and paternal effects). Given our lab’s interest and focus on the mechanistic and developmental origins of variance heterogeneity, these differences are of high interest for future study. 

Comment: The authors highlight in the graphical abstract and again in Figure 4A the formation of lipid droplets following palmitate exposure as evidence of that microglia can process fatty acids. They later suggest that a lack of substantial induction of lipid droplet accumulation suggests that microglia are metabolically wired to release carbon substrates to neighboring cells. Clarification as to the role of lipid droplet formation/accumulation in explaining the results would eliminate any possible confusion.

We modified the wording in the manuscript accordingly:

Results “Microglia take up and metabolize free fatty acids”;

“Based on BODIPY fluorescence, we found that primary microglia increase lipid droplet numbers within 24h of in vitro exposure to palmitate (200uM; Fig 4A), demonstrating a capacity to take up fatty acids.”

Comment: In many bar graphs showing relatively modest effects, it would be helpful to use symbols to also show the distribution of sample and animal replicates (especially behavioral paradigms).

Agreed. Indeed, the results are both modest and impressive given the nature of the intervention (simple change in dietary macronutrient composition). We have now re-plotted the results from the behavioral experiments, accordingly (Fig S7 and Fig S8).

Reviewer #3 (Recommendations For The Authors):

This is a good manuscript deserving of publication assuming some of the concerns posed above are addressed.

We thank Reviewer #3 again for their time, effort, and dedication, and for their objective review of the manuscript.

Author response:

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

All of the reviewers indicate that their major concerns have been adequately addressed, but they each have a few comments that the authors should consider before submitting a final version (without further review) for publication. For example, a statement about the sex of the mice used in the studies and whether any differences were noted if both sexes were used. The idea that the loss of glutamate transport might affect NA loading into vesicles is also worth considering. Finally, the authors might want to mention that the role of neuropeptide release from NA neurons needs further examination. 

As noted in the prior submitted revision, all experiments contained both males and females and this was addressed in our re-submission. In our analysis of breathing and metabolism, sex was included in the analysis and no significant phenotypic difference was observed (The statement of no sex difference is in line 451-456). For the fate map and in situ experiments, although the group size is small, we did not see obvious differences in the expression patterns in the three glutamate transporters between females and males (line 347-350). All the anatomical and phenotypic data in this manuscript are presented as combined graphs (figure 1, figure 1 supplement 1, figure 2, figure 2 supplement 2, figure 4,5,6,7) and we had differentially labeled our data points by sex (female data is pink and male data is blue).

The possibility that loss of Vglut2 might affect NA release has been added in the discussion (line 485-491) of the current revision. Dopamine Beta Hydroxylase (DBH) converts dopamine to noradrenaline in the vesicles, thus, glutamate may not directly affect noradrenaline loading into vesicles. However, since loss of Vglut2 reduced dopamine release in subsets of dopaminergic neurons, it remains possible that glutamate affects dopamine loading in NA neurons and in turn perturbs DA to NA conversion in the vesicle by DBH and subsequent noradrenaline release. Future work could examine this hypothesis using fast-scan cyclic voltammetry (FSCV) or microdialysis.

The further examination of the role of neuropeptide release from NA neurons is mentioned in the discussion (line 491-494 and line 497-499 of the pre).

eLife assessment

Chang et al. provide glutamate co-expression profiles in the central noradrenergic system and test the requirement of Vglut2-based glutamatergic release in respiratory and metabolic activity under physiologically relevant gas challenges. Their experiments provide compelling evidence that conditional deletion of vesicular glutamate transporters from noradrenergic neurons does not impact steady-state breathing or metabolic activity in room air, hypercapnia, or hypoxia. This study provides an important contribution to our understanding of how noradrenergic neurons regulate respiratory homeostasis in conscious adult mice. 

Public Reviews:

Reviewer #1 (Public Review):

Summary:

Chang et al. provide glutamate co-expression profiles in the central noradrenergic system and test the requirement of Vglut2-based glutamatergic release in respiratory and metabolic activity under physiologically relevant gas challenges. Their experiments show that conditional deletion of Vglut2 in NA neurons does not impact steady-state breathing or metabolic activity in room air, hypercapnia, or hypoxia. Their observations challenge the importance of glutamatergic signaling from Vglut2 expressing NA neurons in normal respiratory homeostasis in conscious adult mice. 

Strengths:

The comprehensive Vglut1, Vglut2, and Vglut3 co-expression profiles in the central noradrenergic system and the combined measurements of breathing and oxygen consumption are two major strengths of this study. Observations from these experiments provide previously undescribed insights into (1) expression patterns for subtypes of the vesicular glutamate transporter protein in the noradrenergic system and (2) the dispensable nature of Vglut2dependent glutamate signaling from noradrenergic neurons to breathing responses to physiologically relevant gas challenges in adult conscious mice. 

Weaknesses:

Although the cellular expression profiles for the vesicular glutamate transporters are provided, the study does not document that glutamatergic-based signaling originating from noradrenergic neurons is evident at the cellular level under normal, hypoxic, and/or hypercapnic conditions. The authors effectively recognize this issue and appropriately discuss their findings in this context. 

We thank the reviewer for the positive evaluation of our work.

Reviewer #2 (Public Review):

The authors characterized the recombinase-based cumulative fate maps for vesicular glutamate transporters (Vglut1, Vglut2 and Vglut3) expression and compared those maps to their realtime expression profiles in central NA neurons by RNA in situ hybridization in adult mice. Authors have revealed a new and intriguing expression pattern for Vglut2, along with an entirely uncharted co-expression domain for Vglut3 within central noradrenergic neurons. Interestingly, and in contrast to previous studies, the authors demonstrated that glutamatergic signaling in central noradrenergic neurons does not exert any influence on breathing and metabolic control either under normoxic/normocapnic conditions or after chemoreflex stimulation. Also, they showed for the first-time the Vglut3-expressing NA population in C2/A2 nuclei. In addition, they were also able to demonstrate Vglut2 expression in anterior NA populations, such as LC neurons, by using more refined techniques, unlike previous studies. 

A major strength of the study is the use of a set of techniques to investigate the participation of NA-based glutamatergic signaling in breathing and metabolic control. The authors provided a full characterization of the recombinase-based cumulative fate maps for Vglut transporters. They performed real-time mRNA expression of Vglut transporters in central NA neurons of adult mice. Further, they evaluated the effect of knocking down Vglut2 expression in NA neurons using a DBH-Cre; Vglut2cKO mice on breathing and control in unanesthetized mice. Finally, they injected the AAV virus containing Cre-dependent Td tomato into LC of v-Glut2 Cre mice to verify the VGlut2 expression in LC-NA neurons. A very positive aspect of the article is that the authors combined ventilation with metabolic measurements. This integration holds

particular significance, especially when delving into the exploration of respiratory chemosensitivity. Furthermore, the sample size of the experiments is excellent.  Despite the clear strengths of the paper, some weaknesses exist. It is not clear in the manuscript if the experiments were performed in males and females and if the data were combined. I believe that the study would have benefited from a more comprehensive analysis exploring the sex specific differences. The reason I think this is particularly relevant is the developmental disorders mentioned by the authors, such as SIDS and Rett syndrome, which could potentially arise from disruptions in central noradrenergic (NA) function, exhibit varying degrees of sex predominance. Moreover, some of the noradrenergic cell groups are sexually dimorphic. For instance, female Wistar rats exhibit a larger LC size and more LC-NA neurons than male subjects (Pinos et al., 2001; Garcia-Falgueras et al., 2005). More recently, a detailed transcriptional profiling investigation has unveiled the identities of over 3,000 genes in the LC. This revelation has highlighted significant sexual dimorphisms, with more than 100 genes exhibiting differential expression within LC-NA neurons at the transcript level. Furthermore, this investigation has convincingly showcased that these distinct gene expression patterns have the capacity to elicit disparate behavioral responses between sexes (Mulvey et al., 2018).

Therefore, the authors should compare the fate maps, Vglut transporters in males and females, at least considering LC-NA neurons. Even in the absence of identified sex differences, this information retains significant importance. 

An important point well raised by the authors is that although suggestive, these experiments do not definitively rule out that NA-Vglut2 based glutamatergic signaling has a role in breathing control. Subsequent experiments will be necessary to validate this hypothesis. 

An improvement could be made in terms of measuring body temperature. Opting for implanted sensors over rectal probes would circumvent the need to open the chamber, thereby preventing alterations in gas composition during respiratory measurements. Further, what happens to body temperature phenotype in these animals under different gas exposures? These data should be included in the Tables. 

Is it plausible that another neurotransmitter within NA neurons might be released in higher amounts in DBH-Cre; Vglut2 cKO mice to compensate for the deficiency in glutamate and prevent changes in ventilation? 

Continuing along the same line of inquiry is there a possibility that Vglut2 cKO from NA neurons not only eliminates glutamate release but also reduces NA release? A similar mechanism was previously found in VGLUT2 cKO from DA neurons in previous studies (Alsio et al., 2011; Fortin et al., 2012; Hnasko et al., 2010). Additionally, does glutamate play a role in the vesicular loading of NA? Therefore, could the lack of effect on breathing be explained by the lack of noradrenaline and not glutamate? 

We thank the reviewer for the positive evaluation and further suggestions. Please see our response in “Author Response” to the previous version of Reviewer #2 (Public review).

Reviewer #4 (Public Review): 

Summary:

Although previous research suggested that noradrenergic glutamatergic signaling could influence respiratory control, the work performed by Chang and colleagues reveals that excitatory (specifically Vglut2) neurons is dynamically and widely expressed throughout the central noradrenergic system, but it is not significantly crucial to change baseline breathing as well the hypercapnia and hypoxia ventilatory responses. The central point that will make a significant change in the field is how NA-glutamate transmission may influence breathing control and the dysfunction of NA neurons in respiratory disorders. 

Strengths:

There are several strengths such as the comprehensive analysis of Vglut1, Vglut2, and Vglut3 expression in the central noradrenergic system and the combined measurements of breathing parameters in conscious unrestrained mice. 

Other considerations :

These results strongly suggest that glutamate may not be necessary for modulating breathing under normal conditions or even when faced with high levels of carbon dioxide (hypercapnia) or low oxygen levels (hypoxia). This finding is unexpected, considering many studies have underscored glutamate's vital role in respiratory regulation, more so than catecholamines. This leads us to question the significance of catecholamines in controlling respiration. Moreover, if glutamate is not essential for this function, we need to explore its role in other physiological processes such as sympathetic nerve activity (SNA), thermoregulation, and sensory physiology. 

We thank the reviewer for the positive evaluation and further suggestions. The potential role of noradrenergic-derived glutamate in other processes, which is beyond the scope of this study, should be addressed in the future.

Recommendations for the authors:

Reviewer #2 (Recommendations For The Authors):

All of my concerns were effectively resolved, leading me to accept the paper. However, I suggest that the authors consider investing in a more reliable system for measuring body temperature, as accurate measurements of this parameter are crucial for whole body plethysmography. 

Thank you for the suggestion. The real-time measurement of body temperature is a goal in future studies.

Reviewer #4 (Recommendations For The Authors):

Because I am revising a revised version, I believe the authors have addressed most, if not all, the concerns raised by already 3 reviewers. In my understanding the authors achieved their aims and the results are totally supported by the conclusions. The impact of this work on the respiratory field is significant and is likely to advance the field. The methods and data utilized, which combine standard techniques with genetic tools, will be highly beneficial to the research community. 

In my understanding I still have one concern that if glutamate is not critical, then what is? Could we potentially disable the noradrenergic (NA) system while preserving glutamate functionality to determine if the NA system is indeed crucial for respiratory physiology? This approach might provide clearer insights into the mechanisms underlying respiratory control. 

We agree that there remain several exciting questions about the respective roles of noradrenaline, glutamate, and other neuropeptides such as Neuropeptide Y (NPY) and galanin. We are currently devising strategies to address the respective and combinatorial roles for all these candidates in breathing control. Most simply, we can conditionally, mutagenized each of them in the central noradrenergic system in an acute manner using DBH-CreER mice to determine if any of them are critical to respiratory control with the advantage of minimizing developmental compensatory events.

eLife assessment

Chang et al. provide glutamate co-expression profiles in the central noradrenergic system and test the requirement of Vglut2-based glutamatergic release in respiratory and metabolic activity under physiologically relevant gas challenges. Their experiments provide compelling evidence that conditional deletion of vesicular glutamate transporters from noradrenergic neurons does not impact steady-state breathing or metabolic activity in room air, hypercapnia, or hypoxia. This study provides an important contribution to our understanding of how noradrenergic neurons regulate respiratory homeostasis in conscious adult mice.

Reviewer #1 (Public Review):

Summary:

Chang et al. provide glutamate co-expression profiles in the central noradrenergic system and test the requirement of Vglut2-based glutamatergic release in respiratory and metabolic activity under physiologically relevant gas challenges. Their experiments show that conditional deletion of Vglut2 in NA neurons does not impact steady-state breathing or metabolic activity in room air, hypercapnia, or hypoxia. Their observations challenge the importance of glutamatergic signaling from Vglut2 expressing NA neurons in normal respiratory homeostasis in conscious adult mice.

Strengths:

The comprehensive Vglut1, Vglut2, and Vglut3 co-expression profiles in the central noradrenergic system and the combined measurements of breathing and oxygen consumption are two major strengths of this study. Observations from these experiments provide previously undescribed insights into (1) expression patterns for subtypes of the vesicular glutamate transporter protein in the noradrenergic system and (2) the dispensable nature of Vglut2-dependent glutamate signaling from noradrenergic neurons to breathing responses to physiologically relevant gas challenges in adult conscious mice.

Weaknesses:

Although the cellular expression profiles for the vesicular glutamate transporters are provided, the study does not document that glutamatergic-based signaling originating from noradrenergic neurons is evident at the cellular level under normal, hypoxic, and/or hypercapnic conditions. The authors effectively recognize this issue and appropriately discuss their findings in this context.

Reviewer #2 (Public Review):

The authors characterized the recombinase-based cumulative fate maps for vesicular glutamate transporters (Vglut1, Vglut2 and Vglut3) expression and compared those maps to their real-time expression profiles in central NA neurons by RNA in situ hybridization in adult mice. Authors have revealed a new and intriguing expression pattern for Vglut2, along with an entirely uncharted co-expression domain for Vglut3 within central noradrenergic neurons. Interestingly, and in contrast to previous studies, the authors demonstrated that glutamatergic signaling in central noradrenergic neurons does not exert any influence on breathing and metabolic control either under normoxic/normocapnic conditions or after chemoreflex stimulation. Also, they showed for the first-time the Vglut3-expressing NA population in C2/A2 nuclei. In addition, they were also able to demonstrate Vglut2 expression in anterior NA populations, such as LC neurons, by using more refined techniques, unlike previous studies.

A major strength of the study is the use of a set of techniques to investigate the participation of NA-based glutamatergic signaling in breathing and metabolic control. The authors provided a full characterization of the recombinase-based cumulative fate maps for Vglut transporters. They performed real-time mRNA expression of Vglut transporters in central NA neurons of adult mice. Further, they evaluated the effect of knocking down Vglut2 expression in NA neurons using a DBH-Cre; Vglut2cKO mice on breathing and control in unanesthetized mice. Finally, they injected the AAV virus containing Cre-dependent Td tomato into LC of v-Glut2 Cre mice to verify the VGlut2 expression in LC-NA neurons. A very positive aspect of the article is that the authors combined ventilation with metabolic measurements. This integration holds particular significance, especially when delving into the exploration of respiratory chemosensitivity. Furthermore, the sample size of the experiments is excellent.<br /> Despite the clear strengths of the paper, some weaknesses exist. It is not clear in the manuscript if the experiments were performed in males and females and if the data were combined. I believe that the study would have benefited from a more comprehensive analysis exploring the sex specific differences. The reason I think this is particularly relevant is the developmental disorders mentioned by the authors, such as SIDS and Rett syndrome, which could potentially arise from disruptions in central noradrenergic (NA) function, exhibit varying degrees of sex predominance. Moreover, some of the noradrenergic cell groups are sexually dimorphic. For instance, female Wistar rats exhibit a larger LC size and more LC-NA neurons than male subjects (Pinos et al., 2001; Garcia-Falgueras et al., 2005). More recently, a detailed transcriptional profiling investigation has unveiled the identities of over 3,000 genes in the LC. This revelation has highlighted significant sexual dimorphisms, with more than 100 genes exhibiting differential expression within LC-NA neurons at the transcript level. Furthermore, this investigation has convincingly showcased that these distinct gene expression patterns have the capacity to elicit disparate behavioral responses between sexes (Mulvey et al., 2018). Therefore, the authors should compare the fate maps, Vglut transporters in males and females, at least considering LC-NA neurons. Even in the absence of identified sex differences, this information retains significant importance.<br /> An important point well raised by the authors is that although suggestive, these experiments do not definitively rule out that NA-Vglut2 based glutamatergic signaling has a role in breathing control. Subsequent experiments will be necessary to validate this hypothesis.

An improvement could be made in terms of measuring body temperature. Opting for implanted sensors over rectal probes would circumvent the need to open the chamber, thereby preventing alterations in gas composition during respiratory measurements. Further, what happens to body temperature phenotype in these animals under different gas exposures? These data should be included in the Tables.

Is it plausible that another neurotransmitter within NA neurons might be released in higher amounts in DBH-Cre; Vglut2 cKO mice to compensate for the deficiency in glutamate and prevent changes in ventilation?

Continuing along the same line of inquiry is there a possibility that Vglut2 cKO from NA neurons not only eliminates glutamate release but also reduces NA release? A similar mechanism was previously found in VGLUT2 cKO from DA neurons in previous studies (Alsio et al., 2011; Fortin et al., 2012; Hnasko et al., 2010). Additionally, does glutamate play a role in the vesicular loading of NA? Therefore, could the lack of effect on breathing be explained by the lack of noradrenaline and not glutamate?

Reviewer #4 (Public Review):

Summary:

Although previous research suggested that noradrenergic glutamatergic signaling could influence respiratory control, the work performed by Chang and colleagues reveals that excitatory (specifically Vglut2) neurons is dynamically and widely expressed throughout the central noradrenergic system, but it is not significantly crucial to change baseline breathing as well the hypercapnia and hypoxia ventilatory responses. The central point that will make a significant change in the field is how NA-glutamate transmission may influence breathing control and the dysfunction of NA neurons in respiratory disorders.

Strengths:

There are several strengths such as the comprehensive analysis of Vglut1, Vglut2, and Vglut3 expression in the central noradrenergic system and the combined measurements of breathing parameters in conscious unrestrained mice.

Other considerations :

These results strongly suggest that glutamate may not be necessary for modulating breathing under normal conditions or even when faced with high levels of carbon dioxide (hypercapnia) or low oxygen levels (hypoxia). This finding is unexpected, considering many studies have underscored glutamate's vital role in respiratory regulation, more so than catecholamines. This leads us to question the significance of catecholamines in controlling respiration. Moreover, if glutamate is not essential for this function, we need to explore its role in other physiological processes such as sympathetic nerve activity (SNA), thermoregulation, and sensory physiology.

eLife assessment

This study presents solid evidence to support the effectiveness of the novel eIF2B activator DNL343 in mitigating the integrated stress response (ISR) and reducing neurodegeneration associated with ISR activation in two mouse models. These important findings offer promise for the potential use of DNL343 in treating vanishing white matter disease (VWMD), a rare condition resulting from eIF2B loss of function, and in addressing other neurodegenerative disorders characterized by ISR involvement. The study also identified potential VWMD biomarkers, which hold significance for assessing disease progression and evaluating treatment responses.

Reviewer #3 (Public Review):

Summary:

ISR contributes to the pathogenesis of multiple neurodegenerative diseases, such as ALS, FTD, VWMD, etc. Targeting ISR is a promising avenue for therapeutic intervention. However, all previously identified ways to target ISR have problems. PERK inhibitors suppress ISR by inhibiting eIF2alpha phosphorylation and cause pancreatic toxicity in mice. In order to bypass eIF2alpha, previous studies have identified ISR suppressors that target eIF2B, such as ISRIB and 2BAct. These molecules suppress neurodegeneration but do not cause detrimental effects in mouse models. However, ISRIB is water-insoluble, and 2BAct causes cardiovascular complications in dogs, preventing their use in clinics. Here, the authors showed that DNL343, a new ISR inhibitor targeting eIF2B, suppresses features that can be related to neurodegeneration in mouse models. Combined with their previous results of a clinical phase I trial showing the safety of DNL343, these findings suggest the promise of DNL343 as a potential drug for neurodegenerative diseases in which ISR contributes to pathogenesis.

Strengths:

The finding is important and has disease implications.

Weakness:

The authors did not provide evidence that DNL343 suppresses the demise of nervous systems in their VWMD model.

Reviewer #1 (Public Review):

Summary:

In this study, the authors evaluated a novel eIF2B activator, DNL343, in two mouse models representing different integrated stress response (ISR) forms. They first assessed the pharmacokinetics of DNL343, demonstrating its ability to cross the blood-brain barrier and exhibit good bioavailability. In an acute ISR model induced by optic nerve crush (ONC) injury, DNL343 treatment reduced ISR-induced transcriptional changes and neuronal loss, demonstrating neuroprotective effects. Next, the authors generated an eIF2B loss-of-function mice model by knocking in disease-causing Eif2b5 variants. The model presents a chronic ISR and mimics vanishing white matter disease (VWMD). DNL343 treatment from the pre-symptomatic stage improved body weight and motor functions, corrected transcriptional changes, and reversed proteomic and metabolomic alterations in the brain and cerebrospinal fluid. DNL343 treatment initiated at an advanced disease stage also showed positive effects, restoring body weight gain, suppressing ISR, reducing neurodegeneration biomarkers, and extending lifespan. These findings highlight DNL343 as an effective ISR inhibitor with potential applications in treating VWMD and other neurodegenerative disorders involving ISR.

Strengths:

The study's findings regarding the novel compound DNL343 offer significant promise in addressing VWMD, a condition currently lacking disease-modifying treatment. DNL343 directly targets eIF2B, the disease-causing complex in VWMD, and demonstrates notable efficacy in reversing the integrated stress response (ISR) and mitigating neurodegeneration in a VWMD mouse model. These results raise hope for the potential application of DNL343 in VWMD treatment, a development eagerly anticipated by patients and the VWMD research community. Moreover, the study hints at the broader potential of DNL343 in treating other ISR-related neurodegenerative disorders, such as ALS, a prospect that holds broader interest. Additionally, the study's identification of potential biomarkers for VWMD represents a notable strength, potentially leading to improved disease progression assessment pending further confirmation in future research.

Weaknesses:

Direct biochemical evidence confirming DNL343's activity in eIF2B activation and its toxicity profile have been previously documented in a separate study. It would be beneficial to provide a more detailed introduction to this information, establishing a robust knowledge foundation for the in vivo study described in this work.

Reviewer #2 (Public Review):

Summary:

The authors developed DNL343, a CNS-penetrant small molecule integrated stress response (ISR) inhibitor, to treat neurodegenerative diseases caused by ISR.

Strengths:

DNL343 is an investigational CNS-penetrant small molecule integrated stress response (ISR) inhibitor designed to activate the eukaryotic initiation factor 2B (eIF2B) and suppress aberrant ISR activation. The therapeutic efficacy of DNL343 has been extensively characterized in two animal models. Importantly, plasma biomarkers of neuroinflammation and neurodegeneration can be reversed with DNL343 treatment. Remarkably, several of these biomarkers show differential levels in CSF and plasma from patients with vanishing white matter disease (VWMD) upon DNL343 treatment. Overall, this study is very exciting that targets ISR for therapeutic interventions.

Weaknesses:

My main questions center around the characterization of DNL343.

(1) Is there any biochemical evidence showing DNL343 activates eIF2B, such as binding and in vitro biochemical activity assays? A conference presentation was cited. "Osipov, M. (2022). Discovery of DNL343: a Potent Selective and Brain-penetrant eIF2B Activator Designed for the Treatment of Neurodegenerative Diseases. Medicinal Chemistry Gordon Research Conference. New London, NH." However, there is no public information about this presentation.<br /> (2) How was the selectivity of DNL343 demonstrated? What are the off-targets of DNL343, particularly when DNL343 is administered at a high dose? Thermal-proteasome profiling or photoaffinity labeling experiments could be considered.<br /> (3) What are the total drug concentrations in the brain and plasma? What are the unbound ratios?<br /> (4) If DNL343 is given intravenously, what are the concentrations in the brain and plasma after 5 minutes and 1 h or longer time points? In other words, does DNL343 cross BBB through passive diffusion or an active process?<br /> (5) What is the full PK profile of DNL343 for intravenous and oral dosing?<br /> (6) Are there any major drug metabolites that could be concerns?

Review for Revision:

The companion JMC paper, doi.org/10.1021/acs.jmedchem.3c02422, addressed most of my questions. However, I was unable to find the total concentrations of DNL343 in the brain and plasma or the raw data for the full PK in the JMC paper. Otherwise, the JMC publication addressed all my questions.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study, the authors evaluated a novel eIF2B activator, DNL343, in two mouse models representing different forms of the integrated stress response (ISR). They first assessed the pharmacokinetics of DNL343, demonstrating its ability to cross the blood-brain barrier and exhibit good bioavailability. In an acute ISR model induced by optic nerve crush (ONC) injury, DNL343 treatment reduced ISR-induced transcriptional changes and neuronal loss, demonstrating neuroprotective effects. Next, the authors generated an eIF2B loss-of-function mice model by knocking in disease-causing Eif2b5 variants. The model presents a chronic ISR and mimics vanishing white matter disease (VWMD). DNL343 treatment from the pre-symptomatic stage improved body weight and motor functions corrected transcriptional changes, and reversed proteomic and metabolomic alterations in the brain and cerebrospinal fluid. DNL343 treatment initiated at an advanced disease stage also showed positive effects, restoring body weight gain, suppressing ISR, reducing neurodegeneration biomarkers, and extending lifespan. These findings highlight DNL343 as an effective ISR inhibitor with potential applications in treating VWMD and other neurodegenerative disorders involving ISR.

Strengths:

The study's findings regarding the novel compound DNL343 offer significant promise in addressing VWMD, a condition currently lacking disease-modifying treatment. DNL343 directly targets eIF2B, the disease-causing complex in VWMD, and demonstrates notable efficacy in reversing the integrated stress response (ISR) and mitigating neurodegeneration in a VWMD mouse model. These results raise hope for the potential application of DNL343 in VWMD treatment, a development eagerly anticipated by patients and the VWMD research community. Moreover, the study hints at the broader potential of DNL343 in treating other ISR-related neurodegenerative disorders, such as amyotrophic lateral sclerosis, a prospect that holds broader interest. Additionally, the study's identification of potential biomarkers for VWMD represents a notable strength, potentially leading to improved disease progression assessment pending further confirmation in future research.

Weaknesses:

There are a couple of notable concerns in this study. Firstly, while the in vivo evidence strongly supports the efficacy of DNL343 in mitigating ISR and neurodegeneration, there is a lack of direct biochemical evidence to confirm its activity in eIF2B activation. Secondly, the potential for cardiovascular toxicity, which has been reported for a related eIF2B activator in a canine model (as mentioned in the manuscript), has not been evaluated for DNL343 in this study. This data gap regarding toxicity could be crucial for informing the future development of DNL343 for potential human use. Further investigation into these areas would be valuable for a comprehensive understanding of the compound's mechanisms and safety profile.

We thank the reviewer for the thoughtful feedback and an opportunity to provide further clarification. To address the first question regarding biochemical evidence of the mechanism of action of DNL343, we agree that additional data is helpful to interpreting the results presented in this manuscript. We now include a citation to Craig et al (Craig, R.A., 2nd, J. De Vicente, A.A. Estrada, J.A. Feng, K.W. Lexa, M.J. Canet, W.E. Dowdle, R.I. Erickson, B.N. Flores, P.C.G. Haddick, L.A. Kane, J.W. Lewcock, N.J. Moerke, S.B. Poda, Z. Sweeney, R.H. Takahashi, V. Tong, J. Wang, E. Yulyaningsih, H. Solanoy, K. Scearce-Levie, P.E. Sanchez, L. Tang, M. Xu, R. Zhang and M. Osipov (2024). "Discovery of DNL343: A Potent, Selective, and Brain-Penetrant eIF2B Activator Designed for the Treatment of Neurodegenerative Diseases." J Med Chem.) which includes the full details on the discovery and characterization of DNL343.

On the question of cardiovascular toxicity observed with previous eIF2B activating compounds, Craig et al also provides evidence in a non-human primate (cynomolgus monkey) model that DNL343 dosing did not result in QT prolongation or any functional cardiac changes. We have also completed a Phase 1 (NCT04268784) and Phase 1B double-blind (NCT05006352) trials in healthy and ALS participants, respectively and these trials are referenced on page 4, lines 102-103. The safety profile observed in these clinical studies supported further development of DNL343 for ALS in the Healey Platform trial (NCT04297683, Regimen G).

Reviewer #2 (Public Review):

Summary:

The authors developed DNL343, a CNS-penetrant small molecule integrated stress response (ISR) inhibitor, to treat neurodegenerative diseases caused by ISR.

Strengths:

DNL343 is an investigational CNS-penetrant small molecule integrated stress response (ISR) inhibitor designed to activate the eukaryotic initiation factor 2B (eIF2B) and suppress aberrant ISR activation. The therapeutic efficacy of DNL343 has been extensively characterized in two animal models. Importantly, plasma biomarkers of neuroinflammation and neurodegeneration can be reversed with DNL343 treatment. Remarkably, several of these biomarkers show differential levels in CSF and plasma from patients with vanishing white matter disease (VWMD) upon DNL343 treatment. Overall, this is a very exciting study to target ISR for therapeutic interventions.

Weaknesses:

My main questions center around the characterization of DNL343.

(1) Is there any biochemical evidence showing DNL343 activates eIF2B, such as binding assays or in vitro biochemical activity assays? A conference presentation was cited - "Osipov, M. (2022). Discovery of DNL343: a Potent Selective and Brain-penetrant eIF2B Activator Designed for the Treatment of Neurodegenerative Diseases. Medicinal Chemistry Gordon Research Conference. New London, NH." However, there needs to be public information about this presentation.

Information from this presentation and more details on the discovery and characterization of DNL343 can be found in Craig et al J Med Chem (2024) and this citation has been replaced.

(2) How was the selectivity of DNL343 demonstrated? What are the off-targets of DNL343, in particular when DNL343 is administered at a high dose? Thermal-proteasome profiling or photoaffinity labeling experiments could be considered.

Please see Craig et al J Med Chem (2024) for full details. In brief, there were no significant off target effects observed for DNL343 in a Cerep panel.

(3) What are the total drug concentrations in the brain and plasma? What are the unbound ratios?

Following a single oral dose of DNL343 in mice, unbound brain-to-unbound plasma exposures ratios (Kp,uu) of 0.8 to 1.1 were observed, indicating high CNS penetrance. This was further supported by CSF-to-unbound plasma exposures ratios at 0.9 in the same mouse study. The CNS penetrance was also confirmed in rats and NHP by CSF-to-unbound plasma ratios near unity as reported in Craig et al J Med Chem (2024).

(4) If DNL343 is given intravenously, what are the concentrations in the brain and plasma after 5 minutes and 1 hour or longer time points? In other words, does DNL343 cross BBB through passive diffusion or an active process?

Unbound brain-to-unbound plasma exposure ratios following a single oral dose in the mouse were 0.8 to 1.1 and showed no time dependence. These measurements were made prior to, near, and following plasma tmax of DNL343, indicating unbound DNL343 crosses the BBB through passive diffusion and rapidly reached equilibrium between the brain and systemic circulation. Details can be found in Craig et al J Med Chem (2024).

(5) What is the complete PK profile of DNL343 for intravenous and oral dosing?

DNL343 administered orally to mice as a suspension formulation showed plasma PK consistent with prolonged absorption with tmax ranging from 3 to 4 h, and a terminal elimination half-life (t1/2) of ~10 h. Details can be found in Craig et al J Med Chem (2024).

(6) Are there any major drug metabolites that could be of concern?

DNL343 metabolism is through Phase 1 biotransformation pathways. None of the in vivo circulating metabolites show potency towards eIF2B activation. Given that none of these metabolites are of concern, we believe this information is beyond the scope of the current manuscript.

Reviewer #3 (Public Review):

Summary:

ISR contributes to the pathogenesis of multiple neurodegenerative diseases, such as ALS, FTD, VWMD, etc. Targeting ISR is a promising avenue for potential therapeutics. However, previously identified ways to target ISR present some challenges. PERK inhibitors suppress ISR by inhibiting eIF2alpha phosphorylation and cause pancreatic toxicity in mice. In order to bypass eIF2alpha, previous studies have identified ISR suppressors that target eIF2B, such as ISRIB and 2BAct. These molecules suppress neurodegeneration but do not cause detrimental effects in mouse models. However, ISRIB is water-insoluble, and 2BAct causes cardiovascular complications in dogs, preventing their use in clinics. Here, the authors showed that DNL343, a new ISR inhibitor targeting eIF2B, suppresses neurodegeneration in mouse models. Combined with their previous results of a clinical phase I trial showing the safety of DNL343, these findings suggest the promise of DNL343 as a potential drug for neurodegenerative diseases in which ISR contributes to pathogenesis.

Strengths:

The finding is important and has disease implications, and the conclusion is not surprising.

Weaknesses:

The experimental design and data are hard to comprehend for an audience with a basic research background. This reviewer suggests that the authors use the same way that previous studies on ISRIB and 2BAct (e.g., Wong et al; eLife, 2019) designed experiments and interpret data.

We thank this reviewer for their feedback and recognition that DNL343 has a promising potential as treatment for neurodegenerative diseases. While our studies share some similarities to Wong et al., eLife (2019) and Abbink et al., ACTN (2019), our study design is intentionally distinct (e.g. inclusion of both prevention and treatment dosing paradigms, determining dose-response impact of drug treatment across biomarkers) which necessitates tailored data visualization to effectively communicate our findings. However, we understand the importance of clarity for a broader audience and to this end, we have made a number of changes to the data figures, in particular data from omics experiments in Figures 3 and 5. We also provided additional supplemental tables to aid data interpretation. This would hopefully cater to both audiences familiar with previous work and those with a less specialized background.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

(1) Demyelination is a significant pathological feature in the VWMD mouse model. The authors should clarify whether they observed similar demyelination in their study and if DNL343 had any impact on reversing this demyelination. These findings are crucial for assessing the compound's effectiveness in mitigating neurodegeneration.

Demyelination is indeed an important feature in the eIF2B LOF (VWMD) mouse model. Given that this phenotype and the ability to rescue the histological phenotype with this MOA (Wong et al; eLife, 2019, cited in introduction) is very well characterized, along with our limitation from the size and number of mouse tissues, we prioritized non-histological targeted and unbiased analyses that were aimed at identifying translatable biomarkers. Nonetheless, the totality of our data, in different mouse models and cell types, strongly supports DNL343 as a potent ISR inhibitor that is effective in attenuating neurodegeneration:

· In the optic nerve crush model, DNL343 dose-dependently reduced retinal cell degeneration

· In the VWMD mouse model, DNL343 attenuated the increase in a plasma biomarker of neurodegeneration, neurofilament-light, which corresponded to normalization in motor function.

· Metabolomic and lipidomic analyses in the VWMD mouse model brain showed increases in oxysterols, such as 7-ketocholesterol, and cholesterol esters and these lipids are associated with demyelination (Nugent et al, 2020). DNL343 treatment attenuated the levels of these oxysterols, indicating decreased demyelination.

· When initiated at an advance disease stage, reversal of plasma biomarkers of neurodegeneration (Nf-L) and neuroinflammation (GFAP) by DNL343 in this model was accompanied by extension in the lifespan that is otherwise shortened as the mutant animals succumb to disease.

These data highlight the potential therapeutic benefits of DNL343 in the broader context of ISR-mediated neurodegeneration which can include but may not be limited to VWMD.

(2) Figure 6 presents several biomarkers with significantly increased levels in VWMD mice and patient biofluids. However, these biomarkers are not reflected in the brain proteomics data presented in Figure 3. The discrepancy between these findings should be addressed and discussed in the manuscript to provide a more comprehensive understanding.

Proteins detected in Figure 6 were not detected by TMT proteomics in the CSF. In the brain, only GFAP was detected and the overall abundance in tissue were similar in both genetic groups. Cytokines such as TIMP1, MCP1 are usually present in low abundances and therefore are challenging to detect in broad discovery proteomics method applied in this study. Antibody-based immunoassays are better suited to specifically measure low abundant proteins than mass-spectrometry-based proteomics, while mass-spectrometry based methods offer wider dynamic range to detect more highly abundant proteins. Differences in detection sensitivity between immunoassay vs mass spectrometry assays has been previously noted (Petrera et al, J Proteome Res, 2021). We have added new text to address this point in the revised manuscript (page 7, line 274-277).

(3) Figure 7 discusses the effects of DNL343 treatment initiated at an advanced disease stage. Since the 4-week treatment did not rescue performance in the balance beam test (as shown in Figure 6A), it is important to clarify if a 20-week treatment had any impact on this parameter.

This reviewer raised an important question that we were unfortunately unable test. When the balance beam training was administered after 8 (out of 20) weeks of dosing, most animals of both wildtype and mutant genotypes struggled to remain on or maintain balance on the beam and were unable to progress traversing the beam, making the assay unsuccessful in this cohort. This impairment appeared to be driven by distinct factors in the two genotypes: age-associated obesity in wild-type animals and severe motor impairment in the eIF2B HOM mice, irrespective of treatment. While it is possible that other less demanding and more sensitive assays could reveal more nuanced differences, this, and our earlier data (Figure 4G-I), suggest that DNL343 could prevent but not reverse functional deterioration. This is in line with our understanding of DNL343 mechanism of action that does not include neuronal regeneration, a therapeutic effect that is likely required for functional recuperation. We have added this point to the manuscript (page 8, line 319-326).

Additionally, considering the significant increase in Gdf15 levels in the disease model, it would be valuable to know if DNL343 treatment affected Gdf15 levels. If these assays were conducted, reporting the data would greatly assist in evaluating the compound's efficacy when administered at an advanced disease stage.

We were not able to measure GDF15 levels in the 20-week study due to limitation in the in-life collected plasma samples which was dedicated to assessing biomarkers of neurodegeneration (Figure 7E-F). However, data from our 4-week treatment study, which was initiated at a similar age range to the 20-week treatment study (19-26 and 24-33 weeks of age, respectively), showed that DNL343 was able to reduce GDF15 levels in the brain (mRNA and protein) and CSF (protein) (Supplemental Figure 5A-C), suggesting that DNL343 reduces ISR activation at an advanced disease stage in the model. We expect that this reduction observed at 4 weeks of treatment would persist for the duration of the extended treatment in the 20-week cohort.

(4) A minor point. In Figures 5A, 5C, and 5E, it appears that the red-colored group should likely be labeled as "HOM 0 mg/kg" instead of "HOM 3 mg/kg".

This has been amended, thank you.

Reviewer #3 (Recommendations For The Authors):

Major concerns:

(1) The cellular function of DNL343 needs to be clarified. The authors claim that it activates eIF2B, but no cellular or molecular evidence is provided. Does it bind to eIF2B? Does it not affect eIF2alpha phosphorylation? Does it restore translation upon stress that causes eIF2alpha phosphorylation? Does it suppress stress granule assembly? The authors cited Sun, Tsai et al. 2023 and Osipov et al., 2022. However, these citations are conference abstracts with no published figures available for review.

We agree that additional data outlining the biochemical evidence of the mechanism of action of DNL343 was needed. We now include a citation to Craig et al J Med Chem (2024) that includes the full details on the discovery and molecular characterization of DNL343.

(2) It needs to be clarified how the authors selected the ISR marker genes. ISR genes are more than those selected. How about others? How did the authors measure the mRNA levels, bulk RNA-seq or RT-PCR? If the former, have the authors verified their results using RT-PCR? Have the authors measured the protein levels for nerve crush experiments (by both proteomic and individual protein analyses)? Also, no statistical analyses were found for the heat maps.

The ISR marker genes were selected by a combination of experimental and literature data. Transcriptomics analysis of the eIF2B HOM brains was conducted using untargeted RNAseq (Supplemental Figure 1B). Here, we found an enrichment of transcripts previously reported to be ISR dependent, namely Atf4, Chac1, Ddit3, Eif4ebp1, Ppp1r15a (Larhammar et al., 2017), Atf3, Asns, Mthfd2, Psat1, Sesn2, Slc1a5, Slc7a5, Slc7a11, Trib3 (Wong et al., 2019, Abbink et al., 2019).  These transcripts were assayed using targeted qPCR in the eIF2B HOM brains, spleen and PBMC (Supplemental Figure 1A, C, D) and in the retinas from the ONC experiments (Figure 2C). We have further clarified the analysis method for the gene expression data in the figure legends.

We did not interrogate the proteome of the retina in the ONC model. Our study in this model was intended as a proof-of-concept evaluation of DNL343 effects in this acute ISR-dependent model of neurodegeneration. To this end, we performed gene expression (Figure 2C) and immunofluorescence analyses (Figure 2D-F). Each of these analyses were conducted using dedicated whole retinas; conducting additional protein analyses would necessitate a separate cohort of animals.

We believe that heatmaps provide the best visualization of the data, particularly the dose dependent effects of DNL343 on multiple genes, but we understand the value for also providing statistical analyses. To address this, we provide additional Supplemental tables to show the outcome of statistical analyses undertaken. Statistical data relating to Figure 2C can be found on new Supplemental Tables 1 & 2; those relating to Supplemental Figures 1A, C, and D on new Supplemental Tables 3, 5, 6, respectively; that from Figure 4D on new Supplemental Table 8, and that from Figure 7D on new Supplemental Table 11.

(3) Both the authors and Wong et al. (eLife, 2019) performed transcriptomic analyses on HOM mice. How do the authors compare the two data sets? Are they the same?

In this work, transcriptomic approach was applied to confirm induction of ISR response in our in vivo model. While data are not identical, all of the top annotated genes shown in supplementary figure 1B were also deemed to be significant by Wong and coworkers (Bayes factor > 10). More importantly, as explained in our responses to question #2 from reviewer 3,  ISR genes highlighted in supplementary Figure 1B were also confirmed in two other studies (Larhammar et al., 2017, Abbink et al., 2019). These data support our interpretation that eIF2B HOM have elevated ISR relative to WT mice. We have added new text to line 164 on page 5 to clarify this point.

(4) Can the authors interpret their omic data using volcano plots for HOM rescue experiments, as Wong et al. did in eLife 2019? Heat maps with statistical analyses are more straightforward to comprehend. Can the authors verify some of these data using RT-PCR, Western blot, etc.?

We added additional pathway interpretation in our Figure 3 and 5 to highlight key biological processes altered in the brain and cellular compartment origin of CSF proteins changed in eIF2B HOM at baseline and following treatment with DNL343. Our treatment designed employed multiple dosing levels and as such, summarization by volcano plot would have resulted in creation of many figures that can be more easily captured by a single heat map plot. However, to provide additional quantitative information, we now added supplementary tables showing full statistical analysis for all heat maps for added clarity and transparency.

We demonstrated 100% correlation between the select genes we examined by qPCR in supplemental Figure 1A and those identified from brain by RNA-seq. In addition, question of reliability of RNA-seq data has been previously been examined in great detail (Everaet et al, Sci Rep 2017) and found ~85% concordance between RNA-seq and qPCR data and those that were discordant tended to have < 2 log2FC and were present in low abundance. Given that top core ISR genes identified in our study have >2 log2FC and have been verified by other independent labs (Larhammar et al., 2017, Abbink et al., 2019, Wong et al., 2019). Based on these, we do not think that there is a rationale need for technical confirmation of RNAseq data.

Risks for mis-annotation of proteins in TMT data were further mitigated by removing protein with coverage < 20% and having less than 8 unique peptides detected and setting protein annotation FDR to <1%.

Additionally, TMT-labelling based proteomics offers wider dynamic range and sensitivity than western blotting. Validation of TMT logFC data with western blot technique, which is less quantitative and has lower dynamic ranges of detection may not be very informative. Furthermore, similar trends of changes in key ISR genes and proteins shown in figures 4D and 5A (e.g PSAT, SLC7A11, SLC7A5) provides additional support for the authenticity of proteins identified in this work.

Also, for Figures 4E and F, it is assumed that each line represents an individual animal, but why their body weight gains are so different for the wild type? Can the authors plot the mean and s.e.m.? Also, there are no data about neurodegeneration. The authors need to show microscopy images, count the numbers, and assess the morphology of nerve cells.

The large data spread in the body weight gain in our wild-type mice reflect the normal variability of this endpoint which can be influenced by sex and age. Indeed, both factors are present in our cohorts as animals of both sexes were included and there was a 7-week age-range (10-17 weeks of age at dosing start). Each line in Figures 4E-F indeed represents data sampled from individual animal over time. We chose to represent the data this way for transparency and have provided additional visualization (new Supplemental Figure 3) showing both body weight gain and plasma Nf-L levels as mean ± SEM as requested by this reviewer.

In this study we chose to use a clinically-relevant biomarker of neurodegeneration, plasma neurofilament light chain (NfL) (Figure 4F). This allowed us to prioritize the tissue samples from these studies to execute comprehensive unbiased analyses for more complete characterization of the phenotype of these eIF2B LoF mice. NfL is a biomarker that has been recognized as a sensitive measurement of neuronal/axonal damage regardless of cause (Gaetani et al., 2018, Khalil et al., 2018). Elevated levels of plasma (and CSF) NfL levels has been demonstrated across neurodegenerative conditions such as Alzheimer’s disease (Giacomucci et al., 2022), multiple sclerosis (Ferreira-Atuesta et al., 2021), and in ALS (Huang et al., 2018).

(5) How ISR is connected to metabolomic changes? Can the authors explain it?

ISR caused significant increases in amino acid transporter and serine/glycine/1-carbon metabolism enzymes transcript and protein abundances that were highlighted in Figure 3A and C and lines 237-255 in the main text. Similar patterns were also observed in prior published studies (Larhammar et al., 2017, Abbink et al., 2019, Wong et al., 2019). Consistent with these changes we observed increased levels of Alanine (transported by SLC3A2, SLC7A11, SLC7A3) and decreased cystathionine levels (associated with increased expression of CTH).  ATF4 is one of the main orchestrator of ISR response to stress (e.g., amino acid deprivation) and it is required for expression of amino acid transporters and enzymes required for synthesis non-essential amino acids (PMID: 28494858). ATF4 increases cellular amino acid uptake and deliver AA needed for synthesis of proteins and glutathione needed for survival.

We also observed prominent changes in CE in eIF2B HOM and its normalization with DNL343 treatment shown in Figure 5C. We checked for changes in expression levels of CEL, CES1, LCAT, LIPA, SOAT1, and NCEH1 proteins involved in CE metabolism and failed to detect any changes in protein or RNA abundances.  This  suggests that a rapid demyelination is a more likely trigger for CE accumulation as reported in FTD-GRN (Marian OC et al., 2023 acta neuropathol commun 11, 52), and in experimental demyelination models (Nugent AA et al., 2020 Neuron). We have added new text to the discussion section of the manuscript page 9, lines 408-411 to discuss how these results relate to each other.

(6) It is hard to understand the biomarker part. The authors said "potential translational biomarkers are elevated..." Do the authors mean they are elevated so they can be potential biomarkers? If their levels are unchanged (e.g., TIMP-1), how can they be biomarkers? Also, this part needs a conclusion/summary. Also, what does "reversed biomarkers..." mean?

We have modified the text to clarify and included a concluding sentence for this section of the results (page 7, lines 297-299). In assessing whether a given protein could be a potential translational biomarker for human disease we evaluated if the following two conditions were met: (1) Increased or decreased gene expression or protein levels of the biomarker in the brain or biofluids (CSF or plasma) of Eif2b5 R191H homozygote mice relative to wild-type controls that is modulated or normalized by administration of DNL343 and (2) protein levels in biofluids from VWMD patients that show differential levels than healthy controls in the same directionality as what is seen in the mouse model. GDF-15, GFAP, and NfL meet these criteria, but TIMP-1 and MCP-1 do not.

Minor concerns:

(1) Please explain which multiple comparison tests the authors used.

This information has been further clarified in the figure legends.

(2) Administrating the drug at an advanced stage led to a trend of NfL reduction but did not rescue function. Can the authors discuss what this means?

Further elaboration and discussion about this finding have been added to the results section on page 8, line 319-325.

(3) For statistical analyses on the bar graphs, it would be better if the authors labeled the comparison pairs on the graphs.

We agree that labelling comparisons in bar graphs could aid the readership and have added this modification. Additionally, comparisons are indicated in the figure legend.

(4) The authors need to state clearly that 2BAct's cardiovascular toxicity was observed in dogs, not mice. The current study does not exclude similar DNL343 toxicity. However, previous clinical trials suggest that DNL343 may be safe for humans.

The suggestion to specify cardiovascular toxicity in dogs has been added (page 3, line 101), thank you. We now include a citation to Craig et al J Med Chem (2024) that provides evidence in a non-human primate (cynomolgus monkey) model that DNL343 dosing did not result in QT prolongation or any functional cardiac changes. We have also completed a Phase 1 (NCT04268784) and Phase 1B double-blind (NCT05006352) trials in healthy and ALS participants, respectively and now include reference to these trials on page 4, lines 102-104. The safety profile observed in these clinical studies supported further development of DNL343 for ALS in the Healey Platform trial (NCT04297683, Regimen G).

eLife assessment

This fundamental study addresses the question of how certain zooplankton achieve barotaxis, directed locomotion in response to changes in hydraulic pressure. The authors provide compelling evidence that the response involves ciliary photoreceptors interacting with motoneurons. This work should be of broad interest to scientists working on mechanosensation, cilia, locomotion, and photoreceptors.

Joint Public Review:

In this work, the authors address a fundamental question in the biological physics of many marine organisms, across a range of sizes: what is the mechanism by which they measure and respond to pressure. Such responses are classed under the term "barotaxis", with a specific response termed "barokinesis", in which swimming speed increases with depth (hence with pressure). While macroscopic structures such as gas-filled bladders are known to be relevant in fish, the mechanism for smaller organisms has remained unclear. In this work, the authors use ciliated larvae of the marine annelid Platynereis dumerilii to investigate this question. This organism has previously been of great importance in unravelling the mechanism of multicellular phototaxis associated with a ciliated band of tissue directed by light falling on photoreceptors.

In the present work, the authors use a bespoke system to apply controlled pressure changes to organisms in water and to monitor their transient response in terms of swimming speed and characteristics of swimming trajectories. They establish that those changes are based on relative pressure, and are reflected in changes in the ciliary beating. Significantly, by imaging neuronal activity during pressure stimulation, it was shown that ciliary photoreceptor cells are activated during the pressure response. That these photoreceptors are implicated in the response was verified by the reduced response of certain mutants, which appear to have defective cilia. Finally, serotinin was implicated in the synaptic response of those neurons.

This work is an impressive and synergistic combination of a number of different biological and physical probes into this complex problem. The ultimate result, that ciliary photoreceptors are implicated, is fascinating and suggests and interesting interplay between photoreception and pressure detection.

Future studies ought to address the following three questions opened by this work:

(1) How the off response to decrease of pressure is mediated

(2) Which receptor/channel mediates in photoreceptors the response to increased pressure,

(3) How the integration of light and pressure information is integrated by photoreceptors in order to guide the behavior of the larvae.

Author response:

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

Reviewer #1:

We thank Reviewer #1 for the assessment of our study.

Reviewer #2:

The authors should use DF/F to quantify over time the calcium response in photoreceptors. Furthermore, they should show that there is no concern of motion artifact when the pressure changes - as it could be a concern”.

We used the ΔR/R measure (as defined in Böhm et al. 2016) to correct for motion artifacts due to the larvae moving out of the focal plane at the onset of pressure stimulation. This measure calculates the ratio of the GCaMP signal and a reference fluorescent signal (tdTomato in our case). This ratiometric quantification can better correct for changes in fluorescence that are not related to changes in calcium concentration than the ΔF/F metric, which does not use an independent reference channel.

The authors have not shown

(1) how the off response to decrease of pressure is mediated

(2) which receptor/channel mediates in photoreceptors the response to increased pressure,

(3) nor how the integration of light and pressure information is integrated by photoreceptors in order to guide the behavior of the larvae.

These points are beyond the scope of the study. However, if possible within a short time frame, it would be really interesting to find out whether conflicting stimuli or converging stimuli (light & pressure) can cancel each other out or synergize. In particular since the authors cite unpublished results in the discussion: "Our unpublished results indeed suggest that green light determines the direction of swimming and can override upward swimming induced by pressure, which only influences the speed of swimming (LABC and GJ, unpublished)." Showing in one panel this very cool phenomenon would be exciting & open tons of questions for the field.”

We agree that investigating the interaction of light and pressure is a very exciting direction. However, doing it properly with the rigour we characterised pressure sensation here (across stages, pressure levels and genotypes) and phototaxis and UV avoidance in previous work (across stages, wavelengths, genotypes and stimulus direction; see Randel et al. 2014, Gühmann et al. 2015, Verasztó et al. 2018, Jokura et al. 2023) would require a separate in-depth study.

We agree with points 1-3 regarding the limitations and mentioned these in the discussion.

(1) Although we carried out pressure-release experiments to characterise in more detail the response to pressure OFF, our setup did not allow us to control pressure release as accurately as we could for pressure increase. Therefore, we decided not to address this aspect of the response in more detail in this study.

“Upon a decrease in pressure, three-day-old (but not two-day-old) larvae also show an off-response characterised by downward swimming. We have not analysed in detail the neuronal mechanisms of this response but it may depend on an inverted activation of the cPRC circuit, as happens during UV avoidance (Jokura et al., 2023)”

(2) We decided not to explore this important question in this study, due to the significant effort it would take to test the expression and function of potential candidate channels in pressure transduction mechanism. “The cellular and molecular mechanisms by which cPRCs sense and transduce changes in hydrostatic pressure deserve further enquiry. “ and “The molecular mechanisms of pressure detection remain unclear. Components of the phototransduction cascade may be involved in pressure sensation. Our results indicate that the ciliary opsin required for detecting UV light is not essential for pressure sensation.“ We hypothesise in the discussion that TRP channels may play a role in pressure transduction, due to their diversity, multiple modalities and participation in phototransduction cascades.

(3) We considered that the complexity of this question merits a separate study, where both cues can be accurately titrated and temporally combined to dissect the mechanisms of sensory integration. We have therefore removed the sentence referring to the interaction of phototaxis and the pressure response from the discussion.

“How UV and pressure signals are integrated by the cPRC and how other light responses such as phototaxis interact with pressure responses remain exciting avenues for future research.”

eLife assessment

This valuable study elucidates the essential role of the chromatin regulator KDM6B in the establishment and maintenance of neural stem cells (NSCs) in the mouse hippocampus. While the evidence supporting the authors' claims is largely solid, a more comprehensive investigation into the cellular and molecular events underlying the loss of hippocampal NSCs would have further strengthened the study. Nonetheless, the findings will be of interest to biologists studying neural development and NSCs.

Reviewer #1 (Public Review):

Summary:

The authors have previously studied the function of the lysine demethylase Kdm6b as a positive regulator of neurogenesis from subventricular zone neural precursors. Here they knockout Kdm6b in progenitors of the dentate gyrus and show convincingly that deletion causes precocious differentiation of these stem cells. These data are valuable and show that Kdm6b can have very different functions in distinct populations of neuronal progenitors.

Strengths:

Kdm6b has repeatedly been implicated as a positive regulator of differentiation in the cellular transitions where it has been studied before. By contrast, here the authors show convincingly that it is required for maintenance of the stem cell state in the hippocampus, and that Kdm6b deletion is associated with premature stem cell differentiation and a small dentate gyrus in the adult hippocampus. Inducible deletion of Kdm6b in adult hippocampal stem cells confirms the precocious differentiation and loss of this population in the absence of Kdm6b even when induced at this later age.

Weaknesses:

This is a surprising finding in light of many other papers that are well-cited by the authors, including their own studies of SVZ progenitors where Kdm6b promotes neuronal differentiation. However, the weakness of the study is that the authors shed very little light on why the effects of Kdm6b would be so different (in fact, largely opposite) in the two stem cell populations they have studied.

Reviewer #2 (Public Review):

Summary:

Gil & Lim et al. applied mouse genetic models to study the roles of chromatin regulator KDM6B in regulating the development of the hippocampal dentate gyrus (DG), as well as the establishment and maintenance of DG NSCs. KDM6B is expressed in postnatal DGs. Importantly, conditional knockout of Kdm2b in embryonic DG progenitors leads to a significantly smaller DG with loss of DG NSCs. Hippocampal-dependent behaviors are defective in Kdm6b-cKO mice. Deletion of Kdm6b results in precocious neuronal differentiation and loss of the NSC population in both postnatal and adult DGs. Single-cell RNA-seq reveals disrupted stem cell maintenance gene signature in Kdm6b-deleted NSCs. Moreover, CUT&RUN studies showed that Kdm6b deletion increases H3K27me3 levels at a few NSC maintenance genes.

Strengths:

The conclusions of this paper are mostly well supported by data. The discussion is thorough.

Weaknesses:

I concur with the two reviewing editors who noted that the paper lacks insights into how KDM6B regulates the expression of NSC genes in DG precursors. Additionally, the authors did not provide evidence regarding whether the function of KDM6B is enzymatically dependent.

The Kdm6b-cKO brain exhibited apparently smaller DGs, indicating compromised neurogenesis. While the authors observed an increased number of IPCs in the E17.5 DGs (Figure 4B-4C) and an increased number of BrdU+TBR2+PROX1+ cells in the P0.5 DGs (Figure 5B-5C), it is perplexing why this does not lead to an increased number of PROX1+ DG neurons? Further investigation into the cellular mechanisms underlying these events would enhance the understanding of Kdm6b's role in neurogenesis.

Many data were not of sufficient quality and should be improved.

Reviewer #3 (Public Review):

Gil et al provide novel evidence that the chromatin regulator KDM6B is important for establishing and maintaining the neural stem cell (NSC) pool within the dentate gyrus in development and adulthood. They show compelling evidence that loss of KDM6B promotes precocious neuronal differentiation, resulting in a failure to establish and maintain the dentate gyrus NSC pool. The strongest evidence they provide is their immunohistochemistry analysis, in which they observed precocious expression of later differentiation markers from cells marked by BrdU. However, given that KDM6B is ubiquitously expressed, it is difficult to ascertain if their dysregulation is due to a direct loss of KDM6B within NSCs or caused by dysregulation of other glial cells impacted by KDM6B loss through the hGFAP-Cre. Characterization of mature glia would strengthen the work.

They additionally provide evidence of precocious differentiation through scRNA-seq by highlighting key genes that are dysregulated with KDM6B loss. It appears the clustering analysis into cell types was done with WT and KDM6b-depleted cells together. The evidence for precocious differentiation would be greatly strengthened if they instead determined cell-type specific clusters using their WT samples and then observed if fewer cells are characterized as NSCs and more cells align to later developmental stage clusters with KDM6B depletion.

Gil et al propose that KDM6B loss leads to hippocampus-specific impairments in learning and memory. While KDM6B-depleted mice do show a significant decrease in freezing time in contextual fear conditioning, Figure 2 Supplement 1 shows KDM6B-depleted mice are hyperactive compared to WT in the open field test. Thus, the reduction in freezing could be due to hyperactivity. Plotting freezing time in short bins throughout the duration of the test can help clarify this. It would be additionally helpful to plot the training baseline and the test on the same graph and compare their freezing from baseline to clarify if they completely fail to freeze or show a reduction in freezing compared to the wild-type.

Author response:

We thank the reviewers for their positive evaluation and constructive comments.  In our revision, we will aim to improve the analysis of our existing data and perform new experiments to address questions raised by the reviewers. 

Reviewer 1 found it interesting that Kdm6b-deletion in hippocampal dentate gyrus (DG) neural stem cells causes precocious neuronal differentiation, whereas in contrast, Kdm6b is required for the maturation of neural progenitors in the ventricular-subventricular zone (V-SVZ). In the submitted manuscript, we did not provide much insight into the differences in Kdm6b function in these two neural stem cell populations. We plan on performing new experiments and expanding on our prior V-SVZ studies in a way that allows a direct comparison to the analyses of the DG. We hope that the addition of this data will shed light on why Kdm6b-deletion produces such different phenotypes in postnatal neural stem cells of the mouse brain. 

Reviewer 2 noted that our submitted manuscript lacked insight into how KDM6B regulates gene expression. In particular, this reviewer asked whether the function of KDM6B is mediated by its enzymatic activity. The CUT&RUN experiment in our manuscript revealed an increase in H3K27me3 levels at select neural maintenance genes in the DG of Kdm6b-deleted mice. However, we agree that this data is insufficient to assess the significance of KDM6B-mediated H3K27me3 demethylation in regulating the NSC transcriptome. To address this point, we are performing experiments that can directly test this mechanistic model of KDM6B function and answer the question of whether the H3K27me3 demethylase activity of KDM6B is required for its ability to activate transcription.  Reviewer 2 also had a specific question about the cell types observed in the developing hippocampus after Kdm6b-deletion, and we believe that additional analyses will provide clarity to the overall phenotype.  More generally, we will aim to improve data quality and visualization. 

Reviewer 3 raised the concern that because Kdm6b is not exclusively expressed in neural stem cells, the phenotype of precocious neuronal differentiation in mice with Kdm6b-deletion driven by the hGFAP-Cre transgene may be indirect, such as through changes in mature glial populations.  We will study the mature glia, as suggested by the reviewer.  We will also more thoroughly describe how our experiments targeting Kdm6b-deletion to adult neural stem cells with the tamoxifen-inducible Nestin-CreER provide evidence for the precocious neuronal differentiation phenotype being cell autonomous, at least in adult mice.  Reviewer 3 also had helpful suggestions for analyzing our single-cell RNA-seq data and behavioral studies, and we will address these comments in the revision. 

Again, we thank the editors and reviewers for their time and consideration.  We believe that our manuscript will be greatly improved through this review process and hope to construct a stronger understanding of the role of KDM6B in DG neurogenesis.

Reviewer #3 (Public Review):

Summary:

The goal of this study was to carry out an in-depth granular and unbiased phenotyping of peripheral blood circulating Tfh specific to two malaria vaccine candidates, PfSEA-1A and PfGARP, and correlate these with age (children vs adults) and protection from malaria (antibody titers against Plasmodium antigens.). The authors further attempted to identify any specific differences in the Tfh responses to these two distinct malaria antigens.

Strengths:

The authors had access to peripheral blood samples from children and adults living in a malaria-endemic region of Kenya. The authors studied these samples using in vitro restimulation in the presence of specific malaria antigens. The authors generated a very rich data set from these valuable samples using cutting-edge spectral flow cytometry and a 21-plex panel that included a variety of surface markers, cytokines, and transcription factors.

Weaknesses:

- Quantifying antigen-specific T cells by flow cytometry requires the use of either 1- tetramers or 2- in vitro restimulation with specific antigens followed by identification of TCR-activated cells based on de-novo expression of activation markers (e.g. intracellular cytokine staining and/or surface marker staining). Although authors use an in vitro restimulation strategy, they do not focus their study on cells de-novo expressing activation markers as a result of restimulation; therefore, their study is not really on antigen-specific cTfh. Moreover, the authors report no changes in the expression of activation markers commonly used to identify antigen-specific T cells upon in vitro restimulation (including IFNg and CD40L); therefore, it is not clear if their in vitro restimulation with malaria antigens actually worked.

- CXCR5+CD4+ memory T cells have been shown to present multi-potency and plasticity, capable of differentiating to non-Tfh subsets upon re-challenge. Although authors included in their flow panel a good number of markers commonly used in combination to identify Tfh (CXCR5, PD-1, ICOS, Bcl-6, IL-21), they only used one single marker (CXCR5) as their basis to define Tfh, thus providing a weak definition for Tfh cells and follow up downstream analysis.

- Previous works have used FACS-sorting and in vitro assays for cytokine production and B cell help to study the functional capacity of different cTfh subsets in blood from Plasmodium-infected individuals. In this study, authors do not carry out any such assays to isolate and evaluate the functional capacity of the different Tfh subsets identified. Thus, all the suggestions for the role that these different cTfh subsets may have in vivo in the context of malaria remain highly hypothetical.

- The authors have not included malaria unexposed control groups in their study, and experimental groups are relatively small (n=13).

eLife assessment

Using multiparameter spectral flow cytometry and unbiased clustering analysis, this study provides useful insights into the heterogeneity of antigens-specific circulating T follicular helper cells from children and adults living in malaria-endemic areas of Kenya. Although the study is well-designed, the analysis and interpretation of the potential functional roles for PfSEA-1A- and PfGARP-specific subsets of circulating T follicular helper cells are incomplete.

Reviewer #1 (Public Review):

Summary:

This study aims to understand the malaria antigen-specific cTfh profile of children and adults living in a malaria holoendemic area. PBMC samples from children and adults were unstimulated or stimulated with PfSEA-1A or PfGARP in vitro for 6h and analysed by a cTfh-focused panel. Unsupervised clustering and analysis on cTfh were performed.

The main conclusions are:<br /> (1) the cohort of children has more diverse (cTfh1/2/17) recall responses compared to the cohort of adults (mainly cTfh17) and<br /> (2) Pf-GARP stimulates better cTfh17 responses in adults, thus a promising vaccine candidate.

Strengths:

This study is in general well-designed and with excellent data analysis. The use of unsupervised clustering is a nice attempt to understand the heterogeneity of cTfh cells. Figure 9 is a beautiful summary of the findings.

Weaknesses:

(1) Most of my concerns are related to using PfSEA-1A and PfGARP to analyse cTfh in vitro stimulation response. In vitro, stimulation on cTfh cells has been frequently used (e.g. Dan et al, PMID: 27342848), usually by antigen stimulation for 9h and analysed CD69/CD40L expression, or 18h and CD25/OX40. However, the authors use a different strategy that has not been validated to analyse in vitro stimulated cTfh. Also, they excluded CD25+ cells which might be activated cTfh. I am concerned about whether the conclusions based on these results are reliable.

It has been shown that cTfh cells can hardly produce cytokines by Dan et al. However, in this paper, the authors report the significant secretion of IL-4 and IFNg on some cTfh clusters after 6h stimulation. If the stimulation is antigen-specific through TCR, why cTfh1 cells upregulate IL-4 but not IFNg in Figure 6? I believe including the representative FACS plots of IL-4, IFNg, IL21 staining, and using %positive rather than MFI can make the conclusion more convincing. Similarly, the author should validate whether TCR stimulation under their system for 6h can induce robust BCL6/cMAF expression in cTfh cells. Moreover, there is no CD40L expression. Does this mean TCR stimulation mediated BCl6/cMAF upregulation and cytokine secretion precede CD40L expression?

In summary, I am particularly concerned about the method used to analyse PfSEA-1A and PfGARP-specific cTfh responses because it lacks proper validation. I am unsure if the conclusions related to PfSEA-1A/PfGARP-specific responses are reliable.

(2) The section between lines 246-269 is confusing. Line 249, comparing the abundance after antigen stimulation is improper because 6h stimulation (under Golgi stop) should not induce cell division. I think the major conclusions are contained in Figure 5e, that (A) antigen stimulation will not alter cell number in each cluster and (B) children have more MC03, 06 and fewer MC02, etc.). The authors should consider removing statements between lines 255-259 because the trends are the same regardless of stimulations.

Reviewer #2 (Public Review):

Summary:

Forconi et al explore the heterogeneity of circulating Tfh cell responses in children and adults from malaria-endemic Kenya, and further compare such differences following stimulation with two malaria antigens. In particular, the authors also raised an important consideration for the study of Tfh cells in general, which is the hidden diversity that may exist within the current 'standard' gating strategies for these cells. The utility of multiparametric flow cytometry as well as unbiased clustering analysis provides a potentially potent methodology for exploring this hidden depth. However, the current state of analysis presented does not aid the understanding of this heterogeneity. This main goal of the study could hopefully be achieved by putting all the parameters used in one context, before dissecting such differences into their specific clinical contexts.

Strengths:

Understanding the full heterogeneity of Tfh cells in the context of infection is an important topic of interest to the community. The study included clinical groupings such as age group differences and differences in response to different malaria antigens to further highlight context-dependent heterogeneity, which offers new knowledge to the field. However, improvements in data analyses and presentation strategies should be made in order to fully utilize the potential of this study.

Weaknesses:

In general, most studies using multiparameter analysis coupled with an unbiased grouping/clustering approach aim to describe differences between all the parameters used for defining groupings, prior to exploring differences between these groupings in specific contexts. However, the authors have opted to separate these into sections using "subset chemokine markers", "surface activation markers" and then "cytokine responses", yet nuances within all three of these major groups were taken into account when defining the various Tfh identities. Thus, it would make sense to show how all of these parameters are associated with one another within one specific context to first logically establish to the readers how can we better define Tfh heterogeneity. When presented this way, some of the identities such as those that are less clear such as "MC03/MC04/ MC05/ MC08" may even be better revealed. once established, all of these clusters can then be subsequently explored in further detail to understand cluster-specific differences in children vs adults, and in the various stimulation conditions. Since the authors also showed that many of the activation markers were not significantly altered post-stimulation thus there is no real obstacle for merging the entire dataset for the first part of this study which is to define Tfh heterogeneity in an unbiased manner regardless of age groups or stimulation conditions. Other studies using similar approaches such as Mathew et al 2020 (doi: 10.1126/science.abc8) or Orecchioni et al 2017 (doi: 10.1038/s41467-017-01015-3) can be referred to for more effective data presentation strategies.

Accordingly, the expression of cytokines and transcription factors can only be reliably detected following stimulation. However, the underlying background responses need to be taken into account for understanding "true" positive signals. The only raw data for this was shown in the form of of heatmap where no proper ordering was given to ensure that readers can easily interpret the expression of these markers following stimulation relative to no stimulation. Thus, it is difficult to reliably interpret any real differences reported without this. Finally, the authors report differences in either cluster abundance or cluster-specific cytokine/ transcription factor expression in Tfh cell subsets when comparing children vs adults, and between the two malaria antigens. The comparisons of cytokine/transcription factor between groups will be more clearly highlighted by appropriately combining groupings rather than keeping them separate as in Figures 6 and 7.

eLife assessment

This manuscript makes valuable contributions to our understanding of cell polarisation dynamics and its underlying mechanisms. Through the development of a computational pipeline, the authors provide solid evidence that compensatory actions, whether regulatory or spatial, are essential for the robustness of the polarisation pattern. However, a more comprehensive validation against experimental data and a proper estimation of model parameters are required for further characterization and predictions in natural systems, such as the C. elegans embryo.

Joint Public Review:

The polarisation phenomenon describes how proteins within a signalling network segregate into different spatial domains. This phenomenon holds fundamental importance in biology, contributing to various cellular processes such as cell migration, cell division, and symmetry breaking in embryonic morphogenesis. In this manuscript, the authors assess the robustness of stable asymmetric patterns using both a previously proposed minimal model of a 2-node network and a more realistic 5-node network based on the C. elegans cell polarisation network, which exhibits anterior-posterior asymmetry. They introduce a computational pipeline for numerically exploring the dynamics of a given reaction-diffusion network and evaluate the stability of a polarisation pattern. Typically, the establishment of polarisation requires the mutual inhibition of two groups of proteins, forming a 2-node antagonistic network. Through a reaction-diffusion formulation, the authors initially demonstrate that the widely-used 2-node antagonistic network for creating polarised patterns fails to maintain the polarised pattern in the face of simple modifications. However, the collapsed polarisation can be restored by combining two or more opposing regulations. The position of the interface can be adjusted with spatially varied kinetic parameters. Furthermore, the authors show that the 5-node network utilised by C. elegans is the most stable for maintaining polarisation against parameter changes, identifying key parameters that impact the position of the interface. While the results offer novel and insightful perspectives on the network's robustness for cell polarisation, the manuscript lacks comprehensive validation against experimental data, justified node-node network interactions, and proper estimation of model parameters (based on quantitative measurements or molecular intensity distributions). These limitations significantly restrict the utility of the model in making meaningful predictions or advancing our understanding of cell polarisation and pattern formation in natural systems, such as the C. elegans embryo.<br /> In more detail, the authors demonstrate that the simplified 2-node model requires precise parameter fine-tuning to maintain stable polarisation. Any single modification to this 2-node network disrupts the polarisation pattern, highlighting the model's lack of robustness. However, stability is achieved when two opposite modifications are applied, which also increases the number of parameter sets that sustain the pattern. This robustness is contingent on monotonic correlations between all system parameters.

The study extends its significance by examining how cells maintain pattern stability amid spatial parameter variations, which are common in natural systems due to extracellular and intracellular fluctuations. The authors found that in the 2-node network, varying individual parameters spatially disrupt the pattern, but stability is restored with compensatory variations. Additionally, the polarisation interface stabilises around the step transition between parameter values, making its localisation tunable. This suggests a potential biological mechanism where localisation might be regulated through signalling perception.

Focusing on the C. elegans cell polarisation network, the authors propose a 5-node network based on an exhaustive literature review, summarised in a supplementary table. Using their computational pipeline, they identify several parameter sets capable of achieving stable polarisation and claim that their model replicates experimental behaviour, even when simulating mutants. They also found that among 34 possible network structures, the wild-type network with mutual inhibition is the only one that proves viable in the computational pipeline. Compared with previous studies, which typically considered only 2- or 3-node networks, this analysis provides a more complete and realistic picture of the signalling network behind polarisation in the C. elegans embryo. In particular, the model for C. elegans cell polarisation paves the way for further in silico experiments to investigate the role of the network structure over the polarisation dynamics. The authors suggest that the natural 5-node network of C. elegans is optimised for maintaining cell polarisation, demonstrating the elegance of evolution in finding the optimal network structure to achieve certain functions.

Noteworthy limitations are also found in this work. To simplify the model for numerical exploration, the authors assume several reactions have equivalent dynamics, reducing the parameter space to three independent dimensions. While the authors briefly acknowledge this limitation in the "Discussion and Conclusion" section, further analysis might be required to understand the implications. For instance, it is not clear how the results depend on the particular choice of parameters. The authors showed that adding additional regulation might disrupt the polarised pattern, with the conclusion apparently depending on the strength of the regulation. Even for the 5-node wild-type network, which is the most robust, adding a strong enough self-activation of [A], as done in the 2-node network, will probably cause the polarised pattern to collapse as well.

Additionally, the authors utilise parameter values that are unrealistic, fail to provide units for some of them, and assume unknown parameter values without justification. The model appears to have non-dimensionalised length but not time, resulting in a mix of dimensional and non-dimensional variables that can be confusing. Furthermore, they assume equal values for Hill coefficients and many parameters associated with activation and inhibition pathways, while setting inhibition intensity parameters to 1. These arbitrary choices raise concerns about the fidelity of the proposed model in representing the real system, as their selected values could potentially differ by many orders of magnitude from the actual parameters.

The definition of stability and its evaluation in the proposed pipeline might also be too narrow. Throughout the paper, the authors discuss the stability of the polarised pattern, checked by an exhaustive search of the parameter space where the system reaches a steady state with a polarised pattern instead of a homogeneous pattern. It is not clear if the stability is related to the linear stability analysis of the reaction terms, as conducted in Goehring et al. (Science, 2011), which could indicate if a homogeneous state exists and whether it is stable or unstable. The stability test is performed through a pipeline procedure where they always start from a polarised pattern described by their model and observe how it evolves over time. It is unclear if the conclusions depend on the chosen initial conditions. Particularly, it is unclear what would happen if the initial distribution of posterior molecules is not exactly symmetric with respect to the anterior molecules, or if the initial polarisation is not strong.

Regarding the biological interpretation and relevance of the model, it overlooks some important aspects of the C. elegans polarisation system. The authors focus solely on a reaction-diffusion formulation to reproduce the polarisation pattern. However, the polarisation of the C. elegans zygote consists of two distinct phases: establishment and maintenance, with actomyosin dynamics playing a crucial role in both phases (see Munro et al., Dev Cell 2004; Shivas & Skop, MBoC 2012; Liu et al., Dev Biol 2010; Wang et al., Nat Cell Biol 2017). Both myosin and actin are crucial to maintaining the localisation of PAR proteins during cell polarisation, yet the authors neglect cortical flows during the establishment phase and any effects driven by myosin and actin in their model, failing to capture the system's complexity. How this affects the proposed model and conclusions about the establishment of the polarisation pattern needs careful discussion. Additionally, they assume that diffusion in the cytoplasm is infinitely fast and that cytoplasmic flows do not play any role in cell polarity. Finite cytoplasmic diffusion combined with cytoplasmic flows could compromise the stability of the anterior-posterior molecular distributions. The authors claim that cytoplasmic diffusion coefficients are two orders of magnitude higher than membrane diffusion coefficients, but they seem to differ by only one order of magnitude (Petrášek et al., Biophys. J. 2008). The strength of cytoplasmic flows has been quantified by a few studies, including Cheeks et al., and Curr Biol 2004.

Although the authors compare their model predictions to experimental observations, particularly in reproducing mutant behaviours, they do not explicitly show or discuss these comparisons in detail. Diffusion coefficients and off-rates for some PAR proteins have been measured (Goehring et al., JCB 2011), but the authors seem to use parameter values that differ by many orders of magnitude, perhaps due to applied scaling. To ensure meaningful predictions, whether their proposed model captures the extensive published data should be evaluated. Various cellular/genetic perturbations have been studied to understand their effects on anterior-posterior boundary positioning. Testing these perturbations' responses in the model would be important. For example, comparing the intensity distribution of PAR-6 and PAR-2 with measurements during the maintenance phase by Goehring et al., JCB 2011, or comparing the normalised intensity of PAR-3 and PKC-3 from the model with those measured by Wang et al., Nat Cell Biol 2017, during establishment and maintenance phases (in both wild-type and cdc-42 (RNAi) zygotes) could provide insightful validation. Additionally, in the presence of active CDC-42, it has been observed that PAR-6 extends further into the posterior side (Aceto et al., Dev Biol 2006). Conducting such validation tests is essential to convince readers that the model accurately represents the actual system and provides insights into pattern formation during cell polarisation.

A clear justification, with references, for each network interaction between nodes in the five-node model is needed. Some of the activatory/inhibitory signals proposed by the authors have not been demonstrated (e.g. CDC-42 directly inhibiting CHIN-1). Table S2 provided by the authors is insufficient to justify each node-node interaction, requiring additional explanations. (See the review by Gubieda et al., Phil. Trans. R. Soc. B 2020, for a similar node network that differs from the authors' model.) Additionally, the intensity distributions of cortical PAR-3 and PKC-3 seem to vary significantly during both establishment and maintenance phases (Wang et al., Nat Cell Biol 2017), yet the authors consider the PAR-3/PAR-6/PKC-3 as a single complex. The choices in the model should be justified, as the presence or absence of clustering of these PAR proteins can be crucial during cell polarisation (Wang et al., Nat Cell Biol 2017; Dawes & Munro, Biophys J 2011).

In summary, the authors successfully demonstrate the importance of compensatory actions in maintaining polarisation robustness. Their computational pipeline offers valuable insights into the dynamics of reaction-diffusion networks. However, the lack of detailed experimental validation and realistic parameter estimation limits the model's applicability to real biological systems. While the study provides a solid foundation, further work is needed to fully characterise and validate the model in natural contexts. This work has the potential to significantly impact the field by providing a new perspective on the robustness of cell polarisation networks.

The computational pipeline developed could be a valuable tool for further in silico experiments, allowing researchers to explore the dynamics of more complex networks. To maximise its utility, the model needs comprehensive validation and refinement to ensure it accurately represents biological systems. Addressing these limitations, particularly the need for more detailed experimental validation and realistic parameter choices, will enhance the model's predictive power and its applicability to understanding cell polarisation in natural systems.

eLife assessment

This valuable work advances our understanding of the foraging behaviour of aerial insectivorous birds. Its major strength is the large volume of tracking data and the accuracy of those data. However, the evidence supporting the main claim of optimal foraging is incomplete.

Reviewer #1 (Public Review):

This study tests whether Little Swifts exhibit optimal foraging, which the data seem to indicate is the case. This is unsurprising as most animals would be expected to optimize the energy income:expenditure ratio; however, it hasn't been explicitly quantified before the way it was in this manuscript.

The major strength of this work is the sheer volume of tracking data and the accuracy of those data. The ATLAS tracking system really enhanced this study and allowed for pinpoint monitoring of the tracked birds. These data could be used to ask and answer many questions beyond just the one tested here.

The major weakness of this work lies in the sampling of insect prey abundance at a single point on the landscape, 6.5 km from the colony. This sampling then requires the authors to work under the assumption that prey abundance is simultaneously even across the study region - an assumption that is certainly untrue. The authors recognize this problem and say that sampling in a spatially explicit way was beyond their scope, which I understand, but then at other times try to present this assumption as not being a problem, which it very much is. Further, it is uncertain whether other aspects of the prey data are problematic. For example, the radar only samples insects at 50 m or higher from the ground - how often do Little Swifts forage under 50 m high? Another example might be that the phrases "high abundance" and "low abundance" are often used in the manuscript, but never defined.

It may be fair to say that prey populations might be correlated over space but are not equal. It is this unknown degree of spatial correlation that lends confidence to the findings in the Results. As such, the finding that Little Swifts forage optimally is indeed supported by the data, notwithstanding some of the shortcomings in the prey abundance data. The authors achieved their aims and the results support their conclusions.

At its centre, this work adds to our understanding of Little Swift foraging and extends to a greater understanding of aerial insectivores in general. While unsurprising that Little Swifts act as optimal foragers, it is good to have quantified this and show that the population declines observed in so many aerial insectivores are not necessarily a function of inflexible foraging habits. Further, the methods used in this research have great potential for other work. For example, the ATLAS system poses some real advantages and an exciting challenge to existing systems, like MOTUS. The radar that was used to quantify prey abundance also presents exciting possibilities if multiple units could be deployed to get a more spatially-explicit view.

To improve the context of this work, it is worth noting that the authors suggest that this work is important because it has never been done before for an aerial insectivore; however, that justification is untrue as it has been assessed in several flycatcher and swallow species. A further justification is that this research is needed due to dramatic insect population declines, but the magnitude and extent of such declines are fiercely debated in the literature. Perhaps these justifications are unnecessary, and the work can more simply be couched as just a test of optimality theory.

Reviewer #2 (Public Review):

Summary:

Bloch et al. investigate the relationships between aerial foragers (little swifts) tracked with an automated radio-telemetry system (Atlas) and their prey (flying insects) monitored with a small-scale vertical-looking radar device (BirdScan MR1). The aim of the study was to test whether little swifts optimise their foraging with the abundance of their prey. However, the results provided little evidence of optimal foraging behaviour.

Strengths:

This study addresses fundamental knowledge gaps on the prey-predator dynamics in the airspace. It describes the coincidence between the abundance of flying insects and features derived from tracking individual swifts.

Weaknesses:

The article uses hypotheses broadly derived from optimal foraging theory, but mixes the form of natural selection: parental energetics, parental survival (predation risks), nestling foraging, and breeding success. Results are partly incoherent (e.g., "Thus, even when the birds foraged close to the colony under optimal conditions, the shorter traveling distance is not thought to not confer lower flight-related energetic expenditure because more return trips were made.", L285-287), and confounding factors (e.g., brooding vs. nestling phase) are ignored. Some limits are clearly recognised by the authors (L329 and ff). To illustrate potential confounding effects, the daily flight duration (Prediction 4) should decrease with prey abundance, but how far does the daily flight duration coincide with departure and arrival at sunrise and sunset (note that day length increases between March and May), respectively, and how much do parents vary in the duration of nest attendance during the day across chick ages? To conclude, insufficient analyses are performed to rigorously assess whether little swifts optimize their foraging.

Filters applied on tracking data are necessary but may strongly influence derived features based on maximum or mean values. Providing sensitivity tests or using features less dependent on extreme values may provide more robust results.

Radar insect monitoring is incomplete and strongly size-dependent. What is the favourite prey size of swifts? How does it match with BirdScan MR1 monitoring capability?

eLife assessment

This study investigates the role of Caspar (Casp), an orthologue of human Fas-associated factor-1, in regulating the number of primordial germ cells that form during Drosophila embryogenesis. The findings are important in that they reveal an additional pathway involved in germ cell specification and maintenance. The evidence supporting the conclusions is solid, as the authors identify Casp and its binding partner Transitional endoplasmic reticulum 94 (TER94) as factors that influence germ cell numbers. Minor changes to the title, text, and experimental design are recommended.

Reviewer #1 (Public Review):

Summary:

The authors were seeking to define the roles of the Drosophila caspar gene in embryonic development and primordial germ cell (PGC) formation. They demonstrate that PGC number, and the distribution of the germ cell determinant Oskar, change as a result of changes in caspar expression; reduction of caspar reduces PGC number and the domain of Oskar protein expression, while overexpression of caspar does the reverse. They also observe defects in syncytial nuclear divisions in embryos produced from caspar mutant mothers. Previous work from the same group demonstrated that Caspar protein interacts with two partners, TER94 and Vap33. In this paper, they show that maternal knockdown of TER94 results in embryonic lethality and some overlap of phenotypes with reduction of caspar, supporting the idea may work together in their developmental roles. The authors propose models for how Caspar might carry out its developmental functions. The most specific of these is that Caspar and its partners might regulate oskar mRNA stability by recruiting ubiquitin to the translational regulator Smaug.

Strengths:

The work identifies a new factor that is involved in PGC specification and points toward an additional pathway that may be involved in establishing and maintaining an appropriate distribution of Oskar at the posterior pole of the embryo. It also ties together earlier observations about the presence of TER94 in the pole plasm that have not heretofore been linked to a function.

Weaknesses:

(1) A PiggyBac insertion allele casp[c04227] is used throughout the paper and referred to as a loss-of-function allele (casp[lof]). However, this allele does not appear to act strictly as a loss-of-function. Figure 1E shows that some residual Casp protein is present in early embryos produced by casp[lof]/Df females, and this protein is presumably functional as the PiggyBac insertion does not affect the coding region. Also, Figures 1B and 1C show that the phenotypes of casp[lof] homozygotes and casp[lof]/Df are not the same; surprisingly, the homozygous phenotypes are more severe. These observations are unexplained and inconsistent with the insertion being simply a loss-of-function allele. Might there be a second-site mutation in casp[c04227]?

(2) TER94 knockdown phenotypes have been previously published (Zhang et al 2018 PMID 30012668), and their effects on embryonic viability and syncytial mitotic divisions were described there. This paper is inappropriately not cited, and the data in Figure 4 should be presented in the context of what has been published before.

(3) The peptide counts in the mass spectrometry experiment aimed at finding protein partners for Casp are extremely low, except for Casp itself and TER94. Peptide counts of 1-2 seem to me to be of questionable significance.

(4) The pole bud phenotypes from TER94 knockdown and casp mutant shown in Fig 5 appear to be quite different. These differences are unexplained and seem inconsistent with the model proposed that the two proteins work in a common pathway. Whole embryos should also be shown, as the TER94 KD phenotype could result from a more general dysmorphism.

(5) Figure 6 is not quantitative, lacking even a second control staining to check for intensity variation artifacts. Therefore it shows that the distribution of Oskar protein changes in the various genotypes, but not convincingly that the level of Oskar changes as the paper claims.

(6) The error bars are huge in the graphs in Figure 7H, I, and J, leading me to question whether these changes are statistically significant. Calculations of statistical significance are missing from these graphs and need to be added.

(7) There are many instances of fuzzy and confusing language when describing casp phenotypes. For example, on lines 211-212 it is stated that 'casp[lof] adults are only partially homozygous viable as ~70% embryos laid by the homozygous mutant females failed to hatch into larvae'. Isn't this more accurately described as 'casp[c04227] is a maternal-effect lethal allele with incomplete penetrance'? Another example is on line 1165, what exactly is a 'semi-vital function'?

Reviewer #2 (Public Review):

Summary:

This study investigated the role of the Caspar (Casp) gene, a Drosophila homolog of human Fas-associated factor-1. It revealed that maternal loss of Casp led to centrosomal and cytoskeletal abnormalities during nuclear cycles in Drosophila early embryogenesis, resulting in defective gastrulation. Moreover, Casp regulates PGC numbers, likely by regulating the levels of Smaug and then Oskar. They demonstrate that Casp protein levels are linearly correlated to the PGC number. The partner protein TER94, an ER protein, shows similar but slightly distinct phenotypes. Based on the deletion mutant analysis, TER94 seems functionally relevant for the observed Casp phenotype. Additionally, it is likely involved in regulating protein degradation during PGC specification.

Strengths:

The paper reveals an unexpected function of the maternally produced Casp gene, previously implicated in immune response regulation and NF-kB signaling inhibition, in nuclear division and PGC formation in early fly embryos. Experiments are properly conducted and strongly support the conclusion. The rescue experiment using deletion mutant form is particularly informative as it suggests the requirement of each domain function.

Weaknesses:

Functional relationships among molecules shown here (and other genes known to regulate these processes) are still unclear.

Reviewer #3 (Public Review):

Summary:

Das et al. discovered a maternal role for Caspar (Casp), the Drosophila orthologue of human Fas-associated factor-1 (FAF1), in embryonic development and germ cell formation. They find that Casp interacts with Transitional endoplasmic reticulum 94 (TER94). Loss of Casp or TER94 leads to partial embryonic lethality, correlated with aberrant centrosome behavior and cytoskeletal abnormalities. This suggests that Casp, along with TER94, promotes embryonic development through a still unidentified mechanism. They also find that Casp regulates germ cell number by controlling a key determinant of germ cell formation, Oskar, through its negative regulator, Smaug.

Strengths:

Overall, the experiments are well-conducted, and the conclusions of this paper are mostly well-supported by data.

Weaknesses:

Some additional controls could be included, and the language could be clarified for accuracy.

eLife assessment

This important study investigates the influence of the cingulate cortex on the development of the social vocalizations of marmoset monkeys by making bilateral lesions of this brain area in neonatal animals. The evidence supporting the authors' claims is solid, although including longer-term effects and different social contexts would strengthen the manuscript. The work will be of broad interest to cognitive neuroscientists, speech and language researchers, and primate neuroscientists.

Reviewer #1 (Public Review):

Summary:

This study seeks to quantify changes in vocal behavior during development in marmosets with bilateral anterior cingulate cortex (ACC) lesions. The ACC and its role in social vocal behaviors are of great interest given previous literature on its involvement in the initiation of vocalizations, processing emotional content, and its connectivity to two other critical nodes in the vocal network, the amygdala and the PAG. The authors seek to test the hypothesis that the ACC contributes to the development of mature vocal behaviors during the first few weeks of life by disrupting this process with neonatal ACC lesions. Imaging and histological analyses confirm the extent of the lesion and suggest downstream effects in connected regions. Analysis of call rates and call type proportions show no or slight differences between lesioned and controlled animals. Additional analyses on the proportion of grouped 'social' calls and certain acoustic features of a particular call, the phee, reveal more distinct differences between the groups.

Strengths:

The authors have identified that ACC lesions in early life have no or little influence on certain aspects of vocal behavior (e.g. call rate, call intervals) but larger impacts on other aspects (e.g. acoustic features of phee calls). This data is a valuable addition to the literature on the effects of the ACC on vocal production.

The histological methods and resulting quantification of neural changes in the lesioned area and in downstream areas of interest are intriguing given the large time gap between the lesion and these analyses.

Weaknesses:

The article emphasizes vocal social behavior but none of the experiments involve a social element. Marmosets are recorded in isolation which could be sufficient for examining the development of vocal behavior in that particular context. However, the early-life maturation of vocal behavior is strongly influenced by social interactions with conspecifics. For example, the transition of cries and subharmonic phees which are high-entropy calls to more low-entropy mature phees is affected by social reinforcement from the parents. And this effect extends cross-context where differences in these interaction patterns extend to vocal behavior when the marmosets are alone. From the chord diagrams, cries still consist of a significant proportion of call types in lesioned animals. Additionally, though it is an intriguing finding that the infants' phee calls have acoustic differences being 'blunted of variation, less diverse and more regular,' the suggestion that the social message conveyed by these infants was 'deficient, limited, and/or indiscriminate' is not but can be tested with, for example, playback experiments.

The manuscript would benefit from the addition of more details to be able to better determine if the conclusions are well supported by the data. Understanding that this is very difficult data to get, the number of marmosets and some variability in the collection of the data would allow for the plotting of each individual across figures. For example, in the behavioral figures, which is the marmoset that is in the behavioral data that has a sparing of the ACC lesion in one hemisphere? Certain figures, described below in the recommendations for the authors, could also do with additional description.

Reviewer #2 (Public Review):

Summary:

Nagarajan et al. investigate the role of the anterior cingulate cortex (ACC) in vocal development of infant marmoset monkeys using lesions in this brain area. Many previous studies show that ACC plays an important role in volitional and emotion-driven vocal behavior in mammals. The experiments Nagarajan et al. performed strengthen the long-standing hypothesis that ACC influences the development of social-vocal behavior in non-human primates. Furthermore, their anatomical studies support the idea of cortical structures exerting cognitive control over subcortical networks for innate vocalization, and thus, enabling mammals to perform flexible social-vocal communication.

Strengths:

Many invasive behavioral studies in monkeys often times use 2-3 animals. The authors used a sufficiently high number of animals for their experiments. This increases the power of their conclusions.<br /> The study also investigates the impact of ACC lesions on downstream areas important for innate vocal production. This adds further evidence to the role of ACC in influencing these subcortical regions during vocal development and vocal behavior in general.

Weaknesses:

The authors state that the integrity of white matter tracts at the injection site was impacted but do not show data.

The study only provides data up to the 6th week after birth. Given the plasticity of the cortex, it would be interesting to see if these impairments in vocal behavior persist throughout adulthood or if the lesioned marmosets will recover their social-vocal behavior compared to the control animals.

Even though this study focuses entirely on the development of social vocalizations, providing data about altered social non-vocal behaviors that accompany ACC lesions is missing. This data can provide further insights and generate new hypotheses about the exact role of ACC in social-vocal development. For example, do these marmosets behave differently towards their conspecifics or family members and vice versa, and is this an alternate cause for the observed changes in social-vocal development?

Reviewer #3 (Public Review):

Summary:

In this manuscript, Nagarajan et al. study the impact of early damage to the anterior cingulate cortex (ACC) on the vocal development of marmoset monkeys. AAC lesions were performed on neonatal marmosets and their vocal patterns and the spectrotemporal features of their calls were analyzed compared to control groups during the first six weeks of life. While the vocal repertoire was not significantly affected by ACC lesions, the authors described notable differences in the social contact call, the phee call. Marmosets with ACC damage made fewer social contact calls, and when they did, these calls were shorter, louder, and monotonic. Additionally, the study revealed that ACC damage in infancy led to permanent alterations in downstream brain areas involved in social vocalizations, such as the amygdala and periaqueductal gray.

Strengths:

This study suggests that the ACC plays a crucial role in the normal development of social vocal behavior in infant marmosets. Studying vocal behavior in marmosets can provide insights into the neural mechanisms underlying human speech and communication disorders due to their similarity in brain structure and social behavior.

The methods are robust and reliable with precise localization of the lesions with neuroimaging and histological examination.

Weaknesses:

It is striking to find that the vocal repertoire of infant marmosets was not significantly affected by ACC lesions. During development, the neural circuits are still maturing and the role of different brain regions may evolve over time. While the ACC likely contributes to vocalizations across the lifespan, its relative importance may vary depending on the developmental stage. In neonates, vocalizations may be more reflexive or driven by physiological needs. At this stage, the ACC may play a role in basic socioemotional regulation but may not be as critical for vocal production. Since the animals lived for two years, further analysis might be helpful to elucidate the precise role of ACC in the vocal behavior of marmosets.

- Figure 3D. According to the Introduction "...infant ACC lesions abolish the characteristic cries that infants normally issue when separated from its mother". Are the present results in marmosets showing the opposite effect? Please discuss.

- Figure 3E and Discussion. Phees are mature contact calls and cries immature contact calls (Zhang et al, 2019, Nat Commun). Therefore, I would rather say that the proportion of immature (cries) contact calls increases vs the mature (phee, trill, twitters) contact calls in the ACC group. Cries are also "isolated-induced contact calls" to attract the attention of the caregivers.

- Figure 4D. Animal location and head direction within the recording incubator can have significant effects on the perceived amplitude of a call. Were these factors taken into account?

- Figure 4E. When a phee call has a higher amplitude, as is the case for the ACC group (Figure 4D), the energy of the signal will be concentrated more strongly at the phee call frequency ~8KHz. This concentration of the energy reduces the variability in the frequency distribution, leading to lower entropy. The interpretation of the results should be reconsidered. A faint call (control group) can exhibit more variability in the frequency content since the energy is distributed across a wider range of frequencies contributing to higher entropy. It can still be "fixed, regular, and stereotyped" if the behavior is consistent or predictable with little variation. Also, to define ACC calls as "monotonic" I would rather search for the lack of frequency modulation, amplitude variation, or narrower bandwidth.

- Apart from the changes in the vocal behavior, did the AAC lesions manifest in any other observable cognitive, emotional, or social behavior? ACC plays a role in processing pain and modulating pain perception. Could that be the reason for the observed increase in the proportion of cries in the ACC group and the increase in the phee call amplitude? Did the cries in the ACC group also display a higher amplitude than the cries in the control group?

- Discussion. Louder calls have the potential to travel longer distances compared to fainter calls, possess higher energy levels, and can propagate through the environment more effectively. If the ACC group produced louder phee syllables, how could be the message conveyed over long distances "deficient, limited, and/or indiscriminate"?

eLife assessment

This study provides an important re-evaluation of modality-specific information processing in the thalamus of trained mice. Using an elegant task design that probes competing tactile and visual stimuli, the authors present convincing evidence that behavioral training reshapes the sensitivity of higher-order thalamic nuclei. Despite the innovative methods and significant findings, the conclusions would be strengthened by deeper analyses of the sensory and non-sensory aspects of the modulation of the higher-order thalamic nuclei.

Reviewer #1 (Public Review):

Petty and Bruno investigate how response characteristics in the higher-order thalamic nuclei POm (typically somatosensory) and LP (typically visual) change when a stimulus (whisker air puff or visual drifting grating) of one or the other modality is conditioned to a reward. Using a two-step training procedure, they developed an elegant paradigm, where the distractor stimulus is completely uninformative about the reward, which is reflected in the licking behavior of trained mice. While the animals seem to take on to the tactile stimulus more readily, they can also associate the reward with the visual stimulus, ignoring tactile stimuli. In trained mice, the authors recorded single-unit responses in both POm and LP while presenting the same stimuli. The authors first focused on POm recordings, finding that in animals with tactile conditioning POm units specifically responded to the air puff stimulus but not the visual grating. Unexpectedly, in visually conditioned animals, POm units also responded to the visual grating, suggesting that the responses are not modality-specific but more related to behavioral relevance. These effects seem not not be homogeneously distributed across POm, whereas lateral units maintain tactile specificity and medial units respond more flexibly. The authors further ask if the unexpected cross-modal responses might result from behavioral activity signatures. By regressing behavior-coupled activity out of the responses, they show that late activity indeed can be related to whisking, licking, and pupil size measures. However, cross-modal short latency responses are not clearly related to animal behavior. Finally, LP neurons also seem to change their modality-specificity dependent on conditioning, whereas tactile responses are attenuated in LP if the animal is conditioned to visual stimuli.

The authors make a compelling case that POm neurons are less modality-specific than typically assumed. The training paradigm, employed methods, and analyses are mostly to the point, well supporting the conclusions. The findings importantly widen our understanding of higher-order thalamus processing features with the flexibility to encode multiple modalities and behavioral relevance. The results raise many important questions on the brain-wide representation of conditioned stimuli. E.g. how specific are the responses to the conditioned stimuli? Are thalamic cross-modal neurons recruited for the specific conditioned stimulus or do their responses reflect a more global shift of attention from one modality to another?

To elaborate on higher-order thalamic activity in relationship to conditioned behavior, a trial-by-trial analysis would be very useful. Is neuronal activity predictive of licking and at which relative timing? Furthermore, I wonder why the (in my mind) major and from the data obvious take-away, "POm neurons respond more strongly to visual stimuli if visually conditioned", is not directly tested in the summary statistics in Figure 3h.

The remaining early visual responses in POm in visually conditioned mice after removing behavior-linked activity are very convincing (Figure 5d). It would help, however, to see a representation of this on a single-neuron basis side-by-side. Are individual neurons just coupled to behavior while others are independent, or is behaviorally coupled activity a homogeneous effect on all neurons on top of sensory activity?

The conclusions on flexible response characteristics in LP in general are less strongly supported than those in POm. First, the differentiation between POm and LP relies heavily on the histological alignment of labeled probe depth and recording channel, possibly allowing for wrong assignment. furthermore, it seems surprising, but is not discussed, that putative LP neurons have such strong responses to the air puff stimuli, in both conditioning cases. In tactile conditioning, LP air puff responses seem to be even faster and stronger than POm. In visual conditioning, drifting grating responses paradoxically seem to be later than in tactile conditioning (Fig S2e). These differences in response changes between POm and LP should be discussed in more detail and statements of "similar phenomena" in POm and LP (abstract) should be qualified.

Reviewer #2 (Public Review):

Summary

This manuscript by Petty and Bruno delves into the still poorly understood role of higher-order thalamic nuclei in the encoding of sensory information by examining the activity in the Pom and LP cells in mice performing an associative learning task. They developed an elegant paradigm in which they conditioned head-fixed mice to attend to a stimulus of one sensory modality (visual or tactile) and ignore a second stimulus of the other modality. They recorded simultaneously from POm and LP, using 64-channel electrode arrays, to reveal the context-dependency of the firing activity of cells in higher-order thalamic nuclei. They concluded that behavioral training reshapes activity in these secondary thalamic nuclei. I have no major concerns with the manuscript's conclusions, but some important methodological details are lacking and I feel the manuscript could be improved with the following revisions.

Strengths

The authors developed an original and elegant paradigm in which they conditioned head-fixed mice to attend to a stimulus of one sensory modality, either visual or tactile, and ignore a second stimulus of the other modality. As a tactile stimulus, they applied gentle air puffs on the distal part of the vibrissae, ensuring that the stimulus was innocuous and therefore none aversive which is crucial in their study.

It is commonly viewed that the first-order thalamus performs filtering and re-encoding of the sensory flow; in contrast, the computations taking place in high-order nuclei are poorly understood. They may contribute to cognitive functions. By integrating top-down control, high-order nuclei may participate in generating updated models of the environment based on sensory activity; how this can take place is a key question that Petty and Bruno addressed in the present study.

Weaknesses

(1) Overall, methods, results, and discussion, involving sensory responses, especially for the Pom, are confusing. I have the feeling that throughout the manuscript, the authors are dealing with the sensory and non-sensory aspects of the modulation of the firing activity in the Pom and LP, without a clear definition of what they examined. Making subsections in the results, or a better naming of what is analyzed could convey the authors' message in a clearer way, e.g., baseline, stim-on, reward.

In line #502 in Methods, the authors defined "Sensory Responses. We examined each cell's putative sensory response by comparing its firing rate during a "stimulus period" to its baseline firing rate. We first excluded overlapping stimuli, defined as any stimulus occurring within 6 seconds of a stimulus of a different type. We then counted the number of spikes that occurred within 1 second prior to the onset of each stimulus (baseline period) and within one second of the stimulus onset (stimulus period). The period within +/-50ms of the stimulus was considered ambiguous and excluded from analysis."

Considering that the responses to whisker deflection, while weak and delayed, were shown to occur, when present, before 50 ms in the Pom (Diamond et al., 1992), it is not clear what the authors mean and consider as "Sensory Responses"?

Precise wording may help to clarify the message. For instance, line #134: "Of cells from tactilely conditioned mice, 175 (50.4%) significantly responded to the air puff, as defined by having a firing rate significantly different from baseline within one second from air puff onset (Figure 3d, bottom)", could be written "significantly responded to the air puff" should be written "significantly increased (or modified if some decreased) their firing rate within one second after the air puff onset (baseline: ...)". This will avoid any confusion with the sensory responses per se.

(2) To extend the previous concern, the latency of the modulation of the firing rate of the Pom cells for each modality and each conditioning may be an issue. This latency, given in Figure S2, is rather long, i.e. particularly late latencies for the whisker system, which is completely in favor of non-sensory "responses" per se and the authors' hypothesis that sensory-, arousal-, and movement-evoked activity in Pom are shaped by associative learning. Latency is a key point in this study.

Therefore,<br /> - latencies should be given in the main text, and Figure S2 could be considered for a main figure, at least panels c, d, and e, could be part of Figure 3.

- the Figure S2b points out rather short latency responses to the air puff, at least in some cells, in addition to late ones. The manuscript would highly benefit from an analysis of both early and late latency components of the "responses" to air puffs and drafting grating in both conditions. This analysis may definitely help to clarify the authors' message. Since the authors performed unit recordings, these data are accessible.

- it would be highly instructive to examine the latency of the modulation of Pom cells firing rate in parallel with the onset of each behavior, i.e. modification of pupil radius, whisking amplitude, lick rate (Figures 1e, g and 3a, b). The Figure 1 does not provide the latency of the licks in conditioned mice.

- the authors mention in the discussion low-latency responses, e.g., line #299: "In both tactilely and visually conditioned mice, movement could not explain the increased firing rate at air puff onset. These low-latency responses across conditioning groups is likely due in part to "true" sensory responses driven by S1 and SpVi."; line #306: "Like POm, LP displayed varied stimulus-evoked activity that was heavily dependent on conditioning. LP responded to the air puff robustly and with low latency, despite lacking direct somatosensory inputs."<br /> But which low-latency responses do the authors refer to? Again, this points out that a robust analysis of these latencies is missing in the manuscript but would be helpful to conclude.

(3) Anatomical locations of recordings in the dorsal part of the thalamus. Line #122 "Our recordings covered most of the volume of POm but were clustered primarily in the anterior and medial portions of LP (Figure 2d-f). Cells that were within 50 µm of a region border were excluded from analysis."<br /> How did the authors distinguish the anterior boundary of the LP with the LD nucleus just more anterior to the LP, another higher-order nucleus, where whisker-responsive cells have been isolated (Bezdudnaya and Keller, 2008)?

(4) The mention in the Methods about the approval by an ethics committee is missing.<br /> All the surgery (line #381), i.e., for the implant, the craniotomy, as well as the perfusion, are performed under isoflurane. But isoflurane induces narcosis only and not proper anesthesia. The mention of the use of analgesia is missing.

Reviewer #3 (Public Review):

Petty and Bruno ask whether activity in secondary thalamic nuclei depends on the behavioral relevance of stimulus modality. They recorded from POm and LP, but the weight of the paper is skewed toward POm. They use two cohorts of mice (N=11 and 12), recorded in both nuclei using multi-electrode arrays, while being trained to lick to either a tactile stimulus (air puff against whiskers, first cohort) or a visual stimulus (drifting grating, second cohort), and ignore the respective other. They find that both nuclei, while primarily responsive to their 'home' modality, are more responsive to the relevant modality (i.e. the modality predicting reward).

Strengths:

The paper asks an important question, it is timely and is very well executed. The behavioral method using a delayed lick index (excluding impulsive responses) is well worked out. Electrophysiology methods are state-of-the-art with information about spike quality in Figure S1. The main result is novel and important, convincingly conveying the point that encoding of secondary thalamic nuclei is flexible and clearly includes aspects of the behavioral relevance of a stimulus. The paper explores the mapping of responses within POm, pointing to a complex functional structure, something that has been reported/suggested in earlier studies.

Weaknesses:

Coding: It does not become clear to which aspect of the task POm/LP is responding. There is a motor-related response (whisking, licking, pupil), which, however, after regressing it out leaves a remaining response that the authors speculate could be sensory.

Learning: The paper talks a lot about 'learning', although it is only indirectly addressed. The authors use two differently (over-)trained mice cohorts rather than studying e.g. a rule switch in one and the same mouse, which would allow us to directly assess whether it is the same neurons that undergo rule-dependent encoding.

Mapping: The authors treat and interpret the two nuclei very much in the same vein, although there are clear differences. I would think these differences are mentioned in passing but could be discussed in more depth. Mapping using responses on electrode tracks is done in POm but not LP.

eLife assessment

This study presents convincing evidence of the role of an intestine-released neuropeptide, FLP-2, in the oxidative stress response of C. elegans, as well as for the neural circuit pathway that regulates its release in response to sensing reactive oxygen species (i.e., H2O2). These valuable results advance the understanding of gut-brain signaling and the neural circuit basis of behavioral responses to stress.

Reviewer #1 (Public Review):

Summary:

The main goal of the paper was to identify signals that activate FLP-1 release from AIY neurons in response to H2O2, previously shown by the authors to be an important oxidative stress response in the worm.

Strengths:

This study builds upon the authors' previous work (Jia and Sieburth 2021) by further elucidating the gut-derived signaling mechanisms that coordinate the organism-wide antioxidant stress response in C. elegans.

By detailing how environmental cues like oxidative stress are transduced into gut-derived peptidergic signals, this study represents a valuable advancement in understanding the integrated physiological responses governed by the gut-brain axis.

This work provides valuable mechanistic insights into the gut-specific regulation of the FLP-2 peptide signal.

Weaknesses:

Although the authors identify intestinal FLP-2 as the endocrine signal important for regulating the secretion of the neuronal antioxidant neuropeptide, FLP-1, there is no effort made to identify how FLP-2 levels regulate FLP-1 secretion or identify whether this regulation is occurring directly through the AIY neuron or indirectly. This is brought up in the discussion, but identifying a target for FLP-2 in this pathway seems like a crucial missing piece of information in characterizing this pathway.

Reviewer #2 (Public Review):

Summary:<br /> The core findings demonstrate that the neuropeptide-like protein FLP-2, released from the intestine of C. elegans, is essential for activating the intestinal oxidative stress response. This process is mediated by endogenous hydrogen peroxide (H2O2), which is produced in the mitochondrial matrix by superoxide dismutases SOD-1 and SOD-3. H2O2 facilitates FLP-2 secretion through the activation of protein kinase C family member pkc-2 and the SNAP25 family member aex-4. The study further elucidates that FLP-2 signaling potentiates the release of the antioxidant FLP-1 neuropeptide from neurons, highlighting a bidirectional signaling mechanism between the intestine and the nervous system.

Strengths:

This study presents a significant contribution to the understanding of the gut-brain axis and its role in oxidative stress response and significantly advances our understanding of the intricate mechanisms underlying the gut-brain axis's role in oxidative stress response. By elucidating the role of FLP-2 and its regulation by H2O2, the study provides insights into the molecular basis of inter-tissue communication and antioxidant defense in C. elegans. These findings could have broader implications for understanding similar pathways in more complex organisms, potentially offering new targets for therapeutic intervention in diseases related to oxidative stress and aging.

Weaknesses:

（1）The experimental techniques employed in the study were somewhat simple and could benefit from the incorporation of more advanced methodologies.

（2）The weak identification of the key receptors mediating the interaction between FLP-2 and AIY neurons, as well as the receptors in the gut that respond to FLP-1.

（3）The study could be improved by incorporating a sensor for the direct measurement of hydrogen peroxide levels.

eLife assessment

This important study by Franziska Auer and colleagues examines cerebellar Purkinje cells' role in controlling posture in larval zebrafish using the innovative chemogenetic tool TRPV1/capsaicin. This work will interest neuroscientists studying motor control and cerebellar function. Overall, solid evidence is presented showing that disrupting Purkinje cell function impairs balance in the pitch axis and that this cell population encodes tilt direction. At the same time, some conclusions require more data or better statistical analysis.

Reviewer #1 (Public Review):

This study uses a variety of approaches to explore the role of the cerebellum, and in particular Purkinje cells (PCs), in the development of postural control in larval zebrafish. A chemogenetic approach is used to either ablate PCs or disrupt their normal activity and a powerful, high-throughput behavioural tracking system then enables quantitative assessment of swim kinematics. Using this strategy, convincing evidence is presented that PCs are required for normal postural control in the pitch axis. Calcium imaging further shows that PCs encode tilt direction. Evidence is also presented that suggests the role of the cerebellum changes over the course of early development, although this claim is rather less robust in the current version of the paper. Finally, the authors build on their prior work showing that both axial muscles and pectoral fins contribute to "climbs" and show evidence that suggests PCs are required for correct engagement of the fins during this behaviour. Overall, establishing a role for the cerebellum in postural control is not very surprising. However, a clear motivation of this study was to establish a robust experimental platform to investigate the changing role of cerebellar circuits in the development of postural control in the highly experimentally accessible zebrafish larvae, and in this regard, the authors have certainly succeeded.

Overall, I consider this an excellent paper, with some room for improvement in aspects of presentation, discussion, and some aspects of the data analysis..

Reviewer #2 (Public Review):

Summary:

Franziska Auer et al. investigate the role of cerebellar Purkinje cells in controlling posture in larval zebrafish using the chemogenetic tool TRPV1/capsaicin to bidirectionally manipulate (i.e., activate or ablate) these cells. This tool has been developed for zebrafish previously but has not been applied to Purkinje cells.

High-throughput behavioral experiments are presented to monitor how body posture is affected by these perturbations. The analysis of postural control focuses on a specific subaspect of posture: the body tilt-angle relative to horizontal just before a swim bout is executed, quantified separately for pre-ascent and pre-dive bouts. They report a broad bimodal distribution of pre-ascent bout posture ranging from -20 to +40 degrees, while the pre-dive bout posture was more Gaussian, ranging between -40 and 0 degrees. The treatment effect is quantified as the change in the median of these distributions.

Purkinje cell activation and ablation in 7 days post-fertilization (dpf) fish shifted the median of the ascending bout posture distributions to positive values. The authors hypothesize that the stochastic nature of the activation process might desynchronize Purkinje cell activity, thus abolishing Purkinje cells' role in postural control, similar to ablation. However, this does not explain why dive bout posture decreased upon activation but was unaffected by ablation.

To test whether the role of Purkinje cells in postural control matures over development, the authors repeated the ablation experiments at 14 dpf. They state that "at 14 dpf, the effects of Purkinje cell lesions on posture were more widespread than at 7 dpf." However, this effect size is comparable to that observed at 7 dpf, suggesting no further maturation of the role of Purkinje cells in pre-ascending bout postural control. The median pre-dive bout posture decreased at 14 dpf, contrasting with no effect at 7 dpf, yet this change was comparable in effect size to the activation effect on Purkinje cells at 7 dpf. The current data breadth may not be sufficient to conclude that signatures of emerging cerebellar control of posture across early development were uncovered.

The study's exploration of activating Purkinje cells in freely swimming fish using TRPV1/capsaicin is of special interest, but the practicability of this method is unclear from the current presentation. It would be beneficial to present the distribution of the percentage of activatable Purkinje cells across animals and time points to provide insight into the method's efficiency. Discussing this limitation and potential improvements would aid in evaluating the method, especially since the authors report that the activation experiments were labor-intensive, limiting repeat experiments. This may explain why the activation experiment at 7 dpf is the only data presented with cell activation, with other analyses performed using the cell ablation capabilities of the TRPV1/capsaicin method. Another data point at 14dpf would significantly strengthen the conclusions.

The authors analyze Purkinje cell-controlled fin-trunk coordination by examining ascending bout posture across different swim bout speeds. They make the important finding that pectoral fin movements contribute significant lift for median and fast swim bouts but not for slow ones, and that Purkinje cell ablation disrupts lift generation at all speeds.

Finally, the authors examined whether Purkinje cell activity encodes postural tilt-angle by performing calcium imaging on 31 cells from 8 fish using their Tilt In Place Microscope (TIPM). They report that they could decode the tilt-angle from individual neurons with a highly tuned response, and also from neurons that were not obviously tuned when pooling them and analyzing the population response. However, due to the non-simultaneous recordings across animals, definitive conclusions about population-level encoding should be made cautiously, it might be better to suggest potential population encoding that needs confirmation with more targeted experiments involving simultaneous recordings.

Strengths:

- The study introduces a novel application of the chemogenetic tool TRPV1/capsaicin to study cerebellar function in zebrafish.

- High-throughput behavioral experiments provide detailed analysis of postural control.

- The further investigation of Purkinje cell-controlled fin-trunk coordination offers new insights into motor control mechanisms.

- The use of calcium imaging to decode postural tilt-angle from Purkinje cell activity presents interesting preliminary results on neuronal population encoding.

Weaknesses:

- The term "disruption" for postural control effects may lead to misleading expectations.

- The supporting data show only subtle median shifts in postural angle, raising questions about the significance of observed effects. Statistical methods that account for the hierarchical structure of the data might be required to support the conclusions.

- The study's data breadth may not be sufficient to conclude emerging cerebellar postural control across early development.

- The current presentation does not adequately detail the practicability and efficiency of the TRPV1/capsaicin method for activating Purkinje cells, and the labor-intensive nature of these experiments constrains the ability to replicate and validate the findings.

- Non-simultaneous recordings in calcium imaging necessitate cautious interpretation of population-level encoding results.

Reviewer #3 (Public Review):

Summary:

This paper uses a new chemogenetic tool to investigate the role of cerebellar Purkinje cells in postural control. Using a high-throughput behavioral assay, they show that activation or ablation of Purkinje cells affects various aspects of postural control in zebrafish larvae during spontaneous swimming and that the effects are more pronounced at later developmental time points, where the Purkinje cell number is much greater. Using a sophisticated imaging assay, they record Purkinje cell activity in response to the tilt of the fish and show that some Purkinje cells are tuned to tilt direction and that the direction can even be decoded from untuned neurons.

Strengths:

Overall the study is nice, using a range of tools to address a fundamental question about the role of the cerebellum in postural control in fish.

Weaknesses:

(1) The data in Figure 1 that establishes the method seems to be based on a very small number of experiments and lacks some statistical analysis.

(2) The choice and presentation of the statistical and analysis methods used in Figures 2-5 could be improved.

eLife assessment

In this valuable study, Li et al., set out to understand the mechanisms of audiovisual temporal recalibration - the brain's ability to adjust to the latency differences that emerge due to different (distance-dependent) transduction latencies of auditory and visual signals - through psychophysical measurements and modelling. The analysis supports a role for causal inference in recalibration, though the evidence is incomplete.

Reviewer #1 (Public Review):

This study asks whether the phenomenon of crossmodal temporal recalibration, i.e. the adjustment of time perception by consistent temporal mismatches across the senses, can be explained by the concept of multisensory causal inference. In particular, they ask whether the explanation offered by causal inference better explains temporal recalibration better than a model assuming that crossmodal stimuli are always integrated, regardless of how discrepant they are.

The study is motivated by previous work in the spatial domain, where it has been shown consistently across studies that the use of crossmodal spatial information is explained by the concept of multisensory causal inference. It is also motivated by the observation that the behavioral data showcasing temporal recalibration feature nonlinearities that, by their nature, cannot be explained by a fixed integration model (sometimes also called mandatory fusion).

To probe this the authors implemented a sophisticated experiment that probed temporal recalibration in several sessions. They then fit the data using the two classes of candidate models and rely on model criteria to provide evidence for their conclusion. The study is sophisticated, conceptually and technically state-of-the-art, and theoretically grounded. The data clearly support the authors' conclusions.

I find the conceptual advance somewhat limited. First, by design, the fixed integration model cannot explain data with a nonlinear dependency on multisensory discrepancy, as already explained in many studies on spatial multisensory perception. Hence, it is not surprising that the causal inference model better fits the data. Second, and again similar to studies on spatial paradigms, the causal inference model fails to predict the behavioral data for large discrepancies. The model predictions in Figure 5 show the (expected) vanishing recalibration for large delta, while the behavioral data don't' decay to zero. Either the range of tested SOAs is too small to show that both the model and data converge to the same vanishing effect at large SOAs, or the model's formula is not the best for explaining the data. Again, the studies using spatial paradigms have the same problem, but in my view, this poses the most interesting question here.

In my view there is nothing generally wrong with the study, it does extend the 'known' to another type of paradigm. However, it covers little new ground on the conceptual side.

On that note, the small sample size of n=10 is likely not an issue, but still, it is on the very low end for this type of study.

Reviewer #2 (Public Review):

Summary:

Li et al.'s goal is to understand the mechanisms of audiovisual temporal recalibration. This is an interesting challenge that the brain readily solves in order to compensate for real-world latency differences in the time of arrival of audio/visual signals. To do this they perform a 3-phase recalibration experiment on 9 observers that involves a temporal order judgment (TOJ) pretest and posttest (in which observers are required to judge whether an auditory and visual stimulus were coincident, auditory leading or visual leading) and a conditioning phase in which participants are exposed to a sequence of AV stimuli with a particular temporal disparity. Participants are required to monitor both streams of information for infrequent oddballs, before being tested again in the TOJ, although this time there are 3 conditioning trials for every 1 TOJ trial. Like many previous studies, they demonstrate that conditioning stimuli shift the point of subjective simultaneity (pss) in the direction of the exposure sequence.

These shifts are modest - maxing out at around -50 ms for auditory leading sequences and slightly less than that for visual leading sequences. Similar effects are observed even for the longest offsets where it seems unlikely listeners would perceive the stimuli as synchronous (and therefore under a causal inference model you might intuitively expect no recalibration, and indeed simulations in Figure 5 seem to predict exactly that which isn't what most of their human observers did). Overall I think their data contribute evidence that a causal inference step is likely included within the process of recalibration.

Strengths:

The manuscript performs comprehensive testing over 9 days and 100s of trials and accompanies this with mathematical models to explain the data. The paper is reasonably clearly written and the data appear to support the conclusions.

Weaknesses:

While I believe the data contribute evidence that a causal inference step is likely included within the process of recalibration, this to my mind is not a mechanism but might be seen more as a logical checkpoint to determine whether whatever underlying neuronal mechanism actually instantiates the recalibration should be triggered.

The authors' causal inference model strongly predicts that there should be no recalibration for stimuli at 0.7 ms offset, yet only 3/9 participants appear to show this effect. They note that a significant difference in their design and that of others is the inclusion of longer lags, which are unlikely to originate from the same source, but don't offer any explanation for this key difference between their data and the predictions of a causal inference model.

I'm also not completely convinced that the causal inference model isn't 'best' simply because it has sufficient free parameters to capture the noise in the data. The tested models do not (I think) have equivalent complexity - the causal inference model fits best, but has more parameters with which to fit the data. Moreover, while it fits 'best', is it a good model? Figure S6 is useful in this regard but is not completely clear - are the red dots the actual data or the causal inference prediction? This suggests that it does fit the data very well, but is this based on predicting held-out data, or is it just that by having more parameters it can better capture the noise? Similarly, S7 is a potentially useful figure but it's not clear what is data and what are model predictions (what are the differences between each row for each participant; are they two different models or pre-test post-test or data and model prediction?!).

I'm not an expert on the implementation of such models but my reading of the supplemental methods is that the model is fit using all the data rather than fit and tested on held-out data. This seems problematic.

I would have liked to have seen more individual participant data (which is currently in the supplemental materials, albeit in a not very clear manner as discussed above).

The way that S3 is described in the text (line 141) makes it sound like everyone was in the same direction, however, it is clear that 2 /9 listeners show the opposite pattern, and 2 have confidence intervals close to zero (albeit on the -ve side).

Reviewer #3 (Public Review):

Summary:

Li et al. describe an audiovisual temporal recalibration experiment in which participants perform baseline sessions of ternary order judgments about audiovisual stimulus pairs with various stimulus-onset asynchronies (SOAs). These are followed by adaptation at several adapting SOAs (each on a different day), followed by post-adaptation sessions to assess changes in psychometric functions. The key novelty is the formal specification and application/fit of a causal-inference model for the perception of relative timing, providing simulated predictions for the complete set of psychometric functions both pre and post-adaptation.

Strengths:

(1) Formal models are preferable to vague theoretical statements about a process, and prior to this work, certain accounts of temporal recalibration (specifically those that do not rely on a population code) had only qualitative theoretical statements to explain how/why the magnitude of recalibration changes non-linearly with the stimulus-onset asynchrony of the adaptor.

(2) The experiment is appropriate, the methods are well described, and the average model prediction is a fairly good match to the average data (Figure 4). Conclusions may be overstated slightly, but seem to be essentially supported by the data and modelling.

(3) The work should be impactful. There seems a good chance that this will become the go-to modelling framework for those exploring non-population-code accounts of temporal recalibration (or comparing them with population-code accounts).

(4) A key issue for the generality of the model, specifically in terms of recalibration asymmetries reported by other authors that are inconsistent with those reported here, is properly acknowledged in the discussion.

Weaknesses:

(1) The evidence for the model comes in two forms. First, two trends in the data (non-linearity and asymmetry) are illustrated, and the model is shown to be capable of delivering patterns like these. Second, the model is compared, via AIC, to three other models. However, the main comparison models are clearly not going to fit the data very well, so the fact that the new model fits better does not seem all that compelling. I would suggest that the authors consider a comparison with the atheoretical model they use to first illustrate the data (in Figure 2). This model fits all sessions but with complete freedom to move the bias around (whereas the new model constrains the way bias changes via a principled account). The atheoretical model will obviously fit better, but will have many more free parameters, so a comparison via AIC/BIC or similar should be informative.

(2) It does not appear that some key comparisons have been subjected to appropriate inferential statistical tests. Specifically, lines 196-207 - presumably this is the mean (and SD or SE) change in AIC between models across the group of 9 observers. So are these differences actually significant, for example via t-test?

(3) The manuscript tends to gloss over the population-code account of temporal recalibration, which can already provide a quantitative account of how the magnitude of recalibration varies with adaptor SOA. This could be better acknowledged, and the features a population code may struggle with (asymmetry?) are considered.

(4) The engagement with relevant past literature seems a little thin. Firstly, papers that have applied causal inference modelling to judgments of relative timing are overlooked (see references below). There should be greater clarity regarding how the modelling here builds on or differs from these previous papers (most obviously in terms of additionally modelling the recalibration process, but other details may vary too). Secondly, there is no discussion of previous findings like that in Fujisaki et al.'s seminal work on recalibration, where the spatial overlap of the audio and visual events didn't seem to matter (although admittedly this was an N = 2 control experiment). This kind of finding would seem relevant to a causal inference account.

References:<br /> Magnotti JF, Ma WJ and Beauchamp MS (2013) Causal inference of asynchronous audiovisual speech. Front. Psychol. 4:798. doi: 10.3389/fpsyg.2013.00798<br /> Sato, Y. (2021). Comparing Bayesian models for simultaneity judgement with different causal assumptions. J. Math. Psychol., 102, 102521.

(5) As a minor point, the model relies on simulation, which may limit its take-up/application by others in the field.

(6) There is little in the way of reassurance regarding the model's identifiability and recoverability. The authors might for example consider some parameter recovery simulations or similar.

(7) I don't recall any statements about open science and the availability of code and data.

eLife assessment

This valuable study provides evidence that during learning of a simple detection task, the change in the rate of spike bursts is a signal that is distinct from the change in firing rate, and suggests that the change in bursting is more correlated with learning than other measures of change in activity. However, the evidence for the claim that bursting contributes to learning and attention is currently incomplete, because the authors did not take into account the potentially differential effects of learning-related changes in movement on bursting compared to non-burst spike events, and there is no meaningful way to measure attention in their task. Also, the study used an artificial microstimulation as the stimulus, which limits the generalization of these results to normal sensory-motor learning.

Reviewer #1 (Public Review):

Summary:

In this paper, the authors' study aimed to test existing theories on the role of bursting in learning and attention. They find evidence for both. It is not clear how these two can be reconciled, but this is one of the first studies to explicitly test recent theories of spike multiplexing in the brain. This will pave the way for future investigations, both experimental and theoretical.

Strengths:

(1) A key strength of this study is the fact that it aims to test existing theories of spike multiplexing, finding support for both attention-like and learning-like signals.

(2) The task setup is of particular interest to brain-machine interfaces, and how such setups trigger learning and attention mechanisms.

Weaknesses:

(1) The fact that the teaching signal is an (artificial) stimulation of the primary sensory cortex, makes it unclear how applicable are these results to a more general understanding of learning and attention in the brain.

(2) It would have been useful to more directly compare the results obtained here with existing burst-dependent computational models of learning and attention. This is particularly important since there appears to be an interaction between learning and sharpening signals.

(3) There are inherent limitations in our current ability to read out bursting and non-bursting signals, this is a brave first attempt, but at this point, it is unclear how can one robustly read out this information from noisy data.

Reviewer #2 (Public Review):

Naud et al investigate whether single spikes and bursts encode different information in behavior. To do this, they reanalyze juxtasomal recordings of deep-layer cortical neurons from behaving rats collected in two previous studies by Doron et al. Rats were trained (in a Go-NoGo design) to lick a spout for a water reward in response to electrical microstimulation of the primary somatosensory cortex, which rats quickly learn to do in a single day. Juxtasomal recordings near the site of micro stimuli are then divided up into single spikes ("events") versus high-frequency bursts ("bursts"). Training results in the appearance of bursts, which do not seem to correlate with the rate of events, suggesting that bursts and events carry different information. While the fraction of bursts is elevated during Hit trials, errors appear to uniquely trigger additional bursts. The distribution of burst times appears to shift from long after the stimulus (early in training) to shortly after the stimulus (later in training). Bursts of layer 5 pyramidal neurons in particular are associated with apical tuft activity that could enhance plasticity. The observed increased bursting is therefore suggestive of a potential mechanism by which errors engage plasticity.

This paper has substantial strengths: the experiments appear to be well performed, the dataset is substantial, and the questions and phenomena are interesting.

The exclusion of fast-spike (inhibitory) data, which the experiments seem to have generated, is a weakness as these data could have provided an important control. If the bursts here reflect apical dendrite activity, the same phenomena might be absent in inhibitory cells as they lack apical tufts.

Another weakness is the need to better control movement, which could be an alternative explanation to the top-down modulation of apicals that the authors suspect. For example, the bursts on error trials could be due to the animals moving more when an error occurs. Layer 5 of the somatosensory cortex has increased activity during whisking or body movements. If the mouse fidgets out of frustration that the reward has not occurred or whisks more, bursts are highly likely due to less exotic purely bottom-up inputs.

Reviewer #3 (Public Review):

Summary:

The burst fraction neural code has conceptual interest but has been little examined in vivo. This study examines and compares the burst fraction, the standard firing rate (firing rate) code, and the related event fraction (event rate) code using published data from an experiment where rats learned to lick after detecting electrical microstimulation in the somatosensory (barrel) cortex. Analyzing single-neuron spiking responses, the study reports that the burst fraction identifies more and different neurons showing the effects of training than the firing rate. The study further claims that the burst fraction (1) most promptly responded to false-negative detection errors, (2) during further training of trained animals (from 80% to 90% accuracy, over five days), correlates with behavioral accuracy, and (3) by shifting earlier to align with the (relatively constant) event rate modulation, leads to the observed sharpened firing rate response during this further training. The study concludes that 'a fine-grained separation of spike timing patterns [into burst fraction, firing rate, and event rate] reveals two signals,' an error signal and a sharpening signal.

Strengths:

The burst fraction is shown to discern more (and somewhat different) cells showing significant responses in trained animals and also to reveal a larger absolute difference in the fraction of responsive cells between naïve and trained animals. The Poisson model analysis particularly convincingly shows that the firing rate alone cannot explain either the spiking pattern or the prevalence of burst fraction-ON cells, thereby furnishing strong evidence that the burst fraction conveys independent information from the firing rate. The demonstration of error signals on miss trials in all three neural codes (burst fraction, firing rate, event rate) is interesting. It is also interesting to see that neural responses broadly shift earlier for animals even during further training in an already 'expert' stage and that the burst fraction correlates with further accuracy increases.

Weaknesses:

The evidence is inadequate for the burst fraction as responding more promptly to missed trials.

This key claim seems to rest solely on the timing of the first bins in Figure 3B showing statistically significant differences. This reasoning implicitly draws inferences from the lack of statistical differences, which cannot support a positive claim in general. Specifically, here, the burst fraction is calculated with a division, which can magnify small differences and impact the power of statistical tests. If I trace back from the first bin showing significant differences to the first bin the signal starts rising, the timing seems to be comparable for all three neural codes (~1.6 s).

Pertinently, what is the statistical test used in Figure 3B? A parametric test may be inappropriate for the burst fraction, a ratio that like does not fulfill the normality assumption. An inappropriate test would compound the problem of concluding from the lack of (early) significant differences.

The evidence that burst fraction is responsible for sharpening is opaque due to insufficient statistical reporting. Specifically, it seems there is a correlation between firing rate and accuracy that is reported as non-significant.

Changes in the reaction times (or other movement parameters) over-training may confound the correlation of the burst fraction to the accuracy and firing rate sharpening during further training. Lack of control for changes in movement over training weakens the results.

The claim of independence of burst fraction and event rate/firing rate information is too strong. The authors show a significant negative correlation between burst fraction and firing rate (2D).

The claim that there is no 'functional reorganization' beyond day two is too strong. Although this claim is not a core one to the study, it derives from an absence of statistical significance, especially problematic here as the effect sizes are large. For example, the Spearman correlation is 0.67/0.87 for the analyses with burst fraction. With only five data points, even strong effects may not achieve statistical significance, making negative conclusions problematic. Further, how are the p-values calculated (if using a parametric test, are the assumptions met), and why should these analyses use Spearman's correlation when analogous analyses in Figure 4E, F use Pearson's r?

Does the burst fraction correlate with accuracy in cross-training?

If the burst fraction correlates with accuracy, it should be expected to do so also when the animals progress from the naïve to the trained stage. Moreover, the correlation in Figure 4E can benefit from strengthening as it is now supported by only five points, is driven by only three 'clusters,' and only represents a narrow range of accuracies. If the data is available for this analysis, it should be done to test and potentially strengthen the main claim of the study.

The text and figures contain numerous ambiguities that need to be clarified. These do not include obvious typos, only elements that affect conceptual understanding.

- Some key terms in the main claims are never defined. For example, in the title, it is unclear what 'fast' and 'transients' mean. The abstract uses, but the main text never defines, 'demultiplexing,' 'a *conjunctive* burst code,' 'sparse and succinct [sic],' and 'correlated more *globally*.'

- Some paper components are un(der)explained and, sometimes, apparently internally inconsistent. For example, in Figure 1I, the fraction of firing rate-ON cells does not look like the 6% shown in Figure 1J, left. In Figure 2E-G, what is the total cell number, 279, in Figure 2G legend, why is it different from the 153 total cells in Figure 2E legend, and what is the 'n = 5' within Figure 2G? All n numbers should be explained in general; more examples include the 245 in Figure 3C and the 49 in Figure 3B. In Figure 3C, what is the top horizontal bar (I assume significant differences)? About catch trials, the Figure 3D legend says rewards are given on licks, but the text says licking was not rewarded; which is the case? Figure 4B legend says 'firing rate (left), burst fraction (middle) and event rate (right),' but the plot colors imply a different order.

- The abstract states, 'The alignment of bursting and event rate modulation [...] was strongly associated [sic] behavioral accuracy.' It seems to me it is not the alignment of burst fraction and event rate but rather burst fraction per se that correlates with behavioral accuracy (Figure 4E right). At least, the latter correlation is the only one tested.

eLife assessment

This important study advances our understanding of how FGF13 variants confer seizure susceptibility. By acting in a set of inhibitory interneurons, FGF13 regulates synaptic transmission and excitability. The data presented here are convincing and combine cell type-specific knockouts and electrophysiology, complemented by histology/RNA studies. Collectively, this research will be of interest to a wide audience, particularly those involved in the study of epilepsy, inhibitory neurons, and ion channels.

Reviewer #1 (Public Review):

Summary:

A subset of fibroblast growth factor (FGF) proteins (FGF11-FGF14; often referred to as fibroblast growth factor homologous factors because they are not thought to be secreted and do not seem to act as growth factors) have been implicated in modulating neuronal excitability, however, the exact mechanisms are unclear. In part, this is because it is unclear how different FGF isoforms alter ion channel activity in different neuronal populations. In this study, the authors explore the role of FGF 13 in epilepsy using a variety of FGF13 knock-out mouse models, including several targeted cell-type specific conditional knockout mouse lines. The study is intriguing as it indicates that FGF13 plays an especially important role in inhibitory neurons. Furthermore, although FGF13 has been studied as a regulator of neuronal voltage-gated sodium channels, the authors present data indicating that FGF13 knockout in inhibitory neurons induces seizures not by altering sodium current properties but by reducing voltage-gated potassium currents in inhibitory neurons. While intriguing, the data are incomplete in several aspects and thus the mechanisms by which various FGF13 variants induce Developmental and Epileptic Encephalopathies are not resolved by the data presented.

Strengths:

A major strength is the array of techniques used to assess the mice and the electrical activity of the neurons.

The multiple mouse knock-out models utilized are a strength, clearly demonstrating that FGF13 expression in inhibitory neurons, and possibly specific sub-populations of inhibitory neurons, is critically important.

The data on the increased sensitivity to febrile seizures in KO mice are very nice, provide clear evidence for regulation of excitability in inhibitory neurons by FGF13.

The Gad2Fgf13-KO mice indicated that several Fgf13 splice variants may be expressed in inhibitory neurons and suggest that the Fgf13-VY splice variants may have previously unrecognized specific roles in regulating neuronal excitability.

The data on males and females from the various KO mice lines indicates a clear gene dosage effect for this X-linked gene.

The unbiased metabolomic analysis supports the assertion that Fgf13 expression in inhibitory neurons is important in regulating seizure susceptibility.

Weaknesses:

The knockout approach can be powerful but also has distinct limitations. Multiple missense mutations in FGF13-S have been identified. The knockout models employed here are not appropriate for understanding how these missense variants lead to altered neuronal excitability. While the data show that complete loss of Fgf13 from excitatory forebrain neurons is not sufficient to induce seizure susceptibility, it does not rule out that specific variants (e.g., R11C) might alter the excitability of forebrain neurons. The missense variants may alter excitatory and/or inhibitory neuron excitability in distinct ways from a full FGF13 knockout.

The electrophysiological experiments are intriguing but not comprehensive enough to support all of the conclusions regarding how FGF13 modulates neuronal excitability.

Another concern is the use of different ages of neurons for different experiments. For example, sodium currents in Figures 2 and 5 (and Supplemental Figures 2 and 7) are recorded from cultured neurons, which may have very different properties (including changes in sodium channel complexes) from neurons in vivo that drive the development of seizure activity.

Reviewer #2 (Public Review):

Summary:

The authors address three primary questions:

(1) how FGF13 variants confer seizure susceptibility,<br /> (2) the specific cell types involved, and<br /> (3) the underlying mechanisms, particularly regarding Nav dysfunction.

They use different Cre drivers to generate cell type-specific knockouts (KOs). First, using Nestin-Cre to create a whole-brain Fgf13 KO, they observed spontaneous seizures and premature death. While KO of Fgf13 in excitatory neurons does not lead to spontaneous seizures, KO in inhibitory neurons recapitulates the seizures and premature death observed in the Nestin-Cre KO. They further narrow down the critical cell type to MGE-derived interneurons (INs), demonstrating that MGE-neuron-specific KO partially reproduces the observed phenotypes. "All interneuron" KOs exhibit deficits in synaptic transmission and interneuron excitability, not seen in excitatory neuron-specific KOs. Finally, they rescue the defects in the interneuron-specific KO by expressing specific Fgf13 isoforms. This is an elegant and important study adding to our knowledge of mechanisms that contribute to seizures.

Strengths

• The study provides much-needed cell type-specific KO models.<br /> • The authors use appropriate Cre lines and characterize the phenotypes of the different KOs.<br /> • The metabolomic analysis complements the rest of the data effectively.<br /> • The study confirms and extends previous research using improved approaches (KO lines vs. in vitro KD or antibody infusion).<br /> • The methods and analyses are robust and well-executed.

Weaknesses

• One weakness lies in the use of the Nkx2.1 line (instead of Nkx2.1CreER) in the paper. As a result, some answers to key questions are incomplete. For instance, it remains unclear whether the observed effects are due to Chandelier cells or NGFCs, potentially both MGE and CGE derived, explaining why Nkx2.1 alone does not fully replicate the overall inhibitory KO. Using Nkx2.1CreER could have helped address the cell specificity. With the Nkx2.1 line used in the paper, the answer is partial.

• While the mechanism behind the reduced inhibitory drive in the IN-specific KO is suggested to be presynaptic, the chosen method does not allow them to exactly identify the mechanisms (spontaneous vs mEPSC/mIPSC), and whether it is a loss of inhibitory synapses (potentially axo-axonic) or release probability.

• Some supporting data (e.g. Supplemental Figure 7 and 8) appear to come from only one (or two) WT and one (or two) KO mice. Supplementary data, like main data, should come from at least three mice in total to be considered complete/solid (even if the statistical analysis is done with cells).

General Assessment

The general conclusions of this paper are supported by data. As it is, the claim that "these results enhance our understanding of the molecular mechanisms that drive the pathogenesis of Fgf13-related seizures" is partially supported. A more cautious term may be more appropriate, as the study shows the mechanism is not Nav-mediated and suggests alternative mechanisms without unambiguously identifying them. The conclusion that the findings "expand our understanding of FGF13 functions in different neuron subsets" is supported, although somewhat overstated, as the work is not conclusive about the exact neuron subtypes. However, it does indeed show differential functions for specific neuronal classes, which is a significant result.

Impact and Utility

This paper is undoubtedly valuable. Understanding that excitatory neurons are not the primary contributors to the observed phenotypes is crucial. The finding that the effects are not MGE-unique is also important. This work provides a solid foundation for further research and will be a useful resource for future studies.

Reviewer #3 (Public Review):

Summary:

The authors aimed to determine the mechanism by which seizures emerge in Developmental and Epileptic Encephalopathies caused by variants in the gene FGF13. Loss of FGF13 in excitatory neurons had no effect on seizure phenotype as compared to the loss of FGF13 in GABAergic interneurons, which in contrast caused a dramatic proseizure phenotype and early death in these animals. They were able to show that Fgf13 ablation and consequent loss of FGF13-S and FGF13-VY reduced overall inhibitory input from Fgf13-expressing interneurons onto hippocampal pyramidal neurons. This was shown to occur not via disruption to voltage-gated sodium channels but rather by reducing potassium currents and action potential repolarisation in these interneurons.

Strengths:

The authors employed multiple well-validated, novel mouse lines with FGF13 knocked out in specific cell types including all neurons, all excitatory cells, all GABAergic interneurons, or a subset of MGE-derived interneurons, including axo-axonic chandelier cells. The phenotypes of each of these four mouse lines were carefully characterised to reveal clear differences with the most fundamental being that Interneuron-targeted deletion of FGF13 led to perinatal mortality associated with extensive seizures and impaired the hippocampal inhibitory/excitatory balance while deletion of FGF13 in excitatory neurons caused no detectable seizures and no survival deficits.

The authors made excellent use of western blotting and in situ hybridisation of the different FGF13 isoforms to determine which isoforms are expressed in which cell types, with FGF3-S predominantly in excitatory neurons and FGF13-VY and FGF13-V predominantly in GABAergic neurons.

The authors performed a highly detailed electrophysiological analysis of excitatory neurons and GABAergic interneurons with FGF13 deficits using whole-cell patch clamp. This enabled them to show that FGF13 removal did not affect voltage-gated sodium channels in interneurons, but rather reduced the action of potassium channels, with the resultant effect of making it more likely that interneurons enter depolarisation block. These findings were strengthened by the demonstration that viral re-expression of different Fgf13 splice isoforms could partially rescue deficits in interneuron action potential output and restore K+ channel current size.

Additionally, the discussion was nuanced and demonstrated how the current findings resolved previous apparent contradictions in the field involving the function of FGF13.

These findings will have a significant impact on our understanding of how FGF13 causes seizures and death in DEEs, and the action of different FGF13 isoforms within different neuronal cell types, particularly GABAergic interneurons.

eLife assessment

This study presented a valuable inventory in scoring a neuropsychological test, ROCFT. The level of evidence is compelling. The authors constructed large samples from multi-center international researchers and tested the model using internet data with excellent performance. Their deep learning method could potentially apply to neuropsychological tests as well as other related fields.

Reviewer #1 (Public Review):

Summary:

The authors aimed to develop and validate an automated, deep learning-based system for scoring the Rey-Osterrieth Complex Figure Test (ROCF), a widely used tool in neuropsychology for assessing memory deficits. Their goal was to overcome the limitations of manual scoring, such as subjectivity and time consumption, by creating a model that provides automatic, accurate, objective, and efficient assessments of memory deterioration in individuals with various neurological and psychiatric conditions.

Strengths:

Comprehensive Data Collection:<br /> The authors collected over 20,000 hand-drawn ROCF images from a wide demographic and geographic range, ensuring a robust and diverse dataset. This extensive data collection is critical for training a generalizable and effective deep learning model.

Advanced Deep Learning Approach:<br /> Utilizing a multi-head convolutional neural network to automate ROCF scoring represents a sophisticated application of current AI technologies. This approach allows for detailed analysis of individual figure elements, potentially increasing the accuracy and reliability of assessments.

Validation and Performance Assessment:<br /> The model's performance was rigorously evaluated against crowdsourced human intelligence and professional clinician scores, demonstrating its ability to outperform both groups. The inclusion of an independent prospective validation study further strengthens the credibility of the results.

Robustness Analysis Efficacy:<br /> The model underwent a thorough robustness analysis, testing its adaptability to variations in rotation, perspective, brightness, and contrast. Such meticulous examination ensures the model's consistent performance across different clinical imaging scenarios, significantly bolstering its utility for real-world applications.

Weaknesses:

Insufficient Network Analysis for Explainability:<br /> The paper does not sufficiently delve into network analysis to determine whether the model's predictions are based on accurately identifying and matching the 18 items of the ROCF or if they rely on global, item-irrelevant features. This gap in analysis limits our understanding of the model's decision-making process and its clinical relevance.

Generative Model Consideration:<br /> The critique suggests exploring generative models to model the joint distribution of images and scores, which could offer deeper insights into the relationship between scores and specific visual-spatial disabilities. The absence of this consideration in the study is seen as a missed opportunity to enhance the model's explainability and clinical utility.

Appraisal and discussion:<br /> By leveraging a comprehensive dataset and employing advanced deep learning techniques, they demonstrated the model's ability to outperform both crowdsourced raters and professional clinicians in scoring the ROCF. This achievement represents a significant step forward in automating neuropsychological assessments, potentially revolutionizing how memory deficits are evaluated in clinical settings. Furthermore, the application of deep learning to clinical neuropsychology opens avenues for future research, including the potential automation of other neuropsychological tests and the integration of AI tools into clinical practice. The success of this project may encourage further exploration into how AI can be leveraged to improve diagnostic accuracy and efficiency in healthcare.

However, the critique regarding the lack of detailed analysis across different patient demographics, the inadequacy of network explainability, and concerns about the selection of median crowdsourced scores as ground truth raises questions about the completeness of their objectives. These aspects suggest that while the aims were achieved to a considerable extent, there are areas of improvement that could make the results more robust and the conclusions stronger.

Reviewer #2 (Public Review):

Summary:

The authors aimed to develop and validate a machine-learning-driven neural network capable of automatic scoring of the Rey-Osterrieth Complex Figure. They aimed to further assess the robustness of the model to various parameters such as tilt and perspective shift in real drawings. The authors leveraged the use of a huge sample of lay workers in scoring figures and also a large sample of trained clinicians to score a subsample of figures. Overall, the authors found their model to have exceptional accuracy and perform similarly to crowdsourced workers and clinicians with, in some cases, less degree of error/score dispersion than clinicians.

Strengths:

The authors used very large data; including a large number of Rey-Osterrieth Complex Figures, a huge crowdsourced human worker sample, and a large clinician sample.

The authors deeply describe their model in relatively accessible terms.

The writing style of the paper is accessible, scientific, and thorough.

Pre-registration of the prospectively collected new data was acceptable.

Weaknesses:

There is no detail on how the final scoring app can be accessed and whether it is medical device-regulated.

No discussion on the difference in sample sizes between the pre-registration of the prospective study and the results (e.g., aimed for 500 neurological patients but reported data from 288).

Details in pre-registration and paper regarding samples obtained in the prospective study were lacking.

Demographics for the assessment of the representation of healthy and non-healthy participants were not present.

The authors achieved their aims and their results and conclusions are supported by strong methods and analyses. The resulting app produced in this work, if suitable for clinical practice, will have impact in automated scoring, which many clinicians will be exceptionally happy with.

Reviewer #3 (Public Review):

Summary:

This study presented a valuable inventory of scoring a neuropsychological test, ROCFT, with constructing an artificial intelligence model.

Strengths:

They constructed huge samples collected among multi-center international researchers and tested the model precisely using internet data.<br /> The model scored the test with excellent ability, surpassing even experts. The product can run an application on a tablet, which helps clinicians and patients.<br /> Their method of building the model of deep learning and testing will apply to tests in all fields, not just the psychological field.

Weaknesses:

The considerable effort and cost to make the model only for an existing neuropsychological test.

eLife assessment

This work identifies the molecular function of an orphan human transporter, SLC35G1, providing convincing but somewhat incomplete evidence that this protein is involved in intestinal citrate absorption. This work provides important insight into transporter function and human physiology.

eLife assessment

This is a valuable study on the diffusion rates of drug molecules in human-derived cells, highlighting that their diffusion behavior depends on their charged state. It proposes that blocking drug protonation enhances diffusion and fractional recovery, suggesting improved intracellular availability of weakly basic drugs. The correlation between pKa and intracellular diffusion is solid and well-supported, but the study would benefit from a more rigorous statistical treatment and a balanced comparison across different types of compounds. Despite these limitations, the findings are significant for drug design and understanding the biophysical behavior of small molecules in cells.

eLife assessment

This useful study draws on published single-cell and spatial transcriptomic data of colon cancer liver metastasis to clarify the pro- and anti-tumorigenic properties of NK cells. The authors discover increased GZMK+ resting NK cells in the tumor tissue and reduced abundance of KIR2DL4+ activated NK cells. However, the evidence is currently incomplete, as the models used to validate the hypothesis and claims are inadequate and lack necessary controls.

eLife assessment

This important study examines the relationship between expiratory airflow and vocal pitch in adult mice during the production of ultrasonic vocalizations and also identifies a molecularly defined population of brainstem neurons that regulates mouse vocal production across development. The evidence supporting the study's conclusions that expiratory airflow shapes vocal pitch and that these brainstem neurons preferentially regulate expiratory airflow is novel and compelling. This work will be of interest to neuroscientists working on mechanisms and brainstem circuits that regulate vocal production and vocal-respiratory coordination.

Reviewer #1 (Public Review):

Summary:

In this important work, the authors propose and test a model for the control of murine ultrasonic vocalizations (USV) in which two independent mechanisms involving changes in laryngeal opening or airflow control vocal tone. They present compelling experimental evidence for this dual control model by demonstrating the ability of freely behaving adult mice to generate vocalizations with various intonations by modulating both the breathing pattern and the laryngeal muscles. They also present novel evidence that these mechanisms are encoded in the brainstem vocalization central neural pattern generator, particularly in the component in the medulla called the intermediate reticular oscillator (iRO). The results presented clearly advance understanding of the developmental nature of the iRO, its ability to intrinsically generate and control many of the dynamic features of USV, including those related to intonation, and its coordination with/control of expiratory airflow patterns. This work will interest neuroscientists investigating the neural generation and control of vocalization, breathing, and more generally, neuromotor control mechanisms.

Strengths:

Important features and novelty of this work include:

(1) The study employs an effective combination of anatomical, molecular, and functional/ behavioral approaches to examine the hypothesis and provide novel data indicating that expiratory airflow variations can change adult murine USV's pitch patterns.

(2) The results significantly extend the authors' previous work that identified the iRO in neonatal mice by now presenting data that functionally demonstrates the existence of the critical Penk+Vglut2+ iRO neurons in adult mice, indicating that the iRO neurons maintain their function in generating vocalization throughout development.

(3) The results convincingly demonstrate that the iRO neurons encode and can generate vocalizations by modulating both breathing and the laryngeal muscles.

(4) The anatomical mapping and tracing results establish an important set of input and output circuit connections to the iRO, including input from the vocalization-promoting subregions of the midbrain periaqueductal gray (PAG), as well as output axonal projections to laryngeal motoneurons, and to the respiratory rhythm generator in the preBötzinger complex.

(5) These studies advance the important concept that the brainstem vocalization pattern generator integrates with the medullary respiratory pattern generator to control expiratory airflow, a key mechanism for producing various USV types characterized by different pitch patterns.

Weaknesses:

A limitation is that the cellular and circuit mechanisms by which the vocalization pattern generator integrates with the respiratory pattern generator to control expiratory airflow has not been fully worked out, requiring future studies.

Reviewer #2 (Public Review):

Summary:

Both human and non-human animals modulate the frequency of their vocalizations to communicate important information about context and internal state. While regulation of the size of the laryngeal opening is a well-established mechanism to regulate vocal pitch, the contribution of expiratory airflow to vocal pitch is less clear. To consider this question, this study first characterizes the relationship between the dominant frequency contours of adult mouse ultrasonic vocalizations (USVs) and expiratory airflow using whole-body plethysmography. The authors also include data from a single mouse that combines EMG recordings from the diaphragm and larynx with plethysmography to provide evidence that the respiratory central pattern generator can be re-engaged to drive "mini-breaths" that occur during the expiratory phase of a vocal breath. Next, the authors build off of their previous work characterizing intermediate reticular oscillator (iRO) neurons in mouse pups to establish the existence of a genetically similar population of neurons in adults and show that artificial activation of iRO neurons elicits USV production in adults. Third, the authors examine the acoustic features of USV elicited by optogenetic activation of iRO and find that a majority of natural USV types (as defined by pitch contour) are elicited by iRO activation and that these artificially elicited USVs are more likely than natural USVs to be marked by positive intonation (positive relationship between USV dominant frequency and expiratory airflow).

Strengths:

Strengths of the study include the novel consideration of expiratory airflow as a mechanism to regulate vocal pitch and the use of intersectional methods to identify and activate the iRO in adult mice. The establishment of iRO neurons as a brainstem population that regulates vocal production across development is an important finding.

Weaknesses:

The conclusion that the respiratory CPG is re-engaged during "mini-breaths" throughout a given vocal breath would be strengthened by including analyses from more than one mouse.

Author response:

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

In the revised manuscript we have included an additional study that significantly contributes to the conclusions and models of the original version. Briefly, Figure 3 now describes our characterization of the diaphragm and laryngeal muscle activities (electromyography, EMG) during endogenous vocalizations. These EMGs also serve as representations of the brainstem breathing central pattern generator (CPG) inspiratory and post-inspiratory generating neurons, respectively. In our original submission, we found that many of the vocalizations had changes in pitch that mirrored the change in expiratory airflow (we termed positive intonation), and we proposed that the coordination of breathing muscles (like the inspiratory muscles) and larynx patterned this. This mechanism is akin to our findings for how neonatal cries are rhythmically timed and produced (Wei et al. 2022). The newly presented EMG data re-inforces this idea. We found that for vocalizations with positive intonation, the inspiratory diaphragm muscle has an ectopic burst(s) of activity during the expiration phase which corresponds to a decrease in airflow and pitch, and this is followed by laryngeal muscle activity and increased pitch. This can be cycled throughout the expiration to produce complex vocalizations with oscillations in pitch. A basal breath is hardwired for the laryngeal muscle activity to follow the diaphragm, so the re-cycling of this pattern nested within an expiration (a ‘mini-breath’ in a ‘breath’) demonstrates that the vocalization patterning system engages the entire breathing CPG. This contrasts with the canonical model that activity of the laryngeal premotor neurons control all aspects of producing / patterning vocalizations. Furthermore, this mechanism is exactly how the iRO produces and patterns neonatal vocalizations (Wei et al. 2022) and motivates the likely use of the iRO in adult vocalizations.

Response to recommendations for the authors:

Reviewer #1:

(1) The authors should note in the Discussion that the cellular and circuit mechanisms by which the vocalization pattern generator integrates with the respiratory pattern generator to control expiratory airflow have not been fully worked out, requiring future studies.

This was noted in the discussion section “The iRO likely patterns intonation for endogenous phonation”.

(2) Please change the labeling of the last supplemental figure to Figure Supplemental 5.

Thank you for identifying this.

Reviewer #2:

Major concerns

(1) While it is true that modulation of activity in RAm modulates the laryngeal opening, this statement is an incomplete summary of prior work. Previous studies (Hartmann et al., 2020; Zhang et al., 1992, 1995) found that activation of RAm elicits not just laryngeal adduction but also the production of vocal sounds, albeit vocal sounds that were spectrally dissimilar from speciestypical vocalizations. Moreover, a recent study/preprint that used an activity-dependent labeling approach in mice to optogenetically activate RAm neurons that were active during USV production found that re-activation of these neurons elicits USVs that are acoustically similar to natural USVs (Park et al., 2023). While the authors might not be required to cite that recent preprint (as it is not yet peer-reviewed), the fact that activation of RAm elicits vocal sounds is clear evidence that its effects go beyond modulating the size of the laryngeal opening, as this alone would not result in sound production (i.e., RAm activation must also recruit expiratory airflow). The authors should include these relevant studies in their Introduction. Moreover, the rationale for the model proposed by the authors (that RAm controls laryngeal opening whereas iRO controls expiratory airflow) is unclear with regard to these prior studies. The authors should include a discussion of how these prior findings are consistent with their model (as presented in the Introduction, as well as in Figure 4 and relevant Discussion) that RAm modulates the size of laryngeal opening but not expiratory airflow.

An introduction and discussion of the Veerakumar et. al. 2023 and Park et. al. 2024 manuscripts describing RAm in mice has now been included.

The iRO serves to coordinate the breath airflow and laryngeal adduction to produce sound and the intonation within it that mirrors the breath airflow. This occurs because the iRO can control the breathing CPG (synaptic input to the preBötC inspiratory pacemaker) and is premotor to multiple laryngeal muscles (Wei et. al. 2022). The modulation of the expiratory airflow is by inducing momentary contraction of the diaphragm (via excitation of the preBötC) which opposes (a.k.a. slows) expiration. This change in flow results in a decrease in pitch (Fig. 3 in the revised manuscript, Wei et. al. 2022).

It is our understanding that the basic model for RAm evoked USVs is that RAm evokes laryngeal adduction (and presumed abdominal expiratory muscle activation) and this activity is momentarily stopped during the breath inspiration by inhibition from the preBötC (Park et. al. 2024). So, in this basic model, any change in pitch and expiratory airflow would be controlled by tuning RAm activity (i.e., extent of laryngeal adduction). In this case, the iRO induced inspiratory muscle activity should not occur during expiration, which is not so (Fig. 3). Note, the activity of abdominal expiratory muscles during endogenous and RAm evoked USVs has not been characterized, so the contribution of active expiration remains uncertain. This is an important next step.

We have now included a discussion of this topic which emphasizes that iRO and RAm likely have reciprocal interactions (supported by the evidence of this anatomical structure). These interactions would explain why excitation of either group can evoke USVs and, perhaps, the extent that either group contributes to a USV explains how the pitch / airflow changes. An important future experiment will be to determine the sufficiency of each site in the absence of the other.

(2) The authors provide evidence that the relationship between expiratory airflow and USV pitch is variable (sometimes positive, sometimes negative, and sometimes not related). While the representative spectrograms clearly show examples of all three relationship types, no statistical analyses are included to evaluate whether the relationship between expiratory airflow and USV pitch is different than what one would expect by chance. For example, if USV pitch were actually unrelated to expiratory airflow, one might nonetheless expect spurious periods of positive and negative relationships. The lack of statistical analyses to explicitly compare the observed data to a null model makes it difficult to fully evaluate to what extent the evidence provided by the authors supports their claims.

We have now included two null distributions and compared our observed correlation values to these. The two distributions were created by taking each USV / airflow pair and randomly shuffling either the normalized USV pitch values (pitch shuffled) or the normalized airflow values (airflow shuffled) to simulate the distribution of data should no relationship exist between the USV pitch and airflow.

(3) The relationship between expiratory airflow and USV pitch comes with two important caveats that should be described in the manuscript. First, even in USV types with an overall positive relationship between expiratory airflow and pitch contour, the relationship appears to be relative rather than absolute. For example, in Fig. 2E, both the second and third portions of the illustrated two-step USV have a positive relationship (pitch goes down as expiratory airflow goes down). Nonetheless, the absolute pitch of the third portion of that USV is higher than the second portion, and yet the absolute expiratory airflow is lower. The authors should include an analysis or description of whether the relationship between expiratory airflow and USV pitch is relative vs.

absolute during periods of 'positive intonation'.

The relationship between pitch and airflow is relative and this in now clarified in the text. To determine this, we visualized the relationship between the two variables by scatterplot for each of the USVs syllables and, as the reviewer notes, a given airflow cannot predict the resulting frequency and vice versa.

(4) A second important caveat of the relationship between expiratory airflow and USV pitch is  that changes in expiratory airflow do not appear to account for the pitch jumps that characterize mouse USVs (this lack of relationship also seems clear from the example shown in Fig. 2E). This caveat should also be stated explicitly.

The pitch jumps do not have a corresponding fluctuation in airflow, and this is now stated in the results and discussion.

(5) The authors report that the mode of relationship between expiratory airflow and USV pitch (positive intonation, negative intonation, or no relationship) can change within a single USV. Have the authors considered/analyzed whether the timing of such changes in the mode of relationship coincides with pitch jumps? Perhaps this isn’t the case, but consideration of the question would be a valuable addition to the manuscript.

We analyzed a subset of USVs with pitch jumps that were defined by a change >10 kHz, at least 5ms long, and had one or two jumps. The intonation relationships between the sub-syllables within a USV type were not stereotyped as evidenced by the same syllable being composed of combinations of both modes.

(6) The authors incorrectly state that PAG neurons important for USV production have been localized to the ventrolateral PAG. Tschida et al., 2019 report that PAG-USV neurons are located predominantly in the lateral PAG and to a lesser extent in the ventrolateral PAG (see Fig. 5A from that paper). The finding that iRO neurons receive input from VGlut2+ ventrolateral PAG neurons represents somewhat weak evidence that these neurons reside downstream of PAG-USV neurons. This claim would be strengthened by the inclusion of FOS staining (following USV production), to assess whether the Vglut+ ventrolateral PAG neurons that provide input to iRO are active in association with USV production.

This comment correctly critiques that our PAG à iRO tracing does not demonstrate that the labeled PAG neurons are sufficient nor necessary for vocalization. Directly demonstrating that activation and inhibition the PAG-iRO labeled neurons ectopically drives or prevents endogenous USVs is an important next step. While FOS implies this connectivity, it does not definitely establish it and so this experiment is impacted by some of the caveats of our tracing (e.g. PAG neurons that drive sniffing might be erroneously attributed to vocalization).

Our reading of the literature could not identify an exact anatomical location within the mouse PAG and this site appears to vary within a study and between independent studies (like within and between Tschida et. al. 2019 and Chen et. al. 2021). The labeling we observed aligns with some examples provided in these manuscripts and with the data reported for the retrograde tracing from RAm (Tschida et al 2019).

(7) In Figure S5A, the authors show that USVs are elicited by optogenetic activation of iRO neurons during periods of expiration. In that spectrogram, it also appears that vocalizations were elicited during inspiration. Are these the broadband vocalizations that the authors refer to in the Results? Regardless, if optogenetic activation of iRO neurons in some cases elicits vocalization both during inspiration and during expiration, this should be described and analyzed in the manuscript.

The sound observed on the spectrogram during inspiration is an artefact of laser evoked head movements that resulted in the fiber cable colliding with the plethysmography chamber. In fact, tapping an empty chamber yields the same broad band spectrogram signal. The evoked USV or harmonic band vocalization is distinct from this artefact and highlighted in pink.

(8) Related to the comment above, the authors mention briefly that iRO activation can elicit broadband vocalizations, but no details are provided. The authors should provide a more detailed account of this finding.

The broadband harmonic vocalizations we sometimes observe upon optogenetic stimulation of AAV-ChR2 expressing iRO neurons are akin to those previously described within the mouse vocal repertoire (see Grimsley et. al .2011). We have added this citation and mentioned this within the text. 

(9) The effects of iRO stimulation differ in a couple of interesting ways from the effects of PAGUSV activation. Optogenetic activation of PAG-USV neurons was not found to entrain respiration or to alter the ongoing respiratory rate and instead resulted in the elicitation of USVs at times when laser stimulation overlapped with expiration. In contrast, iRO stimulation increases and entrains respiratory rate, increases expiratory and inspiratory airflow, and elicits USV production (and also potentially vocalization during inspiration, as queried in the comment above). It would be informative for the authors to add some discussion/interpretation of these differences.

We have added a section of discussion to describe the how these different results may be explained by the iRO being a vocal pattern generator versus the PAG as a ‘gating’ signal to turn on the medullary vocalization patterning system (iRO and RAm). See discussion section ‘The iRO likely patterns intonation for endogenous phonation’.

(10) The analysis shown in Fig. 4D is not sufficient to support the author’s conclusion that all USV types elicited by iRO activation are biased to have more positive relationships between pitch and expiratory airflow. The increase in the relative abundance of down fm USVs in the opto condition could account for the average increase in positive relationship when this relationship is considered across all USV types in a pooled fashion. The authors should consider whether each USV type exhibits a positive bias. Although such a comparison is shown visually in Fig. 4G, no statistics are provided. All 7 USV types elicited by optogenetic activation of iRO should be considered collectively in this analysis (rather than only the 5 types currently plotted in Fig. 4G).

In the original submission the statistical analysis of r values between opto and endogenous conditions was included in the figure legend (‘panels E-G, two-way ANOVA with Sidak’s post-hoc test for two-way comparisons was used; all p-values > 0.05), and this has not changed in the revised manuscript. We have now provided the suggested comparison of opto vs endogenous USVs without down fm (Fig. 5D). This positive shift in r is statistically significant (…).

(11) The evidence that supports the author’s model that iRO preferentially regulates airflow and that RAm preferentially regulates laryngeal adduction is unclear. The current study finds that activation of iRO increases expiratory (and inspiratory) airflow and also elicits USVs, which means that iRO activation must also recruit laryngeal adduction to some extent. As the authors hypothesize, this could be achieved by recruitment of RAm through iRO’s axonal projections to that region.

Note, it is more likely that iRO is directly recruiting laryngeal adduction as they are premotor to multiple laryngeal muscles like the thyroarytenoid and cricothyroid (Wei et. al. 2022). The ‘Discussion’ now includes our ideas for how the iRO and RAm likely interact to produce vocalizations.

In the recent preprint from Fan Wang’s group (Park et al., 2023), those authors report that RAm is required for USV production in adults, and that activation of RAm elicits USVs that appear species-typical in their acoustic features and elicits laryngeal adduction (assessed directly via camera). Because RAm activation elicits USVs, though, it must by definition also recruits expiratory airflow. Can the authors add additional clarification of how the evidence at hand supports this distinction in function for iRO vs RAm?

See response to ‘Major Concern #1”.

Minor concerns 

(1) The authors might consider modifying the manuscript title. At present, it primarily reflects the experiments in Figure 2.

We have provided a title that we feel best reflects the major point of the manuscript. We hope that this simplicity enables it to be recognized by a broad audience of neuroscientists as well as specialists in vocalization and language.

(2) The statement in the abstract that "patterns of pitch are used to create distinct 'words' is somewhat unclear. Distinct words are by and large defined by combinations of distinct phonemes. Are the authors referring to the use of "tonemes" in tonal languages? If so, a bit more explanation could be added to clarify this idea. This minor concern includes both the Abstract, as well as the first paragraph of the Introduction.

We have clarified this line in the abstract to avoid the confusing comparison between mouse vocalizations and human speech. In the introduction we have expanded our explanation to clarify that variations in pitch are a component of spoken language that add additional meaning and depth to the underlying, phonemic structure. 

(3) Multiple terms are used throughout the manuscript to refer to expiratory airflow: breath shape (in the title), breath pattern, deviations in exhalation, power of exhalation, exhalation strength, etc. Some of these terms are vague in meaning, and a consolidation of the language would improve the readability of the abstract and introduction.

We have chosen a smaller selection of descriptive words to use when describing these breath features.

(4) Similarly, "exhalation" and "expiration" are both used, and a consistent use of one term would help readability.

See point 3.

(5) In a couple of places in the manuscript, the authors seem to state that RAm contains both laryngeal premotor neurons as well as laryngeal motor neurons. This is not correct to our knowledge., but if we are mistaken, we would ask that the authors add the relevant references that report this finding.

It is our understanding that the RAm is defined as the anatomical region consistent with the murine rostral and caudal ventral respiratory groups composed of multiple premotor neuron pools to inspiratory, expiratory, laryngeal, and other orofacial muscles. This is supported by neurons within RAm that reflect multiple phases of the inspiratory and expiratory cycle (Subramanian et. al. 2018) and excitation of sub-regions within RAm modulating multiple parts of the breathing control system (Subramanian et. al. 2018 and Subramanian 2009). Rabies tracing of the various premotor neurons which define the anatomical region of RAm in the mouse shows that they surround the motor neurons in the loose region of the nucleus ambiguus (the anatomical location of RAm) for multiple muscles of the upper airway system, such as the thyroarytenoid (Wu et. al. 2017, Dempsey et. al. 2021 and Wei et. al. 2022). Given that the name RAm reflects a broad anatomical location, we have used it to describe both the premotor and motor neurons embedded within it. We have now clarified this in the text.

(6) The statistical analysis applied in Figure 1C is somewhat confusing. The authors show two distributions that appear different but report a p-value of 0.98. Was the analysis performed on the mean value of the distributions for each animal, the median, etc.? If each animal has two values (one for USV+ breaths and one for USV- breaths), why not instead compare those with a paired t-test (or Wilcoxon rank sign)? Additional information is needed to understand how this analysis was performed.

The original manuscript version used a two-way anova to compare the normalized histogram of instantaneous frequency for breaths with (USV+) or without (USV-) for each animal (first factor: USV+/-, second factor: Frequency). The p-value for the first factor (USV) was 0.98 showing no statistically significant effect of USV on the distribution of the histogram.

For simplicity, we have instead performed the analysis as suggested and include a bar graph. This analysis shows that the instantaneous frequency of USV breaths is, in fact, statistically significantly lower than those without USVs. We have updated the figure legend and text to reflect this.

(7) The use of the word "syllable" to describe parts of a USV that are produced on a single breath may be confusing to some scientists working on rodent USVs. The term 'syllable' is typically used to describe the entirety of a USV, and the authors appear to use the term to describe parts of a USV that are separated by pitch jumps. The authors might consider calling these parts of USVs "sub-syllables".

We have clarified these descriptions throughout the text. We now refer to the categories as ‘syllable types’, define ‘syllables’ as ‘a continuous USV event’ with no more than 20ms of silence within and finally ‘sub-syllables’ to refer to components of the syllable separated by jumps in frequency (but not gaps in time).

(8) In Figure S3, final row, the authors show a USV produced on a single breath that contains two components separated by a silent period. This type of bi-syllabic USV may be rare in adults and is similar to what the authors showed in their previous work in pups (multiple USVs produced on a single expiration, separated by mini-inspirations). One might assume that the appearance of such USVs in pups and their later reduction in frequency represents a maturation of vocalrespiratory coordination. Nonetheless, the appearance of bi-syllabic USVs has not been reported in adult mice to our knowledge, and the authors might consider further highlighting this finding.

We were also struck by the similarity of these USVs to our study in neonates and such types of similarities sparked an interest in the role of the iRO in patterning adult USVs. We now include a description of the presence and abundance of bi- and tri-syllablic calls observed in our recordings to highlight this finding.

(9) Figure 4 is referenced at the end of the second Results section, but it would seem that the authors intended to reference Figure 2. 

For simplicity we included some of the referenced data within Fig. S5. We appreciate the recommendation.

(10) In the optogenetic stimulation experiments, the authors should clarify why bilateral stimulation was applied. Was unilateral stimulation ineffective or less effective? The rationale provided for the use of bilateral stimulation (to further localize neural activation) is unclear.

The iRO is bilateral and, we presume, functions similarly. So, we attempted to maximally stimulate the system. We have clarified this in the methods.

(11) Figure Supplemental '6' should be '5'.

Thanks!

(12) Last sentence of the Introduction: "Lasty" should be "lastly".

Thanks!

(13) There are two references for Hage et al., 2009. These should be distinguished as 2009a and 2009b for clarity.

Thanks!

Author response:

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

We thank the reviewers and editor for their careful review of our work. We believe the resulting manuscript is much stronger. We agree with the comments made by Reviewer #2 regarding additional histology and neuronal data analysis, which will be presented in subsequent work.

---

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

Reviewer 1 (Public Weaknesses):

It was not always clear what the lesion size was. This information is important for future applica- tions, for example, in the visual cortex, where neurons are organized in retinotopy patterns.

We thank the reviewer for this feedback. While there is some variation in lesion volume for a given parameter set, we have added more details of the volumes of lesions created in our testing (Fig. 4 and Fig. 5).

It would be helpful if the author could add some discussion about whether and how this method could be used in other types of array/multi-contact electrodes, such as passive neuropixels, S- probes, and so on. In addition, though an op-amp was used in the design, it would still be helpful if the author could provide a recommended range for the impedance of the electrodes.

We thank the reviewer for this suggestion. We have both added a demonstration of use in a differ- ent multielectrode probe type (with a U-probe) in Fig. 8, and we have added a discussion about which types of multielectrode probes would be suitable on Page 15, Line 420.

“We demonstrated that our electrolytic lesioning technique works with a linear multicontact probe by testing with a U-Probe in ex vivo rabbit cortex. There are no particular limitations that would prevent our specific electrolytic lesioning technique and device from working with any passive multielectrode probe. The main requirements for use are that the probe has two electrodes that can directly (via whatever necessary adapters) connect to the lesioning device, such that arbitrary current can be passed into them as the anode and cathode. This would limit use of probes, like Neuropixels, where the on-chip acquisition and digitization circuitry generally precludes direct connection to electrodes [1], [2]. The impedance of the multielectrode probe should not be an issue, due to the use of an op amp. We showed use  with a Utah array (20-800 kΩ) and a U-Probe (1-1.5 MΩ). The specific op amp used here has a voltage range of ± 450 V, which assuming a desired output of 150 µA of current would limit electrode impedance to 6 MΩ. Though a different op amp could easily be used to accommodate a higher electrode impedance, it is unlikely that this would be necessary, since most electrodes have impedances between 100 kΩ to 1 MΩ [3].”

Reviewer 2 (Public Weaknesses):

In many of the figures, it is not clear what is shown and the analysis techniques are not well described.

We thank the reviewer for this feedback. We hope that our edits to both the figures and the text have improved clarity for readers.

The flexibility of lesioning/termination location is limited to the implantation site of the multielec- trode array, and thus less flexible compared to some of the other termination methods outlined in Appendix 2.

We thank the reviewer for this point. You are right that the lesioning location is limited to the multielectrode array’s implantation site, while other methods in Appendix 2 do not require prox- imity of the lesion location and the electrophysiology recording site. However, we believe that the closeness of the lesioning location to the microelectrode array is a strength - guaranteeing record- ings from the perilesional area - even with the small negative of reduced flexibility. Multielectrode arrays can be implanted in many areas of cortex. If one wanted to study distal effects of a lesion, additional electrophysiology probes could be implanted to record from those areas. We have noted this on Page 3, Line 117.

“While the link between the lesion location and the multielectrode location technically con- strains the lesion to an area of cortex in which a multielectrode array could be implanted, we see the connection as a positive, because it ensures recording some neuroelectrophysiology from the perilesional area in which recovery is hypothesized to occur (see Appendix 1Data Availabilityappendix.41).”

Although the extent of the damage created through the Utah array will vary based on anatomical structures, it is unclear what is the range of lesion volumes that can be created with this method, given a parameter set. It was also mentioned that they performed a non-exhaustive parameter search for the applied current amplitude and duration (Table S1/S2) to generate the most suitable lesion size but did not present the resulting lesion sizes from these parameter sets listed. Moreover, there’s a lack of histological data suggesting that the lesion size is precise and repeatable given the same current duration/amplitude, at the same location.

We thank the reviewer for this thoughtful feedback. We have added figures (Figs. 4 and 5), where we show the relationship between estimated lesion volume and the current amplitude and duration parameters. These figures include more data from the tests in Supplementary File 1 and Supplementary File 2. While there is some variation in lesion volume for a given current amplitude and duration, there is still a clear relationship between the parameters and lesion volume.

It is unclear what type of behavioral deficits can result from an electrolytic lesion this size and type (∼3 mm in diameter) in rhesus macaques, as the extent of the neuronal loss within the damaged parenchyma can be different from past lesioning studies.

While we appreciate the reviewer’s interest in the behavioral deficits associated with our lesions in rhesus macaques, reporting these falls beyond the scope of this manuscript. Future work will explore the behavioral deficits associated with these lesions

The lesioning procedure was performed in Monkey F while sedated, but no data was presented for Monkey F in terms of lesioning parameters, lesion size, recorded electrophysiology, histological, or behavioral outcomes. It is also unclear if Monkey F was in a terminal study.

We apologize for not being more explicit about the parameters used for the lesion in Monkey F. We have added this in Results on Page 5, Line 209 and in Methods on Page 19, Line 586.

“After this validation and refinement, one proof-of-concept lesion (150 µA direct current passed through adjacent electrodes for 45 seconds) was performed in an in vivo sedated rhe- sus macaque (Monkey F) in order to validate the safety of the procedure.”

“This lesion was created by applying 150 µA of direct current to two adjacent electrodes in the microelectrode array for 45 seconds.”

We also clarified the parameters used for the other lesions in Monkeys H and U in Results on Page 7, Line 233 and in Methods on Page 19, Line 586.

“In all of the fourteen lesions across two awake-behaving rhesus macaques (150 µA direct current passed through adjacent electrodes for 30 or 45 seconds (30s for Monkey U and 45s for Monkey H, except lesion H200120 which was for 50 seconds)), the current source worked as expected, providing a constant current throughout the duration of the procedure.”

“In these lesions, 150 µA of direct current was applied to two adjacent electrodes in the mi- croelectrode array for 30 or 45 seconds (30s for Monkey U, 45s for Monkey H), except in lesion H200120 where current was applied for 50 seconds.”

Monkey F was euthanized shortly after the lesion, so we now mention this on Page 19, Line 583.

“Based on this, and a lack of physiological signs of pain from the anaesthetized pig studies, a lesion was performed on a sedated rhesus macaque who was subsequently euthanized due to unrelated health complications (Monkey F; 16 year-old adult, male rhesus macaque) in order to further verify safety before use in awake-behaving rhesus.”

Because Monkey F was sedated and then euthanized shortly after, there is no behavioral data. As the lesion in sedated Monkey F was used to validate the safety of the procedure, any further data and analysis fall beyond the scope of this manuscript.

As an inactivation method, the electrophysiology recording in Figure 5 only showed a change in pairwise comparisons of clustered action potential waveforms at each electrode (%match) but not a direct measure of neuronal pre and post-lesioning. More evidence is needed to suggest robust neuronal inactivation or termination in rhesus macaques after electrolytic lesioning. Some exam- ples of this can be showing the number of spike clusters identified each day, as well as analyzing local field potential and multi-unit activity.

The reviewer has pointed out some short comings of the original analysis, which we believe have since been addressed with the revised analysis. LFP and spiking activity are functional measures that are more ambiguous in terms of loss and are also the subject of another manuscript currently under revision.

The advantages over recently developed lesioning techniques are not clear and are not discussed.

We thank the reviewer for noting this. We have added a section, also responding to their later request for us to compare our work to Khateeb et al. 2022, by adding a section to the Discussion on Page 16, Line 434.

“Perhaps the most unique advantage of our technique in comparison with other existing inactivation methods lies in Design Consideration #1: stable electrophysiology pre- and post-inactivation (Appendix 1Data Availabilityappendix.41). While several methods exist that allow for localization and size control of the inactivation (Design Consideration #2) and cross compatibility across regions and species (Design Consideration #3), few have achieved compatibility with stable electrophysiology. For example, some studies record electrophysiology only after the creation of the lesion, preventing comparison with baseline neuronal activity [4]. One recent study, Khateeb, et al., 2022, developed an inactivation method that is effectively combined with stable electrophysiology by creating photothrombotic lesions through a chronic cranial window integrated with an electrocorticography (ECoG) array [5], which may be appropriate for applications where local field potential (LFP) recording is sufficient. This approach has trade-offs with regards to the three design considerations presented in Appendix 1Data Availabilityappendix.41.

While Khateeb, et al., present a toolbox with integrated, stable electrophysiology from an ECoG array pre- and post- inactivation (Design Consideration #1), it demonstrated recordings from an ECoG array with limited spatial resolution. While a higher density ECoG array that would provide higher spatial resolution could be used, increasing the density of opaque electrodes might occlude optical penetration and constrain photothrombotic lesions. Further, ECoG arrays are limited to recording LFP, not electrophysiology at single neuron resolution, potentially missing meaningful changes in the neuronal population activity after lesioning. Khateeb, et al., demonstrated localization and control the size of inactivation (Design Consideration #2). In this manuscript, we have shown that the amount and duration of direct current are significant determinants of lesion size and shape, while with photothrombotic lesions, light intensity and aperture diameter are the significantly relevant parameters. One potential advantage of photothrombotic approaches is the use of optical tools to monitor anatomical and physiological changes after lesioning through the cranial window, though the research utility of this monitoring remains to be demonstrated.

Although the method presented by Khateeb, et al., shows some cross-compatibility (Design Consideration #3), it has greater limitations in comparison with the method presented here. For example, while Khateeb, et al., notes that the approach could be adapted for use in smaller organisms, no modification is needed for use in other species with this work’s approach–so long as a multielectrode probe is implantable. In this manuscript we demon- strate electrolytic lesioning spanning two multielectrode probes across rabbits, pigs, sheep, and rhesus macaques, and our same device could be easily used with other smaller species, like rats, in which multielectrode probes have been successfully implanted [6]. Further, the approach in Khateeb, et al., is limited to superficial brain structures, due to the need for opti- cal accessibility. As noted, fiber optics could allow access to deeper structures, which would bring associated additional tissue damage, but deeper structure lesioning was not demon- strated. In contrast, the approach presented here can be used in any region of cortex in which a multielectrode probe can be implanted, which, depending on the probe used, does not limit it to surface structures. For example, we demonstrated use of our lesioning tech- nique with a linear U-probe (Fig. 8figure.caption.25), which could be used to reach deeper layers of cortex or specific deep cortical structures. In both techniques, the location of the lesion is tied to the location of the electrophysiology (for Khateeb et al., wherever the cra- nial window and ECoG array are; for this technique, wherever the multielectrode probe has been implanted), which ensures that the electrophysiology will include recordings from the perilesional area. Neither work addresses the potential of their technique to induce chronic post-lesion behavioral effects, which is a key goal for future work.”

There is a lack of quantitative histological analysis of the change in neuronal morphology and loss.

We appreciate the reviewer’s desire for a quantitative histological analysis, however this falls out- side of the scope of this manuscript. We are not attempting to make strong claims about the number of neurons lost through lesioning or thoroughly characterize morphological changes in the neurons. The histology is intended to show that lesioning did lead to a loss of neurons, but the precise num- ber of neurons lost is neither in scope nor is likely to be highly conserved across lesions.

There is a lack of histology data across animals and on the reliability of their lesioning techniques across animals and experiments.

We thank the reviewer for this point. As stated above, we have now added Fig. 4 and Fig. 5, which includes volume estimates based on the histology from more of our ex vivo and in vivo testing across animals.

There is a lack of data on changes in cortical layers and structures across the lesioning and non- lesioning electrodes.

We acknowledge that the histology does not have the level of detail that is expected from many modern studies. However, the goal here was dramatically different: we sought to calibrate a novel lesion device, ensure it’s safe use in large mammals (specifically, non-human primates) and pro- vide estimates of the lesion size to compare with the literature. The extent of histology that could be performed and the tools available to us prevent such an in depth analysis. We can say based on shank length of the Utah arrays used and known anatomy that we have affected layer 2/3 and maybe a bit of layer 4.

Reviewer 1 (Recommendations For The Authors):

Figure 5b. It would be helpful if the author could plot the delta match separately for the lesion elec- trodes, near neighbor electrodes, and far neighbors. This would help understand the lesion effect, specifically whether the effect is selective (e.g., more potent for the lesion and adjacent electrodes.)

The fact that neuron loss is not particularly selective can already be seen in the spike waveform plots, arranged spatially on the array. Plenty of clear change is observed far from the lesion elec- trodes (marked with black dots) as well as nearby. We have made mention of this localized non- specificity in the main text and have ensured to remphasize in the figure legened. While a nice suggestion, we currently don’t feel this result rises to the level of a figure given it is not highly specific spatially.

Reviewer 2 (Recommendations For The Authors):

Overall the quality of the paper, the figures and the analysis used could be significantly improved. There is a lack of scientific rigor in the presentation of figures and analysis techniques. It is not clear what the authors are trying to communicate through the figures and their choice of figures to show is confusing (see below).

We thank the reviewer for their pointed critiques and believe we have addressed their concerns with many changes to the text, a revamped waveforms analysis, and both the expansion and addition of results.

The neurophysiology data shown doesn’t suggest neuronal loss, it only shows change which needs strong control data to show it is due to a lesion.

As detailed below, we have presented a revised analysis that provides this control. While the reviewer is right to point out we can distinguish actual neuron loss from neuron silencing, we be- lieve the new analysis rigorously indicates new rates of sample turnover beyond those expected from healthy state.

The histology figure should be replaced with a high-quality representation without folds.

We understand the reviewer’s suggestion. While ideally we would have many histology slices from each lesion, due to cost, we were only able to collect one histology slice per lesion. The folds were introduced by the company that performed the H&E staining, and we unfortunately cannot remove the folds. Therefore, despite the folds, this is the best and only image from this lesion. We hope that the markings on the figure and the comment in the caption is sufficient to explain to readers that the folds are not a result of the lesion but instead a result of the histology process.

The authors suggest that this lesioning method will be compatible with any available multielec- trode probe theoretically. Since all testing was done with a Utah array, it will be helpful to add an explanation about potential constraints that will make a given array compatible with this method.

We thank the reviewer for this suggestion. As stated above, we have both added a demonstration of use in a different multielectrode probe type (with a U-probe) in Fig. 8, and we have added a discussion about which types of multielectrode probes would be suitable on Page 15, Line 420.

The authors should cite and discuss previous studies using electrolytic lesioning in awake-behaving animals to study the causal connection between the brain and behavior. (One example study: Morissette MC, Boye SM. Electrolytic lesions of the habenula attenuate brain stimulation reward. Behavioural brain research. 2008 Feb 11;187(1):17-26.)

We thank the reviewers for this suggestion. We have added a mention of existing electrolytic le- sioning studies on Page 2, Line 88.

“Prior termination studies mostly measure behavioral output, with no simultaneous measures of neuronal activity during the behavior, impairing their ability to provide insight into the causal connection between the brain and behavior [7]–[11], or with no baseline (i.e., pre- lesion) measures of neuronal activity [4].”

The authors should compare their technique with other recent lesioning studies in primates (e.g. Khateeb et al, 2022)

We again thank the reviewer for this point. Specifically not mentioning Khateeb et al. 2022 was a submission error on our part; we cited the paper in Appendix 2 in the version uploaded to the eLife submission portal, but we had uploaded the version prior to citing it to bioRxiv. We have combined addressing this with addressing a previous comment, as mentioned above, with a section in the Discussion on Page 16, Line 434.

In Appendix 2, the authors suggest that a major limitation of optogenetics and chemogenetic in- activation methods is the lack of rhesus-compatible constructs. However, several viral constructs have successful implementation in rhesus monkeys so far (e.g. Galvan A, Stauffer WR, Acker L, El-Shamayleh Y, Inoue KI, Ohayon S, Schmid MC. Nonhuman primate optogenetics: recent advances and future directions. Journal of Neuroscience. 2017 Nov 8;37(45):10894-903; Tremblay et al, Neuron 2020)

We thank the reviewer for pointing us to these papers. We have added a more thorough description of what we meant by lack of rhesus-compatible constructs in that Appendix.

“However, other challenges exist with using optogenetics as an inactivation method in nonhu- man primates, including difficulty reliably affecting behavior [12]. While several constructs for rhesus macaques have been developed [13], [14], reports of successfully inducing be- havioral effects have a small effect size and are less numerous than might be expected [12], and several null results have been published [15]–[17]. Other remaining challenges include the need to develop a head-mounted, battery powered light delivery system for multi-day delivery of light and difficulty integrating illumination with simultaneous chronic neuro- electrophysiology.”

For Figure 5b, only pairwise comparison results from monkey U (L11-14) are shown. It is unclear why such results from monkey H were shown in Figure 5a but not in 5b.

We thank the reviewer for pointing out this unconventional one monkey result. As described in the original submission, we previously omitted Monkey H from the analysis in Figure 5b (now Figure 7) since some of the lesions were closely spaced together, preventing well defined pre- and post- lesion rates of turnover. Never-the-less we have included Monkey H in all the revised analysis and believe even the less cleanly separated data shows useful indications of neuron loss or silencing evoked by the lesion.

Behavioral data (during a motor task) from the awake behaving monkeys (U and H) would greatly strengthen the claim that this lesioning method is capable of creating a behavioral effect and can be adopted to study the relationship between neural function and behavior outcomes.

While we are grateful for the reviewer’s interest in the application of our lesioning technique to studies involving behavior, a behavioral analysis of the effects of our electrolytic lesions falls be- yond the scope of this Tools and Resources manuscript. We would also like to point out that we do not claim that we have achieved a behavioral deficit in this manuscript.

Figure 2 would benefit from an illustration of the Utah array placement and the location of the sites used for lesioning. The authors can either overlay the illustrations on the current ex-vivo and histology images or create a separate schematic to demonstrate that for the readers. Also, Figure 2B needs to be replaced with one without the folds to avoid confusion for the readers.

We have added Figure 2 - figure supplement 1, which shows both the location within the Utah array of the two electrodes used to create the lesions as well as the relative size of the surface area of the lesion and the array. Unfortunately, as the lesion was created under the array, the exact location of the array relative to the lesion is unknown.

As mentioned above, Figure 2B is the only histological image from that lesion. We hope that the markings in the image as well as the caption sufficiently explain that the folds are unrelated to the lesion itself.

Figure 3, the conical region is not well delineated. Data across animals and lesion volume with respect to different parameters should be included.

We have included a supplemental figure, Figure 3 - figure supplement 1, where we have used a dashed white line to clearly indicate the area of damaged parenchyma, in case it was not clear in Figure 3a. We have also added volume estimates from lesions across animals and different param- eters. The ex vivo estimates are shown in Figure 4 and the in vivo estimates are shown in Figure 5.

Figure 4: it is not clear what is being communicated, and where the voltage traces are from.

We thank the reviewer for noting this confusion. We have added some lines in the text to explain what the voltage traces show, both in the caption to Fig. 6 and in the text on Page 7, Line 238.

“Traces only capture the values while the lesioning device was turned on (45 seconds for most lesions and 50 seconds for lesion H200120). A) Voltage traces. Discontinuity at the beginning of the traces indicates transient voltages that were too rapid to be captured by the voltmeter, lasting between 0.13 and 0.33 s. The fluctuating voltages, especially the rapid in- crease in voltage at the beginning of lesioning, emphasize the importance of using a current source to deliver consistent amounts of current into the brain.”

“The voltage across the microelectrode array fluctuated much more than the current did, em- phasizing that we made the correct choice in using a current source to ensure delivery of consistent amounts of current into the brain (Fig. 6figure.caption.19).”

Figure 5: why did the authors choose to use matching units as a measure of the lesion? It is surprising that there are still units on the location that the authors claim to be a lesion. To clarify that it would be helpful to show the location of the lesion in Figure 4a. Also, what can we conclude about the lesion induction when we see units on the lesion electrode? The change in unit match shows that there is a change in the network (although the authors need to show control for that so we know those changes don’t happen due to natural dynamics). It is not clear what is the time duration for pre-pre and post-post (i.e. minutes, seconds, hours). Do these comparisons come from the same time frame or are they coming from two fragments of time for both pre and post- conditions?

Aside from post-mortem histology and tissue assays, there is no good way to confirm neuron loss with chronically implanted electrode arrays in nonhuman primates. Waveforms were chosen as they are the one readily isolated physical measure of the system we are injuring. Although functional measures of activity could indicate neuron loss (topic of following papers), there are many conceivable changes in firing rate patterns that could manifest spuriously as loss, making the estimation of loss even more ambiguous and challenging this way.

We believe the new Figure 7 will make the procedure much more clear, while also providing the control requested by the reviewer, illustrating that new statistical categories of altered waveforms emerge during a lesion, beyond those associated with typical changes in waveform composition within multi-unit recordings seen during recording sample turnover fom healthy animals. We further note that by confining this analysis to four day spans at most, we have limited the impact of daily sample turnover described in the literature (Gallego, 2020).

The time duration for pre-session versus pre-session (pre-post and post-post), is some multiple of the approximate 24 hours between each daily recording session. Therefore, since restricting our- selves to four days separation, between 24 and 96 hours. Spikes are sampled from successful trial periods (so on the order of seconds, compiled into minutes across the whole recording session). Although already described in the main text, these points have been reemphasized in the figure legend.

CNO (line 931) needs to be explained.

We thank the reviewer for this point. We have defined CNO and its relevance in Appendix 2.

“Additionally, chronic inactivation over days may be logistically challenging, as the half life of clozapine N-oxide (CNO, a ligand used to activate DREADD receptors) is on the order of hours.”

eLife assessment

This paper reports a valuable new method for creating localized damage to candidate brain regions for functional and behavioral studies. The authors present solid support for their ability to create long-term local lesions with mm spatial resolution. The paper is likely to be of broad interest to brain researchers working to establish causal links between neural circuits and behavior.

Reviewer #1 (Public Review):

In the paper, the authors illustrated a novel method for Electrolytic Lesioning through a microelectronics array. This novel lesioning technique is able to perform long-term micro-scale local lesions with a fine spatial resolution (mm). In addition, it allows a direct comparison of population neural activity patterns before and after the lesions using electrophysiology. This new technique addresses a recent challenge in the field and provides a precious opportunity to study the natural reorganization/recovery at the neuronal population level after long-term lesions. It will help discover new causal insights investigating the neural circuits controlling behavior.

Comments on revised version:

We appreciate the revisions made by the authors in response to our comments on the previous version of their manuscript. They carefully addressed the majority of the concerns and performed additional experiments. The new figure illustrating the lesion volume as a function of electrolytic lesioning parameters provides a valuable reference for future experiments. In addition, the latest results on different versions of passive multielectrode probes, U-probe, demonstrate that the technique is applicable beyond the specific technical setup they employ. Overall, we believe that the revised manuscript is significantly improved.

Reviewer #2 (Public Review):

This work by Bray et al. presented a customized way to induce small electrolytic lesions in the brain using chronically implanted intracortical multielectrode arrays. This type of lesioning technique has the benefit of high spatial precision and low surgical complexity while allowing simultaneous electrophysiology recording before, during, and after the lesion induction. The authors have validated this lesioning method with a Utah array, both ex vivo and in vivo using pig models and awake-behaving rhesus macaques. Given its precision in controlling the lesion size, location, and compatibility with multiple animal models and cortical areas, the authors believe this method can be used to study cortical circuits in the presence of targeted neuronal inactivation or injury and to establish causal relationships before behavior and cortical activity.

Strengths:

- Overall the techniques, parameters, and data analysis methods are better described in the revised version.

- The authors added the section "Relationship Between Applied Current and Lesion Volume" as well as Figure 4 and 5 to address our comments regarding parameter testing. Multiple combinations of current amplitude and duration were tested and the induced lesion volumes were estimated, providing a better picture of why certain parameters were chosen for in vivo studies.

- The authors added Figure 7 which addressed our comment "more evidence is needed to suggest robust neuronal inactivation or termination in rhesus macaques after electrolytic lesioning." They went into more details to explain the observed changes in pairwise comparisons of spike waveforms (difference in projected radii). Particularly in Fig 7C, they identified a new cluster from the pre-post lesioning group, which effectively represented neuronal loss from the<br /> recorded population.

- The authors discussed their method in the context of other literature and stating its strength and limitation.

Major comments:

-The lack of histology limits the validation of lesion induction, ideally cell loss and neuronal loss in vivo needs to be quantified. In addition based on the lack of access to histology, it is not clear how the lesion volumes are calculated which also impacts the scientific rigor of the work. The authors mention that layers 2/3 and maybe 4 have been impacted. The lack of information on the extent of the lesion severely limits the use of their technique for neuroscience experiments.

-The lack of histology in combination with behavioral measures still limits the impact of the paper in the context of NHP research.

- Figure 5 involves fitting an exponential model to the generated lesion volume given the applied current amplitude and duration. However, the data from ex vivo sheep and pig cortex with the same current amplitude & three durations showed very large variability in lesion volume at Time = 2min (larger than the difference from 2 to ~2.2min). Very limited data points exist for the other two parameter combinations. These may suggest that the exponential fit is not the best model in this scenario.

- Regarding the comment on neuronal inactivation, the authors still did not show any evidence of single unit activity loss or changes in local field potential/multi-unit activity from the region being lesioned.

- Regarding this comment "The lesioning procedure was performed in Monkey F while sedated, but no data was presented for Monkey F in terms of lesioning parameters, lesion size, recorded electrophysiology, histological, or behavioral outcomes. It is also unclear if Monkey F was in a terminal study" the authors explained that "a lesion was performed on a sedated rhesus macaque (monkey F) who was subsequently euthanized due to unrelated health complications, in order to further verify safety before use in awake-behaving rhesus" but still no histology data is shown regarding monkey F to demonstrate this verification. Given that NHPs are highly valuable resources, it's important to make use of all collected data and to show that the induced lesion is comparable to those in the pig cortex.

eLife assessment

In this study, camera trapping and species distribution models are used to show that human disturbance in mountain forests in the eastern Himalayas pushes medium-sized and large mammal species into narrower habitat space, thus increasing their co-occurrence. While the collected data provide a useful basis for further work, the study presents incomplete evidence to support the claim that increased co-occurrence may indicate positive interactions between species.

Reviewer #1 (Public Review):

Summary:

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

Strengths:

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

Weaknesses:

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

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

Reviewer #2 (Public Review):

Summary:

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

Strengths:

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

Weaknesses:

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

(1) At the lower level, we would need to better understand what the individual species do and "like" (their environmental niche).

(2) The hypothesis clearly asks for an analysis of the statistical interaction between human disturbance and co-occurrence. Yet, the study is not set up in a way to test this directly.

The hypotheses point towards presenting the spatial and the temporal niche, and how it changes, species for species, under human disturbance. To this, one could then add the layer of interspecific associations.

The change in activity and space use could be analysed by looking at the activity times and spatial distribution directly. If biotic interactions change along the disturbance gradient, then observed data are already the outcome of such changed interactions. We thus cannot use the data to infer them! But we can show, for each species, that the habitat preferences change along the disturbance gradient - or not, as the case may be.

The per-species models are simplistic: the predictors are only linear, and there are no statistical interactions. It is unclear how spatial autocorrelations of residuals were treated, although they form the basis for the association analysis. Why are times of day and day of the year not included as predictors IN INTERACTION with niche predictors and human disturbance, since they represent the temporal dimension on which niches are hypothesised to change?

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

Author response:

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

Reviewer #1 (Public Review):

Summary:

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

Strengths:

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

Weaknesses:

The authors use the terms associations and interactions interchangeably.

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

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

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

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

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

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

Results were considered to be non-random if correlation coefficients (for spatial association) or overlap (for temporal association) fell outside of 95% Confidence Intervals. This is now stated clearly in the Methods section.  In Figure 3—figure supplement 1-3 and Figure 4—figure supplement 1-3, p<0.01 levels are also presented.

The obtained results support the conclusions of the study.

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

Reviewer #2 (Public Review):

Summary:

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

Strengths:

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

Weaknesses:

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

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

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

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

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

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

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

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

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

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

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

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

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

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

The coefficient estimates for response to environmental variables used in the JSDM are provided in Figure 2 and Figure 2—source data 1.

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

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

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

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

Reviewer #1 (Recommendations For The Authors):

L27 indicates instead of "idicates".

We thank the reviewer for catching that error.

L64 I would refer to potential interactions or just associations. It is always hard to provide evidence for the existence of true interactions.

We have revised to “potential interactions” to qualify this statement.

L69 Suggestion: distort instead of upset.

We thank the reviewer for catching that error.

L70-71 Here, authors use the term associations. Please, be consistent with the terminology throughout the manuscript.

We thank the reviewer for raising this important point.  The term “co-occurrence” appears to be used inconsistently in the literature, so we have tried to refer to it only when referencing the work of us. For us, co-occurrence means “spatial overlap” without qualification as to whether it is caused by interaction or simply by similar responses to environmental factors (see Blanchet et al. 2020, Argument 1). In our view, interactions refer to biotic effects like predation, competition, commensalism, etc., while associations are the statistical footprint of these processes.   In keeping with this understanding, in Line 73, we changed "association" to the stronger word "interaction," but in Line 76, we keep the words "spatiotemporal association", which is presumed to be the result of those interactions. In Line 91, we have changed “interactions” to “associations,” as we do not believe interactions were demonstrated in that study. 

L76 "Species associations are not necessarily fixed as positive or negative..." This sentence is misleading. I would say that species associations can vary across time and space, for instance along an environmental gradient.

We thank the reviewer for pointing out the potential for confusion.  In Line 79, we have changed as suggested.

L78 "Associations between free-ranging species are especially context-dependent" Loose sentence. Please, explain a bit further.

We have changed the sentence to be more specific; ”Interactions are known to be context-dependent; for example, gradients in stress are associated with variation in the outcomes of pairwise species interactions.”

L83-85 This would be a good place to introduce the 'stress gradient' hypothesis, which has also been applied to faunal communities in a few studies. According to this hypothesis, the incidence of positive associations should increase as environmental conditions harden.

In our review of the literature, we find that the stress gradient hypothesis is somewhat controversial and does not receive strong support in vertebrates.  We have added the phrase “…the controversial stress-gradient hypothesis predicts that positive associations should increase as environmental conditions become more severe…”

L86-88 Well, overall, the number of studies examining spatiotemporal associations in vertebrates is relatively small. That is, bird associations have not received much more attention than those of mammals. I find this introductory/appealing paragraph a bit rough. I think the authors can do better and find a better justification for their work.

We thank the reviewer for the comments.  We have rewritten the paragraph extensively to make it clearer and to provide a stronger justification for the study.

L106 "[...] resulting in increased positive spatial associations between species" I'd say that habitat shrinking would increase the level of species clustering or co-occurrence, but in my opinion, not necessarily the incidence of positive associations. It is not clear to me if the authors use positive associations as a term analogous to co-occurrence.

We thank the reviewer for raising this very important distinction.  Habitat shrinking would increase levels of species co-occurrence, but this is not particularly interested.  We wanted to test whether there were effects on species interactions, as revealed by associations.  We find that the terms association and co-occurrence are used somewhat loosely in the literature and so have made some new effort to clarify and systematize this in the manuscript.  For example, there appear to be a differences in the way “co-occurrence” is used in Boron 2023 and in Blanchet 2020. We do not use the term "positive spatial association" as analogous to "spatial co-occurrence.". Spatial co-occurrence, which for us has the meaning of spatial overlap, could simply be the result of similar reactions to environmental co-variates, not reflecting any biotic interaction. Joint Species Distribution Models enable the partitioning of spatial overlap and segregation into that which can be explained by responses to known environmental factors, and that which cannot be explained and thus might be the result of biotic interactions.  It is only the latter that we are calling spatial association, which can be positive or negative.   These associations may be the statistical footprint of biotic interactions.

Results:

Difference between random and non-random association patterns. It is not clear to me if the reported associations are significant or not. The authors only report the sign of the association (either positive or negative) but do not clarify if these associations indicate that two species coexist more or less than expected by chance. In my opinion, that is the difference between true ecological associations (e.g., via facilitation or competition effects) and random co-existence patterns. This is paramount and should be addressed in a new version of the manuscript.

This information is provided in Figure 3—figure supplement 1,2,3 and Figure 4—figure supplement 1,2,3.  This is referenced in the text as follows, “… correlation coefficients for 18 species pairs were positive and had a 95 % CI that did not overlap zero, and the number increased to 65 in moderate modifications but dropped to 29 at higher modifications" and so on. This criterion for significance (ie., greater than expected by chance) is now stated at the end of the Materials and methods.  In Figure 3—figure supplement 1,2,3 and Figure 4—figure supplement 1,2,3, those correlations that were significant at p<0.01 are also shown.

I am also missing a more ecological explanation for the observed findings. For instance, the top-ranked species in terms of negative associations is the red fox, whereas the muntjac seems to be the species whose presence can be used as an indicator for that of other species. What are the mechanisms underlying these patterns? Do red foxes compete for food with other species? Do the species that show positive associations (red goral, muntjac) have traits or a diet that are more different from those of other species? More discussion on these aspects (role of traits and the trophic niche) would be necessary to better understand the obtained results.

The purpose of this paper was to test the compression hypotheses, and we have tried to keep that as the focus.  However, the analysis does open up interesting lines of inquiry for future research to decipher the details of the interactions between species and the mechanisms by which human disturbance facilitates or disrupts these interactions. The reviewer raises some interesting possibilities, but at this point, any discussion along these lines would be largely speculation and could lengthen the paper without great benefit. 

Reviewer #2 (Recommendations For The Authors):

The manuscript should be accompanied by all data and code of analysis.

All data and RScripts have been made available in Science Data Bank: https://doi.org/10.57760/sciencedb.11804.

The sentence "not much is known" is weak: it suggests the authors did not bother to quantify what IS known, and simply waved any previous knowledge aside. Surely we have some ideas about who preys on whom, and which species have overlapping resource requirements (e.g., due to jaw width). For those, we would expect a particularly strong signal, if the association is indeed indicative of interactions.

We believe that the reviewer is referring to the statement in Line 90-92 about the lack of understanding of the resilience of terrestrial mammal associations to human disturbance.  We have added a reference to one very recent publication that addresses the issue (Boron et al., 2023), but otherwise we stand by our statement. We have, however, added a qualifier to make it clear that we did indeed look for previous knowledge; "However, a review of the literature indicates that ...."

Figures:

Fig. 1. This reviewer considers that this is too trivial and should be deleted.

This is a graphical statement of the hypotheses and may be helpful to some readers.

Fig. 2. Using points with error bars hides any potential information.

Done as suggested.

That only 4 predictors are presented is unacceptably oversimplified.

Only 4 predictors are included because, in previous work, we found that adding additional predictors or interactions did little to improve the model’s performance (Li et al. 2018, 2021 and 2022) and could lead to over-fitting.

Fig. 5. and 6. aggregate extremely strongly over species; it remains unclear which species contribute to the signal, and I guess most do not.

The number of detection events presented in Table 1 should help to clarify the relative contribution of each species to the data presented in Figures 5 and 6.

This reviewer considers that the introduction 'oversells' the paper.

L55: can you give any such "unique ecological information"

L60: Lyons et al. (Kathleen is the first name) has been challenged by Telford et al. (2016 Nature) as methodologically flawed.

The first name has been deleted.  The methodological flaw has to do with interpretation of the fossil record and choice of samples, not with the need to partition shared environmental preferences and interactions.

L61 contradicts line 64: Blanchet et al. (2022, specifying some arguments from Dormann et al. 2018 GEB) correctly point out that logically one cannot infer the existence or strength from co-occurrence data. It is thus wrong to then claim (citing Boron et al.) that such data "convey key information about interactions". The latter statement is incorrect. A tree and a beetle can have extremely high association and nothing to do with each other. Association does not mean anything in itself. When two species are spatially and temporally non-overlapping, they can exhibit perfect "anti-association", yet, by the authors' own definition, cannot interact.

We believe that the reviewer’s concerns arise from a misunderstanding of how we use the term association.  In our usage, an association is not the same as co-occurrence or overlap, which may simply be the result of shared responses to environmental variables.  The co-occurring tree and beetle would not be found to have any association in our analysis, only shared environmental sensitivities.  In contrast, associations can be the statistical footprint of interactions, and would be overlaid onto any overlap due to similar responses to the environment.  In the case of negative associations, such as might be the result of competitive exclusion or avoidance of predators, the two species would share environmental responses but show lower than expected spatial overlap.  Even though they might be only rarely found in the same vicinity, they would indeed be interacting when they were together.

Joint Species Distribution Models "allow the partitioning of the observed correlation into that which can be explained by species responses to environmental factors... and that which remains unexplained after controlling for environmental effects and which may reflect biotic interactions." (Garcia Navas et al. 2021). It is the latter that we are calling “associations.”

L63: Gilbert reference: Good to have a reference for this statement.

This point is important, but the reviewer’s comments below have made it clear that it is even more important to point out that strong interactions should be expected to lead to significant associations.  We have added a statement to clarify this.

L70-72: Incorrect, interactions play a role, not associations (which are merely statistical).

In this, we agree, and we have revised the statement to refer to interactions, not associations. In our view, an interaction is a biological phenomenon, while an association is the resulting statistical signal that we can detect.

L75: Associations tell us nothing, only interactions do. Since these can not be reliably inferred, this statement and this claim are wrong.

We thank the reviewer for raising this point, but we beg to disagree. Strong interactions should be expected to lead to significant associations that can be detected in the data. Associations, which can be measured reliably, are the evidence of potential interactions, and hence associations can tell us a great deal.  We have added a note to this effect after the Gilbert reference above to clarify this point.

However, we do accept that associations must be interpreted with caution. As Blanchet et al. 2020 explain, " …the co-occurrence signals (e.g. a significant positive or negative correlation value) estimated from these models could originate from any abiotic factors that impact species differently. Therefore, this correlation cannot be systematically interpreted as a signal of biotic interactions, as it could instead express potential non-measured environmental drivers (or combinations of them) that influence species distribution and co-distribution.”  Or alternatively an association could be the result of interaction with a 3rd species. 

L87: Regarding your claim, how would you know you DO understand? For that, you need to formulate an expectation before looking at the data and then show you cannot show what you actually measure. (Jaynes called this the "mind-projection fallacy".)

We are not sure if the reviewer is criticizing our paper or the entire field of community ecology.  Perhaps it is the statement that “….resilience of interspecific spatiotemporal associations of terrestrial mammals to human activity remains poorly understood….”  Since we are confident that the reviewer believes that mammals do interact, we guess that it is the term “association” that is questioned.  We have revised this to “…the impacts of human activity on interspecific interactions of terrestrial mammals remains poorly understood…” 

In this particular case, we did formulate an expectation before looking at the data, in the form of the two formal hypotheses that are clearly stated in the Introduction and illustrated in Figure 1. If the hypotheses had not been supported, then we would have accepted that we do not understand. But as the data are consistent with the hypotheses, we submit that we do understand a bit more now.

eLife assessment

This study provides a valuable resource by thoroughly benchmarking multiple sequencing-based tRNA quantification methods. The suggested best practice is supported by solid evidence from in silico experiments in multiple scenarios. The major weakness of the manuscript is the incomplete validation of newly generated experimental datasets.

Reviewer #1 (Public Review):

Summary:

In the manuscript titled "Benchmarking tRNA-Seq quantification approaches by realistic tRNA-Seq data simulation identifies two novel approaches with higher accuracy," Tom Smith and colleagues conducted a comparative evaluation of various sequencing-based tRNA quantification methods. The inherent challenges in accurately quantifying tRNA transcriptional levels, stemming from their short sequences (70-100nt), extensive redundancy (~600 copies in human genomes with numerous isoacceptors and isodecoders), and potential for over 100 post-transcriptional chemical modifications, necessitate sophisticated approaches. Several wet-experimental methods (QuantM-tRNA, mim-tRNA, YAMAT, DM-tRNA, and ALL-tRNA) combined with bioinformatics tools (bowtie2-based, SHRiMP, and mimseq) have been proposed for this purpose. However, their practical strengths and weaknesses have not been comprehensively explored to date. In this study, the authors systematically assessed and compared these methods, considering factors such as incorrect alignments, multiple alignments, misincorporated bases (experimental errors), truncated reads, and correct assignments. Additionally, the authors introduced their own bioinformatic approaches (referred to as Decision and Salmon), which, while not without flaws (as perfection is unattainable), exhibit significant improvements over existing methods.

Strengths:

The manuscript meticulously compares tRNA quantification methods, offering a comprehensive exploration of each method's relative performance using standardized evaluation criteria. Recognizing the absence of "ground-truth" data, the authors generated in silico datasets mirroring common error profiles observed in real tRNA-seq data. Through the utilization of these datasets, the authors gained insights into prevalent sources of tRNA read misalignment and their implications for accurate quantification. Lastly, the authors proposed their downstream analysis pipelines (Salmon and Decision), enhancing the manuscript's utility.

Weaknesses:

As discussed in the manuscript, the error profiles derived from real-world tRNA-seq datasets may still harbor biases, as reads that failed to "align" in the analysis pipelines were not considered. Additionally, the authors did not validate the efficacy of their "best practice" pipelines on new real-world datasets, preferably those generated by the authors themselves. Such validation would not only confirm the improvements but also demonstrate how these pipelines could alter biological interpretations.<br /> Because tRNA-sequencing methods have not been widely used (compared to mRNA-seq), many readers would not be familiar with the characteristics of different methods introduced in this study (QuantM-tRNA, mim-tRNA, YAMAT, DM-tRNA, and ALL-tRNA; bowtie2-based, SHRiMP, and mimseq; what are the main features of "Salmon?"). The manuscript will read better when the basic features of these methods are described in the manuscript, however brief.

Reviewer #2 (Public Review):

Summary:

The authors provided benchmarking study results on tRNA-seq in terms of read alignment and quantification software with optimal parameterization. This result can be a useful guideline for choosing optimal parameters for tRNA-seq read alignment and quantification.

Strengths:

Benchmarking results for read alignment can be a useful guideline to choose optimal parameters and mapping strategy (mapping to amino acid) for various tRNAseq.

Weaknesses:

The topic is highly specific, and the novelty of the analysis might not be widely useful for general readers.

Some details of the sequencing data analysis pipeline are not clear for general readers:

(1) The explanation of the parameter D for bowtie2 sounds ambiguous. "How much effort to expend" needs to be explained in more detail.

(2) Please provide optimal parameters (L and D) for tRNA-seq alignment.

(3) I think the authors chose L=10 and D=100 based on Figure 1A. Which dataset did you choose for this parameterization among ALL-tRNAseq, DM-tRNAseq, mim-tRNAseq, QuantM-tRNA-seq, and YAMAT-seq?

(4) Salmon does not need a read alignment process such as Bowtie2. Hence, it is not clear "Only results from alignment with bowtie2" in Figure legend for Figure 4a.

Author response:

We thank the reviewers for their critical appraisal of our manuscript. We will address the points of confusion and/or lack of clarity in a revised manuscript. We agree with reviewer 1 that applying the best practice pipeline(s) on new experimental data and comparing this approach with current practices would be a useful demonstration of how this alters the biological interpretation. This is something we are in the process of completing but believe this is best addressed in a separate manuscript where we can focus on the associated biological findings, allowing this manuscript to remain focused on the accurate quantification of tRNA-Seq data.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Eaton et al. examine the regulation of transcription directionality using a powerful genomic approach (more about the methodology below). Their data challenge the notion that the polyadenylation signal-reading Cleavage and Polyadenylation (CPA) complex is responsible for controlling promoter directionality by terminating antisense transcription. Namely, depletion of the required CPA factor RBBP6 has little effect on antisense transcription measured by POINT. They find instead that initiation is intrinsically preferential in the sense direction and additionally maintained by the activities of an alternative processing complex called Integrator, together with the kinase CDK9. In the presence of CDK9 activity, depletion of Integrator endoribonuclease INTS11 leads to globally increased transcription in the antisense direction, and minor effects in the sense direction. However, CDK9 inhibition reveals that sense transcription is also sensitive to INS11 depletion. The authors suggest that CDK9 activity is stronger in the sense direction, preventing INTS11-mediated premature termination of sense transcrpts.

Strengths:

The combination of acute depletion of the studied factors using degron approaches (important to limit possible secondary effects), together with novel and very sensitive nascent transcriptomics methods POINT and sPOINT is very powerful. The applied spike-in normalization means the analysis is more rigorous than most. Using this methodology allowed the authors to revisit the interesting question of how promoter/transcription directionality is determined.

The data quality appears very good and the fact that both global analysis as well as numerous gene-specific examples are shown makes it convincing.

The manuscript is well written and hence a pleasure to read.

We appreciate this positive assessment.

Weaknesses:

I am slightly worried about the reproducibility of the data - it is unclear to me from the manuscript if and which experiments were performed in replicate (lack of table with genomic experiments and GEO access, mentioned in more detail in below recommendations to authors), and the methods could be more detailed.

All sequencing data was deposited with GEO. Multiple biological replicates were performed for each sequencing experiment.  Bigwig files are presented as a table in the GEO submissions. This data has now been made public.

A separate discussion section would be useful, particularly since the data provided challenge some concepts in the field. How do the authors interpret U1 data from the Dreyfuss lab in light of their results? How about the known PAS-density directionality bias (more PAS present in antisense direction than in sense) - could the differential PAS density be still relevant to transcription directionality?

As suggested, we have expanded our discussion to relate our findings to existing data. We think the results from the Dreyfuss lab are very important and highlight the role of U1 snRNA in enforcing transcriptional elongation.  It does this in part by shielding PAS sequences.  Recent work from our lab also shows that U1 snRNA opposes the Restrictor complex and PNUTS, which otherwise suppress transcription (Estell et al., Mol Cell 2023).  Most recently, the Adelman lab has demonstrated that U1 snRNA generally enhances transcription elongation (Mimoso and Adelman., Mol Cell 2023).  Our work does not challenge and is not inconsistent with these studies.

The role of U1 in opposing PAS-dependent termination inspired the idea that antisense transcriptional termination may utilise PASs.  This was because such regions are rich in AAUAAA and comparatively poor in U1 binding sites. However, our RBBP6 depletion and POINT-seq data suggest that PAS-dependent termination is uncommon in the antisense direction. As such, other mechanisms suppress antisense transcription and influence promoter directionality. In our paper, we propose a major role for the Integrator complex.

We do not completely rule out antisense PAS activity and discuss the prior work that identified polyadenylated antisense transcripts. Nevertheless, this was detected by oligo-dT primed RT-PCR/Northern blotting, which cannot determine the fraction of non-polyadenylated RNA that could result from PAS-independent termination (e.g. by Integrator).  To do that requires an analysis of total nascent transcription as achieved by our POINT-seq.  Based on these experiments, Integrator depletion has a greater impact on antisense transcription than RBBP6 depletion. 

I find that the provided evidence for promoter directionality to be for the most part due to preferential initiation in the sense direction should be stressed more. This is in my eyes the strongest effect and is somehow brushed under the rug.

We agree that this is an important finding and incorporated it into the title and abstract.  As the reviewer recommends, we now highlight it further in the new discussion.

References 12-17 report an effect of Integrator on 5' of protein-coding genes, while data in Figure 2 appears contradictory. Then, experiments in Figure 4 show a global effect of INST11 depletion on promoter-proximal sense transcription. In my opinion, data from the 2.5h time-point of depletion should be shown alongside 1.5h in Figure 2 so that it is clear that the authors found an effect similar to the above references. I find the current presentation somehow misleading.

We are grateful for this suggestion and present new analyses demonstrating that our experiment in Figure 2 concurs with previous findings (Supplemental Figures 2A and B). Our original heatmap (Figure 2E) shows a very strong and general antisense effect of INTS11 loss. On the same scale, the effects in the sense direction are not as apparent, which is also the case using metaplots.  New supplemental figure 2A now shows sense transcription from this experiment in isolation and on a lower scale, demonstrating that a subset of genes shows promoter-proximal increases in transcription following INTS11 depletion.  This is smaller and less general than the antisense effect but consistent with previous findings.  Indeed, our new analysis in supplemental figure 2B shows that affected protein-coding genes are lowly expressed, in line with Hu et al., Mol Cell 2023. This explains why a sense effect is not as apparent by metaplot, for which highly expressed genes contribute the most signal.

As a result of our analyses, we are confident that the apparently larger effect at the 2.5hr timepoint (Figure 4) that we initially reported is due to experimental variability and not greater effects of extended INTS11 depletion. Overlaying the 1.5h and 2.5h datasets (Supplemental Figure 4B) revealed a similar number of affected protein-coding genes with a strong (83%) overlap between the affected genes.  To support this, we performed qPCR on four affected protein-coding transcripts which revealed no significant difference in the level of INTS11 effect after 2.5h vs 1.5h (Supplemental Figure 4C).

We now present data for merged replicates in Figures 2 and 4 which reveal very similar average profiles for -INTS11 vs +INTS11 at both timepoints. Overall, we believe that we have resolved this discrepancy by showing that it amounts to experimental variability and because the most acutely affected protein-coding genes are lowly expressed. As detailed above, we show this in multiple ways (and validate by qPCR) We have revised the text accordingly and removed our original speculation that differences reflected the timeframe of INTS11 loss.

Conclusion/assessment:

This important work substantially advances our understanding of the mechanisms governing the directionality of human promoters. The evidence supporting the claims of the authors is compelling, with among others the use of advanced nascent transcriptomics including spike-in normalization controls and acute protein depletion using degron approaches.

In my opinion, the authors' conclusions are in general well supported.

Not only the manuscript but also the data generated will be useful to the wide community of researchers studying transcriptional regulation. Also, the POINT-derived novel sPOINT method described here is very valuable and can positively impact work in the field.

We are grateful for the reviewers' positive assessment of our study.

Reviewer #2 (Public Review):

Summary:

Eaton and colleagues use targeted protein degradation coupled with nascent transcription mapping to highlight a role for the integrator component INST11 in terminating antisense transcription. They find that upon inhibition of CDK9, INST11 can terminate both antisense and sense transcription - leading to a model whereby INST11 can terminate antisense transcription and the activity of CDK9 protects sense transcription from INST11-mediated termination. They further develop a new method called sPOINT which selectively amplifies nascent 5' capped RNAs and find that transcription initiation is more efficient in the sense direction than in the antisense direction. This is an excellent paper that uses elegant experimental design and innovative technologies to uncover a novel regulatory step in the control of transcriptional directionality.

Strengths:

One of the major strengths of this work is that the authors endogenously tag two of their proteins of interest - RBBP6 and INST11. This tag allows them to rapidly degrade these proteins - increasing the likelihood that any effects they see are primary effects of protein depletion rather than secondary effects. Another strength of this work is that the authors immunoprecipitate RNAPII and sequence extracted full-length RNA (POINT-seq) allowing them to map nascent transcription. A technical advance from this work is the development of sPOINT which allows the selective amplification of 5' capped RNAs < 150 nucleotides, allowing the direction of transcription initiation to be resolved.

We appreciate this positive assessment.

Weaknesses:

While the authors provide strong evidence that INST11 and CDK9 play important roles in determining promoter directionality, their data suggests that when INST11 is degraded and CDK9 is inhibited there remains a bias in favour of sense transcription (Figures 4B and C). This suggests that there are other unknown factors that promote sense transcription over antisense transcription and future work could look to identify these.

We agree that other (so far, unknown) factors promote sense transcription over antisense, which was demonstrated by our short POINT.  We have provided an expanded discussion on this in the revision. In our opinion, demonstrating that sense transcription is driven by preferential initiation in that direction is a key finding and we agree that the identification of the underlying mechanism constitutes an interesting avenue for future study.

Reviewer #3 (Public Review):

Summary:

Using a protein degradation approach, Eaton et al show that INST11 can terminate the sense and anti-sense transcription but higher activity of CDK9 in the sense direction protects it from INS11-dependent termination. They developed sPOINT-seq that detects nascent 5'-capped RNA. The technique allowed them to reveal robust transcription initiation of sense-RNA as compared to anti-sense.

Strengths:

The strength of the paper is the acute degradation of proteins, eliminating the off-target effects. Further, the paper uses elegant approaches such as POINT and sPOINT-seq to measure nascent RNA and 5'-capped short RNA. Together, the combination of these three allowed the authors to make clean interpretations of data.

We appreciate this positive assessment.

Weaknesses:

While the manuscript is well written, the details on the panel are not sufficient. The methods could be elaborated to aid understanding. Additional discussion on how the authors' findings contradict the existing model of anti-sense transcription termination should be added.

We have added more detail to the figure panels, which we hope will help readers to navigate the paper more easily. Specifically, the assay employed for each experiment is indicated in each figure panel. As requested, we provide a new and separate discussion section in the revision.

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

Congratulations on this important piece of work!

Some specific suggestions.

MAJOR

-The data are not available (Accession "GSE243266" is currently private and is scheduled to be released on Sep 01, 2026.) This should be corrected and as a minimum, the raw sequencing files as well as the spike-in scaled bigwig files should be provided in GEO.

We have made the data public. Raw and bigwig files are provided as part of the GEO upload.

MINOR

- It would be useful for readers if you could include catalog numbers of the reagents used in the study.

We have included this information in our revision.

- A table in experimental procedures summarizing the genomic experiments performed in this study as well as published ones reanalyzed here would be helpful.

This is now provided as part of the resources table.

- It would be easier for reviewers to evaluate the manuscript if the figure legends were included together with the figures on one page. This is now allowed by most journals.

We have used this formatting in the revision.

- Providing some captions for the results sections would be helpful.

We have included subheadings as suggested.

Reviewer #2 (Recommendations For The Authors):

Generally, I would suggest writing the experiment-type above panels where it is not immediately obvious what they are so a reader can appreciate the figures without referencing the legend. E.g. write POINT-seq on Figure 1B just to make it obvious to someone looking at the figures what methodology they are looking at. Likewise, you could write RNAPII ChIP-seq for Supplementary Figures 3D and 3E.

We have carried out this recommendation.

Can a y-axis be indicated on POINT-seq genome browser tracks? This could make them easier to interpret.

Y-axis scales are provided as RPKM as stated in the figure legends.

The authors could address/speculate in the text why there is less POINT-seq signal for the antisense transcript in the treatment condition in Figure 1B? Or could consider including a different example locus where this is not the case for clarity.

Acute depletion of poly(A) factors (like RBBP6) results in a strong read-through beyond the poly(A) signal of protein-coding genes as Figure 1 shows.  However, it also causes a reduction in transcription levels, which can be seen in the figure and is correctly noted by the reviewer in this comment.  We see this with other poly(A) factor depletions (e.g. CPSF73 and CPSF30 – Eaton et al., 2020 and Estell et al., 2021) and other labs have observed this too (e.g for CPSF73-dTAG depletion (Cugusi et al., Mol Cell 2022)).  Plausible reasons include a limited pool of free RNAPII due to impaired transcriptional termination or limited nucleotide availability due to their incorporation within long read-through transcripts. For these reasons, we have retained the example in Figure 1B as a typical representation of the effect. Moreover, the heatmap in Figure 1D fairly represents the spectrum of effects following RBBP6 loss – highlighting the strong read-through beyond poly(A) signals and the marginal antisense effects.

"The established effect of INTS11 at snRNAs was detected in our POINT-seq data and demonstrates the efficacy of this approach (Figure 2B)." The authors could explain this point more clearly in the text and describe the data - e.g. As expected, depletion of INTS11 leads to increased POINT-seq signal at the 3' end of snRNAs, consistent with defects in transcriptional termination. This is highlighted by the RNU5A-1 and RNU5B-1 loci (Figure 2B).

We agree and have added more context to clarify this.

I would suggest adjusting the scale of the heatmap in Figure 2E - I think it would be easier to interpret if the value of 0 was white - with >0 a gradient of orange and <0 a gradient of blue (as is done in Figure 1C). I think making this change would make the point as written in the text clearer i.e. "heatmap analysis demonstrates the dominant impact of INTS11 on antisense versus sense transcription at most promoters (Figure 2E)." I'm assuming most of the sense transcription would be white (more clearly unchanging) when the scale is adjusted.

We agree and have done this. The reviewer is correct that most sense transcription is unchanged by INTS11 loss.  However, as we alluded to in the original submission, a subset of transcripts shows a promoter-proximal increase after INTS11 depletion. We have expanded the analyses of this effect (see responses to other comments) but stress that it is neither as general nor as large as the antisense effect.

The authors make the point that there is mildly increased transcription over the 5' end of some genes upon INST11 depletion and show a track (Supplementary Fig 2A). It is not immediately obvious from the presentation of the meta-analysis in Figure 2D how generalisable this statement is. Perhaps the size of the panel or thickness of the lines in Figure 2D could be adjusted so that the peak of the control (in blue) could be seen. Perhaps an arrow indicating the peak could be added? I'm assuming the peak at the TSS is slightly lower in the control compared to INST11 depletion based on the authors' statement.

We have provided multiple new analyses of this data to highlight where there are promoter-proximal effects of INTS11 loss in the sense direction.  Please see our response to the public review of reviewer 1 and new supplemental figures 2A, 2B, 4A and 4B which highlight the sense transcription increased in the absence of INTS11.

The authors label Figure 4 "Promoters lose their directionality when CDK9 is inhibited" - but in INST11 depleted cells treated with CDK9i they find that there still is a bias towards sense transcription. Suggested edit "Some promoter directionality is lost when CDK9 is inhibited" or similar.

We agree and have made this change.

The authors conclude that INTS11-mediated effects are the result of perturbation of the catalytic activities of Integrator, the authors should perform rescue experiments with the catalytically dead E203Q-INTS11 mutant.

This is a very good suggestion and something we had intended to pursue.  However, as we will describe below (and shown in Supplemental Figure 4G), there were confounding issues with this experiment.

The E203Q mutant of INTS11 is widely used in the literature to test for catalytic functions of INTS11.  However, we have found that this mutation impairs the ability of INTS11 to bind other Integrator modules in cells. Based on co-immunoprecipitation of flag-tagged WT and E203Q derivatives, INTS1 (backbone module), 10 (tail module), and 8 (phosphatase module) all show reduced binding to E203Q vs. WT. Because E203Q INTS11 is defective in forming Integrator complexes, rescue experiments might not fully distinguish the effects of INTS11 activity from those caused by defects in complex assembly. While this may at first seem unexpected, in the analogous 3’ end processing complex, catalytic mutants of CPSF73 (which is highly related to INTS11) negatively affect its interaction with other complex members (Kolev and Steitz, EMBO Reports 2005).

We hypothesise that INTS11 activity is most likely involved in attenuating promoter-proximal transcription, but we cannot formally rule out other explanations and discuss this in our revision. Regardless of how INTS11 attenuates transcription, our main conclusion is on its requirement to terminate antisense transcription whether this involves its cleavage activity or not.

The authors suggest that CDK9 modulates INTS11 activity/assembly and suggest this may be related to SPT5. Is there an effect of CDK9 inhibition on the snRNA's highlighted in Figure 2B?

We believe that snRNAs are different from protein-coding genes concerning CDK9 function. Shona Murphy’s lab previously showed that, unlike protein-coding genes, snRNA transcription is insensitive to CDK9 inhibition, and that snRNA processing is impaired by CDK9 inhibition (Medlin et al., EMBO 2003 and EMBO 2005).  We reproduce these findings by metaanalysis of 15 highly expressed and well-separated snRNAs and by qRT-PCR of unprocessed RNU1-1, RNU5A-1 and RNU7-1 snRNA following CDK9 inhibition. We observe snRNA read-through by POINT-seq following INTS11 loss whether CDK9 is inhibited or not (left panel, below). Note the higher TES proximal signal in CDK9i conditions, which likely reflects the accumulation of unprocessed snRNA as validated by qPCR for three example snRNAs (right panel, below).

Author response image 1.

For Figure 4, would similar results be observed using inhibitors targeting other transcriptional CDKs such as CDK7,12/13?

In response to this suggestion, we analysed four selected protein-coding transcripts (the same 4 that we used to validate the CDK9i results) by qRT-PCR in a background of CDK7 inhibition using the THZ2 compound (new Supplemental Figure 4E).  THZ2 suppresses transcription from these genes as expected.  Interestingly, expression is restored by co-depleting Integrator, recapitulating our findings with CDK9 inhibition.  As CDK7 is the CDK-activating kinase for CDK9, its inhibition will also inhibit CDK9 so THZ2 may simply hit this pathway upstream of where CDK9 inhibitors.  Second, CDK7 may independently shield transcription from INTS11.  We allude to both interesting possibilities.

What happens to the phosphorylation state of anti-sense engaged RNAPII when INTS11 is acutely depleted and/or CDK9 is inhibited? This could be measured by including Ser5 and Ser2 antibodies in the sPOINT-seq assay and complemented with Western Blot analysis.

We have performed the western blot for Ser5 and Ser2 phosphorylation as suggested.  Both signals are mildly enhanced by INTS11 loss, which is consistent with generally increased transcription.  Ser2p is strongly reduced by CDK9 inhibition, which is consistent with the loss of nascent transcription in this condition.  Interestingly, both modifications are partly recovered when INTS11 is depleted in conjunction with CDK9 inhibition. This is consistent with the effects that we see on POINT-seq and shows that the recovered transcription is associated with some phosphorylation of RNAPII CTD.  This presumably reflects the action(s) of kinases that can act redundantly with CDK9.

We have not performed POINT-seq with Ser5p and Ser2p antibodies under these various conditions.  Our rationale is that our existing data uses an antibody that captures all RNAPII (regardless of its phosphorylation status), which we feel most comprehensively assays transcription in either direction. Moreover, the lab of Fei Chen (Hu et al., Mol Cell 2023) recently published Ser5p and Ser2p ChIP-seq following INTS11 loss. By ChIP-seq, they observe a bigger increase in antisense RNAPII occupancy vs. sense providing independent and orthogonal support for our POINT-seq data.  Interestingly, this antisense increase is not paralleled by proportional increases in Ser5p or Ser2p signals.  This suggests that the unattenuated antisense transcription resulting from INTS11 loss does not have high Ser5p or Ser2p.  Since CDK7 and 9 are major Ser5 and 2 kinases, this supports our model that their activity is less prevalent for antisense transcription.  We now discuss these data in our revision.   

The HIV reporter RNA experiments should be performed with the CDK9 inhibitor added to the experimental conditions. Presumably CDK9 inhibition would result in no upregulation of the reporter upon addition of TAT and/or dTAG. Perhaps the amount of TAT should be reduced to still have a dynamic window in which changes can be detected. It is possible that reporter activation is simply at a maximum. Can anti-sense transcription be measured from the reporter?

We have performed the requested CDK9 inhibitor experiment to confirm that TAT-activated transcription from the HIV promoter is CDK9-dependent (new supplemental figure 4F).  Consistent with previous literature on HIV transcription, CDK9 inhibition attenuates TAT-activated transcription.  Importantly, and in line with our other experiments, depletion of INTS11 results in significant restoration of transcription from the HIV promoter when CDK9 is inhibited. Thus, TAT-activated transcription is CDK9-dependent and, as for endogenous genes, CDK9 prevents attenuation by INTS11.

While TAT-activated transcription is high, we do not think that the plasmid is saturated. When considering this question, we revisited previous experiments using this system to study RNA processing (Dye et al., Mol Cell 1999, Cell 2001, Mol Cell 2006). In these cases, mutations in splice sites or polyadenylation sites have a strong effect on RNA processing and transcription around HIV reporter plasmids. Effects on transcription and RNA processing are; therefore, apparent in the appropriate context. In contrast, we find that the complete elimination of INTS11 has no impact on RNA output from the HIV reporter. Our original experiment assessing the impact of INTS11 loss in +TAT conditions used total RNA.  One possibility is that this allows non-nascent RNA to accumulate which might confound our interpretation of INTS11 effects on ongoing transcription.  However, the new experiment described in the paragraph above was performed on chromatin-associated (nascent) RNA to rule this out.  This again shows no impact of INTS11 loss on HIV promoter-derived transcription in the presence of TAT.

To our knowledge, antisense transcription is not routinely assayed from plasmids. They generally employ very strong promoters (e.g. CMV, HIV) to drive sense transcription.  Crucially, their circular nature means that RNAPII going around the plasmid could interfere with antisense transcription coming the other way which does not happen in a linear genomic context. This is why we restricted our use of plasmids to looking at the effects of stimulated CDK9 recruitment (via TAT) on transcription rather than promoter directionality.   

The authors should clearly state how many replicates were performed for the genomics experiments. Ideally, a signal should be quantified and compared statistically rather than relying on average profiles only.

We have stated the replicate numbers for sequencing experiments in the relevant figure legends. All sequencing experiments were performed in at least two biological replicates, but often three. In addition, we validated their key conclusions by qPCR or with orthogonal sequencing approaches.

Reviewer #3 (Recommendations For The Authors):

The authors provide strong evidence in support of their claims.

ChIP-seq of pol2S5 and S2 upon INST11 and CDK9 inhibition will strengthen the observation that transcription in the sense direction is more efficient.

We view the analysis of total RNAPII as the most unbiased way of establishing how much RNAPII is going one way or the other. Importantly, ChIP-seq was very recently performed for Ser2p and Ser5p RNAPII derivatives in the lab of Fei Chen (Hu et al., Mol Cell 2023). Their data shows that loss of INTS11 increases the occupancy of total RNAPII in the antisense direction more than in the sense direction, which is consistent with our finding. Interestingly, the increased antisense RNAPII was not paralleled with an increase in Ser2p or Ser5p. This suggests that, following INTS11 loss, the unattenuated antisense transcription is not associated with full/normal Ser2p or Ser5p. These modifications are normally established by CDK7 and 9; therefore, this published ChIP-seq suggests that they are not fully active on antisense transcription when INTS11 is lost. This supports our overall model that CDK9 (and potentially CDK7 as suggested for a small number of genes in new Supplemental Figure 4E) is more active in the sense direction to prevent INTS11-dependent attenuation. We now discuss these data in our revision.

In Supplementary Figure 2, the eRNA expression increases upon INST11 degradation, I wonder if the effects of this will be appreciated on cognate promoters? Can the authors test some enhancer:promoter pairs?

We noticed that some genes (e.g. MYC) that are regulated by enhancers show reduced transcription in the absence of INTS11. Whilst this could suggest a correlation, the transcription of other genes (e.g. ACTB and GAPDH) is also reduced by INTS11 loss although they are not regulated by enhancers.  A detailed and extensive analysis would be required to establish any link between INTS11-regulated enhancer transcription and the transcription of genes from their cognate promoters.  We agree that this would be interesting, but it seems beyond the scope of our short report on promoter directionality.

Line 111, meta plot was done of 1316 genes. Details on this number should be provided. Overall, the details of methods and analysis need improvement. The layout of panels and labelling on graphs can be improved.

We have now explained the 1316 gene set.  In essence, these are the genes separated from an expressed neighbour by at least 10kb.  This distance was selected because depletion of RBBP6 induces extensive read-through transcription beyond the polyadenylation site of protein-coding genes.  To avoid including genes affected by transcriptional read-through from nearby transcription units we selected those with a 10kb gap between them. This was the only selection criteria so is unlikely to induce any unintended biases. Finally, we have added more information to the figure panels and their legends, which we hope will make our manuscript more accessible.

eLife assessment

The important study uses a new experimental method to provide compelling evidence on how sense- and anti-sense transcription is differentially regulated. The method described here can generally be used to study the alterations in transcription. This paper will be of interest to scientists working in the gene regulation community.

Reviewer #1 (Public Review):

Summary:

In this manuscript, Eaton et al. examine the regulation of transcription directionality using a powerful genomic approach (more about the methodology below).<br /> Their data challenge the notion that the polyadenylation signal-reading Cleavage and Polyadenylation (CPA) complex is responsible for controlling promoter directionality by terminating antisense transcription. Namely, depletion of the required CPA factor RBBP6 has little effect on antisense transcription measured by POINT. They find instead that initiation is intrinsically preferential in the sense direction and additionally maintained by the activities of an alternative processing complex called Integrator, together with the kinase CDK9. In the presence of CDK9 activity, depletion of Integrator endoribonuclease INTS11 leads to globally increased transcription in the antisense direction, and minor effects in the sense direction. However, CDK9 inhibition reveals that sense transcription is also sensitive to INS11 depletion. The authors suggest that CDK9 activity is stronger in the sense direction, preventing INTS11-mediated premature termination of sense transcripts.

Strengths:

The combination of acute depletion of the studied factors using degron approaches (important to limit possible secondary effects), together with novel and very sensitive nascent transcriptomics methods POINT and sPOINT is very powerful. The applied spike-in normalization means the analysis is more rigorous than most. Using this methodology allowed the authors to revisit the interesting question of how promoter/transcription directionality is determined.

The data quality appears very good and the fact that both global analysis as well as numerous gene-specific examples are shown makes it convincing.

The manuscript is well written and hence a pleasure to read.

Weaknesses:

The bias in transcriptional initiation directionality remains to be elucidated.

Conclusion/assessment:

This important work substantially advances our understanding of the mechanisms governing the directionality of human promoters. The evidence supporting the claims of the authors is compelling, with a.o. the use of advanced nascent transcriptomics including spike-in normalization controls and acute protein depletion using degron approaches.

In my opinion the authors' conclusions are well supported.

Not only the manuscript but also the data generated will be useful to the wide community of researchers studying transcriptional regulation. Also, the POINT-derived novel sPOINT method described here is very valuable and can positively impact work in the field.

Reviewer #2 (Public Review):

Summary:

Eaton and colleagues use targeted protein degradation coupled with nascent transcription mapping to highlight a role for the integrator component INST11 in terminating antisense transcription. They find that upon inhibition of CDK9, INST11 can terminate both antisense and sense transcription - leading to a model whereby INST11 can terminate antisense transcription and the activity of CDK9 protects sense transcription from INST11-mediated termination. They further develop a new method called sPOINT which selectively amplifies nascent 5' capped RNAs and find that transcription initiation is more efficient in the sense direction than in the antisense direction. This is an excellent paper which uses elegant experimental design and innovative technologies to uncover a novel regulatory step in the control of transcriptional directionality.

Strengths:

One of the major strengths of this work is that the authors endogenously tag two of their proteins of interest - RBBP6 and INST11. This tag allows them to rapidly degrade these proteins - increasing the likelihood that any effects they see are primary effects of protein depletion rather than secondary effects. Another strength of this work is that the authors immunoprecipitate RNAPII and sequence extracted full length RNA (POINT-seq) allowing them to map nascent transcription. A technical advance from this work is the development of sPOINT which allows the selective amplification of 5' capped RNAs < 150 nucleotides, allowing the direction of transcription initiation to be resolved.

Weaknesses:

While the authors provide strong evidence that INST11 and CDK9 play important roles in determining promoter directionality, their data suggests that when INST11 is degraded and CDK9 is inhibited there remains a bias in favour of sense transcription (Figure 4B and C). This suggests that there are other unknown factors that promote sense transcription over antisense transcription and future work could look to identify these.

Reviewer #3 (Public Review):

Summary:

Using protein degradation approach, Eaton et al show that INST11 can terminate the sense and anti-sense transcription but higher activity of CDK9 in sense direction protects it from INS11-dependent termination. They developed sPOINT-seq that detects nascent 5'-capped RNA. The technique allowed them to reveal robust transcription initiation of sense-RNA as compared to anti-sense.

Strengths:

The strength of paper is acute degradation of proteins, eliminating the off-target effects. Further, the paper uses elegant approaches such as POINT and sPOINT-seq to measure nascent RNA and 5'-capped short RNA. Together, the combination of these three allowed the authors to make clean interpretations of data.

Weaknesses:

While manuscript is well written, the details on panel is not sufficient. The methods can be more elaborate for better understanding. Additional discussion on how authors findings contradict the existing model of anti-sense transcription termination should be added.

in the revised manuscript, authors have added details on panels and elaborated method and other sections for better understanding.

eLife assessment

The study presents valuable findings on the molecular mechanisms of glucose-stimulated insulin secretion from pancreatic islets, focusing on the main regulatory elements of the signaling pathway in physiological conditions. While the evidence supporting the conclusions is solid, the study can be strengthened by the use of a beta cell line or knockout mice. The work will be of interest to cell biologists and biochemists working on diabetes.

Reviewer #1 (Public Review):

Summary:

This study investigated the mechanism by which PGE2 inhibits the release of insulin from pancreatic beta cells in response to glucose. The researchers used a combination of cell line experiments and studies in mice with genetic ablation of the Kv2.2 channel. Their findings suggest a novel pathway where PGE2 acts through EP2/EP4 receptors to activate PKA, which directly phosphorylates a specific site (S448) on the Kv2.2 channel, inhibiting its activity and reducing GSIS.

Strengths:

- The study elegantly demonstrates a potential pathway connecting PGE2, EP2/EP4 receptors, PKA, and Kv2.2 channel activity, using embryonic cell line.<br /> - Additional experiments in INS1 and primary mouse beta cells with altered Kv2.2 function partially support the inhibitory role of PGE2 on GSIS through Kv2.2 inhibition.

Weaknesses:

- A critical limitation is the use of HEK293T cells, which are not pancreatic beta cells. Functional aspects can differ significantly between these cell types.<br /> - The study needs to address the apparent contradiction of PKA activating insulin secretion in beta cells, while also inhibiting GSIS through the proposed mechanism.<br /> - A more thorough explanation is needed for the discrepancies observed between the effects of PGE2 versus Kv2.2 knockdown/mutation on the electrical activity of beta cells and GSIS.

Reviewer #2 (Public Review):

The authors identified new target elements for prostaglandin E2 (PGE2) through which insulin release can be regulated in pancreatic beta cells under physiological conditions. In vitro extracellular exposure to PGE2 could directly and dose-dependently inhibit the potassium channel Kv2.2. In vitro pharmacology revealed that this inhibition occurs through the EP2/4 receptors, which activate protein kinase A (PKA). By screening specific sites of the Kv2.2 channel, the target phosphorylation site (S448) for PKA regulation was found. The physiological relevance of the described signaling cascade was investigated and confirmed in vivo, using a Kv2.2 knockdown mouse model.

The strength of this manuscript is the novelty of the (EP2/4-PKA-Kv2.2 channel) molecular pathway described and the comprehensive methodological toolkit the authors have relied upon.

The introduction is detailed and contains all the information necessary to place the claims in context. Although the dataset is comprehensive and a logical lead is consistently built, there is one important point to consider: to clarify that the described signaling pathway is characteristic of normal physiological conditions and thus differs from pathological changes. It would be useful to carry out basic experiments in a diabetes model (regardless of whether this is in mice or rats).

Author response:

We thank the reviewers for their positive evaluation and constructive feedback on our study.

We acknowledge the concern regarding the use of HEK293T cells. In the revised manuscript, we will provide a more detailed explanation of the role of the PKA pathway in the regulation of GSIS by PGE2. To validate this regulation through Kv2.2, we will overexpress the Kv2.2 mutant channel in beta cells and assess its impact. Additionally, we will verify the specificity of the antibodies for EP1-EP4 receptors by knockdown. To confirm the receptors involved in PGE2 function, we will use additional EP receptor blockers or perform receptor knockdown experiments.

We will clarify that the described signaling pathway operates under normal physiological conditions and differs from pathological changes.

We once again thank the reviewers for their positive evaluation and constructive suggestions.

eLife assessment

This work describes a novel affinity interactomics approach that allows investigators to identify networks of protein-protein interactions in cells. The important findings presented here describe the application of this technique to the SH3 domain of the membrane remodeling Bridging Integrator 1 (BIN1), the truncation of which leads to centronuclear myopathy. The authors present solid evidence that BIN1 SH3 engages with an unexpectedly high number of cellular proteins, many of which are linked to skeletal muscle disease, and evidence is presented to suggest that BIN1 may play a role in mitosis creating the potential for new avenues in drug development efforts. Some of the findings, however, remain rather preliminary, lack sufficient replicates and may require additional experiments to definitively support the conclusions.

Reviewer #1 (Public Review):

Summary:

In this paper, Zambo and coworkers use a powerful technique, called native holdup, to measure the affinity of the SH3 domain of BIN1 for cellular partners. Using this assay, they combine data using cellular proteins and proline-containing fragments in these proteins to identify 97 distinct direct binding partners of BIN1. They also compare the binding interactome of the BIN1 SH3 domain to the interactome of several other SH3 domains, showing varying levels of promiscuity among SH3 domains. The authors then use pathway analysis of BIN1 binding partners to show that BIN1 may be involved in mitosis. Finally, the authors examine the impact of clinically relevant mutations of the BIN1 SH3 domain on the cellular interactome. The authors were able to compare the interactome of several different SH3 domains and provide novel insight into the cellular function of BIN1. Generally, the data supports the conclusions, although the reliance on one technique and the low number of replicates in each experiment is a weakness of the study.

Strengths:

The major strength of this paper is the use of holdup and native holdup assays to measure the affinity of SH3 domains to cellular partners. The use of both assays using cell-derived proteins and peptides derived from identified binding partners allows the authors to better identify direct binding partners. This assay has some complexity but does hold the possibility of being used to measure the affinity of the cellular interactome of other proteins and protein domains. Beyond the utility of the technique, this study also provides significant insight into the cellular function of BIN1. The authors have strong evidence that BIN1 might have an undiscovered function in cellular mitosis, which potentially highlights BIN1 as a drug target. Finally, the study provides outstanding data on the cellular binding properties and partners of seven distinct SH3 domains, showing surprising differences in the promiscuity of these proteins.

Weaknesses:

There are several weaknesses of the study. First, the authors rely completely on a single technique to measure the affinity of the cellular interactome. The native holdup is a relatively new technique that is powerful yet relatively unproven. However, it appears to have the capacity to measure the relative affinity of proteins and the authors describe the usefulness of the technique. Second, and most important, the authors use a relatively small number of replicates for the holdup assays. The holdup technique will have biological variation in the cellular lysate or purified protein that could impact the results, so more replicates would enhance the reliability of the results.

Reviewer #2 (Public Review):

Summary:

The authors report here interesting data on the interactions mediated by the SH3 domain of BIN1 that expand our knowledge on the role of the SH3 domain of BIN1 in terms of mediating specific interactions with a potentially high number of proteins and how variants in this region alter or prevent these protein-protein interactions. These data provide useful information that will certainly help to further dissect the networks of proteins that are altered in some human myopathies as well as the mechanisms that govern the correct physiological activity of muscle cells.

Strengths:

The work is mostly based on improved biochemical techniques to measure protein-protein interaction and provide solid evidence that the SH3 domain of BIN1 can establish an unexpectedly high number of interactions with at least a hundred cellular proteins, among which the authors underline the presence of other proteins known to be causative of skeletal muscle diseases and not known to interact with BIN1. This represents an unexpected and interesting finding relevant to better define the network of interactions established among different proteins that, if altered, can lead to muscle disease. An interesting contribution is also the detailed identification of the specific sites, namely the Proline-Rich Motifs (PRMs) that in the interacting proteins mediate binding to the BIN1 SH3 domain.

Weaknesses:

Less convincing, or too preliminary in my opinion, are the data supporting BIN1 co-localization with PRC1. Indeed, the affinity of PRC1 is significantly lower than that of DNM2, an established BIN1 interacting protein. Thus, this does not provide compelling evidence to support PRC1 as a significant interactor of BIN1. Similarly, the localization data appears somewhat preliminary to substantiate a role of BIN1 in mitotic processes. These findings may necessitate additional experimental work to be more convincing.

Author response:

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

Reviewer #1

We modified the text regarding PRC1 according to the reviewer’s recommendation.

Reviewer #2

Following the reveiwer’s advise, we introduced the holdup assay, as well as the native holdup assay in more details.

This new part now also discusses the question of replicates in more details. We do not agree with the eLife assessment on this matter, but we think that this assessment was made because analyzing holdup data requires a different approach compared to more conventional interactomic approaches and these differences were not introduced in sufficient depth. We hope that the inclusion of more background reasoning, as well as by providing a more detailed comparison of the measured independent BIN1 interactomes, now included on Figure S4, will eliminate all confusion in the reader.

We thank the reviewer for guiding us to a previous work that was done on Grb2. Indeed, the finding of this earlier work aligns perfectly with our finding suggesting general similarities in SH3 domain mediated interactions.

eLife assessment

This useful manuscript extends prior work to identify OVO as a major transcriptional activator of the female germline gene expression program. Using a combination of solid genomic strategies, the authors demonstrate that OVO binds to the promoters of hundreds of genes in the female germline and promotes their expression.

Reviewer #1 (Public Review):

Summary:

In this manuscript, Benner et al. identify OVO as a transcriptional factor instrumental in promoting expression of hundreds of genes essential for female germline identity and early embryo development. Prior data had identified both ovo and otu as genes activated by OVO binding to the promoters. By combining ChIP-seq, RNA-seq and analysis of prior datasets, the authors extend these data to hundreds of genes and therefore propose that OVO is a master transcriptional regulator of oocyte development. They further speculate that OVO may function to promote chromatin accessibility to facilitate germline gene expression. Overall, the data compellingly demonstrate a much broader role for OVO in activation of genes in the female germline than previously recognized. By contrast, the relationship between OVO, chromatin accessibility and the timing of gene expression is only correlative, and more work will be needed to determine the mechanisms by which OVO promotes transcription.

Strengths

Here Benner at al. convincingly show that OVO is a transcriptional activator that promotes expression of hundreds of genes in the female germline. The ChIP-seq and RNA-seq data included in the manuscript are robust and the analysis is compelling.

Importantly, the set of genes identified are essential for maternal processes, including egg production and patterning of the early embryo. Together, these data identify OVO as a major transcriptional activator of the numerous genes expressed in the female germline, deposited into the oocyte and required for early gene expression. This is an important finding as this is an essential process for development and prior to this study the major drivers of this gene expression program were unknown.

Weaknesses

The novelty of the manuscript is somewhat limited as the authors show that, like two prior, well-studied OVO target genes, OVO binds to promoters of germline genes and activates transcription. The fact that OVO performs this function more broadly is not particularly surprising.

A major challenge to understanding the impact of this manuscript is the fact that the experimental system for the RNA-seq, the tagged constructs, and the expression analysis that provides the rationale for the proposed pioneering function of OVO are all included in a separate manuscript.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this manuscript, Benner et al. identify OVO as a transcriptional factor instrumental in promoting the expression of hundreds of genes essential for female germline identity and early embryo development. Prior data had identified both ovo and otu as genes activated by OVO binding to the promoters. By combining ChIP-seq, RNA-seq, and analysis of prior datasets, the authors extend these data to hundreds of genes and therefore propose that OVO is a master transcriptional regulator of oocyte development. They further speculate that OVO may function to promote chromatin accessibility to facilitate germline gene expression. Overall, the data compellingly demonstrate a much broader role for OVO in the activation of genes in the female germline than previously recognized. By contrast, the relationship between OVO, chromatin accessibility, and the timing of gene expression is only correlative, and more work will be needed to determine the mechanisms by which OVO promotes transcription.

We fully agree with this summary.  

Strengths:

Here Benner et al. convincingly show that OVO is a transcriptional activator that promotes expression of hundreds of genes in the female germline. The ChIP-seq and RNA-seq data included in the manuscript are robust and the analysis is compelling.

Importantly, the set of genes identified is essential for maternal processes, including egg production and patterning of the early embryo. Together, these data identify OVO as a major transcriptional activator of the numerous genes expressed in the female germline, deposited into the oocyte and required for early gene expression. This is an important finding as this is an essential process for development and prior to this study, the major drivers of this gene expression program were unknown.

We are delighted that this aspect of the work came across clearly. Understanding the regulation of maternal effect genes has been something of a black-box, despite the importance of this class of genes in the history of developmental genetics. The repertoire of essential oogenesis/embryonic development genes that are bound by and respond to OVO are well characterized in the literature, but nothing is known about how they are transcriptionally regulated. We feel the manuscript will be of great interest to readers working on these genes.

Weaknesses:

The novelty of the manuscript is somewhat limited as the authors show that, like two prior, well-studied OVO target genes, OVO binds to promoters of germline genes and activates transcription. The fact that OVO performs this function more broadly is not particularly surprising.

Clearly, transcription factors regulate more than one or two genes. Never-the-less we were surprised at how many of the aspects of oogenesis per se and maternal effect genes were OVO targets. It was our hypothesis that OVO would have a transcriptional effect genome-wide, however, it was less clear whether OVO would always bind at the core promoter, as is with the case of ovo and otu. Our results strongly support the idea that core promoter proximal binding is essential for OVO function; a conclusion of work done decades ago, which has not been revisited using modern techniques. 

A major challenge to understanding the impact of this manuscript is the fact that the experimental system for the RNA-seq, the tagged constructs, and the expression analysis that provides the rationale for the proposed pioneering function of OVO are all included in a separate manuscript.

This is a case where we ended up with a very, very long manuscript which included a lot of revisiting of legacy data. It was a tough decision on how to break up all the work we had completed on ovo to date. In our opinion, it was too much to put everything into a single manuscript unless we wanted a manuscript length supplement (we were also worried that supplemental data is often overlooked and sometimes poorly reviewed). We therefore decided to split the work into a developmental localization/characterization paper and a functional genomics paper. As it stands both papers are long. Certainly, readers of this manuscript will benefit from reading our previous OVO paper, which we submitted before this one. The earlier manuscript is under revision at another journal and we hope that this improved manuscript will be published and accessible shortly.

Reviewer #2 (Public Review):

Summary:

In this manuscript, Benner et al. interrogate the transcriptional regulator OVO to identify its targets in the Drosophila germline. The authors perform ChIP-seq in the adult ovary and identify established as well as novel OVO binding motifs in potential transcriptional targets of OVO. Through additional bioinformatic analysis of existing ATAC-seq, CAGE-seq, and histone methylation data, the authors confirm previous reports that OVO is enriched at transcription start sites and suggest that OVO does not act as part of the core RNA polymerase complex. Benner et al. then perform bulk RNA-seq in OVO mutant and "wildtype" (GAL4 mediated expression of OVO under the control of the ovo promoter in OVO mutants) ovaries to identify genes that are differentially expressed in the presence of OVO. This analysis supports previous reports that OVO likely acts at transcription start sites as a transcriptional activator. While the authors propose that OVO activates the expression of genes that are important for egg integrity, maturation, and for embryonic development (nanos, gcl, pgc, bicoid), this hypothesis is based on correlation and is not supported by in vivo analysis of the respective OVO binding sites in some of the key genes. A temporal resolution for OVO's role during germline development and egg chamber maturation in the ovary is also missing. Together, this manuscript contains relevant ChIP-seq and RNA-seq datasets of OVO targets in the Drosophila ovary alongside thorough bioinformatic analysis but lacks important in vivo experimental evidence that would validate the high-quality datasets.

We thank reviewer 2 for the appreciation of the genomics data and analysis. Some of the suggested in vivo experiments are clear next steps, which are well underway. These are beyond the scope of the current manuscript. 

Temporal analysis of ovo function in egg chamber development is not easy, as only the weakest ovo alleles have any egg chambers to examine. However, we will also point out the long-known phenotypes of some of those weak alleles in the text (e.g. ventralized chambers in ovoD3/+). We will need better tools for precise rescue/degradation during egg chamber maturation.     

Strengths:

The manuscript contains relevant ChIP-seq and RNA-seq datasets of OVO targets in the Drosophila ovary alongside thorough bioinformatic analysis

Thank you. We went to great lengths to do our highly replicated experiments in multiple ways (e.g. independent pull-down tags) and spent considerable time coming up with an optimized and robust informatic analysis.

Weaknesses:

(1) The authors propose that OVO acts as a positive regulator of essential germline genes, such as those necessary for egg integrity/maturation and embryonic/germline development. Much of this hypothesis is based on GO term analysis (and supported by the authors' ChIP-seq data). However accurate interpretation of GO term enrichment is highly dependent on using the correct background gene set. What control gene set did the authors use to perform GO term analysis (the information was not in the materials and methods)? If a background gene set was not previously specified, it is essential to perform the analysis with the appropriate background gene set. For this analysis, the total set of genes that were identified in the authors' RNA-seq of OVO-positive ovaries would be an ideal control gene set for which to perform GO term analysis. Alternatively, the total set of genes identified in previous scRNA-seq analysis of ovaries (see Rust et al., 2020, Slaidina et al., 2021 among others) would also be an appropriate control gene set for which to perform GO term analysis. If indeed GO term analysis of the genes bound by OVO compared to all genes expressed in the ovary still produces an enrichment of genes essential for embryonic development and egg integrity, then this hypothesis can be considered.

We feel that this work on OVO as a positive regulator of genes like bcd, osk, nos, png, gnu, plu, etc., is closer to a demonstration than a proposition. These are textbook examples of genes required for egg and early embryonic development. Hopefully, this is not lost on the readers by an over-reliance on GO term analysis, which is required but not always useful in genome-wide studies. 

We used GO term enrichment analysis as a tool to help focus the story on some major pathways that OVO is regulating. To the specific criticism of the reference gene-set, GO term enrichment analysis in this work is robust to gene background set. We will update the GO term enrichment analysis text to indicate this fact and add a table using expressed genes in our RNA-seq dataset to the manuscript and clarify gene set robustness in greater detail in the methods of the revision. We will also try to focus the reader’s attention on the actual target genes rather than the GO terms in the revised text.

We have updated the GO term analysis by including all the expressed genes in our RNA-seq datasets as a background control. Figure 6 has been updated to include the significant GO terms. We have outlined changes in the methods section below.

Lines 794-801:

“Gene ontology enrichment analysis was completed with g:Profiler’s g:GOSt software (Raudvere et al. 2019) on the set of genes overlapping OVO ChIP peaks over the TSS and significantly upregulated in the presence of ectopic OVO (525 genes in total). All genes that were considered to be expressed in our RNA-seq datasets were used as a background control (10,801 genes in total). Default parameters were used for the enrichment analysis except for ‘statistical domain scope’ was set to ‘custom’ (our control background genes were uploaded here), ‘significance threshold’ was set to ‘Bonferroni correction’, and only GO biological process terms were searched for enrichment with the gene list. The GO terms listed in Figure 6 represent the 24 smallest GO term sizes according to Table S5.”

(2) The authors provide important bioinformatic analysis of new and existing datasets that suggest OVO binds to specific motifs in the promoter regions of certain germline genes. While the bioinformatic analysis of these data is thorough and appropriate, the authors do not perform any in vivo validation of these datasets to support their hypotheses. The authors should choose a few important potential OVO targets based on their analysis, such as gcl, nanos, or bicoid (as these genes have well-studied phenotypes in embryogenesis), and perform functional analysis of the OVO binding site in their promoter regions. This may include creating CRISPR lines that do not contain the OVO binding site in the target gene promoter, or reporter lines with and without the OVO binding site, to test if OVO binding is essential for the transcription/function of the candidate genes.

Exploring mechanism using in vivo phenotypic assays is awesome, so this is a very good suggestion. But, it is not essential for this work -- as has been pointed out in the reviews, in vivo validation of OVO binding sites has been comprehensively done for two target genes, ovo and otu. The “rules” appear similar for both genes. That said, we are already following up specific OVO target genes and the detailed mechanism of OVO function at the core promoter. We removed some of our preliminary in vivo figures from the already long current manuscript. We continue to work on OVO and expect to include this type of analysis in a new manuscript.

(3) The authors perform de novo motif analysis to identify novel OVO binding motifs in their ChIP-seq dataset. Motif analysis can be significantly strengthened by comparing DNA sequences within peaks, to sequences that are just outside of peak regions, thereby generating motifs that are specific to peak regions compared to other regions of the promoter/genome. For example, taking the 200 nt sequence on either side of an OVO peak could be used as a negative control sequence set. What control sequence set did the authors use as for their de novo motif analysis? More detail on this is necessary in the materials and methods section. Re-analysis with an appropriate negative control sequence set is suggested if not previously performed.

We apologize for being unclear on negative sequence controls in the methods. We used shuffled OVO ChIP-seq peak sequences as the background for the de novo motif analysis, which we will better outline in the methods of the revision. This is a superior background set of sequences as it exactly balances GC content in the query and background sequences. We are not fond of the idea of using adjacent DNA that won’t be controlled for GC content and shadow motifs. Furthermore, the de novo OVO DNA binding motifs are clear, statistically significant variants of the characterized in vitro OVO DNA binding motifs previously identified (Lu et al., 1998; Lee and Garfinkel, 2000; Bielinska et al., 2005), which lends considerable confidence. We also show that the OVO ChIP-seq read density are highly enriched for all our identified motifs, as well as the in vitro motifs. We provide multiple lines of evidence, through multiple methods, that the core OVO DNA binding motif is 5’-TAACNGT-3’. We have high confidence in the motif data.

We have added the below text to the methods section for further clarity on motif analysis parameters.

Lines 808-812

“The default parameters were used for de novo motif enrichment analysis, including the use of shuffled input sequences as a control. After identifying ‘OVO Motif One’, OVO ChIP peaks that contained that sequence were removed and the resulting ChIP peaks were resubmitted for STREME analysis deriving derivative OVO DNA binding motifs like above.”

(4) The authors mention that OVO binding (based on their ChIP-seq data) is highly associated with increased gene expression (lines 433-434). How many of the 3,094 peaks (conservative OVO binding sites), and what percentage of those peaks, are associated with a significant increase in gene expression from the RNA-seq data? How many are associated with a decrease in gene expression? This information should be added to the results section.

Not including the numbers of the overlapping ChIP peaks and expression changes in the text was an oversight on our part. The numbers that relate to this (666 peaks overlapping genes that significantly increased in expression, significant enrichment according to Fishers exact test, 564 peaks overlapping genes that significantly decreased in expression, significant depletion according to Fishers exact test) are found in figure 4C and will be added to the text.

We have modified the results section to include the overlap between the RNA-seq and ChIP-seq data.

Lines 463-468

“We found that 2,298 genes that were expressed in our RNA-seq data overlapped an OVO ChIP peak. 666 genes significantly increased in expression and were bound by OVO, which is a significant enrichment according to a Fisher’s exact test (Figure 4C, cyan dots, p < 0.01, odds ratio = 2.21). While conversely, 564 genes decreased in expression and were bound by OVO, indicating a significant depletion according to a Fisher’s exact test (Figure 4C, blue dots, p < 0.01, odds ratio = 0.85).”

(5) The authors mention that a change in endogenous OVO expression cannot be determined from the RNA-seq data due to the expression of the OVO-B cDNA rescue construct. Can the authors see a change in endogenous OVO expression based on the presence/absence of OVO introns in their RNA-seq dataset? While intronic sequences are relatively rare in RNA-seq, even a 0.1% capture rate of intronic sequence is likely to be enough to determine the change in endogenous OVO expression in the rescue construct compared to the OVO null.

This is a good point. The GAL4 transcript is downstream of ovo expression in the hypomorphic ovoovo-GAL4 allele. We state in the text that there is a nonsignificant increase in GAL4 expression with ectopic rescue OVO, although the trend is positive. We calculated the RPKM of RNA-seq reads mapping to the intron spanning exon 3 and exon 4 in ovo-RA and found that there is also a nonsignificant increase in intronic RPKM with ectopic rescue OVO (we will add to the results in the revision). We would expect OVO to be autoregulatory and potentially increase the expression of GAL4 and/or intronic reads, but the ovoovoGAL4>UASp-OVOB is not directly autoregulatory like the endogenous locus. It is not clear to us how the intervening GAL4 activity would affect OVOB activity in the artificial circuit. Dampening? Feed-forward? Is there an effect on OVOA activity? Regardless, this result does not change our interpretation of the other OVO target genes.

We have added the analysis of intronic ovo RNA-seq to the results as outlined below.

Lines 512-520

“Transcriptionally, ovo RNA-seq reads are likely derived from the UASp-3xFHA-OVO-B cDNA rescue or are indistinguishable between the genomic locus and rescuing cDNA transgene. We found a nonsignificant increase in exon 3 to exon 4 intronic ovo reads with the expression of ectopic rescue OVO (log2 fold change = 0.76, p-adj = 0.26). These intronic reads would be derived from the endogenous ovo locus, but it is difficult to conclusively determine if the endogenous ovo locus would respond transcriptionally to ectopic OVO downstream of UASp (for example, the pathway for ovo is no longer autoregulatory in ovoovo-GAL4/ovoΔBP; UASp-3xFHA-OVO-B germ cells, there is an additional GAL4>UASp activation step). So, we could not confidently assess whether ovo responded transcriptionally to ectopic rescue OVO.”

(6) The authors conclude with a model of how OVO may participate in the activation of transcription in embryonic pole cells. However, the authors did not carry out any experiments with pole cells that would support/test such a model. It may be more useful to end with a model that describes OVO's role in oogenesis, which is the experimental focus of the manuscript.

We did not complete any experiments in embryonic pole cells in this manuscript and base our discussion on the potential dynamics of OVO transcriptional control and our previous work showing maternal and zygotic OVO protein localization in the developing embryonic germline. Obviously, we are highly interested in this question and continue to work on the role of maternal OVO. We agree that we are extended too far and will remove the embryonic germ cell model in the figure. We will instead focus on the possible mechanisms of OVO gene regulation in light of the evidence we have shown in the adult ovary, as suggested.

We have removed figure 7 and have re-written the last two paragraphs of the discussion as below.

Lines 645-663

“The requirement for OVO at the TSS of target genes has been well characterized at its own locus as well as its downstream target otu. Our OVO ChIP and expression data confirm findings from previous work that OVO is binding to these target promoters, and in the case of otu, strongly responds transcriptionally to the presence of OVO. Although we did not test the requirement for OVO DNA binding motifs at other OVO bound genes in this work, this has been extensively explored before, showing that removal of OVO

DNA binding sites overlapping the TSS results in a strong decrease in reporter expression (Lü et al. 1998; Bielinska et al. 2005; Lü and Oliver 2001). Removal of more distal upstream OVO DNA binding sites also reduces reporter expression to a lesser degree. However, for most cases tested, removal of OVO DNA binding sites while leaving the rest of the enhancer regions intact, never totally abolished reporter expression. These dynamics are highly similar to work that has been completed on the pioneer factor zelda (zld). Adding zld DNA binding motifs to a stochastically expressed transcriptional reporter increases the activity and response of the reporter (Dufourt et al. 2018). Distally located zld DNA binding motifs influenced reporter expression to a lesser degree than proximal sites. A single zld DNA binding site adjacent to the TSS produced the strongest reporter activity. Importantly, just like the activity of OVO transgenic reporters, there is not an absolute requirement for zld DNA binding to activate reporter expression, however, the addition of TSS adjacent zld DNA binding motifs does strongly influence reporter response. We know that zld achieves this reporter response through its pioneering activity (Xu et al. 2014; Harrison et al. 2011), whether OVO achieves this similar effect on gene expression through a shared mechanism, or in cooperation with other transcription factors needs to be further explored.”

Recommendations for the authors:

Reviewer #1 (Recommendations For The Authors):

The Results section could be streamlined by limiting the discussion of analysis to only those details that are unusual or essential for understanding the science. For example, the fact that MACS3 was used to call peaks seems most suitable for the Methods section.

We have removed the below excerpts from the results section to streamline the text.

‘We compared immuno-purified OVO associated DNA with input DNA as a control, for a total of 12 ChIPseq libraries, which we sequenced using the Illumina system. After quality control and alignment to the Drosophila r6.46 genome (Gramates et al. 2022), we used MACS3 (Zhang et al. 2008)’

The Supplemental Tables are referred to out of order. Table S2 is referred to on line 143 while Table S1 is not referred to until the Methods section.

We have reorganized the order of the tables in the manuscript text.

In the analysis of CAGE-seq data, it is unclear whether there is anything distinctive about the ~2000 regions bound by OVO but that is not near TSS in the ovary dataset. Are these TSS that are not active in the ovary or are these non-promoter bound OVO sites? If they are TSS of genes not in the CAGE-seq data set, are these genes expressed in other tissues or just expressed at lower levels in the ovary?

This was a good point that prompted us to take a closer look at the characteristics of OVO binding and its relationships to promoters and other gene elements. 45% of OVO ChIP peaks overlapped the TSS while 55% were either non-overlapping downstream or upstream of the TSS. When plotting OVO ChIP read density, there was still a striking enrichment of OVO binding over the TSS, even though the ChIP peak was not overlapping the TSS (new figure 1K). This is possibly due to weaker direct OVO binding at the TSS that was not considered significant in the peak calling software or were indirect interactions of the distal OVO binding and the TSS. We outline this in the below text added to the results section on the OVO ChIP. To showcase these results, we have included a new panel in figure 1K. We removed the panel showing the enrichment over the cage-seq TSS, but this same data remains in the heatmap shown in figure 1L, so no information is lost. To directly answer the Cage-seq questions considering the OVO bound over the annotated TSS results, we found that 1,047 chip peaks overlapped CAGE-seq TSS, which is only 347 fewer than the annotated TSS overlap (1,394). Of the 1,394 genes that were bound by over the TSS, all of them were considered to be expressed in our RNA-seq dataset, indicating that these might just be more lowly expressed genes that for whatever reason were not considered to be enriched TSSs in the CAGE-seq data. This difference is likely not significant.

Lines 235-251

“Although OVO ChIP peaks overlapping genes showed a strong read density enrichment over the TSS, we found that only 45% (1,394/3,094) of OVO ChIP peaks directly overlapped a TSS. 43% (1,339/3,094) of OVO ChIP peaks were found to overlap the gene body downstream of the TSS (intronic and exonic sequences) and 12% (366/3,094) did not overlap any gene elements, indicating that they were intergenic.

We were interested in the differences between OVO binding directly over the TSS or at more distal upstream and downstream sites. We decided to plot the OVO ChIP read density of these different classes of OVO binding patterns and found that OVO bound over the TSS produced a sharp read density enrichment over the TSS which was consistent with what was found for all OVO bound genes (Figure 1K). OVO binding along the gene body surprisingly also showed a read density enrichment over the TSS, although the magnitude of read density enrichment was notably less than TSS OVO binding. Intergenic OVO binding also showed these same characteristics with a notable upstream read density enrichment possibly indicative of enhancer binding. This indicates that although the significantly called OVO ChIP peaks did not overlap the TSS, there was still a propensity for TSS sequences to be enriched with OVO ChIP over the input control. This could be due to weaker direct in vivo binding of OVO to these TSSs or indirect interactions between the upstream/downstream OVO bound sequences and the TSS, possibly through a looping enhancer-promoter interaction. However, regardless of the location of the OVO ChIP peak, OVO seemed to always be enriched at or in close proximity to TSSs.”

It would be helpful for the authors to provide a bit more detailed analysis of chromatin states of OVObound regions in GSC, 8c NC, and 32c NC (or some more clarity in the current analysis). Are the regions that are bound by OVO accessible in all these cell types or specifically enriched for accessibility in a subset? The authors state that OVO binding is correlated with open chromatin, but whether these are regions that are open in all cell types analyzed or a subset is not clear from the data presented. Promoters are often accessible regardless of cell type, so it is unclear what exactly is to be concluded from this association. Also, is the proximity to open chromatin features for OVO-bound promoters (as shown in Figure 2C) different than non-OVO-bound promoters (the two classes shown Figure 1L, for example)?

We utilized previously published datasets of staged germ cell chromatin status to look at the association of chromatin status and OVO binding. Unfortunately, not all the same germ cell stages were profiled for each chromatin mark from the datasets derived for these two papers. For example, only H3K4me3 data exists for GSCs, and only gsc and 8c data exists for H3K9me3, while the other chromatin marks had more profiles, even including later stages. We focused specifically on gsc and 32c (essentially stage 5 egg chambers) for the other chromatin marks since that is when the ovo hypomorphic egg chambers arrest. A nice control would have been chromatin states in somatic follicle cells of the ovary, since we know germ cell genes such as ovo and otu are not expressed and presumably the chromatin states in somatic cell types would be different than germ cells. However, chromatin states for somatic follicle cells were not published in these two papers and we are not aware of any other existing datasets to compare too. Essentially, we need to determine the changes in chromatin states with and without OVO, which we are currently working on. 

We did further analyze chromatin states and differential OVO binding in respect to gene elements, and found that OVO binding, regardless of the relationship to the gene element, is always open (gsc and 32c ATAC). OVO binding over the gene body shows the same enrichment for open chromatin and transcriptionally active histone marks. We compared the profiles of these chromatin marks and the promoters of OVO bound and not bound genes and consistent with the suggestion that promoters are generally open, we found that this was the case. However, there is an enrichment for open chromatin and transcriptionally active histone marks for OVO bound genes compared to non-OVO bound genes. This could be a consequence of OVO binding or indirect consequence of a downstream OVO target. Regardless, as has been suggested, future experiments directly measuring chromatin status and OVO needs to be performed. The below excerpts have been added to the text to supplement the comments provided above.

Lines 328-343

“The association of OVO binding with active histone marks and open chromatin was striking, but open chromatin is likely a general phenomenon of promoters (Haines and Eisen, 2008). Indeed, when measuring the read density for GSC and 32C ATAC-seq for OVO bound and OVO non-bound promoters, there is an enrichment for open chromatin at the TSS regardless of OVO binding. However, we did notice an increase in enrichment for OVO bound promoters compared to OVO non-bound promoters (Figure S1G), possibly suggesting that OVO bound promoters are more open or have an increase in accessibility when compared to non-OVO bound promoters. This same relationship held true for the transcriptionally active histone mark H3K27ac in GSCs (Figure S1H). Since only 45% of OVO ChIP peaks overlapped TSSs, we plotted the read density of the above chromatin marks over OVO ChIP peak maximums for OVO bound over the TSS, gene body, or intergenic regions (Figure S2A-D). We found that OVO bound regions that were not overlapping the TSS still showed the same propensity for enrichment of open chromatin and active histone marks. Intergenic regions were especially enriched for open chromatin measured through ATAC-seq. Altogether suggesting that OVO binding genome-wide is tightly associated with open chromatin regardless of germ cell stage, and active transcription in GSCs. In other words, chromatin state data suggests OVO is acting positively on its target genes and raises the possibility that OVO-binding and open chromatin are related.”

For clarity, it would help the reader if the authors mentioned the male-specific TATA-associated factors as a rationale for testing the role of OVO binding in core promoter function. This is currently mentioned in the Discussion on lines 575-577, but would help in understanding the motivation behind the detailed analysis of the promoter binding of OVO in the Results and make the negative result more clearly impactful.

We have introduced the male specific tata factors as suggested and have condensed the two intro paragraphs in this section into one, as shown below.

Lines 347-363

“Our data thus far clearly indicates that OVO binding occurs at or very near the core promoter, a region recognized by an enormous collection of factors that associate with RNA polymerase to initiate transcription (Aoyagi and Wassarman 2000; Vo Ngoc, Kassavetis, and Kadonaga 2019). The highly organized polymerase complex has sequence-specific DNA recognition sites with incredibly precise spacing between them, with an overall DNA footprint of a little less than 100bp (Rice, Chamberlin, and Kane 1993; FitzGerald et al. 2006; Ohler et al. 2002). There are upstream binding sites such as TATA, sites at transcription start, such as the initiator (INR), and downstream promoter elements (DPE) (Vo Ngoc, Kassavetis, and Kadonaga 2019). The combinations of these DNA motifs is not random in mammals and Drosophila (FitzGerald et al. 2006), and distinct combinations of different motifs at the TSS of genes expressed in Drosophila are conserved over tens of millions of years of evolution (Chen et al. 2014). The male germline expresses a number of TATA-associated factors that have been implicated in male-specific promoter usage for gene expression (M. Hiller et al. 2004; M. A. Hiller et al. 2001; Lu et al. 2020; V. C. Li et al. 2009). It is possible that OVO is a female germline specific TATA-associated factor, and if so, OVO binding sites at core promoters should share precise spacing with other core promoter elements, suggesting it is likely part of the complex. If not, then OVO is more likely to facilitate binding of the basal transcriptional machinery. Because of the extended footprint of engaged RNA polymerase, OVO and the basal machinery would not be likely to occupy the same region at the same time.”

The description of the system used for the RNA-seq would benefit from additional clarity. It is not clear as written why it is "Lucky" that there is an mRNA isoform with extended exon 2 required for egg chamber development beyond stage 5. How does this requirement compare to the global requirement for OVO, which seems to be required for germ cell development even before stage 5? Understanding this system is essential for interpreting the RNA-seq results. Indeed, the authors have a separate manuscript (currently on bioRxiv) that explains the details of this system. As such, the current description requires that the reader refer to this additional pre-print. Could the authors include a diagram to better illustrate this system? Furthermore, since this RNA-seq is being performed on tissue that includes nurse cells, follicle cells, and germ cells from multiple stages of development, it is important for the authors to clearly state in which cell types OVO is expressed and likely functional. (While this is well beyond this manuscript, this analysis is the type that might benefit from the use of single-cell sequencing as a means to deconvolute the phenotypic effects of OVO loss.)

We have rewritten the text to better describe the system for RNA-seq. We have also included a figure (Figure S1A) showing the alleles used that should help provide clarity for the readers. We agree that moving forward single cell experiments will be critical to have a better understanding of the transcriptional changes and chromatin dynamics with and without OVO. We have included the below changes to the text.

Lines 409-423

“Previous work from our lab has identified a transheterozygous ovo allelic combination (ovoovo-GAL4/ovoΔBP) that greatly reduces OVO activity resulting in sterility, however, female germ cells are able to survive up until at least stage 5 of oogenesis (Benner et al. 2023). ovoovo-GAL4 is a CRISPR/Cas9 derived T2A-GAL43xSTOP insertion upstream of the splice junction of exon 3 in the ovo-RA transcript (Figure S1A).

Importantly, this insertion in the extended exon 3 would disrupt roughly 90% of the ovo-B transcripts. However, since about 10% of ovo-B transcripts utilize an upstream splice junction in exon 3, these transcripts would not be disrupted with the T2A-GAL4-3xSTOP insertion and thus allow for enough OVO activity for germ cell survival (Benner et al. 2023). Since ovoovo-GAL4 expresses GAL4 in place of full length OVO due to the T2A sequences, we can drive expression of a rescuing OVO-B construct downstream of UASp to generate OVO+ female germ cells, which in fact does rescue the arrested germ cell phenotype of ovoovo-GAL4/ovoΔBP ovaries. Therefore, in order to determine genes that are transcriptionally responsive to OVO, we compared the gene expression profiles in sets of ovaries that had the ovo hypomorphic phenotype with a negative control rescue construct (ovoovo-GAL4/ovoΔBP; UASp-GFP)(Figure 4A) versus those that drive expression of the rescue construct expressing OVO-B (ovoovo-GAL4/ovoΔBP; UASp-3xFHAOVO-B)(Figure 4B).”

Lines 427-432

“The adult female ovary contains somatic cells, germline stem cells, and germline derived nurse cells that would be profiled in a bulk ovary tissue RNA-seq experiment. Although OVO is only required and expressed in germline derived cell types, we chose to dissect one day old post-eclosion ovoovoGAL4/ovoΔBP; UASp-3xFHA-OVO-B female ovaries to enrich for early stages of oogenesis and collected only ovarioles containing the germarium through previtellogenic egg chambers.”

On lines 526-532, it is unclear why the genes fs(1)N, fs(1)M3, and closca are particularly sensitive to the ovoD3 allele. What is this allele trans heterozygous with in the assay that allows development through egg laying? Why might these genes be unique in their sensitivity?

These genes are not particularly sensitive, the transheterozygous hypomorphic ovo ovaries are weak enough to reveal the role of OVO for these genes. We rewrote this paragraph to try and provide more clarity to the relationship between OVO+ binding at these vitelline membrane genes and the phenotype of OVOD3 expressing females.

Lines 562-577

“We also found that the genes fs(1)N, fs(1)M3, and closca, were all bound by OVO and responded transcriptionally to the presence of ectopic rescue OVO. These genes are significant because they constitute a set of genes that are expressed in the germline and the encoded proteins are eventually incorporated into the vitelline membrane providing the structural integrity and impermeability of the egg (Mineo, Furriols, and Casanova 2017; Ventura et al. 2010). Loss-of-function of these three genes results in flaccid eggs that are permeable to dye and fail to develop. The loss-of-function phenotype of fs(1)N, fs(1)M3, and closca closely resembles the dominant antimorph ovoD3 phenotype. The ovoD3 allele is the weakest of the original dominant-negative ovo alleles and produces defective eggs allowing us to explore the role of OVO in late stages (Busson et al. 1983; Komitopoulou et al. 1983). ovoD3/ovo+ transheterozygous females express a repressive form of OVO that results in dominant sterility, and importantly, these females lay flaccid eggs with compromised vitelline membranes that are permeable to the dye neutral red (Oliver, Pauli, and Mahowald 1990). Since OVO+ is bound at the TSS of fs(1)N, fs(1)M3, and closca, and these three genes respond transcriptionally to OVO+, then it is plausible that the repressive OVOD3 is negatively regulating these three genes that are required for vitelline membrane formation. This is evidence that OVO is not only involved in regulating the expression of numerous essential maternal pathways for embryonic development, but it is also essential for regulating genes that are required for egg integrity and maturation.”

The Discussion of OVO as a pioneer factor is highly speculative and based only on correlative data. In fact, the expression data in the embryonic germline is not included in this manuscript, but rather in a separate bioRxiv preprint. This makes it challenging to understand, why this is extensively discussed here. However, there are experiments that could begin to test this proposal. OVO could be expressed in an exogenous tissue and test whether it promotes accessibility. Also, mutations could be made (using gene editing) to identify previously known OVO binding sites in the otu and/or other promoters and these could be assayed for accessibility. By selecting promoters of genes that are not essential for germline development, the authors could directly test the role of OVO in promoting chromatin accessibility. Alternatively, are there reasons that the system used for RNA-seq couldn't be similarly used for ATACseq? It is imperfect but could provide insights into chromatin accessibility in the absence of OVO.

We have largely removed the speculation on pioneering activity, reference to embryonic germline OVO dynamics included in the previous work, and Figure 7. These are excellent suggestions for experiments and ones we are currently pursuing. Below is the modified discussion. 

Lines 645-663

“The requirement for OVO at the TSS of target genes has been well characterized at its own locus as well as its downstream target otu. Our OVO ChIP and expression data confirm findings from previous work that OVO is binding to these target promoters, and in the case of otu, strongly responds transcriptionally to the presence of OVO. Although we did not test the requirement for OVO DNA binding motifs at other OVO bound genes in this work, this has been extensively explored before, showing that removal of OVO

DNA binding sites overlapping the TSS results in a strong decrease in reporter expression (Lü et al. 1998; Bielinska et al. 2005; Lü and Oliver 2001). Removal of more distal upstream OVO DNA binding sites also reduces reporter expression to a lesser degree. However, for most cases tested, removal of OVO DNA binding sites while leaving the rest of the enhancer regions intact, never totally abolished reporter expression. These dynamics are highly similar to work that has been completed on the pioneer factor zelda (zld). Adding zld DNA binding motifs to a stochastically expressed transcriptional reporter increases the activity and response of the reporter (Dufourt et al. 2018). Distally located zld DNA binding motifs influenced reporter expression to a lesser degree than proximal sites. A single zld DNA binding site adjacent to the TSS produced the strongest reporter activity. Importantly, just like the activity of OVO transgenic reporters, there is not an absolute requirement for zld DNA binding to activate reporter expression, however, the addition of TSS adjacent zld DNA binding motifs does strongly influence reporter response. We know that zld achieves this reporter response through its pioneering activity (Xu et al. 2014; Harrison et al. 2011), whether OVO achieves this similar effect on gene expression through a shared mechanism, or in cooperation with other transcription factors needs to be further explored.”

The authors suggest that OVO binding is essential for transcriptional activation, but that this may be indirect and that expression of other transcription factors might be necessary for activating gene expression. Did the motif analysis of the OVO-bound regions suggest additional transcription factors that might provide this function?

We did find other motifs significantly enriched in OVO ChIP peaks. We performed XSTREME analysis on the same set of OVO ChIP peaks which allowed us to determine if any of these motifs were significant matches to DNA binding motifs of known transcription factors. Notably, the DNA binding motifs of GAF and CLAMP were enriched in OVO ChIP peaks. GAF is required in germline clones and the potentially for co-regulation of genes is possible. Other enriched motifs did not match any known binding motifs of other transcription factors but we reported some of the most significantly enriched motifs that were alongside of OVO in Figure S1C-F. The below text outlines changes made to the text incorporating these findings.

Lines 170-182

“Along with the OVO DNA binding motif, other motifs were also significantly enriched in OVO ChIP peaks. The motif 5’-GWGMGAGMGAGABRG-3’ (Figure S1C) was found in 18% of OVO ChIP peaks and is a significant match to the DNA binding motifs of the transcription factors GAF (Trl) (Omelina et al. 2011) and CLAMP (Soruco et al. 2013). Trl germline clones are not viable, indicating that GAF activity is required in the germline during oogenesis (Chen et al. 2009). The possibility that OVO binds with and regulates genes alongside of GAF given the enrichment of both transcription factors DNA binding motifs is intriguing. Other significantly enriched motifs 5’-ACACACACACACACA-3’ (29% of peaks, Figure S1D), 5’RCAACAACAACAACA-3’ (26% of peaks, Figure S1E), and 5’-GAAGAAGAAGAAGAR-3’ (17% of peaks,

Figure S1F) were present in OVO ChIP peaks, however, these motifs did not significantly match known

DNA binding motifs of other transcription factors. Determining the factors that bind to these sequences

will certainly help elucidate our understanding of transcriptional control with relationship to OVO in the female germline.”

The figures would benefit from a bit more detail in the legends (see comments below).

Minor comments:

In multiple places throughout the document, the citations are inadvertently italicized (see lines 57-59, 91, and 327 as examples.)

We have changed this in these locations and other instances in the text.

On line 76, when discussing OVO as a transcription factor this is referencing the protein and not the gene. Thus, should be written OVO and not ovo.

We have made the correction ovo to OVO.

On line 349, "core" promoters is likely what is meant rather than "care" promoters.

We have corrected ‘care’ to ‘core’ in the text.

On line 404, the authors state that they wanted to use a "less conservative log2 fold change" but it is not clear what they are comparing to. This is important to understand the motivation.

We are talking about the gene expression comparison between the ectopic ovo rescue and ovo hypomorphic ovaries. “less conservative” was an unfortunate phrasing. We have rewritten the text to state this directly to the reader.

Lines 435-444

“We then performed RNA-seq in quadruplicate and measured the changes in gene expression between ectopic rescue OVO and hypomorphic OVO ovaries. We used a significance level of p-adj < 0.05 and a log2 fold change cutoff of >|0.5| to call differential expression between these two sets of ovaries. We utilized these log2 fold change cutoffs for two reasons. Our control ovary genotype (ovoovo-GAL4/ovoΔBP; UASp-GFP) has hypomorphic OVO activity, hence germ cells can survive but are arrested. With the addition of ectopic rescue OVO in ovoovo-GAL4/ovoΔBP; UASp-3xFHA-OVO-B ovaries, we predicted that genes that were directly regulated by OVO would transcriptionally respond, however, we were unsure as to what degree the response would be in comparison to hypomorphic OVO. We reasoned that if the changes were not significant between genotypes, then minor changes in gene expression would not matter.”

On line 615, it is unclear what is meant by "showing expression with only 10s of bp of sequence in reporters."

This is in reference to some of the previously studied ovo reporter deletion lines, however, we have decided to remove the below text in the revised discussion.

“, despite being remarkably compact. The OVO-dependent ovo core promoter is very compact; showing expression with only 10s of bp of sequence in reporters.” 

It would be useful to cite and discuss Dufourt et al. Nature Communications 2018 (PMID30518940) regarding the role of Zelda in potentiating transcriptional activation when mentioned on line 624.

We have added this and the relationship to previous similar work on OVO in the discussion.

Lines 645-663

“The requirement for OVO at the TSS of target genes has been well characterized at its own locus as well as its downstream target otu. Our OVO ChIP and expression data confirm findings from previous work that OVO is binding to these target promoters, and in the case of otu, strongly responds transcriptionally to the presence of OVO. Although we did not test the requirement for OVO DNA binding motifs at other OVO bound genes in this work, this has been extensively explored before, showing that removal of OVO

DNA binding sites overlapping the TSS results in a strong decrease in reporter expression (Lü et al. 1998; Bielinska et al. 2005; Lü and Oliver 2001). Removal of more distal upstream OVO DNA binding sites also reduces reporter expression to a lesser degree. However, for most cases tested, removal of OVO DNA binding sites while leaving the rest of the enhancer regions intact, never totally abolished reporter expression. These dynamics are highly similar to work that has been completed on the pioneer factor zelda (zld). Adding zld DNA binding motifs to a stochastically expressed transcriptional reporter increases the activity and response of the reporter (Dufourt et al. 2018). Distally located zld DNA binding motifs influenced reporter expression to a lesser degree than proximal sites. A single zld DNA binding site adjacent to the TSS produced the strongest reporter activity. Importantly, just like the activity of OVO transgenic reporters, there is not an absolute requirement for zld DNA binding to activate reporter expression, however, the addition of TSS adjacent zld DNA binding motifs does strongly influence reporter response. We know that zld achieves this reporter response through its pioneering activity (Xu et al. 2014; Harrison et al. 2011), whether OVO achieves this similar effect on gene expression through a shared mechanism, or in cooperation with other transcription factors needs to be further explored.”

On line 1006 (Figure 1 legend), it is unclear what is meant by "The percentage of OVO ChIP peaks each motif was found". Is a word missing?

This was unclear, we have revised the sentence below.

Lines 1035-1036

“The percentage of OVO ChIP peaks containing each motif and their corresponding p-value are indicated to the right.”

In the Figure 1 legend, please include citations for the Garfinkel motif and Oliver motif.

Included, as below.

Lines 1036-1039

“H) OVO ChIP minus input control ChIP-seq read coverage density centered on the location of the four de novo OVO DNA binding motifs and previously defined in vitro OVO DNA binding motifs (Lü et al. 1998, Bielinska et al. 2005, Lee and Garfinkel 2000).”

In Figure 2 legend, it is unclear if B is all instances of a given motif or the DNA motifs that are bound by ChIP. Please clarify.

We meant only the OVO DNA binding motifs that were within significant OVO ChIP peaks. We have revised the legend below.

Lines 1049-1052

“A, B) OVO ChIP minus input control, GSC and 32c ATAC-seq, GSC H3K27ac, H3K4me3, H3K27me3, H3K9me3, 8c NC H3K9me3, 32c NC H3K27ac, and H3K27me3 ChIP-seq read coverage density centered on each OVO peak maximum or OVO DNA binding motif located within a significant OVO ChIP peak.”

The Figure legend for 2D could use more explanation. What do the lines and circles indicate?

These lines and circles indicate the amount of overlapping peaks measured between the two datasets with solid circles. We have included a better description of what these indicate in the figure legend.

Lines 1054-1058

“D) Total number of significant peaks (left) and the total number of overlapping peaks (top) between OVO

ChIP and GSC and 32c ATAC-seq, GSC H3K27ac, H3K4me3, H3K27me3, H3K9me3, 8c NC H3K9me3, 32c NC H3K27ac, and H3K27me3 ChIP-seq. Lines connecting solid dots indicates the amount of overlapping peaks between those two corresponding datasets.”

In Figure 4C, bring the 564 blue dots forward so they are not masked by the yellow dots.

We have brought the colored dots forward in both figure 4C and 4D.

In Figure 4E, what is the order of the heatmaps?

The order is genes with the highest to lowest OVO read density enrichment. We have included this in the figure 4 legend.

Lines 1086-1087

“The order of the heatmap is genes with the highest to lowest amount of OVO ChIP read density.”

In Figure 5, the order of the tracks is not immediately obvious. It appears to be those chromatin features most associated with OVO ChIP and those less correlated. Additional clarity could be provided by showing these tracks (and in Supplemental Figure S2) in different colors with a reference to the figure legend about what the colors might indicate.

We have changed the colors and order of the tracks to be more similar and consistent in both figures.

Lines 1090-1093

“ovo gene level read coverage tracks for OVO ChIP minus input (black), GSC and 32c ATAC-seq (light blue), GSC and 32C H3K27ac (green), H3K4me3 (dark blue), GSC and 32c H3K27me3 (orange), and GSC and 8c H3K9me3 (pink) ChIP-seq, and ovoΔBP/ovoovo-GAL4; UASp-3xFHA-OVO-B minus ovoΔBP/ovoovo-GAL4; UASp-GFP RNA-seq (red).”

In Figure S1 legend, what is the reference to the da-GAL4 X UAS transgene in the title?

This was an error on our part and we have removed it.

Reviewer #2 (Recommendations For The Authors):

Overall, the manuscript would benefit from revisions of the writing style. At times it is difficult to distinguish between hypothesis and results. The use of colloquial phrases/prose was distracting while reading, which the authors may consider revising. Some sentences were confusing or extraneous, and the authors may consider revising those. Occasionally sentences within the results sections seem more appropriate for the materials and methods.

(1) The manuscript is generally clear; however, it is at times difficult to distinguish between hypothesis and results. The use of colloquial phrases/prose was distracting while reading, which the authors may consider revising. Examples include:

a)  Lines 48-49 "While thematic elements of this complex orchestration have been well studied, coordinate regulation of the symphony has not."

We have edited this sentence below.

Lines 48-50

“While the complex interactions between maternally supplied mRNAs and proteins have been well studied, transcriptional regulation driving the expression of these pathways are less well understood.“

b)  Lines 232-233 "In other words, where exactly does transcription start at these genes."

We have removed this sentence.

c)  Line 385, the word "sham" could be changed to "negative control" or "GFP control"

We have rewritten this sentence below.

Lines 419-423

“Therefore, in order to determine genes that are transcriptionally responsive to OVO, we compared the gene expression profiles in sets of ovaries that had the ovo hypomorphic phenotype with a negative control rescue construct (ovoovo-GAL4/ovoΔBP; UASp-GFP)(Figure 4A) versus those that drive expression of the rescue construct expressing OVO-B (ovoovo-GAL4/ovoΔBP; UASp-3xFHA-OVO-B)(Figure 4B)”

d)  Line 490 "For the big picture"

We have removed this and revised with the below sentence.

Lines 530-531

“To do this, we performed Gene Ontology enrichment analysis with gProfiler software (Raudvere et al. 2019).”

(2) Some sentences were confusing or extraneous, and the authors may consider revising them. Examples include:

a)  Lines 195-196 "Therefore, we plotted the significant ChIP (minus input) read density peaks centered on the location of the motif itself."

We have removed the word ‘peaks’ and ‘itself’, as below.

Lines 200-201

“Therefore, we plotted the significant ChIP (minus input) read density centered on the location of the motif.”

b)  Lines 201-203 "... over the location of the motifs, strongly reinforces the idea that our dataset contains regions centered on sequence-specifically bound OVO transcription factor in the ovary."

We have edited this sentence to clarify below.

Lines 204-208

“While it is possible that OVO comes into contact with regions of DNA in three-dimensional nuclear space non-specifically, the presence of OVO motifs within a large percentage of significant ChIP peaks in vivo and enrichment of OVO ChIP read density at the location of the motifs, strongly reinforces the idea that our OVO ChIP dataset contains regions centered on sequences specifically bound by OVO in the ovary.”

c)  Lines 326-328 "The combinations of these elements...tens of millions of years of evolution."

We have revised this sentence below.

Lines 354-357

“The combinations of these DNA motifs is not random in mammals and Drosophila (FitzGerald et al. 2006), and distinct combinations of different motifs at the TSS of genes expressed in Drosophila are conserved over tens of millions of years of evolution (Chen et al. 2014).”

d)  Lines 444-446 "To address this directly, we tested the idea that genes with... and thus downstream of OVO."

We have removed this sentence in its entirety.

e)  Line 579-580 "Where OVO binding in close proximity, in any ...activates transcription"

We have removed this sentence in its entirety.

(3)    Occasionally sentences within the results sections seem more appropriate for the materials and methods. For example, lines 213-218.

(4)    At the end of line 375, do the authors mean "only" instead of "also"?

We have modified this sentence below.

Lines 411-414

“ovoovo-GAL4 is a CRISPR/Cas9 derived T2A-GAL4-3xSTOP insertion upstream of the splice junction of exon 3 in the ovo-RA transcript (Figure S1A). Importantly, this insertion in the extended exon 3 would disrupt roughly 90% of the ovo-B transcripts. However, since about 10% of ovo-B transcripts utilize an upstream splice junction in exon 3, these transcripts would not be disrupted with the T2A-GAL4-3xSTOP insertion and thus allow for enough OVO activity for germ cell survival (Benner et al. 2023).”

(5)    In line 392 the authors say that they dissected ovaries "one day post-eclosion" but the methods section says that ovaries were 3-5 days old. Please clarify.

We meant one day old for the RNAseq experiments. We have changed this in the text.

Lines 679-681

“Twenty, one day old post-eclosion ovoΔBP/ovoovo-GAL4; UASp-GFP and ovoΔBP/ovoovo-GAL4; UASp-3xFHAOVO-B ovaries were dissected and germariums through previtellogenic egg chambers were removed with microdissection scissors and placed in ice cold PBS making up one biological replicate.”

(6)    In line 668 the authors mention CRISPR/Cas9 in the methods, but no such experiment was described.

We have removed this from the Methods header.

Reviewer #1 (Public Review):

(1) Significance of findings and strength of evidence.

(a) The work presented in this manuscript is intended to support the authors' novel idea that HIV DNA integration strongly favors "triple-stranded" R-loops in DNA formed either during transcription of many, but not all, genes or by strand invasion of silent DNA by transcripts made elsewhere, and that HIV infection promotes R-loop formation mediated by incoming virions in the absence of reverse transcription. The authors were able to demonstrate a reverse transcription-independent increase in R-loop formation early during HIV infection, while also demonstrating increased integration into sequences that contain R-loop structures. Furthermore, this manuscript also identifies that R-loops are present in both transcriptionally active and silent regions of the genome and that HIV integrase interacts with R-loops. Although the work presented supports a correlation between R-loop formation and HIV DNA integration, it does not prove the authors' hypothesis that R-loops are directly targeted for integration. Direct experimentation, such as in vitro integration into defined DNA targets, will be required. Further, the authors provide no explanation as to how current sophisticated structural models of concerted retroviral DNA integration into both strands of double-stranded DNA targets can accommodate triple-stranded structures. Finally, there are serious technical concerns with the interpretation of the integration site analyses.

(2) Public review with guidance for readers around how to interpret the work, highlighting important findings but also mentioning caveats.

(a) Introduction: The authors provide an excellent introduction to R-loops but they base the rationale for this study on mis-citation of earlier studies regarding integration in transcriptionally silent regions of the genome. E "most favored locus" cited in the very old reference 6 comprises only 5 events and has not been reproduced in more recent, much larger datasets. For example, see the study of over 300.000 sites in freshly infected PBMC cited in https://doi.org/10.1371/journal.ppat.1009141, which shows a 15-fold preference for integration in expressed genes and no evidence of clustering of sites (as seen in expressed genes) in non-expressed DNA. Further, as far as I can tell, they present no examples in the Results section of R-loops in non-expressed DNA serving as integration targets.

(b) Figure 1: Demonstrates models for HIV infections in both cell lines and primary human CD4+ T cells. R-loop formation was determined through a method called DRIPc-seq which utilizes an antibody specific for DNA-RNA hybrid structures and sequences these regions of the genome using RNaseH treatment to show that when RNA-DNA hybrids are absent then no R-loops are detected. In these models of in vitro and ex vivo infection, the authors show that R-Loop formation increases following HIV infection between 6 hour post-infection and 12 hours post-infection, depending on the cell model. However, these figures lack a mock-infected control for each cell model to assess R-loop formation at the same time points. They would also benefit from a control showing that virus entry is necessary, such as omitting the VSV G protein donor.

Additionally, they use intracellular staining to confirm DRIPc-Seq results, by demonstrating an increase in R-loop formation at 6 hours post-infection in HeLa cells. It would have been more relevant to use primary T cells for this assay, but HeLa cells probably provided easier and clearer imaging.

(c) Figure 2: This figure shows that cells infected with HIV show more R-loops as well as longer sequences containing R-loop structures. Panel B shows that these R-loops were distributed throughout different genomic features, such as both genic and intergenic regions of the genome. However, the data are presented in such a way that it is impossible to determine the proportion of R-loops in each type of genomic feature. The reader has no way to tell, for example, the proportion of R-loops in genic vs intergenic DNA and how this value changes with time. Furthermore, increased R-loop formation due to HIV infection showed poor correlation with gene expression, suggesting that R-loops were not forming due to transcriptional activation, although the difference between 0 and the remaining time points is not apparent, nor is the meaning of the absurd p values.

(d) Figure 3: This figure shows the use of cell lines carrying R-loop inducible (mAIRN) or non-inducible (ECFP) genes to model the association of HIV integration with R-loop structures. The authors demonstrate the functional validation of R-loop induction in the cell line model. Additionally, when R-loops are induced there is a significant increase in HIV integration in the R-loop forming vector sequence when R-loops are induced with doxycycline. This result shows a correlation between expression and integration that is much stronger in the R-loop forming gene than in the unreferenced ECFP gene but does not prove that integration directly targets R-loops. It is possible, for example, that some features of the DNA sequence, such as base composition affect both integration and R-loop formation independently. As described more fully below, there is also a serious concern regarding the method used to quantify the integration frequencies.

(e) Figure 4: This figure shows evidence of increased HIV integration within regions of the genome containing R-loops with an additional preference for integration within the R-loop and a decrease in frequency of integration further from the R-loop. Identifying a preference for R-loops is very intriguing but the authors do also demonstrate that integration does occur when R-loops are not present. Also Panel A, which shows that regions of cell DNA that form R-loops have a higher frequency of Integration sites than those that do not, should also be controlled for the level of gene expression of the two types of region.

(f) Figure 5: In this figure, the authors demonstrate that HIV integrase binds to R-loops through a number of protein assays, but does not show that this binding is associated with enzymatic activity. ESMA of integrase identified increased binding to DNA-RNA over dsDNA. Additionally, precipitation of RNA-DNA hybrids pulled down HIV integrase. A proximity ligation assay detecting R-loops and HIV-integrase showed co-localization within the nucleus of HeLa cells. HeLa cells were probably used due to their efficiency of transduction but are not physiologically relevant cell types.

(g) Discussion: In the discussion, the authors address how their work relates to previous evidence of HIV integration by association of LEDGF/p75 and CPSF6. They also cite that LEDGF/p75 has possible R-loop binding capabilities. They also discuss what possible mechanisms are driving increases in R-loop formation during HIV infection, pointing to possible HIV accessory proteins. They also state that how HIV integrates in transcriptionally silent regions is still unknown but do point out that they were able to show R-loops appear in many different regions of the genome but did not show that R-loops in transcriptional inactive regions are integration targets. More seriously, they failed to make a connection between their work and the current understanding of the biochemical and structural mechanism of the integration reaction.

eLife assessment

Based on largely indirect evidence, this study proposes that genomic integration of HIV targets DNA/RNA hybrids called R-loops. The evidence is indirect because the authors do not use relevant models systems to show integration and because they artificially induce R-loops in the critical experiments. There are two interrelated findings: 1) VSVg-pseudotyped HIV-1 induces R-loops in various cell types, and 2) VSVg-pseudotyped HIV-1 targets R-loops for integration in an artificial Hela cell model in which R-loops are exogenously induced. The induction of R-loops by a pseudotyped HIV-1 is a potentially valuable finding. Critically, however, because of the caveats above, the evidence is inadequate to support the primary claims in the title, abstract, and manuscript. Furthermore, if these claims were true, the authors do not provide context for how they could be reconciled with well-established structural data showing that HIV-1 integrase catalyzes the integration of viral DNA into dsDNA as a substrate.

Reviewer #2 (Public Review):

Retroviral integration in general, and HIV integration in particular, takes place in dsDNA, not in R-loops. Although HIV integration can occur in vitro on naked dsDNA, there is good evidence that, in an infected cell, integration occurs on DNA that is associated with nucleosomes. This review will be presented in two parts. First, a summary will be provided giving some of the reasons to be confident that integration occurs on dsDNA on nucleosomes. The second part will point out some of the obvious problems with the experimental data that are presented in the manuscript.

(1) 2017 Dos Passos Science paper describes the structure of the HIV intasome. The structure makes it clear that the target for integration is dsDNA, not an R-loop, and there are very good reasons to think that structure is physiologically relevant. For example, there is data from the Cherepanov, Engelman, and Lyumkis labs to show that the HIV intasome is quite similar in its overall structure and organization to the structures of the intasomes of other retroviruses. Importantly, these structures explain the way integration creates a small duplication of the host sequences at the integration site. How do the authors propose that an R-loop can replace the dsDNA that was seen in these intasome structures?

(2) As noted above, concerted (two-ended) integration can occur in vitro on a naked dsDNA substrate. However, there is compelling evidence that, in cells, integration preferentially occurs on nucleosomes. Nucleosomes are not found in R loops. In an infected cell, the viral RNA genome of HIV is converted into DNA within the capsid/core which transits the nuclear pore before reverse transcription has been completed. Integration requires the uncoating of the capsid/core, which is linked to the completion of viral DNA synthesis in the nucleus. Two host factors are known to strongly influence integration site selection, CPSF6 and LEDGF. CPSF6 is involved in helping the capsid/core transit the nuclear pore and associate with nuclear speckles. LEDGF is involved in helping the preintegration complex (PIC) find an integration site after it has been released from the capsid/core, most commonly in the bodies of highly expressed genes. In the absence of an interaction of CPSF6 with the core, integration occurs primarily in the lamin-associated domains (LADs). Genes in LADs are usually not expressed or are expressed at low levels. Depending on the cell type, integration in the absence of CPSF6 can be less efficient than normal integration, but that could well be due to a lack of LEDGF (which is associated with expressed genes) in the LADs. In the absence of an interaction of IN with LEDGF (and in cells with low levels of HRP2) integration is less efficient and the obvious preference for integration in highly expressed genes is reduced. Importantly, LEDGF is known to bind histone marks, and will therefore be preferentially associated with nucleosomes, not R-loops. LEDGF fusions, in which the chromatin binding portion of the protein is replaced, can be used to redirect where HIV integrates, and that technique has been used to map the locations of proteins on chromatin. Importantly, LEDGF fusions in which the chromatin binding component of LEDGF is replaced with a module that recognizes specific histone marks direct integration to those marks, confirming integration occurs efficiently on nucleosomes in cells. It is worth noting that it is possible to redirect integration to portions of the host genome that are poorly expressed, which, when taken with the data on integration into LADs (integration in the absence of a CPSF6 interaction) shows that there are circumstances in which there is reasonably efficient integration of HIV DNA in portions of the genome in which there are few if any R-loops.

(3) Given that HIV DNA is known to preferentially integrate into expressed genes and that R-loops must necessarily involve expressed RNA, it is not surprising that there is a correlation between HIV integration and regions of the genome to which R loops have been mapped. However, it is important to remember that correlation does not necessarily imply causation.

If we consider some of the problems in the experiments that are described in the manuscript:

(1) In an infected individual, cells are almost always infected by a single virion and the infecting virion is not accompanied by large numbers of damaged or defective virions. This is a key consideration: the claim that infection by HIV affects R-loop formation in cells was done with a VSVg vector in experiments in which there appears to have been about 6000 virions per cell. Although most of the virions prepared in vitro are defective in some way, that does not mean that a large fraction of the defective virions cannot fuse with cells. In normal in vivo infections, HIV has evolved in ways that avoid signaling infected the cell of its presence. To cite an example, carrying out reverse transcription in the capsid/core prevents the host cell from detecting (free) viral DNA in the cytoplasm. The fact that the large effect on R-loop formation which the authors report still occurs in infections done in the absence of reverse transcription strengthens the probability that the effects are due to the massive amounts of virions present, and perhaps to the presence of VSVg, which is quite toxic. To have physiological relevance, the infections would need to be carried out with virions that contain HIV even under circumstances in which there is at most one virion per cell.

(2) Using the Sso7d version of HIV IN in the in vitro binding assays raises some questions, but that is not the real question/problem. The real problem is that the important question is not what/how HIV IN protein binds to, but where/how an intasome binds. An intasome is formed from a combination of IN bound to the ends of viral DNA. In the absence of viral DNA ends, IN does not have the same structure/organization as it has in an intasome. Moreover, HIV IN (even Sso7d, which was modified to improve its behavior) is notoriously sticky and hard to work with. If viral DNA had been included in the experiment, intasomes would need to be prepared and purified for a proper binding experiment. To make matters worse, there are multiple forms of multimeric HIV IN and it is not clear how many HIV INs are present in the PICs that actually carry out integration in an infected cell.

(3) As an extension of comment 2, the proper association of an HIV intasome/PIC with the host genome requires LEDGF and the appropriate nucleic acid targets need to be chromatinized.

(4) Expressing any form of IN, by itself, in cells to look for what IN associates with is not a valid experiment. A major factor that helps to determine both where integration takes place and the sites chosen for integration is the transport of the viral DNA and IN into the nucleus in the capsid core. However, even if we ignore that important part of the problem, the IN that the authors expressed in HeLa cells won't be bound to the viral DNA ends (see comment 2), even if the fusion protein would be able to form an intasome. As such, the IN that is expressed free in cells will not form a proper intasome/PIC and cannot be expected to bind where/how an intasome/PIC would bind.

(5) As in comment 1, for the PLA experiments presented in Figure 5 to work, the number of virions used per cell (which differs from the MOI measured by the number of cells that express a viral marker) must have a high, which is likely to have affected the cells and the results of the experiment. However, there is the additional question of whether the IN-GFP fusion is functional. The fact that the functional intasome is a complex multimer suggests that this could be a problem. There is an additional problem, even if IN-GFP is fully functional. During a normal infection, the capsid core will have delivered copies of IN (and, in the experiments reported here, the IN-GFP fusion) into the nucleus that is not part of the intasome. These "free" copies of IN (here IN-GFP) are not likely to go to the same sites as an intasome, making this experiment problematic (comment 4).

(6) In the Introduction, the authors state that the site of integration affects the probability that the resulting provirus will be expressed. Although this idea is widely believed in the field, the actual data supporting it are, at best, weak. See, for example, the data from the Bushman lab showing that the distribution of integration sites is the same in cells in which the integrated proviruses are, and are not, expressed. However, given what the authors claim in the introduction, they should be more careful in interpreting enzyme expression levels (luciferase) as a measure of integration efficiency in experiments in which they claim proviruses are integrated in different places.

(7) Using restriction enzymes to create an integration site library introduces biases that derive from the uneven distribution of the recognition sites for the restriction enzymes.

Reviewer #3 (Public Review):

In this manuscript, Park and colleagues describe a series of experiments that investigate the role of R-loops in HIV-1 genome integration. The authors show that during HIV-1 infection, R-loops levels on the host genome accumulate. Using a synthetic R-loop prone gene construct, they show that HIV-1 integration sites target sites with high R-loop levels. They further show that integration sites on the endogenous host genome are correlated with sites prone to R-loops. Using biochemical approaches, as well as in vivo co-IP and proximity ligation experiments, the authors show that HIV-1 integrase physically interacts with R-loop structures.

My primary concern with the paper is with the interpretations the authors make about their genome-wide analyses. I think that including some additional analyses of the genome-wide data, as well as some textual changes can help make these interpretations more congruent with what the data demonstrate. Here are a few specific comments and questions:

(1) I think Figure 1 makes a good case for the conclusion that R-loops are more easily detected HIV-1 infected cells by multiple approaches (all using the S9.6 antibody). The authors show that their signals are RNase H sensitive, which is a critical control. For the DRIPc-Seq, I think including an analysis of biological replicates would greatly strengthen the manuscript. The authors state in the methods that the DRIPc pulldown experiments were done in biological replicates for each condition. Are the increases in DRIPc peaks similar across biological replicates? Are genomic locations of HIV-1-dependent peaks similar across biological replicates? Measuring and reporting the biological variation between replicate experiments is crucial for making conclusions about increases in R-loop peak frequency. This is partially alleviated by the locus-specific data in Figure S3A. However, a better understanding of how the genome-wide data varies across biological replicates will greatly enhance the quality of Figure 1.

(2) I think that the conclusion that R-loops "accumulate" in infected cells is acceptable, given the data presented. However, in line 134 the authors state that "HIV-1 infection induced host genomic R-loop formation". I suggest being very specific about the observation. Accumulation can happen by (a) inducing a higher frequency of the occurrence of individual R-loops and/or (b) stabilizing existing R-loops. I'm not convinced the authors present enough evidence to claim one over the other. It is altogether possible that HIV-1 infection stabilizes R-loops such that they are more persistent (perhaps by interactions with integrase?), and therefore more easily detected. I think rephrasing the conclusions to include this possibility would alleviate my concerns.

(3) A technical problem with using the S9.6 antibody for the detection of R-loops via microscopy is that it cross-reacts with double-stranded RNA. This has been addressed by the work of Chedin and colleagues (as well as others). It is absolutely essential to treat these samples with an RNA:RNA hybrid-specific RNase, which the authors did not include, as far as their methods section states. Therefore, it is difficult to interpret all of the immunofluorescence experiments that depend on S9.6 binding.

(4) Given that there is no clear correlation between expression levels and R-loop peak detection, combined with the data that show increased detection of R-loop frequency in non-genic regions, I think it will be important to show that the R-loop forming regions are indeed transcribed above background levels. This will help alleviate possible concerns that there are technical errors in R-loop peak detection.

(5) In Figures 4C and D the hashed lines are not defined. It is also interesting that the integration sites do not line up with R-loop peaks. This does not necessarily directly refute the conclusions (especially given the scale of the genomic region displayed), but should be addressed in the manuscript. Additionally, it would greatly improve Figure 4 to have some idea about the biological variation across replicates of the data presented 4A.

(6) The authors do not adequately describe the Integrase mutant that they use in their biochemical experiments in Figure 5A. Could this impact the activity of the protein in such a way that interferes with the interpretation of the experiment? The mutant is not used in subsequent experiments for Figure 5 and so even though the data are consistent with each other (and the conclusion that Integrase interacts with R-loops) a more thorough explanation of why that mutant was used and how it impacts the biochemical activity of the protein will help the interpretation of the data presented in Figure 5.

eLife assessment

This valuable work investigates the role of boundary elements in the formation of 3D genome architecture. The authors established a specific model system that allowed them to manipulate boundary elements and examine the resulting genome topology. The work yielded the first demonstration of the existence of stem and circle loops in a genome and confirms a model which had been posited based on extensive prior genetic work, providing insights into how 3D genome topologies affect enhancer-promoter communication. The evidence is solid, although the degree of generalization remains uncertain.

Reviewer #1 (Public Review):

In this study, the authors engineer the endogenous left boundary of the Drosophila eve TAD, replacing the endogenous Nhomie boundary by either a neutral DNA, a wildtype Nhomie boundary, an inverted Nhomie boundary, or a second copy of the Homie boundary. They perform Micro-C on young embryos and conclude that endogenous Nhomie and Homie boundaries flanking eve pair with head-to-tail directionality to form a chromosomal stem loop. Abrogating the Nhomie boundary leads to ectopic activation of genes in the former neighboring TAD by eve embryonic stripe enhancers. Replacing Nhomie by an inverted version or by Homie (which pairs with itself head-to-head) transformed the stem loop into a circle loop. An important finding was that stem and circle loops differentially impact endogenous gene regulation both within the eve TAD and in the TADs bracketing eve. Intriguingly, an eve TAD with a circle loop configuration leads to ectopic activation of flanking genes by eve enhancers - indicating compromised regulatory boundary activity despite the presence of an eve TAD with intact left and right boundaries.

The results obtained are of high-quality and are meticulously discussed. This work advances our fundamental understanding of how 3D genome topologies affect enhancer-promoter communication.

This study raises interesting questions to be addressed in future studies.

First, given the unique specificity with which Nhomie and Homie pair (and exhibit "homing" activity), the generalizability of TAD formation by directional boundary pairing remains unclear. Testing whether boundary pairing is a phenomenon restricted to exceptional loci picked for study, rather than a broader rule of TAD formation, would best be done through the development of untargeted approaches to study boundary pairing.

Second, boundary pairing is one of several mechanisms that may form chromosomal contact domains such as TADs. Other mechanisms include cohesin-mediated chromosomal loop extrusion and the inherent tendency of transcriptionally active and inactive chromatin to segregate (or compartmentalize). The functional interplay between these possible TAD-forming mechanisms remains to be further investigated.

Reviewer #2 (Public Review):

This study reports a set of experiments and subsequent analyses focusing on the role of Drosophila boundary elements in shaping 3D genome structure and regulating gene expression. The authors primarily focus on the region of the fly genome containing the even skipped (eve) gene; eve is expressed in a canonical spatial pattern in fly embryos and its locus is flanked by the well-characterized neighbor of homie (nhomie) and homie boundary elements. The main focus of the investigation is the orientation dependence of these boundary elements, which had been observed previously using reporter assays. In this study, the authors use Crispr/Cas9 editing followed by recombination-mediated cassette exchange to create a series of recombinant fly lines in which the nhomie boundary element is either replaced with exongenous sequence from phage 𝝀, an inversion of nhomie, or a copy of homie that has the same orientation as the endogenous homie sequence. The nhomie sequence is also regenerated in its native orientation to control for effects introduced by the transgenesis process.

The authors then perform high-resolution Micro-C to analyze 3D structure and couple this with fluorescent and colorimetric RNA in situ hybridization experiments to measure the expression of eve and nearby genes during different stages of fly development. The major findings of these experiments are that total loss of boundary sequence (replacement with 𝝀 DNA) results in major 3D structure changes and the most prominent observed gene changes, while inversion of the nhomie boundary or replacement with homie resulted in more modest effects in terms of 3D structure and gene expression changes and a distinct pattern of gene expression change from the 𝝀 DNA replacement. As the samples in which the nhomie boundary is inverted or replaced with homie have similar Micro-C profiles at the eve locus and show similar patterns of a spurious gene activation relative to the control, the observed effects appear to be driven by the relative orientation of the nhomie and homie boundary elements to one another.

Collectively, the findings reported in the manuscript are of broad interest to the 3D genome field. Although extensive work has gone into characterizing the patterns of 3D genome organization in a whole host of species, the underlying mechanisms that structure genomes and their functional consequences are still poorly understood. The perhaps best understood system, mechanistically, is the coordinated action of CTCF with the cohesin complex, which in vertebrates appears to shape 3D contact maps through a loop extrusion-pausing mechanism that relies on orientation-dependent sequence elements found at the boundaries of interacting chromatin loops. Despite having a CTCF paralog and cohesin, the Drosophila genome does not appear to be structured by loop extrusion-pausing. The identification of orientation-dependent elements with pronounced structural effects on genome folding thus may shed light on alternative mechanisms used to regulated genome structure, which in turn may yield insights into the significance of particular folding patterns.

On the whole, this study is comprehensive and represents a useful contribution to the 3D genome field. The transgenic lines and Micro-C datasets generated in the course of the work will be valuable resources for the research community. Moreover, the manuscript, while dense in places, is generally clearly written and comprehensive in its description of the work. However, I have a number of comments and critiques of the manuscript, mainly centering on the framing of the experiments and presentation of the Micro-C results and on the manner in which the data are analyzed and reported.

As this document now reflects my review of a revised version of the initial preprint, I will begin to add the new content at this point. As discussed in detail in the following paragraphs, my initial impression of the manuscript has not changed, so I have accordingly left the above text unaltered.

In my initial review, I provided a number of suggestions to improve the quality of the manuscript. These suggestions, which took the form of six major and three minor points, largely focused on 1) altering the writing in certain places to make the story more broadly accessible to the readership and 2) the inclusion of key, missing methodological detail to increase the rigor and reproducibility of the study. No new experiments were requested, and all of the points could be readily addressed with rather straightforward textual changes.

In their revised manuscript, the authors elected to directly address one of the major points and two of the minor points (major point 4, minor points 1 and 3). The remainder of my suggestions remain entirely unaddressed. A similar level of responsiveness was afforded to the very reasonable critiques of the other Reviewer and the Reviewing Editor. The authors have instead largely chosen to respond to the points raised exclusively in the rebuttal document. This document sprawls across >22 pages, includes numerous in-line figures, and cites dozens of references. The tone of this document, in many places, is at best forceful. In a less generous interpretation, many sections are combative, dismissive, and borderline unprofessional.

It is my opinion that the authors are doing the scientific community a disservice with their response. While it is my understanding that readers will be able see the rebuttal letter, I find that end result far from satisfying. How many readers will take the trouble to access that file, versus the manuscript itself? Skirting the review critiques places an unfair burden on readers, who are expecting peer-reviewed science, to dig into the accessory files to follow the critique and response, rather than seeing in reflected in the final product as they accustomed. Intentionally or not, the tactics the authors have chosen detract from what is otherwise a novel and well-intentioned new publishing model. It is also worth pointing out that peer review is done as an act of service to the scientific community, as the senior authors are doubtless aware. The other reviewer, the Reviewing Editor, and I have all taken time away from advancing our own careers and those of our trainees to offer the thoughtful critiques that were so pointedly dismissed.

In summary, as the vast majority of my critiques remain unaddressed, I have simply reproduced them below.

Major Points:

(1) The authors motivate much of the introduction and results with hypothetical "stem loop" and "circle loop" models of chromosome confirmation, which they argue are reflected in the Micro-C data and help to explain the observed ISH patterns. While such structures may possibly form, the support for these specific models vs. the many alternatives is not in any way justified. For instance, no consideration is given to important biophysical properties such as persistence length, packing/scaling, and conformational entropy. As the biophysical properties of chromatin are a very trafficked topic both in terms of experimentation and computational modeling and generally considered in the analysis of chromosome conformation data, the study would be strengthened by acknowledgement of this body of work and more direct integration of its findings.

(2) Similar to Point 1, while there is a fair amount of discussion of how the observed results are or are not consistent with loop extrusion, there is no discussion of the biophysical forces that are thought to underly compartmentalization such as block-polymer co-segregation and their potential influence. I found this absence surprising, as it is generally accepted that A/B compartmentalization essentially can explain the contact maps observed in Drosophila and other non-vertebrate eukaryotes (Rowley, ..., Corces 2017; PMID 28826674). The manuscript would be strengthened by consideration of this phenomenon.

(3) The contact maps presented in the study represent many cells and distinct cell types. It is clear from single-cell Hi-C and multiplexed FISH experiments that chromosome conformation is highly variable even within populations of the same cell, let alone between cell types, with structures such as TADs being entirely absent at the single cell level and only appearing upon pseudobulking. It is difficult to square these observations with the models of relatively static structures depicted here. The authors should provide commentary on this point.

(4) Related to Point 4, the lack of quantitative details about the Micro-C data make it difficult to evaluate if the changes observed are due to biological or technical factors. It is essential that the authors provide quantitative means of controlling for factors like sampling depth, normalization, and data quality between the samples.

(5) The ISH effects reported are modest, especially in the case of the HCR. The details provided for how the imaging data were acquired and analyzed are minimal, which makes evaluating them challenging. It would strengthen the study to provide much more detail about the acquisition and analysis and to include depiction of intermediates in the analysis process, e.g. the showing segmentation of stripes.

Author response:

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

Public Reviews:

Reviewer #1 (Public Review):

Summary:

In this study, the authors engineer the endogenous left boundary of the Drosophila eve TAD, replacing the endogenous Nhomie boundary by either a neutral DNA, a wildtype Nhomie boundary, an inverted Nhomie boundary, or a second copy of the Homie boundary. They perform Micro-C on young embryos and conclude that endogenous Nhomie and Homie boundaries flanking eve pair with head-to-tail directionality to form a chromosomal stem loop. Abrogating the Nhomie boundary leads to ectopic activation of genes in the former neighboring TAD by eve embryonic stripe enhancers. Replacing Nhomie by an inverted version or by Homie (which pairs with itself head-to-head) transformed the stem loop into a circle loop. An important finding was that stem and circle loops differentially impact endogenous gene regulation both within the eve TAD and in the TADs bracketing eve. Intriguingly, an eve TAD with a circle loop configuration leads to ectopic activation of flanking genes by eve enhancers - indicating compromised regulatory boundary activity despite the presence of an eve TAD with intact left and right boundaries.

Strengths:

Overall, the results obtained are of high-quality and are meticulously discussed. This work advances our fundamental understanding of how 3D genome topologies affect enhancer-promoter communication.

Weaknesses:

Though convincingly demonstrated at eve, the generalizability of TAD formation by directional boundary pairing remains unclear, though the authors propose this mechanism could underly the formation of all TADs in Drosophila and possibly even in mammals. Strong and ample evidence has been obtained to date that cohesin-mediated chromosomal loop extrusion explains the formation of a large fraction of TADs in mammals. 

(1.1) The difficultly with most all of the studies on mammal TADs, cohesin and CTCF roadblocks is that the sequencing depth is not sufficient, and large bin sizes (>1 kb) are needed to visualize chromosome architecture.  The resulting contact profiles show TAD neighborhoods, not actual TADs.

The problem with these studies is illustrated by comparing the contact profiles of mammalian MicroC data sets at different bin sizes in Author response image 1.  In this figure, the darkness of the “pixels” in panels E, F, G and H was enhanced by reducing brightness in photoshop.

Author response image 1.

Mammalian MicroC profiles different bun sizes

Panels A and C show “TADs” using bin sizes typical of most mammalian studies (see Krietenstein et al. (2023) (Krietenstein et al. 2020)).  At this level of resolution, TADs, the “trees” that are the building blocks of chromosomes, are not visible.  Instead, what is seen are TAD neighborhoods or “forests”.  Each neighborhood consists of several dozen individual TADs.  The large bins in these panels also artificially accentuated TAD:TAD interactions, generating a series of “stripes” and “dots” that correspond to TADs bumping into each other and sequences getting crosslinked.  For example, in panel A there is prominent stripe on the edge of a “TAD” (blue arrow).  In panel C, this stripe resolves into a series of dots arranged as parallel, but interrupted “stripes” (green and blue arrows).  At the next level of resolution, it can be seen that the stripe marked by the blue arrow and magenta asterisk is generated by contacts between the left boundary of the TAD indicated by the magenta bar with sequences in a TAD (blue bar) ~180 kb way.  While dots and stripes are prominent features in contact profiles visualized with larger bin sizes (A and C), the actual TADs that are observed with a bin size of 200 bp (examples are underlined by black bars in panel G) are not bordered by stripes, nor are they topped by obvious dots.  The one possible exception is the dot that appears at the top of the volcano triangle underlined with magenta.

The chromosome 1 DNA segment from the MicroC data of Hseih et al. (2023) (Hsieh et al. 2020) shows a putative volcano triangle with a plume (indicated by a V in Author response image 1 panels D, F and H).  Sequences in the V TAD don’t crosslink with their immediate neighbors, and this gives a “plume” above the volcano triangle, as indicate by the light blue asterisk in panels D, F and H.  Interestingly the V TAD does contact two distant TADs, U on the left and W on the right. The U TAD is ~550 kb from V, and the region of contact is indicated by the black arrow.  The W TAD is ~585 kb from V, and the region of contact is indicated by the magenta arrow.  While the plume still seems to be visible with a bin size of 400 bp (light blue asterisk), it is hard to discern when the bin size is 200 bp, as there are not enough reads.

The evidence demonstrating that cohesin is required for TAD formation/maintenance is based on low resolution Hi-C data, and the effects that are observed are on TAD neighborhoods (forests) and not TADs (trees).  In fact, there is published evidence that cohesin is not required in mammals for TAD formation/maintenance.  In an experiment from Goel et al. 2023 the authors depleted the cohesin component Rad21 and then visualized the effects on TAD organization using the high resolution region capture MicroC (RCMC) protocol.  The MicroC contact map in this figure visualizes a ~250 kb DNA segment around the Ppm1pg locus at 250 bp resolution.  On the right side of the diagonal is the untreated control, while the left side shows the MicroC profile of the same region after Rad21 depletion.  The authors indicated that there was a 97% depletion of Rad21 in their experiment.  However, as is evident from a comparison of the experimental and control, loss of Rad21 has no apparent effect on the TAD organization of this mammalian DNA segment.

Several other features are worth noting.  First, unlike the MicroC experiments shown in Author response image 1, there are dots at the apex of the TADs in this chromosomal segment.  In the MicroC protocol, fixed chromatin is digested to mononucleosomes by extensive MNase digestion.  The resulting DNA fragments are then ligated, and dinucleosome-length fragments are isolated and sequenced. 

DNA sequences that are nucleosome free in chromatin (which would be promoters, enhancers, silencers and boundary elements) are typically digested to oligonucleotides in this procedure and won’t be recovered. This means that the dots shown here must correspond to mononucleosome-length elements that are MNase resistant.  This is also true for the dots in the MicroC contact profiles of the Drosophila Abd-B regulatory domain (see Fig. 2B in the paper).  Second, the TADs are connected to each other by 45o stripes (see blue and green arrowheads).  While it is not clear from this experiment whether the stipes are generated by an active mechanism (enzyme) or by some “passive” mechanism (e.g., sliding), the stripes in this chromosomal segment are not generated by cohesin, as they are unperturbed by Rad21 depletion.  Third, there are no volcano triangles with plumes in this chromosomal DNA segment.  Instead, the contact patterns (purple and green asterisks) between neighboring TADs closely resemble those seen for the Abd-B regulatory domains (compare Goel et al. 2023 with Fig. 2B in the paper).  This similarity suggests that the TADs in and around Ppm1g may be circle-loops, not stem-loops.  As volcano triangles with plumes also seem to be rare in the MicroC data sets of Krietenstein et al. (Krietenstein et al. 2020) and Hesih et al. (Hsieh et al. 2020) (with the caveat that these data sets are low resolution: see Author response image 1), it is possible that much of the mammalian genome is assembled into circle-loop TADs, a topology that can’t be generated by the cohesin loop extrusion (bolo tie clip) /CTCF roadblock model.

While Rad21 depletion has no apparent effect on TADs, it does appear to impact TAD neighborhoods.  This is in a supplemental figure in Goel et al. (Goel et al. 2023).  In this figure, TADs in the Ppm1g region of chromosome 5 are visualized with bin sizes of 5 kb and 1 kb.  A 1.2 Mb DNA segment is shown for the 5 kb bin size, while an 800 kb DNA segment is shown for the 1 kb bin size.  As can be seen from comparing the MicroC profiles in Author response image 2 with that in Goel et al. 2023, individual TADs are not visible.  Instead, the individual TADs are binned into large TAD “neighborhoods” that consist of several dozen or more TADs.

Unlike the individual TADs shown in Goel et al. 2023, the TAD neighborhoods in Author response image 2 are sensitive to Rad21 depletion.  The effects of Rad21 depletion can be seen by comparing the relative pixel density inside the blue lines before (above the diagonal) and after (below the diagonal) auxin-induced Rad21 degradation.  The reduction in pixel density is greatest for more distant TAD:TAD contacts (farthest from the diagonal).  By contrast, the TADs themselves are unaffected (Goel et al. 2023), as are contacts between individual TADs and their immediate neighbors.  In addition, contacts between partially overlapping TAD neighborhoods are also lost.  At this point it isn’t clear why contacts between distant TADs in the same neighborhood are lost when Rad21 is depleted; however, a plausible speculation is that it is related to the functioning of cohesin in holding newly replicated DNAs together until mitosis and whatever other role it might have in chromosome condensation.

Author response image 2.

Ppm1g full locus chr5

Moreover, given the unique specificity with which Nhomie and Homie are known to pair (and exhibit "homing" activity), it is conceivable that formation of the eve TAD by boundary pairing represents a phenomenon observed at exceptional loci rather than a universal rule of TAD formation. Indeed, characteristic Micro-C features of the eve TAD are only observed at a restricted number of loci in the fly genome…..

(1.2) The available evidence does not support the claim that nhomie and homie are “exceptional.”  To begin with, nhomie and homie rely on precisely the same set of factors that have been implicated in the functioning of other boundaries in the fly genome.  For example, homie requires (among other factors) the generic boundary protein Su(Hw) for insulation and long-distance interactions (Fujioka et al. 2024).  (This is also true of nhomie: unpublished data.)  The Su(Hw) protein (like other fly polydactyl zinc finger proteins) can engage in distant interactions.  This was first shown by Sigrist and Pirrotta (Sigrist and Pirrotta 1997), who found that the su(Hw) element from the gypsy transposon can mediate long-distance regulatory interactions (PRE dependent silencing) between transgenes inserted at different sites on homologous chromosomes (trans interactions) and at sites on different chromosomes.

The ability to mediate long-distance interactions is not unique to the su(Hw) element, or homie and nhomie.  Muller et al. (Muller et al. 1999) found that the Mcp boundary from the Drosophila BX-C is also able to engage in long-distance regulatory interactions—both PRE-dependent silencing of mini-white and enhancer activation of mini-white and yellow.  The functioning of the Mcp boundary depends upon two other generic insulator proteins, Pita and the fly CTCF homolog (Kyrchanova et al. 2017).  Like Su(Hw) both are polydactyl zinc finger proteins, and they resemble the mammalian CTCF protein in that their N-terminal domain mediates multimerization (Bonchuk et al. 2020; Zolotarev et al. 2016).  Figure 6 from Muller et el. 1999 shows PRE-dependent “pairing sensitive silencing” interactions between transgenes carrying a mini-white reporter, the Mcp and scs’ (Beaf dependent)(Hart et al. 1997) boundary elements, and a PRE closely linked to Mcp.  In this experiment flies homozygous for different transgene inserts were mated and the eye color was examined in their transheterozygous progeny.  As indicated in the figure, the strongest trans-silencing interactions were observed for inserts on the same chromosomal arm; however, transgenes inserted on the left arm of chromosome 3 can interact across the centromere with transgenes inserted on the right arm of chromosome 3. 

Figure 5C (left) from Muller et el. 1999 shows a trans-silencing interaction between w#11.102 at 84D and w#11.16 approximately 5.8 Mb away, at 87D.  Figure 5C (right) shows a trans-silencing interaction across the centromere between w#14.29 on the left arm of chromosome 3 at 78F and w#11.102 on the right arm of chromosome 3 at 84D. The eye color phenotype of mini-white-containing transgenes is usually additive: homozygyous inserts have twice as dark eye color as the corresponding hemizygous inserts.  Likewise, in flies trans-_heterozygous for _mini-white transgenes inserted at different sites, the eye color is equivalent to the sum of the two transgenes.  This is not true when mini-white transgenes are silenced by PREs.  In the combination shown in panel A, the t_rans-_heterozygous fly has a lighter eye color than either of the parents.  In the combination in panel B, the _trans-_heterozygous fly is slightly lighter than either parent.

As evident from the diagram in Figure 6 from Muller et el. 1999, all of the transgenes inserted on the 3rd chromosome that were tested were able to participate in long distance (>Mbs) regulatory interactions.  On the other hand, not all possible pairwise interactions are observed.  This would suggest that potential interactions depend upon the large scale (Mb) 3D folding of the 3rd chromosome.

When the scs boundary (Zw5 dependent) (Gaszner et al. 1999) was added to the transgene to give sMws’, it further enhanced the ability of distant transgenes to find each other and pair.  All eight of the sMws’ inserts that were tested were able to interact with at least one other sMws’ insert on a different chromosome and silence mini-white.  Vazquez et al. () subsequently tagged the sMws’ transgene with LacO sequences (ps0Mws’) and visualized pairing interactions in imaginal discs.  Trans-heterozygous combinations on the same chromosome were found paired in 94-99% of the disc nuclei, while a trans-heterozygous combination on different chromosomes was found paired in 96% of the nuclei (Table 3 from Vazquez et al. 2006).  Vazquez et al. also examined a combination of four transgenes inserted on the same chromosome (two at the same insertion site, and two at different insertion sites).  In this case, all four transgenes were clustered together in 94% of the nuclei (Table 3 from Vazquez et al. 2006).  Their studies also suggest that the distant transgenes remain paired for at least several hours.  A similar experiment was done by Li et al. (Li et al. 2011), except that the transgene contained only a single boundary, Mcp or Fab-7.  While pairing was still observed in trans-heterozygotes, the frequency was reduced without scs and scs’.

It is worth pointing out that there is no plausible mechanism in which cohesin could extrude a loop through hundreds of intervening TADs, across the centromere (ff#13.101_ßà_w#11.102: Figure 6 from Muller et el. 1999; w#14.29_ßà_w#11.02: Figure 6 from Muller et el. 1999 and 5) and come to a halt when it “encounters” Mcp containing transgenes on different homologs.  The same is true for Mcp-dependent pairing interactions in cis (Fig. 7 in Muller et al. (Muller et al. 1999)) or Mcp-dependent pairing interactions between transgenes inserted on different chromosomes (Fig. 8 in Muller et al. (Muller et al. 1999); Line 8 in Table 3 from Vazquez et al. 2006). 

These are not the only boundaries that can engage in long-distance pairing.  Mohana et al. (Mohana et al. 2023) identified nearly 60 meta-loops, many of which appear to be formed by the pairing of TAD boundary elements.  Two examples (at 200 bp resolution from 12-16 hr embryos) are shown in Author response image 3.

Author response image 3.

Metaloops on the 2nd and 3rd chromosomes: circle-loops and multiple stem-loops

One of these meta-loops (panel A) is generated by the pairing of two TAD boundaries on the 2nd chromosome.  The first boundary, blue, (indicated by blue arrow) is located at ~2,006, 500 bp between a small TAD containing the Nplp4 and CG15353 genes and a larger TAD containing 3 genes, CG33543, Obp22a and Npc2a.  Nplp4 encodes a neuropeptide.  The functions of CG15354 and CG33543 are unknown.  Obp22a encodes an odorant binding protein, while Npc2a encodes the Niemann-Pick type C-2a protein which is involved sterol homeostasis.  The other boundary (purple: indicated by purple arrow) is located between two TADs 2.8 Mb away at 4,794,250 bp.  The upstream TAD contains the fipi gene (CG15630) which has neuronal functions in male courtship, while the downstream TAD contains CG3294, which is thought to be a spliceosome component, and schlaff (slf) which encodes a chitin binding protein.  As illustrated in the accompanying diagram, the blue boundary pairs with the purple boundary in a head-to-head orientation, generating a ~2.8 Mb loop with a circle-loop topology.  As a result of this pairing, the multi-gene (CG33543, Obp22a and Npc2a) TAD upstream of the blue boundary interacts with the CG15630 TAD upstream of the purple boundary.  Conversely the small Nplp4:CG15353 TAD downstream of the blue boundary interacts with the CG3294:slf TAD downstream of the purple boundary.  Even if one imagined that the cohesin bolo tie clip was somehow able to extrude 2.8 Mb of chromatin and then know to stop when it encountered the blue and purple boundaries, it would’ve generated a stemloop, not a circle-loop.

The second meta-loop (panel B) is more complicated as it is generated by pairing interactions between four boundary elements.  The blue boundary (blue arrow) located ~4,801,800 bp (3L) separates a large TAD containing the RhoGEF64C gene from a small TAD containing CG7509, which encodes a predicted subunit of an extracellular carboxypeptidase.  As can be seen in the MicroC contact profile and the accompanying diagram, the blue boundary pairs with the purple boundary (purple arrow) which is located at ~7,013, 500 (3L) just upstream of the 2nd internal promoter (indicated by black arrowhead) of the Mp (Multiplexin) gene.  This pairing interaction is head-to-tail and generates a large stem-loop that spans ~2.2 Mb.  The stem-loop brings sequences upstream of the blue boundary and downstream of the purple boundary into contact (the strings below a bolo tie clip), just as was observed in the boundary bypass experiments of Muravyova et al. (Muravyova et al. 2001) and Kyrchanova et al. (Kyrchanova et al. 2008).  The physical interactions result in a box of contacts (right top) between sequences in the large RhoGEF64C TAD and sequences in a large TAD that contains an internal Mp promoter.  The second pairing interaction is between the brown boundary (brown arrow) and the green boundary (green arrow).  The brown boundary is located at ~4 805,600 bp (3L) and separates the TAD containing CG7590 from a large TAD containing CG1808 (predicted to encode an oxidoreductase) and the Dhc64C (Dynein heavy chain 64C) gene.  The green boundary is located at ~6,995,500 bp (3L), and it separates a TAD containing CG32388 and the biniou (bin) transcription factor from a TAD that contains the most distal promoter of the Mp (Multiplexin) gene (blue arrowhead).  As indicated in the diagram, the brown and green boundaries pair with each other head-to-tail, and this generates a small internal loop (and the final configuration would resemble a bolo tie with two tie clips).  This small internal loop brings the CG7590 TAD into contact with the TAD that extends from the distal Mp promoter to the 2nd internal Mp promoter.  The resulting contact profile is a rectangular box with diagonal endpoints corresponding to the paired blue:purple and brown:green boundaries.  The pairing of the brown:green boundaries also brings the TADs immediately downstream of the brown boundary and upstream of the green boundary into contact with each other, and this gives a rectangular box of interactions between the Dhc64C TAD, and sequences in the bin/CG3238 TAD.  This box is located on the lower left side of the contact map.

Since the bin and Mp meta-loops in Author response image 3B are stem-loops, they could have been generated by “sequential” cohesin loop extrusion events.  Besides the fact that cohesin extrusion of 2 Mb of chromatin and breaking through multiple intervening TAD boundaries challenges the imagination, there is no mechanism in the cohesion loop extrusion/CTCF roadblock model to explain why cohesion complex 1 would come to a halt at the purple boundary on one side and the blue boundary on the other, while cohesin complex 2 would instead stop when it hits the brown and green boundaries.  This highlights another problem with the cohesin loop extrusion/CTCF roadblock model, namely that the roadblocks are functionally autonomous: they have an intrinsic ability to block cohesin that is entirely independent of the intrinsic ability of other roadblocks in the neighborhood.  As a result, there is no mechanism for generating specificity in loop formation.  By contrast, boundary pairing interactions are by definition non-autonomous and depend on the ability of individual boundaries to pair with other boundaries: specificity is built into the model. The mechanism for pairing, and accordingly the basis for partner preferences/specificity, are reasonably well understood.  Probably the most common mechanism in flies is based on shared binding sites for architectural proteins that can form dimers or multimers (Bonchuk et al. 2021; Fedotova et al. 2017).  Flies have a large family of polydactyl zinc finger DNA binding proteins, and as noted above, many of these form dimers or multimers and also function as TAD boundary proteins.  This pairing principle was first discovered by Kyrchanova et al. (Kyrchanova et al. 2008).  This paper also showed that orientation-dependent pairing interactions is a common feature of endogenous fly boundaries.  Another mechanism for pairing is specific protein:protein interactions between different DNA binding factors (Blanton et al. 2003).  Yet a third mechanism would be proteins that bridge different DNA binding proteins together.  The boundaries that use these different mechanisms (BX-C boundaries, scs, scs’) depend upon the same sorts of proteins that are used by homie and nhomie.  Likewise, these same set of factors reappear in one combination or another in most other TAD boundaries.  As for the orientation of pairing interactions, this is most likely determined by the order of binding sites for chromosome architectural proteins in the partner boundaries.

…and many TADs lack focal 3D interactions between their boundaries.

(1.3) The idea that flies differ from mammals in that they “lack” focal 3D interactions is simply mistaken.  One of the problems with drawing this distinction is that most all of the “focal 3D interactions” seen mammalian Hi-C experiments are a consequence of binning large DNA segments in low resolution restriction enzyme-dependent experiments.  This is even true in the two “high” resolution MicroC experiments that have been published (Hsieh et al. 2020; Krietenstein et al. 2020).  As illustrated above in Author response image 1, most of the “focal 3D interactions” (the dots at the apex of TAD triangles) seen with large bin sizes (1 kb and greater) disappear when the bin size is 200 bp and TADs rather than TAD neighborhoods are being visualized.

As described in point #1.1, in the MicroC protocol, fixed chromatin is first digested to mononucloesomes by extensive MNase digestion, processed/biotinylated, and ligated to give dinucleosome-length fragments, which are then sequenced.  Regions of chromatin that are nucleosome free (promoters, enhancers, silencers, boundary elements) will typically be reduced to oligonucleotides in this procedure and will not be recovered when dinucleosome-length fragments are sequenced.  The loss of sequences from typical paired boundary elements is illustrated by the lar meta-loop shown in Author response image 4 (at 200 bp resolution).  Panels A and B show the contact profiles generated when the blue boundary (which separates two TADs that span  the Lar (Leukocyteantigen-related-like) transcription unit interacts with the purple boundary (which separates two TADs in a gene poor region ~620 kb away).  The blue and purple boundaries pair with each other head-to-head, and this pairing orientation generates yet another circle-loop.  In the circle-loop topology, sequences in the TADs upstream of both boundaries come into contact with each other, and this gives the small dark rectangular box to the upper left of the paired boundaries (Author response image 4A).  (Note that this small box corresponds to the two small TADs upstream of the blue and purple boundaries, respectively. See panel B.)  Sequences in the TADs downstream of the two boundaries also come into contact with each other, and this gives the large box to the lower right of the paired boundaries.  While this meta-loop is clearly generated by pairing interactions between the blue and purple boundaries, the interacting sequences are degraded in the MicroC protocol, and sequences corresponding to the blue and purple boundaries aren’t recovered.  This can be seen in panel B (red arrow and red arrowheads).  When a different Hi-C procedure is used (dHS-C) that captures nucleosome-free regions of chromatin that are physically linked to each other (Author response image 4C & D), the sequences in the interacting blue and purple boundaries are recovered and generate a prominent “dot” at their physical intersection (blue arrow in panel D).

Author response image 4.

Lar metaloop. Panels A & bB: MicroC. Panels C & D: dHS-C

While sequences corresponding to the blue and purple boundaries are lost in the MicroC procedure, there is at least one class of elements that engage in physical pairing interactions whose sequences are (comparatively) resistant to MNase digestion.  This class of elements includes many PREs ((Kyrchanova et al. 2018); unpublished data), the boundary bypass elements in the Abd-B region of BX-C (Kyrchanova et al. 2023; Kyrchanova et al. 2019a; Kyrchanova et al. 2019b; Postika et al. 2018), and “tethering” elements (Batut et al. 2022; Li et al. 2023).  In all of the cases tested, these elements are bound in nuclear extracts by a large (>1000 kD) GAGA factor-containing multiprotein complex called LBC.  LBC also binds to the hsp70 and eve promoters (unpublished data).  Indirect end-labeling experiments (Galloni et al. 1993; Samal et al. 1981; Udvardy and Schedl 1984) indicate that the LBC protects a ~120-180 bp DNA segment from MNase digestion.  It is likely that this is the reason why LBC-bound sequences can be recovered in MicroC experiments as dots when they are physically linked to each other.  One such example (based on the ChIP signatures of the paired elements) is indicated by the green arrow in panel B and D of Author response image 4.  Note that there are no dots corresponding to these two LBC elements within either of the TADs immediately downstream of the blue and purple boundaries.  Instead the sequences corresponding to the two LBC elements are only recovered when the two elements pair with each other over a distance of ~620 kb.  The fact that these two elements pair with each other is consistent with other findings which indicate that, like classical boundaries, LBC elements exhibit partner preferences.  In fact, LBC elements can sometimes function as TAD boundaries.  For example, the Fab-7 boundary has two LBC elements, and full Fab-7 boundary function can be reconstituted with just these two elements (Kyrchanova et al. 2018).

Reviewer #2 (Public Review):

"Chromatin Structure II: Stem-loops and circle-loops" by Ke*, Fujioka*, Schedl, and Jaynes reports a set of experiments and subsequent analyses focusing on the role of Drosophila boundary elements in shaping 3D genome structure and regulating gene expression. The authors primarily focus on the region of the fly genome containing the even skipped (eve) gene; eve is expressed in a canonical spatial pattern in fly embryos and its locus is flanked by the well-characterized neighbor of homie (nhomie) and homie boundary elements. The main focus of investigation is the orientation dependence of these boundary elements, which had been observed previously using reporter assays. In this study, the authors use Crispr/Cas9 editing followed by recombination-mediated cassette exchange to create a series of recombinant fly lines in which the nhomie boundary element is either replaced with exongenous sequence from phage 𝝀, an inversion of nhomie, or a copy of homie that has the same orientation as the endogenous homie sequence. The nhomie sequence is also regenerated in its native orientation to control for effects introduced by the transgenesis process.

The authors then perform high-resolution Micro-C to analyze 3D structure and couple this with fluorescent and colorimetric RNA in situ hybridization experiments to measure the expression of eve and nearby genes during different stages of fly development. The major findings of these experiments are that total loss of boundary sequence (replacement with 𝝀 DNA) results in major 3D structure changes and the most prominent observed gene changes, while inversion of the nhomie boundary or replacement with homie resulted in more modest effects in terms of 3D structure and gene expression changes and a distinct pattern of gene expression change from the 𝝀 DNA replacement. As the samples in which the nhomie boundary is inverted or replaced with homie have similar Micro-C profiles at the eve locus and show similar patterns of a spurious gene activation relative to the control, the observed effects appear to be driven by the relative orientation of the nhomie and homie boundary elements to one another.

Collectively, the findings reported in the manuscript are of broad interest to the 3D genome field. Although extensive work has gone into characterizing the patterns of 3D genome organization in a whole host of species, the underlying mechanisms that structure genomes and their functional consequences are still poorly understood. The perhaps best understood system, mechanistically, is the coordinated action of CTCF with the cohesin complex, which in vertebrates appears to shape 3D contact maps through a loop extrusion-pausing mechanism that relies on orientation-dependent sequence elements found at the boundaries of interacting chromatin loops.

(2.1) The notion that mammalian genome is shaped in 3D by the coordinate action of cohesin and CTCF has achieved the status of dogma in the field of chromosome structure in vertebrates.  However, as we have pointed out in #1.1, the evidence supporting this dogma is far from convincing.  To begin with, it is based on low resolution Hi-C experiments that rely on large bin sizes to visualize so-called “TADs.”  In fact, the notion that cohesin/CTCF are responsible on their own for shaping the mammalian 3D genome appears to be a result of mistaking a series of forests for the actual trees that populate each of the forests.

As illustrated in Author response image 1 above, the “TADs” that are visualized in these low resolution data sets are not TADs at all, but rather TAD neighborhoods consisting of several dozen or more individual TADs.  Moreover, the “interesting” features that are evident at low resolution (>1 kb)—the dots and stripes—largely disappear at resolutions appropriate for visualizing individual TADs (~200 bp).

In Goel et al. 2023, we presented data from one of the key experiments in Goel et al. (Goel et al. 2023).  In this experiment,  the authors used RCMC to generate high resolution (~250 bp) MicroC contact maps before and after Rad21 depletion.  Contrary to dogma, Rad21 depletion has absolutely no effect on TADs in a ~250 kb DNA segment—and these TADs look very much like the TADs we observe in the Drosophila genome, in particular in the Abd-B region of BX-C that is thought to be assembled into a series of circle-loops (see Fig. 2B).

While Goel et al. (Goel et al. 2023) observed no effect of Rad21 depletion on TADs, they found that loss of Rad21 disturbs long-distance (but not short-distance) contacts in large TAD neighborhoods when their RCMC data set is visualized using bin sizes of 5 kb and I kb.  This is shown in Author response image 2.  The significance of this finding is, however, uncertain.  It could mean that the 3D organization of large TAD neighborhoods have a special requirement for cohesin activity.  On the other hand, since cohesin functions to hold sister chromosomes together after replication until they separate during mitosis (and might also participate in mitotic condensation), it is also possible that the loss of long-range contacts in large TAD neighborhoods when Rad21 is depleted is simply a reflection of this particular activity.  Further studies will be required to address these possibilities.

As for CTCF: a careful inspection of the ChIP data in Goel et al. 2023 indicates that CTCF is not found at each and every TAD boundary.  In fact, the notion that CTCF is the be-all and end-all of TAD boundaries in mammals is truly hard to fathom.  For one, the demands for specificity in TAD formation (and in regulatory interactions) are likely much greater than those in flies, and specificity can’t be generated by a single DNA binding protein.  For another, several dozen chromosomal architectural proteins have already been identified in flies.  This means that (unlike what is thought to be true in mammals) it is possible to use a combinatorial mechanism to generate specificity in, for example, the long distance interactions in RFig 6 and 7.  As noted in #2.1 above, many of the known chromosomal architectural proteins in flies are polydactyl zinc finger proteins (just like CTCF).  There are some 200 different polydactyl zinc finger proteins in flies, and the function of only a hand full of these is known at present.  However, it seems likely that a reasonable fraction of this class of DNA binding proteins will ultimately turn out to have an architectural function of some type (Bonchuk et al. 2021; Fedotova et al. 2017).  The number of different polydactyl zinc finger protein genes in mammals is nearly 3 times that of flies.  It is really possible that of these, only CTCF is involved in shaping the 3D structure of the mammalian genome?

Despite having a CTCF paralog and cohesin, the Drosophila genome does not appear to be structure by loop extrusion-pausing. The identification of orientation-dependent elements with pronounced structural effects on genome folding thus may shed light on alternative mechanisms used to regulated genome structure, which in turn may yield insights into the significance of particular folding patterns.

(2.2) Here we would like to draw the reviewer’s and reader’s attention to Author response image 3, which shows that orientation-dependent pairing interactions have a significant impact on physical interactions between different sequences.  We would also refer the reader to two other publications.  One of these is Kyrchanova et al. (Kyrchanova et al. 2008), which was the first to demonstrate that orientation of pairing interactions matters.  The second is Fujioka et al. (Fujioka et al. 2016), which describes experiments indicating that nhomie and homie pair with each other head-to-tail and with themselves head-to-head.

On the whole, this study is comprehensive and represents a useful contribution to the 3D genome field. The transgenic lines and Micro-C datasets generated in the course of the work will be valuable resources for the research community. Moreover, the manuscript, while dense in places, is generally clearly written and comprehensive in its description of the work. However, I have a number of comments and critiques of the manuscript, mainly centering on the framing of the experiments and presentation of the Micro-C results and on manner in which the data are analyzed and reported. They are as follows:

Major Points:

(1) The authors motivate much of the introduction and results with hypothetical "stem loop" and "circle loop" models of chromosome confirmation, which they argue are reflected in the Micro-C data and help to explain the observed ISH patterns. While such structures may possibly form, the support for these specific models vs. the many alternatives is not in any way justified. For instance, no consideration is given to important biophysical properties such as persistence length, packing/scaling, and conformational entropy. As the biophysical properties of chromatin are a very trafficked topic both in terms of experimentation and computational modeling and generally considered in the analysis of chromosome conformation data, the study would be strengthened by acknowledgement of this body of work and more direct integration of its findings.

(2.3) The reviewer is not correct in claiming that “stem-loops” and “circle-loops” are “hypothetical.”  There is ample evidence that both types of loops are present in eukaryotic genomes, and that loop conformation has significant readouts in terms of not only the physical properties of TADs but also their functional properties.  Here we would draw the reviewer’s attention to Author response image 3 and Author response image 4 for examples of loops formed by the orientation-dependent pairing of yet other TAD boundary elements.  As evident from the MicroC data in these figures, circle-loops and stem-loops have readily distinguishable contact patterns.  The experiments in Fujioka et al. (Fujioka et al. 2016) demonstrate that homie and nhomie pair with each other head-to-tail, while they pair with themselves head-to-head.  The accompany paper (Bing et al. 2024) also provides evidence that loop topology is reflected both in the pattern of activation of reporters and in the MicroC contact profiles.  We would also mention again Kyrchanova et al. (Kyrchanova et al. 2008), who were the first to report orientation-dependent pairing of endogenous fly boundaries.

At this juncture it would premature to try to incorporate computational modeling of chromosome conformation in our studies.  The reason is that the experimental foundations that would be essential for building accurate models are lacking.  As should be evident from RFigs. 1-3 above, studies on mammalian chromosomes are simply not of high enough resolution to draw firm conclusions about chromosome conformation: in most studies only the forests are visible.  While the situation is better in flies, there are still too many unknown.  As just one example, it would be important to know the orientation of the boundary pairing interactions that generate each TAD.  While it is possible to infer loop topology from how TADs interact with their neighbors (a plume versus clouds), a conclusive identification of stem- and circle-loops will require a method to unambiguously determine whether a TAD boundary pairs with its neighbor head-to-head or headto-tail.

(2) Similar to Point 1, while there is a fair amount of discussion of how the observed results are or are not consistent with loop extrusion, there is no discussion of the biophysical forces that are thought to underly compartmentalization such as block-polymer co-segregation and their potential influence. I found this absence surprising, as it is generally accepted that A/B compartmentalization essentially can explain the contact maps observed in Drosophila and other non-vertebrate eukaryotes (Rowley, ..., Corces 2017; PMID 28826674). The manuscript would be strengthened by consideration of this phenomenon.

(2.4) Compartments in mammals have typically been identified and characterized using lowresolution data sets, and these studies have relied on visualizing compartments using quite large bin sizes (>>1 kb).  Our experiments have nothing to do with the large-scale compartments seen in these Hi-C experiments.  Instead, we are studying the properties of individual TADs: how TADs are formed, the relationship between TAD topology and boundary:boundary pairing, and the impact of TAD topology on interactions between TADs in the immediate neighborhood.  There is no evidence to date that these large compartments or “block polymer co-segregation” have a) any impact on the properties of individual boundary elements, b) have a role in determining which boundary elements actually come together to form a given TAD, c) impact the orientation of the interactions between boundaries that generate the TAD or d) determine how TADs tend to interact with their immediate neighbors.  

In more recent publications (c.f., Harris et al. 2023) compartments have shrunk in size and instead of being units of several hundred kb, the median length of the “compartmental” unit in mammalian cells is about12 kb. This is not too much different from the size of fly TADs.  However, the available evidence does not support the idea that block polymer co-segregation/co-repulsion drive the TAD:TAD interactions seen in MicroC experiments.  For example, according to this “micro-compartment” model, the specific patterns of interaction between TADs in the CG3294 meta-loop in Author response image 3 would be driven by block polymer co-segregation and co-repulsion. In this model, the TAD upstream of the blue boundary (which contains CG33543, the odorant binding protein gene Obp22a and the Npc2a gene which encodes a protein involved in sterol homeostasis) would share the same chromatin state/biophysical properties as the TAD upstream of the purple boundary, which has the fipi gene. While it is true that CG33543, Obp22a and also the fipi gene are not expressed in embryos, Npc2a is expressed at high levels during embryogenesis, yet it is part of the TAD that interacts with the fipi TAD.  The TAD downstream of the blue boundary contains CG15353 and Nplp4 and it interacts with the TAD downstream of the purple boundary which contains CG3294 and slf.  CG15353 and Nplp4 are not expressed in the embryo and as such should share a compartment with a TAD that is also silent. However, slf is expressed at a high level in 1216 hr embryos, while CG3294 is expressed at a low level.  In neither case would one conclude that the TADs upstream and downstream of the blue and purple boundaries, respectively, interact because of shared chromatin/biophysical states that drive block polymer co-segregation corepulsion. 

One might also consider several gedanken experiments involving the long-range interactions that generate the CG3294 meta-loop in Author response image 3.    According to the micro-compartment model the patchwork pattern of crosslinking evident in the CG3294 meta-loop arises because the interacting  TADs share the same biochemical/biophysical properties, and this drives block polymer cosegregation and co-repulsion.  If this model is correct, then this patchwork pattern of TAD:TAD interactions would remain unchanged if we were to delete the blue or the purple boundary.  However, given what we know about how boundaries can find and pair with distant boundaries (c.f., Figure 6 from Muller et el. 1999 and the discussion in #1.2), the result of these gedanken experiments seem clear: the patchwork pattern shown in Author response image 3A will disappear.  What would happen if we inverted the blue or the purple boundary? Would the TAD containing CG33543, Obp22a and Npc2a still interact with fipi as would be expected from the compartment model?  Or would the pattern of interactions flip so that the CG33543, Obp22a and Npc2a TAD interacts with the TAD containing CG3294 and slf?  Again we can anticipate the results based on previous studies: the interacting TADs will switch when the CG3294 meta-loop is converted into a stem-loop.  If this happened, the only explanation possible in the compartment model is that the chromatin states change when the boundary is inverted so that TAD upstream of blue boundary now shares the same chromatin state as the TAD downstream of the purple boundary, while the TAD downstream of the blue boundary shares same state as the TAD upstream of the purple boundary.  However, there is no evidence that boundary orientation per se can induce a complete switch in “chromatin states” as would be required in the compartment model. 

While we have not done these experimental manipulations with the CG3294 meta-loop, an equivalent experiment was done in Bing et al. (Bing et al. 2024).  However, instead of deleting a boundary element, we inserted a homie boundary element together with two reporters (gfp and LacZ) 142 kb away from the eve TAD.  The result of this gedanken “reverse boundary deletion” experiment is shown in Author response image 5.  Panel A shows the MicroC contact profile in the region spanning the transgene insertion site and the eve TAD in wild type (read “deletion”) NC14 embryos.  Panel B shows the MicroC contact profile from 12-16 hr embryos carrying the homie dual reporter transgene inserted at -142 kb.  Prior to the “deletion”, the homie element in the transgene pairs with nhomie and homie in the eve TAD and this generates a “mini-metaloop.”  In this particular insert, the homie boundary in the transgene (red arrow) is “pointing” in the opposite orientation from the homie boundary in the eve TAD (red arrow).  In this orientation, the pairing of the transgene homie with eve nhomie/homie brings the LacZ reporter into contact with sequences in the eve TAD.  Since a mini-metaloop is formed by homie_à _nhomie/homie pairing, sequences in TADs upstream and downstream of the transgene insert interact with sequences in TADs close to the eve TAD (Author response image 5B).  Taken together these interactions correspond to the interaction patchwork that is typically seen in “compartments” (see boxed region and inset).  If this patchwork is driven as per the model, by block polymer co-segregation and co-repulsion, then it should still be present when the transgene is deleted.  However, panel A shows that the interactions linking the transgene and the sequences in TADs next to the transgene to eve and TADs next to eve disappear when the homie boundary (plus transgene) is “deleted” in wild type flies.

Author response image 5.

Boundary deletion and compartments

A second experiment would be to invert the homie boundary so that instead of pointing away from eve it points towards eve.  Again, if the compartmental patchwork is driven by block polymer co-segregation and co-repulsion, inverting the homie boundary in the transgene should have no effect on the compartmental contact profile.  Inspection of Fig. 7 in Bing et al. (Bing et al. 2024) will show that this prediction doesn’t hold either.  When homie is inverted, sequences in the eve TAD interact with the gfp reporter not the LacZ reporter.  In addition, there are corresponding changes in how sequences in TADs to either side of eve interact with sequences to either side of the transgene insert.  

Yet another “test” of compartments generated by block polymer co-segregation/co-repulsion is provided by the plume above the eve volcano triangle.  According to the compartment model, sequences in TADs flanking the eve locus form the plume above the eve volcano triangle because their chromatin shares properties that drive block polymer co-segregation.  These same properties result in repulsive interactions with chromatin in the eve TAD, and this would explain why the eve TAD doesn’t crosslink with its neighbors.  If the distinctive chromatin properties of eve and the neighboring TADs drive block polymer co-segregation and co-repulsion, then inverting the nhomie boundary or introducing homie in the forward orientation should have absolutely no effect on the physical interactions between chromatin in the eve TAD and chromatin in the neighboring TADs.  However, Figures 4 and 6 in this paper indicate that boundary pairing orientation, not block polymer co-segregation/co-repulsion, is responsible for forming the plume above the eve TAD. Other findings also appear to be inconsistent with the compartment model. (A) The plume topping the eve volcano triangle is present in NC14 embryos when eve is broadly expressed (and potentially active throughout the embryo).  It is also present in 12-16 hr embryos when eve is only expressed in a very small subset of cells and is subject to PcG silencing everywhere else in the embryo.  B) According to the compartment model the precise patchwork pattern of physical interactions should depend upon the transcriptional program/chromatin state that is characteristic of a particular developmental stage or cell type.  As cell fate decisions are just being made during NC14 one might expect that most nuclei will share similar chromatin states throughout much of the genome.  This would not be true for 12-16 hr embryos.  At this stage the compartmental patchwork would be generated by a complex mixture of interactions in cells that have quite different transcriptional programs and chromatin states.  In this case, the patchwork pattern would be expected to become fuzzy as a given chromosomal segment would be in compartment A in one group of cells and in compartment B in another.   Unlike 12-16 hr embryos,  larval wing discs would be much more homogeneous and likely give a distinct and relatively well resolved compartmental pattern. We’ve examined the compartment patchwork of the same chromosomal segments in NC14 embryos, 12-16 hr embryos and larval wing disc cells.  While there are some differences (e.g., changes in some of the BX-C TADs in the wing disc sample) the compartmental patchwork patterns are surprisingly similar in all three cases. Nor is there any “fuzziness” in the compartmental patterns evident in 12-16 hr embryos, despite the fact that there are many different cell types at this stage of development.  C) TAD interactions with their neighbors and compartmental patchworks are substantially suppressed in salivary gland polytene chromosomes.  This would suggest that features of chromosome structure might be the driving force behind many of the “compartmental” interactions as opposed to distinct biochemical/biophysical of properties of small chromosomal segments that drive polymer co- segregation/co-repulsion.  

(3) The contact maps presented in the study represent many cells and distinct cell types. It is clear from single-cell Hi-C and multiplexed FISH experiments that chromosome conformation is highly variable even within populations of the same cell, let alone between cell types, with structures such as TADs being entirely absent at the single cell level and only appearing upon pseudobulking. It is difficult to square these observations with the models of relatively static structures depicted here. The authors should provide commentary on this point.

(2.5) As should be evident from Author response image 1, single-cell Hi-C experiments would not provide useful information about the physical organization of individual TADs, TAD boundaries or how individual TADs interact with their immediate neighbors.  In addition, since they capture only a very small fraction of the possible contacts within and between TADs, we suspect that these single-cell studies aren’t likely to be useful for making solid conclusions about TAD neighborhoods like those shown in Author response image 1 panels A, B, C and D, or Author response image 2.  While it might be possible to discern relatively stable contacts between pairs of insulators in single cells with the right experimental protocol, the stabilities/dynamics of these interactions may be better judged by the length of time that physical interactions are seen to persist in live imaging studies such as Chen et al. (2018), Vazquez et al. (2006) and Li et al. (2011).

The in situ FISH data we’ve seen also seems problematic in that probe hybridization results in a significant decondensation of chromatin.  For two probe sets complementary to adjacent ~1.2 kb DNA sequences, the measured center-to-center distance that we’ve seen was ~110 nM.  This is about 1/3rd the length that is expected for a 1.2 kb naked DNA fragment, and about 1.7 times larger than that expected for a beads-on-a-string nucleosome array (~60 nM).  However, chromatin is thought to be compacted into a 30 nM fiber, which is estimated to reduce the length of DNA by at least another ~6 fold.  If this estimate is correct, FISH hybridization would appear to result in a ~10 fold decompaction of chromatin.  A decompaction of this magnitude would necessarily be followed by a significant distortion in the actual conformation of chromatin loops.

(4) The analysis of the Micro-C data appears to be largely qualitative. Key information about the number of reads sequenced, reaps mapped, and data quality are not presented. No quantitative framework for identifying features such as the "plumes" is described. The study and its findings would be strengthened by a more rigorous analysis of these rich datasets, including the use of systematic thresholds for calling patterns of organization in the data.

Additional information on the number of reads and data quality have been included in the methods section. 

(5) Related to Point 4, the lack of quantitative details about the Micro-C data make it difficult to evaluate if the changes observed are due to biological or technical factors. It is essential that the authors provide quantitative means of controlling for factors like sampling depth, normalization, and data quality between the samples.

In our view the changes in the MicroC contact patterns for the eve locus and its neighbors when the nhomie boundary is manipulated are not only clear cut and unambiguous but are also readily evident in the Figs that are presented in the manuscript.  If the reviewer believes that there aren’t significant differences between the MicroC contact patterns for the four different nhomie replacements, it seems certain that they would also remain unconvinced by a quantitative analysis.

The reviewer also suggests that biological and/or technical differences between the four samples could account for the observed changes in the MicroC patterns for the eve TAD and its neighbors.  If this were the case, then similar changes in MicroC patterns should be observed elsewhere in the genome.  Since much of the genome is analyzed in these MicroC experiments there is an abundance of internal controls for each experimental manipulation of the nhomie boundary.  For two of the nhomie replacements, nhomie reverse and homie forward, the plume above the eve volcano triangle is replaced by clouds surrounding the eve volcano triangle.  If these changes in the eve MicroC contact patterns are due to significant technical (or biological) factors, we should observe precisely the same sorts of changes in TADs elsewhere in the genome that are volcano triangles with plumes.   Author response image 6 shows the MicroC contact pattern for several genes in the Antennapedia complex.  The deformed gene is included in a TAD which, like eve, is a volcano triangle topped by a plume.  A comparison of the deformed MicroC contact patterns for nhomie forward (panel B) with the MicroC patterns for nhomie reverse (panel C) and homie forward (panel D) indicates that while there are clearly technical differences between the samples, these differences do not result in the conversion of the deformed plume into clouds as is observed for the eve TAD.  The MicroC patterns elsewhere in Antennapedia complex are also very similar in all four samples.  Likewise, comparisons of regions elsewhere in the fly genome indicate that the basic contact patterns are similar in all four samples.   So while there are technical differences which are reflected in the relative pixel density in the TAD triangles and the LDC domains, these differences do not result in converting plumes into clouds nor do the alter the basic patterns of TAD triangles and LDC domains.  As for biological differences— the embryos in each sample are at roughly the same developmental stage and were collected and processed using the same procedures. Thus, the biological factors that could reasonably be expected to impact the organization of specific TADs (e.g., cell type specific differences) are not going to impact the patterns we see in our experiments. 

Author response image 6.

(6) The ISH effects reported are modest, especially in the case of the HCR. The details provided for how the imaging data were acquired and analyzed are minimal, which makes evaluating them

challenging. It would strengthen the study to provide much more detail about the acquisition and analysis and to include depiction of intermediates in the analysis process, e.g. the showing segmentation of stripes.

The imaging analysis is presented in Fig. 5 is just standard confocal microscopy.  Individual embryos were visualized and scored.  An embryo in which stripes could be readily detected was scored as ‘positive’ while an embryo in which stripes couldn’t be detected was scored as ‘negative.’   

Recommendations for the authors:

Editor comments:

It was noted that the Jaynes lab previously published extensive genetic evidence to support the stem loop and circle loop models of Homie-Nhomie interactions (Fujioka 2016 Plos Genetics) that were more convincing than the Micro-C data presented here in proof of their prior model. Maybe the authors could more clearly summarize their prior genetic results to further try to convince the reader about the validity of their model.

Reviewer #1 (Recommendations For The Authors):

Below, I list specific comments to further improve the manuscript for publication. Most importantly, I recommend the authors tone down their proposal that boundary pairing is a universal TAD forming mechanism.

(1) The title is cryptic.

(2) The second sentence in the abstract is an overstatement: "In flies, TADs are formed by physical interactions between neighboring boundaries". Hi-C and Micro-C studies have not provided evidence that most TADs in Drosophila show focal interactions between their bracketing boundaries. The authors rely too strongly on prior studies that used artificial reporter transgenes to show that multimerized insulator protein binding sites or some endogenous fly boundaries can mediate boundary bypass, as evidence that endogenous boundaries pair.

Please see responses #1.1 and #1.3 and figures Author response image 1 and Author response image 3.  Note that using dHS-C, most TADs that we’ve looked at so far are topped by a “dot” at their apex.

(3) Line 64: the references do not cite the stated "studies dating back to the '90's'".

The papers cited for that sentence are reviews which discussed the earlier findings.  The relevant publications are cited at the appropriate places in the same paragraph.  

(4) Line 93: "On the other hand, while boundaries have partner preferences, they are also promiscuous in their ability to establish functional interactions with other boundaries." It was unclear what is meant here.

Boundaries that a) share binding sites for proteins that multimerized, b) have binding sites for proteins that interact with each other, or c) have binding sites for proteins that can be bridged by a third protein can potentially pair with each other.  However, while these mechanisms enable promiscuous pairing interactions, they will also generate partner preferences (through a greater number of a, b and/or c).

(5) It could be interesting to discuss the fact that it remains unclear whether Nhomie and Homie pair in cis or in trans, given that homologous chromosomes are paired in Drosophila.

The studies in Fujioka et al. (Fujioka et al. 2016) show that nhomie and homie can pair both in cis and in trans.  Given the results described in #1.2, we imagine that they are paired in both cis and trans in our experiments.

(6) Line 321: Could the authors further explain why they think that "the nhomie reverse circle-loop also differs from the nhomie deletion (λ DNA) in that there is not such an obvious preference for which eve enhancers activate expression"?

The likely explanation is that the topology/folding of the altered TADs impacts the probability of interactions between the various eve enhancers and the promoters of the flanking genes.  

(7) The manuscript would benefit from shortening the long Discussion by avoiding repeating points described previously in the Results.

(8) Line 495: "If, as seems likely, a significant fraction of the TADs genome-wide are circle loops, this would effectively exclude cohesin-based loop extrusion as a general mechanism for TAD formation in flies". The evidence provided in this manuscript appears insufficient to discard ample evidence from multiple laboratories that TADs form by compartmentalization or loop extrusion. Multiple laboratories have, for example, demonstrated that cohesin depletion disrupts a large fraction of mammalian TADs. 

Points made here and in #9 have been responded to in #1.1, #2.1 and #2.4 above.  We would suggest that the evidence for loop extrusion falls short of compelling (as it is based on the analysis of TAD neighborhoods, not TADs—that is forests, not trees) and given the results reported in Goel et al. (in particular Fig. 4 and Sup Fig. 8) is clearly suspect. This is not to mention the fact that cohesin loop-extrusion can’t generate circle-loops TADs, yet circle-loops clearly exist.  Likewise, as discussed in #2.4, it is not clear to us that the shared chromatin states, polymer co-segregation and co-repulsion account for the compartmental patchwork patterns of TAD;TAD interactions. The results from the  experimental manipulations in this paper and the accompanying paper, together with studies by others (e.g., Kyrchanova et al. (Kyrchanova et al. 2008), Mohana et al. (Mohana et al. 2023) would also seem to be at odds with the model for compartments as currently formulated.  

The unique properties of Nhomie and Homie, namely the remarkable specificity with which they physically pair over large distances (Fujioka et al. 2016) may rather suggest that boundary pairing is a phenomenon restricted to special loci. Moreover, it has not yet been demonstrated that Nhomie or Homie are also able to pair with the TAD boundaries on their left or right, respectively.

Points made here were discussed in detail in #1.2.  As described in detail in #1.2, It is not the case that nhomie and homie are in “unique” or “special.”  Other fly boundaries can do the same things.  As for whether nhomie and homie pair with their neighbors:  We haven’t done transgene experiments (e.g., testing by transvection or boundary bypass).  Likewise, in MicroC experiments there are no obvious dots at the apex of the neighboring TADs that would correspond to nhomie pairing with the neighboring boundary to the left and homie pairing with the neighboring boundary to the right. However, this is to be expected. As we discussed in in #1.3 above, only MNase resistant elements will generate dots in standard MicroC experiments.  On the other hand, when boundary:boundary interactions are analyzed by dHS-C (c.f., Author response image 4), there are dots at the apex of both neighboring TADs.  This would be direct evidence that nhomie pairs with the neighboring boundary to the left and homie pairs with the neighboring boundary to the right.

(9) The comment in point 8 also applies to the concluding 2 sentences (lines 519-524) of the Discussion.

See response to 8 above. Otherwise, the concluding sentences are completely accurate. Validation of the cohesin loop extrusion/CTCF roadblock model will required demonstrating a) that all TADs are either stem-loops or unanchored loops and b) that TAD endpoints are always marked by CTCF. 

The likely presence of circle-loops and evidence that TAD boundaries that don’t have CTCF (c.f.,Goel et al. 2023) already suggests that this model can’t (either fully or not all) account for TAD formation in mammals. 

(10) Figs. 3 and 6: It would be helpful to add the WT screenshot in the same figure, for direct comparison.

It is easy enough to scroll between Figs-especially since nhomie forward looks just like WT.

(11) Fig. 6: It would be helpful to show a cartoon view of a circle loop to the right of the Micro-C screenshot, as was done in Fig. 3.

Good idea.   Added to the Fig.

(12) Fig. 5: It would be helpful to standardize the labelling of the different genotypes throughout the figures and panels ("inverted" versus "reverse" versus an arrow indicating the direction).

Fixed.

Reviewer #2 (Recommendations For The Authors):

Minor Points:

(1) The Micro-C data does not appear to be deposited in an appropriate repository. It would be beneficial to the community to make these data available in this way.

This has been done.

(2) Readers not familiar with Drosophila development would benefit from a gentle introduction to the stages analyzed and some brief discussion on how the phenomenon of somatic homolog pairing might influence the study, if at all.

We included a rough description the stages that were analyzed for both the in situs and MicroC. We thought that an actual description of what is going on at each of the stages wasn’t necessary as the process of development is not a focus of this manuscript.  In other studies, we’ve found that there are only minor differences in MicroC patterns between the blastoderm stage and stage 12-16 embryos.  While these minor differences are clearly interesting, we didn’t discuss them in the text.   In all of experiments chromosomes are likely to be paired.  In NC14 embryos (the stage for visualizing eve stripes and the MicroC contact profiles in Fig. 2) replication of euchromatic sequences is thought to be quite rapid.  While homolog pairing is incomplete at this stage, sister chromosomes are paired.  In stage 12-16 embryos, homologs will be paired and if the cells are arrested in G2, then sister chromosome will also be paired.  So in all of experiments, chromosomes (sisters and/or homologs) are paired. However, since we don’t have examples of unpaired chromosomes, our experiments don’t provide any info on how chromosome pairing might impact MicroC/expression patterns.

(3) "P > 0.01" appears several times. I believe the authors mean to report "P < 0.01".

Fixed.  

References for Response

Batut PJ, Bing XY, Sisco Z, Raimundo J, Levo M, Levine MS. 2022. Genome organization controls transcriptional dynamics during development. Science. 375(6580):566-570.

Bing X, Ke W, Fujioka M, Kurbidaeva A, Levitt S, Levine M, Schedl P, Jaynes JB. 2024. Chromosome structure i: Loop extrusion or boundary:Boundary pairing? eLife.

Blanton J, Gaszner M, Schedl P. 2003. Protein:Protein interactions and the pairing of boundary elements in vivo. Genes Dev. 17(5):664-675.

Bonchuk A, Boyko K, Fedotova A, Nikolaeva A, Lushchekina S, Khrustaleva A, Popov V, Georgiev P. 2021. Structural basis of diversity and homodimerization specificity of zincfinger-associated domains in drosophila. Nucleic Acids Res. 49(4):2375-2389.

Bonchuk A, Kamalyan S, Mariasina S, Boyko K, Popov V, Maksimenko O, Georgiev P. 2020. Nterminal domain of the architectural protein ctcf has similar structural organization and ability to self-association in bilaterian organisms. Sci Rep. 10(1):2677.

Chen H, Levo M, Barinov L, Fujioka M, Jaynes JB, Gregor T. 2018. Dynamic interplay between enhancer–promoter topology and gene activity. Nat Genet. 50(9):1296.

Fedotova AA, Bonchuk AN, Mogila VA, Georgiev PG. 2017. C2h2 zinc finger proteins: The largest but poorly explored family of higher eukaryotic transcription factors. Acta Naturae. 9(2):47-58.

Fujioka M, Ke W, Schedl P, Jaynes JB. 2024. The homie insulator has sub-elements with different insulating and long-range pairing properties. bioRxiv. 2024.02.01.578481.

Fujioka M, Mistry H, Schedl P, Jaynes JB. 2016. Determinants of chromosome architecture: Insulator pairing in cis and in trans. PLoS Genet. 12(2):e1005889.

Galloni M, Gyurkovics H, Schedl P, Karch F. 1993. The bluetail transposon: Evidence for independent cis‐regulatory domains and domain boundaries in the bithorax complex. The EMBO Journal. 12(3):1087-1097.

Gaszner M, Vazquez J, Schedl P. 1999. The zw5 protein, a component of the scs chromatin domain boundary, is able to block enhancer-promoter interaction. Genes Dev. 13(16):2098-2107.

Goel VY, Huseyin MK, Hansen AS. 2023. Region capture micro-c reveals coalescence of enhancers and promoters into nested microcompartments. Nat Genet. 55(6):1048-1056.

Harris HL, Gu H, Olshansky M, Wang A, Farabella I, Eliaz Y, Kalluchi A, Krishna A, Jacobs M, Cauer G et al. 2023. Chromatin alternates between a and b compartments at kilobase scale for subgenic organization. Nat Commun. 14(1):3303.

Hart CM, Zhao K, Laemmli UK. 1997. The scs' boundary element: Characterization of boundary element-associated factors. Mol Cell Biol. 17(2):999-1009.

Hsieh TS, Cattoglio C, Slobodyanyuk E, Hansen AS, Rando OJ, Tjian R, Darzacq X. 2020. Resolving the 3d landscape of transcription-linked mammalian chromatin folding. Mol Cell. 78(3):539-553.e538.

Krietenstein N, Abraham S, Venev SV, Abdennur N, Gibcus J, Hsieh TS, Parsi KM, Yang L, Maehr R, Mirny LA et al. 2020. Ultrastructural details of mammalian chromosome architecture. Mol Cell. 78(3):554-565.e557.

Kyrchanova O, Chetverina D, Maksimenko O, Kullyev A, Georgiev P. 2008. Orientation-dependent interaction between drosophila insulators is a property of this class of regulatory elements. Nucleic Acids Res. 36(22):7019-7028.

Kyrchanova O, Ibragimov A, Postika N, Georgiev P, Schedl P. 2023. Boundary bypass activity in the abdominal-b region of the drosophila bithorax complex is position dependent and regulated. Open Biol. 13(8):230035.

Kyrchanova O, Kurbidaeva A, Sabirov M, Postika N, Wolle D, Aoki T, Maksimenko O, Mogila V, Schedl P, Georgiev P. 2018. The bithorax complex iab-7 polycomb response element has a novel role in the functioning of the fab-7 chromatin boundary. PLoS Genet. 14(8):e1007442. Kyrchanova O, Sabirov M, Mogila V, Kurbidaeva A, Postika N, Maksimenko O, Schedl P, Georgiev P. 2019a. Complete reconstitution of bypass and blocking functions in a minimal artificial fab-7 insulator from drosophila bithorax complex. Proceedings of the National Academy of Sciences.201907190.

Kyrchanova O, Wolle D, Sabirov M, Kurbidaeva A, Aoki T, Maksimenko O, Kyrchanova M, Georgiev P, Schedl P. 2019b. Distinct elements confer the blocking and bypass functions of the bithorax fab-8 boundary. Genetics.genetics. 302694.302019.

Kyrchanova O, Zolotarev N, Mogila V, Maksimenko O, Schedl P, Georgiev P. 2017. Architectural protein pita cooperates with dctcf in organization of functional boundaries in bithorax complex. Development. 144(14):2663-2672.

Li H-B, Muller M, Bahechar IA, Kyrchanova O, Ohno K, Georgiev P, Pirrotta V. 2011. Insulators, not polycomb response elements, are required for long-range interactions between polycomb targets in drosophila melanogaster. Mol Cell Biol. 31(4):616-625.

Li X, Tang X, Bing X, Catalano C, Li T, Dolsten G, Wu C, Levine M. 2023. Gaga-associated factor fosters loop formation in the drosophila genome. Mol Cell. 83(9):1519-1526.e1514.

Mohana G, Dorier J, Li X, Mouginot M, Smith RC, Malek H, Leleu M, Rodriguez D, Khadka J, Rosa P et al. 2023. Chromosome-level organization of the regulatory genome in the drosophila nervous system. Cell. 186(18):3826-3844.e3826.

Muller M, Hagstrom K, Gyurkovics H, Pirrotta V, Schedl P. 1999. The mcp element from the drosophila melanogaster bithorax complex mediates long-distance regulatory interactions. Genetics. 153(3):1333-1356.

Muravyova E, Golovnin A, Gracheva E, Parshikov A, Belenkaya T, Pirrotta V, Georgiev P. 2001. Loss of insulator activity by paired su(hw) chromatin insulators. Science. 291(5503):495498.

Postika N, Metzler M, Affolter M, Müller M, Schedl P, Georgiev P, Kyrchanova O. 2018. Boundaries mediate long-distance interactions between enhancers and promoters in the drosophila bithorax complex. PLoS Genet. 14(12):e1007702.

Samal B, Worcel A, Louis C, Schedl P. 1981. Chromatin structure of the histone genes of d. Melanogaster. Cell. 23(2):401-409.

Sigrist CJ, Pirrotta V. 1997. Chromatin insulator elements block the silencing of a target gene by the drosophila polycomb response element (pre) but allow trans interactions between pres on different chromosomes. Genetics. 147(1):209-221.

Udvardy A, Schedl P. 1984. Chromatin organization of the 87a7 heat shock locus of drosophila melanogaster. J Mol Biol. 172(4):385-403.

Vazquez J, Muller M, Pirrotta V, Sedat JW. 2006. The mcp element mediates stable long-range chromosome-chromosome interactions in drosophila. Molecular Biology of the Cell. 17(5):2158-2165.

Zolotarev N, Fedotova A, Kyrchanova O, Bonchuk A, Penin AA, Lando AS, Eliseeva IA, Kulakovskiy IV, Maksimenko O, Georgiev P. 2016. Architectural proteins pita, zw5,and zipic contain homodimerization domain and support specific long-range interactions in drosophila. Nucleic Acids Res. 44(15):7228-7241.

eLife assessment

This study presents important findings on the different polymorphs of alpha-synuclein filaments that form at various pH's during in vitro assembly reactions with purified recombinant protein. Of particular note is the discovery of two new polymorphs (1M and 5A) that form in PBS buffer at pH 7. The strength of the evidence presented is convincing. The work will be of interest to biochemists and biophysicists working on protein aggregation and amyloids.

Reviewer #1 (Public Review):

Summary:

Frey et al. report the structures of aSyn fibrils that were obtained under a variety of conditions. These include generation of aSyn fibrils without seeds, but in different buffers and at different pH values. These also include the generation of aSyn fibrils in the presence of seeding fibrils, again performed in different buffers and at different pH values, while the seeds were generated at different conditions. The authors find that fibril polymorphs primarily correlate with fibril growth buffer conditions, and not such much with the type of seed. However, the presence of a seed is still required, likely because fibrils can also seed along their lateral surfaces, not only at the blunt ends.

Strengths:

The manuscript includes an excellent review of the numerous available structures of aSyn.<br /> The text is interesting to read, figures are clear and not redundant.

Weaknesses:

My earlier comments have all been addressed to my satisfaction.

Reviewer #2 (Public Review):

The authors have engaged constructively with some of the points raised. In particular the addition of more details about the experimental cryo-EM procedures has strengthened the manuscript.

I do worry that the FSC values of model-vs-map appear to be higher than expected from the corresponding FSCs between the half-maps (e.g. see Fig 13). The implication of this observation is that the atomic models may have been overfitted in the maps, which would have led to a deterioration of their geometry. A table with rmsd on bond lengths, angles, etc would probably show this. In addition, to check for overfitting, the atomic model for each data set could be refined in one of the half-maps, and then that same model could be used to calculate 2 FSC model-vs-map curves: one against the half-map it was refined in and one against the other half-map. Deviations between these two curves are an indication of overfitting.

In addition, the sudden drop in the FSC curves in Figure 16 shows that something unexpected has happened to this refinement. Are the authors sure that only the procedures outlined in the Methods were used to create these curves? The unexpected nature of the FSC curve for this type (2A) raises doubts about the correctness of the reconstruction.

Reviewer #3 (Public Review):

Summary

The high heterogeneity nature of α-synuclein (α-syn) fibrils posed significant challenges in structural reconstruction of the ex vivo conformation. A deeper understanding of the factors influencing the formation of various α-syn polymorphs remains elusive. The manuscript by Frey et al. provides a comprehensive exploration of how pH variations (ranging from 5.8 to 7.4) affect the selection of α-syn polymorphs (specifically, Type1, 2 and 3) in vitro by using cryo-electron microscopy (cryo-EM) and helical reconstruction techniques. Crucially, the authors identify two novel polymorphs at pH 7.0 in PBS. These polymorphs bear resemblance to the structure of patient-derived juvenile-onset synucleinopathy (JOS) polymorph and diseased tissue amplified α-syn fibrils. The revised manuscript more strongly supports the notion that seeding is a non-polymorph-specific in the context of secondary nucleation-dominated aggregation, underscoring the irreplaceable role of pH in polymorph formation.

Strengths

This study systematically investigates the effects of environmental conditions and seeding on the structure of α-syn fibrils. It emphasizes the significant influence of environmental factors, especially pH, in determining the selection of α-syn polymorphs. The high-resolution structures obtained through cryo-EM enable a clear characterization of the composition and proportion of each polymorph in the sample. Collectively, this work provides a strong support for the pronounced sensitivity of α-syn fibril structures to the environmental conditions, and systematically categorizes previously reported α-syn fibril structures. Furthermore, the identification of JOS-like polymorph also demonstrates the possibility of in vitro reconstruction of brain-derived α-syn fibril structures.

Weaknesses

There are two minor points I recommend the authors to address:

(1) In the response to Weakness 1, point (3), the authors state that "the Type 5 represented only 10-20% of the fibrils in the sample." However, this information is not labeled in the corresponding Figure 4. I suggest the authors verify and label all relevant percentages in the figures to prevent misunderstandings.

(2) While the authors have detailed the helical reconstruction procedure in the Methods section, it is necessary to indicate the scale bar or box size in the figure legend of the 2D representative classes to ensure clarity and reproducibility.

Comments on the revised manuscript:

The authors have responded adequately to these critiques in the revised version of the manuscript.

Author response:

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

Reviewer #1:

We thank the reviewer for their careful reading of our manuscript and have taken all of their grammatical corrections into account.

Reviewer #2 (Public Review):

Weaknesses: 

The paper contains multiple instances of non-scientific language, as indicated below. It would also benefit from additional details on the cryo-EM structure determination in the Methods and inclusion of commonly accepted requirements for cryo-EM structures, like examples of 2D class averages, raw micrographs, and FSC curves (between half-maps as well as between rigid-body fitted (or refined) atomic models of the different polymorphs and their corresponding maps). In addition, cryo-EM maps for the control experiments F1 and F2 should be presented in Figure 9.

We tried to correct the non-scientific language and have included the suggested data on the Cryo-EM analyses including new Figures 11-17.  We did not collect data on the sample used for the seeds in the cross seeding experiments because we had already confirmed in multiple datasets that the conditions in F1 and F2 reproducibly produce fibrils of Type 1 and Type 3, respectively. We have now analyzed cryo-EM data for 6 more samples at pH 7.0 and found that several kinds of polymorphs (Types 1A, 1M, 2A, 2B and 5) are accessible at this pH, however the Type 3 polymorphs are not formed at pH 7.0 under the conditions that we used for aggregation.

Reviewer #2 (Recommendations For The Authors):

Remove unscientific language: "it seems that there are about as many unique atomicresolution structures of these aggregates as there are publications describing them"   

We have rephrased this sentence.

For same reason, remove "Obviously, " 

Done

What does this mean? “polymorph-unspecific” 

Rephrased as non-polymorph-specific

What does this mean? "shallow amyloid energy hypersurface"  

By “shallow hypersurface” we mean that the minimum of the multi-dimensional function that describes the energy of the amyloid is not so deep that subtle changes to the environment will not favor another fold/energy minimum. We have left the sentence because while it may not be perfect, it is concise and seems to get the point across.

"The results also confirm the possibility of producing disease-relevant structure in vitro." -> This is incorrect as no disease-relevant structure was replicated in this work. Use another word like “suggest”.

We have changed to “suggest” as suggested.

Remove "historically" 

Done

Rephrase “It has long been understood that all amyloids contain a common structural scaffold” 

Changed to “It has long been established that all amyloids contain a common structural scaffold..” 

"Amyloid polymorphs whose differences lie in both their tertiary structure (the arrangement of the beta-strands) and the quaternary structure (protofilamentprotofilament assembly) have been found to display distinct biological activities [8]" -> I don't think this is true, different biological activities of amyloids have never been linked to their distinct structures.  

We have added 5 new references (8-12) to support this sentence.

Reference 10 is a comment on reference 9; it should be removed. Instead, as for alphasynuclein, all papers describing the tau structures should be included.  

We have removed the reference, but feel that the addition of all Tau structure references is not merited in this manuscript since we are not comparing them.

Rephrase: "is not always 100% faithful"

Removed “100%”

What is pseudo-C2 symmetry? Do the authors mean pseudo 2_1 symmetry (ie a 2-start helical symmetry)?

Thank for pointing this out.  We did indeed mean pseudo 21 helical symmetry.  

Re-phrase: "alpha-Syn's chameleon-like behavior" 

We have removed this phrase.

"In the case of alpha-Syn, the secondary nucleation mechanism is based on the interaction of the positively charged N-terminal region of monomeric alpha-Syn and the disordered, negatively charged C-terminal region of the alpha-Syn amyloid fibrils [54]" -> I would say the mechanisms of secondary nucleation are not that well understood yet, so one may want to tune this down a bit. 

We have changed this to “mechanism has been proposed to be”

The paragraphs describing experiments by others are better suited for a Discussion rather than a Results section. Perhaps re-organize this part? 

We have left the text intact as we are using a Results and Discussion format.

A lot of information about Image processing seems to be missing: what steps were performed after initial model generation? 

We have added more details in the methods section on the EM data processing and model analysis.

Figure 1: Where is Type 4 on the pH scale?

We have adjusted the Fig 1 legend to clarify that pH scale is only applicable to the structures presented in this manuscript. 

Figure 2: This might be better incorporated as a subpanel of Figure 1.

We agree that this figure is somewhat of a loner on its own and we only added it in order to avoid confusion with the somewhat inconsistent naming scheme used for the Type 1B structure. However, we prefer to leave it as a separate figure so that it does not get dilute the impact of figure 1.

Figure 3: What is the extra density at the bottom of Type 3B from pH 5.8 samples 1 and 2. pH 5.8 + 50mM NaCl (but not pH 5.8 + 100 mM NaCl)? Could this be an indication of a local minimum and the pH 5.8 + 100 mM NaCl structure is correct? Or is this a real difference between 0/50mM NaCl and 100 mM NaCl? 

We did not see the extra density to which the reviewer is referring, however the images used in this panel are the based on the output of 3D-classification which is more likely to produce more artifacts than a 3D refinement. With this in mind, we did not see any significant differences in the refined structures and therefore only deposited the better quality map and model for each of the polymorph types.

Figure 3: To what extent is Type 3B of pH 6.5 still a mixture of different types? The density looks poor. In general, in the absence of more details about the cryo-EM maps, it is hard to assess the quality of the structures presented.

In order to improve the quality of the images in this panel, a more complete separation of the particles from each polymorph was achieved via the filament subset selection tool in RELION 5. In each case, an unbiased could be created from the 2D classes via the relion_helix_inimodel2D program, further supporting the coexistence of 4 polymorphs in the pH 6.5 sample. The particles were individually refined to produce the respective maps that are now used in this figure.

Many references are incorrect, containing "Preprint at (20xx)" statements.  

This has been corrected.

Reviewer #3 (Public Review):

Weaknesses: 

(1) The authors reveal that both Type 1 monofilament fibril polymorph (reminiscent of JOSlike polymorph) and Type 5 polymorph (akin to tissue-amplified-like polymorph) can both form under the same condition. Additionally, this condition also fosters the formation of flat ribbon-like fibril across different batches. Notably, at pH 5.8, variations in experimental groups yield disparate abundance ratios between polymorph 3B and 3C, indicating a degree of instability in fibrillar formation. The variability would potentially pose challenges for replicability in subsequent research. In light of these situations, I propose the following recommendations: 

(a) An explicit elucidation of the factors contributing to these divergent outcomes under similar experimental conditions is warranted. This should include an exploration of whether variations in purified protein batches are contributing factors to the observed heterogeneity.

We are in complete agreement that understanding the factors that lead to polymorph variability is of utmost importance (and was the impetus for the manuscript itself). However the number of variables to explore is overwhelming and we will continue to investigate this in our future research. Regarding the variability between batches of purified protein, we also think that this could be a factor in the polymorph variability observed for otherwise “identical” aggregation conditions, particularly at pH 7 where the largest variety of polymorphs have been observed. However, even variation between identical replicates (samples created from the same protein solution and simply aggregated simultaneously in separate tubes) can lead to different outcomes (see datasets 15 and 16 in the revised Table 1) suggesting that there are stochastic processes that can determine the outcome of an individual aggregation experiment. While our data still indicates that Type 1,2 and 3 polymorphs are strongly selected by pH, the selection between interface variants 3B vs. 3C and 2A vs. 2B might also be affected by protein purity. Our standard purification protocol produces a single band by coomassie-stained SDS-PAGE however minor truncations and other impurities below a few percent would go undetected and, given the proposed roles of the N and C-termini in secondary nucleation, could have a large effect on polymorph selection and seeding. In line with the reviewer’s comments we now include a batch number for each EM dataset. While no new conclusions can be drawn from the inclusion of this additional data, we feel that it is important to acknowledge the possible role of batch to batch variability. 

(b) To enhance the robustness of the conclusions, additional replicates of the experiments under the same condition should be conducted, ideally a minimum of three times.  

The pH 5.8 conditions that yield Type 3 fibrils has already been repeated several times in the original manuscript. Since the pH 7.4 conditions produce the most common a-Syn polymorph (Type 1A) and were produced twice in this manuscript (once as an unseeded and once as a cross-seeded fibrilization) we decided to focus on the intermediate condition where the most variability had been seen (pH 7.0). The revised table 1 now has 6 new datasets (11-16) representing 6 independent aggregations at pH 7.0 starting from two different protein purification batches. The results is that we now produce the type 2A/B polymorphs in three samples and in two of these samples we once again observed the type 1M polymorph.  The other samples produced Type 1A or non-twisted fibrils.

(c) Further investigation into whether different polymorphs formed under the same buffer condition could lead to distinct toxicological and pathology effects would be a valuable addition to the study.  

The correlation of toxicity with structure would in principle be interesting. However the Type 1 and Type 3 polymorphs formed at pH 5.8 and 7.4 are not likely to be biologically relevant. The pH 7 polymorphs (Type 5 and 1M) would be more interesting because they form under the same conditions and might be related to some disease relevant structures. Still, it is rare that a single polymorph appears at 7.0 (the Type 5 represented only 10-20% of the fibrils in the sample and the Type 1M also had unidentified double-filament fibrils in the sample). We plan to pursue this line of research and hope to include it in a future publication.

(2) The cross-seeding study presented in the manuscript demonstrates the pivotal role of pH conditions in dictating conformation. However, an intriguing aspect that emerges is the potential role of seed concentration in determining the resultant product structure. This raises a critical question: at what specific seed concentration does the determining factor for polymorph selection shift from pH condition to seed concentration? A methodological robust approach to address this should be conducted through a series of experiments across a range of seed concentrations. Such an approach could delineate a clear boundary at which seed concentration begins to predominantly dictate the conformation, as opposed to pH conditions. Incorporating this aspect into the study would not only clarify the interplay between seed concentration and pH conditions, but also add a fascinating dimension to the understanding of polymorph selection mechanisms.

A more complete analysis of the mechanisms of aggregation, including the effect of seed concentration and the resulting polymorph specificity of the process, are all very important for our understanding of the aggregation pathways of alphasynuclein and are currently the topic of ongoing investigations in our lab.

Furthermore, the study prompts additional queries regarding the behavior of cross-seeding production under the same pH conditions when employing seeds of distinct conformation. Evidence from various studies, such as those involving E46K and G51D cross-seeding, suggests that seed structure plays a crucial role in dictating polymorph selection. A key question is whether these products consistently mirror the structure of their respective seeds. 

We thank the reviewer for reminding us to cite these studies as a clear example of polymorph selection by cross-seeding. Unfortunately, it is not 100% clear from the G51D cross seeding manuscript (https://doi.org/10.1038/s41467-021-26433-2) what conditions were used in the cross-seeding since different conditions were used for the seedless wild-type and mutant aggregations… however it appears that the wildtype without seeds was Tris pH 7.5 (although at 37C the pH could have dropped to 7ish) and the cross-seeded wild-type was in Phosphate buffer at pH 7.0. In the E46K cross-seeding manuscript, it appears that pH 7.5 Tris was used for all fibrilizations (https://doi.org/10.1073/pnas.2012435118).  In any event, both results point to the fact that at pH 7.0-7.5 under low-seed conditions (0.5%) the Type 4 polymorph can propagate in a seed specific manner.

(3) In the Results section of "The buffer environment can dictate polymorph during seeded nucleation", the authors reference previous cell biological and biochemical assays to support the polymorph-specific seeding of MSA and PD patients under the same buffer conditions. This discussion is juxtaposed with recent research that compares the in vivo biological activities of hPFF, ampLB as well as LB, particularly in terms of seeding activity and pathology. Notably, this research suggests that ampLB, rather than hPFF, can accurately model the key aspects of Lewy Body Diseases (LBD) (refer to: https://doi.org/10.1038/s41467-023-42705-5). The critical issue here is the need to reconcile the phenomena observed in vitro with those in in-vivo or in-cell models. Given the low seed concentration reported in these studies, it is imperative for the authors to provide a more detailed explanation as to why the possible similar conformation could lead to divergent pathologies, including differences in cell-type preference and seeding capability.  

We thank the reviewer for bring this recent report to our attention. The findings that ampLB and hPFF have different PK digestion patterns and that only the former is able to model key aspects of Lewy Body disease are in support of the seed-specific nature of some types of alpha-synuclein aggregation.  We have added this to the discussion regarding the significant role that seed type and seed conditions likely play in polymorph selection.

(4) In the Method section of "Image processing", the authors describe the helical reconstruction procedure, without mentioning much detail about the 3D reconstruction and refinement process. For the benefit of reproducibility and to facilitate a deeper understanding among readers, the authors should enrich this part to include more comprehensive information, akin to the level of detail found in similar studies (refer to:

https://doi.org/10.1038/nature23002).

As also suggested by reviewer #2, we have now added more comprehensive information on the 3D reconstruction and refinement process.

(5) The abbreviation of amino acids should be unified. In the Results section "On the structural heterogeneity of Type 1 polymorphs", the amino acids are denoted using three-letter abbreviation. Conversely, in the same section under "On the structural heterogeneity of Type 2 and 3 structures", amino acids are abbreviated using the one-letter format. For clarity and consistency, it is essential that a standardized format for amino acid abbreviations be adopted throughout the manuscript.

That makes perfect sense and had been corrected.

Reviewing Editor:

After discussion among the reviewers, it was decided that point 2 in Reviewer #3's Public Review (about the experiments with different concentrations of seeds) would probably lie outside the scope of a reasonable revision for this work. 

We agree as stated above and will continue to work on this important point.

eLife assessment

This study presents a valuable strategy to co-deliver peptides and adjuvants to antigen-presenting cells by engineering the Virus-like particle (VLP). The evidence supporting the claims of the authors is convincing, but the antitumour efficacy is unimpressive and would benefit from more antitumor experiments. The work will be of broad interest to bioengineers and medical biologists focusing on cancer vaccines.

Reviewer #1 (Public Review):

Tang et al present an important manuscript focused on endogenous virus-like particles (eVLP) for cancer vaccination with solid in vivo studies. The author designed eVLP with high protein loading and transfection efficiency by PEG10 self-assembling while packaging neoantigens inside for cancer immunotherapy. The eVLP was further modified with CpG-ODN for enhanced dendritic cell targeting. The final vaccine ePAC was proven to elicit strong immune stimulation with increased killing effect against tumor cells in 2 mouse models. Below are my specific comments:

(1) The figures were well prepared with minor flaws, such as missed scale bars in Figures 4B, 4K, 5B, and 5C. The author should also add labels representing statistical analysis for Figures 3C, 3D, and 3E. In Figure 6G, the authors should label which cell type is the data for.

(2) In Figure 3H, the antigen-presenting cells (APCs) increased significantly, but there was also a non-negligible 10% of APCs found in the control group, indicating some potential unwanted immune response; the authors need to explain this phenomenon or add a cytotoxic test on the normal liver or other cell lines for confirmation.

(3) In Figure 3I, the ePAC seems to have a very similar effect on cytotoxic T-cell tumor killing compared to the peptides + CpG group. If the concentrations were also the same, based on that, questions will arise as to what is the benefit of using the compact vector other than just free peptide and CpG? Please explain and elaborate.

(4) In the animal experiment in Figures 4F to L, the activation effect of APCs was similar between ePAC and CpG-only groups with no significance, but when it comes to the HCC mouse model in Figure 5, the anti-tumor effect was significantly increased between ePAC and CpG-only group. The authors should explain the difference between these two results.