Add files using upload-large-folder tool

peer_reviews/supplementary_0_Peer Review File__d5b6c7d85843913b9e894d85a549e8a375934c242b73b1338b64274e10565d08/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d5b6c7d85843913b9e894d85a549e8a375934c242b73b1338b64274e10565d08/supplementary_0_Peer Review File__d5b6c7d85843913b9e894d85a549e8a375934c242b73b1338b64274e10565d08.mmd

@@ -0,0 +1,595 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+# Estrogen receptor beta in astrocytes modulates cognitive function in mid-age female mice  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' comments:  
+
+Reviewer #1 (Remarks to the Author):  
+
+A.Key findings:  
+
+Itoh et col. seek to refine hormone replacement therapy (HRT) by gaining insight into brains areas affected by hormone removal, as well as the distinct roles of estrogen receptors (alpha vs beta) and of neurons vs astrocytes. To this end, behavioral and structural analyses of brain areas with magnetic resonance imaging (MRI) are performed in three studies. One is a longitudinal analysis of male and female mice throughout life. Another study explores the effects of gonadectomy at midlife. And a third study examines mice with selective removal of estrogen receptors in astrocytes or neurons. They report age- , sex- and brain region- dependent atrophy that is exacerbated by gonadectomy and mimicked by targeted deletion of estrogen receptor beta in astrocytes. Taken together, the mouse data suggest that the neuroprotective actions of estrogens are cell- and receptor- dependent, primarily affecting dorsal hippocampus. It is also remarkable that gonadectomy results in severe decrease of spines in dorsal hippocampus. The study may inform structural analyses in HRT, as well as therapeutic strategies, with a focus on the emerging avenue of astrocyte- targeted therapies. The authors are to be commended by the hard work and the wealth of tests and animal models. Tests include hippocampal- dependent spatial memory (Morris Water Maze, MWM), working memory (Y- maze) and contextual fear conditioning. Of great value is the use of MRI to dissect out regional changes in association with behavioral deficits. Mouse models include intact and gonadectomized, as well as knock out mice for estrogen receptor types in astrocytes and neurons.  
+
+### B. Limitations:  
+
+## B.1. Insufficient clarity.  
+
+Conceptual: The gap of knowledge ('understanding brain region- specific, cell- specific and receptor- specific mechanisms') that the authors aim to dissipate should be rephrased. As I understand it, the authors aim to clarify why HRT does not mitigate cognitive deficits and brain atrophy in menopausal women (problem statement) and propose that the conflicting results may arise from paradoxical actions of different estrogen receptors in different cell types, and the fact that readout measures are not directed to the right brain areas. There is a confusion in the said gap between actors (cells and receptors) and effects (brain areas and cognition), for, in the study, brain regions are not specifically targeted. The gene knockout affects cells all over the brain, and clinical therapies are systemically administered. The repeated expression 'region- specific targeting' throughout the text should be eliminated or clarified.  
+
+Experimental: Perhaps it is just me, but the study rationale is not clear, particularly which tests are used, and which ages analyzed. Specifically, in figure 1, in intact mice only MWM is performed at three ages, with negative results, while in Fig 2 (gonadectomized mice) Y Maze and contextual fear conditioning are incorporated, but only in two ages. The comparison of young and midlife mice allows to test the interaction between age and absence of hormones, but the absence of old mice makes it difficult to conclude whether gonadectomy at midlife accelerates cognitive and structural decline to match that of old ages. Also, Fig. 2a and Fig. 2c repeat data. Perhaps Fig.1 and Fig. 2 should be condensed into a single figure to present the model of intact (normal scenario) and gonadectomized mice (extreme scenario) and conclude which behavioral and structural analyses are more apt to study the effect of hormones. Likewise, in Fig. 4 (studies in ER knockout mice) Y Maze is not used, only midlife mice are tested, and immunohistochemical analyses of GFAP, Iba1 and PSD95 are not performed.  
+
+Fig. 3 e, f. I guess that it is normal that ' \(\%\) time in other quadrants' is the mirror/inverse version of ' \(\%\) time in target quadrant'. If so, why is the scale in the Y axis of Fig. 3f different from the one in 3e? For example, a \(20\%\) time in target quadrant should correlate with \(80\%\) time in other quadrants.  
+
+B.2. Overinterpretation of correlations.
+
+<--- Page Split --->
+
+Direct results should be differentiated from interpretations. For example, line 212 'Neuroprotection in females at midlife is mediated by ERbeta in astrocytes' is not a correct label. The experimental design will show the effect of ERbeta removal, but it does not demonstrate that the deleterious mechanisms are the same as those mediating damage caused by gonadectomy, which affects many hormones. Such an equivalence is a correlation. Reversal experiments such as addition of ERbeta agonists to gonadectomized mice may beautifully strengthen the connection the authors wish to make.  
+
+### C. Methodologies:  
+
+Supplemental Fig. 1. presents data from estrogen receptor alpha knockout mice that are not described in Methods. These data are important and could be moved to the main text. Likewise, the PCR results confirming successful deletion of estrogen receptor genes should be shown.  
+
+It would be nice to have images of MRI in different experimental scenarios in Fig. 1.  
+
+The quantification of immunohistochemical data is not explained in sufficient detail, including the number and distance between sections, stereological considerations, programs used to quantify, thresholding, background subtraction, etc. Also, images of PSD95 immunostaining, perhaps the extreme cases, would be useful to visually support one of the most interesting finding of the study.  
+
+### D. Style issues:  
+
+The authors may consider removing 'Mind the Gap' from the title, since journalistic expressions should be avoided from scientific reports, and it is not clear which gap should be minded until one reads the introduction. Also, the title should describe the model and main result of the study and not the implications thereof, as noted. Something like 'Selective deletion of ERbeta in astrocytes but not neurons mimics behavioral and structural deficits caused by gonadectomy in female mice at midlife' may work.  
+
+The manuscript is well written, but there is always room for improvement. In the Abstract, the sentence in line 26 starting with 'Some aspects of aging...' is dispensable. I already suggested rephrasing of the problem statement above, in noting that the results should be described in a neutral manner, distinguishing findings from interpretations. For instance, in line 30, an alternative to the current sentence is: Gonadectomy impairs behavioral performance in \*\*\* test (specify) and results in atrophy in \*\*\* but not in \*\*\* (specify) in \*\*\* mice (specify), suggesting that ovarian hormones in \*\*\* mice (specify) protect against hippocampal- dependent cognitive impairment and dorsal hippocampal atrophy. Likewise, in line 33: Deletion of ERbeta in astrocytes but not neurons by \*\*\* technology (specify) in \*\*\* mice (specify) result in \*\*\* (specify), as measured by \*\*\* (specify), suggesting that the protective effects of estrogen in midlife female mice are mediated by ERbeta receptors in astrocytes.  
+
+Line 52 'during health' is unnecessary. The study does not need to touch on the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy. Also, the Marxian sentence (from Groucho) 'A better understanding of the effect of aging can provide insights into the effect of aging' should be rephrased for clarity.  
+
+Line 59. It appears from the sentence that the impact of timing and estrogen type (does it mean receptor type?) has been already clarified according to the literature (refs 6- 9), in contradiction with the problem statement, and with line 88 (refs 8, 28). Please, succinctly specify what is known and what are the areas of conflict.  
+
+Line 62. This sentence belongs to a review, there is no need to defend the need for separation of sexes in a study using, moreover, a reference from 2012. Likewise, the rest of the paragraphs can be summarized, or moved to the Discussion. For example, it should be highlighted in the Discussion that the mice mirror the sex- dependent evolution of cognition and atrophy in humans.  
+
+Line 91. Again, the study does not inform about brain region- specific therapeutic approaches.
+
+<--- Page Split --->
+
+What it does is to unravel which brain regions are more affected by the experimental manipulations; nor the authors address the regional heterogeneity of astrocytes or neurons, for their genetic manipulations are global.  
+
+Lines 212 and 254. These two paragraphs can be condensed into one. Also, the explanation of the contextual fear conditioning (line 256) should be moved to earlier in the text, when the meaning of the different tests is explained.  
+
+In Discussion, the clinical implications of the findings can be summarized. Text condensation is recommended.  
+
+As most articles, including those written by English- speaking authors, the manuscript may benefit from professional copy editing.  
+
+Reviewer #2 (Remarks to the Author):  
+
+In this study, Itoh, Meyer, et al. studied the effect of sexual hormone loss in age- related decline of memory, and the involvement of Estrogen Receptor \(\beta\) (ER \(\beta\) ) in mediating the aggravated hippocampal neuropathy observed in female mice following gonadectomy. The work combines memory assessments, MRI imaging of structures important for memory formation, and immunohistochemistry. The paper is clear, and the methodology thorough, with appropriate conclusions when describing the result. However, I believe the claims to be overstated regarding the potential of astrocytic ER \(\beta\) as neuroprotector, based solely on the presented data, without taking into account the existing literature. The work nicely shows that the deletion of ER \(\beta\) is detrimental for cognition from midlife, but in my opinion, the authors have not thoroughly demonstrated the beneficial effect of ER \(\beta\) activation on cognitive aging. As I elaborate below, I would also advise the authors to highlight the novelty of their finding, and finally to be cautious regarding the interpretation of their research, done in the context of normal aging but not neurodegeneration, as stated several times throughout the text.  
+
+Title: While I understand the underlying idea, i.e. that ER \(\beta\) in astrocytes could be an important candidate for future therapeutic targets in cognitive aging in female. I tend to find it overclaiming. The authors do not demonstrate that ER \(\beta\) solely mediates the potential improvement of memorisation in aging, for example using a specific drug targeting those receptors, or losing a positive effect after the deletion.  
+
+1. Behaviour: I was very surprised to read that the authors did not find an age-related defect of spatial memory in the Morris Water Maze (MWM) experiment (Fig1). It is a well-established fact that hippocampus-dependent types of memory, such as spatial, reference, or episodic memory, are sensitive to aging across species (please see the body of work from Catherine Barnes, Howard Eichenbaum, Michela Gallagher, Michael Yassa, Aline Marighetto, Sara Burke, to name a few). In trying to understand this lack of defects, I came to think that the paper would benefit not only from showing the performance during acquisition, at least in the supplementary material, as a control of learning, but also from providing more details about the paradigm. Indeed, depending on the protocol, the memory can go from spatial to procedural (i.e., hippocampus-independent), which remains relatively intact during aging. Alternatively, the authors could also represent the target quadrant exploration during the test as blocks of 15 sec to clarify the mouse behaviour. That being said, I think the high performance from aged mice is due to the fact that these authors tested the mice 2h after the last training session. This could explain the difference between retrieval in the MWM and the fear conditioning, which was tested 24h post-conditioning, and did reveal the age-related memory decline. I would suggest to the authors to homogenise the conditions of behavioural testing throughout the paper, for clarity of understanding and comparison, or at least to address these discrepancies in the discussion. As the age-related decline of hippocampus-dependent memory is such a long-standing fact, I don't think the study benefits from starting with such a clashing result and could deter the readership. Especially since in GDX mice, the defect appears earlier, at "mid-life", and could be interpreted as early aging. Finally, regarding the behaviour part, I would like to draw the authors' attention to the fact that
+
+<--- Page Split --->
+
+the dichotomy in the role of the hippocampus to sustain recent but not remote memory has been revised in recent years (Goshen et al., Cell 2011) benefiting from the advancement of optogenetics. While I agree with the authors' conclusion regarding their own result, I would advice to modify the introduction of this part to take into account recent literature rather than Frankland and Bontempi's early work. A minor remark on this part, I am sure what is the authors' rationale for making this data as a supplementary figure only, when this is an interesting result.  
+
+2. ERβ cKO. I think it would be helpful to provide more details regarding the origin of the mice and selectivity of the conditional KO. Can you provide a proof of the specificity of the deletion within the astrocyte or the neurons? Could there be a compensation of the Era, that would underlie the result and hinder the interpretation regarding the role of ERβ? Providing an image illustrating the specific deletion in both cell type is, in my opinion, a major control to ensure the validity of study.  
+
+3. Neurodegeneration: One major point that I would like to draw the authors' attention towards, is their multiple reference to neurodegeneration throughout the paper, and the lack of distinction between normal aging and pathological aging. From what I understand, the authors seem to suggest that their study addresses neurodegeneration. However, the study has been conducting in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects), as demonstrated in my previous point, these wildtype aged mice can demonstrate high performance in certain conditions (Fig1), which would not be the case in pathological aging. In normal aging, the memory defects are associated with atrophy, but not neuronal loss, unlike pathological aging Moreover, the underlying mechanisms, and progression of alterations are widely distinct according to the type of aging. Therefore, I do not think it is proper in their condition to refer to neurodegeneration and would reframe the paper as normal aging. They could address the potential importance of their work regarding pathological aging in the discussion, but the use of wildtype mice do not support the direct extrapolation towards neurodegeneration.  
+
+4. Discussion: I think the work would benefit from stressing out throughout the text and in particular the discussion, the novelty of the research, and replacing the work within the literature. It is my belief that aside from the very interesting distinctive role of ERβ in astrocytes vs. neurons, the rest of the work has in some form been already reported. Age-related memory deficits (see authors in part 1 for reference); There is also a large body of work on the sexual dimorphism in cognitive aging (e.g. Frick et al., 2000; for review, Frick et al 2008), and on the role of estrogens in the brain and learning and memory, and effect of ovariectomy, which deserve to be cited in my opinion (L. Galea, K. Frick, E. Waters, M. Adams, C. Wolley, B. McEwen - non-exhaustive list). Similarly, the atrophy of the hippocampus has been demonstrated multiple times (for review Barnes and Burke, 2010). Finally, the reduction in spine density following ovariectomy has been previously reported by the McEwen laboratory (Gould et al., 1990). I do not intend to denigrate the study; I simply believe that the authors would benefit from valorising their findings in light of the existing literature.  
+
+Final point on the discussion, I fail to understand why the authors address hormone replacement and MS in such length, when their study does not use estrogen supplementation, potential effect on alteration similar to MS, and that MS have limited cognitive effect. Rather, I would be very interested in reading the authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes.  
+
+Minor comment: I think the reader would benefit from the homogenisation of the statistical report, which at the moment is alternatively in the legend or the text.  
+
+Looking forward to reading your response,  
+
+AS AI Abed  
+
+Reviewer #3 (Remarks to the Author):  
+
+This article provides original and interesting data regarding the neuroprotective role of the astroglial estrogen receptor (ER) beta against hippocampal- dependent cognitive deficits and
+
+<--- Page Split --->
+
+atrophy in menopausal women. The authors combine molecular approach (constitutive conditional knockout mice) with in vivo MRI and behavioral testing to assess the role of astroglial ER beta in protecting against brain atrophy (in vivo MRI) and cognitive deficits, assessed via spatial reference memory (Morris Water Maze), working memory (Y- maze) and contextual learning (contextual fear conditioning). The data are primarily correlative and descriptive. By using a conditional KO mice, the authors conclude that ER beta in astrocytes could be a novel promising therapeutic target for cognitive deficits and hippocampal atrophy in menopausal women. This is of interest. Yet the conclusions are not well supported by the experimental data mostly because of the molecular approach, which is not appropriate. In addition, the manuscript lacks mechanistic insights. Thus although interesting, this work is preliminary in its present form and presents several major issues.  
+
+1) The authors use GFAP-cre mice to constitutively delete ER beta in astrocytes. These mice cannot be used to conclude about implication of astroglial proteins, when these proteins are also expressed in neurons.  
+
+There are indeed major caveats in using the GFAP-Cre mice, as it has been well documented by numerous groups that deletion occurs in both neurons and astrocytes, since GFAP is expressed in precursor cells during early development. This confounds the authors major conclusion that deletion of ER beta from astrocytes contributes to cognitive deficits and brain atrophy in menopausal female.  
+
+A few papers here listed well describe these caveats : "Looks Can Be Deceiving: Reconsidering the Evidence for Gliotransmission" https://doi.org/10.1016/j.neuron.2014.12.003  
+
+Germ- Line Recombination Activity of the Widely Used hGFAP- Cre and Nestin- Cre Transgenes https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3857304/  
+
+Expression Specificity of GFAP Transgenes https://link.springer.com/article/10.1007/s11064- 004- 6881- 1  
+
+Other experiments should thus be combined with the use of these mice, such as the use of inducible cKO mice and/or rescue experiments targeted to astrocytes (re- expression of ER beta specifically in astrocytes in the astroglial cKO mice).  
+
+2) There is no validation of these newly generated astroglial- and neuronal-specific ER beta transgenic models. One would like to see quantification of ER beta expression in astrocytes and neurons in both cKO mice (gfap-cre:ER betafl/fl and nestin-cre:ER betafl/fl).  
+
+3) There is a lack of details regarding the mice (origin (stock number or laboratory source), breeding strategy controlling germline recombination, appropriate littermate controls. This precludes evaluation of whether the authors have used appropriate controls.  
+
+4) The authors use GFAP immunostaining to assess astroglial reactivity. GFAP expression can vary independently of astroglial reactivity. The authors should thus use additional markers (such as vimentin, stat3, GS...) and analyse astrocyte morphology to conclude about astrocyte reactivity.  
+
+5) The authors report synapse loss in menopausal female. They however only performed PSD95 staining to assess synapses. A synapse is composed of a pre- and a postsynaptic element and can only be identified by the colocalization of both elements. Staining for both pre- and postsynaptic markers should thus be performed to quantify synapse number.  
+
+6) This manuscript completely lacks mechanistic insights. How ER beta in astrocytes contributes to brain structure and cognitive performance in male and female? Does this require a crosstalk with other cell types such as microglia?  
+
+Reviewer #4 (Remarks to the Author):  
+
+The paper "Mind the Gap: Estrogen receptor beta (ERβ) in astrocytes is a therapeutic target to
+
+<--- Page Split --->
+
+prevent cognitive problems at menopause" investigates the impact of estrogen, and more specifically estrogen receptors beta, to the observed gender differences in behavior, hippocampus volume and astrogliosis in rodents at midlife. The authors use a combination of techniques like MRI, histology, behavior on wild type animals, animals which underwent gonadectomy/sham and ERb knock- out animals.  
+
+The narrative is very interesting, the results important and well contextualized. The work is extremely significant and timely. The methodology is rigorous, although a few important pieces of evidence are missing to have a more balanced picture and close the story (detailed below). I also have doubts on some of the statistical approaches chosen. The paper is easy to follow and very well written.  
+
+There are a few aspects that can be improved. Below, my suggestions.  
+
+1 Fig.1 : the information contained in a- b and c- d is redundant. In addition, doing twice a scatter plot, which distributes points random, on the same data generates plots which are visually different, which is awkward. The same applies to Fig. 2 a- b and c- d, and Fig. 4 a- b. I think half of the plots can safely go to supplementary.  
+
+2 Following the previous comment, I think the most correct statistical approach would be a single ANOVA for the water maze data where sex is a factor (in addition to age and quadrant), followed by appropriate post- hoc comparisons. The same applies for MRI volumetric data, where ROI should be another factor (cortex, striatum, HC).  
+
+3 In Fig 1 h, it seems that there is no volume change in HC in male mice across the lifespan. This results somehow weakens the point made by the authors in the rest of the paper, about the importance for neuroprotection across the lifespan of the ERb astrocyte receptors specifically in the dorsal hippocampus, unless the dHC is only important in ageing female, which is hard to digest, unless supported by evidence. Generally speaking, the paper is missing showing and discussing in males at least some of the analyses performed in females, for example Fig 2 g- i.  
+
+4 Fig 1, supplementary excel table: why using Welch test? Is it because the samples have unequal variance? Please specify in the methods.  
+
+5 Fig 1: It is noteworthy that from Fig. 1 there is no relation between structure and function, at least in the context of ageing. If it is a matter of different sensitivities of the two techniques employed (MRI volumetry vs behaviour), as hinted by the authors, I suggest they discuss the issue and report the power analysis used for determining the sample size. Perhaps the behavior is underpowered?  
+
+6 Fig.2: Similarly to comments 1 and 2, the figures of the water maze are redundant, and the statistical model could be just one for each test but include everything (gender, therapy, life stage).  
+
+7 Fig. 2 f: why freezing was only measured at midlife?  
+
+8 Fig. 2 j- k: I suggest to only plot regression line when significant. Also, there seems to be a lot of points in the plots. Are all animals (both sexes, all ages, GDX/sham) pulled together? If so, which is the rationale? Perhaps the different groups could be shown in the plot by coloring the datapoints differently.
+
+<--- Page Split --->
+
+9  
+
+Fig. 3, Neuropathology: the authors talk about "area fraction" but what is exactly measured and how? Can distinct aspect of the morphology be measured (ramifications density, cell body size, cell number etc.? This is important as the term "glia activation" can refer to distinct morphological changes happening to the cell. In general, the method- >histology section is missing details needed for the reader to be able to reproduce the analyses.  
+
+10  
+
+Line 194: the notation is misleading. Perhaps by "direct" and "indirect" correlations the authors mean linear positive and linear negative? Or do the authors really mean indirect as mediated by something else? If so, specify.  
+
+11  
+
+Fig 3 is missing the plots of the correlations between time in TQ and gliosis (data is discussed but not shown).  
+
+12  
+
+Following the rationale of the paper, Fig 4 is missing the neuropathology. How are the astrocytes affected by ERb deletion? Are they more activated at midlife?  
+
+13  
+
+Following the previous comment but more generally, what is ultimately the mechanism that the authors identified as leading to atrophy and worst cognition in ageing females? Is it astrogliosis? If so, the authors should report more neuropathological data (or other equivalent astrogliosis markers) to make a stronger point.  
+
+14  
+
+It is interesting that there is no cognitive effect in young female mice which lacks endogenous hormones or receptors. What about HC volume?  
+
+15  
+
+As the authors point out in the discussion, microglia also have ERb. Can the author show and compare some histology of microglia in animals lacking endogenous receptors and animals lacking ERb in astrocytes?
+
+<--- Page Split --->
+
+## Reviewers' comments:  
+
+Reviewer #1 (Remarks to the Author):  
+
+### A. Key findings:  
+
+The study may inform structural analyses in HRT, as well as therapeutic strategies, with a focus on the emerging avenue of astrocyte- targeted therapies. The authors are to be commended by the hard work and the wealth of tests and animal models. Tests include hippocampal- dependent spatial memory (Morris Water Maze, MWM), working memory (Y- maze) and contextual fear conditioning. Of great value is the use of MRI to dissect out regional changes in association with behavioral deficits. Mouse models include intact and gonadectomized, as well as knock- out mice for estrogen receptor types in astrocytes and neurons.  
+
+We appreciate these comments recognizing the breadth of our experiments.  
+
+### B. Limitations:  
+
+## B.1. Insufficient clarity.  
+
+Conceptual: The gap of knowledge ('understanding brain region- specific, cell- specific and receptor- specific mechanisms') that the authors aim to dissipate should be rephrased. As I understand it, the authors aim to clarify why HRT does not mitigate cognitive deficits and brain atrophy in menopausal women (problem statement) and propose that the conflicting results may arise from paradoxical actions of different estrogen receptors in different cell types, and the fact that readout measures are not directed to the right brain areas. There is a confusion in the said gap between actors (cells and receptors) and effects (brain areas and cognition), for, in the study, brain regions are not specifically targeted. The gene knockout affects cells all over the brain, and clinical therapies are systemically administered. The repeated expression 'region- specific targeting' throughout the text should be eliminated or clarified.  
+
+As suggested by the reviewer, we have now removed "gap" from the title as well as throughout the revised manuscript. Also, our phrasing, "Region- specific targeting", warrants clarification. Here, we show functional, pathology, and MRI effects that support future treatments targeting a specific cell and brain region (astrocytes in dorsal hippocampus). Future therapeutics could target this cell and region in midlife females. This does not limit treatments to those affecting only astrocytes in dorsal hippocampus (no other cells and no other regions). That is a pharmacologic bar too high for existing technologies. However, future clinical trials could tailor estrogen type, inclusion criteria, and outcome measures (specific cognitive domains, brain substructure atrophy, PET imaging, pathology of post- mortem or biopsy tissues, etc.) to assess whether a systemic treatment has an effect on astrocytes in dorsal hippocampus in midlife females. This manuscript identifies astrocytes in dorsal hippocampus as such a candidate cell and region.  
+
+Experimental: Perhaps it is just me, but the study rationale is not clear, particularly which tests are used, and which ages analyzed. Specifically, in figure 1, in intact mice only MWM is performed at three ages, with negative results, while in Fig 2 (gonadectomized mice) Y Maze and contextual fear conditioning are incorporated, but only in two ages. The comparison of young and midlife mice allows to test the interaction between age and
+
+<--- Page Split --->
+
+absence of hormones, but the absence of old mice makes it difficult to conclude whether gonadectomy at midlife accelerates cognitive and structural decline to match that of old ages. Also, Fig. 2a and Fig. 2c repeat data. Perhaps Fig.1 and Fig. 2 should be condensed into a single figure to present the model of intact (normal scenario) and gonadectomized mice (extreme scenario) and conclude which behavioral and structural analyses are more apt to study the effect of hormones. Likewise, in Fig. 4 (studies in ER knockout mice) Y Maze is not used, only midlife mice are tested, and immunohistochemical analyses of GFAP, Iba1 and PSD95 are not performed.  
+
+See response to the Editor. Our experimental strategy was focused on midlife females with a set of informative comparators.  
+
+Combining former Fig. 1 and 2 (current Fig. 2 and 3) would not align with each question investigated in each figure. An assessment of sex differences is the focus of current Fig. 2, while investigation of midlife females \(+ / - \mathsf{GDX}\) is the focus of current Fig. 3.  
+
+As suggested by the reviewer, MWM data are now shown as a merged graph with both between group comparisons of \(\%\) Time in Target Quadrant and within group comparisons of \(\%\) Time in Target Quadrant compared to \(\%\) Time in Other Quadrants. Former Fig. 2a and 2c (new Fig. 3b and 3c) make different comparisons of midlife females, one to midlife males and the other to young females, respectively.  
+
+ERbeta knock- outs are shown not only at midlife in former Fig. 4 (new Fig. 5), they are also shown at young age (see Suppl Fig. 1).  
+
+As suggested by the reviewer, additional immunohistochemical images and analyses for the ERbeta cKOs are now added to former Fig. 4 (new Fig. 5) of the revised manuscript.  
+
+Fig. 3 e,f. I guess that it is normal that \(\%\) time in other quadrants' is the mirror/inverse version of \(\%\) time in target quadrant'. If so, why is the scale in the Y axis of Fig. 3f different from the one in 3e? For example, a \(20\%\) time in target quadrant should correlate with \(80\%\) time in other quadrants.  
+
+We are not reporting \(\%\) time in the other quadrant, instead it is \(\%\) time in the other quadrants plural (an average of the \(\%\) time in the other 3 quadrants). Thus, \(\%\) time in other quadrants is not the mirror/inverse version of \(\%\) time in the target quadrant. For example, \(20\%\) time in TQ would align with \(80\%\) total time in the other three quadrants, with the average of the three other quadrants \(= 26.7\%\) .  
+
+## B.2. Overinterpretation of correlations.  
+
+Direct results should be differentiated from interpretations. For example, line 212 'Neuroprotection in females at midlife is mediated by ERbeta in astrocytes' is not a correct label. The experimental design will show the effect of ERbeta removal, but it does not demonstrate that the deleterious mechanisms are the same as those mediating damage caused by gonadectomy, which affects many hormones. Such an equivalence is a correlation. Reversal experiments such as addition of ERbeta agonists to gonadectomized mice may beautifully strengthen the connection the authors wish to make.  
+
+As suggested by the reviewer, to beautifully strengthen the connection, "reversal experiments such as addition of ERbeta agonists to gonadectomized mice", are now done (see new Fig. 7 of the revised manuscript).
+
+<--- Page Split --->
+
+The statement, "Neuroprotection in females at midlife is mediated by ERbeta in astrocytes" is a fair based on our data. When selective deletion of a receptor (in the cKO) causes a deleterious effect, then the presence of that receptor prevents the deleterious effect (in the WT). Here, cKO and WT were females at midlife.  
+
+### C. Methodologies:  
+
+Supplemental Fig. 1. presents data from estrogen receptor alpha knockout mice that are not described in Methods. These data are important and could be moved to the main text. Likewise, the PCR results confirming successful deletion of estrogen receptor genes should be shown.  
+
+As suggested by the reviewer, confirmation of specific deletion of estrogen receptor beta is now shown (see new Fig. 4 of the revised manuscript). Supplemental data on ERalpha knock- out mice are now removed since repeating all outcomes in another knock- out would take focus away from the role ERbeta. Instead we provide additional data on ERbeta including 1) treatment with the ERbeta ligand to reverse hippocampal outcomes, and 2) RNA- sequencing and gene expression analysis of the astrocyte transcriptome in midlife females with deletion of ERbeta in astrocytes.  
+
+It would be nice to have images of MRI in different experimental scenarios in Fig. 1.  
+
+Other than the substructure delineations overlaid onto the brain image in new Fig. 2 (former Fig. 1), it is unclear what image is being requested. Notably, MRI images from young to midlife to old mice will not appear to differ upon visual inspection. The standard approach required is quantification of neuroanatomical substructure volumes in 3D using atlas- based morphometry, which is what we did.  
+
+The quantification of immunohistochemical data is not explained in sufficient detail, including the number and distance between sections, stereological considerations, programs used to quantify, thresholding, background subtraction, etc. Also, images of PSD95 immunostaining, perhaps the extreme cases, would be useful to visually support one of the most interesting finding of the study.  
+
+As suggested by the reviewer, the revised manuscript now has detailed descriptions of immunohistochemical analyses, including antibody information and the analysis program used. Representative images of reactive astrocytes, microglia activation, disease- associated microglia, and synapses have each been added (see new Fig. 3 and Fig. 5 of the revised manuscript). Regarding synapses, we have now shown interesting immunostaining and analyses for SYN1 (including images) in the revised manuscript).  
+
+### D. Style issues:  
+
+The authors may consider removing 'Mind the Gap' from the title.  
+
+As suggested by the reviewer, the title is revised, removing 'Mind the Gap'.  
+
+The manuscript is well written, but there is always room for improvement. In the Abstract ...  
+
+The Abstract is rewritten.
+
+<--- Page Split --->
+
+Line 52 'during health' is unnecessary. The study does not need to touch on the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy. Also, the Marxian sentence (from Groucho) 'A better understanding of the effect of aging can provide insights into the effect of aging' should be rephrased for clarity.  
+
+We agree with the reviewer that our study does not need to touch on the controversy of healthy or unhealthy aging. Our use of the term disease means neurological disease (not aging in otherwise healthy). The full sentence is: "A better understanding of the effect of brain aging during health can provide insights into the effect of brain aging during disease". "During health" is necessary to contrast with "during disease". This refers two different aging scenarios, whereby mechanisms in one can inform the other.  
+
+Line 59. It appears from the sentence that the impact of timing and estrogen type (does it mean receptor type?) has been already clarified according to the literature (refs 6- 9), in contradiction with the problem statement, and with line 88 (refs 8, 28). Please, succinctly specify what is known and what are the areas of conflict.  
+
+Line 62. This sentence belongs to a review, there is no need to defend the need for separation of sexes in a study using, moreover, a reference from 2012. Likewise, the rest of the paragraphs can be summarized, or moved to the Discussion. For example, it should be highlighted in the Discussion that the mice mirror the sex- dependent evolution of cognition and atrophy in humans.  
+
+Lines 212 and 254. These two paragraphs can be condensed into one. Also, the explanation of the contextual fear conditioning (line 256) should be moved to earlier in the text, when the meaning of the different tests is explained.  
+
+## Wording is changed in the revised manuscript.  
+
+Line 91. Again, the study does not inform about brain region- specific therapeutic approaches. What it does is to unravel which brain regions are more affected by the experimental manipulations; nor the authors address the regional heterogeneity of astrocytes or neurons, for their genetic manipulations are global.  
+
+See above regarding clarification of "Region- specific targeting".  
+
+In Discussion, the clinical implications of the findings can be summarized. Text condensation is recommended.  
+
+## Condensation is now done in the revised manuscript.  
+
+As most articles, including those written by English- speaking authors, the manuscript may benefit from professional copy editing.  
+
+Thank you very much for these suggestions which have substantially improved the manuscript!  
+
+Reviewer #2 (Remarks to the Author):
+
+<--- Page Split --->
+
+Title: While I understand the underlying idea, i.e. that ERβ in astrocytes could be an important candidate for future therapeutic targets in cognitive aging in female. I tend to find it overclaiming. The authors do not demonstrate that ERβ solely mediates the potential improvement of memorisation in aging, for example using a specific drug targeting those receptors, or losing a positive effect after the deletion.  
+
+As suggested by the reviewer, we have now added the use of "a specific drug targeting those receptors", namely ER beta ligand treatment (see new Fig. 7 of the revised manuscript).  
+
+Results using the astrocyte- specific knock- out show that ER beta in astrocytes is necessary to prevent deleterious effects on cognitive outcomes at midlife in females. We do not claim that "ERβ solely mediates the potential improvement ...". Our finding is not mutually exclusive of additional potentially protective effects of other molecules or cells.  
+
+1. Behaviour: I was very surprised to read that the authors did not find an age-related defect of spatial memory in the Morris Water Maze (MWM) experiment (Fig1). It is a well-established fact that hippocampus-dependent types of memory, such as spatial, reference, or episodic memory, are sensitive to aging across species (please see the body of work from Catherine Barnes, Howard Eichenbaum, Michela Gallagher, Michael Yassa, Aline Marighetto, Sara Burke, to name a few). In trying to understand this lack of defects, I came to think that the paper would benefit not only from showing the performance during acquisition, at least in the supplementary material, as a control of learning, but also from providing more details about the paradigm. Indeed, depending on the protocol, the memory can go from spatial to procedural (i.e., hippocampus-independent), which remains relatively intact during aging. Alternatively, the authors could also represent the target quadrant exploration during the test as blocks of 15 sec to clarify the mouse behaviour. That being said, I think the high performance from aged mice is due to the fact that the authors tested the mice 2h after the last training session. This could explain the difference between retrieval in the MWM and the fear conditioning, which was tested 24h post-conditioning, and did reveal the age-related memory decline. I would suggest to the authors to homogenise the conditions of behavioural testing throughout the paper, for clarity of understanding and comparison, or at least to address these discrepancies in the discussion. As the age-related decline of hippocampus-dependent memory is such a long-standing fact, I don't think the study benefits from starting with such a clashing result and could deter the readership. Especially since in GDX mice, the defect appears earlier, at "mid-life", and could be interpreted as early aging. Finally, regarding the behaviour part, I would like to draw the authors' attention to the fact that the dichotomy in the role of the hippocampus to sustain recent but not remote memory has been revised in recent years (Goshen et al., Cell 2011) benefiting from the advancement of optogenetics. While I agree with the authors' conclusion regarding their own result, I would advice to modify the introduction of this part to take into account recent literature rather than Frankland and Bontempi's early work. A minor remark on this part, I am sure what is the authors' rationale for making this data as a supplementary figure only, when this is an interesting result.  
+
+We agree with these insightful comments of the reviewer. Various protocol paradigms (timing, etc.) can affect behavioral performance. Here, using the MWM protocol as described, with the sample size as indicated, we did not observe a significant defect in spatial reference memory in the Morris Water Maze in the 2hr delay probe trial in gonadally intact, healthy females or males up to the ages tested in the environment of the UCLA Behavioral Core Facility. This lack of a significant deficit has been seen before in healthy control groups of studies focusing on AD mice where healthy mice served as
+
+<--- Page Split --->
+
+the cognitively normal control, albeit often tested at ages a few months earlier. In agreement with the reviewer, we observed that older ages had a learning speed difference (older slower than younger), so a training learning graph could be added as a supplemental figure upon request. However, our focus here is not to do additional behavioral tests to show significant deficits in gonadally intact mice using a given paradigm. Instead, our focus is on the significant worsening observed in GDX females compared to gonadally intact females using a standard MWM protocol.  
+
+2. ERβ cKO. I think it would be helpful to provide more details regarding the origin of the mice and selectivity of the conditional KO.  
+
+As suggested by the reviewer, this is now done (see new Fig. 4 in the revised manuscript).  
+
+3. Neurodegeneration: One major point that I would like to draw the authors' attention towards, is their multiple reference to neurodegeneration throughout the paper, and the lack of distinction between normal aging and pathological aging. From what I understand, the authors seem to suggest that their study addresses neurodegeneration. However, the study has been conducting in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects), as demonstrated in my previous point, these wildtype aged mice can demonstrate high performance in certain conditions (Fig1), which would not be the case in pathological aging. In normal aging, the memory defects are associated with atrophy, but not neuronal loss, unlike pathological aging Moreover, the underlying mechanisms, and progression of alterations are widely distinct according to the type of aging. Therefore, I do not think it is proper in their condition to refer to neurodegeneration and would reframe the paper as normal aging. They could address the potential importance of their work regarding pathological aging in the discussion, but the use of wildtype mice do not support the direct extrapolation towards neurodegeneration.  
+
+It is unclear what is meant by the need to "reframe the paper as normal aging". As the reviewer points out: 1) Our study was "conducted in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects)", and 2) "these wildtype aged mice can demonstrate high performance in certain conditions". Our manuscript is framed on the mice we used, those with "normal aging", and these two correct statements by the reviewer show that we have conveyed that.  
+
+We did not make a "distinction between normal aging and pathological aging". As stated by Rev #1, we are not entering "the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy",  
+
+Use of the word "neurodegeneration" is not problematic when describing a condition characterized by cognitive decline, glial activation, synaptic loss, and brain atrophy, indeed what occurs in midlife females either GDX or with selective deletion of ER beta in astrocytes.  
+
+4. Discussion: I think the work would benefit from stressing out throughout the text and in particular the discussion, the novelty of the research, and replacing the work within the literature. It is my belief that aside from the very interesting distinctive role of ERβ in astrocytes vs. neurons, the rest of the work has in some form been already reported. Age-related memory deficits (see authors in part 1 for reference); There is also a large body of work on the sexual dimorphism in cognitive aging (e.g. Frick et al., 2000; for review, Frick et al 2008), and on the role of estrogens in the brain and learning and memory, and effect of ovariectomy, which deserve to be cited in my opinion (L. Galea, K. Frick, E. Waters, M. Adams, C. Wolley, B.
+
+<--- Page Split --->
+
+McEwen – non- exhaustive list). Similarly, the atrophy of the hippocampus has been demonstrated multiple times (for review Barnes and Burke, 2010). Finally, the reduction in spine density following ovariectomy has been previously reported by the McEwen laboratory (Gould et al., 1990). I do not intend to denigrate the study; I simply believe that the authors would benefit from valorising their findings in light of the existing literature.  
+
+Final point on the discussion, I fail to understand why the authors address hormone replacement and MS in such length, when their study does not use estrogen supplementation, potential effect on alteration similar to MS, and that MS have limited cognitive effect. Rather, I would be very interested in reading the authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes.  
+
+The revised manuscript now includes estrogen supplementation (see new Fig. 7). We thank the reviewer for highlighting our finding of a "very interesting distinctive role of ERβ in astrocytes vs. neurons". We have now further added additional novel findings on gene expression analyses of astrocyte transcriptomes in midlife females with ER beta selectively deleted. This revealed an exciting link to previous work by others on the role of glucose metabolism in brain during menopause, albeit previously not studied in hippocampal astrocytes.  
+
+The literature cited by this knowledgeable reviewer is excellent and could be the basis for a stand- alone review article. We agree with the reviewer that some of our work has been done in parts in papers by others. The strength here is integration of a breadth of complementary studies (some done before, some not) in the same paper, namely behavioral studies, in vivo MRI for regional atrophy, underlying neuropathology, cell- specific knock outs, RNA- sequencing, and estrogen receptor specific treatment reversal). Rather than reviewing parts published in the past, we took our discussion in a future direction. We discussed implications of the totality of our findings on clinical translation. This addresses the unmet need for a hormone replacement therapy designed more specifically and based on neuroscience mechanisms underlying cognitive outcomes in midlife females.  
+
+Regarding justification of the discussion of MS, these patients are predominantly female, have substantial cognitive disability, and menopause confers overall disability worsening. In addition, clinical trials testing a naturally occurring ER beta ligand (estriol) treatment showed neuroprotective effects in women with MS. This published work warrants discussion of the potential for repurposing this treatment approach to cognitive deficits in healthy women with menopause.  
+
+We thank the reviewer very much for their interest in getting the "authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes". Dr. Voskuhl is an expert on the study of sex differences, and this was the starting point of this manuscript (new Fig. 1 and 2). However, the trajectory of this manuscript is to use initial sex differences findings as a basis for subsequent research focused on gaining cellular and molecular insights relevant to midlife females with cognitive deficits. A discussion of sexual dimorphism as described by the reviewer would align better with a manuscript that studied both females and males with the full breadth of our outcomes from young to midlife to old ages. While that is interesting, it would be a different paper.  
+
+Minor comment: I think the reader would benefit from the homogenisation of the statistical report, which at the moment is alternatively in the legend or the text.
+
+<--- Page Split --->
+
+As the reviewer suggested, this is now done in the revised manuscript. We also have a statistical summary ready for submission pending the response to this appeal to proceed.  
+
+Looking forward to reading your response, AS Al Abed  
+
+Thank you for your insightful comments.  
+
+Reviewer #3 (Remarks to the Author):  
+
+1) The authors use GFAP-cre mice to constitutively delete ER beta in astrocytes. These mice cannot be used to conclude about implication of astroglial proteins, when these proteins are also expressed in neurons.  
+
+There are indeed major caveats in using the GFAP-Cre mice, as it has been well documented by numerous groups that deletion occurs in both neurons and astrocytes, since GFAP is expressed in precursor cells during early development. This confounds the authors major conclusion that deletion of ER beta from astrocytes contributes to cognitive deficits and brain atrophy in menopausal female.  
+
+A few papers here listed well describe these caveats : "Looks Can Be Deceiving: Reconsidering the Evidence for Gliotransmission" https://doi.org/10.1016/j.neuron.2014.12.003  
+
+Germ- Line Recombination Activity of the Widely Used hGFAP- Cre and Nestin- Cre Transgenes https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3857304/  
+
+Expression Specificity of GFAP Transgenes https://link.springer.com/article/10.1007/s11064- 004- 6881- 1  
+
+Other experiments should thus be combined with the use of these mice, such as the use of inducible cKO mice and/or rescue experiments targeted to astrocytes (re- expression of ER beta specifically in astrocytes in the astroglial cKO mice).  
+
+There are different subtypes of mGFAP- Cre lines. The mGFAP- Cre line 77.6 used in this manuscript has now been characterized, and specificity for astrocytes and not other cell types (including neurons) was shown (see new Fig. 4 in the revised manuscript).  
+
+Recent publications used the mGFAP- Cre 77.6 line for selective gene deletion in astrocytes: 1) Deerhake ME et al, Immunity, 2021, and 2) Jia YF et al, Behav. Brain Res. 2021. They also showed specificity of deletion in astrocytes.  
+
+The reviewer states that GFAP- Cre driven deletion in our experiments is not specific to astrocytes, that ERbeta was also deleted from neurons, so our effects on cognitive outcomes at midlife may be due to ERbeta deletion in neurons (not astrocytes). If that were true, then one would see effects on cognitive outcomes in the neuronal ERbeta cKO (NSE- Cre), but that was not the case (see former Fig 4, new Fig. 5).  
+
+Specificity of GFAP- Cre driven selective gene deletion can differ not only between different mGFAP- Cre lines, but also between mGFAP- Cre versus hGFAP- Cre, and between sexes. The mGFAP- Cre 77.6 line has shown some expression of Cre recombinase in the male, but not the female, germline
+
+<--- Page Split --->
+
+(https://www.jax.org/strain/024098). To avoid this ectopic recombination, the mGFAP-Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP-Cre 77.6: ERβ<sup>ff</sup> females were crossed with ERβ<sup>ff</sup> males (without the mGFAP-Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP-Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed.  
+
+Regarding use of an inducible Cre mouse line, cognitive and hippocampal outcomes observed at midlife were not observed in astrocyte ERbeta cKO mice at age 3 months (see Suppl. Fig. 1). Thus, cognitive and hippocampal outcomes observed in astrocyte ERbeta cKO mice at midlife were not due to effects on brain development.  
+
+2) There is no validation of these newly generated astroglial- and neuronal-specific ER beta transgenic models. One would like to see quantification of ER beta expression in astrocytes and neurons in both cKO mice (gfap-cre:ER betafl/fl and nestin-cre:ER betafl/fl).  
+
+As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript).  
+
+Validation of ERbeta deletion specifically in neurons using the NSE-Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript).  
+
+3) There is a lack of details regarding the mice (origin (stock number or laboratory source), breeding strategy controlling germline recombination, appropriate littermate controls. This precludes evaluation of whether the authors have used appropriate controls.  
+
+As suggested by the reviewer, details of information about mice, including references, have now been added to the Methods section. Also, schematics of breeding have been added to new Fig. 4a and 4d of the revised manuscript.  
+
+4) The authors use GFAP immunostaining to assess astroglial reactivity. GFAP expression can vary independently of astroglial reactivity. The authors should thus use additional markers (such as vimentin, stat3, GS...) and analyse astrocyte morphology to conclude about astrocyte reactivity.  
+
+As suggested by the reviewer, reactive astrocytes are now stained as double positive: LCN2\*GFAP\* (see Fig. 3, Fig. 5, and Fig. 7 of the revised manuscript.  
+
+5) The authors report synapse loss in menopausal female. They however only performed PSD95 staining to assess synapses. A synapse is composed of a pre- and a postsynaptic element and can only be identified by the colocalization of both elements. Staining for both pre- and postsynaptic markers should thus be performed to quantify synapse number.  
+
+Pre- synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)  
+
+6) This manuscript completely lacks mechanistic insights. How ER beta in astrocytes contributes to brain structure and cognitive performance in male and female? Does this require a crosstalk with other cell types such as microglia?
+
+<--- Page Split --->
+
+Mechanistic insights are shown by a series of causality experiments. 1) An ovarian hormone by age interaction was found, whereby both midlife aging and loss of ovarian hormones caused cognitive decline, neuropathology, and atrophy of dorsal hippocampus by in vivo MRI. 2) Selective deletion of estrogen receptor beta (ERβ) in astrocytes, but not neurons, caused cognitive decline, neuropathology, and atrophy of dorsal hippocampus by in vivo MRI. 3) Complementing our two loss- of- function causality experiments (gonadectomy and specific deletion of ERβ in astrocytes), gain- of- function experiments using ERβ ligand treatment of midlife females showed that treatment caused reversal of outcomes. 4) Gene expression differences in hippocampal astrocyte transcriptomes from midlife females with selective deletion of ERβ vs WT in astrocytes revealed Gluconeogenesis I and Glycolysis I pathways as the most differentially expressed pathways. Further studies of a key gene in the Gluconeogenesis I pathway, Eno1, were also done (see new Fig. 6 in the revised manuscript).  
+
+Additional mechanisms related to cross talk between astrocytes and microglia are interesting and discussed. Further experiments addressing this possibility are beyond scope here.  
+
+Reviewer #4 (Remarks to the Author):  
+
+The narrative is very interesting, the results important and well contextualized. The work is extremely significant and timely. The methodology is rigorous, although a few important pieces of evidence are missing to have a more balanced picture and close the story (detailed below). I also have doubts on some of the statistical approaches chosen. The paper is easy to follow and very well written.  
+
+There are a few aspects that can be improved. Below, my suggestions.  
+
+## Thank you!  
+
+1  
+
+Fig.1 : the information contained in a- b and c- d is redundant. In addition, doing twice a scatter plot, which distributes points random, on the same data generates plots which are visually different, which is awkward. The same applies to Fig. 2 a- b and c- d, and Fig. 4 a- b. I think half of the plots can safely go to supplementary.  
+
+As suggested by the reviewer, to remove redundancy MwM data are now shown as a merged graph with both between group comparisons of % Time in Target Quadrant and within group comparisons of % Time in Target Quadrant compared to % Time in Other Quadrants. This is done in former Fig. 2 (new Fig. 3) and in former Fig. 4 (new Fig. 5)  
+
+2  
+
+Following the previous comment, I think the most correct statistical approach would be a single ANOVA for the water maze data where sex is a factor (in addition to age and quadrant), followed by appropriate post- hoc comparisons. The same applies for MRI volumetric data, where ROI should be another factor (cortex, striatum, HC).
+
+<--- Page Split --->
+
+In the revised manuscript, given the relatively small sample size and unequal variance, MWM was analyzed using a nonparametric test, Mann Whitney \(U\) test. All volumetric MRI data were analyzed using a two- way ANOVA (age, sex) and then followed up with two- tailed Welch's t- tests, appropriate for these comparisons. See item 4 below.  
+
+3  
+
+In Fig 1 h, it seems that there is no volume change in HC in male mice across the lifespan. This results somehow weakens the point made by the authors in the rest of the paper, about the importance for neuroprotection across the lifespan of the ERb astrocyte receptors specifically in the dorsal hippocampus, unless the dHC is only important in ageing female, which is hard to digest, unless supported by evidence. Generally speaking, the paper is missing showing and discussing in males at least some of the analyses performed in females, for example Fig 2 g- i.  
+
+The reviewer is correct. While females had dorsal hippocampal atrophy from midlife to old age, males did not.  
+
+Analysis of data from the Genotype- Tissue Expression (GTEx) project examined sex differences in gene expression across 44 tissues in humans and showed that \(37\%\) of all genes exhibit sex- biased expression in at least one tissue (Oliva et al., Science, 2020). In another study using the same dataset focusing on 29 human healthy tissues, whole- genome expression profiles showed distinct sex- biased regulatory networks in each tissue (Lopes- Ramos, Cell Rep. 2020). Finally, sex differences in gene expression are region- specific and cell- specific within the brain (Kim- Hellmuth, et al., Science, 2020). These studies underscore the pervasiveness and complexity of sex differences in gene expression during health, which can be distinct depending on the brain region and cell involved.  
+
+Sex differences in substructure volumes and rates of atrophy have previously been shown in MRI studies in humans and mice. Dorsal hippocampus and ventral hippocampus differ in regard to which is larger in females versus males, each when normalized for brain size.  
+
+Given the above evidence in the sex differences field, we did not find it particularly surprising that dorsal hippocampus may undergo atrophy with aging in females and not males. That said, this manuscript did not pursue why dorsal atrophy did not occur in males with aging. A related question is why did males have gradual atrophy with aging in frontal cortex and striatum, but not in dorsal hippocampus? It could be due to direct effects of testosterone or dihydrotestosterone acting on androgen receptors in hippocampus in males. This is only one of several possibilities. Space limitations do not permit our speculations about males during aging since our studies focused on females during aging.  
+
+Fig 1, supplementary excel table: why using Welch test? Is it because the samples have unequal variance? Please specify in the methods.  
+
+Yes, the samples do exhibit unequal variances, as is the case in almost all biological measures. Thus, the use of Welch's t- test is appropriate here since it performs better than the Student's t- test when sample sizes and variances are unequal between groups, and it gives identical results when sample sizes and variances are equal. We routinely use Welch's t- test for volumetric analyses in atlas- based morphometry.  
+
+Fig 1: It is noteworthy that from Fig. 1 there is no relation between structure and function, at
+
+<--- Page Split --->
+
+least in the context of ageing. If it is a matter of different sensitivities of the two techniques employed (MRI volumetry vs behaviour), as hinted by the authors, I suggest they discuss the issue and report the power analysis used for determining the sample size. Perhaps the behavior is underpowered?  
+
+See response to Reviewer #2 regarding not observing a significant defect in spatial reference memory in the Morris Water Maze in the 2hr delay probe trial in gonadally intact, healthy females or males up to the ages tested in the conditions of the UCLA Behavioral Core Facility given the sample size used. Sample sizes were driven by the number of mice needed to generate a robust minimum deformation atlas (MDA) for atlas- based morphometry (MacKenzie- Graham Neurolmage 2012). Logistically, with several groups (various ages, two sexes, and/or GDX vs sham surgery, cKOs vs WT, treatments) and several outcomes, sample size was also limited by what was practically feasible by personnel and costs. Data from in vivo MRI is expected to be more sensitive than clinical testing in mice. This aligns with in vivo MRI for brain atrophy being a more sensitive outcome than clinical outcomes in humans. Indeed, MRI is very frequently used as a more sensitive biomarker in Phase 2 clinical trials prior to designing Phase 3 trials which require much larger sample sizes since they have a clinical primary outcome (such as cognitive performance).  
+
+6 Fig.2: Similarly to comments 1 and 2, the figures of the water maze are redundant, and the statistical model could be just one for each test but include everything (gender, therapy, life stage).  
+
+As suggested by the reviewer, redundancy of MwM data has been decreased by making a single merged panel for between groups and within groups comparisons. See comments above regarding statistical tests used.  
+
+7 Fig. 2 f: why freezing was only measured at midlife?  
+
+Fear conditioning was done to expand upon MwM test performance at midlife in females, the group and age of focus. It was not done at all ages, in both sexes, in + / - GDX, or in cKOs. Therefore, fear conditioning has now been removed from the revised manuscript.  
+
+8  
+
+Fig. 2 j- k: I suggest to only plot regression line when significant. Also, there seems to be a lot of points in the plots. Are all animals (both sexes, all ages, GDX/sham) pulled together? If so, which is the rationale? Perhaps the different groups could be shown in the plot by coloring the datapoints differently.  
+
+As suggested by the reviewer, dot plots in former Fig. 2 (new Fig. 3 p,q) have now been made using distinct coloring of datapoints (see below). This can replace the panel with all black dots upon request. The rationale of pooling groups is to increase the sample size when asking if cognitive performance (% Time in Target Quadrant) correlates with dorsal (p) or ventral (q) hippocampal volume by in vivo MRI. The plots nicely show a positive correlation between better cognitive performance and higher volumes of dorsal (but not ventral) hippocampus. This is presented in our manuscript since in vivo MRI for hippocampal atrophy is not usually done in papers showing behavioral testing in mice.
+
+<--- Page Split --->
+![PLACEHOLDER_20_0]
+  
+
+Possible panel replacement in new Fig. 3. % time in TQ correlated with (p) dorsal hippocampus volume \((r = 0.28\) ; \(p = 0.011\) ), not with (q) ventral hippocampus volume \((r = 0.027\) ; \(p = 0.809\) ). Dot colors: black (astrocyte ERβ cKO, old), yellow (astrocyte ERβ cKO, young), blue (neuron ERβ cKO, old), green (neuron ERβ cKO, young), red (WT, old), pink (WT, young).  
+
+9  
+
+Fig. 3, Neuropathology: the authors talk about "area fraction" but what is exactly measured and how? Can distinct aspect of the morphology be measured (ramifications density, cell body size, cell number etc.? This is important as the term "glia activation" can refer to distinct morphological changes happening to the cell. In general, the method- >histology section is missing details needed for the reader to be able to reproduce the analyses.  
+
+The term "area fraction" is widely used during double-labelling. For example, positive labeling in a specific cell type was quantified and graphed as the percentage labeling of LCN2+ area within the GFAP+ area. A level of detail needed for the reader to reproduce histology (information about antibodies, dilution factors) is now in the Methods section of the revised manuscript.  
+
+10  
+
+Line 194: the notation is misleading. Perhaps by "direct" and "indirect" correlations the authors mean linear positive and linear negative? Or do the authors really mean indirect as mediated by something else? If so, specify.  
+
+The reviewer is correct. Direct is positive and indirect is negative.  
+
+11  
+
+Fig 3 is missing the plots of the correlations between time in TQ and gliosis (data is discussed but not shown).  
+
+The correlation between % Time in Target Quadrant and glial activation markers are stated in the results section. Plots showing the correlations between % Time in Target Quadrant and LNC2+GFAP+ astrocytes, as well as % Time in Target Quadrant and MHCII+ IBA1+ microglia are presented below. This can be added as additional supplemental figure upon request.
+
+<--- Page Split --->
+![PLACEHOLDER_21_0]
+  
+
+Possible additional supplemental figure. Correlation of MWM cognitive performance with glia activation. a- b) % Time in target quadrant is negatively correlated with a) LNC2\*GFAP\* astrocytes \((r = - 0.46203\) , \(p = 0.017\) ) and b) MHCII\* \(\mathrm{IBA1^{+}}\) microglia \((r = - 0.40723\) , \(p = 0.048\) ). red dot, Young Sham; red dot with black rim, Young GDX; black square, Midlife Sham; white square, Midlife GDX females. Pearson correlation analyses.  
+
+12  
+
+Following the rationale of the paper, Fig 4 is missing the neuropathology. How are the astrocytes affected by ERb deletion? Are they more activated at midlife?  
+
+As suggested by the reviewer, this neuropathology has now been done (see new Fig. 5e- l in the revised manuscript).  
+
+13  
+
+Following the previous comment but more generally, what is ultimately the mechanism that the authors identified as leading to atrophy and worst cognition in ageing females? Is it astrogliosis? If so, the authors should report more neuropathological data (or other equivalent astrogliosis markers) to make a stronger point.  
+
+In addition to more neuropathology, we have also added new Fig. 6 showing gene expression in astrocytes by RNA- sequencing. See response #6 to Reviewer #3.  
+
+14  
+
+It is interesting that there is no cognitive effect in young female mice which lacks endogenous hormones or receptors. What about HC volume?  
+
+There were no differences in HC volumes in young female GDX vs young female sham. See the figure below, which can be added as a supplemental figure upon request. This, together with our other findings, reveals an ovarian sex hormone by age interaction. In essence, it is deleterious to have BOTH aging to midlife AND to lose ovarian hormones.
+
+<--- Page Split --->
+![PLACEHOLDER_22_0]
+  
+
+Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Substructure volumes, assessed by MRI, taken as a percentage of intercranial volume (ICV) are shown for (a) hippocampus, (b) dorsal hippocampus, and (c) ventral hippocampus. There were no differences in substructure volumes between groups at young age. \(n = 8\) for each group.  
+
+15  
+
+As the authors point out in the discussion, microglia also have ERb. Can the author show and compare some histology of microglia in animals lacking endogenous receptors and animals lacking ERb in astrocytes?  
+
+As suggested by the reviewer, this is now added to the revised manuscript. See double- labelling stains for MHCII & IBA1 as well as for CLEC7A & P2RY12. This is done in sham vs GDX (new Fig. 3), cKO vs WT (new Fig. 5), and ER beta ligand treatment (new Fig. 7).
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+Reviewer #3 (Remarks to the Author):  
+
+The authors have addressed several of my comments in the revised manuscript. Yet there are still a few points that were not satisfactorily addressed and needs to be done:  
+
+1) "To avoid this ectopic recombination, the mGFAP-Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP-Cre 77.6: ERβf/f females were crossed with ERβf/f males (without the mGFAP-Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP-Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed."  
+
+Please add in the methods section the breeding strategy.  
+
+2) "As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript). Validation of ERbeta deletion specifically in neurons using the NSE-Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript)."  
+
+To be fully convinced by the astroglial cKO mice in, the authors need to illustrate that the expression of ER beta is intact in neurons from the gfap- cre:ER betafl/fl.  
+
+3) "Pre-synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)"  
+
+I understand the SYN1 staining is done and illustrated in these figures, but it is not used in the manuscript to quantify synapse number by counting the number of colocalized SYN1 and PSD95 puncta. This needs to be done.  
+
+Reviewer #4 (Remarks to the Author):  
+
+The authors have successfully addressed my concerns and those from other reviewers. Minor comments:  
+
+Possible panel replacement in new Fig. 3: I think that the plot can stay by/w.  
+
+"Direct" and "indirect" to qualify correlation is used but not precise. Indirect strictly speaking means "mediated by". I recommend using positive or negative.  
+
+Plots showing the correlations between \(\%\) Time in Target Quadrant and LNC2+GFAP+ astrocytes, as well as \(\%\) Time in Target Quadrant and MHCII+ IBA1+ microglia: I would add them in the supplementary.  
+
+Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Also suitable for supplementary material.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS followed by Point-by-point Responses.  
+
+## Reviewer #3 (Remarks to the Author):  
+
+Comment 1) "To avoid this ectopic recombination, the mGFAP- Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP- Cre 77.6: ERβf/f females were crossed with ERβf/f males (without the mGFAP- Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP- Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed." Please add in the methods section the breeding strategy.  
+
+Response 1) Now added to the methods section as well as referred to in the results section of the revised manuscript is the following:  
+
+In breeding of our mouse lines, Cre recombinase alleles were always inherited from females (mother), not from males (father). This is because one Cre line showed ectopic expression of Cre recombinase in the male germline (mGFAP- Cre 77.6 line; https://www.jax.org/strain/024098). Thus, paternal inheritance of Cre recombinase should be avoided. To this end, our breeding pairs were: a) GFAP- Cre:ERβf/f females with ERβf/f males (to generate astrocyte ERβ cKO mice), b) NSE- Cre:ERβf/f females with ERβf/f males (to generate neuron ERβ cKO mice), c) GFAP- Cre:RiboTag females with RiboTag males (to generate astrocyte RiboTag mice), and d) GFAP- Cre:RiboTag:ERβf/f females with RiboTag: ERβf/f males (to generate astrocyte ERβ cKO RiboTag mice).  
+
+Labels of Male or Female to parents during breeding have also been added to Fig. 4 of the revised manuscript.  
+
+Comment 2) "As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript). Validation of ERbeta deletion specifically in neurons using the NSE- Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript)." To be fully convinced by the astroglial cKO mice in, the authors need to illustrate that the expression of ER beta is intact in neurons from the gfap- cre:ER beta/f/f.  
+
+Response 2) In the revised manuscript a supplemental figure is added focusing on ERβ expression in neurons. ERβ expression was intact in neurons of the astrocyte- ERβ cKO (see Supplemental Fig. 3).  
+
+Comment 3) "Pre- synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)". I understand the SYN1 staining is done and illustrated in these figures, but it is not used in the manuscript to quantify synapse number by counting the number of colocalized SYN1 and PSD95 puncta. This needs to be done.  
+
+Response 3) Co- localization of pre- synaptic staining for SYN1 and post- synaptic staining for PSD95 is now done in the revised manuscript (see Fig. 3h, Fig. 5l, and Fig. 7f).  
+
+## Reviewer #4 (Remarks to the Author):  
+
+Comment 1) The authors have successfully addressed my concerns and those from other reviewers. Response 1) Thank you.  
+
+Comment 2) Minor comments:  
+
+Possible panel replacement in Fig. 3: I think that the plot can stay b/w.  
+
+Response 2) As recommended by the reviewer, panels (in former Fig. 3p,q; current Fig. 2m,n) showing correlation analyses between Dorsal Hippocampus volume and % Time in Target Quadrant or between Ventral Hippocampus volume and % Time in Target Quadrant will stay black and white.  
+
+Comment 3) "Direct" and "indirect" to qualify correlation is used but not precise. Indirect strictly speaking means "mediated by". I recommend using positive or negative.  
+
+Response 3) The word "direct" correlation has now been revised to "positive" correlation to describe results in Fig. 2m. The word "indirect" correlation has now been revised to "negative" correlation to describe results in Supplemental Fig. 2.  
+
+Comment 4) Plots showing the correlations between % Time in Target Quadrant and LNC2+GFAP+
+
+<--- Page Split --->
+
+astrocytes, as well as % Time in Target Quadrant and MHCII+ IBA1+ microglia: I would add them in the supplementary.  
+
+Response 4) These two plots showing correlations have now been added to supplementary materials (Supplemental Fig. 2).  
+
+Comment 5) Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Also suitable for supplementary material.  
+
+Response 5) This figure has now been added to supplementary materials (Supplemental Fig. 1).
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d5b6c7d85843913b9e894d85a549e8a375934c242b73b1338b64274e10565d08/supplementary_0_Peer Review File__d5b6c7d85843913b9e894d85a549e8a375934c242b73b1338b64274e10565d08_det.mmd

@@ -0,0 +1,835 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[68, 110, 360, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>title<|/ref|><|det|>[[104, 159, 888, 213]]<|/det|>
+# Estrogen receptor beta in astrocytes modulates cognitive function in mid-age female mice  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 784]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 912, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 85, 290, 97]]<|/det|>
+Reviewers' comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 111, 415, 125]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[119, 139, 234, 152]]<|/det|>
+A.Key findings:  
+
+<|ref|>text<|/ref|><|det|>[[117, 166, 878, 420]]<|/det|>
+Itoh et col. seek to refine hormone replacement therapy (HRT) by gaining insight into brains areas affected by hormone removal, as well as the distinct roles of estrogen receptors (alpha vs beta) and of neurons vs astrocytes. To this end, behavioral and structural analyses of brain areas with magnetic resonance imaging (MRI) are performed in three studies. One is a longitudinal analysis of male and female mice throughout life. Another study explores the effects of gonadectomy at midlife. And a third study examines mice with selective removal of estrogen receptors in astrocytes or neurons. They report age- , sex- and brain region- dependent atrophy that is exacerbated by gonadectomy and mimicked by targeted deletion of estrogen receptor beta in astrocytes. Taken together, the mouse data suggest that the neuroprotective actions of estrogens are cell- and receptor- dependent, primarily affecting dorsal hippocampus. It is also remarkable that gonadectomy results in severe decrease of spines in dorsal hippocampus. The study may inform structural analyses in HRT, as well as therapeutic strategies, with a focus on the emerging avenue of astrocyte- targeted therapies. The authors are to be commended by the hard work and the wealth of tests and animal models. Tests include hippocampal- dependent spatial memory (Morris Water Maze, MWM), working memory (Y- maze) and contextual fear conditioning. Of great value is the use of MRI to dissect out regional changes in association with behavioral deficits. Mouse models include intact and gonadectomized, as well as knock out mice for estrogen receptor types in astrocytes and neurons.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 433, 230, 446]]<|/det|>
+### B. Limitations:  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 461, 297, 475]]<|/det|>
+## B.1. Insufficient clarity.  
+
+<|ref|>text<|/ref|><|det|>[[118, 489, 877, 630]]<|/det|>
+Conceptual: The gap of knowledge ('understanding brain region- specific, cell- specific and receptor- specific mechanisms') that the authors aim to dissipate should be rephrased. As I understand it, the authors aim to clarify why HRT does not mitigate cognitive deficits and brain atrophy in menopausal women (problem statement) and propose that the conflicting results may arise from paradoxical actions of different estrogen receptors in different cell types, and the fact that readout measures are not directed to the right brain areas. There is a confusion in the said gap between actors (cells and receptors) and effects (brain areas and cognition), for, in the study, brain regions are not specifically targeted. The gene knockout affects cells all over the brain, and clinical therapies are systemically administered. The repeated expression 'region- specific targeting' throughout the text should be eliminated or clarified.  
+
+<|ref|>text<|/ref|><|det|>[[118, 643, 878, 813]]<|/det|>
+Experimental: Perhaps it is just me, but the study rationale is not clear, particularly which tests are used, and which ages analyzed. Specifically, in figure 1, in intact mice only MWM is performed at three ages, with negative results, while in Fig 2 (gonadectomized mice) Y Maze and contextual fear conditioning are incorporated, but only in two ages. The comparison of young and midlife mice allows to test the interaction between age and absence of hormones, but the absence of old mice makes it difficult to conclude whether gonadectomy at midlife accelerates cognitive and structural decline to match that of old ages. Also, Fig. 2a and Fig. 2c repeat data. Perhaps Fig.1 and Fig. 2 should be condensed into a single figure to present the model of intact (normal scenario) and gonadectomized mice (extreme scenario) and conclude which behavioral and structural analyses are more apt to study the effect of hormones. Likewise, in Fig. 4 (studies in ER knockout mice) Y Maze is not used, only midlife mice are tested, and immunohistochemical analyses of GFAP, Iba1 and PSD95 are not performed.  
+
+<|ref|>text<|/ref|><|det|>[[118, 826, 872, 882]]<|/det|>
+Fig. 3 e, f. I guess that it is normal that ' \(\%\) time in other quadrants' is the mirror/inverse version of ' \(\%\) time in target quadrant'. If so, why is the scale in the Y axis of Fig. 3f different from the one in 3e? For example, a \(20\%\) time in target quadrant should correlate with \(80\%\) time in other quadrants.  
+
+<|ref|>text<|/ref|><|det|>[[119, 896, 415, 910]]<|/det|>
+B.2. Overinterpretation of correlations.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 97, 873, 196]]<|/det|>
+Direct results should be differentiated from interpretations. For example, line 212 'Neuroprotection in females at midlife is mediated by ERbeta in astrocytes' is not a correct label. The experimental design will show the effect of ERbeta removal, but it does not demonstrate that the deleterious mechanisms are the same as those mediating damage caused by gonadectomy, which affects many hormones. Such an equivalence is a correlation. Reversal experiments such as addition of ERbeta agonists to gonadectomized mice may beautifully strengthen the connection the authors wish to make.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 210, 250, 223]]<|/det|>
+### C. Methodologies:  
+
+<|ref|>text<|/ref|><|det|>[[118, 238, 855, 280]]<|/det|>
+Supplemental Fig. 1. presents data from estrogen receptor alpha knockout mice that are not described in Methods. These data are important and could be moved to the main text. Likewise, the PCR results confirming successful deletion of estrogen receptor genes should be shown.  
+
+<|ref|>text<|/ref|><|det|>[[118, 293, 759, 308]]<|/det|>
+It would be nice to have images of MRI in different experimental scenarios in Fig. 1.  
+
+<|ref|>text<|/ref|><|det|>[[118, 321, 876, 378]]<|/det|>
+The quantification of immunohistochemical data is not explained in sufficient detail, including the number and distance between sections, stereological considerations, programs used to quantify, thresholding, background subtraction, etc. Also, images of PSD95 immunostaining, perhaps the extreme cases, would be useful to visually support one of the most interesting finding of the study.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 392, 238, 405]]<|/det|>
+### D. Style issues:  
+
+<|ref|>text<|/ref|><|det|>[[118, 419, 869, 504]]<|/det|>
+The authors may consider removing 'Mind the Gap' from the title, since journalistic expressions should be avoided from scientific reports, and it is not clear which gap should be minded until one reads the introduction. Also, the title should describe the model and main result of the study and not the implications thereof, as noted. Something like 'Selective deletion of ERbeta in astrocytes but not neurons mimics behavioral and structural deficits caused by gonadectomy in female mice at midlife' may work.  
+
+<|ref|>text<|/ref|><|det|>[[117, 517, 878, 672]]<|/det|>
+The manuscript is well written, but there is always room for improvement. In the Abstract, the sentence in line 26 starting with 'Some aspects of aging...' is dispensable. I already suggested rephrasing of the problem statement above, in noting that the results should be described in a neutral manner, distinguishing findings from interpretations. For instance, in line 30, an alternative to the current sentence is: Gonadectomy impairs behavioral performance in \*\*\* test (specify) and results in atrophy in \*\*\* but not in \*\*\* (specify) in \*\*\* mice (specify), suggesting that ovarian hormones in \*\*\* mice (specify) protect against hippocampal- dependent cognitive impairment and dorsal hippocampal atrophy. Likewise, in line 33: Deletion of ERbeta in astrocytes but not neurons by \*\*\* technology (specify) in \*\*\* mice (specify) result in \*\*\* (specify), as measured by \*\*\* (specify), suggesting that the protective effects of estrogen in midlife female mice are mediated by ERbeta receptors in astrocytes.  
+
+<|ref|>text<|/ref|><|det|>[[118, 685, 855, 742]]<|/det|>
+Line 52 'during health' is unnecessary. The study does not need to touch on the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy. Also, the Marxian sentence (from Groucho) 'A better understanding of the effect of aging can provide insights into the effect of aging' should be rephrased for clarity.  
+
+<|ref|>text<|/ref|><|det|>[[118, 756, 875, 811]]<|/det|>
+Line 59. It appears from the sentence that the impact of timing and estrogen type (does it mean receptor type?) has been already clarified according to the literature (refs 6- 9), in contradiction with the problem statement, and with line 88 (refs 8, 28). Please, succinctly specify what is known and what are the areas of conflict.  
+
+<|ref|>text<|/ref|><|det|>[[118, 825, 872, 881]]<|/det|>
+Line 62. This sentence belongs to a review, there is no need to defend the need for separation of sexes in a study using, moreover, a reference from 2012. Likewise, the rest of the paragraphs can be summarized, or moved to the Discussion. For example, it should be highlighted in the Discussion that the mice mirror the sex- dependent evolution of cognition and atrophy in humans.  
+
+<|ref|>text<|/ref|><|det|>[[115, 895, 833, 910]]<|/det|>
+Line 91. Again, the study does not inform about brain region- specific therapeutic approaches.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 83, 855, 126]]<|/det|>
+What it does is to unravel which brain regions are more affected by the experimental manipulations; nor the authors address the regional heterogeneity of astrocytes or neurons, for their genetic manipulations are global.  
+
+<|ref|>text<|/ref|><|det|>[[118, 140, 880, 183]]<|/det|>
+Lines 212 and 254. These two paragraphs can be condensed into one. Also, the explanation of the contextual fear conditioning (line 256) should be moved to earlier in the text, when the meaning of the different tests is explained.  
+
+<|ref|>text<|/ref|><|det|>[[118, 196, 847, 224]]<|/det|>
+In Discussion, the clinical implications of the findings can be summarized. Text condensation is recommended.  
+
+<|ref|>text<|/ref|><|det|>[[118, 239, 870, 266]]<|/det|>
+As most articles, including those written by English- speaking authors, the manuscript may benefit from professional copy editing.  
+
+<|ref|>text<|/ref|><|det|>[[119, 321, 415, 335]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 350, 875, 530]]<|/det|>
+In this study, Itoh, Meyer, et al. studied the effect of sexual hormone loss in age- related decline of memory, and the involvement of Estrogen Receptor \(\beta\) (ER \(\beta\) ) in mediating the aggravated hippocampal neuropathy observed in female mice following gonadectomy. The work combines memory assessments, MRI imaging of structures important for memory formation, and immunohistochemistry. The paper is clear, and the methodology thorough, with appropriate conclusions when describing the result. However, I believe the claims to be overstated regarding the potential of astrocytic ER \(\beta\) as neuroprotector, based solely on the presented data, without taking into account the existing literature. The work nicely shows that the deletion of ER \(\beta\) is detrimental for cognition from midlife, but in my opinion, the authors have not thoroughly demonstrated the beneficial effect of ER \(\beta\) activation on cognitive aging. As I elaborate below, I would also advise the authors to highlight the novelty of their finding, and finally to be cautious regarding the interpretation of their research, done in the context of normal aging but not neurodegeneration, as stated several times throughout the text.  
+
+<|ref|>text<|/ref|><|det|>[[118, 530, 870, 600]]<|/det|>
+Title: While I understand the underlying idea, i.e. that ER \(\beta\) in astrocytes could be an important candidate for future therapeutic targets in cognitive aging in female. I tend to find it overclaiming. The authors do not demonstrate that ER \(\beta\) solely mediates the potential improvement of memorisation in aging, for example using a specific drug targeting those receptors, or losing a positive effect after the deletion.  
+
+<|ref|>text<|/ref|><|det|>[[118, 614, 875, 910]]<|/det|>
+1. Behaviour: I was very surprised to read that the authors did not find an age-related defect of spatial memory in the Morris Water Maze (MWM) experiment (Fig1). It is a well-established fact that hippocampus-dependent types of memory, such as spatial, reference, or episodic memory, are sensitive to aging across species (please see the body of work from Catherine Barnes, Howard Eichenbaum, Michela Gallagher, Michael Yassa, Aline Marighetto, Sara Burke, to name a few). In trying to understand this lack of defects, I came to think that the paper would benefit not only from showing the performance during acquisition, at least in the supplementary material, as a control of learning, but also from providing more details about the paradigm. Indeed, depending on the protocol, the memory can go from spatial to procedural (i.e., hippocampus-independent), which remains relatively intact during aging. Alternatively, the authors could also represent the target quadrant exploration during the test as blocks of 15 sec to clarify the mouse behaviour. That being said, I think the high performance from aged mice is due to the fact that these authors tested the mice 2h after the last training session. This could explain the difference between retrieval in the MWM and the fear conditioning, which was tested 24h post-conditioning, and did reveal the age-related memory decline. I would suggest to the authors to homogenise the conditions of behavioural testing throughout the paper, for clarity of understanding and comparison, or at least to address these discrepancies in the discussion. As the age-related decline of hippocampus-dependent memory is such a long-standing fact, I don't think the study benefits from starting with such a clashing result and could deter the readership. Especially since in GDX mice, the defect appears earlier, at "mid-life", and could be interpreted as early aging. Finally, regarding the behaviour part, I would like to draw the authors' attention to the fact that
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 83, 874, 168]]<|/det|>
+the dichotomy in the role of the hippocampus to sustain recent but not remote memory has been revised in recent years (Goshen et al., Cell 2011) benefiting from the advancement of optogenetics. While I agree with the authors' conclusion regarding their own result, I would advice to modify the introduction of this part to take into account recent literature rather than Frankland and Bontempi's early work. A minor remark on this part, I am sure what is the authors' rationale for making this data as a supplementary figure only, when this is an interesting result.  
+
+<|ref|>text<|/ref|><|det|>[[118, 181, 875, 252]]<|/det|>
+2. ERβ cKO. I think it would be helpful to provide more details regarding the origin of the mice and selectivity of the conditional KO. Can you provide a proof of the specificity of the deletion within the astrocyte or the neurons? Could there be a compensation of the Era, that would underlie the result and hinder the interpretation regarding the role of ERβ? Providing an image illustrating the specific deletion in both cell type is, in my opinion, a major control to ensure the validity of study.  
+
+<|ref|>text<|/ref|><|det|>[[118, 265, 878, 448]]<|/det|>
+3. Neurodegeneration: One major point that I would like to draw the authors' attention towards, is their multiple reference to neurodegeneration throughout the paper, and the lack of distinction between normal aging and pathological aging. From what I understand, the authors seem to suggest that their study addresses neurodegeneration. However, the study has been conducting in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects), as demonstrated in my previous point, these wildtype aged mice can demonstrate high performance in certain conditions (Fig1), which would not be the case in pathological aging. In normal aging, the memory defects are associated with atrophy, but not neuronal loss, unlike pathological aging Moreover, the underlying mechanisms, and progression of alterations are widely distinct according to the type of aging. Therefore, I do not think it is proper in their condition to refer to neurodegeneration and would reframe the paper as normal aging. They could address the potential importance of their work regarding pathological aging in the discussion, but the use of wildtype mice do not support the direct extrapolation towards neurodegeneration.  
+
+<|ref|>text<|/ref|><|det|>[[118, 461, 875, 643]]<|/det|>
+4. Discussion: I think the work would benefit from stressing out throughout the text and in particular the discussion, the novelty of the research, and replacing the work within the literature. It is my belief that aside from the very interesting distinctive role of ERβ in astrocytes vs. neurons, the rest of the work has in some form been already reported. Age-related memory deficits (see authors in part 1 for reference); There is also a large body of work on the sexual dimorphism in cognitive aging (e.g. Frick et al., 2000; for review, Frick et al 2008), and on the role of estrogens in the brain and learning and memory, and effect of ovariectomy, which deserve to be cited in my opinion (L. Galea, K. Frick, E. Waters, M. Adams, C. Wolley, B. McEwen - non-exhaustive list). Similarly, the atrophy of the hippocampus has been demonstrated multiple times (for review Barnes and Burke, 2010). Finally, the reduction in spine density following ovariectomy has been previously reported by the McEwen laboratory (Gould et al., 1990). I do not intend to denigrate the study; I simply believe that the authors would benefit from valorising their findings in light of the existing literature.  
+
+<|ref|>text<|/ref|><|det|>[[118, 644, 861, 714]]<|/det|>
+Final point on the discussion, I fail to understand why the authors address hormone replacement and MS in such length, when their study does not use estrogen supplementation, potential effect on alteration similar to MS, and that MS have limited cognitive effect. Rather, I would be very interested in reading the authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes.  
+
+<|ref|>text<|/ref|><|det|>[[118, 728, 874, 756]]<|/det|>
+Minor comment: I think the reader would benefit from the homogenisation of the statistical report, which at the moment is alternatively in the legend or the text.  
+
+<|ref|>text<|/ref|><|det|>[[120, 770, 444, 784]]<|/det|>
+Looking forward to reading your response,  
+
+<|ref|>text<|/ref|><|det|>[[118, 798, 203, 811]]<|/det|>
+AS AI Abed  
+
+<|ref|>text<|/ref|><|det|>[[120, 854, 415, 868]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 883, 822, 911]]<|/det|>
+This article provides original and interesting data regarding the neuroprotective role of the astroglial estrogen receptor (ER) beta against hippocampal- dependent cognitive deficits and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 83, 875, 238]]<|/det|>
+atrophy in menopausal women. The authors combine molecular approach (constitutive conditional knockout mice) with in vivo MRI and behavioral testing to assess the role of astroglial ER beta in protecting against brain atrophy (in vivo MRI) and cognitive deficits, assessed via spatial reference memory (Morris Water Maze), working memory (Y- maze) and contextual learning (contextual fear conditioning). The data are primarily correlative and descriptive. By using a conditional KO mice, the authors conclude that ER beta in astrocytes could be a novel promising therapeutic target for cognitive deficits and hippocampal atrophy in menopausal women. This is of interest. Yet the conclusions are not well supported by the experimental data mostly because of the molecular approach, which is not appropriate. In addition, the manuscript lacks mechanistic insights. Thus although interesting, this work is preliminary in its present form and presents several major issues.  
+
+<|ref|>text<|/ref|><|det|>[[118, 252, 860, 293]]<|/det|>
+1) The authors use GFAP-cre mice to constitutively delete ER beta in astrocytes. These mice cannot be used to conclude about implication of astroglial proteins, when these proteins are also expressed in neurons.  
+
+<|ref|>text<|/ref|><|det|>[[118, 293, 866, 364]]<|/det|>
+There are indeed major caveats in using the GFAP-Cre mice, as it has been well documented by numerous groups that deletion occurs in both neurons and astrocytes, since GFAP is expressed in precursor cells during early development. This confounds the authors major conclusion that deletion of ER beta from astrocytes contributes to cognitive deficits and brain atrophy in menopausal female.  
+
+<|ref|>text<|/ref|><|det|>[[118, 364, 696, 406]]<|/det|>
+A few papers here listed well describe these caveats : "Looks Can Be Deceiving: Reconsidering the Evidence for Gliotransmission" https://doi.org/10.1016/j.neuron.2014.12.003  
+
+<|ref|>text<|/ref|><|det|>[[118, 420, 828, 448]]<|/det|>
+Germ- Line Recombination Activity of the Widely Used hGFAP- Cre and Nestin- Cre Transgenes https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3857304/  
+
+<|ref|>text<|/ref|><|det|>[[118, 463, 595, 491]]<|/det|>
+Expression Specificity of GFAP Transgenes https://link.springer.com/article/10.1007/s11064- 004- 6881- 1  
+
+<|ref|>text<|/ref|><|det|>[[118, 504, 850, 546]]<|/det|>
+Other experiments should thus be combined with the use of these mice, such as the use of inducible cKO mice and/or rescue experiments targeted to astrocytes (re- expression of ER beta specifically in astrocytes in the astroglial cKO mice).  
+
+<|ref|>text<|/ref|><|det|>[[118, 560, 850, 602]]<|/det|>
+2) There is no validation of these newly generated astroglial- and neuronal-specific ER beta transgenic models. One would like to see quantification of ER beta expression in astrocytes and neurons in both cKO mice (gfap-cre:ER betafl/fl and nestin-cre:ER betafl/fl).  
+
+<|ref|>text<|/ref|><|det|>[[118, 616, 820, 658]]<|/det|>
+3) There is a lack of details regarding the mice (origin (stock number or laboratory source), breeding strategy controlling germline recombination, appropriate littermate controls. This precludes evaluation of whether the authors have used appropriate controls.  
+
+<|ref|>text<|/ref|><|det|>[[118, 672, 870, 714]]<|/det|>
+4) The authors use GFAP immunostaining to assess astroglial reactivity. GFAP expression can vary independently of astroglial reactivity. The authors should thus use additional markers (such as vimentin, stat3, GS...) and analyse astrocyte morphology to conclude about astrocyte reactivity.  
+
+<|ref|>text<|/ref|><|det|>[[118, 728, 870, 784]]<|/det|>
+5) The authors report synapse loss in menopausal female. They however only performed PSD95 staining to assess synapses. A synapse is composed of a pre- and a postsynaptic element and can only be identified by the colocalization of both elements. Staining for both pre- and postsynaptic markers should thus be performed to quantify synapse number.  
+
+<|ref|>text<|/ref|><|det|>[[118, 799, 874, 840]]<|/det|>
+6) This manuscript completely lacks mechanistic insights. How ER beta in astrocytes contributes to brain structure and cognitive performance in male and female? Does this require a crosstalk with other cell types such as microglia?  
+
+<|ref|>text<|/ref|><|det|>[[120, 870, 415, 884]]<|/det|>
+Reviewer #4 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 898, 844, 912]]<|/det|>
+The paper "Mind the Gap: Estrogen receptor beta (ERβ) in astrocytes is a therapeutic target to
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 868, 153]]<|/det|>
+prevent cognitive problems at menopause" investigates the impact of estrogen, and more specifically estrogen receptors beta, to the observed gender differences in behavior, hippocampus volume and astrogliosis in rodents at midlife. The authors use a combination of techniques like MRI, histology, behavior on wild type animals, animals which underwent gonadectomy/sham and ERb knock- out animals.  
+
+<|ref|>text<|/ref|><|det|>[[118, 168, 870, 239]]<|/det|>
+The narrative is very interesting, the results important and well contextualized. The work is extremely significant and timely. The methodology is rigorous, although a few important pieces of evidence are missing to have a more balanced picture and close the story (detailed below). I also have doubts on some of the statistical approaches chosen. The paper is easy to follow and very well written.  
+
+<|ref|>text<|/ref|><|det|>[[119, 252, 660, 267]]<|/det|>
+There are a few aspects that can be improved. Below, my suggestions.  
+
+<|ref|>text<|/ref|><|det|>[[118, 281, 870, 351]]<|/det|>
+1 Fig.1 : the information contained in a- b and c- d is redundant. In addition, doing twice a scatter plot, which distributes points random, on the same data generates plots which are visually different, which is awkward. The same applies to Fig. 2 a- b and c- d, and Fig. 4 a- b. I think half of the plots can safely go to supplementary.  
+
+<|ref|>text<|/ref|><|det|>[[118, 365, 880, 435]]<|/det|>
+2 Following the previous comment, I think the most correct statistical approach would be a single ANOVA for the water maze data where sex is a factor (in addition to age and quadrant), followed by appropriate post- hoc comparisons. The same applies for MRI volumetric data, where ROI should be another factor (cortex, striatum, HC).  
+
+<|ref|>text<|/ref|><|det|>[[118, 449, 880, 547]]<|/det|>
+3 In Fig 1 h, it seems that there is no volume change in HC in male mice across the lifespan. This results somehow weakens the point made by the authors in the rest of the paper, about the importance for neuroprotection across the lifespan of the ERb astrocyte receptors specifically in the dorsal hippocampus, unless the dHC is only important in ageing female, which is hard to digest, unless supported by evidence. Generally speaking, the paper is missing showing and discussing in males at least some of the analyses performed in females, for example Fig 2 g- i.  
+
+<|ref|>text<|/ref|><|det|>[[118, 561, 860, 603]]<|/det|>
+4 Fig 1, supplementary excel table: why using Welch test? Is it because the samples have unequal variance? Please specify in the methods.  
+
+<|ref|>text<|/ref|><|det|>[[118, 616, 880, 700]]<|/det|>
+5 Fig 1: It is noteworthy that from Fig. 1 there is no relation between structure and function, at least in the context of ageing. If it is a matter of different sensitivities of the two techniques employed (MRI volumetry vs behaviour), as hinted by the authors, I suggest they discuss the issue and report the power analysis used for determining the sample size. Perhaps the behavior is underpowered?  
+
+<|ref|>text<|/ref|><|det|>[[118, 714, 825, 771]]<|/det|>
+6 Fig.2: Similarly to comments 1 and 2, the figures of the water maze are redundant, and the statistical model could be just one for each test but include everything (gender, therapy, life stage).  
+
+<|ref|>text<|/ref|><|det|>[[118, 785, 520, 814]]<|/det|>
+7 Fig. 2 f: why freezing was only measured at midlife?  
+
+<|ref|>text<|/ref|><|det|>[[118, 828, 880, 898]]<|/det|>
+8 Fig. 2 j- k: I suggest to only plot regression line when significant. Also, there seems to be a lot of points in the plots. Are all animals (both sexes, all ages, GDX/sham) pulled together? If so, which is the rationale? Perhaps the different groups could be shown in the plot by coloring the datapoints differently.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 130, 96]]<|/det|>
+9  
+
+<|ref|>text<|/ref|><|det|>[[118, 98, 879, 169]]<|/det|>
+Fig. 3, Neuropathology: the authors talk about "area fraction" but what is exactly measured and how? Can distinct aspect of the morphology be measured (ramifications density, cell body size, cell number etc.? This is important as the term "glia activation" can refer to distinct morphological changes happening to the cell. In general, the method- >histology section is missing details needed for the reader to be able to reproduce the analyses.  
+
+<|ref|>text<|/ref|><|det|>[[118, 183, 140, 195]]<|/det|>
+10  
+
+<|ref|>text<|/ref|><|det|>[[118, 196, 856, 239]]<|/det|>
+Line 194: the notation is misleading. Perhaps by "direct" and "indirect" correlations the authors mean linear positive and linear negative? Or do the authors really mean indirect as mediated by something else? If so, specify.  
+
+<|ref|>text<|/ref|><|det|>[[118, 253, 140, 265]]<|/det|>
+11  
+
+<|ref|>text<|/ref|><|det|>[[118, 266, 866, 295]]<|/det|>
+Fig 3 is missing the plots of the correlations between time in TQ and gliosis (data is discussed but not shown).  
+
+<|ref|>text<|/ref|><|det|>[[118, 309, 140, 321]]<|/det|>
+12  
+
+<|ref|>text<|/ref|><|det|>[[118, 322, 861, 351]]<|/det|>
+Following the rationale of the paper, Fig 4 is missing the neuropathology. How are the astrocytes affected by ERb deletion? Are they more activated at midlife?  
+
+<|ref|>text<|/ref|><|det|>[[118, 365, 140, 377]]<|/det|>
+13  
+
+<|ref|>text<|/ref|><|det|>[[118, 378, 875, 435]]<|/det|>
+Following the previous comment but more generally, what is ultimately the mechanism that the authors identified as leading to atrophy and worst cognition in ageing females? Is it astrogliosis? If so, the authors should report more neuropathological data (or other equivalent astrogliosis markers) to make a stronger point.  
+
+<|ref|>text<|/ref|><|det|>[[118, 449, 140, 461]]<|/det|>
+14  
+
+<|ref|>text<|/ref|><|det|>[[118, 462, 840, 491]]<|/det|>
+It is interesting that there is no cognitive effect in young female mice which lacks endogenous hormones or receptors. What about HC volume?  
+
+<|ref|>text<|/ref|><|det|>[[118, 505, 140, 517]]<|/det|>
+15  
+
+<|ref|>text<|/ref|><|det|>[[118, 519, 872, 561]]<|/det|>
+As the authors point out in the discussion, microglia also have ERb. Can the author show and compare some histology of microglia in animals lacking endogenous receptors and animals lacking ERb in astrocytes?
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[174, 91, 355, 107]]<|/det|>
+## Reviewers' comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 122, 419, 139]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 153, 237, 170]]<|/det|>
+### A. Key findings:  
+
+<|ref|>text<|/ref|><|det|>[[115, 185, 881, 298]]<|/det|>
+The study may inform structural analyses in HRT, as well as therapeutic strategies, with a focus on the emerging avenue of astrocyte- targeted therapies. The authors are to be commended by the hard work and the wealth of tests and animal models. Tests include hippocampal- dependent spatial memory (Morris Water Maze, MWM), working memory (Y- maze) and contextual fear conditioning. Of great value is the use of MRI to dissect out regional changes in association with behavioral deficits. Mouse models include intact and gonadectomized, as well as knock- out mice for estrogen receptor types in astrocytes and neurons.  
+
+<|ref|>text<|/ref|><|det|>[[170, 313, 824, 331]]<|/det|>
+We appreciate these comments recognizing the breadth of our experiments.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 345, 230, 362]]<|/det|>
+### B. Limitations:  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 377, 295, 394]]<|/det|>
+## B.1. Insufficient clarity.  
+
+<|ref|>text<|/ref|><|det|>[[115, 409, 880, 570]]<|/det|>
+Conceptual: The gap of knowledge ('understanding brain region- specific, cell- specific and receptor- specific mechanisms') that the authors aim to dissipate should be rephrased. As I understand it, the authors aim to clarify why HRT does not mitigate cognitive deficits and brain atrophy in menopausal women (problem statement) and propose that the conflicting results may arise from paradoxical actions of different estrogen receptors in different cell types, and the fact that readout measures are not directed to the right brain areas. There is a confusion in the said gap between actors (cells and receptors) and effects (brain areas and cognition), for, in the study, brain regions are not specifically targeted. The gene knockout affects cells all over the brain, and clinical therapies are systemically administered. The repeated expression 'region- specific targeting' throughout the text should be eliminated or clarified.  
+
+<|ref|>text<|/ref|><|det|>[[114, 584, 876, 795]]<|/det|>
+As suggested by the reviewer, we have now removed "gap" from the title as well as throughout the revised manuscript. Also, our phrasing, "Region- specific targeting", warrants clarification. Here, we show functional, pathology, and MRI effects that support future treatments targeting a specific cell and brain region (astrocytes in dorsal hippocampus). Future therapeutics could target this cell and region in midlife females. This does not limit treatments to those affecting only astrocytes in dorsal hippocampus (no other cells and no other regions). That is a pharmacologic bar too high for existing technologies. However, future clinical trials could tailor estrogen type, inclusion criteria, and outcome measures (specific cognitive domains, brain substructure atrophy, PET imaging, pathology of post- mortem or biopsy tissues, etc.) to assess whether a systemic treatment has an effect on astrocytes in dorsal hippocampus in midlife females. This manuscript identifies astrocytes in dorsal hippocampus as such a candidate cell and region.  
+
+<|ref|>text<|/ref|><|det|>[[115, 809, 868, 896]]<|/det|>
+Experimental: Perhaps it is just me, but the study rationale is not clear, particularly which tests are used, and which ages analyzed. Specifically, in figure 1, in intact mice only MWM is performed at three ages, with negative results, while in Fig 2 (gonadectomized mice) Y Maze and contextual fear conditioning are incorporated, but only in two ages. The comparison of young and midlife mice allows to test the interaction between age and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 880, 229]]<|/det|>
+absence of hormones, but the absence of old mice makes it difficult to conclude whether gonadectomy at midlife accelerates cognitive and structural decline to match that of old ages. Also, Fig. 2a and Fig. 2c repeat data. Perhaps Fig.1 and Fig. 2 should be condensed into a single figure to present the model of intact (normal scenario) and gonadectomized mice (extreme scenario) and conclude which behavioral and structural analyses are more apt to study the effect of hormones. Likewise, in Fig. 4 (studies in ER knockout mice) Y Maze is not used, only midlife mice are tested, and immunohistochemical analyses of GFAP, Iba1 and PSD95 are not performed.  
+
+<|ref|>text<|/ref|><|det|>[[114, 244, 840, 278]]<|/det|>
+See response to the Editor. Our experimental strategy was focused on midlife females with a set of informative comparators.  
+
+<|ref|>text<|/ref|><|det|>[[114, 277, 870, 340]]<|/det|>
+Combining former Fig. 1 and 2 (current Fig. 2 and 3) would not align with each question investigated in each figure. An assessment of sex differences is the focus of current Fig. 2, while investigation of midlife females \(+ / - \mathsf{GDX}\) is the focus of current Fig. 3.  
+
+<|ref|>text<|/ref|><|det|>[[114, 341, 872, 421]]<|/det|>
+As suggested by the reviewer, MWM data are now shown as a merged graph with both between group comparisons of \(\%\) Time in Target Quadrant and within group comparisons of \(\%\) Time in Target Quadrant compared to \(\%\) Time in Other Quadrants. Former Fig. 2a and 2c (new Fig. 3b and 3c) make different comparisons of midlife females, one to midlife males and the other to young females, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[114, 420, 875, 454]]<|/det|>
+ERbeta knock- outs are shown not only at midlife in former Fig. 4 (new Fig. 5), they are also shown at young age (see Suppl Fig. 1).  
+
+<|ref|>text<|/ref|><|det|>[[115, 454, 867, 503]]<|/det|>
+As suggested by the reviewer, additional immunohistochemical images and analyses for the ERbeta cKOs are now added to former Fig. 4 (new Fig. 5) of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 516, 881, 581]]<|/det|>
+Fig. 3 e,f. I guess that it is normal that \(\%\) time in other quadrants' is the mirror/inverse version of \(\%\) time in target quadrant'. If so, why is the scale in the Y axis of Fig. 3f different from the one in 3e? For example, a \(20\%\) time in target quadrant should correlate with \(80\%\) time in other quadrants.  
+
+<|ref|>text<|/ref|><|det|>[[115, 596, 875, 677]]<|/det|>
+We are not reporting \(\%\) time in the other quadrant, instead it is \(\%\) time in the other quadrants plural (an average of the \(\%\) time in the other 3 quadrants). Thus, \(\%\) time in other quadrants is not the mirror/inverse version of \(\%\) time in the target quadrant. For example, \(20\%\) time in TQ would align with \(80\%\) total time in the other three quadrants, with the average of the three other quadrants \(= 26.7\%\) .  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 692, 420, 708]]<|/det|>
+## B.2. Overinterpretation of correlations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 723, 878, 837]]<|/det|>
+Direct results should be differentiated from interpretations. For example, line 212 'Neuroprotection in females at midlife is mediated by ERbeta in astrocytes' is not a correct label. The experimental design will show the effect of ERbeta removal, but it does not demonstrate that the deleterious mechanisms are the same as those mediating damage caused by gonadectomy, which affects many hormones. Such an equivalence is a correlation. Reversal experiments such as addition of ERbeta agonists to gonadectomized mice may beautifully strengthen the connection the authors wish to make.  
+
+<|ref|>text<|/ref|><|det|>[[115, 858, 870, 908]]<|/det|>
+As suggested by the reviewer, to beautifully strengthen the connection, "reversal experiments such as addition of ERbeta agonists to gonadectomized mice", are now done (see new Fig. 7 of the revised manuscript).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 879, 154]]<|/det|>
+The statement, "Neuroprotection in females at midlife is mediated by ERbeta in astrocytes" is a fair based on our data. When selective deletion of a receptor (in the cKO) causes a deleterious effect, then the presence of that receptor prevents the deleterious effect (in the WT). Here, cKO and WT were females at midlife.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 170, 256, 186]]<|/det|>
+### C. Methodologies:  
+
+<|ref|>text<|/ref|><|det|>[[115, 201, 878, 250]]<|/det|>
+Supplemental Fig. 1. presents data from estrogen receptor alpha knockout mice that are not described in Methods. These data are important and could be moved to the main text. Likewise, the PCR results confirming successful deletion of estrogen receptor genes should be shown.  
+
+<|ref|>text<|/ref|><|det|>[[115, 265, 848, 378]]<|/det|>
+As suggested by the reviewer, confirmation of specific deletion of estrogen receptor beta is now shown (see new Fig. 4 of the revised manuscript). Supplemental data on ERalpha knock- out mice are now removed since repeating all outcomes in another knock- out would take focus away from the role ERbeta. Instead we provide additional data on ERbeta including 1) treatment with the ERbeta ligand to reverse hippocampal outcomes, and 2) RNA- sequencing and gene expression analysis of the astrocyte transcriptome in midlife females with deletion of ERbeta in astrocytes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 392, 777, 409]]<|/det|>
+It would be nice to have images of MRI in different experimental scenarios in Fig. 1.  
+
+<|ref|>text<|/ref|><|det|>[[115, 424, 881, 505]]<|/det|>
+Other than the substructure delineations overlaid onto the brain image in new Fig. 2 (former Fig. 1), it is unclear what image is being requested. Notably, MRI images from young to midlife to old mice will not appear to differ upon visual inspection. The standard approach required is quantification of neuroanatomical substructure volumes in 3D using atlas- based morphometry, which is what we did.  
+
+<|ref|>text<|/ref|><|det|>[[115, 520, 881, 602]]<|/det|>
+The quantification of immunohistochemical data is not explained in sufficient detail, including the number and distance between sections, stereological considerations, programs used to quantify, thresholding, background subtraction, etc. Also, images of PSD95 immunostaining, perhaps the extreme cases, would be useful to visually support one of the most interesting finding of the study.  
+
+<|ref|>text<|/ref|><|det|>[[115, 616, 870, 730]]<|/det|>
+As suggested by the reviewer, the revised manuscript now has detailed descriptions of immunohistochemical analyses, including antibody information and the analysis program used. Representative images of reactive astrocytes, microglia activation, disease- associated microglia, and synapses have each been added (see new Fig. 3 and Fig. 5 of the revised manuscript). Regarding synapses, we have now shown interesting immunostaining and analyses for SYN1 (including images) in the revised manuscript).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 743, 240, 759]]<|/det|>
+### D. Style issues:  
+
+<|ref|>text<|/ref|><|det|>[[115, 774, 632, 791]]<|/det|>
+The authors may consider removing 'Mind the Gap' from the title.  
+
+<|ref|>text<|/ref|><|det|>[[171, 806, 815, 824]]<|/det|>
+As suggested by the reviewer, the title is revised, removing 'Mind the Gap'.  
+
+<|ref|>text<|/ref|><|det|>[[115, 838, 840, 855]]<|/det|>
+The manuscript is well written, but there is always room for improvement. In the Abstract ...  
+
+<|ref|>text<|/ref|><|det|>[[174, 870, 391, 886]]<|/det|>
+The Abstract is rewritten.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 105, 874, 170]]<|/det|>
+Line 52 'during health' is unnecessary. The study does not need to touch on the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy. Also, the Marxian sentence (from Groucho) 'A better understanding of the effect of aging can provide insights into the effect of aging' should be rephrased for clarity.  
+
+<|ref|>text<|/ref|><|det|>[[115, 185, 875, 300]]<|/det|>
+We agree with the reviewer that our study does not need to touch on the controversy of healthy or unhealthy aging. Our use of the term disease means neurological disease (not aging in otherwise healthy). The full sentence is: "A better understanding of the effect of brain aging during health can provide insights into the effect of brain aging during disease". "During health" is necessary to contrast with "during disease". This refers two different aging scenarios, whereby mechanisms in one can inform the other.  
+
+<|ref|>text<|/ref|><|det|>[[115, 313, 875, 377]]<|/det|>
+Line 59. It appears from the sentence that the impact of timing and estrogen type (does it mean receptor type?) has been already clarified according to the literature (refs 6- 9), in contradiction with the problem statement, and with line 88 (refs 8, 28). Please, succinctly specify what is known and what are the areas of conflict.  
+
+<|ref|>text<|/ref|><|det|>[[115, 377, 880, 456]]<|/det|>
+Line 62. This sentence belongs to a review, there is no need to defend the need for separation of sexes in a study using, moreover, a reference from 2012. Likewise, the rest of the paragraphs can be summarized, or moved to the Discussion. For example, it should be highlighted in the Discussion that the mice mirror the sex- dependent evolution of cognition and atrophy in humans.  
+
+<|ref|>text<|/ref|><|det|>[[115, 457, 875, 506]]<|/det|>
+Lines 212 and 254. These two paragraphs can be condensed into one. Also, the explanation of the contextual fear conditioning (line 256) should be moved to earlier in the text, when the meaning of the different tests is explained.  
+
+<|ref|>sub_title<|/ref|><|det|>[[173, 521, 574, 538]]<|/det|>
+## Wording is changed in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 553, 870, 618]]<|/det|>
+Line 91. Again, the study does not inform about brain region- specific therapeutic approaches. What it does is to unravel which brain regions are more affected by the experimental manipulations; nor the authors address the regional heterogeneity of astrocytes or neurons, for their genetic manipulations are global.  
+
+<|ref|>text<|/ref|><|det|>[[173, 632, 723, 650]]<|/det|>
+See above regarding clarification of "Region- specific targeting".  
+
+<|ref|>text<|/ref|><|det|>[[115, 681, 870, 714]]<|/det|>
+In Discussion, the clinical implications of the findings can be summarized. Text condensation is recommended.  
+
+<|ref|>sub_title<|/ref|><|det|>[[173, 728, 631, 745]]<|/det|>
+## Condensation is now done in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 760, 833, 794]]<|/det|>
+As most articles, including those written by English- speaking authors, the manuscript may benefit from professional copy editing.  
+
+<|ref|>text<|/ref|><|det|>[[115, 808, 857, 842]]<|/det|>
+Thank you very much for these suggestions which have substantially improved the manuscript!  
+
+<|ref|>text<|/ref|><|det|>[[115, 873, 419, 890]]<|/det|>
+Reviewer #2 (Remarks to the Author):
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 105, 880, 187]]<|/det|>
+Title: While I understand the underlying idea, i.e. that ERβ in astrocytes could be an important candidate for future therapeutic targets in cognitive aging in female. I tend to find it overclaiming. The authors do not demonstrate that ERβ solely mediates the potential improvement of memorisation in aging, for example using a specific drug targeting those receptors, or losing a positive effect after the deletion.  
+
+<|ref|>text<|/ref|><|det|>[[115, 202, 842, 250]]<|/det|>
+As suggested by the reviewer, we have now added the use of "a specific drug targeting those receptors", namely ER beta ligand treatment (see new Fig. 7 of the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 251, 877, 315]]<|/det|>
+Results using the astrocyte- specific knock- out show that ER beta in astrocytes is necessary to prevent deleterious effects on cognitive outcomes at midlife in females. We do not claim that "ERβ solely mediates the potential improvement ...". Our finding is not mutually exclusive of additional potentially protective effects of other molecules or cells.  
+
+<|ref|>text<|/ref|><|det|>[[112, 328, 881, 780]]<|/det|>
+1. Behaviour: I was very surprised to read that the authors did not find an age-related defect of spatial memory in the Morris Water Maze (MWM) experiment (Fig1). It is a well-established fact that hippocampus-dependent types of memory, such as spatial, reference, or episodic memory, are sensitive to aging across species (please see the body of work from Catherine Barnes, Howard Eichenbaum, Michela Gallagher, Michael Yassa, Aline Marighetto, Sara Burke, to name a few). In trying to understand this lack of defects, I came to think that the paper would benefit not only from showing the performance during acquisition, at least in the supplementary material, as a control of learning, but also from providing more details about the paradigm. Indeed, depending on the protocol, the memory can go from spatial to procedural (i.e., hippocampus-independent), which remains relatively intact during aging. Alternatively, the authors could also represent the target quadrant exploration during the test as blocks of 15 sec to clarify the mouse behaviour. That being said, I think the high performance from aged mice is due to the fact that the authors tested the mice 2h after the last training session. This could explain the difference between retrieval in the MWM and the fear conditioning, which was tested 24h post-conditioning, and did reveal the age-related memory decline. I would suggest to the authors to homogenise the conditions of behavioural testing throughout the paper, for clarity of understanding and comparison, or at least to address these discrepancies in the discussion. As the age-related decline of hippocampus-dependent memory is such a long-standing fact, I don't think the study benefits from starting with such a clashing result and could deter the readership. Especially since in GDX mice, the defect appears earlier, at "mid-life", and could be interpreted as early aging. Finally, regarding the behaviour part, I would like to draw the authors' attention to the fact that the dichotomy in the role of the hippocampus to sustain recent but not remote memory has been revised in recent years (Goshen et al., Cell 2011) benefiting from the advancement of optogenetics. While I agree with the authors' conclusion regarding their own result, I would advice to modify the introduction of this part to take into account recent literature rather than Frankland and Bontempi's early work. A minor remark on this part, I am sure what is the authors' rationale for making this data as a supplementary figure only, when this is an interesting result.  
+
+<|ref|>text<|/ref|><|det|>[[115, 793, 882, 907]]<|/det|>
+We agree with these insightful comments of the reviewer. Various protocol paradigms (timing, etc.) can affect behavioral performance. Here, using the MWM protocol as described, with the sample size as indicated, we did not observe a significant defect in spatial reference memory in the Morris Water Maze in the 2hr delay probe trial in gonadally intact, healthy females or males up to the ages tested in the environment of the UCLA Behavioral Core Facility. This lack of a significant deficit has been seen before in healthy control groups of studies focusing on AD mice where healthy mice served as
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 869, 203]]<|/det|>
+the cognitively normal control, albeit often tested at ages a few months earlier. In agreement with the reviewer, we observed that older ages had a learning speed difference (older slower than younger), so a training learning graph could be added as a supplemental figure upon request. However, our focus here is not to do additional behavioral tests to show significant deficits in gonadally intact mice using a given paradigm. Instead, our focus is on the significant worsening observed in GDX females compared to gonadally intact females using a standard MWM protocol.  
+
+<|ref|>text<|/ref|><|det|>[[115, 217, 860, 251]]<|/det|>
+2. ERβ cKO. I think it would be helpful to provide more details regarding the origin of the mice and selectivity of the conditional KO.  
+
+<|ref|>text<|/ref|><|det|>[[115, 265, 835, 300]]<|/det|>
+As suggested by the reviewer, this is now done (see new Fig. 4 in the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[114, 313, 882, 539]]<|/det|>
+3. Neurodegeneration: One major point that I would like to draw the authors' attention towards, is their multiple reference to neurodegeneration throughout the paper, and the lack of distinction between normal aging and pathological aging. From what I understand, the authors seem to suggest that their study addresses neurodegeneration. However, the study has been conducting in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects), as demonstrated in my previous point, these wildtype aged mice can demonstrate high performance in certain conditions (Fig1), which would not be the case in pathological aging. In normal aging, the memory defects are associated with atrophy, but not neuronal loss, unlike pathological aging Moreover, the underlying mechanisms, and progression of alterations are widely distinct according to the type of aging. Therefore, I do not think it is proper in their condition to refer to neurodegeneration and would reframe the paper as normal aging. They could address the potential importance of their work regarding pathological aging in the discussion, but the use of wildtype mice do not support the direct extrapolation towards neurodegeneration.  
+
+<|ref|>text<|/ref|><|det|>[[114, 553, 881, 649]]<|/det|>
+It is unclear what is meant by the need to "reframe the paper as normal aging". As the reviewer points out: 1) Our study was "conducted in C57Bl/6J mice, which do not develop pathological aging (i.e., dementia, Alzheimer's-like memory defects)", and 2) "these wildtype aged mice can demonstrate high performance in certain conditions". Our manuscript is framed on the mice we used, those with "normal aging", and these two correct statements by the reviewer show that we have conveyed that.  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 878, 700]]<|/det|>
+We did not make a "distinction between normal aging and pathological aging". As stated by Rev #1, we are not entering "the controversy of whether there is healthy or unhealthy aging, or aging is always unhealthy",  
+
+<|ref|>text<|/ref|><|det|>[[115, 699, 858, 763]]<|/det|>
+Use of the word "neurodegeneration" is not problematic when describing a condition characterized by cognitive decline, glial activation, synaptic loss, and brain atrophy, indeed what occurs in midlife females either GDX or with selective deletion of ER beta in astrocytes.  
+
+<|ref|>text<|/ref|><|det|>[[114, 778, 870, 907]]<|/det|>
+4. Discussion: I think the work would benefit from stressing out throughout the text and in particular the discussion, the novelty of the research, and replacing the work within the literature. It is my belief that aside from the very interesting distinctive role of ERβ in astrocytes vs. neurons, the rest of the work has in some form been already reported. Age-related memory deficits (see authors in part 1 for reference); There is also a large body of work on the sexual dimorphism in cognitive aging (e.g. Frick et al., 2000; for review, Frick et al 2008), and on the role of estrogens in the brain and learning and memory, and effect of ovariectomy, which deserve to be cited in my opinion (L. Galea, K. Frick, E. Waters, M. Adams, C. Wolley, B.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 881, 170]]<|/det|>
+McEwen – non- exhaustive list). Similarly, the atrophy of the hippocampus has been demonstrated multiple times (for review Barnes and Burke, 2010). Finally, the reduction in spine density following ovariectomy has been previously reported by the McEwen laboratory (Gould et al., 1990). I do not intend to denigrate the study; I simply believe that the authors would benefit from valorising their findings in light of the existing literature.  
+
+<|ref|>text<|/ref|><|det|>[[114, 170, 867, 267]]<|/det|>
+Final point on the discussion, I fail to understand why the authors address hormone replacement and MS in such length, when their study does not use estrogen supplementation, potential effect on alteration similar to MS, and that MS have limited cognitive effect. Rather, I would be very interested in reading the authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes.  
+
+<|ref|>text<|/ref|><|det|>[[114, 281, 880, 394]]<|/det|>
+The revised manuscript now includes estrogen supplementation (see new Fig. 7). We thank the reviewer for highlighting our finding of a "very interesting distinctive role of ERβ in astrocytes vs. neurons". We have now further added additional novel findings on gene expression analyses of astrocyte transcriptomes in midlife females with ER beta selectively deleted. This revealed an exciting link to previous work by others on the role of glucose metabolism in brain during menopause, albeit previously not studied in hippocampal astrocytes.  
+
+<|ref|>text<|/ref|><|det|>[[114, 393, 877, 569]]<|/det|>
+The literature cited by this knowledgeable reviewer is excellent and could be the basis for a stand- alone review article. We agree with the reviewer that some of our work has been done in parts in papers by others. The strength here is integration of a breadth of complementary studies (some done before, some not) in the same paper, namely behavioral studies, in vivo MRI for regional atrophy, underlying neuropathology, cell- specific knock outs, RNA- sequencing, and estrogen receptor specific treatment reversal). Rather than reviewing parts published in the past, we took our discussion in a future direction. We discussed implications of the totality of our findings on clinical translation. This addresses the unmet need for a hormone replacement therapy designed more specifically and based on neuroscience mechanisms underlying cognitive outcomes in midlife females.  
+
+<|ref|>text<|/ref|><|det|>[[114, 569, 880, 666]]<|/det|>
+Regarding justification of the discussion of MS, these patients are predominantly female, have substantial cognitive disability, and menopause confers overall disability worsening. In addition, clinical trials testing a naturally occurring ER beta ligand (estriol) treatment showed neuroprotective effects in women with MS. This published work warrants discussion of the potential for repurposing this treatment approach to cognitive deficits in healthy women with menopause.  
+
+<|ref|>text<|/ref|><|det|>[[114, 666, 874, 825]]<|/det|>
+We thank the reviewer very much for their interest in getting the "authors' take on the integration of their findings about sexual dimorphism, from cognition to synaptic loss and the specific role of glial cells in these processes". Dr. Voskuhl is an expert on the study of sex differences, and this was the starting point of this manuscript (new Fig. 1 and 2). However, the trajectory of this manuscript is to use initial sex differences findings as a basis for subsequent research focused on gaining cellular and molecular insights relevant to midlife females with cognitive deficits. A discussion of sexual dimorphism as described by the reviewer would align better with a manuscript that studied both females and males with the full breadth of our outcomes from young to midlife to old ages. While that is interesting, it would be a different paper.  
+
+<|ref|>text<|/ref|><|det|>[[114, 841, 830, 873]]<|/det|>
+Minor comment: I think the reader would benefit from the homogenisation of the statistical report, which at the moment is alternatively in the legend or the text.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 868, 139]]<|/det|>
+As the reviewer suggested, this is now done in the revised manuscript. We also have a statistical summary ready for submission pending the response to this appeal to proceed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 154, 454, 186]]<|/det|>
+Looking forward to reading your response, AS Al Abed  
+
+<|ref|>text<|/ref|><|det|>[[173, 201, 524, 218]]<|/det|>
+Thank you for your insightful comments.  
+
+<|ref|>text<|/ref|><|det|>[[115, 249, 420, 265]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 281, 860, 328]]<|/det|>
+1) The authors use GFAP-cre mice to constitutively delete ER beta in astrocytes. These mice cannot be used to conclude about implication of astroglial proteins, when these proteins are also expressed in neurons.  
+
+<|ref|>text<|/ref|><|det|>[[115, 329, 863, 410]]<|/det|>
+There are indeed major caveats in using the GFAP-Cre mice, as it has been well documented by numerous groups that deletion occurs in both neurons and astrocytes, since GFAP is expressed in precursor cells during early development. This confounds the authors major conclusion that deletion of ER beta from astrocytes contributes to cognitive deficits and brain atrophy in menopausal female.  
+
+<|ref|>text<|/ref|><|det|>[[115, 426, 725, 474]]<|/det|>
+A few papers here listed well describe these caveats : "Looks Can Be Deceiving: Reconsidering the Evidence for Gliotransmission" https://doi.org/10.1016/j.neuron.2014.12.003  
+
+<|ref|>text<|/ref|><|det|>[[115, 490, 870, 522]]<|/det|>
+Germ- Line Recombination Activity of the Widely Used hGFAP- Cre and Nestin- Cre Transgenes https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3857304/  
+
+<|ref|>text<|/ref|><|det|>[[115, 538, 591, 570]]<|/det|>
+Expression Specificity of GFAP Transgenes https://link.springer.com/article/10.1007/s11064- 004- 6881- 1  
+
+<|ref|>text<|/ref|><|det|>[[115, 586, 875, 634]]<|/det|>
+Other experiments should thus be combined with the use of these mice, such as the use of inducible cKO mice and/or rescue experiments targeted to astrocytes (re- expression of ER beta specifically in astrocytes in the astroglial cKO mice).  
+
+<|ref|>text<|/ref|><|det|>[[115, 650, 866, 714]]<|/det|>
+There are different subtypes of mGFAP- Cre lines. The mGFAP- Cre line 77.6 used in this manuscript has now been characterized, and specificity for astrocytes and not other cell types (including neurons) was shown (see new Fig. 4 in the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 715, 861, 762]]<|/det|>
+Recent publications used the mGFAP- Cre 77.6 line for selective gene deletion in astrocytes: 1) Deerhake ME et al, Immunity, 2021, and 2) Jia YF et al, Behav. Brain Res. 2021. They also showed specificity of deletion in astrocytes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 764, 877, 843]]<|/det|>
+The reviewer states that GFAP- Cre driven deletion in our experiments is not specific to astrocytes, that ERbeta was also deleted from neurons, so our effects on cognitive outcomes at midlife may be due to ERbeta deletion in neurons (not astrocytes). If that were true, then one would see effects on cognitive outcomes in the neuronal ERbeta cKO (NSE- Cre), but that was not the case (see former Fig 4, new Fig. 5).  
+
+<|ref|>text<|/ref|><|det|>[[115, 844, 877, 907]]<|/det|>
+Specificity of GFAP- Cre driven selective gene deletion can differ not only between different mGFAP- Cre lines, but also between mGFAP- Cre versus hGFAP- Cre, and between sexes. The mGFAP- Cre 77.6 line has shown some expression of Cre recombinase in the male, but not the female, germline
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 880, 202]]<|/det|>
+(https://www.jax.org/strain/024098). To avoid this ectopic recombination, the mGFAP-Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP-Cre 77.6: ERβ<sup>ff</sup> females were crossed with ERβ<sup>ff</sup> males (without the mGFAP-Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP-Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 202, 870, 267]]<|/det|>
+Regarding use of an inducible Cre mouse line, cognitive and hippocampal outcomes observed at midlife were not observed in astrocyte ERbeta cKO mice at age 3 months (see Suppl. Fig. 1). Thus, cognitive and hippocampal outcomes observed in astrocyte ERbeta cKO mice at midlife were not due to effects on brain development.  
+
+<|ref|>text<|/ref|><|det|>[[115, 297, 879, 348]]<|/det|>
+2) There is no validation of these newly generated astroglial- and neuronal-specific ER beta transgenic models. One would like to see quantification of ER beta expression in astrocytes and neurons in both cKO mice (gfap-cre:ER betafl/fl and nestin-cre:ER betafl/fl).  
+
+<|ref|>text<|/ref|><|det|>[[115, 362, 812, 396]]<|/det|>
+As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 396, 870, 444]]<|/det|>
+Validation of ERbeta deletion specifically in neurons using the NSE-Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 458, 835, 507]]<|/det|>
+3) There is a lack of details regarding the mice (origin (stock number or laboratory source), breeding strategy controlling germline recombination, appropriate littermate controls. This precludes evaluation of whether the authors have used appropriate controls.  
+
+<|ref|>text<|/ref|><|det|>[[115, 522, 867, 571]]<|/det|>
+As suggested by the reviewer, details of information about mice, including references, have now been added to the Methods section. Also, schematics of breeding have been added to new Fig. 4a and 4d of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 586, 878, 650]]<|/det|>
+4) The authors use GFAP immunostaining to assess astroglial reactivity. GFAP expression can vary independently of astroglial reactivity. The authors should thus use additional markers (such as vimentin, stat3, GS...) and analyse astrocyte morphology to conclude about astrocyte reactivity.  
+
+<|ref|>text<|/ref|><|det|>[[115, 666, 835, 699]]<|/det|>
+As suggested by the reviewer, reactive astrocytes are now stained as double positive: LCN2\*GFAP\* (see Fig. 3, Fig. 5, and Fig. 7 of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 714, 872, 779]]<|/det|>
+5) The authors report synapse loss in menopausal female. They however only performed PSD95 staining to assess synapses. A synapse is composed of a pre- and a postsynaptic element and can only be identified by the colocalization of both elements. Staining for both pre- and postsynaptic markers should thus be performed to quantify synapse number.  
+
+<|ref|>text<|/ref|><|det|>[[115, 794, 864, 827]]<|/det|>
+Pre- synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)  
+
+<|ref|>text<|/ref|><|det|>[[115, 842, 870, 890]]<|/det|>
+6) This manuscript completely lacks mechanistic insights. How ER beta in astrocytes contributes to brain structure and cognitive performance in male and female? Does this require a crosstalk with other cell types such as microglia?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 881, 302]]<|/det|>
+Mechanistic insights are shown by a series of causality experiments. 1) An ovarian hormone by age interaction was found, whereby both midlife aging and loss of ovarian hormones caused cognitive decline, neuropathology, and atrophy of dorsal hippocampus by in vivo MRI. 2) Selective deletion of estrogen receptor beta (ERβ) in astrocytes, but not neurons, caused cognitive decline, neuropathology, and atrophy of dorsal hippocampus by in vivo MRI. 3) Complementing our two loss- of- function causality experiments (gonadectomy and specific deletion of ERβ in astrocytes), gain- of- function experiments using ERβ ligand treatment of midlife females showed that treatment caused reversal of outcomes. 4) Gene expression differences in hippocampal astrocyte transcriptomes from midlife females with selective deletion of ERβ vs WT in astrocytes revealed Gluconeogenesis I and Glycolysis I pathways as the most differentially expressed pathways. Further studies of a key gene in the Gluconeogenesis I pathway, Eno1, were also done (see new Fig. 6 in the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 302, 880, 351]]<|/det|>
+Additional mechanisms related to cross talk between astrocytes and microglia are interesting and discussed. Further experiments addressing this possibility are beyond scope here.  
+
+<|ref|>text<|/ref|><|det|>[[116, 382, 418, 398]]<|/det|>
+Reviewer #4 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 413, 877, 494]]<|/det|>
+The narrative is very interesting, the results important and well contextualized. The work is extremely significant and timely. The methodology is rigorous, although a few important pieces of evidence are missing to have a more balanced picture and close the story (detailed below). I also have doubts on some of the statistical approaches chosen. The paper is easy to follow and very well written.  
+
+<|ref|>text<|/ref|><|det|>[[116, 495, 680, 511]]<|/det|>
+There are a few aspects that can be improved. Below, my suggestions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[175, 526, 271, 543]]<|/det|>
+## Thank you!  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 120, 588]]<|/det|>
+1  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 864, 655]]<|/det|>
+Fig.1 : the information contained in a- b and c- d is redundant. In addition, doing twice a scatter plot, which distributes points random, on the same data generates plots which are visually different, which is awkward. The same applies to Fig. 2 a- b and c- d, and Fig. 4 a- b. I think half of the plots can safely go to supplementary.  
+
+<|ref|>text<|/ref|><|det|>[[114, 685, 868, 767]]<|/det|>
+As suggested by the reviewer, to remove redundancy MwM data are now shown as a merged graph with both between group comparisons of % Time in Target Quadrant and within group comparisons of % Time in Target Quadrant compared to % Time in Other Quadrants. This is done in former Fig. 2 (new Fig. 3) and in former Fig. 4 (new Fig. 5)  
+
+<|ref|>text<|/ref|><|det|>[[115, 784, 120, 796]]<|/det|>
+2  
+
+<|ref|>text<|/ref|><|det|>[[115, 799, 870, 862]]<|/det|>
+Following the previous comment, I think the most correct statistical approach would be a single ANOVA for the water maze data where sex is a factor (in addition to age and quadrant), followed by appropriate post- hoc comparisons. The same applies for MRI volumetric data, where ROI should be another factor (cortex, striatum, HC).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 875, 156]]<|/det|>
+In the revised manuscript, given the relatively small sample size and unequal variance, MWM was analyzed using a nonparametric test, Mann Whitney \(U\) test. All volumetric MRI data were analyzed using a two- way ANOVA (age, sex) and then followed up with two- tailed Welch's t- tests, appropriate for these comparisons. See item 4 below.  
+
+<|ref|>text<|/ref|><|det|>[[115, 171, 129, 185]]<|/det|>
+3  
+
+<|ref|>text<|/ref|><|det|>[[115, 186, 876, 283]]<|/det|>
+In Fig 1 h, it seems that there is no volume change in HC in male mice across the lifespan. This results somehow weakens the point made by the authors in the rest of the paper, about the importance for neuroprotection across the lifespan of the ERb astrocyte receptors specifically in the dorsal hippocampus, unless the dHC is only important in ageing female, which is hard to digest, unless supported by evidence. Generally speaking, the paper is missing showing and discussing in males at least some of the analyses performed in females, for example Fig 2 g- i.  
+
+<|ref|>text<|/ref|><|det|>[[115, 297, 832, 331]]<|/det|>
+The reviewer is correct. While females had dorsal hippocampal atrophy from midlife to old age, males did not.  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 875, 490]]<|/det|>
+Analysis of data from the Genotype- Tissue Expression (GTEx) project examined sex differences in gene expression across 44 tissues in humans and showed that \(37\%\) of all genes exhibit sex- biased expression in at least one tissue (Oliva et al., Science, 2020). In another study using the same dataset focusing on 29 human healthy tissues, whole- genome expression profiles showed distinct sex- biased regulatory networks in each tissue (Lopes- Ramos, Cell Rep. 2020). Finally, sex differences in gene expression are region- specific and cell- specific within the brain (Kim- Hellmuth, et al., Science, 2020). These studies underscore the pervasiveness and complexity of sex differences in gene expression during health, which can be distinct depending on the brain region and cell involved.  
+
+<|ref|>text<|/ref|><|det|>[[115, 489, 844, 554]]<|/det|>
+Sex differences in substructure volumes and rates of atrophy have previously been shown in MRI studies in humans and mice. Dorsal hippocampus and ventral hippocampus differ in regard to which is larger in females versus males, each when normalized for brain size.  
+
+<|ref|>text<|/ref|><|det|>[[115, 554, 876, 699]]<|/det|>
+Given the above evidence in the sex differences field, we did not find it particularly surprising that dorsal hippocampus may undergo atrophy with aging in females and not males. That said, this manuscript did not pursue why dorsal atrophy did not occur in males with aging. A related question is why did males have gradual atrophy with aging in frontal cortex and striatum, but not in dorsal hippocampus? It could be due to direct effects of testosterone or dihydrotestosterone acting on androgen receptors in hippocampus in males. This is only one of several possibilities. Space limitations do not permit our speculations about males during aging since our studies focused on females during aging.  
+
+<|ref|>text<|/ref|><|det|>[[115, 714, 819, 763]]<|/det|>
+Fig 1, supplementary excel table: why using Welch test? Is it because the samples have unequal variance? Please specify in the methods.  
+
+<|ref|>text<|/ref|><|det|>[[115, 777, 868, 858]]<|/det|>
+Yes, the samples do exhibit unequal variances, as is the case in almost all biological measures. Thus, the use of Welch's t- test is appropriate here since it performs better than the Student's t- test when sample sizes and variances are unequal between groups, and it gives identical results when sample sizes and variances are equal. We routinely use Welch's t- test for volumetric analyses in atlas- based morphometry.  
+
+<|ref|>text<|/ref|><|det|>[[115, 872, 840, 905]]<|/det|>
+Fig 1: It is noteworthy that from Fig. 1 there is no relation between structure and function, at
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 880, 154]]<|/det|>
+least in the context of ageing. If it is a matter of different sensitivities of the two techniques employed (MRI volumetry vs behaviour), as hinted by the authors, I suggest they discuss the issue and report the power analysis used for determining the sample size. Perhaps the behavior is underpowered?  
+
+<|ref|>text<|/ref|><|det|>[[114, 169, 880, 395]]<|/det|>
+See response to Reviewer #2 regarding not observing a significant defect in spatial reference memory in the Morris Water Maze in the 2hr delay probe trial in gonadally intact, healthy females or males up to the ages tested in the conditions of the UCLA Behavioral Core Facility given the sample size used. Sample sizes were driven by the number of mice needed to generate a robust minimum deformation atlas (MDA) for atlas- based morphometry (MacKenzie- Graham Neurolmage 2012). Logistically, with several groups (various ages, two sexes, and/or GDX vs sham surgery, cKOs vs WT, treatments) and several outcomes, sample size was also limited by what was practically feasible by personnel and costs. Data from in vivo MRI is expected to be more sensitive than clinical testing in mice. This aligns with in vivo MRI for brain atrophy being a more sensitive outcome than clinical outcomes in humans. Indeed, MRI is very frequently used as a more sensitive biomarker in Phase 2 clinical trials prior to designing Phase 3 trials which require much larger sample sizes since they have a clinical primary outcome (such as cognitive performance).  
+
+<|ref|>text<|/ref|><|det|>[[115, 425, 840, 491]]<|/det|>
+6 Fig.2: Similarly to comments 1 and 2, the figures of the water maze are redundant, and the statistical model could be just one for each test but include everything (gender, therapy, life stage).  
+
+<|ref|>text<|/ref|><|det|>[[115, 504, 870, 555]]<|/det|>
+As suggested by the reviewer, redundancy of MwM data has been decreased by making a single merged panel for between groups and within groups comparisons. See comments above regarding statistical tests used.  
+
+<|ref|>text<|/ref|><|det|>[[115, 570, 530, 604]]<|/det|>
+7 Fig. 2 f: why freezing was only measured at midlife?  
+
+<|ref|>text<|/ref|><|det|>[[115, 616, 877, 667]]<|/det|>
+Fear conditioning was done to expand upon MwM test performance at midlife in females, the group and age of focus. It was not done at all ages, in both sexes, in + / - GDX, or in cKOs. Therefore, fear conditioning has now been removed from the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 682, 130, 696]]<|/det|>
+8  
+
+<|ref|>text<|/ref|><|det|>[[115, 697, 875, 762]]<|/det|>
+Fig. 2 j- k: I suggest to only plot regression line when significant. Also, there seems to be a lot of points in the plots. Are all animals (both sexes, all ages, GDX/sham) pulled together? If so, which is the rationale? Perhaps the different groups could be shown in the plot by coloring the datapoints differently.  
+
+<|ref|>text<|/ref|><|det|>[[115, 776, 870, 907]]<|/det|>
+As suggested by the reviewer, dot plots in former Fig. 2 (new Fig. 3 p,q) have now been made using distinct coloring of datapoints (see below). This can replace the panel with all black dots upon request. The rationale of pooling groups is to increase the sample size when asking if cognitive performance (% Time in Target Quadrant) correlates with dorsal (p) or ventral (q) hippocampal volume by in vivo MRI. The plots nicely show a positive correlation between better cognitive performance and higher volumes of dorsal (but not ventral) hippocampus. This is presented in our manuscript since in vivo MRI for hippocampal atrophy is not usually done in papers showing behavioral testing in mice.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[123, 128, 581, 312]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 870, 397]]<|/det|>
+Possible panel replacement in new Fig. 3. % time in TQ correlated with (p) dorsal hippocampus volume \((r = 0.28\) ; \(p = 0.011\) ), not with (q) ventral hippocampus volume \((r = 0.027\) ; \(p = 0.809\) ). Dot colors: black (astrocyte ERβ cKO, old), yellow (astrocyte ERβ cKO, young), blue (neuron ERβ cKO, old), green (neuron ERβ cKO, young), red (WT, old), pink (WT, young).  
+
+<|ref|>text<|/ref|><|det|>[[115, 428, 130, 441]]<|/det|>
+9  
+
+<|ref|>text<|/ref|><|det|>[[115, 442, 872, 523]]<|/det|>
+Fig. 3, Neuropathology: the authors talk about "area fraction" but what is exactly measured and how? Can distinct aspect of the morphology be measured (ramifications density, cell body size, cell number etc.? This is important as the term "glia activation" can refer to distinct morphological changes happening to the cell. In general, the method- >histology section is missing details needed for the reader to be able to reproduce the analyses.  
+
+<|ref|>text<|/ref|><|det|>[[115, 538, 867, 619]]<|/det|>
+The term "area fraction" is widely used during double-labelling. For example, positive labeling in a specific cell type was quantified and graphed as the percentage labeling of LCN2+ area within the GFAP+ area. A level of detail needed for the reader to reproduce histology (information about antibodies, dilution factors) is now in the Methods section of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 641, 135, 655]]<|/det|>
+10  
+
+<|ref|>text<|/ref|><|det|>[[115, 657, 876, 706]]<|/det|>
+Line 194: the notation is misleading. Perhaps by "direct" and "indirect" correlations the authors mean linear positive and linear negative? Or do the authors really mean indirect as mediated by something else? If so, specify.  
+
+<|ref|>text<|/ref|><|det|>[[115, 720, 675, 737]]<|/det|>
+The reviewer is correct. Direct is positive and indirect is negative.  
+
+<|ref|>text<|/ref|><|det|>[[115, 752, 135, 766]]<|/det|>
+11  
+
+<|ref|>text<|/ref|><|det|>[[115, 768, 863, 801]]<|/det|>
+Fig 3 is missing the plots of the correlations between time in TQ and gliosis (data is discussed but not shown).  
+
+<|ref|>text<|/ref|><|det|>[[115, 816, 869, 898]]<|/det|>
+The correlation between % Time in Target Quadrant and glial activation markers are stated in the results section. Plots showing the correlations between % Time in Target Quadrant and LNC2+GFAP+ astrocytes, as well as % Time in Target Quadrant and MHCII+ IBA1+ microglia are presented below. This can be added as additional supplemental figure upon request.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 103, 600, 300]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[115, 315, 878, 397]]<|/det|>
+Possible additional supplemental figure. Correlation of MWM cognitive performance with glia activation. a- b) % Time in target quadrant is negatively correlated with a) LNC2\*GFAP\* astrocytes \((r = - 0.46203\) , \(p = 0.017\) ) and b) MHCII\* \(\mathrm{IBA1^{+}}\) microglia \((r = - 0.40723\) , \(p = 0.048\) ). red dot, Young Sham; red dot with black rim, Young GDX; black square, Midlife Sham; white square, Midlife GDX females. Pearson correlation analyses.  
+
+<|ref|>text<|/ref|><|det|>[[115, 428, 137, 443]]<|/det|>
+12  
+
+<|ref|>text<|/ref|><|det|>[[115, 444, 800, 477]]<|/det|>
+Following the rationale of the paper, Fig 4 is missing the neuropathology. How are the astrocytes affected by ERb deletion? Are they more activated at midlife?  
+
+<|ref|>text<|/ref|><|det|>[[115, 491, 880, 525]]<|/det|>
+As suggested by the reviewer, this neuropathology has now been done (see new Fig. 5e- l in the revised manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[115, 540, 137, 555]]<|/det|>
+13  
+
+<|ref|>text<|/ref|><|det|>[[115, 556, 880, 621]]<|/det|>
+Following the previous comment but more generally, what is ultimately the mechanism that the authors identified as leading to atrophy and worst cognition in ageing females? Is it astrogliosis? If so, the authors should report more neuropathological data (or other equivalent astrogliosis markers) to make a stronger point.  
+
+<|ref|>text<|/ref|><|det|>[[115, 635, 816, 670]]<|/det|>
+In addition to more neuropathology, we have also added new Fig. 6 showing gene expression in astrocytes by RNA- sequencing. See response #6 to Reviewer #3.  
+
+<|ref|>text<|/ref|><|det|>[[115, 685, 137, 699]]<|/det|>
+14  
+
+<|ref|>text<|/ref|><|det|>[[115, 700, 860, 733]]<|/det|>
+It is interesting that there is no cognitive effect in young female mice which lacks endogenous hormones or receptors. What about HC volume?  
+
+<|ref|>text<|/ref|><|det|>[[115, 745, 861, 811]]<|/det|>
+There were no differences in HC volumes in young female GDX vs young female sham. See the figure below, which can be added as a supplemental figure upon request. This, together with our other findings, reveals an ovarian sex hormone by age interaction. In essence, it is deleterious to have BOTH aging to midlife AND to lose ovarian hormones.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[112, 85, 572, 310]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[114, 325, 852, 407]]<|/det|>
+Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Substructure volumes, assessed by MRI, taken as a percentage of intercranial volume (ICV) are shown for (a) hippocampus, (b) dorsal hippocampus, and (c) ventral hippocampus. There were no differences in substructure volumes between groups at young age. \(n = 8\) for each group.  
+
+<|ref|>text<|/ref|><|det|>[[116, 434, 138, 448]]<|/det|>
+15  
+
+<|ref|>text<|/ref|><|det|>[[116, 448, 863, 499]]<|/det|>
+As the authors point out in the discussion, microglia also have ERb. Can the author show and compare some histology of microglia in animals lacking endogenous receptors and animals lacking ERb in astrocytes?  
+
+<|ref|>text<|/ref|><|det|>[[115, 512, 870, 563]]<|/det|>
+As suggested by the reviewer, this is now added to the revised manuscript. See double- labelling stains for MHCII & IBA1 as well as for CLEC7A & P2RY12. This is done in sham vs GDX (new Fig. 3), cKO vs WT (new Fig. 5), and ER beta ligand treatment (new Fig. 7).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[119, 83, 313, 98]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[120, 111, 415, 125]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 139, 870, 167]]<|/det|>
+The authors have addressed several of my comments in the revised manuscript. Yet there are still a few points that were not satisfactorily addressed and needs to be done:  
+
+<|ref|>text<|/ref|><|det|>[[118, 182, 867, 266]]<|/det|>
+1) "To avoid this ectopic recombination, the mGFAP-Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP-Cre 77.6: ERβf/f females were crossed with ERβf/f males (without the mGFAP-Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP-Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed."  
+
+<|ref|>text<|/ref|><|det|>[[119, 280, 554, 294]]<|/det|>
+Please add in the methods section the breeding strategy.  
+
+<|ref|>text<|/ref|><|det|>[[118, 322, 864, 378]]<|/det|>
+2) "As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript). Validation of ERbeta deletion specifically in neurons using the NSE-Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript)."  
+
+<|ref|>text<|/ref|><|det|>[[118, 392, 805, 420]]<|/det|>
+To be fully convinced by the astroglial cKO mice in, the authors need to illustrate that the expression of ER beta is intact in neurons from the gfap- cre:ER betafl/fl.  
+
+<|ref|>text<|/ref|><|det|>[[116, 434, 850, 462]]<|/det|>
+3) "Pre-synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)"  
+
+<|ref|>text<|/ref|><|det|>[[118, 476, 856, 518]]<|/det|>
+I understand the SYN1 staining is done and illustrated in these figures, but it is not used in the manuscript to quantify synapse number by counting the number of colocalized SYN1 and PSD95 puncta. This needs to be done.  
+
+<|ref|>text<|/ref|><|det|>[[119, 588, 415, 602]]<|/det|>
+Reviewer #4 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 616, 783, 644]]<|/det|>
+The authors have successfully addressed my concerns and those from other reviewers. Minor comments:  
+
+<|ref|>text<|/ref|><|det|>[[118, 658, 700, 673]]<|/det|>
+Possible panel replacement in new Fig. 3: I think that the plot can stay by/w.  
+
+<|ref|>text<|/ref|><|det|>[[118, 686, 830, 714]]<|/det|>
+"Direct" and "indirect" to qualify correlation is used but not precise. Indirect strictly speaking means "mediated by". I recommend using positive or negative.  
+
+<|ref|>text<|/ref|><|det|>[[118, 728, 866, 770]]<|/det|>
+Plots showing the correlations between \(\%\) Time in Target Quadrant and LNC2+GFAP+ astrocytes, as well as \(\%\) Time in Target Quadrant and MHCII+ IBA1+ microglia: I would add them in the supplementary.  
+
+<|ref|>text<|/ref|><|det|>[[115, 784, 870, 813]]<|/det|>
+Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Also suitable for supplementary material.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[86, 42, 620, 62]]<|/det|>
+## REVIEWER COMMENTS followed by Point-by-point Responses.  
+
+<|ref|>sub_title<|/ref|><|det|>[[87, 76, 409, 94]]<|/det|>
+## Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[85, 107, 935, 206]]<|/det|>
+Comment 1) "To avoid this ectopic recombination, the mGFAP- Cre 77.6 allele should not be in male parents in the breeding scheme. In our experiments, mGFAP- Cre 77.6: ERβf/f females were crossed with ERβf/f males (without the mGFAP- Cre 77.6 allele), thereby avoiding germline ectopic recombination. This breeding instruction about the mGFAP- Cre 77.6 line is available from Jackson Labs, but could be added to our Methods section upon request. We can also clarify the sex of the parents in our breeding strategy in the Methods section if needed." Please add in the methods section the breeding strategy.  
+
+<|ref|>text<|/ref|><|det|>[[87, 204, 930, 238]]<|/det|>
+Response 1) Now added to the methods section as well as referred to in the results section of the revised manuscript is the following:  
+
+<|ref|>text<|/ref|><|det|>[[85, 236, 930, 365]]<|/det|>
+In breeding of our mouse lines, Cre recombinase alleles were always inherited from females (mother), not from males (father). This is because one Cre line showed ectopic expression of Cre recombinase in the male germline (mGFAP- Cre 77.6 line; https://www.jax.org/strain/024098). Thus, paternal inheritance of Cre recombinase should be avoided. To this end, our breeding pairs were: a) GFAP- Cre:ERβf/f females with ERβf/f males (to generate astrocyte ERβ cKO mice), b) NSE- Cre:ERβf/f females with ERβf/f males (to generate neuron ERβ cKO mice), c) GFAP- Cre:RiboTag females with RiboTag males (to generate astrocyte RiboTag mice), and d) GFAP- Cre:RiboTag:ERβf/f females with RiboTag: ERβf/f males (to generate astrocyte ERβ cKO RiboTag mice).  
+
+<|ref|>text<|/ref|><|det|>[[85, 363, 930, 397]]<|/det|>
+Labels of Male or Female to parents during breeding have also been added to Fig. 4 of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[85, 410, 936, 492]]<|/det|>
+Comment 2) "As suggested by the reviewer, validation of ERbeta deletion specifically in astrocytes is now done (see new Fig. 4 of the revised manuscript). Validation of ERbeta deletion specifically in neurons using the NSE- Cre has been validated and published, see Spence, et al, J. Neuroscience, 2013 (as cited in this manuscript)." To be fully convinced by the astroglial cKO mice in, the authors need to illustrate that the expression of ER beta is intact in neurons from the gfap- cre:ER beta/f/f.  
+
+<|ref|>text<|/ref|><|det|>[[85, 491, 904, 526]]<|/det|>
+Response 2) In the revised manuscript a supplemental figure is added focusing on ERβ expression in neurons. ERβ expression was intact in neurons of the astrocyte- ERβ cKO (see Supplemental Fig. 3).  
+
+<|ref|>text<|/ref|><|det|>[[85, 540, 933, 605]]<|/det|>
+Comment 3) "Pre- synaptic staining for SYN1 is done in the revised manuscript (see Fig. 3, Fig. 5, and Fig. 7)". I understand the SYN1 staining is done and illustrated in these figures, but it is not used in the manuscript to quantify synapse number by counting the number of colocalized SYN1 and PSD95 puncta. This needs to be done.  
+
+<|ref|>text<|/ref|><|det|>[[85, 605, 907, 639]]<|/det|>
+Response 3) Co- localization of pre- synaptic staining for SYN1 and post- synaptic staining for PSD95 is now done in the revised manuscript (see Fig. 3h, Fig. 5l, and Fig. 7f).  
+
+<|ref|>sub_title<|/ref|><|det|>[[87, 653, 409, 671]]<|/det|>
+## Reviewer #4 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[85, 684, 892, 719]]<|/det|>
+Comment 1) The authors have successfully addressed my concerns and those from other reviewers. Response 1) Thank you.  
+
+<|ref|>text<|/ref|><|det|>[[86, 733, 335, 749]]<|/det|>
+Comment 2) Minor comments:  
+
+<|ref|>text<|/ref|><|det|>[[85, 749, 644, 766]]<|/det|>
+Possible panel replacement in Fig. 3: I think that the plot can stay b/w.  
+
+<|ref|>text<|/ref|><|det|>[[85, 766, 912, 815]]<|/det|>
+Response 2) As recommended by the reviewer, panels (in former Fig. 3p,q; current Fig. 2m,n) showing correlation analyses between Dorsal Hippocampus volume and % Time in Target Quadrant or between Ventral Hippocampus volume and % Time in Target Quadrant will stay black and white.  
+
+<|ref|>text<|/ref|><|det|>[[85, 828, 920, 862]]<|/det|>
+Comment 3) "Direct" and "indirect" to qualify correlation is used but not precise. Indirect strictly speaking means "mediated by". I recommend using positive or negative.  
+
+<|ref|>text<|/ref|><|det|>[[85, 861, 888, 910]]<|/det|>
+Response 3) The word "direct" correlation has now been revised to "positive" correlation to describe results in Fig. 2m. The word "indirect" correlation has now been revised to "negative" correlation to describe results in Supplemental Fig. 2.  
+
+<|ref|>text<|/ref|><|det|>[[85, 923, 884, 942]]<|/det|>
+Comment 4) Plots showing the correlations between % Time in Target Quadrant and LNC2+GFAP+
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[85, 43, 911, 76]]<|/det|>
+astrocytes, as well as % Time in Target Quadrant and MHCII+ IBA1+ microglia: I would add them in the supplementary.  
+
+<|ref|>text<|/ref|><|det|>[[85, 77, 901, 110]]<|/det|>
+Response 4) These two plots showing correlations have now been added to supplementary materials (Supplemental Fig. 2).  
+
+<|ref|>text<|/ref|><|det|>[[85, 124, 925, 157]]<|/det|>
+Comment 5) Possible additional supplement figure. Gonadectomy did not induce hippocampal atrophy at young age. Also suitable for supplementary material.  
+
+<|ref|>text<|/ref|><|det|>[[85, 157, 861, 175]]<|/det|>
+Response 5) This figure has now been added to supplementary materials (Supplemental Fig. 1).
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d5cd42613e42af1cb6af7d1aa2807cdfdd9681d47c9231fe8f1b4ff73a62a0d1/images_list.json

@@ -0,0 +1,25 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Distribution of azimuthal and tilt angles sampled during MD simulations, and a scatter plot showing minimal correlation between the dynamics of these two angles during MD.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Bending angle by replica for simulations using \\(40^{\\circ}\\) map",
+    "footnote": [],
+    "bbox": [
+      [
+        125,
+        110,
+        568,
+        355
+      ]
+    ],
+    "page_idx": 20
+  }
+]

peer_reviews/supplementary_0_Peer Review File__d5cd42613e42af1cb6af7d1aa2807cdfdd9681d47c9231fe8f1b4ff73a62a0d1/supplementary_0_Peer Review File__d5cd42613e42af1cb6af7d1aa2807cdfdd9681d47c9231fe8f1b4ff73a62a0d1.mmd

@@ -0,0 +1,379 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Structural insights into the modulation of coronavirus spike tilting and infectivity by hinge glycans  
+
+![](images/Figure_1.jpg)
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+HCoV- NL63 is an endemic seasonal human coronavirus which shares the same huACE2 receptor with SARS- CoV- 2. Compared to SARS- CoV- 2, HCoV- NL63 shows a lack of adaptive evolution, and is more glycosylated on its spikes. In this manuscript, Chmielewski et al. have integrated multiple experimental and computational approaches to discover and demonstrate the impact of a single glycan on the dynamics and infectivity of HCoV- NL63. The authors have imaged unfixed virions by cryo- ET and analyzed its spike protein structures and statistics. Combined with MS, the glycan compositions were analyzed and an atomic model of the most abundant glycan topology was built into the density of their previous single particle cryoEM map, giving them the opportunity in analyzing the glycan shielding. Atomic models of the ectodomain were built for glycosylated spike resolved at tilt angles of 100, 200...700, which were subjected to MD analysis. The simulation has revealed the bending profile of the spike, revealing that two N- glycans (N1242 &1247) near this disordered hinge region may have important roles in spike tilting. Further MD and infectivity assay using a VSV pseudovirus have revealed a role of the conserved N- linked glycan N1242 in both spike bending and infectivity.  
+
+Overall, the discoveries presented in this work is novel, and the logic among various experimental approaches is fluent. The authors have also discussed the limitations in this work, such as uncertainties in building the atomic model for MD and short simulation duration, which is helpful for readers. My main concerns locate at the sample preparation procedure and infectivity assay using the VSV pseudovirus. The manuscript needs major revision before being recommend for publication in Nature Communications.  
+
+Major points:  
+
+1. The authors pelted the unfixed virions through \(20\%\) sucrose cushion followed by OptiPrep gradient, then filtered the virions through a membrane filter to exchange the buffer. The pelleting and filtering raise concerns on the intactness of the virions, as the thermal stability of coronaviruses and their spike proteins are known to be weak (especially when they are not chemically fixed). Also, for how long and at what temperature was the virion sample stored during each step before plunge-freezing? This is not clearly written in the method part on virus preparation. These possibly result in only \(\sim 9\) spikes (according to line 130) found per virion and some virions look bald (fig.1A), this number is significantly less than that of SARS-CoV-2 and PEDV (\~30 spikes/virion). This suggests that the virions are possibly damaged during sample preparation. If so, the average distance between neighboring spikes is overestimated, and this will impact the spike dynamics analysis, since the spikes can contact each other on an intact virion. The authors shall analyze this possibility and provide a control for the virion intactness by cryo-ET of freshly prepared & unconcentrated HCoV-NL63 in the supernatant.
+
+<--- Page Split --->
+
+2. The cryo-ET data processing part is not clearly stated. Detailing this part in the M&M session will not only justify the robustness of their structures, but also help the readers in learning cryo-ET:  
+
+(1) In the M&M part, the authors wrote: "approximately 15,000 individual spike particles were manually picked... Subtomogram averaging was then performed using \(\sim 10,000\) particles". However, in line 133, they wrote "Subtomogram averaging analysis of 18,356 spikes". The numbers are inconsistent.  
+
+(2) What software did they use for subtomogram averaging? How was the "focused classification" (as stated in line 153) done?  
+
+(3) A supplementary table containing detailed information on cryo-ET data acquisition and processing must be provided.  
+
+3. Why is the cryo-ET determined spike with a bending angle of 80o not modeled for MD?  
+
+4. The influence of the hinge glycan N1242 on infectivity was analyzed using VSV-Luc reporter pseudovirus. While coronaviruses bud in ERGIC, where N-glycosylation gets further modified, VSV buds at the plasma membrane. In addition, coronaviral E protein has a function of intracellular retention, which helps concentrating viral structural proteins at the ERGIC to favor viral particle assembly. Therefore, the difference in budding and the missing E from the VSV pseudovirus can result in differences in glycosylation. The authors shall analyze the glycan compositions of the VSV pseudovirus, especially for the N1242 and N1247 sites, to control for the possible differences in glycosylation.  
+
+Minor points:  
+
+1. Figure 4D is never mentioned in the manuscript.  
+
+2. Fig. S2B is not mentioned in the main text. There is no discussion on why/how two spikes are connected in the stalk.  
+
+3. Fig. 1G: the angles are boxed in colors, however, the colors are not explained.  
+
+4. Adding a comparison animation of the MD simulation of S-trimers with and without the N1242 glycan to the manuscript will be highly appreciated.  
+
+5. I'm wondering if the most favorable spike tilt 56o is the optimal tilt conformation, what would be the virological implications of this angle? Would the spike mediate receptor binding better so as to influence viral infectivity? A little discussion expansion on correlation between the spike tilting and viral infectivity will be appreciated.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript by Chmielewski et al. provides important knowledge on the molecular dynamics of virion- bound HCoV- NL63 spike glycoprotein (S) and proposes that a specific N- glycan within a flexible hinge region at the stalk- crown interface confers some of the conformational flexibility. By applying
+
+<--- Page Split --->
+
+focused classification, the authors were able to resolve the otherwise orientationally flexible stalk domain, indiscernible at the complete cryo- EM map. By further isolating image classes with various combinations of crown and stalk domain orientations, the authors also elegantly demonstrate highly variable tilting angles of virion- bound spike, with the crown domain not only tilting, but also rotating relative to the stalk domain, with this flexibility imposed by the hinge region. Intriguingly, in silico removal of the hinge- resident N- glycan at N1242, or all the S glycans, results in a shift of most energetically favorable tilting angles from \(56^{\circ}\) to around \(25^{\circ}\) , decreasing S flexibility, and possibly increasing exposure to antibodies. Moreover, HCoV- NL63 S- pseudotyped VSVdeltaG with said N- glycan deletion exhibits substantially decreased infectivity. The work provides extremely valuable insight into how glycans affect the conformational stability of proteins – a subject that is still underexplored. The authors provide compelling evidence for plausible roles of glycans in HCoV- NL63 biology, which could be further strengthened by in vitro experiments. The glycoproteomic analysis of virion- derived S is a welcome addition for determining predominant site- specific glycan compositions, however, a few details should be clarified.  
+
+1. The mass spectrometry sample and data analysis approach should be clarified. The multiple digests are typically employed to enhance coverage, and it is not unusual to pool them prior to analysis. Is that what was performed here, or were they analyzed separately? Given the semi-quantitative nature of spectral counting, were any measures taken to estimate variation (biological or technical replicates)?  
+
+2. The increased access to antibody molecules upon removal of N-glycans in the flexible hinge region makes sense, as suggested by the molecular modelling experiments (Figure 4C, 4D). However, the software-determined approach measuring the accessible area of the probe, and not of the exposed spike area is somewhat counterintuitive. It would strengthen the manuscript, if an in vitro approach was employed to further test the validity of the claims. For example, suitable neutralizing antibodies, or sera from seropositive individuals could be used to measure antibody binding capacity, or neutralization of pseudotyped virions.  
+
+3. Figure 5C. The plot elements should be described in the legend (measure of center, error bars, and statistical test). If the data is normalized to the infectivity of wild type, why are there error bars on the wild type data? What is the reproducibility of the data? Pseudovirus infectivity data can be rather variable, and the information on numbers of independent experiments is missing.  
+
+Minor:  
+
+Lines 254- 255. The term "monosaccharide" should be used instead of "glycan".
+
+<--- Page Split --->
+
+Line 319. It would be useful to mention exact numbers for glycan shield density for the other coronaviruses.  
+
+Lines 484- 485. How was the "conservation of glycosylation" determined as opposed to amino acid conservation (previous sentence)? Is it based on experimental data from the literature? If so, how many and which viruses? Or is it algorithm- based prediction?  
+
+Figure S7D. Please add amino acid residue numbers for the start/end of the shown ranges.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+HCov- NL63 is an endemic seasonal human coronavirus which shares the same huACE2 receptor with SARS- CoV- 2. Compared to SARS- CoV- 2, HCoV- NL63 shows a lack of adaptive evolution, and is more glycosylated on its spikes. In this manuscript, Chmielewski et al. have integrated multiple experimental and computational approaches to discover and demonstrate the impact of a single glycan on the dynamics and infectivity of HCoV- NL63. The authors have imaged unfixed virions by cryo- ET and analyzed its spike protein structures and statistics. Combined with MS, the glycan compositions were analyzed and an atomic model of the most abundant glycan topology was built into the density of their previous single particle cryoEM map, giving them the opportunity in analyzing the glycan shielding. Atomic models of the ectodomain were built for glycosylated spike resolved at tilt angles of 10o, 20o...70o, which were subjected to MD analysis. The simulation has revealed the bending profile of the spike, revealing that two N- glycans (N1242 &1247) near this disordered hinge region may have important roles in spike tilting. Further MD and infectivity assay using a VSV pseudovirus have revealed a role of the conserved N- linked glycan N1242 in both spike bending and infectivity. Overall, the discoveries presented in this work is novel, and the logic among various experimental approaches is fluent. The authors have also discussed the limitations in this work, such as uncertainties in building the atomic model for MD and short simulation duration, which is helpful for readers. My main concerns locate at the sample preparation procedure and infectivity assay using the VSV pseudovirus. The manuscript needs major revision before being recommend for publication in Nature Communications.  
+
+Major points:  
+
+1. The authors pelleted the unfixed virions through \(20\%\) sucrose cushion followed by OptiPrep gradient, then filtered the virions through a membrane filter to exchange the buffer. The pelleting and filtering raise concerns on the intactness of the virions, as the thermal stability of coronaviruses and their spike proteins are known to be weak (especially when they are not chemically fixed). Also, for how long and at what temperature was the virion sample stored during each step before plunge-freezing? This is not clearly written in the method part on virus preparation. These possibly result in only \(\sim 9\) spikes (according to line 130) found per virion and some virions look bald (fig.1A), this number is significantly less than that of SARS-CoV-2 and PEDV ( \(\sim 30\) spikes/virion). This suggests that the virions are possibly damaged during sample preparation. If so, the average distance between neighboring spikes is over-estimated, and this will impact the spike dynamics analysis, since the spikes can contact each other on an intact
+
+<--- Page Split --->
+
+virion. The authors shall analyze this possibility and provide a control for the virion intactness by cryo- ET of freshly prepared & unconcentrated HCoV- NL63 in the supernatant.  
+
+We thank the reviewer for pointing out the missing details of virus preparation. We have added a detailed description about temperature/time during virus purification in the Materials and Methods. Our previous draft may have led to a possible misinterpretation of the number of virions in our tomograms from which spikes were actually analyzed by subtomogram averaging. We returned to the data and determined the number of spikes per particle by manual identification of spikes within tomograms. In our purified HCoV- NL63 virion, there are \(20 \pm 13\) spikes/HCoV- NL63 virions in a selected pool of 154 intact virus particles. This number is similar to what is reported for spikes in infusion conformation on SARS- CoV- 2 virion: \(23 \pm 9\) in (Ke et al. 2020) and \(26 \pm 15\) in (Yao et al. 2020), but lower than \(\sim 40\) spikes per virion reported in (Turoňová et al. 2020). Unlike betacoronaviruses, HCoV- NL63 spike is not cleaved between S1 and S2, and the spike displays only one closed prefusion conformation observed in both purified soluble spike proteins (Walls et al. 2016) and in situ on purified virions (Zhang et al. 2020). On SARS- CoV- 2 virus particles, not only prefusion but also postfusion spikes were observed, and the prefusion spikes display both open and closed conformations. The high flexibility of the SARS- CoV- 2 spikes may explain the loss of prefusion spikes in open conformation after centrifugation through sucrose cushion as reported in (Ke et al. 2020). The purity of our HCoV- NL63 virus prep and stability of HCoV- NL63 spike are shown in the coomassie gel in new Figure S3B, appearing to be more stable than betacoronavirus spikes.  
+
+2. The cryo-ET data processing part is not clearly stated. Detailing this part in the M&M session will not only justify the robustness of their structures, but also help the readers in learning cryo-ET:  
+
+(1) In the M&M part, the authors wrote: "approximately 15,000 individual spike particles were manually picked... Subtomogram averaging was then performed using \(\sim 10,000\) particles". However, in line 133, they wrote "Subtomogram averaging analysis of 18,356 spikes". The numbers are inconsistent.  
+
+We have revised our manuscript to clarify the data processing method. The same dataset was in fact processed twice using different versions of the EMAN2 software package resulting in different numbers of spikes. The results shown in the paper are from the final version, but some of the numbers in the original paper were not updated and thus incorrect. We have now updated the numbers in the main text and the method section. Our final analysis for subtomogram averaging included 18,356 spikes in both intact and apparently broken virions.  
+
+(2) What software did they use for subtomogram averaging? How was the "focused classification" (as stated in line 153) done?
+
+<--- Page Split --->
+
+We revised the method of subtomogram averaging and provided a supplemental figure (new Figure S1) showing the processing details of focused classification.  
+
+(3) A supplementary table containing detailed information on cryo-ET data acquisition and processing must be provided.  
+
+As suggested, we provided a supplemental table showing detailed information on cryoET data acquisition and processing.  
+
+3. Why is the cryo-ET determined spike with a bending angle of \(80^{\circ}\) not modeled for MD?  
+
+The simulations starting from the \(70^{\circ}\) bend had \(73.5\%\) of the trajectory going towards lower angles, and only \(26.5\%\) going to higher bent conformations. Furthermore, Of the trajectories with bending angles transitioning above \(70^{\circ}\) , only \(7\%\) (out of the aforementioned \(26.5\%\) ) reached all the way to \(80^{\circ}\) ( \(1.8\%\) of all simulations starting at \(70^{\circ}\) or \(0.2\%\) of all measured frames). None of the trajectories starting from other angles (i.e. \(< 70^{\circ}\) ) make it to \(80^{\circ}\) . On the other end of the bending range, \(5.8\%\) of all trajectories reached less than \(10^{\circ}\) (i.e. \(10^{\circ}\) to \(0^{\circ}\) ) which is still an order of magnitude more than that observed over the range from \(70^{\circ}\) to \(80^{\circ}\) . Therefore, our simulations encompassed extremities of bent conformations, even without starting from a low- probability model (i.e. \(80^{\circ}\) ).  
+
+4. The influence of the hinge glycan N1242 on infectivity was analyzed using VSV-Luc reporter pseudovirus. While coronaviruses bud in ERGIC, where N-glycosylation gets further modified, VSV buds at the plasma membrane. In addition, coronaviral E protein has a function of intracellular retention, which helps concentrating viral structural proteins at the ERGIC to favor viral particle assembly. Therefore, the difference in budding and the missing E from the VSV pseudovirus can result in differences in glycosylation. The authors shall analyze the glycan compositions of the VSV pseudovirus, especially for the N1242 and N1247 sites, to control for the possible differences in glycosylation.  
+
+We concur with the reviewer's observation that the processing of glycoproteins incorporated in the pseudo- VSV may differ from those in coronaviruses due to the difference in virus budding sites. Our study employed the VSV pseudovirus system to investigate the role of N1242 and N1247 glycans in virus infection. Our findings revealed that the removal of the N1242 glycan specifically reduced pseudovirus infectivity, supporting our hypothesis that N1242 glycan plays a role in HCoV- NL63 infection based on structural and computational observations.  
+
+However, we respectfully disagree with the suggestion to conduct glycan analysis of the pseudo- VSV as we believe that such an approach would not provide additional evidence to support our conclusion. Our manuscript utilized an integrated approach using cryoEM, cryoET, MD, and
+
+<--- Page Split --->
+
+glycan analysis of the same HCoV- NL63 virion to arrive at our conclusions. We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology.  
+
+We are currently working on generating mutant HCoV- NL63 viruses using reverse genetics for an independent and rigorous investigation that goes beyond the scope of the current manuscript. Our aim is to conduct integrative biochemical and structural/computational analyses to further elucidate the role of N1242 and N1247 glycans in coronavirus infection. We believe that our current study is technically sound and biologically relevant to coronavirus researchers, given the global health concerns surrounding these viruses.  
+
+Minor points:  
+
+1. Figure 4D is never mentioned in the manuscript.  
+
+Thanks for the suggestion. Figure 4C and 4D together showed different epitope shielding by hinge glycans at different bending angles. We made the correction in the text.  
+
+2. Fig. S2B is not mentioned in the main text. There is no discussion on why/how two spikes are connected in the stalk.  
+
+Thanks for pointing this out and why/how two spikes are connected in the stalk is not clear but it is very rare. We removed Figure S2B in the revised manuscript.  
+
+3. Fig. 1G: the angles are boxed in colors, however, the colors are not explained.  
+
+Thanks for pointing this out and we removed colors of the angle boxes in the revised manuscript.  
+
+4. Adding a comparison animation of the MD simulation of S-trimers with and without the N1242 glycan to the manuscript will be highly appreciated.  
+
+We provided a supplemental movie (supplemental video 2) comparing the bending of HCoV- NL63 spike and of the spike with the two hinge glycans removed (ΔglycN1242/N1247) that has similar profile as spike with N1242 glycan removed (ΔglycN1242).  
+
+5. I'm wondering if the most favorable spike tilt 56o is the optimal tilt conformation, what would be the virological implications of this angle? Would the spike mediate receptor binding better so as to influence viral infectivity? A little discussion expansion on correlation between the spike tilting and viral infectivity will be appreciated.
+
+<--- Page Split --->
+
+We do not yet completely understand the mechanism of how the optimal spike tilt links to virus infectivity. The hypothesis of epitope accessibility and glycan shielding is purely structural. We are in the process of generating mutant HCoV- NL63 viruses for detailed structural, biochemical and virological analyses. In the discussion, we emphasized the necessity of such future studies.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript by Chmielewski et al. provides important knowledge on the molecular dynamics of virion- bound HCoV- NL63 spike glycoprotein (S) and proposes that a specific N- glycan within a flexible hinge region at the stalk- crown interface confers some of the conformational flexibility. By applying focused classification, the authors were able to resolve the otherwise orientationally flexible stalk domain, indiscernible at the complete cryo- EM map. By further isolating image classes with various combinations of crown and stalk domain orientations, the authors also elegantly demonstrate highly variable tilting angles of virion- bound spike, with the crown domain not only tilting, but also rotating relative to the stalk domain, with this flexibility imposed by the hinge region. Intriguingly, in silico removal of the hinge- resident N- glycan at N1242, or all the S glycans, results in a shift of most energetically favorable tilting angles from \(56^{\circ}\) to around \(25^{\circ}\) , decreasing S flexibility, and possibly increasing exposure to antibodies. Moreover, HCoV- NL63 S- pseudotyped VSVdeltaG with said N- glycan deletion exhibits substantially decreased infectivity. The work provides extremely valuable insight into how glycans affect the conformational stability of proteins – a subject that is still underexplored. The authors provide compelling evidence for plausible roles of glycans in HCoV- NL63 biology, which could be further strengthened by in vitro experiments. The glycoproteomic analysis of virion- derived S is a welcome addition for determining predominant site- specific glycan compositions, however, a few details should be clarified.  
+
+1. The mass spectrometry sample and data analysis approach should be clarified. The multiple digests are typically employed to enhance coverage, and it is not unusual to pool them prior to analysis. Is that what was performed here, or were they analyzed separately? Given the semiquantitative nature of spectral counting, were any measures taken to estimate variation (biological or technical replicates)?  
+
+The multiple digests were analyzed separately using the same instrument acquisition and data analysis method. After manual validation, the spectral counts from each digest were combined for quantitation. We consider the multiple digests as technical replicates.  
+
+2. The increased access to antibody molecules upon removal of N-glycans in the flexible hinge region makes sense, as suggested by the molecular modeling experiments (Figure 4C, 4D). However, the software-determined approach measuring the accessible area of the probe, and not of the exposed spike area is somewhat counterintuitive. It would strengthen the manuscript, if an
+
+<--- Page Split --->
+
+in vitro approach was employed to further test the validity of the claims. For example, suitable neutralizing antibodies, or sera from seropositive individuals could be used to measure antibody binding capacity, or neutralization of pseudotyped virions.  
+
+We'd love to do the suggested experiment if there is any neutralizing antibody targeting at the hinge loop available. Due to the lack of anti- HCoV- NL63 neutralizing antibodies, we can only predict the glycan mediated immune evasion using molecular dynamic simulation. We are planning to design molecular binders targeting the hinge loop and test their neutralizing activities against wt and glycan mutants. Like most structural results, they are used to guide better and functionally meaningful experiments.  
+
+3. Figure 5C. The plot elements should be described in the legend (measure of center, error bars, and statistical test). If the data is normalized to the infectivity of wild type, why are there error bars on the wild type data? What is the reproducibility of the data? Pseudovirus infectivity data can be rather variable, and the information on numbers of independent experiments is missing.  
+
+Thank you very much for pointing it out and we have revised the figure legend for Figure 5C accordingly. The bars represent means+SD of triplicate samples. The wild type infectivity was also tested in triplicate and the mean+SD of the wild type was shown. The mutants were normalized to the mean of the wild type. The experiments were repeated at least three times and data from one representative experiment was shown.  
+
+Minor: Lines 254- 255. The term "monosaccharide" should be used instead of "glycan".  
+
+Thanks for the correction.  
+
+Line 319. It would be useful to mention exact numbers for glycan shield density for the other coronaviruses.  
+
+As suggested, we added panels G and H in Figure 3 showing the number of residues on the surface of different coronavirus spikes that are accessible to the Fab fragment of IgG. The less number of residues accessible to Fab, the more immune evasive the virus is. The analysis showed that HCoV- NL63 is more immune evasive (or glycan shielded) than other human coronaviruses.  
+
+Lines 484- 485. How was the "conservation of glycosylation" determined as opposed to amino acid conservation (previous sentence)? Is it based on experimental data from the literature? If so, how many and which viruses? Or is it algorithm- based prediction?
+
+<--- Page Split --->
+
+We analyzed spike sequences from 35 different coronaviruses, at position 1242 and 1247, \(89\%\) and \(74\%\) sequences have ASN and at position 1244 and 1249, \(91\%\) and \(69\%\) of the sequences have either Ser or Thr. The conservation of N- glycosylation is based on the NxS/T sequenc conservation at 1242- 1244 and 1247- 1249. We changed N- linked glycan to N- linked glycosylation sequon in the revised manuscript.  
+
+Figure S7D. Please add amino acid residue numbers for the start/end of the shown ranges.  
+
+Thanks for the suggestion, we added amino acid residue numbers in the sequence alignment shown in the new Figure S9D.  
+
+References Cited  
+
+Ke, Zunlong, Joaquin Oton, Kun Qu, Mirko Cortese, Vojtech Zila, Lesley McKeane, Takanori Nakane, et al. 2020. "Structures and Distributions of SARS- CoV- 2 Spike Proteins on Intact Virions." Nature 588 (7838): 498- 502. Turoňová, Beata, Mateusz Sikora, Christoph Schürmann, Wim J. H. Hagen, Sonja Welsch, Florian E. C. Blanc, Sören von Bülow, et al. 2020. "In Situ Structural Analysis of SARS- CoV- 2 Spike Reveals Flexibility Mediated by Three Hinges." Science 370 (6513): 203- 8. Walls, Alexandra C., M. Alejandra Tortorici, Brandon Frenz, Joost Snijder, Wentao Li, Félix A. Rey, Frank DiMaio, Berend- Jan Bosch, and David Veesler. 2016. "Glycan Shield and Epitope Masking of a Coronavirus Spike Protein Observed by Cryo- Electron Microscopy." Nature Structural & Molecular Biology 23 (10): 899- 905. Yao, Hangping, Yutong Song, Yong Chen, Nanping Wu, Jialu Xu, Chujie Sun, Jiaxing Zhang, et al. 2020. "Molecular Architecture of the SARS- CoV- 2 Virus." Cell 183 (3): 730- 38. e13. Zhang, Kaiming, Shanshan Li, Grigore Pintilie, David Chmielewski, Michael F. Schmid, Graham Simmons, Jing Jin, and Wah Chiu. 2020. "A 3.4- Å Cryo- Electron Microscopy Structure of the Human Coronavirus Spike Trimer Computationally Derived from Vitrified NL63 Virus Particles." ORB Discovery 1 (November): e11.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors have improved the manuscript, however, several concerns are still not fully addressed.  
+
+1. The authors have added the details for virus production and purification. They found a possible misinterpretation of the number of virions in their data. They re-checked the data, and based on a selected pool of 154 intact virions, their counts changed to \(20 \pm 13\) spikes/virion from the original \(\sim 9\) spikes/virion. This number is compared to that of SARS-CoV-2.  
+
+Despite the change of statistical analysis of the spikes/virion, the authors still report "On the virion surface, the average distance between the nearest neighboring spikes is \(\sim 34 \text{nm}\) (Figure 1E) compared to \(\sim 15 \text{nm}\) average nearest distance between prefusion spikes on SARS-CoV-2 virion" (line 175- 178). Also, the purity of viral prep evidenced by the coomassie gel does not exclude the possibility that the spikes may detach from the virions during purification. Therefore, I am not fully convinced unless the authors control for the virion intactness (more specifically the No. of spikes/virion) by cryo- ET of freshly prepared & unconcentrated & not chemically fixed HCoV- NL63 in the supernatant.  
+
+2. The cryo-ET technical notes have now been improved. There are still confusions, e.g. supplementary table 2 says the Voltage is 200kV, however, the MM part says "Cryo-ET data was collected by loading frozen grids into a Thermo Fisher Titan Krios transmission electron microscope operated at 300kV". Also, the tilt scheme shall be added to the table.  
+
+3. Why is the title of the manuscript "Integrated analyses reveal a hinge glycan regulates coronavirus spike tilting and virus infectivity" different from that of the SI "Bending of coronavirus spike regulated by a hinge glycan"?  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors have addressed my concerns.
+
+<--- Page Split --->
+
+Reviewer #3 (Remarks to the Author):  
+
+Chmielewski et al. presented a significant advance in understanding the structural dynamics of the spike protein of HCoV- NL63. Using cryo- ET and subtomogram averaging, the authors revealed a distribution of distinct titled conformations for the spike crown relative to its stalk region. As glycans are essential for the dynamics of spike, the authors used mass spectrometry to determine the site- specific occupancy and percentage of the N- linked glycan types at each site. The results revealed that different from other coronavirus such as SARS- CoV- 2, the glycans on the spike of HCoV- NL63 are predominantly high- mannose glycans. The higher glycan shielding on HCoV- NL63 leads to less accessible residues for Fab fragment of antibody, suggesting that HCoV- NL63 is more immune evasive than other coronaviruses.  
+
+Using structural prediction tools like I- TASSER and AlphaFold, the author constructed a predicted model for the stalk region. MD simulations of the glycosylated spike at different tilt angles generated a similar profile of the spike bending dynamics to the results collected by cryo- ET. There are two glycans, N1242 and N1247, near the hinge region of the spike stalk. The authors found that these two glycans generated the most favorable protein- glycan interactions and the minimum accessible surface area around the most probable bending conformation (\~56o) captured in the cryo- ET results. Removing these glycans, particularly N1242- linked glycan, shifted the distribution of bending conformations and decreased the protein- glycan contacts in the MD simulation. Finally, the authors found that mutation on N1242 (but not N1247) of spike reduced the infection of pseudo- VSV by \~70%.  
+
+This is a comprehensive study of the structural dynamics regarding the spike of HCoV- NL63 and provides important information on its glycans and immune evasion properties. However, there are some concerns that need to be addressed.  
+
+Major concern:  
+
+The evidence for structural dynamics is very compelling in the manuscript. However, the link between the structural dynamics of spike and infectivity is not convincing. Did the N1242D mutant spike generated on VSV- pseudovirus show different distributions on the bending conformations when compared to WT spike if measured by cryo- ET? Furthermore, the reduction of infectivity for N1242 mutant could be due to many reasons. Except for the MD simulation results (generated using the HCoV- NL63 spike, not the pseudo- VSV spike), there is no other evidence indicating the infection reduction is due to loss of the glycan- dependent motion. The response of the authors to Reviewer 1 brings more concerns – if ‘We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology’, then is the N1242D mutant spike generated using pseudo- VSV really a good model to validate the connection between the glycan- dependent motion and the infectivity? In the abstract, the authors quite emphasize the functional aspects of structural dynamics. For example, the abstract starts with ‘How the structure dynamics of the full- length spikes incorporated in viral lipid envelope correlates with the virus infectivity remains poorly understood.’ and ends with ‘Subsequent infectivity assays support the hypothesis that this glycan- dependent motion impacts virus entry.’ There is, in fact, no clear evidence linking the glycan- dependent motion to the virus entry in the manuscript. It is ideal that the authors can provide more evidence, but if the authors think it ‘goes beyond the scope of the current manuscript’, then at least they need to point out the limitation of the pseudo- VSV model
+
+<--- Page Split --->
+
+and tune down the description of the functional aspects (i.e., the infectivity) for the glycan- dependent structural dynamics.  
+
+Other concerns:  
+
+1) In the cryo-ET results, the authors found there are also different azimuthal directions along with different tilting angles. Because there is no preference for the azimuthal direction, the authors focused their analysis on the tilting angles. The authors did not state which azimuthal direction they used for the following MD simulations. Would different azimuthal directions impact the simulation result?  
+
+2) For the distributions of the bending angle probabilities (Fig. 4A and Supplementary Figure 10), the starting angle of 40o is quite close to the most probable angle observed in the cryo-ET results (~56o); however, the distribution peaks at around ~10o after MD simulation, any explanation for that?
+
+<--- Page Split --->
+
+## Re: RESPONSES TO REVIEWER COMMENTS (MS # NCOMMS-23-05304A)  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors have improved the manuscript, however, several concerns are still not fully addressed.  
+
+1. The authors have added the details for virus production and purification. They found a possible misinterpretation of the number of virions in their data. They re-checked the data, and based on a selected pool of 154 intact virions, their counts changed to \(20 \pm 13\) spikes/virion from the original "9 spikes/virion. This number is compared to that of SARS-CoV-2. Despite the change of statistical analysis of the spikes/virion, the authors still report "On the virion surface, the average distance between the nearest neighboring spikes is "34 nm (Figure 1E) compared to "15 nm average nearest distance between prefusion spikes on SARS-CoV-2 virion" (line 175-178). Also, the purity of viral prep evidenced by the Coomassie gel does not exclude the possibility that the spikes may detach from the virions during purification. Therefore, I am not fully convinced unless the authors control for the virion intactness (more specifically the No. of spikes/virion) by cryo-ET of freshly prepared & unconcentrated & not chemically fixed HCoV-NL63 in the supernatant.  
+
+## Responses:  
+
+The concern raised by the reviewer regards the potential fragility of coronavirus spikes leading to spike loss during virus purification, and consequently, the possible lack of representativeness in the observed spike dynamics on the virion surface. We respectfully disagree with this critique and offer the following rationales for our stance.  
+
+A. The employed technique of gradient-purification followed by concentration of virions is a well-established biochemical protocol for preparing coronaviruses suitable for cryogenic electron microscopy (cryo-EM) and tomography (cryo-ET) investigations. This method has been successfully employed in the analysis of the structures of other coronaviruses, such as SARS-CoV, FCoV, MHV (Neuman et al., 2006), and PEDV (Huang et al., 2022). These previous studies, akin to our current approach, utilized virions purified through gradient ultracentrifugation, which effectively maintained the pre-fusion conformation of the spikes without experiencing S1 shedding during the purification process. It is important to note that this purification procedure is necessary
+
+<--- Page Split --->
+
+to yield enough virus particles suitable for high- resolution cryo- EM and cryo- ET structural analysis of spikes.  
+
+B. We would like to underscore that HCoV-NL63 presents distinct challenges in terms of in vitro culture when juxtaposed with SARS-CoV-2, despite sharing the same receptor (Zhu et al., 2020). The diminished in vitro growth efficiency of HCoV-NL63 in comparison to other human coronaviruses that cause common cold, as noted by (Dijkman et al., 2013), makes it arduous to attain sufficiently high titer for direct imaging. It is impractical to carry out the cryo-ET experiment without concentration, as suggested by the reviewer.  
+
+C. Importantly, the bending dynamics of HCoV-NL63 spikes predicted by the MD simulations are in good agreement with our cryo-ET observation of individual spikes on the purified virions in our study (Figure 4A). The quantitative agreement between experimental and in silico estimates of spike bending dynamics suggested our MD simulation system was aptly capturing the conformational ensemble of the spike in situ. This suggests that the bending dynamics of HCoV-NL63 spikes is intrinsic to the individual spike itself. We observed variations in spike numbers on the virion and distances between neighboring spikes. Thus, such biochemical variations observed in our experimental settings appear not to impact on the spike bending dynamics which is one of the primary conclusions in our paper.  
+
+D. We thank the reviewer for accepting the corrected number of spikes/virion based on our first revision. However, with this correction, the spike-spike distance we reported will not change, as it is unrelated to the number of virions in the dataset that the reviewer correctly noted was incorrectly reported in the original manuscript Methods. Instead, the spike-spike distance analysis utilizes manually picked and refined spike subvolumes, where each virion in the tomograms was visually inspected in 2D slices for complete identification of all spikes prior to subtomogram processing. By utilizing refined subvolume orientations of spike stalks, which are anchored in the membrane, for determining nearest neighbor distances on the virion surface, our method represents an advanced quantification method for proteins attached to membranes. In contrast to earlier studies that approximated spike coordinates, our approach offers greater accuracy and is poised to contribute to future investigations into spike-antibody binding interactions and spike-receptor associations. Since there is no ground truth on the number of spikes and their spatial distributions known in any infectious coronavirus, our studies can only be validated by traditional biochemical assays like SDS gel and infectivity. Furthermore, the detailed infection steps are likely to be different between SARS-CoV-2 and HCoV-NL63, as exemplified by the lack of postfusion particles detected in our system as in SARS-CoV-2. HCoV-NL63 S protein is a single chain transmembrane glycoprotein with no cleavage site at S1/S2 boundary that is demonstrated by a single S protein band resolved by SDS-PAGE analysis of purified HCoV-NL63 (Fig. S3B in the
+
+<--- Page Split --->
+
+manuscript) in contrast to mixed S, S1 and S2 protein bands shown in SARS- CoV- 2 (Ke et al., 2020). Trimerized HCoV- NL63 S proteins form spikes in single closed pre- fusion conformation that was observed in both purified soluble spike proteins (Walls et al., 2016) and in situ on purified virions (Zhang et al., 2020).  
+
+E. Finally, we wish to clarify that we do not assert any claims regarding no potential loss of spikes from HCoV-NL63 virions during the standard purification protocol. While we acknowledge the reviewer's suggestion to examine unpurified CoV virions within CoV-infected cells, possibly involving various cell types and distinct biochemical conditions, as an important avenue of research, we concur that this lies beyond the confines of our current study's scope. Our primary message in this manuscript remains the detailed exploration of the structural dynamics of spikes on purified and concentrated HCoV-NL63 virions. Serendipitously, our analysis suggests that the bending of the spikes may modulate the infectivity of a pseudo virus. We believe that our study will inspire many follow-up experiments to validate our hypothetical model of the structure and function relationship of the spike in situ.  
+
+2. The cryo-ET technical notes have now been improved. There are still confusions, e.g. supplementary table 2 says the Voltage is 200kV, however, the MM part says "Cryo-ET data was collected by loading frozen grids into a Thermo Fisher Titan Krios transmission electron microscope operated at 300kV". Also, the tilt scheme shall be added to the table.  
+
+We thank reviewer 1 for finding this inconsistency and have corrected the methods.  
+
+3. Why is the title of the manuscript "Integrated analyses reveal a hinge glycan regulates coronavirus spike tilting and virus infectivity" different from that of the SI "Bending of coronavirus spike regulated by a hinge glycan"?  
+
+We thank reviewer 1 for finding this inconsistency and have corrected the title of the SI. To tone down the claim, we changed our title to "Cryo-ET, Chemical and Computational Analyses Suggest That Hinge Glycans Modulate Coronavirus Spike Tilting and Infectivity".  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors have addressed my concerns. Response: We thank reviewer 2 for all the previous comments to improve the manuscript.  
+
+Reviewer #3 (Remarks to the Author):
+
+<--- Page Split --->
+
+Chmielewski et al. presented a significant advance in understanding the structural dynamics of the spike protein of HCoV- NL63. Using cryo- ET and subtomogram averaging, the authors revealed a distribution of distinct titled conformations for the spike crown relative to its stalk region. As glycans are essential for the dynamics of spike, the authors used mass spectrometry to determine the site- specific occupancy and percentage of the N- linked glycan types at each site. The results revealed that different from other coronavirus such as SARS- CoV- 2, the glycans on the spike of HCoV- NL63 are predominantly high- mannose glycans. The higher glycan shielding on HCoV- NL63 leads to less accessible residues for Fab fragment of antibody, suggesting that HCoV- NL63 is more immune evasive than other coronaviruses. Using structural prediction tools like I- TASSER and AlphaFold, the author constructed a predicted model for the stalk region. MD simulations of the glycosylated spike at different tilt angles generated a similar profile of the spike bending dynamics to the results collected by cryo- ET. There are two glycans, N1242 and N1247, near the hinge region of the spike stalk. The authors found that these two glycans generated the most favorable protein- glycan interactions and the minimum accessible surface area around the most probable bending conformation (\~56o) captured in the cryo- ET results. Removing these glycans, particularly N1242- linked glycan, shifted the distribution of bending conformations and decreased the protein- glycan contacts in the MD simulation. Finally, the authors found that mutation on N1242 (but not N1247) of spike reduced the infection of pseudo- VSV by \~70%.  
+
+This is a comprehensive study of the structural dynamics regarding the spike of HCoV- NL63 and provides important information on its glycans and immune evasion properties. However, there are some concerns that need to be addressed.  
+
+Major concern:  
+
+The evidence for structural dynamics is very compelling in the manuscript.  
+
+Response: We thank the Reviewer for the overall positive evaluation of our work.  
+
+However, the link between the structural dynamics of spike and infectivity is not convincing. Did the N1242D mutant spike generated on VSV- pseudovirus show different distributions on the bending conformations when compared to WT spike if measured by cryo- ET? Furthermore, the reduction of infectivity for N1242 mutant could be due to many reasons. Except for the MD simulation results (generated using the HCoV- NL63 spike, not the pseudo- VSV spike), there is no other evidence indicating the infection reduction is due to loss of the glycan- dependent motion.  
+
+Response: We agree with the Reviewer that detailed analysis of a potential causal relationship between HCoV- NL63 infectivity and the bending of the spike is yet to be established. The correlation between these events is purely empirical as we have clearly stated in the
+
+<--- Page Split --->
+
+Limitations section about the modeling. Following the Reviewer's cogent thought, we have further toned down this narrative across the article. In addition, we changed our manuscript title to reflect our approach and conclusion as "Cryo- ET, Chemical and Computational Analyses Suggest That Hinge Glycans Modulate Coronavirus Spike Tilting and Infectivity".  
+
+The response of the authors to Reviewer 1 brings more concerns – if 'We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology', then is the N1242D mutant spike generated using pseudo- VSV really a good model to validate the connection between the glycan- dependent motion and the infectivity? In the abstract, the authors quite emphasize the functional aspects of structural dynamics. For example, the abstract starts with 'How the structure dynamics of the full- length spikes incorporated in viral lipid envelope correlates with the virus infectivity remains poorly understood.' and ends with 'Subsequent infectivity assays support the hypothesis that this glycan- dependent motion impacts virus entry.' There is, in fact, no clear evidence linking the glycan- dependent motion to the virus entry in the manuscript. It is ideal that the authors can provide more evidence, but if the authors think it 'goes beyond the scope of the current manuscript', then at least they need to point out the limitation of the pseudo- VSV model and tune down the description of the functional aspects (i.e., the infectivity) for the glycan- dependent structural dynamics.  
+
+Responses: Thank you for the question on the statement we used in the manuscript and the original rebuttal. We have modified the sentence in the revised abstract as the following: "Subsequent infectivity assays implicated involvement of N1242- glyan in virus entry". The assay we used is indeed a standard procedure to measure viral spike- mediated entry. Hope that our revision clarifies the misunderstanding due to our original wording.  
+
+Following the Reviewer's suggestions, we have further revised the article to ensure that no statements remain that structural dynamics of the spike are correlated with the infectivity assays.  
+
+Other concerns:  
+
+1) In the cryo-ET results, the authors found there are also different azimuthal directions along with different tilting angles. Because there is no preference for the azimuthal direction, the authors focused their analysis on the tilting angles. The authors did not state which azimuthal direction they used for the following MD simulations. Would different azimuthal directions impact the simulation result?  
+
+Responses: We thank the Reviewer for this insightful suggestion. Indeed, MD simulations were initially performed with a range of tilt angles but a single azimuthal angle which was arbitrarily
+
+<--- Page Split --->
+
+chosen. Our results were shown in Fig 4A in the manuscript. Subsequently, using 3.5 microsecond- long MD simulations, a broad range of azimuthal angles (between 0 to \(60^{\circ}\) ) are sampled that are now presented in Figure 1 below (Fig. S11 in the SI), which is also broader relative to the ones previously observed in MD simulations of SARS- CoV- 2 spikes (Fig. 3B from (Kapoor et al., 2022)). However, given the higher diffusion barriers associated with azimuthal over tilting changes, the entire range of the azimuthal space (0 to \(120^{\circ}\) as seen in cryo- ET) is not captured in our finite time MD simulations. This limited sampling has further led to an artificial minimum around the azimuthal angle of \(\sim 20^{\circ}\) , which will potentially even out with longer simulations. Nonetheless, as stated in the subsequent response the dynamics of the azimuthal angle is highly decoupled from that of the tilt. This decoupling between the two angular changes is observed from the maps, where despite an even distribution of the azimuthal angle, the tilt was prominently peaked around \(56^{\circ}\) . These caveats are now mentioned in the Limitations section.  
+
+The azimuthal angle controls the direction along which the bending occurs, while the tilt quantifies the magnitude of this bending. Our cryo- EM data shows that the spikes are dispersed so that the inter- spike dynamics (and hence the azimuthal changes) does not affect spike bending. In additional analysis that we now provide (Figure 1), we find that despite finite sampling there is little correlation between the azimuthal and tilt dynamics in the MD simulations.  
+
+![](images/Figure_2.jpg)
+
+<center>Figure 1. Distribution of azimuthal and tilt angles sampled during MD simulations, and a scatter plot showing minimal correlation between the dynamics of these two angles during MD. </center>  
+
+2) For the distributions of the bending angle probabilities (Fig. 4A and Supplementary Figure 10), the starting angle of \(40^{\circ}\) is quite close to the most probable angle observed in the cryo-ET results ( \(\sim 56^{\circ}\) ); however, the distribution peaks at around \(\sim 10\) after MD simulation, any explanation for that?
+
+<--- Page Split --->
+![PLACEHOLDER_21_0]
+
+<center>Figure 2. Bending angle by replica for simulations using \(40^{\circ}\) map </center>  
+
+Response: We have now investigated this data further. Presented in Figure 2, we find that two of the three replicas were indeed moving towards higher angles. There was one replica nonetheless that moved towards lower angles creating the peak. The cumulative area under the curve for drift towards \(56^{\circ}\) is still higher making the \(\sim 56^{\circ}\) region more probable.  
+
+## References Cited  
+
+Dijkman, R., Jebbink, M. F., Koekkoek, S. M., Deijs, M., Jonsdottir, H. R., Molenkamp, R., leven, M., Goossens, H., Thiel, V., & van der Hoek, L. (2013). Isolation and characterization of current human coronavirus strains in primary human epithelial cell cultures reveal differences in target cell tropism. Journal of Virology, 87(11), 6081- 6090. Huang, C.- Y., Draczkowski, P., Wang, Y.- S., Chang, C.- Y., Chien, Y.- C., Cheng, Y.- H., Wu, Y.- M., Wang, C.- H., Chang, Y.- C., Chang, Y.- C., Yang, T.- J., Tsai, Y.- X., Khoo, K.- H., Chang, H.- W., & Hsu, S.- T. D. (2022). In situ structure and dynamics of an alphacoronavirus spike protein by cryo- ET and cryo- EM. Nature Communications, 13(1), 4877. Kapoor, K., Chen, T., & Tajkhorshid, E. (2022). Posttranslational modifications optimize the ability of SARS- CoV- 2 spike for effective interaction with host cell receptors. Proceedings of the National Academy of Sciences of the United States of America, 119(28), e2119761119. Ke, Z., Oton, J., Qu, K., Cortese, M., Zila, V., McKeane, L., Nakane, T., Zivanov, J., Neufeldt, C. J., Cerikan, B., Lu, J. M., Peukes, J., Xiong, X., Krausslich, H.- G., Scheres, S. H. W., Bartenschlager, R., & Briggs, J. A. G. (2020). Structures and distributions of SARS- CoV- 2 spike proteins on intact virions. Nature, 588(7838), 498- 502. Neuman, B. W., Adair, B. D., Yoshioka, C., Quispe, J. D., Orca, G., Kuhn, P., Milligan, R. A., Yeager, M., & Buchmeier, M. J. (2006). Supramolecular architecture of severe acute respiratory syndrome
+
+<--- Page Split --->
+
+coronavirus revealed by electron cryomicroscopy. Journal of Virology, 80(16), 7918- 7928. Walls, A. C., Tortorici, M. A., Frenz, B., Snijder, J., Li, W., Rey, F. A., DiMaio, F., Bosch, B.- J., & Veesler, D. (2016). Glycan shield and epitope masking of a coronavirus spike protein observed by cryo- electron microscopy. Nature Structural & Molecular Biology, 23(10), 899- 905. Zhang, K., Li, S., Pintilie, G., Chmielewski, D., Schmid, M. F., Simmons, G., Jin, J., & Chiu, W. (2020). A 3.4- Å cryo- electron microscopy structure of the human coronavirus spike trimer computationally derived from vitrified NL63 virus particles. QRB Discovery, 1, e11. Zhu, N., Wang, W., Liu, Z., Liang, C., Wang, W., Ye, F., Huang, B., Zhao, L., Wang, H., Zhou, W., Deng, Y., Mao, L., Su, C., Qiang, G., Jiang, T., Zhao, J., Wu, G., Song, J., & Tan, W. (2020). Morphogenesis and cytopathic effect of SARS- CoV- 2 infection in human airway epithelial cells. Nature Communications, 11(1), 3910.
+
+<--- Page Split --->
+
+
+<--- Page Split --->
+
+REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+Filtering the coronavirus through the membrane filter is unconventional and can detach the spikes. The control experiment to clarify this is cryo- ET of unconcentrated HCoV- NL63 in the supernatant, images of a few virions would do. According to our experiences, this can be done on SARS- CoV- 2 and other seasonal human coronaviruses. If the authors can not supply this, they shall add a statement of the above risks to the Limitations part.  
+
+I have no further comments to the other parts.  
+
+Reviewer #3 (Remarks to the Author):  
+
+The authors addressed all my concerns in the revised manuscript.
+
+<--- Page Split --->
+
+## Re: RESPONSES TO REVIEWER COMMENTS (MS # NCOMMS-23-05304B)  
+
+Reviewer #1 (Remarks to the Author):  
+
+Filtering the coronavirus through the membrane filter is unconventional and can detach the spikes. The control experiment to clarify this is cryo- ET of unconcentrated HCoV- NL63 in the supernatant, images of a few virions would do. According to our experiences, this can be done on SARS- CoV- 2 and other seasonal human coronaviruses. If the authors can not supply this, they shall add a statement of the above risks to the Limitations part.  
+
+I have no further comments to the other parts.  
+
+## Responses:  
+
+We want to thank this reviewer for the suggestion. Unfortunately, the virus titer in unconcentrated culture supernatant of HCoV- NL63 infected cells is too low to direct image without virus concentration/purification. We added this limitation in the discussion of the revised manuscript.  
+
+Reviewer #3 (Remarks to the Author):  
+
+The authors addressed all my concerns in the revised manuscript.  
+
+Responses:  
+
+Thanks.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d5cd42613e42af1cb6af7d1aa2807cdfdd9681d47c9231fe8f1b4ff73a62a0d1/supplementary_0_Peer Review File__d5cd42613e42af1cb6af7d1aa2807cdfdd9681d47c9231fe8f1b4ff73a62a0d1_det.mmd

@@ -0,0 +1,512 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[68, 110, 361, 140]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[70, 154, 885, 212]]<|/det|>
+Structural insights into the modulation of coronavirus spike tilting and infectivity by hinge glycans  
+
+<|ref|>image<|/ref|><|det|>[[56, 732, 240, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 912, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 147, 393, 163]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[113, 202, 879, 440]]<|/det|>
+HCoV- NL63 is an endemic seasonal human coronavirus which shares the same huACE2 receptor with SARS- CoV- 2. Compared to SARS- CoV- 2, HCoV- NL63 shows a lack of adaptive evolution, and is more glycosylated on its spikes. In this manuscript, Chmielewski et al. have integrated multiple experimental and computational approaches to discover and demonstrate the impact of a single glycan on the dynamics and infectivity of HCoV- NL63. The authors have imaged unfixed virions by cryo- ET and analyzed its spike protein structures and statistics. Combined with MS, the glycan compositions were analyzed and an atomic model of the most abundant glycan topology was built into the density of their previous single particle cryoEM map, giving them the opportunity in analyzing the glycan shielding. Atomic models of the ectodomain were built for glycosylated spike resolved at tilt angles of 100, 200...700, which were subjected to MD analysis. The simulation has revealed the bending profile of the spike, revealing that two N- glycans (N1242 &1247) near this disordered hinge region may have important roles in spike tilting. Further MD and infectivity assay using a VSV pseudovirus have revealed a role of the conserved N- linked glycan N1242 in both spike bending and infectivity.  
+
+<|ref|>text<|/ref|><|det|>[[115, 450, 872, 560]]<|/det|>
+Overall, the discoveries presented in this work is novel, and the logic among various experimental approaches is fluent. The authors have also discussed the limitations in this work, such as uncertainties in building the atomic model for MD and short simulation duration, which is helpful for readers. My main concerns locate at the sample preparation procedure and infectivity assay using the VSV pseudovirus. The manuscript needs major revision before being recommend for publication in Nature Communications.  
+
+<|ref|>text<|/ref|><|det|>[[115, 599, 215, 615]]<|/det|>
+Major points:  
+
+<|ref|>text<|/ref|><|det|>[[114, 655, 881, 874]]<|/det|>
+1. The authors pelted the unfixed virions through \(20\%\) sucrose cushion followed by OptiPrep gradient, then filtered the virions through a membrane filter to exchange the buffer. The pelleting and filtering raise concerns on the intactness of the virions, as the thermal stability of coronaviruses and their spike proteins are known to be weak (especially when they are not chemically fixed). Also, for how long and at what temperature was the virion sample stored during each step before plunge-freezing? This is not clearly written in the method part on virus preparation. These possibly result in only \(\sim 9\) spikes (according to line 130) found per virion and some virions look bald (fig.1A), this number is significantly less than that of SARS-CoV-2 and PEDV (\~30 spikes/virion). This suggests that the virions are possibly damaged during sample preparation. If so, the average distance between neighboring spikes is overestimated, and this will impact the spike dynamics analysis, since the spikes can contact each other on an intact virion. The authors shall analyze this possibility and provide a control for the virion intactness by cryo-ET of freshly prepared & unconcentrated HCoV-NL63 in the supernatant.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 870, 126]]<|/det|>
+2. The cryo-ET data processing part is not clearly stated. Detailing this part in the M&M session will not only justify the robustness of their structures, but also help the readers in learning cryo-ET:  
+
+<|ref|>text<|/ref|><|det|>[[115, 136, 877, 191]]<|/det|>
+(1) In the M&M part, the authors wrote: "approximately 15,000 individual spike particles were manually picked... Subtomogram averaging was then performed using \(\sim 10,000\) particles". However, in line 133, they wrote "Subtomogram averaging analysis of 18,356 spikes". The numbers are inconsistent.  
+
+<|ref|>text<|/ref|><|det|>[[115, 201, 861, 237]]<|/det|>
+(2) What software did they use for subtomogram averaging? How was the "focused classification" (as stated in line 153) done?  
+
+<|ref|>text<|/ref|><|det|>[[115, 248, 861, 283]]<|/det|>
+(3) A supplementary table containing detailed information on cryo-ET data acquisition and processing must be provided.  
+
+<|ref|>text<|/ref|><|det|>[[115, 294, 772, 312]]<|/det|>
+3. Why is the cryo-ET determined spike with a bending angle of 80o not modeled for MD?  
+
+<|ref|>text<|/ref|><|det|>[[114, 322, 870, 451]]<|/det|>
+4. The influence of the hinge glycan N1242 on infectivity was analyzed using VSV-Luc reporter pseudovirus. While coronaviruses bud in ERGIC, where N-glycosylation gets further modified, VSV buds at the plasma membrane. In addition, coronaviral E protein has a function of intracellular retention, which helps concentrating viral structural proteins at the ERGIC to favor viral particle assembly. Therefore, the difference in budding and the missing E from the VSV pseudovirus can result in differences in glycosylation. The authors shall analyze the glycan compositions of the VSV pseudovirus, especially for the N1242 and N1247 sites, to control for the possible differences in glycosylation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 462, 216, 478]]<|/det|>
+Minor points:  
+
+<|ref|>text<|/ref|><|det|>[[115, 490, 486, 507]]<|/det|>
+1. Figure 4D is never mentioned in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 517, 814, 553]]<|/det|>
+2. Fig. S2B is not mentioned in the main text. There is no discussion on why/how two spikes are connected in the stalk.  
+
+<|ref|>text<|/ref|><|det|>[[115, 564, 701, 582]]<|/det|>
+3. Fig. 1G: the angles are boxed in colors, however, the colors are not explained.  
+
+<|ref|>text<|/ref|><|det|>[[115, 593, 875, 629]]<|/det|>
+4. Adding a comparison animation of the MD simulation of S-trimers with and without the N1242 glycan to the manuscript will be highly appreciated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 639, 880, 712]]<|/det|>
+5. I'm wondering if the most favorable spike tilt 56o is the optimal tilt conformation, what would be the virological implications of this angle? Would the spike mediate receptor binding better so as to influence viral infectivity? A little discussion expansion on correlation between the spike tilting and viral infectivity will be appreciated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 781, 393, 797]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 836, 857, 890]]<|/det|>
+The manuscript by Chmielewski et al. provides important knowledge on the molecular dynamics of virion- bound HCoV- NL63 spike glycoprotein (S) and proposes that a specific N- glycan within a flexible hinge region at the stalk- crown interface confers some of the conformational flexibility. By applying
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 880, 345]]<|/det|>
+focused classification, the authors were able to resolve the otherwise orientationally flexible stalk domain, indiscernible at the complete cryo- EM map. By further isolating image classes with various combinations of crown and stalk domain orientations, the authors also elegantly demonstrate highly variable tilting angles of virion- bound spike, with the crown domain not only tilting, but also rotating relative to the stalk domain, with this flexibility imposed by the hinge region. Intriguingly, in silico removal of the hinge- resident N- glycan at N1242, or all the S glycans, results in a shift of most energetically favorable tilting angles from \(56^{\circ}\) to around \(25^{\circ}\) , decreasing S flexibility, and possibly increasing exposure to antibodies. Moreover, HCoV- NL63 S- pseudotyped VSVdeltaG with said N- glycan deletion exhibits substantially decreased infectivity. The work provides extremely valuable insight into how glycans affect the conformational stability of proteins – a subject that is still underexplored. The authors provide compelling evidence for plausible roles of glycans in HCoV- NL63 biology, which could be further strengthened by in vitro experiments. The glycoproteomic analysis of virion- derived S is a welcome addition for determining predominant site- specific glycan compositions, however, a few details should be clarified.  
+
+<|ref|>text<|/ref|><|det|>[[115, 411, 872, 484]]<|/det|>
+1. The mass spectrometry sample and data analysis approach should be clarified. The multiple digests are typically employed to enhance coverage, and it is not unusual to pool them prior to analysis. Is that what was performed here, or were they analyzed separately? Given the semi-quantitative nature of spectral counting, were any measures taken to estimate variation (biological or technical replicates)?  
+
+<|ref|>text<|/ref|><|det|>[[114, 522, 880, 652]]<|/det|>
+2. The increased access to antibody molecules upon removal of N-glycans in the flexible hinge region makes sense, as suggested by the molecular modelling experiments (Figure 4C, 4D). However, the software-determined approach measuring the accessible area of the probe, and not of the exposed spike area is somewhat counterintuitive. It would strengthen the manuscript, if an in vitro approach was employed to further test the validity of the claims. For example, suitable neutralizing antibodies, or sera from seropositive individuals could be used to measure antibody binding capacity, or neutralization of pseudotyped virions.  
+
+<|ref|>text<|/ref|><|det|>[[115, 690, 866, 763]]<|/det|>
+3. Figure 5C. The plot elements should be described in the legend (measure of center, error bars, and statistical test). If the data is normalized to the infectivity of wild type, why are there error bars on the wild type data? What is the reproducibility of the data? Pseudovirus infectivity data can be rather variable, and the information on numbers of independent experiments is missing.  
+
+<|ref|>text<|/ref|><|det|>[[115, 831, 166, 846]]<|/det|>
+Minor:  
+
+<|ref|>text<|/ref|><|det|>[[115, 858, 694, 875]]<|/det|>
+Lines 254- 255. The term "monosaccharide" should be used instead of "glycan".
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 794, 125]]<|/det|>
+Line 319. It would be useful to mention exact numbers for glycan shield density for the other coronaviruses.  
+
+<|ref|>text<|/ref|><|det|>[[115, 165, 877, 219]]<|/det|>
+Lines 484- 485. How was the "conservation of glycosylation" determined as opposed to amino acid conservation (previous sentence)? Is it based on experimental data from the literature? If so, how many and which viruses? Or is it algorithm- based prediction?  
+
+<|ref|>text<|/ref|><|det|>[[112, 286, 777, 304]]<|/det|>
+Figure S7D. Please add amino acid residue numbers for the start/end of the shown ranges.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 146, 350, 164]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 167, 420, 184]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[113, 204, 875, 605]]<|/det|>
+HCov- NL63 is an endemic seasonal human coronavirus which shares the same huACE2 receptor with SARS- CoV- 2. Compared to SARS- CoV- 2, HCoV- NL63 shows a lack of adaptive evolution, and is more glycosylated on its spikes. In this manuscript, Chmielewski et al. have integrated multiple experimental and computational approaches to discover and demonstrate the impact of a single glycan on the dynamics and infectivity of HCoV- NL63. The authors have imaged unfixed virions by cryo- ET and analyzed its spike protein structures and statistics. Combined with MS, the glycan compositions were analyzed and an atomic model of the most abundant glycan topology was built into the density of their previous single particle cryoEM map, giving them the opportunity in analyzing the glycan shielding. Atomic models of the ectodomain were built for glycosylated spike resolved at tilt angles of 10o, 20o...70o, which were subjected to MD analysis. The simulation has revealed the bending profile of the spike, revealing that two N- glycans (N1242 &1247) near this disordered hinge region may have important roles in spike tilting. Further MD and infectivity assay using a VSV pseudovirus have revealed a role of the conserved N- linked glycan N1242 in both spike bending and infectivity. Overall, the discoveries presented in this work is novel, and the logic among various experimental approaches is fluent. The authors have also discussed the limitations in this work, such as uncertainties in building the atomic model for MD and short simulation duration, which is helpful for readers. My main concerns locate at the sample preparation procedure and infectivity assay using the VSV pseudovirus. The manuscript needs major revision before being recommend for publication in Nature Communications.  
+
+<|ref|>text<|/ref|><|det|>[[115, 627, 222, 644]]<|/det|>
+Major points:  
+
+<|ref|>text<|/ref|><|det|>[[114, 664, 878, 886]]<|/det|>
+1. The authors pelleted the unfixed virions through \(20\%\) sucrose cushion followed by OptiPrep gradient, then filtered the virions through a membrane filter to exchange the buffer. The pelleting and filtering raise concerns on the intactness of the virions, as the thermal stability of coronaviruses and their spike proteins are known to be weak (especially when they are not chemically fixed). Also, for how long and at what temperature was the virion sample stored during each step before plunge-freezing? This is not clearly written in the method part on virus preparation. These possibly result in only \(\sim 9\) spikes (according to line 130) found per virion and some virions look bald (fig.1A), this number is significantly less than that of SARS-CoV-2 and PEDV ( \(\sim 30\) spikes/virion). This suggests that the virions are possibly damaged during sample preparation. If so, the average distance between neighboring spikes is over-estimated, and this will impact the spike dynamics analysis, since the spikes can contact each other on an intact
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 880, 128]]<|/det|>
+virion. The authors shall analyze this possibility and provide a control for the virion intactness by cryo- ET of freshly prepared & unconcentrated HCoV- NL63 in the supernatant.  
+
+<|ref|>text<|/ref|><|det|>[[113, 150, 877, 510]]<|/det|>
+We thank the reviewer for pointing out the missing details of virus preparation. We have added a detailed description about temperature/time during virus purification in the Materials and Methods. Our previous draft may have led to a possible misinterpretation of the number of virions in our tomograms from which spikes were actually analyzed by subtomogram averaging. We returned to the data and determined the number of spikes per particle by manual identification of spikes within tomograms. In our purified HCoV- NL63 virion, there are \(20 \pm 13\) spikes/HCoV- NL63 virions in a selected pool of 154 intact virus particles. This number is similar to what is reported for spikes in infusion conformation on SARS- CoV- 2 virion: \(23 \pm 9\) in (Ke et al. 2020) and \(26 \pm 15\) in (Yao et al. 2020), but lower than \(\sim 40\) spikes per virion reported in (Turoňová et al. 2020). Unlike betacoronaviruses, HCoV- NL63 spike is not cleaved between S1 and S2, and the spike displays only one closed prefusion conformation observed in both purified soluble spike proteins (Walls et al. 2016) and in situ on purified virions (Zhang et al. 2020). On SARS- CoV- 2 virus particles, not only prefusion but also postfusion spikes were observed, and the prefusion spikes display both open and closed conformations. The high flexibility of the SARS- CoV- 2 spikes may explain the loss of prefusion spikes in open conformation after centrifugation through sucrose cushion as reported in (Ke et al. 2020). The purity of our HCoV- NL63 virus prep and stability of HCoV- NL63 spike are shown in the coomassie gel in new Figure S3B, appearing to be more stable than betacoronavirus spikes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 530, 874, 588]]<|/det|>
+2. The cryo-ET data processing part is not clearly stated. Detailing this part in the M&M session will not only justify the robustness of their structures, but also help the readers in learning cryo-ET:  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 866, 669]]<|/det|>
+(1) In the M&M part, the authors wrote: "approximately 15,000 individual spike particles were manually picked... Subtomogram averaging was then performed using \(\sim 10,000\) particles". However, in line 133, they wrote "Subtomogram averaging analysis of 18,356 spikes". The numbers are inconsistent.  
+
+<|ref|>text<|/ref|><|det|>[[115, 691, 872, 809]]<|/det|>
+We have revised our manuscript to clarify the data processing method. The same dataset was in fact processed twice using different versions of the EMAN2 software package resulting in different numbers of spikes. The results shown in the paper are from the final version, but some of the numbers in the original paper were not updated and thus incorrect. We have now updated the numbers in the main text and the method section. Our final analysis for subtomogram averaging included 18,356 spikes in both intact and apparently broken virions.  
+
+<|ref|>text<|/ref|><|det|>[[115, 831, 774, 870]]<|/det|>
+(2) What software did they use for subtomogram averaging? How was the "focused classification" (as stated in line 153) done?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 842, 128]]<|/det|>
+We revised the method of subtomogram averaging and provided a supplemental figure (new Figure S1) showing the processing details of focused classification.  
+
+<|ref|>text<|/ref|><|det|>[[115, 150, 835, 188]]<|/det|>
+(3) A supplementary table containing detailed information on cryo-ET data acquisition and processing must be provided.  
+
+<|ref|>text<|/ref|><|det|>[[115, 210, 857, 249]]<|/det|>
+As suggested, we provided a supplemental table showing detailed information on cryoET data acquisition and processing.  
+
+<|ref|>text<|/ref|><|det|>[[119, 270, 833, 288]]<|/det|>
+3. Why is the cryo-ET determined spike with a bending angle of \(80^{\circ}\) not modeled for MD?  
+
+<|ref|>text<|/ref|><|det|>[[114, 310, 881, 489]]<|/det|>
+The simulations starting from the \(70^{\circ}\) bend had \(73.5\%\) of the trajectory going towards lower angles, and only \(26.5\%\) going to higher bent conformations. Furthermore, Of the trajectories with bending angles transitioning above \(70^{\circ}\) , only \(7\%\) (out of the aforementioned \(26.5\%\) ) reached all the way to \(80^{\circ}\) ( \(1.8\%\) of all simulations starting at \(70^{\circ}\) or \(0.2\%\) of all measured frames). None of the trajectories starting from other angles (i.e. \(< 70^{\circ}\) ) make it to \(80^{\circ}\) . On the other end of the bending range, \(5.8\%\) of all trajectories reached less than \(10^{\circ}\) (i.e. \(10^{\circ}\) to \(0^{\circ}\) ) which is still an order of magnitude more than that observed over the range from \(70^{\circ}\) to \(80^{\circ}\) . Therefore, our simulations encompassed extremities of bent conformations, even without starting from a low- probability model (i.e. \(80^{\circ}\) ).  
+
+<|ref|>text<|/ref|><|det|>[[114, 510, 875, 670]]<|/det|>
+4. The influence of the hinge glycan N1242 on infectivity was analyzed using VSV-Luc reporter pseudovirus. While coronaviruses bud in ERGIC, where N-glycosylation gets further modified, VSV buds at the plasma membrane. In addition, coronaviral E protein has a function of intracellular retention, which helps concentrating viral structural proteins at the ERGIC to favor viral particle assembly. Therefore, the difference in budding and the missing E from the VSV pseudovirus can result in differences in glycosylation. The authors shall analyze the glycan compositions of the VSV pseudovirus, especially for the N1242 and N1247 sites, to control for the possible differences in glycosylation.  
+
+<|ref|>text<|/ref|><|det|>[[114, 691, 870, 808]]<|/det|>
+We concur with the reviewer's observation that the processing of glycoproteins incorporated in the pseudo- VSV may differ from those in coronaviruses due to the difference in virus budding sites. Our study employed the VSV pseudovirus system to investigate the role of N1242 and N1247 glycans in virus infection. Our findings revealed that the removal of the N1242 glycan specifically reduced pseudovirus infectivity, supporting our hypothesis that N1242 glycan plays a role in HCoV- NL63 infection based on structural and computational observations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 831, 875, 888]]<|/det|>
+However, we respectfully disagree with the suggestion to conduct glycan analysis of the pseudo- VSV as we believe that such an approach would not provide additional evidence to support our conclusion. Our manuscript utilized an integrated approach using cryoEM, cryoET, MD, and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 866, 150]]<|/det|>
+glycan analysis of the same HCoV- NL63 virion to arrive at our conclusions. We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology.  
+
+<|ref|>text<|/ref|><|det|>[[114, 169, 874, 289]]<|/det|>
+We are currently working on generating mutant HCoV- NL63 viruses using reverse genetics for an independent and rigorous investigation that goes beyond the scope of the current manuscript. Our aim is to conduct integrative biochemical and structural/computational analyses to further elucidate the role of N1242 and N1247 glycans in coronavirus infection. We believe that our current study is technically sound and biologically relevant to coronavirus researchers, given the global health concerns surrounding these viruses.  
+
+<|ref|>text<|/ref|><|det|>[[115, 311, 224, 328]]<|/det|>
+Minor points:  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 516, 348]]<|/det|>
+1. Figure 4D is never mentioned in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 369, 842, 409]]<|/det|>
+Thanks for the suggestion. Figure 4C and 4D together showed different epitope shielding by hinge glycans at different bending angles. We made the correction in the text.  
+
+<|ref|>text<|/ref|><|det|>[[115, 429, 875, 469]]<|/det|>
+2. Fig. S2B is not mentioned in the main text. There is no discussion on why/how two spikes are connected in the stalk.  
+
+<|ref|>text<|/ref|><|det|>[[115, 490, 877, 529]]<|/det|>
+Thanks for pointing this out and why/how two spikes are connected in the stalk is not clear but it is very rare. We removed Figure S2B in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 550, 748, 569]]<|/det|>
+3. Fig. 1G: the angles are boxed in colors, however, the colors are not explained.  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 877, 609]]<|/det|>
+Thanks for pointing this out and we removed colors of the angle boxes in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 630, 832, 670]]<|/det|>
+4. Adding a comparison animation of the MD simulation of S-trimers with and without the N1242 glycan to the manuscript will be highly appreciated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 690, 872, 750]]<|/det|>
+We provided a supplemental movie (supplemental video 2) comparing the bending of HCoV- NL63 spike and of the spike with the two hinge glycans removed (ΔglycN1242/N1247) that has similar profile as spike with N1242 glycan removed (ΔglycN1242).  
+
+<|ref|>text<|/ref|><|det|>[[115, 771, 880, 850]]<|/det|>
+5. I'm wondering if the most favorable spike tilt 56o is the optimal tilt conformation, what would be the virological implications of this angle? Would the spike mediate receptor binding better so as to influence viral infectivity? A little discussion expansion on correlation between the spike tilting and viral infectivity will be appreciated.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 872, 169]]<|/det|>
+We do not yet completely understand the mechanism of how the optimal spike tilt links to virus infectivity. The hypothesis of epitope accessibility and glycan shielding is purely structural. We are in the process of generating mutant HCoV- NL63 viruses for detailed structural, biochemical and virological analyses. In the discussion, we emphasized the necessity of such future studies.  
+
+<|ref|>text<|/ref|><|det|>[[115, 190, 419, 207]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[113, 230, 875, 590]]<|/det|>
+The manuscript by Chmielewski et al. provides important knowledge on the molecular dynamics of virion- bound HCoV- NL63 spike glycoprotein (S) and proposes that a specific N- glycan within a flexible hinge region at the stalk- crown interface confers some of the conformational flexibility. By applying focused classification, the authors were able to resolve the otherwise orientationally flexible stalk domain, indiscernible at the complete cryo- EM map. By further isolating image classes with various combinations of crown and stalk domain orientations, the authors also elegantly demonstrate highly variable tilting angles of virion- bound spike, with the crown domain not only tilting, but also rotating relative to the stalk domain, with this flexibility imposed by the hinge region. Intriguingly, in silico removal of the hinge- resident N- glycan at N1242, or all the S glycans, results in a shift of most energetically favorable tilting angles from \(56^{\circ}\) to around \(25^{\circ}\) , decreasing S flexibility, and possibly increasing exposure to antibodies. Moreover, HCoV- NL63 S- pseudotyped VSVdeltaG with said N- glycan deletion exhibits substantially decreased infectivity. The work provides extremely valuable insight into how glycans affect the conformational stability of proteins – a subject that is still underexplored. The authors provide compelling evidence for plausible roles of glycans in HCoV- NL63 biology, which could be further strengthened by in vitro experiments. The glycoproteomic analysis of virion- derived S is a welcome addition for determining predominant site- specific glycan compositions, however, a few details should be clarified.  
+
+<|ref|>text<|/ref|><|det|>[[115, 610, 860, 710]]<|/det|>
+1. The mass spectrometry sample and data analysis approach should be clarified. The multiple digests are typically employed to enhance coverage, and it is not unusual to pool them prior to analysis. Is that what was performed here, or were they analyzed separately? Given the semiquantitative nature of spectral counting, were any measures taken to estimate variation (biological or technical replicates)?  
+
+<|ref|>text<|/ref|><|det|>[[115, 730, 860, 789]]<|/det|>
+The multiple digests were analyzed separately using the same instrument acquisition and data analysis method. After manual validation, the spectral counts from each digest were combined for quantitation. We consider the multiple digests as technical replicates.  
+
+<|ref|>text<|/ref|><|det|>[[115, 810, 875, 889]]<|/det|>
+2. The increased access to antibody molecules upon removal of N-glycans in the flexible hinge region makes sense, as suggested by the molecular modeling experiments (Figure 4C, 4D). However, the software-determined approach measuring the accessible area of the probe, and not of the exposed spike area is somewhat counterintuitive. It would strengthen the manuscript, if an
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 870, 150]]<|/det|>
+in vitro approach was employed to further test the validity of the claims. For example, suitable neutralizing antibodies, or sera from seropositive individuals could be used to measure antibody binding capacity, or neutralization of pseudotyped virions.  
+
+<|ref|>text<|/ref|><|det|>[[114, 170, 880, 289]]<|/det|>
+We'd love to do the suggested experiment if there is any neutralizing antibody targeting at the hinge loop available. Due to the lack of anti- HCoV- NL63 neutralizing antibodies, we can only predict the glycan mediated immune evasion using molecular dynamic simulation. We are planning to design molecular binders targeting the hinge loop and test their neutralizing activities against wt and glycan mutants. Like most structural results, they are used to guide better and functionally meaningful experiments.  
+
+<|ref|>text<|/ref|><|det|>[[114, 310, 875, 389]]<|/det|>
+3. Figure 5C. The plot elements should be described in the legend (measure of center, error bars, and statistical test). If the data is normalized to the infectivity of wild type, why are there error bars on the wild type data? What is the reproducibility of the data? Pseudovirus infectivity data can be rather variable, and the information on numbers of independent experiments is missing.  
+
+<|ref|>text<|/ref|><|det|>[[114, 409, 875, 509]]<|/det|>
+Thank you very much for pointing it out and we have revised the figure legend for Figure 5C accordingly. The bars represent means+SD of triplicate samples. The wild type infectivity was also tested in triplicate and the mean+SD of the wild type was shown. The mutants were normalized to the mean of the wild type. The experiments were repeated at least three times and data from one representative experiment was shown.  
+
+<|ref|>text<|/ref|><|det|>[[114, 531, 744, 568]]<|/det|>
+Minor: Lines 254- 255. The term "monosaccharide" should be used instead of "glycan".  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 320, 607]]<|/det|>
+Thanks for the correction.  
+
+<|ref|>text<|/ref|><|det|>[[115, 629, 850, 668]]<|/det|>
+Line 319. It would be useful to mention exact numbers for glycan shield density for the other coronaviruses.  
+
+<|ref|>text<|/ref|><|det|>[[114, 690, 859, 789]]<|/det|>
+As suggested, we added panels G and H in Figure 3 showing the number of residues on the surface of different coronavirus spikes that are accessible to the Fab fragment of IgG. The less number of residues accessible to Fab, the more immune evasive the virus is. The analysis showed that HCoV- NL63 is more immune evasive (or glycan shielded) than other human coronaviruses.  
+
+<|ref|>text<|/ref|><|det|>[[114, 811, 877, 870]]<|/det|>
+Lines 484- 485. How was the "conservation of glycosylation" determined as opposed to amino acid conservation (previous sentence)? Is it based on experimental data from the literature? If so, how many and which viruses? Or is it algorithm- based prediction?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 89, 860, 189]]<|/det|>
+We analyzed spike sequences from 35 different coronaviruses, at position 1242 and 1247, \(89\%\) and \(74\%\) sequences have ASN and at position 1244 and 1249, \(91\%\) and \(69\%\) of the sequences have either Ser or Thr. The conservation of N- glycosylation is based on the NxS/T sequenc conservation at 1242- 1244 and 1247- 1249. We changed N- linked glycan to N- linked glycosylation sequon in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[114, 209, 822, 228]]<|/det|>
+Figure S7D. Please add amino acid residue numbers for the start/end of the shown ranges.  
+
+<|ref|>text<|/ref|><|det|>[[114, 249, 841, 288]]<|/det|>
+Thanks for the suggestion, we added amino acid residue numbers in the sequence alignment shown in the new Figure S9D.  
+
+<|ref|>text<|/ref|><|det|>[[115, 310, 253, 328]]<|/det|>
+References Cited  
+
+<|ref|>text<|/ref|><|det|>[[112, 364, 880, 647]]<|/det|>
+Ke, Zunlong, Joaquin Oton, Kun Qu, Mirko Cortese, Vojtech Zila, Lesley McKeane, Takanori Nakane, et al. 2020. "Structures and Distributions of SARS- CoV- 2 Spike Proteins on Intact Virions." Nature 588 (7838): 498- 502. Turoňová, Beata, Mateusz Sikora, Christoph Schürmann, Wim J. H. Hagen, Sonja Welsch, Florian E. C. Blanc, Sören von Bülow, et al. 2020. "In Situ Structural Analysis of SARS- CoV- 2 Spike Reveals Flexibility Mediated by Three Hinges." Science 370 (6513): 203- 8. Walls, Alexandra C., M. Alejandra Tortorici, Brandon Frenz, Joost Snijder, Wentao Li, Félix A. Rey, Frank DiMaio, Berend- Jan Bosch, and David Veesler. 2016. "Glycan Shield and Epitope Masking of a Coronavirus Spike Protein Observed by Cryo- Electron Microscopy." Nature Structural & Molecular Biology 23 (10): 899- 905. Yao, Hangping, Yutong Song, Yong Chen, Nanping Wu, Jialu Xu, Chujie Sun, Jiaxing Zhang, et al. 2020. "Molecular Architecture of the SARS- CoV- 2 Virus." Cell 183 (3): 730- 38. e13. Zhang, Kaiming, Shanshan Li, Grigore Pintilie, David Chmielewski, Michael F. Schmid, Graham Simmons, Jing Jin, and Wah Chiu. 2020. "A 3.4- Å Cryo- Electron Microscopy Structure of the Human Coronavirus Spike Trimer Computationally Derived from Vitrified NL63 Virus Particles." ORB Discovery 1 (November): e11.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 144, 393, 160]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 202, 840, 219]]<|/det|>
+The authors have improved the manuscript, however, several concerns are still not fully addressed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 259, 850, 330]]<|/det|>
+1. The authors have added the details for virus production and purification. They found a possible misinterpretation of the number of virions in their data. They re-checked the data, and based on a selected pool of 154 intact virions, their counts changed to \(20 \pm 13\) spikes/virion from the original \(\sim 9\) spikes/virion. This number is compared to that of SARS-CoV-2.  
+
+<|ref|>text<|/ref|><|det|>[[114, 342, 875, 470]]<|/det|>
+Despite the change of statistical analysis of the spikes/virion, the authors still report "On the virion surface, the average distance between the nearest neighboring spikes is \(\sim 34 \text{nm}\) (Figure 1E) compared to \(\sim 15 \text{nm}\) average nearest distance between prefusion spikes on SARS-CoV-2 virion" (line 175- 178). Also, the purity of viral prep evidenced by the coomassie gel does not exclude the possibility that the spikes may detach from the virions during purification. Therefore, I am not fully convinced unless the authors control for the virion intactness (more specifically the No. of spikes/virion) by cryo- ET of freshly prepared & unconcentrated & not chemically fixed HCoV- NL63 in the supernatant.  
+
+<|ref|>text<|/ref|><|det|>[[114, 509, 882, 581]]<|/det|>
+2. The cryo-ET technical notes have now been improved. There are still confusions, e.g. supplementary table 2 says the Voltage is 200kV, however, the MM part says "Cryo-ET data was collected by loading frozen grids into a Thermo Fisher Titan Krios transmission electron microscope operated at 300kV". Also, the tilt scheme shall be added to the table.  
+
+<|ref|>text<|/ref|><|det|>[[114, 620, 881, 675]]<|/det|>
+3. Why is the title of the manuscript "Integrated analyses reveal a hinge glycan regulates coronavirus spike tilting and virus infectivity" different from that of the SI "Bending of coronavirus spike regulated by a hinge glycan"?  
+
+<|ref|>text<|/ref|><|det|>[[115, 772, 393, 788]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 829, 424, 845]]<|/det|>
+The authors have addressed my concerns.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 91, 393, 106]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 145, 882, 292]]<|/det|>
+Chmielewski et al. presented a significant advance in understanding the structural dynamics of the spike protein of HCoV- NL63. Using cryo- ET and subtomogram averaging, the authors revealed a distribution of distinct titled conformations for the spike crown relative to its stalk region. As glycans are essential for the dynamics of spike, the authors used mass spectrometry to determine the site- specific occupancy and percentage of the N- linked glycan types at each site. The results revealed that different from other coronavirus such as SARS- CoV- 2, the glycans on the spike of HCoV- NL63 are predominantly high- mannose glycans. The higher glycan shielding on HCoV- NL63 leads to less accessible residues for Fab fragment of antibody, suggesting that HCoV- NL63 is more immune evasive than other coronaviruses.  
+
+<|ref|>text<|/ref|><|det|>[[115, 302, 882, 467]]<|/det|>
+Using structural prediction tools like I- TASSER and AlphaFold, the author constructed a predicted model for the stalk region. MD simulations of the glycosylated spike at different tilt angles generated a similar profile of the spike bending dynamics to the results collected by cryo- ET. There are two glycans, N1242 and N1247, near the hinge region of the spike stalk. The authors found that these two glycans generated the most favorable protein- glycan interactions and the minimum accessible surface area around the most probable bending conformation (\~56o) captured in the cryo- ET results. Removing these glycans, particularly N1242- linked glycan, shifted the distribution of bending conformations and decreased the protein- glycan contacts in the MD simulation. Finally, the authors found that mutation on N1242 (but not N1247) of spike reduced the infection of pseudo- VSV by \~70%.  
+
+<|ref|>text<|/ref|><|det|>[[116, 478, 880, 532]]<|/det|>
+This is a comprehensive study of the structural dynamics regarding the spike of HCoV- NL63 and provides important information on its glycans and immune evasion properties. However, there are some concerns that need to be addressed.  
+
+<|ref|>text<|/ref|><|det|>[[116, 544, 229, 560]]<|/det|>
+Major concern:  
+
+<|ref|>text<|/ref|><|det|>[[115, 570, 882, 880]]<|/det|>
+The evidence for structural dynamics is very compelling in the manuscript. However, the link between the structural dynamics of spike and infectivity is not convincing. Did the N1242D mutant spike generated on VSV- pseudovirus show different distributions on the bending conformations when compared to WT spike if measured by cryo- ET? Furthermore, the reduction of infectivity for N1242 mutant could be due to many reasons. Except for the MD simulation results (generated using the HCoV- NL63 spike, not the pseudo- VSV spike), there is no other evidence indicating the infection reduction is due to loss of the glycan- dependent motion. The response of the authors to Reviewer 1 brings more concerns – if ‘We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology’, then is the N1242D mutant spike generated using pseudo- VSV really a good model to validate the connection between the glycan- dependent motion and the infectivity? In the abstract, the authors quite emphasize the functional aspects of structural dynamics. For example, the abstract starts with ‘How the structure dynamics of the full- length spikes incorporated in viral lipid envelope correlates with the virus infectivity remains poorly understood.’ and ends with ‘Subsequent infectivity assays support the hypothesis that this glycan- dependent motion impacts virus entry.’ There is, in fact, no clear evidence linking the glycan- dependent motion to the virus entry in the manuscript. It is ideal that the authors can provide more evidence, but if the authors think it ‘goes beyond the scope of the current manuscript’, then at least they need to point out the limitation of the pseudo- VSV model
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 864, 125]]<|/det|>
+and tune down the description of the functional aspects (i.e., the infectivity) for the glycan- dependent structural dynamics.  
+
+<|ref|>text<|/ref|><|det|>[[115, 137, 234, 153]]<|/det|>
+Other concerns:  
+
+<|ref|>text<|/ref|><|det|>[[115, 165, 877, 237]]<|/det|>
+1) In the cryo-ET results, the authors found there are also different azimuthal directions along with different tilting angles. Because there is no preference for the azimuthal direction, the authors focused their analysis on the tilting angles. The authors did not state which azimuthal direction they used for the following MD simulations. Would different azimuthal directions impact the simulation result?  
+
+<|ref|>text<|/ref|><|det|>[[115, 248, 857, 302]]<|/det|>
+2) For the distributions of the bending angle probabilities (Fig. 4A and Supplementary Figure 10), the starting angle of 40o is quite close to the most probable angle observed in the cryo-ET results (~56o); however, the distribution peaks at around ~10o after MD simulation, any explanation for that?
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 110, 696, 129]]<|/det|>
+## Re: RESPONSES TO REVIEWER COMMENTS (MS # NCOMMS-23-05304A)  
+
+<|ref|>text<|/ref|><|det|>[[115, 154, 419, 172]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 196, 819, 235]]<|/det|>
+The authors have improved the manuscript, however, several concerns are still not fully addressed.  
+
+<|ref|>text<|/ref|><|det|>[[113, 259, 875, 514]]<|/det|>
+1. The authors have added the details for virus production and purification. They found a possible misinterpretation of the number of virions in their data. They re-checked the data, and based on a selected pool of 154 intact virions, their counts changed to \(20 \pm 13\) spikes/virion from the original "9 spikes/virion. This number is compared to that of SARS-CoV-2. Despite the change of statistical analysis of the spikes/virion, the authors still report "On the virion surface, the average distance between the nearest neighboring spikes is "34 nm (Figure 1E) compared to "15 nm average nearest distance between prefusion spikes on SARS-CoV-2 virion" (line 175-178). Also, the purity of viral prep evidenced by the Coomassie gel does not exclude the possibility that the spikes may detach from the virions during purification. Therefore, I am not fully convinced unless the authors control for the virion intactness (more specifically the No. of spikes/virion) by cryo-ET of freshly prepared & unconcentrated & not chemically fixed HCoV-NL63 in the supernatant.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 537, 207, 554]]<|/det|>
+## Responses:  
+
+<|ref|>text<|/ref|><|det|>[[115, 579, 882, 662]]<|/det|>
+The concern raised by the reviewer regards the potential fragility of coronavirus spikes leading to spike loss during virus purification, and consequently, the possible lack of representativeness in the observed spike dynamics on the virion surface. We respectfully disagree with this critique and offer the following rationales for our stance.  
+
+<|ref|>text<|/ref|><|det|>[[144, 685, 875, 874]]<|/det|>
+A. The employed technique of gradient-purification followed by concentration of virions is a well-established biochemical protocol for preparing coronaviruses suitable for cryogenic electron microscopy (cryo-EM) and tomography (cryo-ET) investigations. This method has been successfully employed in the analysis of the structures of other coronaviruses, such as SARS-CoV, FCoV, MHV (Neuman et al., 2006), and PEDV (Huang et al., 2022). These previous studies, akin to our current approach, utilized virions purified through gradient ultracentrifugation, which effectively maintained the pre-fusion conformation of the spikes without experiencing S1 shedding during the purification process. It is important to note that this purification procedure is necessary
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[170, 90, 808, 130]]<|/det|>
+to yield enough virus particles suitable for high- resolution cryo- EM and cryo- ET structural analysis of spikes.  
+
+<|ref|>text<|/ref|><|det|>[[144, 133, 879, 256]]<|/det|>
+B. We would like to underscore that HCoV-NL63 presents distinct challenges in terms of in vitro culture when juxtaposed with SARS-CoV-2, despite sharing the same receptor (Zhu et al., 2020). The diminished in vitro growth efficiency of HCoV-NL63 in comparison to other human coronaviruses that cause common cold, as noted by (Dijkman et al., 2013), makes it arduous to attain sufficiently high titer for direct imaging. It is impractical to carry out the cryo-ET experiment without concentration, as suggested by the reviewer.  
+
+<|ref|>text<|/ref|><|det|>[[144, 258, 877, 471]]<|/det|>
+C. Importantly, the bending dynamics of HCoV-NL63 spikes predicted by the MD simulations are in good agreement with our cryo-ET observation of individual spikes on the purified virions in our study (Figure 4A). The quantitative agreement between experimental and in silico estimates of spike bending dynamics suggested our MD simulation system was aptly capturing the conformational ensemble of the spike in situ. This suggests that the bending dynamics of HCoV-NL63 spikes is intrinsic to the individual spike itself. We observed variations in spike numbers on the virion and distances between neighboring spikes. Thus, such biochemical variations observed in our experimental settings appear not to impact on the spike bending dynamics which is one of the primary conclusions in our paper.  
+
+<|ref|>text<|/ref|><|det|>[[144, 472, 879, 896]]<|/det|>
+D. We thank the reviewer for accepting the corrected number of spikes/virion based on our first revision. However, with this correction, the spike-spike distance we reported will not change, as it is unrelated to the number of virions in the dataset that the reviewer correctly noted was incorrectly reported in the original manuscript Methods. Instead, the spike-spike distance analysis utilizes manually picked and refined spike subvolumes, where each virion in the tomograms was visually inspected in 2D slices for complete identification of all spikes prior to subtomogram processing. By utilizing refined subvolume orientations of spike stalks, which are anchored in the membrane, for determining nearest neighbor distances on the virion surface, our method represents an advanced quantification method for proteins attached to membranes. In contrast to earlier studies that approximated spike coordinates, our approach offers greater accuracy and is poised to contribute to future investigations into spike-antibody binding interactions and spike-receptor associations. Since there is no ground truth on the number of spikes and their spatial distributions known in any infectious coronavirus, our studies can only be validated by traditional biochemical assays like SDS gel and infectivity. Furthermore, the detailed infection steps are likely to be different between SARS-CoV-2 and HCoV-NL63, as exemplified by the lack of postfusion particles detected in our system as in SARS-CoV-2. HCoV-NL63 S protein is a single chain transmembrane glycoprotein with no cleavage site at S1/S2 boundary that is demonstrated by a single S protein band resolved by SDS-PAGE analysis of purified HCoV-NL63 (Fig. S3B in the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[170, 90, 877, 172]]<|/det|>
+manuscript) in contrast to mixed S, S1 and S2 protein bands shown in SARS- CoV- 2 (Ke et al., 2020). Trimerized HCoV- NL63 S proteins form spikes in single closed pre- fusion conformation that was observed in both purified soluble spike proteins (Walls et al., 2016) and in situ on purified virions (Zhang et al., 2020).  
+
+<|ref|>text<|/ref|><|det|>[[144, 175, 877, 406]]<|/det|>
+E. Finally, we wish to clarify that we do not assert any claims regarding no potential loss of spikes from HCoV-NL63 virions during the standard purification protocol. While we acknowledge the reviewer's suggestion to examine unpurified CoV virions within CoV-infected cells, possibly involving various cell types and distinct biochemical conditions, as an important avenue of research, we concur that this lies beyond the confines of our current study's scope. Our primary message in this manuscript remains the detailed exploration of the structural dynamics of spikes on purified and concentrated HCoV-NL63 virions. Serendipitously, our analysis suggests that the bending of the spikes may modulate the infectivity of a pseudo virus. We believe that our study will inspire many follow-up experiments to validate our hypothetical model of the structure and function relationship of the spike in situ.  
+
+<|ref|>text<|/ref|><|det|>[[115, 429, 882, 513]]<|/det|>
+2. The cryo-ET technical notes have now been improved. There are still confusions, e.g. supplementary table 2 says the Voltage is 200kV, however, the MM part says "Cryo-ET data was collected by loading frozen grids into a Thermo Fisher Titan Krios transmission electron microscope operated at 300kV". Also, the tilt scheme shall be added to the table.  
+
+<|ref|>text<|/ref|><|det|>[[115, 536, 785, 555]]<|/det|>
+We thank reviewer 1 for finding this inconsistency and have corrected the methods.  
+
+<|ref|>text<|/ref|><|det|>[[115, 579, 822, 639]]<|/det|>
+3. Why is the title of the manuscript "Integrated analyses reveal a hinge glycan regulates coronavirus spike tilting and virus infectivity" different from that of the SI "Bending of coronavirus spike regulated by a hinge glycan"?  
+
+<|ref|>text<|/ref|><|det|>[[115, 662, 866, 725]]<|/det|>
+We thank reviewer 1 for finding this inconsistency and have corrected the title of the SI. To tone down the claim, we changed our title to "Cryo-ET, Chemical and Computational Analyses Suggest That Hinge Glycans Modulate Coronavirus Spike Tilting and Infectivity".  
+
+<|ref|>text<|/ref|><|det|>[[115, 748, 420, 766]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 791, 835, 831]]<|/det|>
+The authors have addressed my concerns. Response: We thank reviewer 2 for all the previous comments to improve the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 855, 420, 873]]<|/det|>
+Reviewer #3 (Remarks to the Author):
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 880, 500]]<|/det|>
+Chmielewski et al. presented a significant advance in understanding the structural dynamics of the spike protein of HCoV- NL63. Using cryo- ET and subtomogram averaging, the authors revealed a distribution of distinct titled conformations for the spike crown relative to its stalk region. As glycans are essential for the dynamics of spike, the authors used mass spectrometry to determine the site- specific occupancy and percentage of the N- linked glycan types at each site. The results revealed that different from other coronavirus such as SARS- CoV- 2, the glycans on the spike of HCoV- NL63 are predominantly high- mannose glycans. The higher glycan shielding on HCoV- NL63 leads to less accessible residues for Fab fragment of antibody, suggesting that HCoV- NL63 is more immune evasive than other coronaviruses. Using structural prediction tools like I- TASSER and AlphaFold, the author constructed a predicted model for the stalk region. MD simulations of the glycosylated spike at different tilt angles generated a similar profile of the spike bending dynamics to the results collected by cryo- ET. There are two glycans, N1242 and N1247, near the hinge region of the spike stalk. The authors found that these two glycans generated the most favorable protein- glycan interactions and the minimum accessible surface area around the most probable bending conformation (\~56o) captured in the cryo- ET results. Removing these glycans, particularly N1242- linked glycan, shifted the distribution of bending conformations and decreased the protein- glycan contacts in the MD simulation. Finally, the authors found that mutation on N1242 (but not N1247) of spike reduced the infection of pseudo- VSV by \~70%.  
+
+<|ref|>text<|/ref|><|det|>[[115, 494, 877, 555]]<|/det|>
+This is a comprehensive study of the structural dynamics regarding the spike of HCoV- NL63 and provides important information on its glycans and immune evasion properties. However, there are some concerns that need to be addressed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 559, 239, 576]]<|/det|>
+Major concern:  
+
+<|ref|>text<|/ref|><|det|>[[115, 580, 707, 599]]<|/det|>
+The evidence for structural dynamics is very compelling in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 621, 763, 640]]<|/det|>
+Response: We thank the Reviewer for the overall positive evaluation of our work.  
+
+<|ref|>text<|/ref|><|det|>[[114, 663, 878, 809]]<|/det|>
+However, the link between the structural dynamics of spike and infectivity is not convincing. Did the N1242D mutant spike generated on VSV- pseudovirus show different distributions on the bending conformations when compared to WT spike if measured by cryo- ET? Furthermore, the reduction of infectivity for N1242 mutant could be due to many reasons. Except for the MD simulation results (generated using the HCoV- NL63 spike, not the pseudo- VSV spike), there is no other evidence indicating the infection reduction is due to loss of the glycan- dependent motion.  
+
+<|ref|>text<|/ref|><|det|>[[115, 833, 868, 895]]<|/det|>
+Response: We agree with the Reviewer that detailed analysis of a potential causal relationship between HCoV- NL63 infectivity and the bending of the spike is yet to be established. The correlation between these events is purely empirical as we have clearly stated in the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 874, 172]]<|/det|>
+Limitations section about the modeling. Following the Reviewer's cogent thought, we have further toned down this narrative across the article. In addition, we changed our manuscript title to reflect our approach and conclusion as "Cryo- ET, Chemical and Computational Analyses Suggest That Hinge Glycans Modulate Coronavirus Spike Tilting and Infectivity".  
+
+<|ref|>text<|/ref|><|det|>[[113, 196, 880, 470]]<|/det|>
+The response of the authors to Reviewer 1 brings more concerns – if 'We do not believe that pursuing a glycan analysis of the pseudo- VSV would yield new insights into coronavirus biology', then is the N1242D mutant spike generated using pseudo- VSV really a good model to validate the connection between the glycan- dependent motion and the infectivity? In the abstract, the authors quite emphasize the functional aspects of structural dynamics. For example, the abstract starts with 'How the structure dynamics of the full- length spikes incorporated in viral lipid envelope correlates with the virus infectivity remains poorly understood.' and ends with 'Subsequent infectivity assays support the hypothesis that this glycan- dependent motion impacts virus entry.' There is, in fact, no clear evidence linking the glycan- dependent motion to the virus entry in the manuscript. It is ideal that the authors can provide more evidence, but if the authors think it 'goes beyond the scope of the current manuscript', then at least they need to point out the limitation of the pseudo- VSV model and tune down the description of the functional aspects (i.e., the infectivity) for the glycan- dependent structural dynamics.  
+
+<|ref|>text<|/ref|><|det|>[[114, 494, 881, 599]]<|/det|>
+Responses: Thank you for the question on the statement we used in the manuscript and the original rebuttal. We have modified the sentence in the revised abstract as the following: "Subsequent infectivity assays implicated involvement of N1242- glyan in virus entry". The assay we used is indeed a standard procedure to measure viral spike- mediated entry. Hope that our revision clarifies the misunderstanding due to our original wording.  
+
+<|ref|>text<|/ref|><|det|>[[115, 622, 845, 682]]<|/det|>
+Following the Reviewer's suggestions, we have further revised the article to ensure that no statements remain that structural dynamics of the spike are correlated with the infectivity assays.  
+
+<|ref|>text<|/ref|><|det|>[[115, 708, 245, 724]]<|/det|>
+Other concerns:  
+
+<|ref|>text<|/ref|><|det|>[[115, 728, 866, 830]]<|/det|>
+1) In the cryo-ET results, the authors found there are also different azimuthal directions along with different tilting angles. Because there is no preference for the azimuthal direction, the authors focused their analysis on the tilting angles. The authors did not state which azimuthal direction they used for the following MD simulations. Would different azimuthal directions impact the simulation result?  
+
+<|ref|>text<|/ref|><|det|>[[115, 856, 870, 895]]<|/det|>
+Responses: We thank the Reviewer for this insightful suggestion. Indeed, MD simulations were initially performed with a range of tilt angles but a single azimuthal angle which was arbitrarily
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 881, 365]]<|/det|>
+chosen. Our results were shown in Fig 4A in the manuscript. Subsequently, using 3.5 microsecond- long MD simulations, a broad range of azimuthal angles (between 0 to \(60^{\circ}\) ) are sampled that are now presented in Figure 1 below (Fig. S11 in the SI), which is also broader relative to the ones previously observed in MD simulations of SARS- CoV- 2 spikes (Fig. 3B from (Kapoor et al., 2022)). However, given the higher diffusion barriers associated with azimuthal over tilting changes, the entire range of the azimuthal space (0 to \(120^{\circ}\) as seen in cryo- ET) is not captured in our finite time MD simulations. This limited sampling has further led to an artificial minimum around the azimuthal angle of \(\sim 20^{\circ}\) , which will potentially even out with longer simulations. Nonetheless, as stated in the subsequent response the dynamics of the azimuthal angle is highly decoupled from that of the tilt. This decoupling between the two angular changes is observed from the maps, where despite an even distribution of the azimuthal angle, the tilt was prominently peaked around \(56^{\circ}\) . These caveats are now mentioned in the Limitations section.  
+
+<|ref|>text<|/ref|><|det|>[[114, 386, 883, 512]]<|/det|>
+The azimuthal angle controls the direction along which the bending occurs, while the tilt quantifies the magnitude of this bending. Our cryo- EM data shows that the spikes are dispersed so that the inter- spike dynamics (and hence the azimuthal changes) does not affect spike bending. In additional analysis that we now provide (Figure 1), we find that despite finite sampling there is little correlation between the azimuthal and tilt dynamics in the MD simulations.  
+
+<|ref|>image<|/ref|><|det|>[[308, 520, 694, 716]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 728, 876, 768]]<|/det|>
+<center>Figure 1. Distribution of azimuthal and tilt angles sampled during MD simulations, and a scatter plot showing minimal correlation between the dynamics of these two angles during MD. </center>  
+
+<|ref|>text<|/ref|><|det|>[[114, 790, 867, 873]]<|/det|>
+2) For the distributions of the bending angle probabilities (Fig. 4A and Supplementary Figure 10), the starting angle of \(40^{\circ}\) is quite close to the most probable angle observed in the cryo-ET results ( \(\sim 56^{\circ}\) ); however, the distribution peaks at around \(\sim 10\) after MD simulation, any explanation for that?
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[125, 110, 568, 355]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[132, 358, 650, 395]]<|/det|>
+<center>Figure 2. Bending angle by replica for simulations using \(40^{\circ}\) map </center>  
+
+<|ref|>text<|/ref|><|det|>[[114, 413, 870, 497]]<|/det|>
+Response: We have now investigated this data further. Presented in Figure 2, we find that two of the three replicas were indeed moving towards higher angles. There was one replica nonetheless that moved towards lower angles creating the peak. The cumulative area under the curve for drift towards \(56^{\circ}\) is still higher making the \(\sim 56^{\circ}\) region more probable.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 541, 244, 558]]<|/det|>
+## References Cited  
+
+<|ref|>text<|/ref|><|det|>[[111, 573, 872, 907]]<|/det|>
+Dijkman, R., Jebbink, M. F., Koekkoek, S. M., Deijs, M., Jonsdottir, H. R., Molenkamp, R., leven, M., Goossens, H., Thiel, V., & van der Hoek, L. (2013). Isolation and characterization of current human coronavirus strains in primary human epithelial cell cultures reveal differences in target cell tropism. Journal of Virology, 87(11), 6081- 6090. Huang, C.- Y., Draczkowski, P., Wang, Y.- S., Chang, C.- Y., Chien, Y.- C., Cheng, Y.- H., Wu, Y.- M., Wang, C.- H., Chang, Y.- C., Chang, Y.- C., Yang, T.- J., Tsai, Y.- X., Khoo, K.- H., Chang, H.- W., & Hsu, S.- T. D. (2022). In situ structure and dynamics of an alphacoronavirus spike protein by cryo- ET and cryo- EM. Nature Communications, 13(1), 4877. Kapoor, K., Chen, T., & Tajkhorshid, E. (2022). Posttranslational modifications optimize the ability of SARS- CoV- 2 spike for effective interaction with host cell receptors. Proceedings of the National Academy of Sciences of the United States of America, 119(28), e2119761119. Ke, Z., Oton, J., Qu, K., Cortese, M., Zila, V., McKeane, L., Nakane, T., Zivanov, J., Neufeldt, C. J., Cerikan, B., Lu, J. M., Peukes, J., Xiong, X., Krausslich, H.- G., Scheres, S. H. W., Bartenschlager, R., & Briggs, J. A. G. (2020). Structures and distributions of SARS- CoV- 2 spike proteins on intact virions. Nature, 588(7838), 498- 502. Neuman, B. W., Adair, B. D., Yoshioka, C., Quispe, J. D., Orca, G., Kuhn, P., Milligan, R. A., Yeager, M., & Buchmeier, M. J. (2006). Supramolecular architecture of severe acute respiratory syndrome
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 884, 303]]<|/det|>
+coronavirus revealed by electron cryomicroscopy. Journal of Virology, 80(16), 7918- 7928. Walls, A. C., Tortorici, M. A., Frenz, B., Snijder, J., Li, W., Rey, F. A., DiMaio, F., Bosch, B.- J., & Veesler, D. (2016). Glycan shield and epitope masking of a coronavirus spike protein observed by cryo- electron microscopy. Nature Structural & Molecular Biology, 23(10), 899- 905. Zhang, K., Li, S., Pintilie, G., Chmielewski, D., Schmid, M. F., Simmons, G., Jin, J., & Chiu, W. (2020). A 3.4- Å cryo- electron microscopy structure of the human coronavirus spike trimer computationally derived from vitrified NL63 virus particles. QRB Discovery, 1, e11. Zhu, N., Wang, W., Liu, Z., Liang, C., Wang, W., Ye, F., Huang, B., Zhao, L., Wang, H., Zhou, W., Deng, Y., Mao, L., Su, C., Qiang, G., Jiang, T., Zhao, J., Wu, G., Song, J., & Tan, W. (2020). Morphogenesis and cytopathic effect of SARS- CoV- 2 infection in human airway epithelial cells. Nature Communications, 11(1), 3910.
+
+<--- Page Split --->
+
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 91, 300, 106]]<|/det|>
+REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 147, 393, 163]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 203, 879, 293]]<|/det|>
+Filtering the coronavirus through the membrane filter is unconventional and can detach the spikes. The control experiment to clarify this is cryo- ET of unconcentrated HCoV- NL63 in the supernatant, images of a few virions would do. According to our experiences, this can be done on SARS- CoV- 2 and other seasonal human coronaviruses. If the authors can not supply this, they shall add a statement of the above risks to the Limitations part.  
+
+<|ref|>text<|/ref|><|det|>[[115, 334, 460, 350]]<|/det|>
+I have no further comments to the other parts.  
+
+<|ref|>text<|/ref|><|det|>[[115, 419, 393, 435]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 475, 595, 492]]<|/det|>
+The authors addressed all my concerns in the revised manuscript.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 110, 695, 129]]<|/det|>
+## Re: RESPONSES TO REVIEWER COMMENTS (MS # NCOMMS-23-05304B)  
+
+<|ref|>text<|/ref|><|det|>[[115, 154, 419, 171]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 195, 879, 291]]<|/det|>
+Filtering the coronavirus through the membrane filter is unconventional and can detach the spikes. The control experiment to clarify this is cryo- ET of unconcentrated HCoV- NL63 in the supernatant, images of a few virions would do. According to our experiences, this can be done on SARS- CoV- 2 and other seasonal human coronaviruses. If the authors can not supply this, they shall add a statement of the above risks to the Limitations part.  
+
+<|ref|>text<|/ref|><|det|>[[115, 314, 459, 330]]<|/det|>
+I have no further comments to the other parts.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 355, 206, 371]]<|/det|>
+## Responses:  
+
+<|ref|>text<|/ref|><|det|>[[115, 397, 841, 474]]<|/det|>
+We want to thank this reviewer for the suggestion. Unfortunately, the virus titer in unconcentrated culture supernatant of HCoV- NL63 infected cells is too low to direct image without virus concentration/purification. We added this limitation in the discussion of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 508, 419, 525]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 550, 595, 567]]<|/det|>
+The authors addressed all my concerns in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 586, 206, 602]]<|/det|>
+Responses:  
+
+<|ref|>text<|/ref|><|det|>[[115, 628, 178, 643]]<|/det|>
+Thanks.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d5ef8fed2a8e3dbecb979f0481b24ab88c243101d124d78b8ab1cdf9509460c5/images_list.json

@@ -0,0 +1,243 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample.",
+    "footnote": [],
+    "bbox": [
+      [
+        241,
+        175,
+        768,
+        545
+      ]
+    ],
+    "page_idx": 5
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_7.jpg",
+    "caption": "Sup. Figure 7. Penetration of anti-calbidin scFv into fixed HEK cells or live cells.",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        264,
+        880,
+        860
+      ]
+    ],
+    "page_idx": 6
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_8.jpg",
+    "caption": "Sup. Figure 8. Penetration of anti-calbindin scFv into fixed COS-1 cells.",
+    "footnote": [],
+    "bbox": [
+      [
+        260,
+        191,
+        736,
+        556
+      ]
+    ],
+    "page_idx": 8
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_6.jpg",
+    "caption": "Sup. Figure 6. Ultrastructure comparison between samples incubated for one day or seven days.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        90,
+        880,
+        400
+      ]
+    ],
+    "page_idx": 9
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_21.jpg",
+    "caption": "Sup. Figure 21. Well-preserved ultrastructure from the surface (a) to the middle (d) of the \\(120 - \\mu m\\) section.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        90,
+        877,
+        565
+      ]
+    ],
+    "page_idx": 11
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents.",
+    "footnote": [],
+    "bbox": [
+      [
+        118,
+        120,
+        877,
+        750
+      ]
+    ],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_9.jpg",
+    "caption": "Sup. Figure 9. Validation of anti-GFAP scFv with Immunofluorescence immunocytochemistry.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        559,
+        875,
+        840
+      ]
+    ],
+    "page_idx": 14
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_10.jpg",
+    "caption": "Sup. Figure 10. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (cerebellum).",
+    "footnote": [],
+    "bbox": [
+      [
+        253,
+        148,
+        741,
+        526
+      ]
+    ],
+    "page_idx": 17
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_11.jpg",
+    "caption": "Sup. Figure 11. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (hippocampus).",
+    "footnote": [],
+    "bbox": [
+      [
+        252,
+        92,
+        742,
+        468
+      ]
+    ],
+    "page_idx": 18
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_14.jpg",
+    "caption": "Sup. Figure 14. Validation of anti-calbindin and anti-parvalbumin scFvs with Immunofluorescence immunohistochemistry.",
+    "footnote": [],
+    "bbox": [
+      [
+        252,
+        92,
+        744,
+        468
+      ]
+    ],
+    "page_idx": 19
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Sup. Figure 2. Pixel correlation scatter plots comparing the native YFP fluorescence signal and the red fluorescence from the labeling of the GFP-specific scFv.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_12.jpg",
+    "caption": "Sup. Figure 12. Validation of immunofluorescence by scFv probes and their parental mAbs (part 1).",
+    "footnote": [],
+    "bbox": [
+      [
+        116,
+        500,
+        879,
+        799
+      ]
+    ],
+    "page_idx": 22
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_1.jpg",
+    "caption": "Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure.",
+    "footnote": [],
+    "bbox": [
+      [
+        115,
+        88,
+        880,
+        744
+      ]
+    ],
+    "page_idx": 24
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Sup. Figure 3. Immunolabeling results of anti-calbindin scFv and tissue penetration comparison.",
+    "footnote": [],
+    "bbox": [
+      [
+        116,
+        160,
+        872,
+        840
+      ]
+    ],
+    "page_idx": 25
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample.",
+    "footnote": [],
+    "bbox": [
+      [
+        240,
+        163,
+        767,
+        530
+      ]
+    ],
+    "page_idx": 27
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents.",
+    "footnote": [],
+    "bbox": [
+      [
+        117,
+        90,
+        877,
+        725
+      ]
+    ],
+    "page_idx": 28
+  }
+]

peer_reviews/supplementary_0_Peer Review File__d5ef8fed2a8e3dbecb979f0481b24ab88c243101d124d78b8ab1cdf9509460c5/supplementary_0_Peer Review File__d5ef8fed2a8e3dbecb979f0481b24ab88c243101d124d78b8ab1cdf9509460c5.mmd

@@ -0,0 +1,352 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Multiplexed volumetric CLEM enabled by scFvs provides new insights into the cytology of cerebellar cortex  
+
+![](images/Figure_1.jpg)
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors developed eight smaller single- chain variable fragments (scFvs) based on eight well- characterized mAbs and conjugated them with various fluorescent dyes. This idea is great for volume correlative light and electron microscopy. The authors reported that each scFv proved effective as a detergent- free immunofluorescent probe. The approach was believed to be promising for routine linking of molecular information to connectomic information from the same material since the quality of data from the volumetric fluorescent and electron microscopy is good. However, some critical aspects of the experimental design need to be addressed, and additional experiments are required to draw conclusive findings.  
+
+1. One concern in this study is to choose Triton X-100 as a detergent for comparison. While Triton X-100 is commonly used to aid antibody penetration in light microscopy, it is not widely used for EM or immuno-EM studies. For EM studies, saponin is one of organic solvents which dissolve lipids from cell membranes making them permeable to antibodies. Furthermore, organic solvents can be used to fix and permeabilize cells at the same time by coagulating proteins. Saponin interacts with membrane cholesterol, selectively removing it and leaving holes in the membrane. Most researchers in the EM field choose low concentration of saponin to balance the antibody penetration and excellent membrane morphology in their immuno-EM experiments. Majority of their immuno-EM images showed great morphology of membrane at the ultrastructural level with the use of saponin. In some immuno-EM studies, Triton X-100 has been used. But those studies don't care the cell membrane, most of them were interested in some subcellular organelles.  
+
+2. Another point of concern is the lack of clarity on how the antibodies penetrate the cell through the cell membrane in detergent-free immunofluorescence labeling. A more thorough investigation and explanation of this process are needed to provide a comprehensive understanding of the technique's efficacy.  
+
+3. The animals were perfused with the fixative (4% paraformaldehyde + 0.1% glutaraldehyde). The low concentration of glutaraldehyde (0.1%) may not be sufficient to adequately preserve the lipids in the cell membrane. However, for detergent-free immunofluorescence labeling, scFv probes or nanobody probes were incubated for 3 days (50 \(\mu \mathrm{m}\) ) or 7 days (120 \(\mu \mathrm{m}\) ). This could potentially affect the ultrastructural morphology.  
+
+4. This study lack of novelty. The use of scFv has been well-established over the years, making it a solid foundation for the volume CLEM study. However, the choice of detergent-free immunofluorescence labeling raises questions about its suitability. Considering alternative approaches, such as utilizing low concentrations of saponin, might offer a more effective option for preserving membrane morphology during immuno-EM experiments.  
+
+5. The choice of 0.3% Triton X-100 to demonstrate detergent issues in EM seems excessive. Most EM studies recommend not exceeding 0.1% Triton X-100, making it unnecessary to use a higher concentration for this purpose.  
+
+Reviewer #2 (Remarks to the Author):  
+
+This is a technology development manuscript describing the generation and implementation of an assortment of single chain antibody- based probes (scFvs) against different brain proteins to label tissue for correlative fluorescence/volumetric electron microscopy. The key advance is the nature of the labeling probes, which can diffuse deep into fixed brain tissue and penetrate cells without the need for detergent permeabilization. The authors do nice side- by- side comparisons with whole IgG antibodies to highlight this. Detergent- free labeling thus allows processing for ultrastructural analysis
+
+<--- Page Split --->
+
+by EM, with excellent membrane preservation. Furthermore, the scFv labeling reagents can be easily labeled with different fluorescent dyes allowing the authors to visualize numerous (here they show 6) different labels in the same sample using spectral unmixing confocal microscopy. Serial section EM images of the same samples were then reconstructed and aligned with the fluorescence images to achieve correlated fluorescence/ultrastructure. Overall the data were compelling, with many beautiful examples of correlated fluorescence localization with 3D ultrastructure, nicely demonstrating the power of the technique. While the manuscript primarily focuses on tool development and offers little in the way of novel biological insight, I feel the potential future impact of the technique (i.e. ability to assign neural identities to volumetric connectomics EM datasets, ultrastructural localization of channels, receptors, etc.) will have broad appeal. There are several specific points that deserve attention:  
+
+- Nowhere in the manuscript do the authors validate their labeling reagents in a knockout background. In many cases the labeling pattern is distinct, consistent with previously published work and the localization of the signal makes sense with the correlated ultrastructure (i.e. vGLUT labels presynaptic terminals), but in some cases it is more ambiguous. For example, in Fig. S2a,e the authors argue that the CB and PV scFvs label more of the target proteins in the cell nucleus, is this real signal or are these probes picking up something non-specific in the nucleus that the IgG does not?  
+
+- While many of the images are quite striking, overall the manuscript lacked any sort of quantitative analysis. Just as one example, in Fig. 1, showing a simple pixel correlation scatter plot comparing the YFP and GFP-scFv signal would give readers a better idea of how evenly the scFv is penetrating cells to label YFP.  
+
+- p. 9 "....immunofluorescence patterns that were similar to or in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex (Figure 2 a; Sup. Figure 2; Sup. Figure 3). In many cases the comparisons between mAb and scFv is not entirely fair since the mAb is labeled in the 488/green channel (in which brain tissue notoriously has more autofluorescence) and the scFv in the red channel e.g. NPY signal in Sup. 2b,d; PSD95 in Sup. 2f. Is the labeling really that much cleaner or is the background signal in the green channel making the mAb appear worse than it is?  
+
+- p. 9 "....found that the anti-calbindin scFv penetrated to a depth of \(\sim 150 \mu m\) in a 300-μm tissue slice". So the probe labeled throughout the entire slice?
+
+<--- Page Split --->
+
+## 1 REVIEWER COMMENTS  
+
+2  
+
+## Reviewer #1 (Remarks to the Author):  
+
+The authors developed eight smaller single- chain variable fragments (scFvs) based on eight well- characterized mAbs and conjugated them with various fluorescent dyes. This idea is great for volume correlative light and electron microscopy. The authors reported that each scFv proved effective as a detergent- free immunofluorescent probe. The approach was believed to be promising for routine linking of molecular information to connectomic information from the same material since the quality of data from the volumetric fluorescent and electron microscopy is good. However, some critical aspects of the experimental design need to be addressed, and additional experiments are required to draw conclusive findings.  
+
+1. One concern in this study is to choose Triton X-100 as a detergent for comparison. While Triton X-100 is commonly used to aid antibody penetration in light microscopy, it is not widely used for EM or immuno-EM studies. For EM studies, saponin is one of organic solvents which dissolve lipids from cell membranes making them permeable to antibodies. Furthermore, organic solvents can be used to fix and permeabilize cells at the same time by coagulating proteins. Saponin interacts with membrane cholesterol, selectively removing it and leaving holes in the membrane. Most researchers in the EM field choose low concentration of saponin to balance the antibody penetration and excellent membrane morphology in their immuno-EM experiments. Majority of their immuno-EM images showed great morphology of membrane at the ultrastructural level with the use of saponin. In some immuno-EM studies, Triton X-100 has been used. But those studies don't care the cell membrane, most of them were interested in some subcellular organelles.  
+
+We thank the reviewer for pointing out the important fact that saponin is a far better detergent for electron microscopy ultrastructural studies than Triton X- 100. In response to this suggestion, we did a new series of experiments analyzing twenty- two tissue blocks at various saponin concentrations, sample thicknesses, and durations of antibody incubation. Moreover, directly conjugated mAbs have become available, so we moved the results related to the penetration tests with secondary antibody labeling (original Figure 1 d and e) to Supplementary Figure 3 b and c). The new results are now presented in new Figure 1 panels f- i, new Supplementary Figure 4 (plus additional new Supplementary Figure 5 that we will describe below in point 4). The key result is that saponin at \(0.05\%\) concentration does not allow fluorescently labeled monoclonal antibodies to penetrate into the middle \(500 \mu \mathrm{m}\) of a 1- mm block even after a 1- week incubation with the labeled antibody (Figure 1 g; Supplementary Figure 4). In contrast, the fluorescently labeled scFv penetrated throughout a block in the absence of detergent (Figure 1 f; Supplementary Figure 4). In different experiments, we did examine if higher concentrations of saponin could aid in deeper penetration (see new Supplementary Figure 5) but for our purposes, even saponin at \(0.05\%\) was problematic. The reason was that we found small breaks in the plasma membranes of neuronal processes (arrows, Fig 1 i) that were not present in samples not treated with detergent (Fig 1 h). While these ultrastructural breaks are small and, for many kinds of studies, would be of no consequence, for connectomics they are serious. This seriousness is related to the requirement for automatic algorithms to segment each nerve cell process. When two adjacent objects have a
+
+<--- Page Split --->
+
+continuity between them, this is often interpreted by the algorithms erroneously as the same object. Such merge errors are far more difficult to find and correct than split errors, so avoiding them at all costs is necessary (Shapson- Coe et al. 2021; Januszewski et al. 2018). With newer techniques like multicolor 2- photon microscopy (Mahou et al. 2012; Blanc et al. 2023; Pudavar et al. 2024), lightsheet microscopy in uncleared tissue (Schmid et al. 2013), and confocal done with clearing approaches compatible with electron microscopy (Furuta et al. 2022), we think having fluorescent scFv penetration hundreds of microns into tissue blocks will be of great use in CLEM studies. We have modified the text to make these points clearer (line 150).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure. </center>  
+
+a, Schematic representations of a full-length IgG antibody and an scFv probe with a conjugated fluorescent dye. b, Confocal images from the cerebral cortex of a YFP-H mouse labeled using a GFP-specific scFv probe conjugated with the red dye 5-TAMRA. Arrows show thinner neuronal processes, perhaps myelinated, that are not labeled by scFv. c, Layer ⅔ of the cerebral cortex labeled with a calbindin-specific scFv probe. d, Cerebellum cortex of Crus 1 labeled with the PSD-95 specific scFv. Right panel is the enlarged boxed inset from left. e, Cerebral cortex labeled with the NPY-specific scFv. f and g, Tissue penetration depth comparison of a parvalbumin-specific scFv without detergent and its parental
+
+<--- Page Split --->
+
+mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation. h and i, Comparison of ultrastructure from samples incubated 7 days without detergent and with \(0.05\%\) saponin. Arrows indicate membrane breaks. Asterisks indicate abnormal appearing vesicle- filled axonal terminals.  
+
+![](images/Figure_7.jpg)
+
+<center>Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample. </center>  
+
+Tissue penetration depth comparison of a calbindin- specific scFv without detergent and its parental mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation.  
+
+2. Another point of concern is the lack of clarity on how the antibodies penetrate the cell through the cell membrane in detergent-free immunofluorescence labeling. A more thorough investigation and explanation of this process are needed to provide a comprehensive understanding of the technique's efficacy.  
+
+We agree with the reviewer on the importance of investigating how scFvs penetrate the cell membrane in the absence of detergent. We are interested in determining the mechanism as well. We think there are at least two possible mechanisms:  
+
+First, because all the immunolabeling experiments in this study were performed on brain tissue samples from animals perfused and fixed with \(4\%\) formaldehyde (prepared fresh from paraformaldehyde) \(+0.1\%\) glutaraldehyde in PBS, the cell membrane penetration by scFvs may be simply explained by the fact that formaldehyde and glutaraldehyde permeate the lipid bilayer. formaldehyde and glutaraldehyde are commonly used as chemical fixatives by crosslinking amino groups of proteins (Fischer et al. 2008). It has been known that formaldehyde also
+
+<--- Page Split --->
+
+dissolves lipids (Fox et al. 1985; Kiernan 2000; Thavarajah et al. 2012). A recent study using surface plasmon resonance (SPR) (Cheng et al. 2019) showed that fixation with formaldehyde perturbed the integrity of membranes ( \(10 \pm 5\%\) mass loss), and they showed increased permeability of sucrose. In another recent study using atomic force microscopy (Ichikawa et al. 2022), both formaldehyde and glutaraldehyde were shown to increase the size of nanoscopic protrusions on cell membranes. These protrusions were generated by membrane protein aggregates induced by crosslinking via formaldehyde or glutaraldehyde. The aggregated membrane proteins may create gaps between them and their nearby lipids providing a permeability pore. Additionally, two extracellular space- preserving fixation methods employing formaldehyde and glutaraldehyde (Fulton and Briggman 2021; Lu et al. 2023) showed that full- length antibodies can penetrate cell membranes albeit with lower diffusivity than scFvs, supporting the idea that the formaldehyde plus glutaraldehyde treated membranes do have gaps caused by the fixation.  
+
+We have tested this idea as well by using scFv immunolabeling on HEK293T cells cultured as a single layer on a petri dish with a coverglass bottom. HEK293T cells allowed us to avoid the issue of cut/fragmented cells in tissue sections where scFv could penetrate into cells via a cut surface rather than through a membrane. After transfecting the HEK293T cells with a plasmid encoding calbindin, we fixed the cells with the same fixative (4% formaldehyde + 0.1% glutaraldehyde in PBS) for 15 min and then washed with PBS. Overnight immunolabeling of the anti- calbindin scFv was then performed without or with 0.1% Triton- X. The results showed that in both conditions (without or with 0.1% Triton- X), the scFv can penetrate and label its intracellular target (we have added a new Supplementary Figure 7 a, b to show this data). This result provides evidence consistent with the idea that the cell membranes fixed with 4% formaldehyde + 0.1% glutaraldehyde allow scFvs to penetrate into intracellular spaces. Additionally, we also tested a 1- hour immunolabeling protocol using scFvs and full- size mAbs directed to transfected calbindin in COS- 1 cells both without and with detergent permeabilization. Similarly, we found that the scFvs were able to penetrate COS- 1 cells and label intracellular targets. However, the mAb was unable to penetrate at least at 1 hour (see new Supplementary Figure 8). In another experiment we did find that an overnight incubation with a mAb did label fixed cells that were not permeabilized with detergent. From all of these experiments we infer that due to their small size the scFvs are better to penetrate fixed cells than larger immunoprobes. We have modified the text to make these points clearer (line 169).  
+
+It is also possible that scFvs by virtue of their small size could permeate unfixed lipid bilayers. Indeed (Li et al. 2016) showed that anti- pTau nanobodies when injected into the blood of live mice could cross the blood- brain barrier and also cross neuronal cell membranes to label intracellular pTau. In (Bernard et al. 2016), after transgenically inducing expression of an anti- Otx2 scFv to express in cells of the choroid plexus cells, scFv in the CSF can cross the blood- brain barrier and neutralize Otx2 in the cortex, perhaps via transcytosis. In (Thiel et al. 2002), scFvs were shown to be able to pass through live cornea with an intact epithelium. (Im, Chung, and Jang 2017) showed that scFvs can enter live, unfixed culture cells.  
+
+Based on these results, we were motivated to see if the scFvs we generated could cross into living cells. We attempted to immunolabel the transfected HEK293T cell for calbindin with the anti- calbindin scFv using live HEK293T cells. We found that after a one- hour incubation, the scFv could penetrate cells (Supplementary Figure 7 c, arrow). However, unlike the penetration of fixed cells described above, the labeling was more punctate. This labeling was most likely
+
+<--- Page Split --->
+
+explained by endocytosis as has been previously seen for extracellular dye molecules (see for example, (Tsuriel et al. 2015)). We have added a Supplementary Figure 7 c to show this result. Consistent with this, it has been documented that both nanobodies and scFvs can be internalized into cells via endocytosis (de Beer and Giepmans 2020; Wittrup et al. 2009; Alric et al. 2018; Kim et al. 2020). We have modified the text to make this point clearer (line 176). This is a potentially important route of entry because it provides an option to achieve immunolabeling of larger tissue samples, such as a whole mouse brain, by introducing these small immunoprobes in live animals.  
+
+![](images/Figure_8.jpg)
+
+<center>Sup. Figure 7. Penetration of anti-calbidin scFv into fixed HEK cells or live cells. </center>
+
+<--- Page Split --->
+
+Immunofluorescence immunocytochemistry on transiently transfected cells. HEK cells were transfected with a plasmid encoding Flag-tagged human calbindin. a, Chemically fixed cells were labeled overnight with Alexa594 anti-calbindin L109/57 scFv in the absence of detergent. b, Chemically fixed cells were labeled overnight with Alexa594 anti-calbindin L109/57. scFv with \(0.1\%\) Triton-X. c, Live cells were labeled with Alexa594 anti-calbindin L109/57 scFv. The arrow indicates the cell that has intracellular scFv labeling. Arrowheads indicate puncta labeling in some cells.  
+
+![](images/Figure_6.jpg)
+
+<center>Sup. Figure 8. Penetration of anti-calbindin scFv into fixed COS-1 cells. </center>  
+
+Immunofluorescence immunocytochemistry on transiently transfected cells. COS- 1 cells were transfected with a plasmid encoding Flag- tagged human calbindin. Cells in panels A and B were labeled for 1 hour after fixation and prior to detergent permeabilization with (A) Alexa594 anti- calbindin L109/57 scFv or (B) anti- calbindin mouse mAb L109/39 (scFv and mAb labeling in red). After permeabilization, cells were labeled with rabbit anti- Flag (green) to detect calbindin, and Hoechst nuclear dye (blue). For cells in panels C and D all immunolabeling was performed after fixation and detergent permeabilization with (C) Alexa594 anti- calbindin L109/57 scFv or (D) anti- calbindin mouse mAb L109/39 (scFv and mAb labeling in red). Cells were simultaneously labeled with rabbit anti- Flag (green), and Hoechst (blue). Cells in all panels were imaged at the same exposure.  
+
+3. The animals were perfused with the fixative (4% paraformaldehyde + 0.1% glutaraldehyde). The low concentration of glutaraldehyde (0.1%) may not be sufficient to adequately preserve the lipids in the cell membrane. However, for detergent-free immunofluorescence labeling, scFv probes or nanobody probes were incubated for 3 days (50 μm) or 7 days (120 μm). This could potentially affect the ultrastructural morphology.  
+
+We agree this is a reasonable concern that the use of 0.1% glutaraldehyde does not sufficiently preserve lipids in the cell membrane, which may cause the ultrastructure to deteriorate when tissue samples are incubated with immuno-probes for prolonged periods like
+
+<--- Page Split --->
+
+three or seven days. We were aware of this potential problem. The reasons we used \(0.1\%\) glutaraldehyde instead of a higher concentration was: first, glutaraldehyde is a harsher fixative which may modify epitopes on target proteins (Fischer et al. 2008), preventing immuno- probe labeling; Second, glutaraldehyde has higher autofluorescence than formaldehyde (Fischer et al. 2008), which causes high background in fluorescence microscopy. Because we only used \(0.1\%\) glutaraldehyde, we always postfixed the perfused brain for many hours (overnight). To prevent reversal of the formaldehyde fixation (Fischer et al. 2008) the brain samples were then sliced in ice- cold fixative (4% formaldehyde \(+0.1\%\) glutaraldehyde) and stored in the same fixative at \(4^{\circ}C\) . The only exception to this protocol was our work with the neuropeptide NPY, which we, as others, have found to be difficult to label if the fixation is too extensive. In this case, we stored the slices in PBS at \(4^{\circ}C\) . We also performed all the incubations, including the washing steps, at \(4^{\circ}C\) to prevent ultrastructural degradation. In the manuscript, in Supplementary Figure 21, we examined the ultrastructure of a \(2\mathrm{mm}\) , \(2\mathrm{mm}\) , \(120 - \mu \mathrm{m}\) thick cerebellum tissue sample incubated with scFv probes for seven days after light fixation (described above). As shown in the figure, ultrastructure at four locations across the cerebellar cortex layers including regions that are near the center of the block, is preserved well. After careful examination of the images during the revision process, we have noticed some abnormalities. In the superficial layer (the molecular layer) of the cerebellar cortex, which is mainly composed of neuronal processes and close to the surface of the block, we did observe some artifacts (new arrows in Supplementary Figure 21). We are unsure whether these artifacts are explained by mechanical or chemical or thermal factors that are different at the surface vs. the interior of the block. We have also modified the manuscript (line 217) to make readers aware of this issue. If reviewers are interested in examining the ultrastructure directly, we encourage reviewers to visit the Neuroglancer link of our vCLEM dataset at Neuroglancer LINK.  
+
+In addition, in a new set of experiments we performed for the revision that we will discuss in detail below in point 4, we showed that instead of 3- day or 7- day incubations, the anti- calbindin scFv can penetrate to the center of a \(300 - \mu \mathrm{m}\) vibratome section with incubation of only one day. We found fewer tissue artifacts in the ultrastructure of 1- day incubated samples than the 7- day samples (see an example in new Supplementary Figure 6, arrows indicating artifacts). We have modified the text to make this point clearer (line 166). So, we conclude that, at least for some scFvs, 1- day incubations are sufficient.
+
+<--- Page Split --->
+![](images/Figure_21.jpg)
+
+<center>Sup. Figure 6. Ultrastructure comparison between samples incubated for one day or seven days. </center>  
+
+Ultrastructure of locations close to the surfaces of 300- μm cerebral cortex sections immunolabeled for one day (a) or seven days (b). Arrows indicate artifacts.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+
+<center>Sup. Figure 21. Well-preserved ultrastructure from the surface (a) to the middle (d) of the \(120 - \mu m\) section. </center>  
+
+Panel 1- 4 in a- d show the ultrastructure at the locations labeled by the red circles in the right panels. Arrows indicate the artifacts potentially caused by prolonged incubation with scFvs for immunolabeling.  
+
+4. This study lack of novelty. The use of scFv has been well-established over the years, making it a solid foundation for the volume CLEM study. However, the choice of detergent-free immunofluorescence labeling raises questions about its suitability. Considering alternative approaches, such as utilizing low concentrations of saponin, might offer a more effective option for preserving membrane morphology during immuno-EM experiments.  
+
+We agree with the reviewer that the use of scFvs is well- established (Bird et al. 1988; Huston et al. 1988; Monnier, Vigouroux, and Tassew 2013; Ahmad et al. 2012). The use of scFvs as immuno- probes for CLEM has been raised in a number of papers in discussion (de Beer and Giepmans 2020; Franek et al. 2024) but, to the best of our knowledge, this is the first actual demonstration of scFvs in volumetric CLEM.  
+
+We agree with the reviewer that when performing volumetric CLEM, alternative immunolabeling approaches other than those that employ scFvs should be considered, such as
+
+<--- Page Split --->
+
+fluorescently tagged primary IgG antibodies with saponin permeabilization. Therefore, in new experiments we compared detergent- free immunolabeling with scFv and Triton- X or saponin- enabled immunolabeling with a dye- directly conjugated monoclonal antibody (mAb) at various detergent concentrations (0.1% and 0.3% Triton- X; 0.05%, 0.1%, and 0.2% saponin) and with two different incubation times (1 day and 7 days). The experiments were performed on 300- μm cerebral cortex tissue blocks with an anti- calbindin L109/57 scFv and a dye- directly conjugated anti- calbindin L109/57 mAb. The epitope binding site of the mAb and the scFv were the same. As shown in new Supplementary Figure 5 a, after 1 day of incubation, only the scFv and the mAb with 0.3% Triton- X penetrated to the middle (i.e., 150 μm) of the section. The other experimental conditions showed various degrees of penetration: 0.1% Triton- X, \(\sim 50\) μm; 0.05% saponin, \(\sim 30\) μm; 0.1% saponin, \(\sim 80\) μm; 0.2% saponin, \(\sim 100\) μm. In all cases with saponin permeabilization, there was a lack of labeling in the cell nuclei (indicated by arrows). When we examined the ultrastructure of these labeled samples, the samples treated with detergent- free scFv labeling showed the best quality. The sample treated with 0.05% saponin showed good- quality EM ultrastructure. All the other samples showed compromised EM ultrastructure, the severity of which increased with the increase of detergent concentration. The membrane breaks in these samples would make automatic segmentation for connectomics challenging, as stated above in our answer to point 1. Although the sample treated with 0.05% saponin for one day showed no obvious ultrastructural artifacts, the mAb penetration was far shallower than the scFv ( \(\sim 30\) μm vs. 150 μm) making volumetric CLEM on the samples larger than the penetration depth difficult.  
+
+As shown in new Supplementary Figure 5 b, after 7 days of incubation, scFvs without detergent and mAb with various concentrations of Triton- X or saponin can penetrate the middle of the \(300 - \mu m\) . However, we still observed in the case of 0.05% saponin a lack of labeling in the cell nuclei (indicated by arrows). Again, when examining the ultrastructure of these labeled samples, the sample treated with detergent- free scFv labeling showed the best quality and is similar to the one- day sample (which we have also mentioned in our answer to reviewer's point 3). All the other samples showed compromised EM ultrastructure, which was much worse when compared with the 1- day samples. Even the 0.05% saponin now showed membrane breaks. We also noticed after 7- day saponin incubation a new artifact: the vesicle- filled axonal profiles in samples treated with saponin for seven days showed a granular texture (indicated by arrowheads in Supplementary Figure 5 b). We think the protein- coagulating function of saponin, as the reviewer stated previously, may be the cause. These granules could pose challenges when synaptic vesicles need to be automatically detected and analyzed (as we did in the later part of this paper) for connectomic studies.  
+
+In addition, as we have mentioned in our answer to point 1, scFvs can penetrate 1- mm tissue blocks while saponin at 0.05% concentration only allows mAbs to penetrate into \(250 \mu m\) after a seven- day incubation (Figure 1 g; new Supplementary Figure 4). These new results suggest that if a researcher wants to do a small- scale volumetric CLEM on a smaller tissue sample (such as several \(\mu m\) to \(50 - \mu m\) ), directly dye- conjugated primary antibodies with a low concentration (0.05%) of saponin with a shorter incubation (one day) may be an option. However, should a researcher need to conduct large- scale volumetric CLEM on larger tissue samples ( \(\sim 1\) mm in thickness), using scFvs for detergent- free immunolabeling is more advantageous. Large- scale volumetric CLEM is especially important for connectomics because a smaller volume is very likely to have fragmented cells/processes that prevent the mapping of the neural circuits. We have modified the text to make these points clearer (line 150; line 162).
+
+<--- Page Split --->
+![](images/Figure_9.jpg)
+
+<center>Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents. </center>  
+
+300- \(\mu \mathrm{m}\) cerebral cortex sections were immunolabeled for one day (a) or seven days (b) with a calbindin- specific scFv conjugated with 5- TAMRA in the absence of detergent or with the scFv's parental mAb conjugated with FL550 in the presence of \(0.1\%\) , \(0.3\%\) Triton- X, or \(0.05\%\) , \(0.1\%\) , \(0.2\%\) saponin. Arrows indicate unlabeled cell nuclei. Arrowheads indicate granular textures associated with the treatment of saponin.
+
+<--- Page Split --->
+
+5. The choice of \(0.3\%\) Triton X-100 to demonstrate detergent issues in EM seems excessive. Most EM studies recommend not exceeding \(0.1\%\) Triton X-100, making it unnecessary to use a higher concentration for this purpose.  
+
+We agree with the reviewer that choosing \(0.3\%\) Triton X- 100 is excessive to demonstrate the detergent's issue on EM ultrastructure. We have changed Figure 1 h and i so the comparison is with a sample treated with \(0.05\%\) saponin.
+
+<--- Page Split --->
+
+## Reviewer #2 (Remarks to the Author):  
+
+This is a technology development manuscript describing the generation and implementation of an assortment of single chain antibody- based probes (scFvs) against different brain proteins to label tissue for correlative fluorescence/volumetric electron microscopy. The key advance is the nature of the labeling probes, which can diffuse deep into fixed brain tissue and penetrate cells without the need for detergent permeabilization. The authors do nice side- by- side comparisons with whole IgG antibodies to highlight this. Detergent- free labeling thus allows processing for ultrastructural analysis by EM, with excellent membrane preservation. Furthermore, the scFv labeling reagents can be easily labeled with different fluorescent dyes allowing the authors to visualize numerous (here they show 6) different labels in the same sample using spectral unmixing confocal microscopy. Serial section EM images of the same samples were then reconstructed and aligned with the fluorescence images to achieve correlated fluorescence/ultrastructure. Overall the data were compelling, with many beautiful examples of correlated fluorescence localization with 3D ultrastructure, nicely demonstrating the power of the technique. While the manuscript primarily focuses on tool development and offers little in the way of novel biological insight, I feel the potential future impact of the technique (i.e. ability to assign neural identities to volumetric connectomics EM datasets, ultrastructural localization of channels, receptors, etc.) will have broad appeal. There are several specific points that deserve attention:  
+
+- Nowhere in the manuscript do the authors validate their labeling reagents in a knockout background. In many cases the labeling pattern is distinct, consistent with previously published work and the localization of the signal makes sense with the correlated ultrastructure (i.e. vGLUT labels presynaptic terminals), but in some cases it is more ambiguous. For example, in Fig. S2a,e the authors argue that the CB and PV scFvs label more of the target proteins in the cell nucleus, is this real signal or are these probes picking up something non-specific in the nucleus that the IgG does not?  
+
+Concerning validation, we agree with the reviewer that the most crucial concern for immuno- probes or any similar affinity probes is whether they label or detect the actual target they are supposed to bind to. There are many cases when antibodies working in ELISA or Western blot settings fail to label their targets or have off- target labeling that creates abnormal background signals in immunohistochemistry (IHC). The parental (aka. progenitor) monoclonal antibodies (mAbs) from the UC Davis/NIH NeuroMab facility, whose sequences were used to generate the eight scFvs in this study, have undergone a strict validation process. In all but one case (the anti- NPY mAb), the mAbs have passed by at least three of the following: immunofluorescence on transfected COS- 1 cells, Western blots on homogenized rat and mouse brains, IHC on rat and mouse brain sections, and IHC on mouse sections in a knockout background. These were accomplished in co- author James Trimmer's lab (for more details, see (Gong, Murray, and Trimmer 2016)). The validation tests of the eight mAbs used in the paper are now shown in new Supplementary Table 4). Although limited by the availability of KO brain samples, the three that we were able to test of them (N206Bb/9, GFAP R416WT; K28/43, PSD- 95; K14/16, Kv 1.2) have passed the test of IHC on WT versus KO mouse brain sections, in that all detectable labeling observed in WT sections was eliminated in KO sections (all three also passed on WT/KO comparison by immunoblot) in a knockout background. While we want to test all the mAbs in a knockout background, but we hope the reviewer understands that it is
+
+<--- Page Split --->
+
+challenging to gather KO animals brain samples for all seven endogenous targets because some may be lethal mutations. 
+
+We also validated the scFvs in each case via IHC on rat and mouse brain sections and by immunofluorescence immunocytochemistry on transiently transfected COS-1 cells (also summarized in Supplementary Table 4; we also provide representative images in new Supplementary Figure 9, 10, 11 for the validation of the N206b/9, anti-GFAP R416WT scFv. Details of the methods of the validation tests for the other scFvs in this paper (and other scFvs) can be found in (Mitchell et al. 2023; Gong, Murray, and Trimmer 2016). We have modified the text to make these points clearer (line 139; line 143). 
+
+Sup. Table 4. Information on the validation of the scFvs and their parental mAbs. 
+
+<table><tr><td>Target</td><td>Clone No.</td><td colspan="4">mAb validation</td><td>scFv validation</td></tr><tr><td></td><td></td><td>COS-IF</td><td>Brain IB</td><td>Brain IHC</td><td>KO Brain IHC</td><td>Method</td></tr><tr><td>GFP</td><td>N86/38</td><td>Pass</td><td>NA</td><td>NA</td><td>NA</td><td>COS-IF</td></tr><tr><td>Calbindin</td><td>L109/57</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>GFAP</td><td>N206B/9</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>R416WT</td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>VGlut1</td><td>N28/9</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>PSD-95</td><td>K28/43</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>Kv 1.2</td><td>K14/16</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>Parvalbumin</td><td>L114/81</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>NPY</td><td>L115/13</td><td>Pass</td><td>Fail</td><td>Pass</td><td>ND</td><td>COS-IF</td></tr></table>
+
+![](images/Figure_10.jpg)
+
+ 
+
+<center>Sup. Figure 9. Validation of anti-GFAP scFv with Immunofluorescence immunocytochemistry.</center> 
+
+Immunofluorescence immunocytochemistry on transiently transfected cells. COS-1 cells (top row) and HEK293T cells (bottom row). Cells were transfected with a plasmid mEmerald-tagged human GFAP
+
+<--- Page Split --->
+
+(green) and double immunolabeled with the 5-TAMRA-labeled anti-GFAP N206B/9 scFv (red) and the anti-GFAP N206A/8 mouse IgG1 mAb (white). Hoechst nuclear labeling is shown in blue.  
+
+![](images/Figure_11.jpg)
+
+<center>Sup. Figure 10. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (cerebellum). </center>  
+
+GFAP scFv and original monoclonal antibody from which it was derived display the same tissue labeling pattern of a sagittal section through the rat cerebellum. A) Glial cells throughout the cerebellar granule cell layer (GCL) and prominent Bergmann glial process in molecular layer (ML) are labeled with hybridoma derived monoclonal antibody N206B/9. B) merged image includes labeling with a polyclonal rabbit antibody (KC) against the neuronal potassium channel Kv2.1, monoclonal antibody targeting glial specific RNA binding protein QKI (N147/6) and nuclear specific Hoechst labeling. C) An adjacent section labeled with N206B/9 derived scFv shows the same pattern of labeling. D) merged image with the same additional labeling as B.
+
+<--- Page Split --->
+![](images/Figure_14.jpg)
+
+<center>Sup. Figure 11. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (hippocampus). </center>  
+
+Validation of scFv labeling pattern against hybridoma- generated monoclonal antibody N206B/9 from which it was derived. Multiplex immunofluorescent labeling of a sagittal section through rat hippocampal region CA1. A) 5- TAMRA conjugated N206B/9 derived scFv, B) Hybridoma derived monoclonal antibody N206B/9 indirectly labeled with Alexa fluor 647 conjugated goat anti- mouse IgG1 secondary antibody, C) merged images from A and B illustrating co- labeled astroglial cells (e.g arrowheads). D) Same multiplex image shown in C with additional neuronal specific potassium channel Kv2.1, glial specific pan- QKI RNA binding protein, and DNA marker Hoechst 33342 labeling.  
+
+The second issue raised is concern over the ambiguous signals detected by the anti- calbindin and anti- parvalbumin scFvs in the cell nuclei of Purkinje cells. Clearly, in the original Supplementary Figure 2, the mAbs for calbindin and parvalbumin did not detect signals in the cell nuclei as the scFvs did. In the revised paper we now include the validation IHC images of the anti- calbindin and anti- parvalbumin mAbs from experiments done in James Trimmer's lab (new Supplementary Figure 14). These images showed the normal expected labeling, that included signal in some cell nuclei of Purkinje cells in the lateral hemisphere of the rat cerebellum. In the original manuscript, we mentioned previous studies reporting the detection of calbindin and parvalbumin in cell nuclei of Purkinje cells (Celio 1990; Brandenburg et al. 2021; German et al. 1997; Schmidt et al. 2007). We believe calbindin and parvalbumin are present in the cell nuclei. We therefore suspect that the reason our mAb labeling didn't detect signals in the cell nuclei is insufficient antibody incubation time.
+
+<--- Page Split --->
+
+To address this problem, in Triton- X- treated samples, we first did immunolabeling with the same anti- calbindin and anti- parvalbumin mAbs directly conjugated with the red fluorophore FL- 550 in distinction to using secondary antibodies as we did previously. The directly conjugated mAbs exclude the requirement for secondary antibodies allowing for only one round of incubation. We extended the incubation time to seven days (versus 2 days of incubation with primary antibodies previously). The results showed that both mAbs can detect signals in the cell nuclei of most ( \(> - 90\%\) ) Purkinje cells (new Supplementary Figure 15 c, I). This staining is similar to the scFv labeling except that scFvs were detected in all Purkinje cells. The second approach we attempted was immunolabeling with commercial polyclonal antibodies (pAbs) against calbindin and parvalbumin. These antibodies are not directly conjugated with fluorophores, so we utilized (Fab)2 as fluorescently tagged secondaries, which are smaller than conventional secondaries and supposedly can diffuse in tissue better. Our results showed that, again, both calbindin and parvalbumin could be detected by the pAbs in the cell nuclei in most but not all Purkinje cells (new Supplementary Figure 15 d and m). The third approach we used, was to slice the section in a different orientation to cut through the nuclei of most Purkinje cells in order to gain direct access to the nuclei in the vibratome section (see new Supplementary Figure 15 e). We immunolabeled with the same anti- calbindin or anti- parvalbumin mAbs directly conjugated with the fluorophore FL- 550 with a seven- day incubation. This time, we observed labeling in nearly all Purkinje cells (Supplementary Figure 15 f and n). These results indicate that the signals detected in the cell nuclei by scFvs are real signals. The reason it is relatively harder for full- length antibodies (mAbs or pAb) to detect these signals can be attributed to their relatively weaker penetration ability, even in the presence of Triton- X. This can be addressed in ways like extending incubation time or slicing sections to expose the internal epitopes in cell nuclei better. We have modified the text to make these points clearer (line 148).
+
+<--- Page Split --->
+![](images/Figure_2.jpg)
+
+<center>Sup. Figure 14. Validation of anti-calbindin and anti-parvalbumin scFvs with Immunofluorescence immunohistochemistry. </center>  
+
+Labeling pattern of original mAbs used to generate scFvs against Parvalbumin and Calbindin in rat cerebellum. Sagittal section through cerebellum labeled with monoclonal antibodies L127/8 (A, GAD1), L114/8 R (B, PARV), and L109/57 (C, CALB1). The merged image (D) shows the colocalized pattern of labeling within Purkinje cell layer (PCL). ML, molecular layer, GCL, granule cell layer.
+
+<--- Page Split --->
+![](images/Figure_12.jpg)
+
+
+<--- Page Split --->
+
+Sup. Figure 15. Validation of immunofluorescence by scFv probes and their parental mAbs (part 3).  
+
+Cerebellum Crus 1 sections were immunolabeled with a calbindin- specific scFv (a), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (b), the mAb conjugated with FL550 (c), or a commercial calbindin- specific pAb and secondary (Fab)2 conjugated with Alexa Fluor 594 (d). e, Schematics showing the cutting orientation that is parallel to the lobule of Crus 1, which intersects perpendicular to the planer Purkinje cells in Crus 1. f, Sections cut in this orientation immunolabeled with the mAb conjugated with FL550. The boxed inset is shown enlarged in the adjacent panel. Whole- section images of cerebellum Crus 1 sections immunolabeled with a calbindin- specific scFv (g), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (h), or the mAb conjugated with FL550 (i). Arrows indicate labeled cell nuclei of Purkinje cells. Arrowheads indicate the labeled axons.  
+
+Cerebellum Crus 1 sections were immunolabeled with a parvalbumin- specific scFv (j), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (k), the mAb conjugated with FL550 (l), or a commercial parvalbumin- specific pAb and secondary (Fab)2 conjugated with Alexa Fluor 594 (m). f, Sections cut in this orientation in a immunolabeled with the mAb conjugated with FL550. The boxed inset is shown enlarged in the adjacent panel. Whole- section images of cerebellum Crus 1 sections immunolabeled with a parvalbumin- specific scFv (o), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (p), or the mAb conjugated with FL550 (q). Arrows indicate labeled cell nuclei of Purkinje cells. Arrowheads indicate the labeled axons.  
+
+- While many of the images are quite striking, overall the manuscript lacked any sort of quantitative analysis. Just as one example, in Fig. 1, showing a simple pixel correlation scatter plot comparing the YFP and GFP-scFv signal would give readers a better idea of how evenly the scFv is penetrating cells to label YFP.  
+
+We thank the reviewer for highlighting the lack of quantitative analysis in comparing the specificity of scFvs and mAbs. The paper does contain other quantitative analyses (see Figure 7; Supplementary Figure 29; Supplementary Table 5, 6) but in response to the specific question raised, we have now created pixel correlation scatter plots for three images from two cortical and one hippocampal section from YFP-H mice, which were also immunolabeled with the anti- GFP scFv (new Supplementary Figure 2). Supplementary Figure 2 is the raw image of Figure 1 b. As is shown in all three pixel correlating scatter plots, the signals from the scFv labeling (red) correlate with the native YFP fluorescence signal (green). There are pixels that only have values in the green channels, which correspond to the insufficiently labeled axons pointed out in Figure 2 a. There are very few pixels that only have values in the red (scFv) channel, which indicates that there is minimal off-target labeling of this anti-GFP scFv. This analysis gives us confidence in the specify of the scFv for green fluorescent protein. Doing this kind of double labeling is more problematic when comparing scFvs to mAbs that have the identical paratope as they compete for the same site. So, in these cases, as described in detail above, we had to be content with the comparative labeling of different tissue sections. We have modified the text to make this point clearer (line 128).
+
+<--- Page Split --->
+![](images/Figure_1.jpg)
+
+<center>Sup. Figure 2. Pixel correlation scatter plots comparing the native YFP fluorescence signal and the red fluorescence from the labeling of the GFP-specific scFv. </center>  
+
+Cerebral cortex samples (a and b) and hippocampus (c) from YFP- H mice were immunolabeled with a GFP- specific scFv conjugated with the red fluorophore 5- TAMRA. The images are raw data without any brightness/contrast adjustment. a is the raw image data of Figure 1 b.  
+
+- p. 9 "...immunofluorescence patterns that were similar to or in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex (Figure 2 a; Sup. Figure 2; Sup. Figure 3). In many cases the comparisons between mAb and scFv is not entirely fair since the mAb is labeled in the 488/green channel (in which brain tissue notoriously has more autofluorescence) and the scFv in the red channel e.g. NPY signal in Sup. 2b,d; PSD95 in Sup. 2f. Is the labeling
+
+<--- Page Split --->
+
+really that much cleaner or is the background signal in the green channel making the mAb appear worse than it is?  
+
+We apologize that the phrasing in our original manuscript may have caused a misunderstanding. In this sentence, "in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex." only refers to the cases of calbindin and parvalbumin, as discussed in the previous point. In the other cases (GFAP, VGlut1, Kv 1.2, PSD- 95, and NPY), we believe our results suggest that the labeling of scFvs and mAbs are comparably good in terms of both signal and background. We understand the legitimate concern that the tissue sections fixed with formaldehyde and glutaraldehyde may have a higher background in the 488/green channel. Glutaraldehyde especially has stronger autofluorescence (Fischer et al. 2008). However, we only used \(0.1\%\) glutaraldehyde in our preparation prior to osmium staining. After adequate washing with PBS, the brain tissue sections do not show strong autofluorescence in the 488/green channel (as now shown in new Supplementary Figure 12 a). There is some autofluorescence, mostly from lipofuscin granules in cell bodies (arrows in new Supplementary Figure 12 a) with broad excitation/emission spectra (Di Guardo 2015; Marmorstein et al. 2002). But this autofluorescence is found in all channels. We emphasize that we are not attempting to make a case that scFv labeling is cleaner than that obtained with mAb, as the results from both are very similar. Indeed we also performed new experiments with red fluorophore- conjugated mAbs for calbindin and parvalbumin, as discussed in our answer to the reviewer's point 1, and didn't find any difference in the background level (new Supplementary Figure 15). We have adjusted the phrasing in the manuscript (line 145) to avoid any further misunderstanding.  
+
+![](images/Figure_3.jpg)
+
+<center>Sup. Figure 12. Validation of immunofluorescence by scFv probes and their parental mAbs (part 1). </center>  
+
+a, Confocal images from unlabeled the cerebral cortex and cerebellar cortex of a wild-type mouse showed limited background in the 488/green channel. Arrows indicate background signals lipofuscin granule. b- d, Cerebellum Crus 1 sections were immunolabeled with scFvs targeting VGlut1, GFAP, and
+
+<--- Page Split --->
+
+Kv 1.2; or these scFvs' parental mAbs and secondary antibodies conjugated with Alexa Fluor 488.  
+
+-p. 9 "......found that the anti-calbindin scFv penetrated to a depth of \(\sim 150 \mu m\) in a 300- \(\mu m\) tissue slice". So the probe labeled throughout the entire slice?  
+
+We apologize for the lack of clarity, yes, we meant that they labeled through the entire 300 \(\mu m\) slice (150 \(\mu m\) from each side). Given the recent availability of directly conjugated mAbs we have removed Figure 1 d and e to Supplementary Figure 3 b (we remade figures from raw images showing the penetration across the 300 \(\mu m\) thickness) and c, and added new Figure 1 f- i, Supplementary Figure 4 and 5 of the results of a comparable experiment of scFv labeling on 300- \(\mu m\) and 1- mm thick brain tissue sections in the absence of detergent. In Figure 1 and Supplementary Figure 4, scFvs are shown to label throughout a 1- mm thickness with a seven- day incubation; in Supplementary Figure 5, scFvs can label a 300- \(\mu m\) thickness sample with either a 1- or 3- day incubation. EM ultrastructure of all these samples was good. We have modified the text to make these points clearer (line 150).
+
+<--- Page Split --->
+![](images/Figure_4.jpg)
+
+<center>Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure. </center>  
+
+a, Schematic representations of a full-length IgG antibody and an scFv probe with a conjugated fluorescent dye. b, Confocal images from the cerebral cortex of a YFP-H mouse labeled using a GFP-specific scFv probe conjugated with the red dye 5-TAMRA. Arrows show thinner neuronal processes, perhaps myelinated, that are not labeled by scFv. c, Layer ⅔ of the cerebral cortex labeled with a calbindin-specific scFv probe. d, Cerebellum cortex of Crus 1 labeled with the PSD-95 specific scFv. Right panel is the enlarged boxed inset from left. e, Cerebral cortex labeled with the NPY-specific scFv. f and g, Tissue penetration depth comparison of a parvalbumin-specific scFv without detergent and its parental
+
+<--- Page Split --->
+
+mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1-mm cerebral cortex tissue sections with a 7-day incubation. h and i, Comparison of ultrastructure from samples incubated 7 days without detergent and with \(0.05\%\) saponin. Arrows indicate membrane breaks. Asterisks indicate abnormal appearing vesicle-filled axonal terminals.  
+
+![](images/Figure_5.jpg)
+
+<center>Sup. Figure 3. Immunolabeling results of anti-calbindin scFv and tissue penetration comparison. </center>  
+
+a, Additional brain regions labeled with a calbindin-specific scFv probe conjugated with 5-TAMRA. The arrow in the left panel shows myelinated Purkinje cell axons in the granule layer. b, Tissue penetration
+
+<--- Page Split --->
+
+depth comparison of scFvs, mAbs (plus secondary antibodies), and the role of detergents in improving the depth of labeling. c, Comparison of ultrastructure with and without \(0.5\%\) Triton X- 100 on scFv labeled samples. Boxed insets are shown at higher magnification in adjacent panels. of \(\sim 30 \mu m\) ; the nanobody can penetrate into a depth of \(\sim 150 \mu m\) .  
+
+![PLACEHOLDER_29_0]
+
+<center>Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample. </center>  
+
+Tissue penetration depth comparison of a calbindin- specific scFv without detergent and its parental mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation.
+
+<--- Page Split --->
+![PLACEHOLDER_30_0]
+
+<center>Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents. </center>  
+
+300- \(\mu \mathrm{m}\) cerebral cortex sections were immunolabeled for one day (a) or seven days (b) with a calbindin- specific scFv conjugated with 5- TAMRA in the absence of detergent or with the scFv's parental mAb conjugated with FL550 in the presence of \(0.1\%\) , \(0.3\%\) Triton- X, or \(0.05\%\) , \(0.1\%\) , \(0.2\%\) saponin. Arrows indicate unlabeled cell nuclei. Arrowheads indicate granular textures associated with the treatment of saponin.
+
+<--- Page Split --->
+
+597 Ahmad, Zuhaila Asra, Swee Keong Yeap, Abdul Manaf Ali, Wan Yong Ho, Noorjahan Banu 598 Mohamed Alitheen, and Muhajir Hamid. 2012. "scFv Antibody: Principles and Clinical 599 Application." Clinical & Developmental Immunology 2012 (March): 980250. 600 Alric, Christophe, Katel Hervé-Aubert, Nicolas Aubrey, Souad Melouk, Laurie Lajoie, William 601 Même, Sandra Même, et al. 2018. "Targeting HER2-Breast Tumors with scFv-Decorated 602 Bimodal Nanoprobes." Journal of Nanobiotechnology 16 (1): 18. 603 Beer, Marit A. de, and Ben N. G. Giepmans. 2020. "Nanobody-Based Probes for Subcellular 604 Protein Identification and Visualization." Frontiers in Cellular Neuroscience 14 (November): 573278. 605 Bernard, Clémence, Clémentine Vincent, Damien Testa, Eva Bertini, Jérôme Ribot, Ariel A. Di 606 Nardo, Michel Volovitch, and Alain Prochiantz. 2016. "A Mouse Model for Conditional 607 Secretion of Specific Single-Chain Antibodies Provides Genetic Evidence for Regulation of 608 Cortical Plasticity by a Non-Cell Autonomous Homeoprotein Transcription Factor." PLoS 609 Genetics 12 (5): e1006035. 610 Bird, R. E., K. D. Hardman, J. W. Jacobson, S. Johnson, B. M. Kaufman, S. M. Lee, T. Lee, S. 611 H. Pope, G. S. Riordan, and M. Whitlow. 1988. "Single-Chain Antigen-Binding Proteins." 612 Science 242 (4877): 423-26. 613 Blanc, Hugo, Gabriel Kaddour, Nicolas B. David, Willy Supatto, Jean Livet, Emmanuel 614 Beaurepaire, and Pierre Mahou. 2023. "Chromatically Corrected Multicolor Multiphoton 615 Microscopy." ACS Photonics 10 (12): 4104-11. 616 Brandenburg, Cheryl, Lindsey A. Smith, Michaela B. C. Kilander, Morgan S. Bridi, Yu-Chih Lin, 617 Shiyong Huang, and Gene J. Blatt. 2021. "Parvalbumin Subtypes of Cerebellar Purkinje 618 Cells Contribute to Differential Intrinsic Firing Properties." Molecular and Cellular 619 Neurosciences 115 (September): 103650. 620 Celio, M. R. 1990. "Calbindin D-28k and Parvalbumin in the Rat Nervous System." 621 Neuroscience 35 (2): 375-475. 622 Cheng, Ruoyu, Feng Zhang, Meng Li, Xiang Wo, Yu-Wen Su, and Wei Wang. 2019. "Influence 623 of Fixation and Permeabilization on the Mass Density of Single Cells: A Surface Plasmon 624 Resonance Imaging Study." Frontiers in Chemistry 7 (August): 588. 625 Di Guardo, G. 2015. "Lipofuscin, Lipofuscin-like Pigments and Autofluorescence." European 626 Journal of Histochemistry: EJH 59 (1): 2485. 627 Fischer, Andrew H., Kenneth A. Jacobson, Jack Rose, and Rolf Zeller. 2008. "Fixation and 628 Permeabilization of Cells and Tissues." CSH Protocols 2008 (May): db.top36. 629 Fox, C. H., F. B. Johnson, J. Whiting, and P. P. Roller. 1985. "Formaldehyde Fixation." The 630 Journal of Histochemistry and Cytochemistry: Official Journal of the Histochemistry Society 631 33 (8): 845-53. 632 Franek, Michal, Lenka Koptasikova, Jiri Miksátko, David Liebl, Eliška Macíčková, Jakub 633 Pospíšil, Milan Esner, Martina Dvořáčková, and Jiří Fajkus. 2024. "In-Section Click-iT 634 Detection and Super-Resolution CLEM Analysis of Nucleolar Ultrastructure and Replication 635 in Plants." Nature Communications 15 (1): 2445. 636 Fulton, Kara A., and Kevin L. Briggman. 2021. "Permeabilization-Free En Bloc 637 Immunohistochemistry for Correlative Microscopy." eLife 10 (May). 638 https://doi.org/10.7554/eLife.63392. 639 Furuta, Takahiro, Kenta Yamauchi, Shinichiro Okamoto, Megumu Takahashi, Soichiro Kakuta, 640 Yoko Ishida, Aya Takenaka, et al. 2022. "Multi-Scale Light Microscopy/electron Microscopy 641 Neuronal Imaging from Brain to Synapse with a Tissue Clearing Method, ScaleSF." 642 iScience 25 (1): 103601. 643 German, D. C., M. C. Ng, C. L. Liang, A. McMahon, and A. M. Iacopino. 1997. "Calbindin-D28k 644 in Nerve Cell Nuclei." Neuroscience 81 (3): 735-43.
+
+<--- Page Split --->
+
+646 Gong, Belvin, Karl D. Murray, and James S. Trimmer. 2016. “Developing High- Quality Mouse Monoclonal Antibodies for Neuroscience Research - Approaches, Perspectives and Opportunities.” New Biotechnology 33 (5 Pt A): 551–64. 647 Huston, J. S., D. Levinson, M. Mudgett- Hunter, M. S. Tai, J. Novotný, M. N. Margolies, R. J. Ridge, R. E. Bruccoleri, E. Haber, and R. Crea. 1988. “Protein Engineering of Antibody Binding Sites: Recovery of Specific Activity in an Anti- Digoxin Single- Chain Fv Analogue Produced in Escherichia Coli.” Proceedings of the National Academy of Sciences of the United States of America 85 (16): 5879–83. 654 Ichikawa, Takehiko, Dong Wang, Keisuke Miyazawa, Kazuki Miyata, Masanobu Oshima, and Takeshi Fukuma. 2022. “Chemical Fixation Creates Nanoscale Clusters on the Cell Surface by Aggregating Membrane Proteins.” Communications Biology 5 (1): 487. 657 Im, Sun- Woo, Hee Yong Chung, and Young- Ju Jang. 2017. “Development of Single- Chain Fv of Antibody to DNA as Intracellular Delivery Vehicle.” Animal Cells and Systems 21 (6): 382–87. 660 Januszewski, Michał, Jörgen Kornfeld, Peter H. Li, Art Pope, Tim Blakely, Larry Lindsey, Jeremy Maitin- Shepard, Mike Tyka, Winfried Denk, and Viren Jain. 2018. “High- Precision Automated Reconstruction of Neurons with Flood- Filling Networks.” Nature Methods 15 (8): 605–10. 664 Kiernan, John A. 2000. “Formaldehyde, Formalin, Paraformaldehyde And Glutaraldehyde: What They Are And What They Do.” Microscopy Today 8 (1): 8–13. 666 Kim, Eunhee G., Jieun Jeong, Junghyeon Lee, Hyeryeon Jung, Minho Kim, Yi Zhao, Eugene C. Yi, and Kristine M. Kim. 2020. “Rapid Evaluation of Antibody Fragment Endocytosis for Antibody Fragment- Drug Conjugates.” Biomolecules 10 (6). 669 https://doi.org/10.3390/biom10060955. 670 Li, Tengfei, Matthias Vandesquille, Fani Koukouli, Clémence Dudeffant, Ihsen Youssef, Pascal Lenormand, Christelle Ganneau, et al. 2016. “Camelid Single- Domain Antibodies: A Versatile Tool for in Vivo Imaging of Extracellular and Intracellular Brain Targets.” Journal of Controlled Release: Official Journal of the Controlled Release Society 243 (December): 1–10. 674 Lu, Xiaotang, Xiaomeng Han, Yaron Meirovitch, Evelina Sjöstedt, Richard L. Schalek, and Jeff W. Lichtman. 2023. “Preserving Extracellular Space for High- Quality Optical and Ultrastructural Studies of Whole Mammalian Brains.” Cell Reports Methods 3 (7): 100520. 678 Mahou, Pierre, Maxwell Zimmerley, Karine Loulier, Katherine S. Matho, Guillaume Labroille, Xavier Morin, Willy Supatto, Jean Livet, Delphine Débarre, and Emmanuel Beaurepaire. 2012. “Multicolor Two- Photon Tissue Imaging by Wavelength Mixing.” Nature Methods 9 (8): 815–18. 682 Marmorstein, Alan D., Lihua Y. Marmorstein, Hirokazu Sakaguchi, and Joe G. Hollyfield. 2002. “Spectral Profiling of Autofluorescence Associated with Lipofuscin, Bruch’s Membrane, and Sub- RPE Deposits in Normal and AMD Eyes.” Investigative Ophthalmology & Visual Science 43 (7): 2435–41. 686 Mitchell, Keith G., Belvin Gong, Samuel S. Hunter, Diana Burkart- Waco, Clara E. Gavira- O’Neill, Kayla M. Templeton, Madeline E. Goethel, et al. 2023. “High- Volume Hybridoma Sequencing on the NeuroMabSeq Platform Enables Efficient Generation of Recombinant Monoclonal Antibodies and scFvs for Neuroscience Research.” Scientific Reports 13 (1): 16200. 691 Monnier, Philippe P., Robin J. Vigouroux, and Nardos G. Tassew. 2013. “In Vivo Applications of Single Chain Fv (Variable Domain) (scFv) Fragments.” Antibodies 2 (2): 193–208. 693 Pudavar, Haridas, Judith Reddington, Jason A. Junge, Scott E. Fraser, and Giulia Ossato. 2024. “STELLARIS 8 DIVE: A Rainbow of Possibilities with Multiphoton Excitation and Lifetime- Based Information.” In Multiphoton Microscopy in the Biomedical Sciences XXIV, 12847:32–33. SPIE.
+
+<--- Page Split --->
+
+697 Schmid, Benjamin, Gopi Shah, Nico Scherf, Michael Weber, Konstantin Thierbach, Citlali Pérez 698 Campos, Ingo Roeder, Pia Aanstad, and Jan Huisken. 2013. “High-Speed Panoramic Light- 699 Sheet Microscopy Reveals Global Endodermal Cell Dynamics.” Nature Communications 4: 700 2207. 701 Schmidt, Hartmut, Oliver Arendt, Edward B. Brown, Beat Schwaller, and Jens Eilers. 2007. 702 “Parvalbumin Is Freely Mobile in Axons, Somata and Nuclei of Cerebellar Purkinje 703 Neurones.” Journal of Neurochemistry 100 (3): 727–35. 704 Shapson-Coe, Alexander, Michał Januszewski, Daniel R. Berger, Art Pope, Yuelong Wu, Tim 705 Blakely, Richard L. Schalek, et al. 2021. “A Connectomic Study of a Petascale Fragment of 706 Human Cerebral Cortex.” bioRxiv. https://doi.org/10.1101/2021.05.29.446289. 707 Thavarajah, Rooban, Vidya Kazhiyur Mudimbaimannar, Joshua Elizabeth, Umadevi 708 Krishnamohan Rao, and Kannan Ranganathan. 2012. “Chemical and Physical Basics of 709 Routine Formaldehyde Fixation.” Journal of Oral and Maxillofacial Pathology: JOMFP 16 710 (3): 400–405. 711 Thiel, M. A., D. J. Coster, S. D. Standfield, H. M. Brereton, C. Mavrangelos, H. Zola, S. Taylor, 712 A. Yusim, and K. A. Williams. 2002. “Penetration of Engineered Antibody Fragments into 713 the Eye.” Clinical and Experimental Immunology 128 (1): 67–74. 714 Tsuriel, Shlomo, Sagi Gudes, Ryan W. Draft, Alexander M. Binshtok, and Jeff W. Lichtman. 715 2015. “Multispectral Labeling Technique to Map Many Neighboring Axonal Projections in 716 the Same Tissue.” Nature Methods 12 (6): 547–52. 717 Weisser, Nina E., and J. Christopher Hall. 2009. “Applications of Single-Chain Variable 718 Fragment Antibodies in Therapeutics and Diagnostics.” Biotechnology Advances 27 (4): 719 502–20. 720 Wittrup, Anders, Si-He Zhang, Gerdy B. ten Dam, Toin H. van Kuppevelt, Per Bengtson, Maria 721 Johansson, Johanna Welch, Matthias Mörgelin, and Mattias Belting. 2009. “ScFv Antibody- 722 Induced Translocation of Cell-Surface Heparan Sulfate Proteoglycan to Endocytic 723 Vesicles.” The Journal of Biological Chemistry 284 (47): 32959–67.
+
+<--- Page Split --->
+
+## REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors have developed a method employing fluorescent single- chain variable fragments (scFvs) for conducting multiplexed detergent- free immunolabeling and volumetric- correlated- light- and- electron- microscopy (vCLEM) on the same samples. In this manuscript, they detail the development of eight fluorescent scFvs targeting specific markers crucial for brain studies. Through experimentation, six fluorescent probes were successfully visualized in the cerebellum using confocal microscopy with spectral unmixing, followed by vEM analysis of the identical sample. The outcomes reveal an exceptional blend of ultrastructure alongside multiple fluorescence channels, offering valuable insights into cellular composition and organization.  
+
+This approach facilitated the documentation of a previously poorly characterized cell type and the precise subcellular localization. Leveraging scFvs derived from existing monoclonal antibodies opens up the possibility of generating numerous such probes, which could significantly enhance molecular overlays for connectomic investigations.  
+
+This study represents a significant advancement in the field of vCLEM and holds great promise for researchers seeking to unravel intricate neuronal networks. The revised version of the manuscript effectively addresses key questions and concerns. I don't have further comments.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The revised manuscript thoroughly addresses all of the concerns raised in my initial review. I think the manuscript will have broad appeal and I appreciate the effort put in to address the referees' comments. I strongly recommend publication in Nature Communications.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+The authors have developed a method employing fluorescent single- chain variable fragments (scFvs) for conducting multiplexed detergent- free immunolabeling and volumetric- correlated- light- and- electron- microscopy (vCLEM) on the same samples. In this manuscript, they detail the development of eight fluorescent scFvs targeting specific markers crucial for brain studies. Through experimentation, six fluorescent probes were successfully visualized in the cerebellum using confocal microscopy with spectral unmixing, followed by vEM analysis of the identical sample. The outcomes reveal an exceptional blend of ultrastructure alongside multiple fluorescence channels, offering valuable insights into cellular composition and organization.  
+
+This approach facilitated the documentation of a previously poorly characterized cell type and the precise subcellular localization. Leveraging scFvs derived from existing monoclonal antibodies opens up the possibility of generating numerous such probes, which could significantly enhance molecular overlays for connectomic investigations.  
+
+This study represents a significant advancement in the field of vCLEM and holds great promise for researchers seeking to unravel intricate neuronal networks. The revised version of the manuscript effectively addresses key questions and concerns. I don't have further comments.  
+
+We thank the reviewer's comments. There's nothing to be further addressed.  
+
+## Reviewer #2 (Remarks to the Author):  
+
+The revised manuscript thoroughly addresses all of the concerns raised in my initial review. I think the manuscript will have broad appeal and I appreciate the effort put in to address the referees' comments. I strongly recommend publication in Nature Communications.  
+
+We thank the reviewer's comments. There's nothing to be further addressed.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d5ef8fed2a8e3dbecb979f0481b24ab88c243101d124d78b8ab1cdf9509460c5/supplementary_0_Peer Review File__d5ef8fed2a8e3dbecb979f0481b24ab88c243101d124d78b8ab1cdf9509460c5_det.mmd

@@ -0,0 +1,431 @@
+<|ref|>title<|/ref|><|det|>[[99, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[106, 110, 373, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[108, 162, 872, 219]]<|/det|>
+Multiplexed volumetric CLEM enabled by scFvs provides new insights into the cytology of cerebellar cortex  
+
+<|ref|>image<|/ref|><|det|>[[95, 732, 262, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[271, 732, 880, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[116, 88, 305, 104]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[116, 121, 402, 137]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 872, 280]]<|/det|>
+The authors developed eight smaller single- chain variable fragments (scFvs) based on eight well- characterized mAbs and conjugated them with various fluorescent dyes. This idea is great for volume correlative light and electron microscopy. The authors reported that each scFv proved effective as a detergent- free immunofluorescent probe. The approach was believed to be promising for routine linking of molecular information to connectomic information from the same material since the quality of data from the volumetric fluorescent and electron microscopy is good. However, some critical aspects of the experimental design need to be addressed, and additional experiments are required to draw conclusive findings.  
+
+<|ref|>text<|/ref|><|det|>[[115, 280, 880, 456]]<|/det|>
+1. One concern in this study is to choose Triton X-100 as a detergent for comparison. While Triton X-100 is commonly used to aid antibody penetration in light microscopy, it is not widely used for EM or immuno-EM studies. For EM studies, saponin is one of organic solvents which dissolve lipids from cell membranes making them permeable to antibodies. Furthermore, organic solvents can be used to fix and permeabilize cells at the same time by coagulating proteins. Saponin interacts with membrane cholesterol, selectively removing it and leaving holes in the membrane. Most researchers in the EM field choose low concentration of saponin to balance the antibody penetration and excellent membrane morphology in their immuno-EM experiments. Majority of their immuno-EM images showed great morphology of membrane at the ultrastructural level with the use of saponin. In some immuno-EM studies, Triton X-100 has been used. But those studies don't care the cell membrane, most of them were interested in some subcellular organelles.  
+
+<|ref|>text<|/ref|><|det|>[[115, 456, 870, 520]]<|/det|>
+2. Another point of concern is the lack of clarity on how the antibodies penetrate the cell through the cell membrane in detergent-free immunofluorescence labeling. A more thorough investigation and explanation of this process are needed to provide a comprehensive understanding of the technique's efficacy.  
+
+<|ref|>text<|/ref|><|det|>[[115, 521, 872, 600]]<|/det|>
+3. The animals were perfused with the fixative (4% paraformaldehyde + 0.1% glutaraldehyde). The low concentration of glutaraldehyde (0.1%) may not be sufficient to adequately preserve the lipids in the cell membrane. However, for detergent-free immunofluorescence labeling, scFv probes or nanobody probes were incubated for 3 days (50 \(\mu \mathrm{m}\) ) or 7 days (120 \(\mu \mathrm{m}\) ). This could potentially affect the ultrastructural morphology.  
+
+<|ref|>text<|/ref|><|det|>[[115, 601, 876, 681]]<|/det|>
+4. This study lack of novelty. The use of scFv has been well-established over the years, making it a solid foundation for the volume CLEM study. However, the choice of detergent-free immunofluorescence labeling raises questions about its suitability. Considering alternative approaches, such as utilizing low concentrations of saponin, might offer a more effective option for preserving membrane morphology during immuno-EM experiments.  
+
+<|ref|>text<|/ref|><|det|>[[116, 681, 870, 729]]<|/det|>
+5. The choice of 0.3% Triton X-100 to demonstrate detergent issues in EM seems excessive. Most EM studies recommend not exceeding 0.1% Triton X-100, making it unnecessary to use a higher concentration for this purpose.  
+
+<|ref|>text<|/ref|><|det|>[[116, 777, 403, 792]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 809, 880, 904]]<|/det|>
+This is a technology development manuscript describing the generation and implementation of an assortment of single chain antibody- based probes (scFvs) against different brain proteins to label tissue for correlative fluorescence/volumetric electron microscopy. The key advance is the nature of the labeling probes, which can diffuse deep into fixed brain tissue and penetrate cells without the need for detergent permeabilization. The authors do nice side- by- side comparisons with whole IgG antibodies to highlight this. Detergent- free labeling thus allows processing for ultrastructural analysis
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 88, 880, 265]]<|/det|>
+by EM, with excellent membrane preservation. Furthermore, the scFv labeling reagents can be easily labeled with different fluorescent dyes allowing the authors to visualize numerous (here they show 6) different labels in the same sample using spectral unmixing confocal microscopy. Serial section EM images of the same samples were then reconstructed and aligned with the fluorescence images to achieve correlated fluorescence/ultrastructure. Overall the data were compelling, with many beautiful examples of correlated fluorescence localization with 3D ultrastructure, nicely demonstrating the power of the technique. While the manuscript primarily focuses on tool development and offers little in the way of novel biological insight, I feel the potential future impact of the technique (i.e. ability to assign neural identities to volumetric connectomics EM datasets, ultrastructural localization of channels, receptors, etc.) will have broad appeal. There are several specific points that deserve attention:  
+
+<|ref|>text<|/ref|><|det|>[[114, 280, 877, 377]]<|/det|>
+- Nowhere in the manuscript do the authors validate their labeling reagents in a knockout background. In many cases the labeling pattern is distinct, consistent with previously published work and the localization of the signal makes sense with the correlated ultrastructure (i.e. vGLUT labels presynaptic terminals), but in some cases it is more ambiguous. For example, in Fig. S2a,e the authors argue that the CB and PV scFvs label more of the target proteins in the cell nucleus, is this real signal or are these probes picking up something non-specific in the nucleus that the IgG does not?  
+
+<|ref|>text<|/ref|><|det|>[[114, 392, 881, 456]]<|/det|>
+- While many of the images are quite striking, overall the manuscript lacked any sort of quantitative analysis. Just as one example, in Fig. 1, showing a simple pixel correlation scatter plot comparing the YFP and GFP-scFv signal would give readers a better idea of how evenly the scFv is penetrating cells to label YFP.  
+
+<|ref|>text<|/ref|><|det|>[[114, 472, 875, 568]]<|/det|>
+- p. 9 "....immunofluorescence patterns that were similar to or in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex (Figure 2 a; Sup. Figure 2; Sup. Figure 3). In many cases the comparisons between mAb and scFv is not entirely fair since the mAb is labeled in the 488/green channel (in which brain tissue notoriously has more autofluorescence) and the scFv in the red channel e.g. NPY signal in Sup. 2b,d; PSD95 in Sup. 2f. Is the labeling really that much cleaner or is the background signal in the green channel making the mAb appear worse than it is?  
+
+<|ref|>text<|/ref|><|det|>[[114, 583, 852, 615]]<|/det|>
+- p. 9 "....found that the anti-calbindin scFv penetrated to a depth of \(\sim 150 \mu m\) in a 300-μm tissue slice". So the probe labeled throughout the entire slice?
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[68, 88, 323, 107]]<|/det|>
+## 1 REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[70, 120, 88, 133]]<|/det|>
+2  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 144, 438, 162]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 171, 876, 310]]<|/det|>
+The authors developed eight smaller single- chain variable fragments (scFvs) based on eight well- characterized mAbs and conjugated them with various fluorescent dyes. This idea is great for volume correlative light and electron microscopy. The authors reported that each scFv proved effective as a detergent- free immunofluorescent probe. The approach was believed to be promising for routine linking of molecular information to connectomic information from the same material since the quality of data from the volumetric fluorescent and electron microscopy is good. However, some critical aspects of the experimental design need to be addressed, and additional experiments are required to draw conclusive findings.  
+
+<|ref|>text<|/ref|><|det|>[[115, 345, 880, 553]]<|/det|>
+1. One concern in this study is to choose Triton X-100 as a detergent for comparison. While Triton X-100 is commonly used to aid antibody penetration in light microscopy, it is not widely used for EM or immuno-EM studies. For EM studies, saponin is one of organic solvents which dissolve lipids from cell membranes making them permeable to antibodies. Furthermore, organic solvents can be used to fix and permeabilize cells at the same time by coagulating proteins. Saponin interacts with membrane cholesterol, selectively removing it and leaving holes in the membrane. Most researchers in the EM field choose low concentration of saponin to balance the antibody penetration and excellent membrane morphology in their immuno-EM experiments. Majority of their immuno-EM images showed great morphology of membrane at the ultrastructural level with the use of saponin. In some immuno-EM studies, Triton X-100 has been used. But those studies don't care the cell membrane, most of them were interested in some subcellular organelles.  
+
+<|ref|>text<|/ref|><|det|>[[112, 563, 875, 909]]<|/det|>
+We thank the reviewer for pointing out the important fact that saponin is a far better detergent for electron microscopy ultrastructural studies than Triton X- 100. In response to this suggestion, we did a new series of experiments analyzing twenty- two tissue blocks at various saponin concentrations, sample thicknesses, and durations of antibody incubation. Moreover, directly conjugated mAbs have become available, so we moved the results related to the penetration tests with secondary antibody labeling (original Figure 1 d and e) to Supplementary Figure 3 b and c). The new results are now presented in new Figure 1 panels f- i, new Supplementary Figure 4 (plus additional new Supplementary Figure 5 that we will describe below in point 4). The key result is that saponin at \(0.05\%\) concentration does not allow fluorescently labeled monoclonal antibodies to penetrate into the middle \(500 \mu \mathrm{m}\) of a 1- mm block even after a 1- week incubation with the labeled antibody (Figure 1 g; Supplementary Figure 4). In contrast, the fluorescently labeled scFv penetrated throughout a block in the absence of detergent (Figure 1 f; Supplementary Figure 4). In different experiments, we did examine if higher concentrations of saponin could aid in deeper penetration (see new Supplementary Figure 5) but for our purposes, even saponin at \(0.05\%\) was problematic. The reason was that we found small breaks in the plasma membranes of neuronal processes (arrows, Fig 1 i) that were not present in samples not treated with detergent (Fig 1 h). While these ultrastructural breaks are small and, for many kinds of studies, would be of no consequence, for connectomics they are serious. This seriousness is related to the requirement for automatic algorithms to segment each nerve cell process. When two adjacent objects have a
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 88, 870, 230]]<|/det|>
+continuity between them, this is often interpreted by the algorithms erroneously as the same object. Such merge errors are far more difficult to find and correct than split errors, so avoiding them at all costs is necessary (Shapson- Coe et al. 2021; Januszewski et al. 2018). With newer techniques like multicolor 2- photon microscopy (Mahou et al. 2012; Blanc et al. 2023; Pudavar et al. 2024), lightsheet microscopy in uncleared tissue (Schmid et al. 2013), and confocal done with clearing approaches compatible with electron microscopy (Furuta et al. 2022), we think having fluorescent scFv penetration hundreds of microns into tissue blocks will be of great use in CLEM studies. We have modified the text to make these points clearer (line 150).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 88, 880, 744]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 752, 844, 782]]<|/det|>
+<center>Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 793, 880, 897]]<|/det|>
+a, Schematic representations of a full-length IgG antibody and an scFv probe with a conjugated fluorescent dye. b, Confocal images from the cerebral cortex of a YFP-H mouse labeled using a GFP-specific scFv probe conjugated with the red dye 5-TAMRA. Arrows show thinner neuronal processes, perhaps myelinated, that are not labeled by scFv. c, Layer ⅔ of the cerebral cortex labeled with a calbindin-specific scFv probe. d, Cerebellum cortex of Crus 1 labeled with the PSD-95 specific scFv. Right panel is the enlarged boxed inset from left. e, Cerebral cortex labeled with the NPY-specific scFv. f and g, Tissue penetration depth comparison of a parvalbumin-specific scFv without detergent and its parental
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 89, 870, 150]]<|/det|>
+mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation. h and i, Comparison of ultrastructure from samples incubated 7 days without detergent and with \(0.05\%\) saponin. Arrows indicate membrane breaks. Asterisks indicate abnormal appearing vesicle- filled axonal terminals.  
+
+<|ref|>image<|/ref|><|det|>[[241, 175, 768, 545]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 560, 681, 577]]<|/det|>
+<center>Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 588, 879, 633]]<|/det|>
+Tissue penetration depth comparison of a calbindin- specific scFv without detergent and its parental mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 660, 880, 728]]<|/det|>
+2. Another point of concern is the lack of clarity on how the antibodies penetrate the cell through the cell membrane in detergent-free immunofluorescence labeling. A more thorough investigation and explanation of this process are needed to provide a comprehensive understanding of the technique's efficacy.  
+
+<|ref|>text<|/ref|><|det|>[[115, 739, 880, 789]]<|/det|>
+We agree with the reviewer on the importance of investigating how scFvs penetrate the cell membrane in the absence of detergent. We are interested in determining the mechanism as well. We think there are at least two possible mechanisms:  
+
+<|ref|>text<|/ref|><|det|>[[115, 800, 882, 904]]<|/det|>
+First, because all the immunolabeling experiments in this study were performed on brain tissue samples from animals perfused and fixed with \(4\%\) formaldehyde (prepared fresh from paraformaldehyde) \(+0.1\%\) glutaraldehyde in PBS, the cell membrane penetration by scFvs may be simply explained by the fact that formaldehyde and glutaraldehyde permeate the lipid bilayer. formaldehyde and glutaraldehyde are commonly used as chemical fixatives by crosslinking amino groups of proteins (Fischer et al. 2008). It has been known that formaldehyde also
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 88, 870, 315]]<|/det|>
+dissolves lipids (Fox et al. 1985; Kiernan 2000; Thavarajah et al. 2012). A recent study using surface plasmon resonance (SPR) (Cheng et al. 2019) showed that fixation with formaldehyde perturbed the integrity of membranes ( \(10 \pm 5\%\) mass loss), and they showed increased permeability of sucrose. In another recent study using atomic force microscopy (Ichikawa et al. 2022), both formaldehyde and glutaraldehyde were shown to increase the size of nanoscopic protrusions on cell membranes. These protrusions were generated by membrane protein aggregates induced by crosslinking via formaldehyde or glutaraldehyde. The aggregated membrane proteins may create gaps between them and their nearby lipids providing a permeability pore. Additionally, two extracellular space- preserving fixation methods employing formaldehyde and glutaraldehyde (Fulton and Briggman 2021; Lu et al. 2023) showed that full- length antibodies can penetrate cell membranes albeit with lower diffusivity than scFvs, supporting the idea that the formaldehyde plus glutaraldehyde treated membranes do have gaps caused by the fixation.  
+
+<|ref|>text<|/ref|><|det|>[[110, 323, 877, 654]]<|/det|>
+We have tested this idea as well by using scFv immunolabeling on HEK293T cells cultured as a single layer on a petri dish with a coverglass bottom. HEK293T cells allowed us to avoid the issue of cut/fragmented cells in tissue sections where scFv could penetrate into cells via a cut surface rather than through a membrane. After transfecting the HEK293T cells with a plasmid encoding calbindin, we fixed the cells with the same fixative (4% formaldehyde + 0.1% glutaraldehyde in PBS) for 15 min and then washed with PBS. Overnight immunolabeling of the anti- calbindin scFv was then performed without or with 0.1% Triton- X. The results showed that in both conditions (without or with 0.1% Triton- X), the scFv can penetrate and label its intracellular target (we have added a new Supplementary Figure 7 a, b to show this data). This result provides evidence consistent with the idea that the cell membranes fixed with 4% formaldehyde + 0.1% glutaraldehyde allow scFvs to penetrate into intracellular spaces. Additionally, we also tested a 1- hour immunolabeling protocol using scFvs and full- size mAbs directed to transfected calbindin in COS- 1 cells both without and with detergent permeabilization. Similarly, we found that the scFvs were able to penetrate COS- 1 cells and label intracellular targets. However, the mAb was unable to penetrate at least at 1 hour (see new Supplementary Figure 8). In another experiment we did find that an overnight incubation with a mAb did label fixed cells that were not permeabilized with detergent. From all of these experiments we infer that due to their small size the scFvs are better to penetrate fixed cells than larger immunoprobes. We have modified the text to make these points clearer (line 169).  
+
+<|ref|>text<|/ref|><|det|>[[113, 660, 877, 799]]<|/det|>
+It is also possible that scFvs by virtue of their small size could permeate unfixed lipid bilayers. Indeed (Li et al. 2016) showed that anti- pTau nanobodies when injected into the blood of live mice could cross the blood- brain barrier and also cross neuronal cell membranes to label intracellular pTau. In (Bernard et al. 2016), after transgenically inducing expression of an anti- Otx2 scFv to express in cells of the choroid plexus cells, scFv in the CSF can cross the blood- brain barrier and neutralize Otx2 in the cortex, perhaps via transcytosis. In (Thiel et al. 2002), scFvs were shown to be able to pass through live cornea with an intact epithelium. (Im, Chung, and Jang 2017) showed that scFvs can enter live, unfixed culture cells.  
+
+<|ref|>text<|/ref|><|det|>[[113, 808, 879, 896]]<|/det|>
+Based on these results, we were motivated to see if the scFvs we generated could cross into living cells. We attempted to immunolabel the transfected HEK293T cell for calbindin with the anti- calbindin scFv using live HEK293T cells. We found that after a one- hour incubation, the scFv could penetrate cells (Supplementary Figure 7 c, arrow). However, unlike the penetration of fixed cells described above, the labeling was more punctate. This labeling was most likely
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 879, 228]]<|/det|>
+explained by endocytosis as has been previously seen for extracellular dye molecules (see for example, (Tsuriel et al. 2015)). We have added a Supplementary Figure 7 c to show this result. Consistent with this, it has been documented that both nanobodies and scFvs can be internalized into cells via endocytosis (de Beer and Giepmans 2020; Wittrup et al. 2009; Alric et al. 2018; Kim et al. 2020). We have modified the text to make this point clearer (line 176). This is a potentially important route of entry because it provides an option to achieve immunolabeling of larger tissue samples, such as a whole mouse brain, by introducing these small immunoprobes in live animals.  
+
+<|ref|>image<|/ref|><|det|>[[117, 264, 880, 860]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 872, 737, 889]]<|/det|>
+<center>Sup. Figure 7. Penetration of anti-calbidin scFv into fixed HEK cells or live cells. </center>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 882, 180]]<|/det|>
+Immunofluorescence immunocytochemistry on transiently transfected cells. HEK cells were transfected with a plasmid encoding Flag-tagged human calbindin. a, Chemically fixed cells were labeled overnight with Alexa594 anti-calbindin L109/57 scFv in the absence of detergent. b, Chemically fixed cells were labeled overnight with Alexa594 anti-calbindin L109/57. scFv with \(0.1\%\) Triton-X. c, Live cells were labeled with Alexa594 anti-calbindin L109/57 scFv. The arrow indicates the cell that has intracellular scFv labeling. Arrowheads indicate puncta labeling in some cells.  
+
+<|ref|>image<|/ref|><|det|>[[260, 191, 736, 556]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 575, 668, 591]]<|/det|>
+<center>Sup. Figure 8. Penetration of anti-calbindin scFv into fixed COS-1 cells. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 602, 877, 735]]<|/det|>
+Immunofluorescence immunocytochemistry on transiently transfected cells. COS- 1 cells were transfected with a plasmid encoding Flag- tagged human calbindin. Cells in panels A and B were labeled for 1 hour after fixation and prior to detergent permeabilization with (A) Alexa594 anti- calbindin L109/57 scFv or (B) anti- calbindin mouse mAb L109/39 (scFv and mAb labeling in red). After permeabilization, cells were labeled with rabbit anti- Flag (green) to detect calbindin, and Hoechst nuclear dye (blue). For cells in panels C and D all immunolabeling was performed after fixation and detergent permeabilization with (C) Alexa594 anti- calbindin L109/57 scFv or (D) anti- calbindin mouse mAb L109/39 (scFv and mAb labeling in red). Cells were simultaneously labeled with rabbit anti- Flag (green), and Hoechst (blue). Cells in all panels were imaged at the same exposure.  
+
+<|ref|>text<|/ref|><|det|>[[113, 760, 883, 846]]<|/det|>
+3. The animals were perfused with the fixative (4% paraformaldehyde + 0.1% glutaraldehyde). The low concentration of glutaraldehyde (0.1%) may not be sufficient to adequately preserve the lipids in the cell membrane. However, for detergent-free immunofluorescence labeling, scFv probes or nanobody probes were incubated for 3 days (50 μm) or 7 days (120 μm). This could potentially affect the ultrastructural morphology.  
+
+<|ref|>text<|/ref|><|det|>[[113, 857, 864, 908]]<|/det|>
+We agree this is a reasonable concern that the use of 0.1% glutaraldehyde does not sufficiently preserve lipids in the cell membrane, which may cause the ultrastructure to deteriorate when tissue samples are incubated with immuno-probes for prolonged periods like
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 881, 504]]<|/det|>
+three or seven days. We were aware of this potential problem. The reasons we used \(0.1\%\) glutaraldehyde instead of a higher concentration was: first, glutaraldehyde is a harsher fixative which may modify epitopes on target proteins (Fischer et al. 2008), preventing immuno- probe labeling; Second, glutaraldehyde has higher autofluorescence than formaldehyde (Fischer et al. 2008), which causes high background in fluorescence microscopy. Because we only used \(0.1\%\) glutaraldehyde, we always postfixed the perfused brain for many hours (overnight). To prevent reversal of the formaldehyde fixation (Fischer et al. 2008) the brain samples were then sliced in ice- cold fixative (4% formaldehyde \(+0.1\%\) glutaraldehyde) and stored in the same fixative at \(4^{\circ}C\) . The only exception to this protocol was our work with the neuropeptide NPY, which we, as others, have found to be difficult to label if the fixation is too extensive. In this case, we stored the slices in PBS at \(4^{\circ}C\) . We also performed all the incubations, including the washing steps, at \(4^{\circ}C\) to prevent ultrastructural degradation. In the manuscript, in Supplementary Figure 21, we examined the ultrastructure of a \(2\mathrm{mm}\) , \(2\mathrm{mm}\) , \(120 - \mu \mathrm{m}\) thick cerebellum tissue sample incubated with scFv probes for seven days after light fixation (described above). As shown in the figure, ultrastructure at four locations across the cerebellar cortex layers including regions that are near the center of the block, is preserved well. After careful examination of the images during the revision process, we have noticed some abnormalities. In the superficial layer (the molecular layer) of the cerebellar cortex, which is mainly composed of neuronal processes and close to the surface of the block, we did observe some artifacts (new arrows in Supplementary Figure 21). We are unsure whether these artifacts are explained by mechanical or chemical or thermal factors that are different at the surface vs. the interior of the block. We have also modified the manuscript (line 217) to make readers aware of this issue. If reviewers are interested in examining the ultrastructure directly, we encourage reviewers to visit the Neuroglancer link of our vCLEM dataset at Neuroglancer LINK.  
+
+<|ref|>text<|/ref|><|det|>[[114, 513, 880, 636]]<|/det|>
+In addition, in a new set of experiments we performed for the revision that we will discuss in detail below in point 4, we showed that instead of 3- day or 7- day incubations, the anti- calbindin scFv can penetrate to the center of a \(300 - \mu \mathrm{m}\) vibratome section with incubation of only one day. We found fewer tissue artifacts in the ultrastructure of 1- day incubated samples than the 7- day samples (see an example in new Supplementary Figure 6, arrows indicating artifacts). We have modified the text to make this point clearer (line 166). So, we conclude that, at least for some scFvs, 1- day incubations are sufficient.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 90, 880, 400]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 419, 866, 436]]<|/det|>
+<center>Sup. Figure 6. Ultrastructure comparison between samples incubated for one day or seven days. </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 446, 856, 478]]<|/det|>
+Ultrastructure of locations close to the surfaces of 300- μm cerebral cortex sections immunolabeled for one day (a) or seven days (b). Arrows indicate artifacts.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 90, 877, 565]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 578, 874, 610]]<|/det|>
+<center>Sup. Figure 21. Well-preserved ultrastructure from the surface (a) to the middle (d) of the \(120 - \mu m\) section. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 620, 860, 653]]<|/det|>
+Panel 1- 4 in a- d show the ultrastructure at the locations labeled by the red circles in the right panels. Arrows indicate the artifacts potentially caused by prolonged incubation with scFvs for immunolabeling.  
+
+<|ref|>text<|/ref|><|det|>[[115, 675, 875, 762]]<|/det|>
+4. This study lack of novelty. The use of scFv has been well-established over the years, making it a solid foundation for the volume CLEM study. However, the choice of detergent-free immunofluorescence labeling raises questions about its suitability. Considering alternative approaches, such as utilizing low concentrations of saponin, might offer a more effective option for preserving membrane morphology during immuno-EM experiments.  
+
+<|ref|>text<|/ref|><|det|>[[115, 772, 870, 860]]<|/det|>
+We agree with the reviewer that the use of scFvs is well- established (Bird et al. 1988; Huston et al. 1988; Monnier, Vigouroux, and Tassew 2013; Ahmad et al. 2012). The use of scFvs as immuno- probes for CLEM has been raised in a number of papers in discussion (de Beer and Giepmans 2020; Franek et al. 2024) but, to the best of our knowledge, this is the first actual demonstration of scFvs in volumetric CLEM.  
+
+<|ref|>text<|/ref|><|det|>[[115, 870, 875, 904]]<|/det|>
+We agree with the reviewer that when performing volumetric CLEM, alternative immunolabeling approaches other than those that employ scFvs should be considered, such as
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 879, 452]]<|/det|>
+fluorescently tagged primary IgG antibodies with saponin permeabilization. Therefore, in new experiments we compared detergent- free immunolabeling with scFv and Triton- X or saponin- enabled immunolabeling with a dye- directly conjugated monoclonal antibody (mAb) at various detergent concentrations (0.1% and 0.3% Triton- X; 0.05%, 0.1%, and 0.2% saponin) and with two different incubation times (1 day and 7 days). The experiments were performed on 300- μm cerebral cortex tissue blocks with an anti- calbindin L109/57 scFv and a dye- directly conjugated anti- calbindin L109/57 mAb. The epitope binding site of the mAb and the scFv were the same. As shown in new Supplementary Figure 5 a, after 1 day of incubation, only the scFv and the mAb with 0.3% Triton- X penetrated to the middle (i.e., 150 μm) of the section. The other experimental conditions showed various degrees of penetration: 0.1% Triton- X, \(\sim 50\) μm; 0.05% saponin, \(\sim 30\) μm; 0.1% saponin, \(\sim 80\) μm; 0.2% saponin, \(\sim 100\) μm. In all cases with saponin permeabilization, there was a lack of labeling in the cell nuclei (indicated by arrows). When we examined the ultrastructure of these labeled samples, the samples treated with detergent- free scFv labeling showed the best quality. The sample treated with 0.05% saponin showed good- quality EM ultrastructure. All the other samples showed compromised EM ultrastructure, the severity of which increased with the increase of detergent concentration. The membrane breaks in these samples would make automatic segmentation for connectomics challenging, as stated above in our answer to point 1. Although the sample treated with 0.05% saponin for one day showed no obvious ultrastructural artifacts, the mAb penetration was far shallower than the scFv ( \(\sim 30\) μm vs. 150 μm) making volumetric CLEM on the samples larger than the penetration depth difficult.  
+
+<|ref|>text<|/ref|><|det|>[[113, 461, 880, 702]]<|/det|>
+As shown in new Supplementary Figure 5 b, after 7 days of incubation, scFvs without detergent and mAb with various concentrations of Triton- X or saponin can penetrate the middle of the \(300 - \mu m\) . However, we still observed in the case of 0.05% saponin a lack of labeling in the cell nuclei (indicated by arrows). Again, when examining the ultrastructure of these labeled samples, the sample treated with detergent- free scFv labeling showed the best quality and is similar to the one- day sample (which we have also mentioned in our answer to reviewer's point 3). All the other samples showed compromised EM ultrastructure, which was much worse when compared with the 1- day samples. Even the 0.05% saponin now showed membrane breaks. We also noticed after 7- day saponin incubation a new artifact: the vesicle- filled axonal profiles in samples treated with saponin for seven days showed a granular texture (indicated by arrowheads in Supplementary Figure 5 b). We think the protein- coagulating function of saponin, as the reviewer stated previously, may be the cause. These granules could pose challenges when synaptic vesicles need to be automatically detected and analyzed (as we did in the later part of this paper) for connectomic studies.  
+
+<|ref|>text<|/ref|><|det|>[[113, 712, 872, 903]]<|/det|>
+In addition, as we have mentioned in our answer to point 1, scFvs can penetrate 1- mm tissue blocks while saponin at 0.05% concentration only allows mAbs to penetrate into \(250 \mu m\) after a seven- day incubation (Figure 1 g; new Supplementary Figure 4). These new results suggest that if a researcher wants to do a small- scale volumetric CLEM on a smaller tissue sample (such as several \(\mu m\) to \(50 - \mu m\) ), directly dye- conjugated primary antibodies with a low concentration (0.05%) of saponin with a shorter incubation (one day) may be an option. However, should a researcher need to conduct large- scale volumetric CLEM on larger tissue samples ( \(\sim 1\) mm in thickness), using scFvs for detergent- free immunolabeling is more advantageous. Large- scale volumetric CLEM is especially important for connectomics because a smaller volume is very likely to have fragmented cells/processes that prevent the mapping of the neural circuits. We have modified the text to make these points clearer (line 150; line 162).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[118, 120, 877, 750]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 758, 844, 789]]<|/det|>
+<center>Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents. </center>  
+
+<|ref|>text<|/ref|><|det|>[[114, 800, 870, 875]]<|/det|>
+300- \(\mu \mathrm{m}\) cerebral cortex sections were immunolabeled for one day (a) or seven days (b) with a calbindin- specific scFv conjugated with 5- TAMRA in the absence of detergent or with the scFv's parental mAb conjugated with FL550 in the presence of \(0.1\%\) , \(0.3\%\) Triton- X, or \(0.05\%\) , \(0.1\%\) , \(0.2\%\) saponin. Arrows indicate unlabeled cell nuclei. Arrowheads indicate granular textures associated with the treatment of saponin.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 876, 142]]<|/det|>
+5. The choice of \(0.3\%\) Triton X-100 to demonstrate detergent issues in EM seems excessive. Most EM studies recommend not exceeding \(0.1\%\) Triton X-100, making it unnecessary to use a higher concentration for this purpose.  
+
+<|ref|>text<|/ref|><|det|>[[114, 152, 866, 204]]<|/det|>
+We agree with the reviewer that choosing \(0.3\%\) Triton X- 100 is excessive to demonstrate the detergent's issue on EM ultrastructure. We have changed Figure 1 h and i so the comparison is with a sample treated with \(0.05\%\) saponin.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 438, 107]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 117, 879, 427]]<|/det|>
+This is a technology development manuscript describing the generation and implementation of an assortment of single chain antibody- based probes (scFvs) against different brain proteins to label tissue for correlative fluorescence/volumetric electron microscopy. The key advance is the nature of the labeling probes, which can diffuse deep into fixed brain tissue and penetrate cells without the need for detergent permeabilization. The authors do nice side- by- side comparisons with whole IgG antibodies to highlight this. Detergent- free labeling thus allows processing for ultrastructural analysis by EM, with excellent membrane preservation. Furthermore, the scFv labeling reagents can be easily labeled with different fluorescent dyes allowing the authors to visualize numerous (here they show 6) different labels in the same sample using spectral unmixing confocal microscopy. Serial section EM images of the same samples were then reconstructed and aligned with the fluorescence images to achieve correlated fluorescence/ultrastructure. Overall the data were compelling, with many beautiful examples of correlated fluorescence localization with 3D ultrastructure, nicely demonstrating the power of the technique. While the manuscript primarily focuses on tool development and offers little in the way of novel biological insight, I feel the potential future impact of the technique (i.e. ability to assign neural identities to volumetric connectomics EM datasets, ultrastructural localization of channels, receptors, etc.) will have broad appeal. There are several specific points that deserve attention:  
+
+<|ref|>text<|/ref|><|det|>[[115, 465, 872, 585]]<|/det|>
+- Nowhere in the manuscript do the authors validate their labeling reagents in a knockout background. In many cases the labeling pattern is distinct, consistent with previously published work and the localization of the signal makes sense with the correlated ultrastructure (i.e. vGLUT labels presynaptic terminals), but in some cases it is more ambiguous. For example, in Fig. S2a,e the authors argue that the CB and PV scFvs label more of the target proteins in the cell nucleus, is this real signal or are these probes picking up something non-specific in the nucleus that the IgG does not?  
+
+<|ref|>text<|/ref|><|det|>[[113, 595, 879, 905]]<|/det|>
+Concerning validation, we agree with the reviewer that the most crucial concern for immuno- probes or any similar affinity probes is whether they label or detect the actual target they are supposed to bind to. There are many cases when antibodies working in ELISA or Western blot settings fail to label their targets or have off- target labeling that creates abnormal background signals in immunohistochemistry (IHC). The parental (aka. progenitor) monoclonal antibodies (mAbs) from the UC Davis/NIH NeuroMab facility, whose sequences were used to generate the eight scFvs in this study, have undergone a strict validation process. In all but one case (the anti- NPY mAb), the mAbs have passed by at least three of the following: immunofluorescence on transfected COS- 1 cells, Western blots on homogenized rat and mouse brains, IHC on rat and mouse brain sections, and IHC on mouse sections in a knockout background. These were accomplished in co- author James Trimmer's lab (for more details, see (Gong, Murray, and Trimmer 2016)). The validation tests of the eight mAbs used in the paper are now shown in new Supplementary Table 4). Although limited by the availability of KO brain samples, the three that we were able to test of them (N206Bb/9, GFAP R416WT; K28/43, PSD- 95; K14/16, Kv 1.2) have passed the test of IHC on WT versus KO mouse brain sections, in that all detectable labeling observed in WT sections was eliminated in KO sections (all three also passed on WT/KO comparison by immunoblot) in a knockout background. While we want to test all the mAbs in a knockout background, but we hope the reviewer understands that it is
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 92, 839, 123]]<|/det|>
+challenging to gather KO animals brain samples for all seven endogenous targets because some may be lethal mutations. 
+
+<|ref|>text<|/ref|><|det|>[[115, 135, 872, 256]]<|/det|>
+We also validated the scFvs in each case via IHC on rat and mouse brain sections and by immunofluorescence immunocytochemistry on transiently transfected COS-1 cells (also summarized in Supplementary Table 4; we also provide representative images in new Supplementary Figure 9, 10, 11 for the validation of the N206b/9, anti-GFAP R416WT scFv. Details of the methods of the validation tests for the other scFvs in this paper (and other scFvs) can be found in (Mitchell et al. 2023; Gong, Murray, and Trimmer 2016). We have modified the text to make these points clearer (line 139; line 143). 
+
+<|ref|>table_caption<|/ref|><|det|>[[184, 293, 812, 308]]<|/det|>
+Sup. Table 4. Information on the validation of the scFvs and their parental mAbs. 
+
+<|ref|>table<|/ref|><|det|>[[115, 319, 883, 530]]<|/det|>
+<table><tr><td>Target</td><td>Clone No.</td><td colspan="4">mAb validation</td><td>scFv validation</td></tr><tr><td></td><td></td><td>COS-IF</td><td>Brain IB</td><td>Brain IHC</td><td>KO Brain IHC</td><td>Method</td></tr><tr><td>GFP</td><td>N86/38</td><td>Pass</td><td>NA</td><td>NA</td><td>NA</td><td>COS-IF</td></tr><tr><td>Calbindin</td><td>L109/57</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>GFAP</td><td>N206B/9</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>R416WT</td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td>VGlut1</td><td>N28/9</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>PSD-95</td><td>K28/43</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>Kv 1.2</td><td>K14/16</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Pass</td><td>Brain IHC and COS-IF</td></tr><tr><td>Parvalbumin</td><td>L114/81</td><td>Pass</td><td>Pass</td><td>Pass</td><td>ND</td><td>Brain IHC and COS-IF</td></tr><tr><td>NPY</td><td>L115/13</td><td>Pass</td><td>Fail</td><td>Pass</td><td>ND</td><td>COS-IF</td></tr></table>
+
+<|ref|>image<|/ref|><|det|>[[115, 559, 875, 840]]<|/det|>
+ 
+
+<|ref|>image_caption<|/ref|><|det|>[[115, 846, 844, 861]]<|/det|>
+<center>Sup. Figure 9. Validation of anti-GFAP scFv with Immunofluorescence immunocytochemistry.</center> 
+
+<|ref|>text<|/ref|><|det|>[[115, 873, 855, 902]]<|/det|>
+Immunofluorescence immunocytochemistry on transiently transfected cells. COS-1 cells (top row) and HEK293T cells (bottom row). Cells were transfected with a plasmid mEmerald-tagged human GFAP
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 852, 121]]<|/det|>
+(green) and double immunolabeled with the 5-TAMRA-labeled anti-GFAP N206B/9 scFv (red) and the anti-GFAP N206A/8 mouse IgG1 mAb (white). Hoechst nuclear labeling is shown in blue.  
+
+<|ref|>image<|/ref|><|det|>[[253, 148, 741, 526]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 530, 855, 562]]<|/det|>
+<center>Sup. Figure 10. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (cerebellum). </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 572, 872, 688]]<|/det|>
+GFAP scFv and original monoclonal antibody from which it was derived display the same tissue labeling pattern of a sagittal section through the rat cerebellum. A) Glial cells throughout the cerebellar granule cell layer (GCL) and prominent Bergmann glial process in molecular layer (ML) are labeled with hybridoma derived monoclonal antibody N206B/9. B) merged image includes labeling with a polyclonal rabbit antibody (KC) against the neuronal potassium channel Kv2.1, monoclonal antibody targeting glial specific RNA binding protein QKI (N147/6) and nuclear specific Hoechst labeling. C) An adjacent section labeled with N206B/9 derived scFv shows the same pattern of labeling. D) merged image with the same additional labeling as B.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[252, 92, 742, 468]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 485, 855, 515]]<|/det|>
+<center>Sup. Figure 11. Validation of anti-GFAP scFv with Immunofluorescence immunohistochemistry (hippocampus). </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 525, 874, 630]]<|/det|>
+Validation of scFv labeling pattern against hybridoma- generated monoclonal antibody N206B/9 from which it was derived. Multiplex immunofluorescent labeling of a sagittal section through rat hippocampal region CA1. A) 5- TAMRA conjugated N206B/9 derived scFv, B) Hybridoma derived monoclonal antibody N206B/9 indirectly labeled with Alexa fluor 647 conjugated goat anti- mouse IgG1 secondary antibody, C) merged images from A and B illustrating co- labeled astroglial cells (e.g arrowheads). D) Same multiplex image shown in C with additional neuronal specific potassium channel Kv2.1, glial specific pan- QKI RNA binding protein, and DNA marker Hoechst 33342 labeling.  
+
+<|ref|>text<|/ref|><|det|>[[112, 667, 879, 875]]<|/det|>
+The second issue raised is concern over the ambiguous signals detected by the anti- calbindin and anti- parvalbumin scFvs in the cell nuclei of Purkinje cells. Clearly, in the original Supplementary Figure 2, the mAbs for calbindin and parvalbumin did not detect signals in the cell nuclei as the scFvs did. In the revised paper we now include the validation IHC images of the anti- calbindin and anti- parvalbumin mAbs from experiments done in James Trimmer's lab (new Supplementary Figure 14). These images showed the normal expected labeling, that included signal in some cell nuclei of Purkinje cells in the lateral hemisphere of the rat cerebellum. In the original manuscript, we mentioned previous studies reporting the detection of calbindin and parvalbumin in cell nuclei of Purkinje cells (Celio 1990; Brandenburg et al. 2021; German et al. 1997; Schmidt et al. 2007). We believe calbindin and parvalbumin are present in the cell nuclei. We therefore suspect that the reason our mAb labeling didn't detect signals in the cell nuclei is insufficient antibody incubation time.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 878, 504]]<|/det|>
+To address this problem, in Triton- X- treated samples, we first did immunolabeling with the same anti- calbindin and anti- parvalbumin mAbs directly conjugated with the red fluorophore FL- 550 in distinction to using secondary antibodies as we did previously. The directly conjugated mAbs exclude the requirement for secondary antibodies allowing for only one round of incubation. We extended the incubation time to seven days (versus 2 days of incubation with primary antibodies previously). The results showed that both mAbs can detect signals in the cell nuclei of most ( \(> - 90\%\) ) Purkinje cells (new Supplementary Figure 15 c, I). This staining is similar to the scFv labeling except that scFvs were detected in all Purkinje cells. The second approach we attempted was immunolabeling with commercial polyclonal antibodies (pAbs) against calbindin and parvalbumin. These antibodies are not directly conjugated with fluorophores, so we utilized (Fab)2 as fluorescently tagged secondaries, which are smaller than conventional secondaries and supposedly can diffuse in tissue better. Our results showed that, again, both calbindin and parvalbumin could be detected by the pAbs in the cell nuclei in most but not all Purkinje cells (new Supplementary Figure 15 d and m). The third approach we used, was to slice the section in a different orientation to cut through the nuclei of most Purkinje cells in order to gain direct access to the nuclei in the vibratome section (see new Supplementary Figure 15 e). We immunolabeled with the same anti- calbindin or anti- parvalbumin mAbs directly conjugated with the fluorophore FL- 550 with a seven- day incubation. This time, we observed labeling in nearly all Purkinje cells (Supplementary Figure 15 f and n). These results indicate that the signals detected in the cell nuclei by scFvs are real signals. The reason it is relatively harder for full- length antibodies (mAbs or pAb) to detect these signals can be attributed to their relatively weaker penetration ability, even in the presence of Triton- X. This can be addressed in ways like extending incubation time or slicing sections to expose the internal epitopes in cell nuclei better. We have modified the text to make these points clearer (line 148).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[252, 92, 744, 468]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 485, 872, 515]]<|/det|>
+<center>Sup. Figure 14. Validation of anti-calbindin and anti-parvalbumin scFvs with Immunofluorescence immunohistochemistry. </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 526, 866, 585]]<|/det|>
+Labeling pattern of original mAbs used to generate scFvs against Parvalbumin and Calbindin in rat cerebellum. Sagittal section through cerebellum labeled with monoclonal antibodies L127/8 (A, GAD1), L114/8 R (B, PARV), and L109/57 (C, CALB1). The merged image (D) shows the colocalized pattern of labeling within Purkinje cell layer (PCL). ML, molecular layer, GCL, granule cell layer.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[113, 87, 880, 920]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 866, 123]]<|/det|>
+Sup. Figure 15. Validation of immunofluorescence by scFv probes and their parental mAbs (part 3).  
+
+<|ref|>text<|/ref|><|det|>[[113, 131, 876, 264]]<|/det|>
+Cerebellum Crus 1 sections were immunolabeled with a calbindin- specific scFv (a), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (b), the mAb conjugated with FL550 (c), or a commercial calbindin- specific pAb and secondary (Fab)2 conjugated with Alexa Fluor 594 (d). e, Schematics showing the cutting orientation that is parallel to the lobule of Crus 1, which intersects perpendicular to the planer Purkinje cells in Crus 1. f, Sections cut in this orientation immunolabeled with the mAb conjugated with FL550. The boxed inset is shown enlarged in the adjacent panel. Whole- section images of cerebellum Crus 1 sections immunolabeled with a calbindin- specific scFv (g), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (h), or the mAb conjugated with FL550 (i). Arrows indicate labeled cell nuclei of Purkinje cells. Arrowheads indicate the labeled axons.  
+
+<|ref|>text<|/ref|><|det|>[[113, 275, 879, 395]]<|/det|>
+Cerebellum Crus 1 sections were immunolabeled with a parvalbumin- specific scFv (j), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (k), the mAb conjugated with FL550 (l), or a commercial parvalbumin- specific pAb and secondary (Fab)2 conjugated with Alexa Fluor 594 (m). f, Sections cut in this orientation in a immunolabeled with the mAb conjugated with FL550. The boxed inset is shown enlarged in the adjacent panel. Whole- section images of cerebellum Crus 1 sections immunolabeled with a parvalbumin- specific scFv (o), or its parental mAb and secondary antibody conjugated with Alexa Fluor 488 (p), or the mAb conjugated with FL550 (q). Arrows indicate labeled cell nuclei of Purkinje cells. Arrowheads indicate the labeled axons.  
+
+<|ref|>text<|/ref|><|det|>[[113, 418, 869, 488]]<|/det|>
+- While many of the images are quite striking, overall the manuscript lacked any sort of quantitative analysis. Just as one example, in Fig. 1, showing a simple pixel correlation scatter plot comparing the YFP and GFP-scFv signal would give readers a better idea of how evenly the scFv is penetrating cells to label YFP.  
+
+<|ref|>text<|/ref|><|det|>[[112, 497, 880, 775]]<|/det|>
+We thank the reviewer for highlighting the lack of quantitative analysis in comparing the specificity of scFvs and mAbs. The paper does contain other quantitative analyses (see Figure 7; Supplementary Figure 29; Supplementary Table 5, 6) but in response to the specific question raised, we have now created pixel correlation scatter plots for three images from two cortical and one hippocampal section from YFP-H mice, which were also immunolabeled with the anti- GFP scFv (new Supplementary Figure 2). Supplementary Figure 2 is the raw image of Figure 1 b. As is shown in all three pixel correlating scatter plots, the signals from the scFv labeling (red) correlate with the native YFP fluorescence signal (green). There are pixels that only have values in the green channels, which correspond to the insufficiently labeled axons pointed out in Figure 2 a. There are very few pixels that only have values in the red (scFv) channel, which indicates that there is minimal off-target labeling of this anti-GFP scFv. This analysis gives us confidence in the specify of the scFv for green fluorescent protein. Doing this kind of double labeling is more problematic when comparing scFvs to mAbs that have the identical paratope as they compete for the same site. So, in these cases, as described in detail above, we had to be content with the comparative labeling of different tissue sections. We have modified the text to make this point clearer (line 128).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[120, 90, 875, 700]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 707, 850, 737]]<|/det|>
+<center>Sup. Figure 2. Pixel correlation scatter plots comparing the native YFP fluorescence signal and the red fluorescence from the labeling of the GFP-specific scFv. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 747, 866, 794]]<|/det|>
+Cerebral cortex samples (a and b) and hippocampus (c) from YFP- H mice were immunolabeled with a GFP- specific scFv conjugated with the red fluorophore 5- TAMRA. The images are raw data without any brightness/contrast adjustment. a is the raw image data of Figure 1 b.  
+
+<|ref|>text<|/ref|><|det|>[[115, 819, 874, 904]]<|/det|>
+- p. 9 "...immunofluorescence patterns that were similar to or in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex (Figure 2 a; Sup. Figure 2; Sup. Figure 3). In many cases the comparisons between mAb and scFv is not entirely fair since the mAb is labeled in the 488/green channel (in which brain tissue notoriously has more autofluorescence) and the scFv in the red channel e.g. NPY signal in Sup. 2b,d; PSD95 in Sup. 2f. Is the labeling
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 833, 124]]<|/det|>
+really that much cleaner or is the background signal in the green channel making the mAb appear worse than it is?  
+
+<|ref|>text<|/ref|><|det|>[[112, 134, 880, 464]]<|/det|>
+We apologize that the phrasing in our original manuscript may have caused a misunderstanding. In this sentence, "in some cases stronger than their parental mAbs in Crus 1 of the cerebellar cortex." only refers to the cases of calbindin and parvalbumin, as discussed in the previous point. In the other cases (GFAP, VGlut1, Kv 1.2, PSD- 95, and NPY), we believe our results suggest that the labeling of scFvs and mAbs are comparably good in terms of both signal and background. We understand the legitimate concern that the tissue sections fixed with formaldehyde and glutaraldehyde may have a higher background in the 488/green channel. Glutaraldehyde especially has stronger autofluorescence (Fischer et al. 2008). However, we only used \(0.1\%\) glutaraldehyde in our preparation prior to osmium staining. After adequate washing with PBS, the brain tissue sections do not show strong autofluorescence in the 488/green channel (as now shown in new Supplementary Figure 12 a). There is some autofluorescence, mostly from lipofuscin granules in cell bodies (arrows in new Supplementary Figure 12 a) with broad excitation/emission spectra (Di Guardo 2015; Marmorstein et al. 2002). But this autofluorescence is found in all channels. We emphasize that we are not attempting to make a case that scFv labeling is cleaner than that obtained with mAb, as the results from both are very similar. Indeed we also performed new experiments with red fluorophore- conjugated mAbs for calbindin and parvalbumin, as discussed in our answer to the reviewer's point 1, and didn't find any difference in the background level (new Supplementary Figure 15). We have adjusted the phrasing in the manuscript (line 145) to avoid any further misunderstanding.  
+
+<|ref|>image<|/ref|><|det|>[[116, 500, 879, 799]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 810, 864, 842]]<|/det|>
+<center>Sup. Figure 12. Validation of immunofluorescence by scFv probes and their parental mAbs (part 1). </center>  
+
+<|ref|>text<|/ref|><|det|>[[113, 852, 875, 899]]<|/det|>
+a, Confocal images from unlabeled the cerebral cortex and cerebellar cortex of a wild-type mouse showed limited background in the 488/green channel. Arrows indicate background signals lipofuscin granule. b- d, Cerebellum Crus 1 sections were immunolabeled with scFvs targeting VGlut1, GFAP, and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 825, 108]]<|/det|>
+Kv 1.2; or these scFvs' parental mAbs and secondary antibodies conjugated with Alexa Fluor 488.  
+
+<|ref|>text<|/ref|><|det|>[[113, 145, 881, 181]]<|/det|>
+-p. 9 "......found that the anti-calbindin scFv penetrated to a depth of \(\sim 150 \mu m\) in a 300- \(\mu m\) tissue slice". So the probe labeled throughout the entire slice?  
+
+<|ref|>text<|/ref|><|det|>[[113, 191, 880, 365]]<|/det|>
+We apologize for the lack of clarity, yes, we meant that they labeled through the entire 300 \(\mu m\) slice (150 \(\mu m\) from each side). Given the recent availability of directly conjugated mAbs we have removed Figure 1 d and e to Supplementary Figure 3 b (we remade figures from raw images showing the penetration across the 300 \(\mu m\) thickness) and c, and added new Figure 1 f- i, Supplementary Figure 4 and 5 of the results of a comparable experiment of scFv labeling on 300- \(\mu m\) and 1- mm thick brain tissue sections in the absence of detergent. In Figure 1 and Supplementary Figure 4, scFvs are shown to label throughout a 1- mm thickness with a seven- day incubation; in Supplementary Figure 5, scFvs can label a 300- \(\mu m\) thickness sample with either a 1- or 3- day incubation. EM ultrastructure of all these samples was good. We have modified the text to make these points clearer (line 150).
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[115, 88, 880, 744]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 752, 844, 782]]<|/det|>
+<center>Figure 1. Fluorescent scFv probes label brain tissues without detergents to preserve electron microscopy ultrastructure. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 793, 880, 898]]<|/det|>
+a, Schematic representations of a full-length IgG antibody and an scFv probe with a conjugated fluorescent dye. b, Confocal images from the cerebral cortex of a YFP-H mouse labeled using a GFP-specific scFv probe conjugated with the red dye 5-TAMRA. Arrows show thinner neuronal processes, perhaps myelinated, that are not labeled by scFv. c, Layer ⅔ of the cerebral cortex labeled with a calbindin-specific scFv probe. d, Cerebellum cortex of Crus 1 labeled with the PSD-95 specific scFv. Right panel is the enlarged boxed inset from left. e, Cerebral cortex labeled with the NPY-specific scFv. f and g, Tissue penetration depth comparison of a parvalbumin-specific scFv without detergent and its parental
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 868, 151]]<|/det|>
+mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1-mm cerebral cortex tissue sections with a 7-day incubation. h and i, Comparison of ultrastructure from samples incubated 7 days without detergent and with \(0.05\%\) saponin. Arrows indicate membrane breaks. Asterisks indicate abnormal appearing vesicle-filled axonal terminals.  
+
+<|ref|>image<|/ref|><|det|>[[116, 160, 872, 840]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[112, 840, 857, 856]]<|/det|>
+<center>Sup. Figure 3. Immunolabeling results of anti-calbindin scFv and tissue penetration comparison. </center>  
+
+<|ref|>text<|/ref|><|det|>[[112, 866, 864, 898]]<|/det|>
+a, Additional brain regions labeled with a calbindin-specific scFv probe conjugated with 5-TAMRA. The arrow in the left panel shows myelinated Purkinje cell axons in the granule layer. b, Tissue penetration
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 90, 875, 150]]<|/det|>
+depth comparison of scFvs, mAbs (plus secondary antibodies), and the role of detergents in improving the depth of labeling. c, Comparison of ultrastructure with and without \(0.5\%\) Triton X- 100 on scFv labeled samples. Boxed insets are shown at higher magnification in adjacent panels. of \(\sim 30 \mu m\) ; the nanobody can penetrate into a depth of \(\sim 150 \mu m\) .  
+
+<|ref|>image<|/ref|><|det|>[[240, 163, 767, 530]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 550, 680, 566]]<|/det|>
+<center>Sup. Figure 4. Penetration of anti-calbidin scFv into 1-mm tissue sample. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 577, 878, 621]]<|/det|>
+Tissue penetration depth comparison of a calbindin- specific scFv without detergent and its parental mAbs directly conjugated with fluorophores with \(0.05\%\) saponin on 1- mm cerebral cortex tissue sections with a 7- day incubation.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[117, 90, 877, 725]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[113, 731, 844, 763]]<|/det|>
+<center>Sup. Figure 5. Tissue penetration depth comparison of scFvs in the absence of detergent and fluorophore-conjugated mAbs with the treatments of various concentrations of detergents. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 772, 870, 848]]<|/det|>
+300- \(\mu \mathrm{m}\) cerebral cortex sections were immunolabeled for one day (a) or seven days (b) with a calbindin- specific scFv conjugated with 5- TAMRA in the absence of detergent or with the scFv's parental mAb conjugated with FL550 in the presence of \(0.1\%\) , \(0.3\%\) Triton- X, or \(0.05\%\) , \(0.1\%\) , \(0.2\%\) saponin. Arrows indicate unlabeled cell nuclei. Arrowheads indicate granular textures associated with the treatment of saponin.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 117, 875, 905]]<|/det|>
+597 Ahmad, Zuhaila Asra, Swee Keong Yeap, Abdul Manaf Ali, Wan Yong Ho, Noorjahan Banu 598 Mohamed Alitheen, and Muhajir Hamid. 2012. "scFv Antibody: Principles and Clinical 599 Application." Clinical & Developmental Immunology 2012 (March): 980250. 600 Alric, Christophe, Katel Hervé-Aubert, Nicolas Aubrey, Souad Melouk, Laurie Lajoie, William 601 Même, Sandra Même, et al. 2018. "Targeting HER2-Breast Tumors with scFv-Decorated 602 Bimodal Nanoprobes." Journal of Nanobiotechnology 16 (1): 18. 603 Beer, Marit A. de, and Ben N. G. Giepmans. 2020. "Nanobody-Based Probes for Subcellular 604 Protein Identification and Visualization." Frontiers in Cellular Neuroscience 14 (November): 573278. 605 Bernard, Clémence, Clémentine Vincent, Damien Testa, Eva Bertini, Jérôme Ribot, Ariel A. Di 606 Nardo, Michel Volovitch, and Alain Prochiantz. 2016. "A Mouse Model for Conditional 607 Secretion of Specific Single-Chain Antibodies Provides Genetic Evidence for Regulation of 608 Cortical Plasticity by a Non-Cell Autonomous Homeoprotein Transcription Factor." PLoS 609 Genetics 12 (5): e1006035. 610 Bird, R. E., K. D. Hardman, J. W. Jacobson, S. Johnson, B. M. Kaufman, S. M. Lee, T. Lee, S. 611 H. Pope, G. S. Riordan, and M. Whitlow. 1988. "Single-Chain Antigen-Binding Proteins." 612 Science 242 (4877): 423-26. 613 Blanc, Hugo, Gabriel Kaddour, Nicolas B. David, Willy Supatto, Jean Livet, Emmanuel 614 Beaurepaire, and Pierre Mahou. 2023. "Chromatically Corrected Multicolor Multiphoton 615 Microscopy." ACS Photonics 10 (12): 4104-11. 616 Brandenburg, Cheryl, Lindsey A. Smith, Michaela B. C. Kilander, Morgan S. Bridi, Yu-Chih Lin, 617 Shiyong Huang, and Gene J. Blatt. 2021. "Parvalbumin Subtypes of Cerebellar Purkinje 618 Cells Contribute to Differential Intrinsic Firing Properties." Molecular and Cellular 619 Neurosciences 115 (September): 103650. 620 Celio, M. R. 1990. "Calbindin D-28k and Parvalbumin in the Rat Nervous System." 621 Neuroscience 35 (2): 375-475. 622 Cheng, Ruoyu, Feng Zhang, Meng Li, Xiang Wo, Yu-Wen Su, and Wei Wang. 2019. "Influence 623 of Fixation and Permeabilization on the Mass Density of Single Cells: A Surface Plasmon 624 Resonance Imaging Study." Frontiers in Chemistry 7 (August): 588. 625 Di Guardo, G. 2015. "Lipofuscin, Lipofuscin-like Pigments and Autofluorescence." European 626 Journal of Histochemistry: EJH 59 (1): 2485. 627 Fischer, Andrew H., Kenneth A. Jacobson, Jack Rose, and Rolf Zeller. 2008. "Fixation and 628 Permeabilization of Cells and Tissues." CSH Protocols 2008 (May): db.top36. 629 Fox, C. H., F. B. Johnson, J. Whiting, and P. P. Roller. 1985. "Formaldehyde Fixation." The 630 Journal of Histochemistry and Cytochemistry: Official Journal of the Histochemistry Society 631 33 (8): 845-53. 632 Franek, Michal, Lenka Koptasikova, Jiri Miksátko, David Liebl, Eliška Macíčková, Jakub 633 Pospíšil, Milan Esner, Martina Dvořáčková, and Jiří Fajkus. 2024. "In-Section Click-iT 634 Detection and Super-Resolution CLEM Analysis of Nucleolar Ultrastructure and Replication 635 in Plants." Nature Communications 15 (1): 2445. 636 Fulton, Kara A., and Kevin L. Briggman. 2021. "Permeabilization-Free En Bloc 637 Immunohistochemistry for Correlative Microscopy." eLife 10 (May). 638 https://doi.org/10.7554/eLife.63392. 639 Furuta, Takahiro, Kenta Yamauchi, Shinichiro Okamoto, Megumu Takahashi, Soichiro Kakuta, 640 Yoko Ishida, Aya Takenaka, et al. 2022. "Multi-Scale Light Microscopy/electron Microscopy 641 Neuronal Imaging from Brain to Synapse with a Tissue Clearing Method, ScaleSF." 642 iScience 25 (1): 103601. 643 German, D. C., M. C. Ng, C. L. Liang, A. McMahon, and A. M. Iacopino. 1997. "Calbindin-D28k 644 in Nerve Cell Nuclei." Neuroscience 81 (3): 735-43.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[58, 90, 886, 905]]<|/det|>
+646 Gong, Belvin, Karl D. Murray, and James S. Trimmer. 2016. “Developing High- Quality Mouse Monoclonal Antibodies for Neuroscience Research - Approaches, Perspectives and Opportunities.” New Biotechnology 33 (5 Pt A): 551–64. 647 Huston, J. S., D. Levinson, M. Mudgett- Hunter, M. S. Tai, J. Novotný, M. N. Margolies, R. J. Ridge, R. E. Bruccoleri, E. Haber, and R. Crea. 1988. “Protein Engineering of Antibody Binding Sites: Recovery of Specific Activity in an Anti- Digoxin Single- Chain Fv Analogue Produced in Escherichia Coli.” Proceedings of the National Academy of Sciences of the United States of America 85 (16): 5879–83. 654 Ichikawa, Takehiko, Dong Wang, Keisuke Miyazawa, Kazuki Miyata, Masanobu Oshima, and Takeshi Fukuma. 2022. “Chemical Fixation Creates Nanoscale Clusters on the Cell Surface by Aggregating Membrane Proteins.” Communications Biology 5 (1): 487. 657 Im, Sun- Woo, Hee Yong Chung, and Young- Ju Jang. 2017. “Development of Single- Chain Fv of Antibody to DNA as Intracellular Delivery Vehicle.” Animal Cells and Systems 21 (6): 382–87. 660 Januszewski, Michał, Jörgen Kornfeld, Peter H. Li, Art Pope, Tim Blakely, Larry Lindsey, Jeremy Maitin- Shepard, Mike Tyka, Winfried Denk, and Viren Jain. 2018. “High- Precision Automated Reconstruction of Neurons with Flood- Filling Networks.” Nature Methods 15 (8): 605–10. 664 Kiernan, John A. 2000. “Formaldehyde, Formalin, Paraformaldehyde And Glutaraldehyde: What They Are And What They Do.” Microscopy Today 8 (1): 8–13. 666 Kim, Eunhee G., Jieun Jeong, Junghyeon Lee, Hyeryeon Jung, Minho Kim, Yi Zhao, Eugene C. Yi, and Kristine M. Kim. 2020. “Rapid Evaluation of Antibody Fragment Endocytosis for Antibody Fragment- Drug Conjugates.” Biomolecules 10 (6). 669 https://doi.org/10.3390/biom10060955. 670 Li, Tengfei, Matthias Vandesquille, Fani Koukouli, Clémence Dudeffant, Ihsen Youssef, Pascal Lenormand, Christelle Ganneau, et al. 2016. “Camelid Single- Domain Antibodies: A Versatile Tool for in Vivo Imaging of Extracellular and Intracellular Brain Targets.” Journal of Controlled Release: Official Journal of the Controlled Release Society 243 (December): 1–10. 674 Lu, Xiaotang, Xiaomeng Han, Yaron Meirovitch, Evelina Sjöstedt, Richard L. Schalek, and Jeff W. Lichtman. 2023. “Preserving Extracellular Space for High- Quality Optical and Ultrastructural Studies of Whole Mammalian Brains.” Cell Reports Methods 3 (7): 100520. 678 Mahou, Pierre, Maxwell Zimmerley, Karine Loulier, Katherine S. Matho, Guillaume Labroille, Xavier Morin, Willy Supatto, Jean Livet, Delphine Débarre, and Emmanuel Beaurepaire. 2012. “Multicolor Two- Photon Tissue Imaging by Wavelength Mixing.” Nature Methods 9 (8): 815–18. 682 Marmorstein, Alan D., Lihua Y. Marmorstein, Hirokazu Sakaguchi, and Joe G. Hollyfield. 2002. “Spectral Profiling of Autofluorescence Associated with Lipofuscin, Bruch’s Membrane, and Sub- RPE Deposits in Normal and AMD Eyes.” Investigative Ophthalmology & Visual Science 43 (7): 2435–41. 686 Mitchell, Keith G., Belvin Gong, Samuel S. Hunter, Diana Burkart- Waco, Clara E. Gavira- O’Neill, Kayla M. Templeton, Madeline E. Goethel, et al. 2023. “High- Volume Hybridoma Sequencing on the NeuroMabSeq Platform Enables Efficient Generation of Recombinant Monoclonal Antibodies and scFvs for Neuroscience Research.” Scientific Reports 13 (1): 16200. 691 Monnier, Philippe P., Robin J. Vigouroux, and Nardos G. Tassew. 2013. “In Vivo Applications of Single Chain Fv (Variable Domain) (scFv) Fragments.” Antibodies 2 (2): 193–208. 693 Pudavar, Haridas, Judith Reddington, Jason A. Junge, Scott E. Fraser, and Giulia Ossato. 2024. “STELLARIS 8 DIVE: A Rainbow of Possibilities with Multiphoton Excitation and Lifetime- Based Information.” In Multiphoton Microscopy in the Biomedical Sciences XXIV, 12847:32–33. SPIE.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[111, 90, 881, 522]]<|/det|>
+697 Schmid, Benjamin, Gopi Shah, Nico Scherf, Michael Weber, Konstantin Thierbach, Citlali Pérez 698 Campos, Ingo Roeder, Pia Aanstad, and Jan Huisken. 2013. “High-Speed Panoramic Light- 699 Sheet Microscopy Reveals Global Endodermal Cell Dynamics.” Nature Communications 4: 700 2207. 701 Schmidt, Hartmut, Oliver Arendt, Edward B. Brown, Beat Schwaller, and Jens Eilers. 2007. 702 “Parvalbumin Is Freely Mobile in Axons, Somata and Nuclei of Cerebellar Purkinje 703 Neurones.” Journal of Neurochemistry 100 (3): 727–35. 704 Shapson-Coe, Alexander, Michał Januszewski, Daniel R. Berger, Art Pope, Yuelong Wu, Tim 705 Blakely, Richard L. Schalek, et al. 2021. “A Connectomic Study of a Petascale Fragment of 706 Human Cerebral Cortex.” bioRxiv. https://doi.org/10.1101/2021.05.29.446289. 707 Thavarajah, Rooban, Vidya Kazhiyur Mudimbaimannar, Joshua Elizabeth, Umadevi 708 Krishnamohan Rao, and Kannan Ranganathan. 2012. “Chemical and Physical Basics of 709 Routine Formaldehyde Fixation.” Journal of Oral and Maxillofacial Pathology: JOMFP 16 710 (3): 400–405. 711 Thiel, M. A., D. J. Coster, S. D. Standfield, H. M. Brereton, C. Mavrangelos, H. Zola, S. Taylor, 712 A. Yusim, and K. A. Williams. 2002. “Penetration of Engineered Antibody Fragments into 713 the Eye.” Clinical and Experimental Immunology 128 (1): 67–74. 714 Tsuriel, Shlomo, Sagi Gudes, Ryan W. Draft, Alexander M. Binshtok, and Jeff W. Lichtman. 715 2015. “Multispectral Labeling Technique to Map Many Neighboring Axonal Projections in 716 the Same Tissue.” Nature Methods 12 (6): 547–52. 717 Weisser, Nina E., and J. Christopher Hall. 2009. “Applications of Single-Chain Variable 718 Fragment Antibodies in Therapeutics and Diagnostics.” Biotechnology Advances 27 (4): 719 502–20. 720 Wittrup, Anders, Si-He Zhang, Gerdy B. ten Dam, Toin H. van Kuppevelt, Per Bengtson, Maria 721 Johansson, Johanna Welch, Matthias Mörgelin, and Mattias Belting. 2009. “ScFv Antibody- 722 Induced Translocation of Cell-Surface Heparan Sulfate Proteoglycan to Endocytic 723 Vesicles.” The Journal of Biological Chemistry 284 (47): 32959–67.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[116, 88, 320, 104]]<|/det|>
+## REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[116, 121, 403, 137]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 881, 280]]<|/det|>
+The authors have developed a method employing fluorescent single- chain variable fragments (scFvs) for conducting multiplexed detergent- free immunolabeling and volumetric- correlated- light- and- electron- microscopy (vCLEM) on the same samples. In this manuscript, they detail the development of eight fluorescent scFvs targeting specific markers crucial for brain studies. Through experimentation, six fluorescent probes were successfully visualized in the cerebellum using confocal microscopy with spectral unmixing, followed by vEM analysis of the identical sample. The outcomes reveal an exceptional blend of ultrastructure alongside multiple fluorescence channels, offering valuable insights into cellular composition and organization.  
+
+<|ref|>text<|/ref|><|det|>[[115, 296, 861, 360]]<|/det|>
+This approach facilitated the documentation of a previously poorly characterized cell type and the precise subcellular localization. Leveraging scFvs derived from existing monoclonal antibodies opens up the possibility of generating numerous such probes, which could significantly enhance molecular overlays for connectomic investigations.  
+
+<|ref|>text<|/ref|><|det|>[[116, 375, 846, 424]]<|/det|>
+This study represents a significant advancement in the field of vCLEM and holds great promise for researchers seeking to unravel intricate neuronal networks. The revised version of the manuscript effectively addresses key questions and concerns. I don't have further comments.  
+
+<|ref|>text<|/ref|><|det|>[[116, 473, 402, 488]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 504, 878, 552]]<|/det|>
+The revised manuscript thoroughly addresses all of the concerns raised in my initial review. I think the manuscript will have broad appeal and I appreciate the effort put in to address the referees' comments. I strongly recommend publication in Nature Communications.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 89, 323, 108]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 145, 438, 163]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 172, 880, 310]]<|/det|>
+The authors have developed a method employing fluorescent single- chain variable fragments (scFvs) for conducting multiplexed detergent- free immunolabeling and volumetric- correlated- light- and- electron- microscopy (vCLEM) on the same samples. In this manuscript, they detail the development of eight fluorescent scFvs targeting specific markers crucial for brain studies. Through experimentation, six fluorescent probes were successfully visualized in the cerebellum using confocal microscopy with spectral unmixing, followed by vEM analysis of the identical sample. The outcomes reveal an exceptional blend of ultrastructure alongside multiple fluorescence channels, offering valuable insights into cellular composition and organization.  
+
+<|ref|>text<|/ref|><|det|>[[115, 346, 863, 416]]<|/det|>
+This approach facilitated the documentation of a previously poorly characterized cell type and the precise subcellular localization. Leveraging scFvs derived from existing monoclonal antibodies opens up the possibility of generating numerous such probes, which could significantly enhance molecular overlays for connectomic investigations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 454, 872, 507]]<|/det|>
+This study represents a significant advancement in the field of vCLEM and holds great promise for researchers seeking to unravel intricate neuronal networks. The revised version of the manuscript effectively addresses key questions and concerns. I don't have further comments.  
+
+<|ref|>text<|/ref|><|det|>[[172, 515, 785, 534]]<|/det|>
+We thank the reviewer's comments. There's nothing to be further addressed.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 568, 438, 586]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 596, 857, 648]]<|/det|>
+The revised manuscript thoroughly addresses all of the concerns raised in my initial review. I think the manuscript will have broad appeal and I appreciate the effort put in to address the referees' comments. I strongly recommend publication in Nature Communications.  
+
+<|ref|>text<|/ref|><|det|>[[172, 657, 785, 675]]<|/det|>
+We thank the reviewer's comments. There's nothing to be further addressed.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d62429608af455dfb3898d1ec1dbf35ffe750b30c81658917f22585b63bf7d8a/images_list.json

@@ -0,0 +1,124 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_5.jpg",
+    "caption": "Supplementary Figure 5| Temperature-dependent excitation spectra of \\(\\mathrm{SMO:20\\%Yb / 1\\%Er}\\) phosphor.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2. Temperature-dependent proximate distances of RE-RE (RE=Sc/Yb/Er) along different axes derived from the unit cell structure.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  },
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_7.jpg",
+    "caption": "Supplementary Figure 7| Downshifting emission spectra of \\(\\mathrm{SMO:20\\%Yb^{3 + } / 1\\%Er^{3 + }}\\) phosphor and \\(\\mathrm{YF}_3:20\\% \\mathrm{Yb}^{3 + } / 1\\% \\mathrm{Er}^{3 + }\\) phosphor. a. and b. Downshifting emission spectra of the \\(\\mathrm{YF}_3:20\\% \\mathrm{Yb} / 1\\% \\mathrm{Er}\\) and \\(\\mathrm{SMO:20\\%Yb / 1\\%Er}\\) phosphor as a function of temperature under 980 nm excitation. c. Comparison of the relative integrated intensity of the downshifting emission spectra of the \\(\\mathrm{YF}_3:20\\% \\mathrm{Yb} / 1\\% \\mathrm{Er}\\) and \\(\\mathrm{SMO:20\\%Yb / 1\\%Er}\\) phosphor. The enlarged data of \\(\\mathrm{YF}_3:20\\% \\mathrm{Yb} / 1\\% \\mathrm{Er}\\) is displayed on the right side.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 10
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5| Lifetime-based luminescence thermometry in 20% \\(\\mathrm{Yb^{3 + } / 1\\% Er^{3 + }}\\) -codoped SMO. a. Experimentally measured and exponentially fitted plots of lifetime \\(\\tau\\) of SMO:20%Yb/1%Er at different temperatures. b. Calculated absolute sensitivity \\(S_{a}\\) and relative sensitivity \\(S_{r}\\) versus temperature based on the SMO:20%Yb/1%Er. d. Thermal dependence of temperature uncertainty for SMO:20%Yb/1%Er.",
+    "footnote": [],
+    "bbox": [
+      [
+        148,
+        484,
+        848,
+        636
+      ]
+    ],
+    "page_idx": 11
+  },
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_15.jpg",
+    "caption": "Supplementary Figure 15| Temperature-dependent luminescence decay curves of \\(^{4}I_{132}\\) excited states of \\(\\mathrm{Er}^{3 + }\\) in \\(20\\% \\mathrm{Yb}^{3 + } / 1\\% \\mathrm{Er}^{3 + }\\) -codoped SMO at different temperatures from 298 K to 623 K. Each temperature point was measured by ten consecutive measurements.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 12
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_3.jpg",
+    "caption": "Figure 3. e. Rigid unit mode model for \\(\\mathrm{Sc_2Mo_3O_{12}:20\\%Yb^{3 + } / 1\\%Er^{3 + }}\\) extracted from the unit cell structure. f. Temperature-dependence ratio \\(\\delta D\\) of RE-Mo distances marked in the model \\(\\begin{array}{r}\\delta D = \\frac{D_T - D_{420}}{D_{420}}\\times 100\\% \\end{array}\\) \\(D_{T}\\) and \\(D_{420}\\) stand for RE-Mo distances of the given temperature and \\(420\\mathrm{K}\\) , respectively). g. Temperature-dependence ratio \\(\\delta \\theta\\) of RE-Mo-RE angles marked in the model \\(\\begin{array}{r}\\delta \\theta = \\frac{\\theta_T - \\theta_{420}}{\\theta_{420}}\\times 100\\% \\end{array}\\) \\(\\theta_{T}\\) and \\(\\theta_{420}\\) stand for RE-Mo-RE angles of the given temperature and \\(420\\mathrm{K}\\) , respectively).",
+    "footnote": [],
+    "bbox": [
+      [
+        168,
+        90,
+        848,
+        243
+      ]
+    ],
+    "page_idx": 16
+  },
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_3.jpg",
+    "caption": "Supplementary Figure 3| Water molecules analysis. a. Thermogravimetry curves of the \\(\\mathrm{SMO:20\\%Yb / 1\\%Er}\\) phosphor. b. Temperature-dependent Infrared spectra of \\(\\mathrm{SMO:20\\%Yb / 1\\%Er}\\) phosphor.",
+    "footnote": [],
+    "bbox": [
+      [
+        147,
+        384,
+        850,
+        599
+      ]
+    ],
+    "page_idx": 19
+  },
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_4.jpg",
+    "caption": "Supplementary Figure 4 c. Temperature-dependent changes of the unit cell volumes.",
+    "footnote": [],
+    "bbox": [
+      [
+        365,
+        667,
+        644,
+        858
+      ]
+    ],
+    "page_idx": 20
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_2.jpg",
+    "caption": "Figure 2 h. Temperature-dependent in situ Raman spectra within the temperature range from 298 to 773 K.",
+    "footnote": [],
+    "bbox": [
+      [
+        333,
+        118,
+        670,
+        333
+      ]
+    ],
+    "page_idx": 21
+  },
+  {
+    "type": "image",
+    "img_path": "images/Figure_5.jpg",
+    "caption": "Figure 5| Thermally enhanced photoluminescence mechanism a. Temperature-dependent luminescence lifetime of \\(^{2}\\mathrm{F}_{5 / 2}\\) excited state of \\(\\mathrm{Yb^{3 + }}\\) in \\(\\mathrm{Yb^{3 + } / Er^{3 + }}\\) -codoped and \\(\\mathrm{Yb^{3 + }}\\) -doped SMO, respectively. Temperature-dependent energy transfer efficiency of \\(\\mathrm{Yb^{3 + }}\\) -to-Er \\(^{3 + }\\) . b. Temperature-dependent lifetimes of \\(^{2}\\mathrm{H}_{11 / 2}\\) (522 nm) and \\(^{4}\\mathrm{I}_{13 / 2}\\) (1538 nm) excited states of \\(\\mathrm{Er^{3 + }}\\) in SMO: \\(20\\% \\mathrm{Yb} / 1\\% \\mathrm{Er}\\) , respectively. c. Energy level diagram of green UC and NIR DS emission showing the proposed temperature dependence of electronic transition and energy-transfer processes in SMO: \\(\\mathrm{Yb / Er}\\) with two-dimensional negative thermal expansion.",
+    "footnote": [],
+    "bbox": [
+      [
+        187,
+        92,
+        844,
+        455
+      ]
+    ],
+    "page_idx": 22
+  }
+]

peer_reviews/supplementary_0_Peer Review File__d62429608af455dfb3898d1ec1dbf35ffe750b30c81658917f22585b63bf7d8a/supplementary_0_Peer Review File__d62429608af455dfb3898d1ec1dbf35ffe750b30c81658917f22585b63bf7d8a.mmd

@@ -0,0 +1,616 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Thermally boosted upconversion and downshifting luminescence in Sc2(MoO4)3:Yb/Er with twodimensional negative thermal expansion  
+
+![](images/Supplementary_Figure_5.jpg)
+  
+
+![](images/Figure_2.jpg)
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to  
+
+the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+The work concerns a very interesting and popular topic - luminescence thermometry. The presented strategy using a material with a negative temperature expansion coefficient as a material with a positivistic thermal luminescence coefficient seems to be a very promising approach offering very high relative sensitivities. Therefore, I believe that the work has an important aspect of novelty. I am not sure that it will be of interest to a wide readership of this journal. Nevertheless, some issues should be clarified before the work can be considered for publication in Nature Communications:  
+
+1. Caption of Figure 6 is misleading. Actually, the authors do not show any application there.  
+
+2. Authors should, in accordance with current standards, use an uppercase "K" to designate temperature units  
+
+3. Although the summary of results presented in Figure 6d is very informative, I encourage the authors to contrast the results presented with respect to recently published work for which very high relative sensitivities were obtained such as 10.1016/j.cej.2021.131165  
+
+4. The interionic distance contraction is one of the very specific mechanisms responsible for the thermal enhancement of the luminescence. There are several other processes that may lead to the similar thermal behaviors like i.e. change of the local ion symmetry etc.  
+
+5. The purpose of the data shown in Figures 1 a, b, c is to describe three types of temperature effect trends on luminescence intensity. Therefore, I would suggest removing the "diagrams" placed under the graphs to maintain the universality of the description presented.  
+
+6. Authors should use consistent temperature units throughout the paper (e.g. Fig.1 Celsius and Figure 6 Kelvins).  
+
+7. Please remove the information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and b  
+
+8. The increase in emission intensity with increasing temperature may be due to an increase in the absorption cross section associated with the spectral broadening of the absorption band. This occurs when the excitation line etches into the slope of the absorption band. Therefore, the authors should present the excitation spectra of Yb3+ ions measured as a function of temperature
+
+<--- Page Split --->
+
+9. The magnitude of the thermal contraction of the crystalographic unit cell depends on the difference in the ionic radii between dopants and replaces ions and on the concentration of dopants. Therefore, I would like the authors to comment if it is possible to modulate the thermometric parameters of this type of thermometer by changing the dopant concentration?  
+
+10. Table S1- the comparing the thermometric properties of the luminescent thermometry operating in different readout modes is unjustified and unreasonable.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors of "Thermally boosted upconversion and downshifting luminescence in Sc2(MoO4)3:Yb/Er with two- dimensional negative thermal expansion" presented a very interesting research of the famous upconverting pair codoped in more than astounding host matrix. However, for the following reasons I recommend major revision and reinspection, as there are numerous problems with novelty, significance and support for the conclusions. Briefly these problems can be summed as:  
+
+- The same host matrix was investigated for thermometry by the Yb/Ho pair and LIR with similar conclusions. This just slightly diminishes the stated novelty.  
+
+- The discussion about inter-atomic differences is possibly erroneous, and need either serious revision or removal.  
+
+- The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.  
+
+- This high relative sensitivity occurs at the points where the signal is borderline detectable. This means that the temperature resolutions will be low at those points, i.e. the significance of the research is not as high as initially suggested. The authors need to include temperature resolution discussion in the most correct manner, and accordingly adjust their conclusions.  
+
+Issues in detail:  
+
+In the first reading the novelty of this article seams immense. However, the exactly the same host, co- doped with an upconverting pair of Yb3+/Ho3+ has been already explored for luminescence thermometry in Ref: https://pubs.acs.org/doi/10.1021/acs.jpclett.0c00628. Although the authors did reference this work, their contribution and similarity needs to be more stressed. That being said, the differences in employing the more important upconverting pair, Yb3+/Er3+, and lifetime instead
+
+<--- Page Split --->
+
+of luminescence intensity ratio method, should be stated as the novelty of this work. In my opinion, this subtle differences are important and enough novelty for the publication of this work.  
+
+For the equation 1 the referenced is the famous short work by Blasse (Ref. 22). However, in that reference the Blasse is not reporting on a ways of calculation of the average distance, but on the critical distance, i.e. when the concentration quenching starts to dominate. The c in the subscript of xc does not abbreviate concentration, but the word critical. Up to my knowledge, this equation cannot be applied for calculation of distance between interacting ions if that distance is not critical distance, and even more, of the ions of different types. Unless the authors can provide another, a very strong reference to support their claims of usage of equation 1, I suggest removing the Figure 2d, Equation 1, and any corresponding discussion.  
+
+There is another issue with concentrations and distances. In figure 2 no concentrations of dopants are mentioned and yet there is a graph of their distances. To additionally confirm the assumption that the author's analysis of interatomic distances is incorrect is the un- logical trends in presented distances. Although both the codopants substitute the same ions in the SMO matrix, the authors claim that Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase. All three claims cannot be possible, if I understood well, if not, then the paragraph needs to be revised for clarity. As for the red emission of Er3+, the authors should see the book by Kaminskii, Crystalline lasers, where the pathways for population of the 4F9/2 are explained, and they are related to the Yb3+ ion back- transfer more than the Er3+ cross- relaxation.  
+
+Lifetime method for thermometry is erroneously described as ratiometric multiple times in the abstract and throughout the article. The definition of ratiometric fluorescence is where intensities of at least two bands are measured and compared. Thus, the word ratio in ratiometric. FIR or LIR methods, of which the authors are certainly aware of, are the ratiometric luminescence thermometry methods, while lifetime, although also self- referenced, is not. This needs to be corrected in the caption of Table S1 as well.  
+
+Throughout the article there needs to be more linear story of using concentrations and even hosts. It is hard to follow which concentrations are being used, and when it is compared to the other host. For each spectra the precise concentrations of codopants should be given.  
+
+The authors discuss the 4111/2 emission of Er3+, however in the spectra and in the text the 1538 nm emission is mentioned. Did you mean the emission from 4113/2 instead, as the emission from 4111/2 to the ground level is at about 1000 nm? See for example Figure S9b: caption says 4111/2 and Eem = 1538 nm, the same in Figure 5.  
+
+For other journals of lower esteem than the Nature series, the comparison with other hosts would not be necessary. However, I feel that in this host the emission intensity is much lower than in other famous upconverting hosts, for example YF3 or NaYF4. I would like to see the comparison in either intensities or quantum yields between SMO and one of these prominent hosts.  
+
+Lastly but not the least, this amazing sensitivity occurs at the exact same temperatures where the emission intensities are very low. Thus, at the points of the highest relative sensitivity, the sensing might be just possible, as the uncertainty in measurement will be great. This was discussed in the article: https://onlinelibrary.wiley.com/doi/full/10.1002/adts.202000176 . This last point
+
+<--- Page Split --->
+
+significantly diminishes the firstly claimed importance of the obtained results. I invite the authors to be honest in reporting the sensitivities only on the practically usable range. Also, taking everything in consideration, there is a must that the authors report on the temperature resolutions at the whole temperature range, and unfortunately this means reporting on the uncertainties that should be evaluated by at least several temperatures. Uncertainties should be given as the standard deviation of at least 10 consecutive measurements at the same temperature. This temperature resolutions will show the true usability of the sensor, and my assumption is that the temperature resolution curve will be different than the sensitivity curves.  
+
+Reviewer #3 (Remarks to the Author):  
+
+The authors present a two- dimensional negative thermal phosphor SMO:Yb/Er systematic study. They present a temperature dependent Raman, synchrotron X- ray diffraction and luminescence investigation. The thermally boosted UC and DS luminescence mechanism was investigated. The authors claim that the luminescence lifetime temperature dependency enables ratiometric thermometry with high relative sensitivity; an important topic in current research.  
+
+This is a comprehensive work, very well- written and with insights into the \(\mathsf{Ln3 + }\) - doped phosphors for potential applications.  
+
+Although they address an important topic some issues, that I refer below, need further clarification.  
+
+Authors should comment, if known, what is the mechanism for the NTE behaviour in this system, whether it is structural, or electronically based for instance?  
+
+What is supposed to drive the transition from PTE to NTE around \(100^{\circ}C\)  
+
+The transition from PTE to NTE occurs at ca \(100^{\circ}C\) (figure 2c). How this transition impacts the luminescence properties and how this is compatible with the suggested thermally enhanced photoluminescence mechanism. It is important to clarify this issue.  
+
+Some Raman modes appear to alter above \(100K\) , can they be related with the NTE to PTE transition?
+
+<--- Page Split --->
+
+From lines 182- 183: "Moreover, the distance of \(\mathsf{Yb3 + - Er3 + }\) becomes shorter with an increase of the temperature. The ET processes between sensitizer \((\mathsf{Yb3 + })\) to activator \((\mathsf{Er3 + })\) are usually considered to occur through dipolar- dipolar interactions, whose ET efficiency is proportional to r- 6 (r is the average donor- acceptor distance). The result indicates that the distances of \(\mathsf{Yb3 + - Er3 + }\) ions (13.975 \(\approx\) , at \(25^{\circ}C\) ) decreased at higher temperature, which may benefit the improvement of the ET efficiency between \(\mathsf{Yb3 + - Er3 + }\) ."  
+
+And line 236 "Based on the above analysis, the thermally enhanced UC emission with the temperature from 25 to \(200^{\circ}C\) is mainly governed by the ET from \(\mathsf{Yb3 + }\) to \(\mathsf{Er3 + }\) "  
+
+Authors should clarify the nature for thermally enhanced UC emission within the temperature range from 25 to \(200^{\circ}C\) , if this effect is mainly governed by the ET from \(\mathsf{Yb3 + }\) to \(\mathsf{Er3 + }\) , how it relates with the PTE of the system shown in Fig. 2 observed between \(250 - C\) and \(1000C\) .  
+
+From lines 46- 49: "Most of such abnormal thermally enhanced UC luminescence is observed in \(\mathsf{Ln3 + }\) - doped inorganic materials with three- dimensional negative- thermal expansion characteristics (NTE)9, 10, where all the three cell parameters of the doped crystals shrink at elevated temperature."  
+
+And below, lines 58- 59 "with a unique two- dimensional NET coefficient \((\alpha a = - 8.62\times 10 - 6 / K,\) \(\alpha b = 4.25\times 10 - 6 / K,\) \(\alpha c = - 6.35\times 10 - 6 / K)\) "  
+
+In the context of this work, is there any effective advantage in two- dimensional negative- thermal expansion materials when compared with the three- dimensional ones? The authors could also mention recent works with uniaxial NTE.  
+
+Can the authors explain in what way the Energy migration between \(\mathsf{Ln3 + }\) dopants is confined to two dimensions?  
+
+From lines 231- 233. The promoted Ar and inhibited Anr in the \(\mathsf{Ln3 + }\) dopant indicate that lattice distortion enhances a crystal field with odd parity and modifies local symmetry of activator ions via the NTE effect.  
+
+The authors mention the existence of lattice distortion and a change in the local symmetry of activator ions via the NTE effect. Can these local distortions be substantiated via the temperature- dependent Raman data?  
+
+Can this effect be witnessed by the other techniques that would perceive the negative thermal expansion at the local scale? The authors could comment on that.  
+
+Along the manuscript:
+
+<--- Page Split --->
+
+What is the degree of Sc substitution in the samples that were measured in the presented results Figure 2,3,4,5? This is only specified in the labels of the Figures presented in the Supplementary material.  
+
+line 18 "decrease at the higher temperature" author should substitute "decrease for high temperatures"  
+
+line 58 and 103 Where is the abbreviation "NET" defined? Do the authors mean "NTE"?  
+
+Line 121- 122: What is the value of Xc used for the calculations?  
+
+Line 133 - 135 "The observed mode with a frequency of 341 cm- 1 corresponds to the bending mode of MoO4 tetrahedra. When the temperature increases, this Raman peak exhibits a blue shift (Fig. 2f). The median frequency B1g mode of SMO is observed at 511 cm- 1"  
+
+Is this a typo? Figure 2f is not from the 341cm- 1 mode. The sentence should be improved.  
+
+The authors refer the studied systems as "the structures of the SMO:Yb/Er samples with different dopant concentrations", doping generally refers to a small concentrations, here the degree of ionic Sc substitution by Yb and Er cations reaches the high value of \(25\%\) or \(8\%\) . The term chemical substitution would be more appropriated.
+
+<--- Page Split --->
+
+## Reply to the Comments of Reviewer #1  
+
+## General comment:  
+
+The work concerns a very interesting and popular topic - luminescence thermometry. The presented strategy using a material with a negative temperature expansion coefficient as a material with a positivistic thermal luminescence coefficient seems to be a very promising approach offering very high relative sensitivities. Therefore, I believe that the work has an important aspect of novelty. I am not sure that it will be of interest to a wide readership of this journal. Nevertheless, some issues should be clarified before the work can be considered for publication in Nature Communications:  
+
+## Response:  
+
+We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. A point- by- point response is noted below.  
+
+## Comment #1:  
+
+Caption of Figure 6 is misleading. Actually, the authors do not show any application there.  
+
+## Response:  
+
+Thank you for your kind reminding. Caption of Figure 6 has been corrected as "Lifetime- based luminescence thermometry" in the revised manuscript. Please see also line 1 of p. 15 in the revised manuscript.  
+
+## Comment #2:  
+
+Authors should, in accordance with current standards, use an uppercase "K" to designate temperature units  
+
+## Response:  
+
+Temperature units have been revised as uppercase "K" in the revised manuscript.  
+
+## Comment #3:  
+
+Although the summary of results presented in Figure 6d is very informative, I encourage the authors to contrast the results presented with respect to recently published work for which very high relative sensitivities were obtained such as 10.1016/j.cej.2021.131165.  
+
+## Response:
+
+<--- Page Split --->
+
+The suggested reference has been cited as Ref. 54, which was compared with our results and discussed in the revised manuscript. Specifically, the optimal \(S_{r}\) value of \(\mathrm{SMO:20\%Yb / 1\%Er}\) \((12.3\% \mathrm{K}^{- 1})\) is higher than that \((8.83\% \mathrm{K}^{- 1})\) of \(\mathrm{SrTiO_3:Tb^{3 + }}\) nanocrystals in the suggested reference (Chem. Eng. J. 2022, 428, 131165). Please see also line 27 of p.14 in the revised manuscript.  
+
+## Comment #4:  
+
+The interionic distance contraction is one of the very specific mechanisms responsible for the thermal enhancement of the luminescence. There are several other processes that may lead to the similar thermal behaviors like i.e. change of the local ion symmetry etc.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Indeed, the interionic distance contraction of the different axis is one of the very specific mechanisms responsible for the thermal enhancement of luminescence. Besides, the change of the local ion symmetry may also affect the optical properties of \(\mathrm{RE}^{3 + }\) ions like luminescence intensity and radiative transition. In this work, we propose a new class of NTE phosphors based on \(\mathrm{SMO:Yb / Er}\) exhibiting a unique two- dimensional NTE coefficient. With increasing the temperature, the \(\mathrm{SMO:Yb / Er}\) phosphors experience an anisotropic two- dimensional shrinkage along \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions increases along \(b\) axis. As such, the detrimental energy migration of \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions can be effectively confined within the two- dimensional layered structures (e.g., (010) lattice plane) at elevated temperature. More discussion has been added. Please see also lines 7- 16 of p. 5 in the revised manuscript.  
+
+## Comment #5:  
+
+The purpose of the data shown in Figures 1 a, b, c is to describe three types of temperature effect trends on luminescence intensity. Therefore, I would suggest removing the "diagrams" placed under the graphs to maintain the universality of the description presented.  
+
+## Response:  
+
+Thanks a lot. We have removed the "diagrams" placed under the graphs in the revised manuscript. The captions of Figures 1 a, b, c have been corrected as " Figure 1 Scheme of thermal- dependence effects in phosphor. a. Positive thermal quenching phenomenon. b. Zero thermal quenching phenomenon. c. Negative- thermal quenching (thermal- enhanced) phenomenon". Please see also line 1 of p. 3 in the revised manuscript.
+
+<--- Page Split --->
+
+## Comment #6:  
+
+Comment #6:Authors should use consistent temperature units throughout the paper (e.g. Figure1 Celsius and Figure 6 Kelvins).  
+
+## Response:  
+
+Response:All the temperature units have been corrected as Kelvins in the revised manuscript.  
+
+## Comment #7:  
+
+Comment #7:Please remove the information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and b.  
+
+## Response:  
+
+Response:Thanks. The information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and 2b has been removed in the revised manuscript. Please see also line 1 of p. 6 in the revised manuscript.  
+
+## Comment #8:  
+
+Comment #8:The increase in emission intensity with increasing temperature may be due to an increase in the absorption cross section associated with the spectral broadening of the absorption band. This occurs when the excitation line etches into the slope of the absorption band. Therefore, the authors should present the excitation spectra of \(\mathrm{Yb^{3 + }}\) ions measured as a function of temperature.  
+
+## Response:  
+
+Response:Many thanks for this valuable suggestion. We have added the excitation spectra of \(\mathrm{Yb^{3 + }}\) ions measured as a function of temperature in the revised manuscript (Supplementary Figure 5 newly added). Accordingly, the absorption band centered at 980 nm was hardly broadened with increasing temperature. The intensity of the excitation peak increased with increasing the temperature from 298 to 573 K, which may benefit the luminescence emission of \(\mathrm{Yb^{3 + }}\) at the higher temperature. Please see also lines 7- 11 of p. 8 in the revised manuscript.
+
+<--- Page Split --->
+![](images/Supplementary_Figure_7.jpg)
+
+<center>Supplementary Figure 5| Temperature-dependent excitation spectra of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. </center>  
+
+## Comment #9:  
+
+The magnitude of the thermal contraction of the crystallographic unit cell depends on the difference in the ionic radii between dopants and, replaces ions and on the concentration of dopants. Therefore, I would like the authors to comment if it is possible to modulate the thermometric parameters of this type of thermometer by changing the dopant concentration?  
+
+## Response:  
+
+Many thanks for this valuable suggestion. It is true that the magnitude of the thermal contraction of the crystallographic unit cell depends on the difference in the ionic radii between dopants and replaces ions and on the concentration of dopants. Therefore, we investigated the thermometric parameters of \(\mathrm{SMO:xYb / 1\%Er}\) with different concentrations of sensitizer \(\mathrm{(Yb^{3 + })}\) . Following the reviewer's suggestion, we have measured the temperature- dependent decay curves of \(\mathrm{SMO:Yb / 1\%Er}\) and phosphors with different \(\mathrm{Yb^{3 + }}\) concentrations (10%, 20% and 25%), which were newly added as Supplementary Figure 14 in the Supporting Information. As indicated in Supplementary Table 1, \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors exhibit the optimal absolute sensitivity \(\mathrm{S_a}\) (53.0 \(\mu \mathrm{sK}^{- 1}\) ) and relative sensitivity \(\mathrm{S_r}\) (12.3% \(\mathrm{K}^{- 1}\) ). Please see also lines 20- 23 of p. 14 in the revised manuscript.
+
+<--- Page Split --->
+![](images/Figure_5.jpg)
+  
+
+Supplementary Figure/14] Lifetime- based luminescence thermometry in \(\mathbf{x}\% \mathbf{Yb}^{3 + } / 1\% \mathbf{Er}^{3 + }\) - codoped SMO with different \(\mathbf{Yb}^{3 + }\) concentrations. a. Temperature- dependent luminescence decay curves of \(^{4}\mathrm{I}_{13 / 2}\) excited states of \(\mathrm{Er}^{3 + }\) in \(10\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) - codoped SMO. b. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO: \(10\% \mathrm{Yb} / 1\% \mathrm{Er}\) at different temperatures. c. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO: \(10\% \mathrm{Yb} / 1\% \mathrm{Er}\) . d. Temperature- dependent luminescence decay curves of \(^{4}\mathrm{I}_{13 / 2}\) excited states of \(\mathrm{Er}^{3 + }\) in \(25\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) - codoped SMO. e. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO: \(25\% \mathrm{Yb} / 1\% \mathrm{Er}\) at different temperatures. f. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO: \(25\% \mathrm{Yb} / 1\% \mathrm{Er}\) .
+
+<--- Page Split --->
+
+Supplementary Table 1]. Lifetime-based luminescence thermometry parameters in x%Yb3+/1%3+-codoped SMO with different Yb3+ concentrations.
+
+<table><tr><td>Sample</td><td>\(S_{a}(\mu \mathrm {s}/\mathrm {K})\)</td><td>\(S_{r}(\%/K)\)</td></tr><tr><td>\(10\%Yb/1\%Er:SMO\)</td><td>31.8</td><td>3.0</td></tr><tr><td>\(20\%Yb/1\%Er:SMO\)</td><td>53.0</td><td>12.3</td></tr><tr><td>\(25\%Yb/1\%Er:SMO\)</td><td>31.2</td><td>7.1</td></tr></table>
+
+# Comment #10:
+
+Table S1- the comparing the thermometric properties of the luminescent thermometry operating in different readout modes is unjustified and unreasonable.
+
+# Response:
+
+Thank you. Table S1 has been deleted in the revised manuscript.
+
+![](images/Supplementary_Figure_15.jpg)
+
+
+<--- Page Split --->
+
+## Reply to the Comments of Reviewer #2  
+
+## General comment:  
+
+The authors of "Thermally boosted upconversion and downshifting luminescence in \(Sc_{2}(MoO_{4})_{3}\cdot Yb / Er\) with two- dimensional negative thermal expansion" presented a very interesting research of the famous upconverting pair codoped in more than astounding host matrix. However, for the following reasons I recommend major revision and reinspection, as there are numerous problems with novelty, significance and support for the conclusions.  
+
+## Response:  
+
+We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. We have made all the requested changes in the revised manuscript according to the reviewer's comments. Particularly, more control experiments and analyses have been added to highlight the novelty of our work. A point- by- point response is noted below.  
+
+## Comment #1:  
+
+The same host matrix was investigated for thermometry by the Yb/Ho pair and LIR with similar conclusions. This just slightly diminishes the stated novelty.  
+
+## Response:  
+
+Thank you for pointing out this important issue. It is true that the same host matrix was investigated as traditional LIR thermometry by the Yb/Ho pair (Ref. 10). Nevertheless, the content and novelty of our work are totally different from the previous report. For the first time, we have revealed the unique two- dimensional NTE of SMO:Yb/Er phosphors, which exhibited simultaneous enhancements of upconversion and downshifting photoluminescence (PL) of \(\mathrm{Er}^{3 + }\) by 45- fold and 450- fold from 298 to 773 K, respectively. Besides, the strategy of NIR downshifting luminescence lifetime instead of the conventional visible UC luminescence intensity ratio method was applied for temperature sensing. The near- infrared luminescence lifetime of \(\mathrm{Er}^{3 + }\) in SMO:Yb/Er was determined to span two orders of magnitude from 298 to 623 K, which enables their application in lifetime- based luminescent thermometry with high absolute temperature sensitivity and relative temperature sensitivity. Our work presents a substantial advance as compared to previous work and thus could be a breakthrough in the development of \(\mathrm{RE}^{3 + }\) - doped NTE phosphors. Please see also lines 20- 26 of p. 16 in the revised manuscript.  
+
+## Comment #2:  
+
+The discussion about inter- atomic differences is possibly erroneous, and need either serious revision or removal.
+
+<--- Page Split --->
+
+## Response:  
+
+Response:Many thanks for this valuable suggestion. We have corrected the discussion about inter- atomic differences, which were estimated by the crystallographic information file (CIF) of Rietveld refinement of the in situ temperature- dependent SXRD. Please see also lines 7- 16 of p. 5 in the revised manuscript.  
+
+## Comment #3:  
+
+Comment #3:The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.  
+
+## Response:  
+
+Comment #3:The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.Response:Many thanks for this valuable suggestion. We have added the discussion about the overall intensity of this phosphor. To explicitly indicate the overall intensity of this phosphor, the famous Orthorhombic- phase \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor with the same crystallographic system of SMO is selected as the control samples. To compare the emission intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor, we have measured the temperature- dependent UC/DS luminescence spectra of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) under otherwise identical conditions (Supplementary Figures. 6 and 7 newly added). It can be observed that the UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor decreased continuously with increasing temperature, while the UC/DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor increased markedly with increasing the temperature. Specifically, the overall UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is much higher than that of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) at 298 K. Nevertheless, the overall UC and DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor are 4.5 and 12.9 times higher than that of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) counterpart at 773 K, respectively. These results explicitly validate the superiority of the \(\mathrm{SMO}:\mathrm{Yb} / \mathrm{Er}\) as novel luminescent materials over existing PTE phosphors, particularly at high temperatures. The above discussions have been added in lines 1- 13 of p. 10 and Supplementary Figures 6- 7 newly added in the Supporting Information.  
+
+## Comment #4:  
+
+Comment #4:This high relative sensitivity occurs at the points where the signal is borderline detectable. This means that the temperature resolutions will be low at those points, i.e. the significance of the research is not as high as initially suggested. The authors need to include temperature resolution discussion in the most correct manner, and accordingly adjust their conclusions.  
+
+## Response:  
+
+Response:Thank you for pointing out this important issue. To determine the relative sensitivity of our phosphors at other temperatures, we have measured the luminescence lifetimes of our phosphors at four temperature points (303, 313, 333, 363 K). All the PL decays
+
+<--- Page Split --->
+
+for each temperature point were measured independently for ten times under identical conditions to yield the average value. Temperature resolution discussion via data analysis of the 10 consecutive measurements has been newly added in lines 1- 14 in p. 16 in the revised manuscript. A slight change in sensitivity was obtained by re- fitting the data by adding data pointing. From Figures 5 a and 5b, it is obvious that the data become dense at the temperature range from 298 to 373K, which may improve the data accuracy via decreasing the fitting error. The optimal \(S_{a}\) and \(S_{r}\) were determined to be as high as \(53.0 \mu \mathrm{sK}^{- 1}\) and \(12.3\% \mathrm{K}^{- 1}\) , respectively. The fitting results have been adjusted in the revised manuscript.  
+
+## Comment #5:  
+
+Issues in detail:  
+
+In the first reading the novelty of this article seems immense. However, the exactly the same host, co- doped with an upconverting pair of \(\mathrm{Yb}^{3 + } / \mathrm{Ho}^{3 + }\) has been already explored for luminescence thermometry in Ref: https://pubs.acs.org/doi/10.1021/acs.jpclett.0c00628. Although the authors did reference this work, their contribution and similarity needs to be more stressed. That being said, the differences in employing the more important upconverting pair, \(\mathrm{Yb}^{3 + } / \mathrm{Er}^{3 + }\) , and lifetime instead of luminescence intensity ratio method, should be stated as the novelty of this work. In my opinion, this subtle differences are important and enough novelty for the publication of this work.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Previously, SMO:Yb/Ho with same host was used as a ratiometric thermometer based on the UC red- to- green emission intensity ratio. Their contribution and similarity about the same SMO:Yb/Ho used as luminescence thermometry in Ref. 10 have been stressed in the revised manuscript. The above discussions have been added in lines 29- 30 of p. 13 in the revised manuscript.  
+
+It should be noted that the content and novelty in this work are totally different from the previous report. For the first time, we have revealed the unique two- dimensional NTE of SMO:Yb/Er phosphors, which exhibited simultaneous enhancements of upconversion and downshifting PL of \(\mathrm{Er}^{3 + }\) by 45- fold and 450- fold from 298 to 773 K, respectively. Besides, the strategy of NIR downshifting luminescence lifetime instead of the conventional visible UC luminescence intensity ratio method was applied for temperature sensing. The near- infrared luminescence lifetime of \(\mathrm{Er}^{3 + }\) in SMO:Yb/Er was determined to span two orders of magnitude from 298 to 623 K, which enables their application in lifetime- based luminescent thermometry with high absolute temperature sensitivity and relative temperature sensitivity. Our work presents a substantial advance as compared to previous work and thus could be a breakthrough in the development of \(\mathrm{RE}^{3 + }\) - doped NTE
+
+<--- Page Split --->
+
+phosphors. We sincerely hope the reviewer concurs after reading this clarification.  
+
+## Comment #6:  
+
+For the equation 1 the referenced is the famous short work by Blasse (Ref. 22). However, in that reference the Blasse is not reporting on a ways of calculation of the average distance, but on the critical distance, i.e. when the concentration quenching starts to dominate. The c in the subscript of xc does not abbreviate concentration, but the word critical. Up to my knowledge, this equation cannot be applied for calculation of distance between interacting ions if that distance is not critical distance, and even more, of the ions of different types. Unless the authors can provide another, a very strong reference to support their claims of usage of equation 1, I suggest removing the Figure 2d, Equation 1, and any corresponding discussion.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we have updated Figure 2d. For the case of homogeneous substitution of \(\mathrm{Sc^{3 + }}\) by \(\mathrm{RE^{3 + }}\) ions in SMO, the changing trend of the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions ( \(\mathrm{RE} = \mathrm{Sc / Yb / Er}\) ) along different crystal axis- direction in the structure diagram for different temperatures will reflect the changing trend of the distance between the \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) ions. With the increase of temperature, the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(a\) - axis (6.7061 Å, at 298 K) contracts, and the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(b\) - axis (10.8288 Å, at 298 K) expands steadily, while the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(c\) - axis first expands in the temperature range (298- 373 K) and then contracts. The results indicate that the distances of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(a / c\) - axis decreased at higher temperatures, which may benefit the improvement of the energy- transfer efficiency between \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) . See also the updated discussion in lines 7- 16 on page 5 of the revised manuscript.  
+
+![](images/Figure_3.jpg)
+
+<center>Figure 2. Temperature-dependent proximate distances of RE-RE (RE=Sc/Yb/Er) along different axes derived from the unit cell structure. </center>
+
+<--- Page Split --->
+
+## Comment #7:  
+
+There is another issue with concentrations and distances. In figure 2 no concentrations of dopants are mentioned and yet there is a graph of their distances. To additionally confirm the assumption that the author's analysis of interatomic distances is incorrect is the un- logical trends in presented distances. Although both the codopants substitute the same ions in the SMO matrix, the authors claim that Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase. All three claims cannot be possible, if I understood well, if not, then the paragraph needs to be revised for clarity. As for the red emission of \(\mathrm{Er}^{3 + }\) , the authors should see the book by Kaminskii, Crystalline lasers, where the pathways for population of the \(^4\mathrm{F}_{9 / 2}\) are explained, and they are related to the \(\mathrm{Yb}^{3 + }\) ion back- transfer more than the \(\mathrm{Er}^{3 + }\) cross- relaxation.  
+
+Response: Thank you for pointing out our mistakes. In Figure 2, the \(\mathrm{Yb}^{3 + } / \mathrm{Er}^{3 + }\) dopant concentrations are \(20\%\) and \(1\%\) , respectively, which was added in the revised manuscript.  
+
+For the incorrect description "Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase.", we have updated Figure 2d, wherein the distances of \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions at different planes were determined by the crystallographic information file (CIF) of Rietveld refinement on the basis of the SXRD patterns (Supplementary Figure 3c). See also the updated discussion in lines 1- 10 on page 6 of the revised manuscript.  
+
+The confusion of the temperature- dependent luminescence mechanism in this work was perhaps caused by our unclear writing in the manuscript. Regarding the red emission of \(\mathrm{Er}^{3 + }\) , it is true that the shortened \(\mathrm{Yb}^{3 + } - \mathrm{Er}^{3 + }\) interatomic distance may contribute to the back energy transfer from \(\mathrm{Er}^{3 + }\) to \(\mathrm{Yb}^{3 + }\) ions: \(^4\mathrm{S}_{3 / 2}(\mathrm{Er}^{3 + }) + ^2\mathrm{F}_{7 / 2}(\mathrm{Yb}^{3 + })\) \(\rightarrow {}^{4}\mathrm{I}_{13 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{5 / 2}(\mathrm{Yb}^{3 + })\) , followed by energy transfer from \(\mathrm{Yb}^{3 + }\) to \(\mathrm{Er}^{3 + }\) through \(^4\mathrm{I}_{13 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{5 / 2}(\mathrm{Yb}^{3 + })\rightarrow {}^{4}\mathrm{F}_{9 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{7 / 2}(\mathrm{Yb}^{3 + })\) . As a result, the population of the \(\mathrm{F}_{9 / 2}\) states is enhanced to produce stronger red UC emissions (Ref. 34 Kaminskii A., Crystalline Lasers: Physical Processes and Operating Schemes. (CRC Press, 2020).). The above discussion has been added in lines 20- 24 of p. 7 in the revised manuscript.  
+
+## Comment #8:  
+
+Lifetime method for thermometry is erroneously described as ratiometric multiple times in the abstract and throughout the article. The definition of ratiometric fluorescence is where intensities of at least two bands are measured and compared. Thus, the word ratio in ratiometric. FIR or LIR methods, of which the authors are certainly aware of, are the ratiometric luminescence thermometry methods, while lifetime, although also self- referenced, is not. This needs to be corrected in the caption of Table S1 as well.
+
+<--- Page Split --->
+
+Response: Thanks. We have carefully checked the manuscript and revised similar mistakes. In the abstract "the luminescence lifetime of \(^{4}\mathrm{I}_{11 / 2}\) of \(\mathrm{Er}^{3 + }\) in \(\mathrm{SMO:Yb / Er}\) displays a strong temperature dependence, enabling ratiometric thermometry with the highest relative sensitivity of \(12.3\% / \mathrm{K}\) at \(298\mathrm{K}\) has corrected as "the luminescence lifetime of \(^{4}\mathrm{I}_{11 / 2}\) of \(\mathrm{Er}^{3 + }\) in \(\mathrm{SMO:Yb / Er}\) displays a strong temperature dependence, enabling luminescence thermometry with the highest relative sensitivity of \(12.3\% / \mathrm{K}\) at \(298\mathrm{K}\) ". Please see also line 30 p. 1 in the revised manuscript. Besides, Table S1 has been deleted according to the comment of the first reviewer.  
+
+## Comment #9:  
+
+Throughout the article there needs to be more linear story of using concentrations and even hosts. It is hard to follow which concentrations are being used, and when it is compared to the other host. For each spectra the precise concentrations of codopants should be given.  
+
+## Response:  
+
+Thank you for your kind reminding. For each spectrum, the precise concentration of dopants has been given in the revised manuscript.  
+
+## Comment #10:  
+
+The authors discuss the \(^{4}\mathrm{I}_{11 / 2}\) emission of \(\mathrm{Er}^{3 + }\) , however in the spectra and in the text the \(1538\mathrm{nm}\) emission is mentioned. Did you mean the emission from \(^{4}\mathrm{I}_{13 / 2}\) instead, as the emission from \(^{4}\mathrm{I}_{11 / 2}\) to the ground level is at about \(1000\mathrm{nm}\) ? See for example Figure S9b: caption says \(^{4}\mathrm{I}_{11 / 2}\) and \(\mathrm{Eem} = 1538\mathrm{nm}\) , the same in Figure 5.  
+
+## Response:  
+
+Thank you for your kind reminding. Indeed, the emission peak of \(1538\mathrm{nm}\) originated from \(^{4}\mathrm{I}_{13 / 2}\) of \(\mathrm{Er}^{3 + }\) . All the " \(^{4}\mathrm{I}_{11 / 2}\) " have been revised as " \(^{4}\mathrm{I}_{13 / 2}\) " in the revised manuscript.  
+
+## Comment #11:  
+
+For other journals of lower esteem than the Nature series, the comparison with other hosts would not be necessary. However, I feel that in this host the emission intensity is much lower than in other famous upconverting hosts, for example YF3 or NaYF4. I would like to see the comparison in either intensities or quantum yields between SMO and one of these prominent hosts.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we have selected \(\mathrm{YF}_{3}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor as the control group. As we know, the
+
+<--- Page Split --->
+
+famous Orthorhombic- phase \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor with the same crystallographic system of SMO is considered to be one of the most efficient UC/DS luminescent materials. To compare the emission intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor, we have measured the temperature- dependent UC/DS spectra of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) under otherwise identical conditions (Supplementary Figures 6 and 7 newly added). It can be observed that the UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor decreased with increasing temperature, while the UC/DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor increased markedly with increasing the temperature. Specifically, the overall UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is much higher than that of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) at 298 K. Nevertheless, the overall UC and DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor are 4.5 and 12.9 times higher than that of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) counterpart at 773 K, respectively. These results explicitly validate the superiority of the SMO:Yb/Er as novel luminescent materials over existing PTE phosphors, particularly at high temperatures. The above discussions have been added in lines 1- 13 of p. 10 and Supplementary Figures 6 and 7 newly added in the Supporting Information.  
+
+![](images/Supplementary_Figure_3.jpg)
+
+
+<--- Page Split --->
+
+Supplementary Figure 6| Upconversion emission spectra of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor and \(\mathrm{YF}_3:20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) phosphor. a. XRD patterns of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor. b. Upconversion emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) as a function of temperature under 980 nm excitation. c. Comparison of the relative integrated intensity of the upconversion emission of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor, wherein the data of \(\mathrm{SMO:20\%Yb / 1\%Er}\) originate from Figure 3b.  
+
+![](images/Supplementary_Figure_4.jpg)
+
+<center>Supplementary Figure 7| Downshifting emission spectra of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor and \(\mathrm{YF}_3:20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) phosphor. a. and b. Downshifting emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor as a function of temperature under 980 nm excitation. c. Comparison of the relative integrated intensity of the downshifting emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. The enlarged data of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is displayed on the right side. </center>  
+
+## Comment #12:  
+
+Lastly but not the least, this amazing sensitivity occurs at the exact same temperatures where the emission intensities are very low. Thus, at the points of the highest relative sensitivity, the sensing might be just possible, as the uncertainty in measurement will be great. This was discussed in the article: https://onlinelibrary.wiley.com/doi/full/10.1002/adts.202000176. This last point significantly diminishes the firstly claimed importance of the obtained results. I invite the authors to be honest in reporting the sensitivities only on the practically usable range. Also, taking everything in consideration, there is a must that the authors
+
+<--- Page Split --->
+
+report on the temperature resolutions at the whole temperature range, and unfortunately this means reporting on the uncertainties that should be evaluated by at least several temperatures. Uncertainties should be given as the standard deviation of at least 10 consecutive measurements at the same temperature. This temperature resolutions will show the true usability of the sensor, and my assumption is that the temperature resolution curve will be different than the sensitivity curves.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we measured the luminescence lifetimes of our phosphors at four temperature points (303, 313, 333, 363 K). All the PL decays for each temperature point were measured independently for ten times under identical conditions to yield the average value. Temperature resolution discussion via data analysis of the 10 consecutive measurements has been newly added at lines 1- 14 in p. 16 of the revised manuscript. A slight change in sensitivity was obtained by re- fitting the data by adding data pointing. From Figures 5a and 5b, it is obvious that the data become dense at the temperature range from 298 to 373 K, which can improve the data accuracy via decreasing the fitting error. The optimal \(S_{a}\) and \(S_{r}\) were determined to be as high as 53.0 \(\mu \mathrm{sK}^{- 1}\) and 12.3% \(\mathrm{K}^{- 1}\) , respectively. The fitting results have been updated in the revised manuscript. The suggested reference (Adc, Theory Simul. 3, 2000176 (2020)) has been cited as ref. 45 for the discussion of temperature uncertainty.  
+
+![](images/Figure_2.jpg)
+
+<center>Figure 5| Lifetime-based luminescence thermometry in 20% \(\mathrm{Yb^{3 + } / 1\% Er^{3 + }}\) -codoped SMO. a. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO:20%Yb/1%Er at different temperatures. b. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO:20%Yb/1%Er. d. Thermal dependence of temperature uncertainty for SMO:20%Yb/1%Er. </center>  
+
+Moreover, the temperature uncertainty ( \(\delta \mathrm{T}\) ) is an important parameter to assess the performance of a thermometer since it includes not only the relative sensitivity but also the error on the luminescence lifetime ( \(\delta \tau\) ) \(^{55,56}\) . \(\delta \mathrm{T}\) is calculated as follows:  
+
+\[\delta T = \frac{1}{S_{r}}\cdot \frac{\delta\tau}{\tau} \quad (6)\]  
+
+Where \(\delta \tau /\tau\) is the uncertainty in the calculation of \(\tau\) (determined as a standard deviation in ten measurements of \(\tau\) at the same temperature), \(S_{r}\) is the relative
+
+<--- Page Split --->
+
+sensitivity of luminescence thermometer. Temperature- dependent luminescence decay curves of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors by ten measurements at the same temperature are shown in the Supplementary Figure 15 newly added. Thermal dependence of temperature uncertainty for \(\mathrm{SMO:20\%Yb / 1\%Er}\) is presented in Figure 6d. The minimum value of \(\delta \mathrm{T}\) is \(0.11\mathrm{K}\) even at \(623\mathrm{K}\) . As such, although there is a relatively low value of \(\mathbf{S}_{\mathrm{r}}\) at high- temperature range, the obviously longer luminescence lifetime \((\tau)\) can effectively reduce the value of \(\delta \tau /\tau\) , resulting in lower \(\delta T\) . It should be noted that it is possible to keep the temperature uncertainty below a threshold of \(0.33\mathrm{K}\) throughout the whole studied temperature range (298- 623 K). The \(\delta T\) threshold of \(\mathrm{SMO:20\%Yb / 1\%Er}\) for LLT is much lower than that (0.7 K) of cubic- phase \(\mathrm{LiLuF_4:18\%Yb^{3 + } / 2\%Er^{3 + }}\) nanocrystals for the conventional intensity- based thermometry. The \(\delta T\) threshold of \(\mathrm{SMO:20\%Yb / 1\%Er}\) is comparative to that (0.46) of \(\mathrm{LiLuF_4:18\%Yb^{3 + } / 2\%Er^{3 + }@SiO_2}\) (three shells) synthesized with the complicated condition. All these results indicate that \(\mathrm{SMO:Yb / Er}\) phosphor can be explored as a kind of ideal lifetime- based luminescence thermometry with high \(S\) , and low \(\delta T\) , which validate the superiority and applicability of the proposed \(\mathrm{SMO:Yb / Er}\) for luminescence thermometry.  
+
+![](images/Figure_5.jpg)
+
+
+<--- Page Split --->
+![PLACEHOLDER_23_0]
+
+<center>Supplementary Figure 15| Temperature-dependent luminescence decay curves of \(^{4}I_{132}\) excited states of \(\mathrm{Er}^{3 + }\) in \(20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) -codoped SMO at different temperatures from 298 K to 623 K. Each temperature point was measured by ten consecutive measurements. </center>
+
+<--- Page Split --->
+
+## Reply to the Comments of Reviewer #3  
+
+## General comment:  
+
+The authors present a two- dimensional negative thermal phosphor SMO:Yb/Er systematic study. They present a temperature dependent Raman, synchrotron X- ray diffraction and luminescence investigation. The thermally boosted UC and DS luminescence mechanism was investigated. The authors claim that the luminescence lifetime temperature dependency enables ratiometric thermometry with high relative sensitivity; an important topic in current research. This is a comprehensive work, very well written and with insights into the \(\mathrm{Ln}^{3 + }\) - doped phosphors for potential applications.  
+
+Although they address an important topic some issues, that I refer below, need further clarification.  
+
+Response: We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. We have made all the requested changes in the revised manuscript according to the reviewer's comments. Particularly, the mechanism for the NTE behaviour of SMO:Yb \(^{3 + }\) /Er \(^{3 + }\) has been clarified by means of more characterizations. Some inappropriate statement or unclear writing has been rephrased throughout the manuscript. A point- by- point response is noted below.  
+
+## Comment #1:  
+
+Authors should comment, if known, what is the mechanism for the NTE behaviour in this system, whether it is structural, or electronically based for instance?  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, the structural rigid unit mode has been added to explain the mechanism for the NTE behavior of SMO:Yb \(^{3 + }\) /Er \(^{3 + }\) . The NTE behavior in this system is attributed to the structural variability with temperature. To shed more light on the structure of RE \(^{3 + }\) - doped SMO, rigid unit model for RE \(^{3 + }\) - doped SMO has been given as Figure 2e- 2g (Front. Phys., 16, 53302 (2021); Rep. Prog. Phys., 79, 066503 (2016)). With the temperature increased from 420 to 623 K, the distortion of the \(\mathrm{REO}_6\) octahedron and twist of the RE- O- Mo may induce a variation in the rigid unit mode, resulting in a decrease ratio in the distance of RE- Mo along the \(a\) and \(c\) axes by \(0.26\%\) and \(0.10\%\) , as well as an increased ratio in the distance of RE- Mo along \(b\) axis by \(0.02\%\) (Figure 3d). Note that the decreased amplitudes in the distance of RE- Mo along the \(a\) and \(c\) axes are one order larger than that along the \(b\) axis. Such an alteration causes not only shrinkage of lattice but also a local distortion of RE with increasing the temperature. The angle of RE- Mo- RE nearly keeps unchanged at the temperature range of 420- 623 K, which indicates that the structure of \(\mathrm{Sc}_2\mathrm{Mo}_3\mathrm{O}_{12}\) is rigid. The above discussions have been added in lines 17- 24 of p. 5 in the revised manuscript.
+
+<--- Page Split --->
+![PLACEHOLDER_25_0]
+
+<center>Figure 3. e. Rigid unit mode model for \(\mathrm{Sc_2Mo_3O_{12}:20\%Yb^{3 + } / 1\%Er^{3 + }}\) extracted from the unit cell structure. f. Temperature-dependence ratio \(\delta D\) of RE-Mo distances marked in the model \(\begin{array}{r}\delta D = \frac{D_T - D_{420}}{D_{420}}\times 100\% \end{array}\) \(D_{T}\) and \(D_{420}\) stand for RE-Mo distances of the given temperature and \(420\mathrm{K}\) , respectively). g. Temperature-dependence ratio \(\delta \theta\) of RE-Mo-RE angles marked in the model \(\begin{array}{r}\delta \theta = \frac{\theta_T - \theta_{420}}{\theta_{420}}\times 100\% \end{array}\) \(\theta_{T}\) and \(\theta_{420}\) stand for RE-Mo-RE angles of the given temperature and \(420\mathrm{K}\) , respectively). </center>  
+
+## Comment #2:  
+
+What is supposed to drive the transition from PTE to NTE around \(100^{\circ}\mathrm{C}\) ? The transition from PTE to NTE occurs at ca \(100^{\circ}\mathrm{C}\) (figure 2c). How this transition impacts the luminescence properties and how this is compatible with the suggested thermally enhanced photoluminescence mechanism. It is important to clarify this issue. Some Raman modes appear to alter above \(100^{\circ}\mathrm{C}\) , can they be related with the NTE to PTE transition?  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Such a transformation from PTE to NTE can be attributed to the existence of water molecules in the \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors as revealed by the thermogravimetry (TG) analysis, in situ temperature- dependent Fourier transform infrared (FTIR) spectroscopy (Supplementary Figure 3 newly added), and temperature- dependent Raman spectra (Figure 2h).  
+
+As shown in Supplementary Figure 3a, a weight loss of about \(1\%\) for \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) was detected by heating from 298 to \(420\mathrm{K}\) . Generally, \(\mathrm{A_2Mo_3O_{12}}\) compounds with large \(\mathrm{A^{3 + }}\) cation size are highly hygroscopic and easily hydrated at ambient conditions. Therefore, \(\mathrm{Sc_2Mo_3O_{12}}\) compound with small \(\mathrm{Sc^{3 + }}\) (0.0745 nm, \(\mathrm{CN} = 6\) ) is nonhygroscopic. For \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors with the substitution of \(\mathrm{Sc^{3 + }}\) by the big \(\mathrm{Yb^{3 + }}\) (0.0868 nm) and \(\mathrm{Er^{3 + }}\) (0.089 nm, \(\mathrm{CN} = 6\) ), water molecules may be easily absorbed into the relatively large microchannels (Ref. 23, Effect of water species on the phonon modes in orthorhombic \(\mathrm{Y_2(MoO_4)_3}\) revealed by Raman spectroscopy. J. Phys. Chem. C, 112, 6577- 6581 (2008)). Accordingly, it
+
+<--- Page Split --->
+
+can be deduced that weight loss of about \(1\%\) for \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) may result from the removal of water molecules.  
+
+In the FTIR spectra, the typical absorbance peak at \(3400–3500 \mathrm{cm}^{- 1}\) , corresponding to the asymmetric vibration of - OH was markedly suppressed with the temperature from 298 to \(420 \mathrm{K}\) (Supplementary Figure 3b), which further verified the existence and removal of water molecules. As such, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors with water molecules in the microchannels exhibit PTE behavior at low temperature (Ref.16, Negative thermal expansion: mechanisms and materials. Front. Phys., 16, 53302 (2021)). With the temperature rising from 298 to \(353 \mathrm{K}\) , the water molecules remain in the microchannels. With an increase in temperature to around \(353 \mathrm{K}\) , water molecules begin to escape from the microchannels. As such, the \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors without water molecules in the microchannels exhibit intrinsic NTE character. When the temperature reaches \(420 \mathrm{K}\) , water molecules are completely removed and thus \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor may completely recover the NTE property.  
+
+![PLACEHOLDER_26_0]
+
+<center>Supplementary Figure 3| Water molecules analysis. a. Thermogravimetry curves of the \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. b. Temperature-dependent Infrared spectra of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. </center>  
+
+![PLACEHOLDER_26_1]
+
+<center>Supplementary Figure 4 c. Temperature-dependent changes of the unit cell volumes. </center>
+
+<--- Page Split --->
+![PLACEHOLDER_27_0]
+
+<center>Figure 2 h. Temperature-dependent in situ Raman spectra within the temperature range from 298 to 773 K. </center>  
+
+For the in situ temperature- dependent Raman spectra (Figure 2h), the Raman peaks of 330, 836, and \(945~\mathrm{cm^{- 1}}\) are characteristic of the hydrated orthorhombic structure, indicative of water species residing in the microchannels of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) . As the temperature increases from 298 to 348 K, the intensity and position of these peaks remain unchanged, which reveals that the water molecules remain in the microchannels. As the temperature increases from 348 to 423 K, these peaks become weaker. Above 423 K, the peak at \(945~\mathrm{cm^{- 1}}\) vanished, which reveals the complete removal of water molecules. These results are in agreement with the obtained results of the TG, FTIR/spectra, as well as in situ temperature- dependent SXRD patterns (Supplementary Figures 3 and 4). Thus, it can be deduced that the temperature transition points of the onset and complete removal of water molecules from microchannels are \(\sim 348\) and \(\sim 423\) K, respectively. The above discussions have been added in lines 25- 29 of p. 5 in the revised manuscript.  
+
+Based on these results, the photoluminescence mechanisms of \(\mathrm{SMO:Yb / Er}\) phosphors at different temperatures have been updated. For the temperature range of 298- 353 K, the \(\mathrm{SMO:Yb / Er}\) phosphors exhibit PTE due to the existence of water molecules in the microchannels. At higher temperatures above 353 K, the \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibited thermally enhanced photoluminescence behavior due to the typical NTE properties of \(\mathrm{A_2Mo_3O_{12}}\) compounds (Supplementary Figure 4c). Considering our topic is thermally boosted upconversion and downshifting luminescence in \(\mathrm{SMO:Yb / Er}\) with two- dimensional negative thermal expansion, we have displayed the temperature range (353- 773) of the thermally enhanced photoluminescence mechanism in the energy level diagram (Figure 5c). Please see also lines 2- 5 of p. 12 in the revised manuscript and Supplementary Figure 3 newly added in the Supporting Information.
+
+<--- Page Split --->
+![PLACEHOLDER_28_0]
+
+<center>Figure 5| Thermally enhanced photoluminescence mechanism a. Temperature-dependent luminescence lifetime of \(^{2}\mathrm{F}_{5 / 2}\) excited state of \(\mathrm{Yb^{3 + }}\) in \(\mathrm{Yb^{3 + } / Er^{3 + }}\) -codoped and \(\mathrm{Yb^{3 + }}\) -doped SMO, respectively. Temperature-dependent energy transfer efficiency of \(\mathrm{Yb^{3 + }}\) -to-Er \(^{3 + }\) . b. Temperature-dependent lifetimes of \(^{2}\mathrm{H}_{11 / 2}\) (522 nm) and \(^{4}\mathrm{I}_{13 / 2}\) (1538 nm) excited states of \(\mathrm{Er^{3 + }}\) in SMO: \(20\% \mathrm{Yb} / 1\% \mathrm{Er}\) , respectively. c. Energy level diagram of green UC and NIR DS emission showing the proposed temperature dependence of electronic transition and energy-transfer processes in SMO: \(\mathrm{Yb / Er}\) with two-dimensional negative thermal expansion. </center>  
+
+## Comment #3:  
+
+From lines 182- 183: "Moreover, the distance of \(\mathrm{Yb^{3 + } - Er^{3 + }}\) becomes shorter with an increase of the temperature. The ET processes between sensitizer \((\mathrm{Yb^{3 + }})\) to activator \((\mathrm{Er^{3 + }})\) are usually considered to occur through dipolar- dipolar interactions, whose ET efficiency is proportional to \(r^{- 6}\) ( \(r\) is the average donor- acceptor distance). The result indicates that the distances of \(\mathrm{Yb^{3 + } - Er^{3 + }}\) ions (13.975 \(\approx\) , at 25 \(^\circ \mathrm{C}\) ) decreased at higher temperature, which may benefit the improvement of the ET efficiency between \(\mathrm{Yb^{3 + } - Er^{3 + }}\) ."  
+
+And line 236 "Based on the above analysis, the thermally enhanced UC emission with the temperature from 25 to \(200^{\circ}\mathrm{C}\) is mainly governed by the ET from \(\mathrm{Yb^{3 + }}\) to \(\mathrm{Er^{3 + }}\) " Authors should clarify the nature for thermally enhanced UC emission within the temperature range from 25 to \(200^{\circ}\mathrm{C}\) , if this effect is mainly governed by the ET from \(\mathrm{Yb^{3 + }}\) to \(\mathrm{Er^{3 + }}\) , how it relates with the PTE of the system shown in Figure 2 observed between \(25^{\circ}\mathrm{C}\) and \(100^{\circ}\mathrm{C}\) .
+
+<--- Page Split --->
+
+## Response:  
+
+Many thanks for pointing out this problem. As indicated in the above analyses, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibit PTE behavior at the temperature below 353 K due to the existence of water molecules in the microchannels. When the temperature increases up to 353 K, water molecules begin to escape from the microchannels. As such, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibit thermally enhanced photoluminescence behavior due to the typical NTE properties of \(\mathrm{A_2Mo_3O_{12}}\) compounds. Please see also lines 22- 29 of p. 4 in the revised manuscript and Supplementary Figure 3 newly added in the Supporting Information.  
+
+## Comment #4:  
+
+From lines 46- 49: "Most of such abnormal thermally enhanced UC luminescence is observed in \(\mathrm{Ln3 + }\) - doped inorganic materials with three- dimensional negative- thermal expansion characteristics (NTE)9, 10, where all the three cell parameters of the doped crystals shrink at elevated temperature."  
+
+And below, lines 58- 59 "with a unique two- dimensional NTE coefficient \((\alpha_{a} = - 8.62\times 10^{- 6} / \mathrm{K}, \alpha_{b} = 4.25\times 10^{- 6} / \mathrm{K}, \alpha_{c} = - 6.35\times 10^{- 6} / \mathrm{K})\) "  
+
+In the context of this work, is there any effective advantage in two- dimensional negative- thermal expansion materials when compared with the three- dimensional ones? The authors could also mention recent works with uniaxial NTE.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. The advantage of two- dimensional (2D) negative- thermal expansion materials relative to the three- dimensional (3D) ones was not explicitly elucidated perhaps because of our unclear writing in the manuscript. For the previously reported \(\mathrm{RE^{3 + }}\) doped phosphors, the existence of 3D compression at elevated temperature can promote the energy transfer between sensitizers and activators. Meanwhile, such a 3D compression may also benefit the dissipation of the excitation energy in all directions of crystal sublattice to the lattice/surface defects, which deteriorates the luminescent emission of \(\mathrm{RE^{3 + }}\) ions. In this work, we propose a new class of NTE phosphors based on \(\mathrm{SMO:Yb / Er}\) exhibiting a unique two- dimensional NTE coefficient. With increasing the temperature, the \(\mathrm{SMO:Yb / Er}\) phosphors experience an anisotropic two- dimensional shrinkage along the \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE^{3 + }}\) ions increases along the \(b\) axis. As such, the detrimental energy migration of \(\mathrm{RE^{3 + }}\) dopants can be effectively confined within the two- dimensional layered structures (e.g., (020) lattice plane) at elevated temperatures. Please see also lines 14- 16 of p. 5 in the revised manuscript. Recent works with uniaxial NTE behavior have been cited as Refs. 12- 15 (Adv. Sci. 2016, 3, 1600108; Phys. Rev. B 2020, 101, 104305; Inorg. Chem., 56, 15101- 15109 (2017); Phys. Rev. B, 96, 134113- 134120 (2017)).
+
+<--- Page Split --->
+
+## Comment #5:  
+
+Can the authors explain in what way the Energy migration between \(Ln^{3 + }\) dopants is confined to two dimensions?  
+
+## Response:  
+
+Thank you for pointing out this important issue. For \(\mathrm{RE}^{3 + }\) - doped inorganic phosphors, the long- distance energy migration between \(\mathrm{RE}^{3 + }\) dopants depends on the distance between lattice positions occupied by \(\mathrm{RE}^{3 + }\) ions. As such, it is possible to minimize the depletion of excitation energy by employing a crystal lattice disfavoring the long- distance energy migration (Nat. Mater. 2014, 13, 157- 162). In our work, the SMO:Yb/Er phosphors experience an anisotropic two- dimensional shrinkage along the \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE}^{3 + }\) ions increases along with the \(b\) axis. Under such circumstances, the migration of excitation energy between \(\mathrm{RE}^{3 + }\) dopants may be effectively minimized long the \(b\) axis. Thus, more energy migration between \(\mathrm{RE}^{3 + }\) dopants is confined to the two- dimensional layered structures (e.g., (020) lattice plane) at elevated temperatures. Please see also lines 21- 23 of p. 2 in the revised manuscript.  
+
+## Comment #6:  
+
+From lines 231- 233. The promoted Ar and inhibited \(A_{nr}\) in the \(Ln^{3 + }\) dopant indicate that lattice distortion enhances a crystal field with odd parity and modifies local symmetry of activator ions via the NTE effect. The authors mention the existence of lattice distortion and a change in the local symmetry of activator ions, via the NTE effect. Can these local distortions be substantiated via the temperature- dependent Raman data?  
+
+## Response:  
+
+Many thanks for this valuable suggestion. Raman spectroscopy is a valuable tool to study the phonon modes of NTE materials and evaluate the change of local structure. In situ temperature- dependent Raman spectra are shown in Figure 2h. The Raman peaks of 330, 836, and \(945~\mathrm{cm^{- 1}}\) are characteristic of the hydrated orthorhombic structure, indicative of water species residing in the microchannels of SMO: \(20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) . As the temperature increases from 298 to 348 K, the intensity and position of these peaks nearly remain unchanged, which reveals that the water molecules were not removed from the microchannels. As the temperature increases from 348 to \(423~\mathrm{K}\) , these peaks become weaker. Above \(423~\mathrm{K}\) , the peak at \(945~\mathrm{cm^{- 1}}\) vanished, which demonstrated that water molecules are completely removed. These results are in agreement with the obtained results of the TG, FTIR spectra, as well as in situ temperature- dependent SXRD patterns (Supplementary Figures 3 and 4). Thus, it can be deduced that the temperature transition points of the onset and complete removal of water molecules from microchannels are \(\sim 348\) and \(\sim 423~\mathrm{K}\) , respectively. In addition, as the temperature increases, a Raman peak with a frequency of \(341~\mathrm{cm^{- 1}}\)
+
+<--- Page Split --->
+
+\(\mathrm{(v_4)}\) exhibits a blue shift (Figure 2h), suggesting that this mode is the origin of NTE in \(\mathrm{SMO:Yb / Er}\) . The median frequency at \(510~\mathrm{cm^{- 1}}\) and \(579~\mathrm{cm^{- 1}}\) is induced by the disorder of \(\mathrm{MoO_4}\) tetrahedra because of the incorporation of the \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) into the lattice, which will affect the local structure of the doped \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) ions. Note that the temperature coefficients of high phonon frequencies (511, 579, 813, \(979~\mathrm{cm^{- 1}}\) ) are negative above \(348~\mathrm{K}\) (Figures 2i and 2j), verifying strong anharmonic stretching/bending of \(\mathrm{MoO_4}\) tetrahedra. These results indicate that high phonon frequencies contribute also to the NTE behavior of the sample. Moreover, the reduction of maximum phonon energy \((979~\mathrm{cm^{- 1}})\) may benefit the luminescence of \(\mathrm{RE^{3 + }}\) at elevated temperatures due to the suppressed nonradiative relaxation. Please see also lines 1- 17 of p. 7 in the revised manuscript.  
+
+## Comment #7  
+
+Can this effect be witnessed by the other techniques that would perceive the negative thermal expansion at the local scale? The authors could comment on that.  
+
+## Response:  
+
+The negative thermal expansion at the local scale has also been witnessed by the other techniques like atomically resolved scanning tunnel microscope (STM) topographic image. We have cited the corresponding reference about the characterization of negative thermal expansion as Ref.13 (Phys. Rev. B, 101, 104305 (2020)) in the revised manuscript.  
+
+## Comment #8:  
+
+Along the manuscript:  
+
+What is the degree of Sc substitution in the samples that were measured in the presented results Figure 2,3,4,5? This is only specified in the labels of the Figures presented in the Supplementary material.  
+
+## Response:  
+
+Many thanks for this valuable suggestion. We have provided the degree of Sc substitution (Sc substituted by \(20\%\) Yb and \(1\%\) Er) in the samples for Figures 2- 5 in the revised manuscript.  
+
+## Comment #9:  
+
+line 18 "decrease at the higher temperature" author should substitute "decrease for high temperatures"  
+
+## Response:  
+
+"decrease at the higher temperature" has been revised as "decrease for high
+
+<--- Page Split --->
+
+temperatures" in the revised manuscript. Please see also line 20 of p. 1 in the revised manuscript.  
+
+## Comment #10:  
+
+Comment #10:line 58 and 103 Where is the abbreviation "NET" defined? Do the authors mean "NET"?  
+
+## Response:  
+
+Response:Thank you for pointing out this typo. The abbreviation "NET" has been revised as "NET" in the revised manuscript.  
+
+## Comment #11:  
+
+Comment #11:Line 121- 122: What is the value of Xc used for the calculations?  
+
+## Response:  
+
+Response:According to the second reviewer's suggestion, the ionic distance was updated, which was determined by the crystallographic information file (CIF) of Rietveld refinement on the basis of the SXRD patterns. Therefore, Equation 1 used for the calculation has been deleted.  
+
+## Comment #12:  
+
+Comment #12:Line 133 - 135 "The observed mode with a frequency of \(341 \mathrm{cm}^{- 1}\) corresponds to the bending mode of MoO4 tetrahedra. When the temperature increases, this Raman peak exhibits a blue shift (Figure 2f). The median frequency B1g mode of SMO is observed at \(511 \mathrm{cm}^{- 1}\) . Is this a typo? Figure 2f is not from the \(341 \mathrm{cm}^{- 1}\) mode. The sentence should be improved.  
+
+## Response:  
+
+Response:Thank you for pointing out this typo. "this Raman peak exhibits a blue shift (Figure 2f)" has been revised as "this Raman peak exhibits a blue shift (Figure 2h)". Please see also line 10 of p. 7 in the revised manuscript.  
+
+## Comment #13:  
+
+Comment #13:The authors refer the studied systems as "the structures of the SMO:Yb/Er samples with different dopant concentrations", doping generally refers to a small concentrations, here the degree of ionic Sc substitution by Yb and Er cations reaches the high value of \(25\%\) or \(8\%\) . The term chemical substitution would be more appropriated.
+
+<--- Page Split --->
+
+## Response:  
+
+Thanks. "the structures of the SMO:Yb/Er samples with different dopant concentrations" has been revised as "the structures of the SMO:Yb/Er samples with different substituted concentrations".
+
+<--- Page Split --->
+
+REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors responded correctly to my comments. The work does not contain any factual errors. In my opinion, it can be accepted for publication in its current form  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors have adequately addressed all of the reviewer's comments, thus I recommend this article for publication.  
+
+Reviewer #3 (Remarks to the Author):  
+
+The paper is much improved, and it addresses the issues that concerned me. In my opinion is ready for publication.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d62429608af455dfb3898d1ec1dbf35ffe750b30c81658917f22585b63bf7d8a/supplementary_0_Peer Review File__d62429608af455dfb3898d1ec1dbf35ffe750b30c81658917f22585b63bf7d8a_det.mmd

@@ -0,0 +1,829 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 507, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[70, 110, 362, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[73, 154, 912, 258]]<|/det|>
+Thermally boosted upconversion and downshifting luminescence in Sc2(MoO4)3:Yb/Er with twodimensional negative thermal expansion  
+
+<|ref|>image<|/ref|><|det|>[[273, 350, 720, 655]]<|/det|>  
+
+<|ref|>image<|/ref|><|det|>[[57, 733, 240, 782]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 733, 912, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to  
+
+<|ref|>text<|/ref|><|det|>[[57, 784, 936, 924]]<|/det|>
+the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 150, 404, 165]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 205, 876, 313]]<|/det|>
+The work concerns a very interesting and popular topic - luminescence thermometry. The presented strategy using a material with a negative temperature expansion coefficient as a material with a positivistic thermal luminescence coefficient seems to be a very promising approach offering very high relative sensitivities. Therefore, I believe that the work has an important aspect of novelty. I am not sure that it will be of interest to a wide readership of this journal. Nevertheless, some issues should be clarified before the work can be considered for publication in Nature Communications:  
+
+<|ref|>text<|/ref|><|det|>[[116, 355, 812, 373]]<|/det|>
+1. Caption of Figure 6 is misleading. Actually, the authors do not show any application there.  
+
+<|ref|>text<|/ref|><|det|>[[118, 386, 810, 420]]<|/det|>
+2. Authors should, in accordance with current standards, use an uppercase "K" to designate temperature units  
+
+<|ref|>text<|/ref|><|det|>[[118, 434, 848, 490]]<|/det|>
+3. Although the summary of results presented in Figure 6d is very informative, I encourage the authors to contrast the results presented with respect to recently published work for which very high relative sensitivities were obtained such as 10.1016/j.cej.2021.131165  
+
+<|ref|>text<|/ref|><|det|>[[118, 502, 866, 556]]<|/det|>
+4. The interionic distance contraction is one of the very specific mechanisms responsible for the thermal enhancement of the luminescence. There are several other processes that may lead to the similar thermal behaviors like i.e. change of the local ion symmetry etc.  
+
+<|ref|>text<|/ref|><|det|>[[118, 569, 874, 623]]<|/det|>
+5. The purpose of the data shown in Figures 1 a, b, c is to describe three types of temperature effect trends on luminescence intensity. Therefore, I would suggest removing the "diagrams" placed under the graphs to maintain the universality of the description presented.  
+
+<|ref|>text<|/ref|><|det|>[[118, 636, 840, 671]]<|/det|>
+6. Authors should use consistent temperature units throughout the paper (e.g. Fig.1 Celsius and Figure 6 Kelvins).  
+
+<|ref|>text<|/ref|><|det|>[[118, 684, 867, 719]]<|/det|>
+7. Please remove the information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and b  
+
+<|ref|>text<|/ref|><|det|>[[118, 762, 877, 835]]<|/det|>
+8. The increase in emission intensity with increasing temperature may be due to an increase in the absorption cross section associated with the spectral broadening of the absorption band. This occurs when the excitation line etches into the slope of the absorption band. Therefore, the authors should present the excitation spectra of Yb3+ ions measured as a function of temperature
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 878, 157]]<|/det|>
+9. The magnitude of the thermal contraction of the crystalographic unit cell depends on the difference in the ionic radii between dopants and replaces ions and on the concentration of dopants. Therefore, I would like the authors to comment if it is possible to modulate the thermometric parameters of this type of thermometer by changing the dopant concentration?  
+
+<|ref|>text<|/ref|><|det|>[[118, 199, 878, 234]]<|/det|>
+10. Table S1- the comparing the thermometric properties of the luminescent thermometry operating in different readout modes is unjustified and unreasonable.  
+
+<|ref|>text<|/ref|><|det|>[[119, 370, 404, 385]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 429, 866, 538]]<|/det|>
+The authors of "Thermally boosted upconversion and downshifting luminescence in Sc2(MoO4)3:Yb/Er with two- dimensional negative thermal expansion" presented a very interesting research of the famous upconverting pair codoped in more than astounding host matrix. However, for the following reasons I recommend major revision and reinspection, as there are numerous problems with novelty, significance and support for the conclusions. Briefly these problems can be summed as:  
+
+<|ref|>text<|/ref|><|det|>[[118, 551, 846, 586]]<|/det|>
+- The same host matrix was investigated for thermometry by the Yb/Ho pair and LIR with similar conclusions. This just slightly diminishes the stated novelty.  
+
+<|ref|>text<|/ref|><|det|>[[118, 600, 827, 634]]<|/det|>
+- The discussion about inter-atomic differences is possibly erroneous, and need either serious revision or removal.  
+
+<|ref|>text<|/ref|><|det|>[[118, 648, 870, 683]]<|/det|>
+- The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.  
+
+<|ref|>text<|/ref|><|det|>[[118, 697, 850, 769]]<|/det|>
+- This high relative sensitivity occurs at the points where the signal is borderline detectable. This means that the temperature resolutions will be low at those points, i.e. the significance of the research is not as high as initially suggested. The authors need to include temperature resolution discussion in the most correct manner, and accordingly adjust their conclusions.  
+
+<|ref|>text<|/ref|><|det|>[[118, 783, 234, 798]]<|/det|>
+Issues in detail:  
+
+<|ref|>text<|/ref|><|det|>[[118, 812, 875, 902]]<|/det|>
+In the first reading the novelty of this article seams immense. However, the exactly the same host, co- doped with an upconverting pair of Yb3+/Ho3+ has been already explored for luminescence thermometry in Ref: https://pubs.acs.org/doi/10.1021/acs.jpclett.0c00628. Although the authors did reference this work, their contribution and similarity needs to be more stressed. That being said, the differences in employing the more important upconverting pair, Yb3+/Er3+, and lifetime instead
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 868, 119]]<|/det|>
+of luminescence intensity ratio method, should be stated as the novelty of this work. In my opinion, this subtle differences are important and enough novelty for the publication of this work.  
+
+<|ref|>text<|/ref|><|det|>[[117, 132, 870, 278]]<|/det|>
+For the equation 1 the referenced is the famous short work by Blasse (Ref. 22). However, in that reference the Blasse is not reporting on a ways of calculation of the average distance, but on the critical distance, i.e. when the concentration quenching starts to dominate. The c in the subscript of xc does not abbreviate concentration, but the word critical. Up to my knowledge, this equation cannot be applied for calculation of distance between interacting ions if that distance is not critical distance, and even more, of the ions of different types. Unless the authors can provide another, a very strong reference to support their claims of usage of equation 1, I suggest removing the Figure 2d, Equation 1, and any corresponding discussion.  
+
+<|ref|>text<|/ref|><|det|>[[117, 291, 878, 456]]<|/det|>
+There is another issue with concentrations and distances. In figure 2 no concentrations of dopants are mentioned and yet there is a graph of their distances. To additionally confirm the assumption that the author's analysis of interatomic distances is incorrect is the un- logical trends in presented distances. Although both the codopants substitute the same ions in the SMO matrix, the authors claim that Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase. All three claims cannot be possible, if I understood well, if not, then the paragraph needs to be revised for clarity. As for the red emission of Er3+, the authors should see the book by Kaminskii, Crystalline lasers, where the pathways for population of the 4F9/2 are explained, and they are related to the Yb3+ ion back- transfer more than the Er3+ cross- relaxation.  
+
+<|ref|>text<|/ref|><|det|>[[118, 468, 877, 576]]<|/det|>
+Lifetime method for thermometry is erroneously described as ratiometric multiple times in the abstract and throughout the article. The definition of ratiometric fluorescence is where intensities of at least two bands are measured and compared. Thus, the word ratio in ratiometric. FIR or LIR methods, of which the authors are certainly aware of, are the ratiometric luminescence thermometry methods, while lifetime, although also self- referenced, is not. This needs to be corrected in the caption of Table S1 as well.  
+
+<|ref|>text<|/ref|><|det|>[[118, 590, 878, 644]]<|/det|>
+Throughout the article there needs to be more linear story of using concentrations and even hosts. It is hard to follow which concentrations are being used, and when it is compared to the other host. For each spectra the precise concentrations of codopants should be given.  
+
+<|ref|>text<|/ref|><|det|>[[118, 657, 875, 729]]<|/det|>
+The authors discuss the 4111/2 emission of Er3+, however in the spectra and in the text the 1538 nm emission is mentioned. Did you mean the emission from 4113/2 instead, as the emission from 4111/2 to the ground level is at about 1000 nm? See for example Figure S9b: caption says 4111/2 and Eem = 1538 nm, the same in Figure 5.  
+
+<|ref|>text<|/ref|><|det|>[[118, 742, 874, 814]]<|/det|>
+For other journals of lower esteem than the Nature series, the comparison with other hosts would not be necessary. However, I feel that in this host the emission intensity is much lower than in other famous upconverting hosts, for example YF3 or NaYF4. I would like to see the comparison in either intensities or quantum yields between SMO and one of these prominent hosts.  
+
+<|ref|>text<|/ref|><|det|>[[118, 828, 858, 900]]<|/det|>
+Lastly but not the least, this amazing sensitivity occurs at the exact same temperatures where the emission intensities are very low. Thus, at the points of the highest relative sensitivity, the sensing might be just possible, as the uncertainty in measurement will be great. This was discussed in the article: https://onlinelibrary.wiley.com/doi/full/10.1002/adts.202000176 . This last point
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 880, 230]]<|/det|>
+significantly diminishes the firstly claimed importance of the obtained results. I invite the authors to be honest in reporting the sensitivities only on the practically usable range. Also, taking everything in consideration, there is a must that the authors report on the temperature resolutions at the whole temperature range, and unfortunately this means reporting on the uncertainties that should be evaluated by at least several temperatures. Uncertainties should be given as the standard deviation of at least 10 consecutive measurements at the same temperature. This temperature resolutions will show the true usability of the sensor, and my assumption is that the temperature resolution curve will be different than the sensitivity curves.  
+
+<|ref|>text<|/ref|><|det|>[[119, 334, 404, 350]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 393, 847, 485]]<|/det|>
+The authors present a two- dimensional negative thermal phosphor SMO:Yb/Er systematic study. They present a temperature dependent Raman, synchrotron X- ray diffraction and luminescence investigation. The thermally boosted UC and DS luminescence mechanism was investigated. The authors claim that the luminescence lifetime temperature dependency enables ratiometric thermometry with high relative sensitivity; an important topic in current research.  
+
+<|ref|>text<|/ref|><|det|>[[118, 528, 875, 564]]<|/det|>
+This is a comprehensive work, very well- written and with insights into the \(\mathsf{Ln3 + }\) - doped phosphors for potential applications.  
+
+<|ref|>text<|/ref|><|det|>[[118, 576, 870, 593]]<|/det|>
+Although they address an important topic some issues, that I refer below, need further clarification.  
+
+<|ref|>text<|/ref|><|det|>[[118, 636, 850, 672]]<|/det|>
+Authors should comment, if known, what is the mechanism for the NTE behaviour in this system, whether it is structural, or electronically based for instance?  
+
+<|ref|>text<|/ref|><|det|>[[118, 685, 666, 702]]<|/det|>
+What is supposed to drive the transition from PTE to NTE around \(100^{\circ}C\)  
+
+<|ref|>text<|/ref|><|det|>[[118, 744, 821, 800]]<|/det|>
+The transition from PTE to NTE occurs at ca \(100^{\circ}C\) (figure 2c). How this transition impacts the luminescence properties and how this is compatible with the suggested thermally enhanced photoluminescence mechanism. It is important to clarify this issue.  
+
+<|ref|>text<|/ref|><|det|>[[115, 841, 877, 859]]<|/det|>
+Some Raman modes appear to alter above \(100K\) , can they be related with the NTE to PTE transition?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 868, 192]]<|/det|>
+From lines 182- 183: "Moreover, the distance of \(\mathsf{Yb3 + - Er3 + }\) becomes shorter with an increase of the temperature. The ET processes between sensitizer \((\mathsf{Yb3 + })\) to activator \((\mathsf{Er3 + })\) are usually considered to occur through dipolar- dipolar interactions, whose ET efficiency is proportional to r- 6 (r is the average donor- acceptor distance). The result indicates that the distances of \(\mathsf{Yb3 + - Er3 + }\) ions (13.975 \(\approx\) , at \(25^{\circ}C\) ) decreased at higher temperature, which may benefit the improvement of the ET efficiency between \(\mathsf{Yb3 + - Er3 + }\) ."  
+
+<|ref|>text<|/ref|><|det|>[[118, 206, 790, 241]]<|/det|>
+And line 236 "Based on the above analysis, the thermally enhanced UC emission with the temperature from 25 to \(200^{\circ}C\) is mainly governed by the ET from \(\mathsf{Yb3 + }\) to \(\mathsf{Er3 + }\) "  
+
+<|ref|>text<|/ref|><|det|>[[118, 255, 873, 308]]<|/det|>
+Authors should clarify the nature for thermally enhanced UC emission within the temperature range from 25 to \(200^{\circ}C\) , if this effect is mainly governed by the ET from \(\mathsf{Yb3 + }\) to \(\mathsf{Er3 + }\) , how it relates with the PTE of the system shown in Fig. 2 observed between \(250 - C\) and \(1000C\) .  
+
+<|ref|>text<|/ref|><|det|>[[118, 351, 867, 423]]<|/det|>
+From lines 46- 49: "Most of such abnormal thermally enhanced UC luminescence is observed in \(\mathsf{Ln3 + }\) - doped inorganic materials with three- dimensional negative- thermal expansion characteristics (NTE)9, 10, where all the three cell parameters of the doped crystals shrink at elevated temperature."  
+
+<|ref|>text<|/ref|><|det|>[[118, 437, 797, 472]]<|/det|>
+And below, lines 58- 59 "with a unique two- dimensional NET coefficient \((\alpha a = - 8.62\times 10 - 6 / K,\) \(\alpha b = 4.25\times 10 - 6 / K,\) \(\alpha c = - 6.35\times 10 - 6 / K)\) "  
+
+<|ref|>text<|/ref|><|det|>[[118, 485, 852, 538]]<|/det|>
+In the context of this work, is there any effective advantage in two- dimensional negative- thermal expansion materials when compared with the three- dimensional ones? The authors could also mention recent works with uniaxial NTE.  
+
+<|ref|>text<|/ref|><|det|>[[118, 583, 870, 617]]<|/det|>
+Can the authors explain in what way the Energy migration between \(\mathsf{Ln3 + }\) dopants is confined to two dimensions?  
+
+<|ref|>text<|/ref|><|det|>[[118, 661, 866, 714]]<|/det|>
+From lines 231- 233. The promoted Ar and inhibited Anr in the \(\mathsf{Ln3 + }\) dopant indicate that lattice distortion enhances a crystal field with odd parity and modifies local symmetry of activator ions via the NTE effect.  
+
+<|ref|>text<|/ref|><|det|>[[118, 728, 860, 781]]<|/det|>
+The authors mention the existence of lattice distortion and a change in the local symmetry of activator ions via the NTE effect. Can these local distortions be substantiated via the temperature- dependent Raman data?  
+
+<|ref|>text<|/ref|><|det|>[[118, 795, 833, 830]]<|/det|>
+Can this effect be witnessed by the other techniques that would perceive the negative thermal expansion at the local scale? The authors could comment on that.  
+
+<|ref|>text<|/ref|><|det|>[[118, 875, 286, 890]]<|/det|>
+Along the manuscript:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 114, 852, 167]]<|/det|>
+What is the degree of Sc substitution in the samples that were measured in the presented results Figure 2,3,4,5? This is only specified in the labels of the Figures presented in the Supplementary material.  
+
+<|ref|>text<|/ref|><|det|>[[118, 211, 790, 245]]<|/det|>
+line 18 "decrease at the higher temperature" author should substitute "decrease for high temperatures"  
+
+<|ref|>text<|/ref|><|det|>[[118, 260, 775, 276]]<|/det|>
+line 58 and 103 Where is the abbreviation "NET" defined? Do the authors mean "NTE"?  
+
+<|ref|>text<|/ref|><|det|>[[118, 320, 596, 336]]<|/det|>
+Line 121- 122: What is the value of Xc used for the calculations?  
+
+<|ref|>text<|/ref|><|det|>[[118, 380, 878, 434]]<|/det|>
+Line 133 - 135 "The observed mode with a frequency of 341 cm- 1 corresponds to the bending mode of MoO4 tetrahedra. When the temperature increases, this Raman peak exhibits a blue shift (Fig. 2f). The median frequency B1g mode of SMO is observed at 511 cm- 1"  
+
+<|ref|>text<|/ref|><|det|>[[118, 448, 795, 465]]<|/det|>
+Is this a typo? Figure 2f is not from the 341cm- 1 mode. The sentence should be improved.  
+
+<|ref|>text<|/ref|><|det|>[[118, 507, 866, 579]]<|/det|>
+The authors refer the studied systems as "the structures of the SMO:Yb/Er samples with different dopant concentrations", doping generally refers to a small concentrations, here the degree of ionic Sc substitution by Yb and Er cations reaches the high value of \(25\%\) or \(8\%\) . The term chemical substitution would be more appropriated.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[329, 95, 666, 112]]<|/det|>
+## Reply to the Comments of Reviewer #1  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 123, 310, 139]]<|/det|>
+## General comment:  
+
+<|ref|>text<|/ref|><|det|>[[148, 150, 850, 297]]<|/det|>
+The work concerns a very interesting and popular topic - luminescence thermometry. The presented strategy using a material with a negative temperature expansion coefficient as a material with a positivistic thermal luminescence coefficient seems to be a very promising approach offering very high relative sensitivities. Therefore, I believe that the work has an important aspect of novelty. I am not sure that it will be of interest to a wide readership of this journal. Nevertheless, some issues should be clarified before the work can be considered for publication in Nature Communications:  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 309, 237, 325]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 336, 850, 390]]<|/det|>
+We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. A point- by- point response is noted below.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 428, 267, 445]]<|/det|>
+## Comment #1:  
+
+<|ref|>text<|/ref|><|det|>[[148, 455, 850, 490]]<|/det|>
+Caption of Figure 6 is misleading. Actually, the authors do not show any application there.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 503, 237, 519]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 530, 850, 584]]<|/det|>
+Thank you for your kind reminding. Caption of Figure 6 has been corrected as "Lifetime- based luminescence thermometry" in the revised manuscript. Please see also line 1 of p. 15 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 623, 267, 639]]<|/det|>
+## Comment #2:  
+
+<|ref|>text<|/ref|><|det|>[[148, 650, 850, 685]]<|/det|>
+Authors should, in accordance with current standards, use an uppercase "K" to designate temperature units  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 697, 237, 713]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 724, 801, 742]]<|/det|>
+Temperature units have been revised as uppercase "K" in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 780, 267, 797]]<|/det|>
+## Comment #3:  
+
+<|ref|>text<|/ref|><|det|>[[148, 808, 850, 880]]<|/det|>
+Although the summary of results presented in Figure 6d is very informative, I encourage the authors to contrast the results presented with respect to recently published work for which very high relative sensitivities were obtained such as 10.1016/j.cej.2021.131165.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 892, 237, 908]]<|/det|>
+## Response:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 84, 850, 177]]<|/det|>
+The suggested reference has been cited as Ref. 54, which was compared with our results and discussed in the revised manuscript. Specifically, the optimal \(S_{r}\) value of \(\mathrm{SMO:20\%Yb / 1\%Er}\) \((12.3\% \mathrm{K}^{- 1})\) is higher than that \((8.83\% \mathrm{K}^{- 1})\) of \(\mathrm{SrTiO_3:Tb^{3 + }}\) nanocrystals in the suggested reference (Chem. Eng. J. 2022, 428, 131165). Please see also line 27 of p.14 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 215, 267, 231]]<|/det|>
+## Comment #4:  
+
+<|ref|>text<|/ref|><|det|>[[148, 242, 850, 315]]<|/det|>
+The interionic distance contraction is one of the very specific mechanisms responsible for the thermal enhancement of the luminescence. There are several other processes that may lead to the similar thermal behaviors like i.e. change of the local ion symmetry etc.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 327, 237, 343]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 353, 850, 595]]<|/det|>
+Many thanks for this valuable suggestion. Indeed, the interionic distance contraction of the different axis is one of the very specific mechanisms responsible for the thermal enhancement of luminescence. Besides, the change of the local ion symmetry may also affect the optical properties of \(\mathrm{RE}^{3 + }\) ions like luminescence intensity and radiative transition. In this work, we propose a new class of NTE phosphors based on \(\mathrm{SMO:Yb / Er}\) exhibiting a unique two- dimensional NTE coefficient. With increasing the temperature, the \(\mathrm{SMO:Yb / Er}\) phosphors experience an anisotropic two- dimensional shrinkage along \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions increases along \(b\) axis. As such, the detrimental energy migration of \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions can be effectively confined within the two- dimensional layered structures (e.g., (010) lattice plane) at elevated temperature. More discussion has been added. Please see also lines 7- 16 of p. 5 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 633, 266, 649]]<|/det|>
+## Comment #5:  
+
+<|ref|>text<|/ref|><|det|>[[148, 659, 850, 732]]<|/det|>
+The purpose of the data shown in Figures 1 a, b, c is to describe three types of temperature effect trends on luminescence intensity. Therefore, I would suggest removing the "diagrams" placed under the graphs to maintain the universality of the description presented.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 744, 237, 760]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 770, 850, 880]]<|/det|>
+Thanks a lot. We have removed the "diagrams" placed under the graphs in the revised manuscript. The captions of Figures 1 a, b, c have been corrected as " Figure 1 Scheme of thermal- dependence effects in phosphor. a. Positive thermal quenching phenomenon. b. Zero thermal quenching phenomenon. c. Negative- thermal quenching (thermal- enhanced) phenomenon". Please see also line 1 of p. 3 in the revised manuscript.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 85, 267, 101]]<|/det|>
+## Comment #6:  
+
+<|ref|>text<|/ref|><|det|>[[148, 113, 848, 149]]<|/det|>
+Comment #6:Authors should use consistent temperature units throughout the paper (e.g. Figure1 Celsius and Figure 6 Kelvins).  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 160, 237, 177]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 187, 820, 205]]<|/det|>
+Response:All the temperature units have been corrected as Kelvins in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 243, 267, 260]]<|/det|>
+## Comment #7:  
+
+<|ref|>text<|/ref|><|det|>[[148, 270, 848, 306]]<|/det|>
+Comment #7:Please remove the information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and b.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 318, 237, 335]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 344, 850, 400]]<|/det|>
+Response:Thanks. The information about the wavelength of the radiation used in the brackets of the axis captions in Figures 2a and 2b has been removed in the revised manuscript. Please see also line 1 of p. 6 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 437, 267, 454]]<|/det|>
+## Comment #8:  
+
+<|ref|>text<|/ref|><|det|>[[148, 464, 850, 556]]<|/det|>
+Comment #8:The increase in emission intensity with increasing temperature may be due to an increase in the absorption cross section associated with the spectral broadening of the absorption band. This occurs when the excitation line etches into the slope of the absorption band. Therefore, the authors should present the excitation spectra of \(\mathrm{Yb^{3 + }}\) ions measured as a function of temperature.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 567, 237, 583]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 594, 850, 724]]<|/det|>
+Response:Many thanks for this valuable suggestion. We have added the excitation spectra of \(\mathrm{Yb^{3 + }}\) ions measured as a function of temperature in the revised manuscript (Supplementary Figure 5 newly added). Accordingly, the absorption band centered at 980 nm was hardly broadened with increasing temperature. The intensity of the excitation peak increased with increasing the temperature from 298 to 573 K, which may benefit the luminescence emission of \(\mathrm{Yb^{3 + }}\) at the higher temperature. Please see also lines 7- 11 of p. 8 in the revised manuscript.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[340, 84, 655, 305]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[197, 316, 799, 353]]<|/det|>
+<center>Supplementary Figure 5| Temperature-dependent excitation spectra of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 391, 266, 408]]<|/det|>
+## Comment #9:  
+
+<|ref|>text<|/ref|><|det|>[[148, 418, 850, 512]]<|/det|>
+The magnitude of the thermal contraction of the crystallographic unit cell depends on the difference in the ionic radii between dopants and, replaces ions and on the concentration of dopants. Therefore, I would like the authors to comment if it is possible to modulate the thermometric parameters of this type of thermometer by changing the dopant concentration?  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 522, 237, 539]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 548, 850, 752]]<|/det|>
+Many thanks for this valuable suggestion. It is true that the magnitude of the thermal contraction of the crystallographic unit cell depends on the difference in the ionic radii between dopants and replaces ions and on the concentration of dopants. Therefore, we investigated the thermometric parameters of \(\mathrm{SMO:xYb / 1\%Er}\) with different concentrations of sensitizer \(\mathrm{(Yb^{3 + })}\) . Following the reviewer's suggestion, we have measured the temperature- dependent decay curves of \(\mathrm{SMO:Yb / 1\%Er}\) and phosphors with different \(\mathrm{Yb^{3 + }}\) concentrations (10%, 20% and 25%), which were newly added as Supplementary Figure 14 in the Supporting Information. As indicated in Supplementary Table 1, \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors exhibit the optimal absolute sensitivity \(\mathrm{S_a}\) (53.0 \(\mu \mathrm{sK}^{- 1}\) ) and relative sensitivity \(\mathrm{S_r}\) (12.3% \(\mathrm{K}^{- 1}\) ). Please see also lines 20- 23 of p. 14 in the revised manuscript.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[150, 87, 840, 620]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[147, 622, 850, 825]]<|/det|>
+Supplementary Figure/14] Lifetime- based luminescence thermometry in \(\mathbf{x}\% \mathbf{Yb}^{3 + } / 1\% \mathbf{Er}^{3 + }\) - codoped SMO with different \(\mathbf{Yb}^{3 + }\) concentrations. a. Temperature- dependent luminescence decay curves of \(^{4}\mathrm{I}_{13 / 2}\) excited states of \(\mathrm{Er}^{3 + }\) in \(10\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) - codoped SMO. b. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO: \(10\% \mathrm{Yb} / 1\% \mathrm{Er}\) at different temperatures. c. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO: \(10\% \mathrm{Yb} / 1\% \mathrm{Er}\) . d. Temperature- dependent luminescence decay curves of \(^{4}\mathrm{I}_{13 / 2}\) excited states of \(\mathrm{Er}^{3 + }\) in \(25\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) - codoped SMO. e. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO: \(25\% \mathrm{Yb} / 1\% \mathrm{Er}\) at different temperatures. f. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO: \(25\% \mathrm{Yb} / 1\% \mathrm{Er}\) .
+
+<--- Page Split --->
+<|ref|>table_caption<|/ref|><|det|>[[147, 85, 850, 120]]<|/det|>
+Supplementary Table 1]. Lifetime-based luminescence thermometry parameters in x%Yb3+/1%3+-codoped SMO with different Yb3+ concentrations.
+
+<|ref|>table<|/ref|><|det|>[[221, 130, 775, 247]]<|/det|>
+
+<table><tr><td>Sample</td><td>\(S_{a}(\mu \mathrm {s}/\mathrm {K})\)</td><td>\(S_{r}(\%/K)\)</td></tr><tr><td>\(10\%Yb/1\%Er:SMO\)</td><td>31.8</td><td>3.0</td></tr><tr><td>\(20\%Yb/1\%Er:SMO\)</td><td>53.0</td><td>12.3</td></tr><tr><td>\(25\%Yb/1\%Er:SMO\)</td><td>31.2</td><td>7.1</td></tr></table>
+
+<|ref|>title<|/ref|><|det|>[[148, 284, 280, 296]]<|/det|>
+# Comment #10:
+
+<|ref|>text<|/ref|><|det|>[[148, 312, 850, 344]]<|/det|>
+Table S1- the comparing the thermometric properties of the luminescent thermometry operating in different readout modes is unjustified and unreasonable.
+
+<|ref|>title<|/ref|><|det|>[[148, 359, 240, 371]]<|/det|>
+# Response:
+
+<|ref|>text<|/ref|><|det|>[[148, 385, 666, 399]]<|/det|>
+Thank you. Table S1 has been deleted in the revised manuscript.
+
+<|ref|>image<|/ref|><|det|>[[270, 355, 717, 657]]<|/det|>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[330, 85, 667, 102]]<|/det|>
+## Reply to the Comments of Reviewer #2  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 113, 310, 129]]<|/det|>
+## General comment:  
+
+<|ref|>text<|/ref|><|det|>[[148, 140, 850, 250]]<|/det|>
+The authors of "Thermally boosted upconversion and downshifting luminescence in \(Sc_{2}(MoO_{4})_{3}\cdot Yb / Er\) with two- dimensional negative thermal expansion" presented a very interesting research of the famous upconverting pair codoped in more than astounding host matrix. However, for the following reasons I recommend major revision and reinspection, as there are numerous problems with novelty, significance and support for the conclusions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 262, 237, 278]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 289, 850, 379]]<|/det|>
+We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. We have made all the requested changes in the revised manuscript according to the reviewer's comments. Particularly, more control experiments and analyses have been added to highlight the novelty of our work. A point- by- point response is noted below.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 419, 266, 435]]<|/det|>
+## Comment #1:  
+
+<|ref|>text<|/ref|><|det|>[[148, 446, 849, 482]]<|/det|>
+The same host matrix was investigated for thermometry by the Yb/Ho pair and LIR with similar conclusions. This just slightly diminishes the stated novelty.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 493, 237, 509]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 520, 850, 797]]<|/det|>
+Thank you for pointing out this important issue. It is true that the same host matrix was investigated as traditional LIR thermometry by the Yb/Ho pair (Ref. 10). Nevertheless, the content and novelty of our work are totally different from the previous report. For the first time, we have revealed the unique two- dimensional NTE of SMO:Yb/Er phosphors, which exhibited simultaneous enhancements of upconversion and downshifting photoluminescence (PL) of \(\mathrm{Er}^{3 + }\) by 45- fold and 450- fold from 298 to 773 K, respectively. Besides, the strategy of NIR downshifting luminescence lifetime instead of the conventional visible UC luminescence intensity ratio method was applied for temperature sensing. The near- infrared luminescence lifetime of \(\mathrm{Er}^{3 + }\) in SMO:Yb/Er was determined to span two orders of magnitude from 298 to 623 K, which enables their application in lifetime- based luminescent thermometry with high absolute temperature sensitivity and relative temperature sensitivity. Our work presents a substantial advance as compared to previous work and thus could be a breakthrough in the development of \(\mathrm{RE}^{3 + }\) - doped NTE phosphors. Please see also lines 20- 26 of p. 16 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 835, 266, 851]]<|/det|>
+## Comment #2:  
+
+<|ref|>text<|/ref|><|det|>[[148, 863, 849, 898]]<|/det|>
+The discussion about inter- atomic differences is possibly erroneous, and need either serious revision or removal.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 86, 237, 102]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 112, 850, 186]]<|/det|>
+Response:Many thanks for this valuable suggestion. We have corrected the discussion about inter- atomic differences, which were estimated by the crystallographic information file (CIF) of Rietveld refinement of the in situ temperature- dependent SXRD. Please see also lines 7- 16 of p. 5 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 224, 266, 241]]<|/det|>
+## Comment #3:  
+
+<|ref|>text<|/ref|><|det|>[[148, 251, 850, 287]]<|/det|>
+Comment #3:The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 298, 237, 315]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 325, 850, 658]]<|/det|>
+Comment #3:The overall intensity of this phosphor is unknown. For this the authors must expand their research as suggested.Response:Many thanks for this valuable suggestion. We have added the discussion about the overall intensity of this phosphor. To explicitly indicate the overall intensity of this phosphor, the famous Orthorhombic- phase \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor with the same crystallographic system of SMO is selected as the control samples. To compare the emission intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor, we have measured the temperature- dependent UC/DS luminescence spectra of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) under otherwise identical conditions (Supplementary Figures. 6 and 7 newly added). It can be observed that the UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor decreased continuously with increasing temperature, while the UC/DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor increased markedly with increasing the temperature. Specifically, the overall UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is much higher than that of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) at 298 K. Nevertheless, the overall UC and DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor are 4.5 and 12.9 times higher than that of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) counterpart at 773 K, respectively. These results explicitly validate the superiority of the \(\mathrm{SMO}:\mathrm{Yb} / \mathrm{Er}\) as novel luminescent materials over existing PTE phosphors, particularly at high temperatures. The above discussions have been added in lines 1- 13 of p. 10 and Supplementary Figures 6- 7 newly added in the Supporting Information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 696, 266, 712]]<|/det|>
+## Comment #4:  
+
+<|ref|>text<|/ref|><|det|>[[148, 723, 850, 815]]<|/det|>
+Comment #4:This high relative sensitivity occurs at the points where the signal is borderline detectable. This means that the temperature resolutions will be low at those points, i.e. the significance of the research is not as high as initially suggested. The authors need to include temperature resolution discussion in the most correct manner, and accordingly adjust their conclusions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 826, 237, 843]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 853, 849, 907]]<|/det|>
+Response:Thank you for pointing out this important issue. To determine the relative sensitivity of our phosphors at other temperatures, we have measured the luminescence lifetimes of our phosphors at four temperature points (303, 313, 333, 363 K). All the PL decays
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 85, 850, 251]]<|/det|>
+for each temperature point were measured independently for ten times under identical conditions to yield the average value. Temperature resolution discussion via data analysis of the 10 consecutive measurements has been newly added in lines 1- 14 in p. 16 in the revised manuscript. A slight change in sensitivity was obtained by re- fitting the data by adding data pointing. From Figures 5 a and 5b, it is obvious that the data become dense at the temperature range from 298 to 373K, which may improve the data accuracy via decreasing the fitting error. The optimal \(S_{a}\) and \(S_{r}\) were determined to be as high as \(53.0 \mu \mathrm{sK}^{- 1}\) and \(12.3\% \mathrm{K}^{- 1}\) , respectively. The fitting results have been adjusted in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 290, 266, 306]]<|/det|>
+## Comment #5:  
+
+<|ref|>text<|/ref|><|det|>[[148, 317, 277, 333]]<|/det|>
+Issues in detail:  
+
+<|ref|>text<|/ref|><|det|>[[147, 344, 850, 512]]<|/det|>
+In the first reading the novelty of this article seems immense. However, the exactly the same host, co- doped with an upconverting pair of \(\mathrm{Yb}^{3 + } / \mathrm{Ho}^{3 + }\) has been already explored for luminescence thermometry in Ref: https://pubs.acs.org/doi/10.1021/acs.jpclett.0c00628. Although the authors did reference this work, their contribution and similarity needs to be more stressed. That being said, the differences in employing the more important upconverting pair, \(\mathrm{Yb}^{3 + } / \mathrm{Er}^{3 + }\) , and lifetime instead of luminescence intensity ratio method, should be stated as the novelty of this work. In my opinion, this subtle differences are important and enough novelty for the publication of this work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 522, 237, 538]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 549, 850, 660]]<|/det|>
+Many thanks for this valuable suggestion. Previously, SMO:Yb/Ho with same host was used as a ratiometric thermometer based on the UC red- to- green emission intensity ratio. Their contribution and similarity about the same SMO:Yb/Ho used as luminescence thermometry in Ref. 10 have been stressed in the revised manuscript. The above discussions have been added in lines 29- 30 of p. 13 in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[147, 669, 850, 890]]<|/det|>
+It should be noted that the content and novelty in this work are totally different from the previous report. For the first time, we have revealed the unique two- dimensional NTE of SMO:Yb/Er phosphors, which exhibited simultaneous enhancements of upconversion and downshifting PL of \(\mathrm{Er}^{3 + }\) by 45- fold and 450- fold from 298 to 773 K, respectively. Besides, the strategy of NIR downshifting luminescence lifetime instead of the conventional visible UC luminescence intensity ratio method was applied for temperature sensing. The near- infrared luminescence lifetime of \(\mathrm{Er}^{3 + }\) in SMO:Yb/Er was determined to span two orders of magnitude from 298 to 623 K, which enables their application in lifetime- based luminescent thermometry with high absolute temperature sensitivity and relative temperature sensitivity. Our work presents a substantial advance as compared to previous work and thus could be a breakthrough in the development of \(\mathrm{RE}^{3 + }\) - doped NTE
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 849, 121]]<|/det|>
+phosphors. We sincerely hope the reviewer concurs after reading this clarification.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 159, 266, 175]]<|/det|>
+## Comment #6:  
+
+<|ref|>text<|/ref|><|det|>[[147, 187, 850, 352]]<|/det|>
+For the equation 1 the referenced is the famous short work by Blasse (Ref. 22). However, in that reference the Blasse is not reporting on a ways of calculation of the average distance, but on the critical distance, i.e. when the concentration quenching starts to dominate. The c in the subscript of xc does not abbreviate concentration, but the word critical. Up to my knowledge, this equation cannot be applied for calculation of distance between interacting ions if that distance is not critical distance, and even more, of the ions of different types. Unless the authors can provide another, a very strong reference to support their claims of usage of equation 1, I suggest removing the Figure 2d, Equation 1, and any corresponding discussion.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 365, 237, 381]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 390, 850, 630]]<|/det|>
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we have updated Figure 2d. For the case of homogeneous substitution of \(\mathrm{Sc^{3 + }}\) by \(\mathrm{RE^{3 + }}\) ions in SMO, the changing trend of the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions ( \(\mathrm{RE} = \mathrm{Sc / Yb / Er}\) ) along different crystal axis- direction in the structure diagram for different temperatures will reflect the changing trend of the distance between the \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) ions. With the increase of temperature, the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(a\) - axis (6.7061 Å, at 298 K) contracts, and the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(b\) - axis (10.8288 Å, at 298 K) expands steadily, while the distance of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(c\) - axis first expands in the temperature range (298- 373 K) and then contracts. The results indicate that the distances of \(\mathrm{RE^{3 + } - RE^{3 + }}\) ions along \(a / c\) - axis decreased at higher temperatures, which may benefit the improvement of the energy- transfer efficiency between \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) . See also the updated discussion in lines 7- 16 on page 5 of the revised manuscript.  
+
+<|ref|>image<|/ref|><|det|>[[168, 644, 820, 844]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[171, 854, 825, 890]]<|/det|>
+<center>Figure 2. Temperature-dependent proximate distances of RE-RE (RE=Sc/Yb/Er) along different axes derived from the unit cell structure. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[149, 85, 266, 101]]<|/det|>
+## Comment #7:  
+
+<|ref|>text<|/ref|><|det|>[[148, 112, 850, 315]]<|/det|>
+There is another issue with concentrations and distances. In figure 2 no concentrations of dopants are mentioned and yet there is a graph of their distances. To additionally confirm the assumption that the author's analysis of interatomic distances is incorrect is the un- logical trends in presented distances. Although both the codopants substitute the same ions in the SMO matrix, the authors claim that Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase. All three claims cannot be possible, if I understood well, if not, then the paragraph needs to be revised for clarity. As for the red emission of \(\mathrm{Er}^{3 + }\) , the authors should see the book by Kaminskii, Crystalline lasers, where the pathways for population of the \(^4\mathrm{F}_{9 / 2}\) are explained, and they are related to the \(\mathrm{Yb}^{3 + }\) ion back- transfer more than the \(\mathrm{Er}^{3 + }\) cross- relaxation.  
+
+<|ref|>text<|/ref|><|det|>[[149, 325, 849, 380]]<|/det|>
+Response: Thank you for pointing out our mistakes. In Figure 2, the \(\mathrm{Yb}^{3 + } / \mathrm{Er}^{3 + }\) dopant concentrations are \(20\%\) and \(1\%\) , respectively, which was added in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[148, 390, 850, 500]]<|/det|>
+For the incorrect description "Yb- Yb and Yb- Er distances decrease with temperature, but the Er- Er distances increase.", we have updated Figure 2d, wherein the distances of \(\mathrm{RE}^{3 + } - \mathrm{RE}^{3 + }\) ions at different planes were determined by the crystallographic information file (CIF) of Rietveld refinement on the basis of the SXRD patterns (Supplementary Figure 3c). See also the updated discussion in lines 1- 10 on page 6 of the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[147, 509, 855, 696]]<|/det|>
+The confusion of the temperature- dependent luminescence mechanism in this work was perhaps caused by our unclear writing in the manuscript. Regarding the red emission of \(\mathrm{Er}^{3 + }\) , it is true that the shortened \(\mathrm{Yb}^{3 + } - \mathrm{Er}^{3 + }\) interatomic distance may contribute to the back energy transfer from \(\mathrm{Er}^{3 + }\) to \(\mathrm{Yb}^{3 + }\) ions: \(^4\mathrm{S}_{3 / 2}(\mathrm{Er}^{3 + }) + ^2\mathrm{F}_{7 / 2}(\mathrm{Yb}^{3 + })\) \(\rightarrow {}^{4}\mathrm{I}_{13 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{5 / 2}(\mathrm{Yb}^{3 + })\) , followed by energy transfer from \(\mathrm{Yb}^{3 + }\) to \(\mathrm{Er}^{3 + }\) through \(^4\mathrm{I}_{13 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{5 / 2}(\mathrm{Yb}^{3 + })\rightarrow {}^{4}\mathrm{F}_{9 / 2}(\mathrm{Er}^{3 + }) + {}^{2}\mathrm{F}_{7 / 2}(\mathrm{Yb}^{3 + })\) . As a result, the population of the \(\mathrm{F}_{9 / 2}\) states is enhanced to produce stronger red UC emissions (Ref. 34 Kaminskii A., Crystalline Lasers: Physical Processes and Operating Schemes. (CRC Press, 2020).). The above discussion has been added in lines 20- 24 of p. 7 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 735, 266, 751]]<|/det|>
+## Comment #8:  
+
+<|ref|>text<|/ref|><|det|>[[147, 761, 850, 890]]<|/det|>
+Lifetime method for thermometry is erroneously described as ratiometric multiple times in the abstract and throughout the article. The definition of ratiometric fluorescence is where intensities of at least two bands are measured and compared. Thus, the word ratio in ratiometric. FIR or LIR methods, of which the authors are certainly aware of, are the ratiometric luminescence thermometry methods, while lifetime, although also self- referenced, is not. This needs to be corrected in the caption of Table S1 as well.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 85, 850, 232]]<|/det|>
+Response: Thanks. We have carefully checked the manuscript and revised similar mistakes. In the abstract "the luminescence lifetime of \(^{4}\mathrm{I}_{11 / 2}\) of \(\mathrm{Er}^{3 + }\) in \(\mathrm{SMO:Yb / Er}\) displays a strong temperature dependence, enabling ratiometric thermometry with the highest relative sensitivity of \(12.3\% / \mathrm{K}\) at \(298\mathrm{K}\) has corrected as "the luminescence lifetime of \(^{4}\mathrm{I}_{11 / 2}\) of \(\mathrm{Er}^{3 + }\) in \(\mathrm{SMO:Yb / Er}\) displays a strong temperature dependence, enabling luminescence thermometry with the highest relative sensitivity of \(12.3\% / \mathrm{K}\) at \(298\mathrm{K}\) ". Please see also line 30 p. 1 in the revised manuscript. Besides, Table S1 has been deleted according to the comment of the first reviewer.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 272, 266, 288]]<|/det|>
+## Comment #9:  
+
+<|ref|>text<|/ref|><|det|>[[149, 298, 850, 370]]<|/det|>
+Throughout the article there needs to be more linear story of using concentrations and even hosts. It is hard to follow which concentrations are being used, and when it is compared to the other host. For each spectra the precise concentrations of codopants should be given.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 382, 237, 399]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[149, 408, 850, 444]]<|/det|>
+Thank you for your kind reminding. For each spectrum, the precise concentration of dopants has been given in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 484, 277, 500]]<|/det|>
+## Comment #10:  
+
+<|ref|>text<|/ref|><|det|>[[148, 510, 850, 584]]<|/det|>
+The authors discuss the \(^{4}\mathrm{I}_{11 / 2}\) emission of \(\mathrm{Er}^{3 + }\) , however in the spectra and in the text the \(1538\mathrm{nm}\) emission is mentioned. Did you mean the emission from \(^{4}\mathrm{I}_{13 / 2}\) instead, as the emission from \(^{4}\mathrm{I}_{11 / 2}\) to the ground level is at about \(1000\mathrm{nm}\) ? See for example Figure S9b: caption says \(^{4}\mathrm{I}_{11 / 2}\) and \(\mathrm{Eem} = 1538\mathrm{nm}\) , the same in Figure 5.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 595, 237, 611]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[149, 621, 850, 675]]<|/det|>
+Thank you for your kind reminding. Indeed, the emission peak of \(1538\mathrm{nm}\) originated from \(^{4}\mathrm{I}_{13 / 2}\) of \(\mathrm{Er}^{3 + }\) . All the " \(^{4}\mathrm{I}_{11 / 2}\) " have been revised as " \(^{4}\mathrm{I}_{13 / 2}\) " in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 715, 276, 731]]<|/det|>
+## Comment #11:  
+
+<|ref|>text<|/ref|><|det|>[[149, 742, 850, 833]]<|/det|>
+For other journals of lower esteem than the Nature series, the comparison with other hosts would not be necessary. However, I feel that in this host the emission intensity is much lower than in other famous upconverting hosts, for example YF3 or NaYF4. I would like to see the comparison in either intensities or quantum yields between SMO and one of these prominent hosts.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 845, 237, 861]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[149, 872, 850, 907]]<|/det|>
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we have selected \(\mathrm{YF}_{3}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor as the control group. As we know, the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 850, 380]]<|/det|>
+famous Orthorhombic- phase \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor with the same crystallographic system of SMO is considered to be one of the most efficient UC/DS luminescent materials. To compare the emission intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor, we have measured the temperature- dependent UC/DS spectra of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) under otherwise identical conditions (Supplementary Figures 6 and 7 newly added). It can be observed that the UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor decreased with increasing temperature, while the UC/DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor increased markedly with increasing the temperature. Specifically, the overall UC/DS intensity of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is much higher than that of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) at 298 K. Nevertheless, the overall UC and DS intensity of \(\mathrm{SMO}:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor are 4.5 and 12.9 times higher than that of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) counterpart at 773 K, respectively. These results explicitly validate the superiority of the SMO:Yb/Er as novel luminescent materials over existing PTE phosphors, particularly at high temperatures. The above discussions have been added in lines 1- 13 of p. 10 and Supplementary Figures 6 and 7 newly added in the Supporting Information.  
+
+<|ref|>image<|/ref|><|det|>[[278, 384, 680, 906]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 852, 195]]<|/det|>
+Supplementary Figure 6| Upconversion emission spectra of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor and \(\mathrm{YF}_3:20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) phosphor. a. XRD patterns of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) phosphor. b. Upconversion emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) as a function of temperature under 980 nm excitation. c. Comparison of the relative integrated intensity of the upconversion emission of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor, wherein the data of \(\mathrm{SMO:20\%Yb / 1\%Er}\) originate from Figure 3b.  
+
+<|ref|>image<|/ref|><|det|>[[147, 201, 846, 580]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 585, 850, 694]]<|/det|>
+<center>Supplementary Figure 7| Downshifting emission spectra of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor and \(\mathrm{YF}_3:20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) phosphor. a. and b. Downshifting emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor as a function of temperature under 980 nm excitation. c. Comparison of the relative integrated intensity of the downshifting emission spectra of the \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) and \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. The enlarged data of \(\mathrm{YF}_3:20\% \mathrm{Yb} / 1\% \mathrm{Er}\) is displayed on the right side. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 734, 277, 750]]<|/det|>
+## Comment #12:  
+
+<|ref|>text<|/ref|><|det|>[[147, 761, 850, 909]]<|/det|>
+Lastly but not the least, this amazing sensitivity occurs at the exact same temperatures where the emission intensities are very low. Thus, at the points of the highest relative sensitivity, the sensing might be just possible, as the uncertainty in measurement will be great. This was discussed in the article: https://onlinelibrary.wiley.com/doi/full/10.1002/adts.202000176. This last point significantly diminishes the firstly claimed importance of the obtained results. I invite the authors to be honest in reporting the sensitivities only on the practically usable range. Also, taking everything in consideration, there is a must that the authors
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 85, 852, 195]]<|/det|>
+report on the temperature resolutions at the whole temperature range, and unfortunately this means reporting on the uncertainties that should be evaluated by at least several temperatures. Uncertainties should be given as the standard deviation of at least 10 consecutive measurements at the same temperature. This temperature resolutions will show the true usability of the sensor, and my assumption is that the temperature resolution curve will be different than the sensitivity curves.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 207, 237, 222]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 233, 852, 473]]<|/det|>
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, we measured the luminescence lifetimes of our phosphors at four temperature points (303, 313, 333, 363 K). All the PL decays for each temperature point were measured independently for ten times under identical conditions to yield the average value. Temperature resolution discussion via data analysis of the 10 consecutive measurements has been newly added at lines 1- 14 in p. 16 of the revised manuscript. A slight change in sensitivity was obtained by re- fitting the data by adding data pointing. From Figures 5a and 5b, it is obvious that the data become dense at the temperature range from 298 to 373 K, which can improve the data accuracy via decreasing the fitting error. The optimal \(S_{a}\) and \(S_{r}\) were determined to be as high as 53.0 \(\mu \mathrm{sK}^{- 1}\) and 12.3% \(\mathrm{K}^{- 1}\) , respectively. The fitting results have been updated in the revised manuscript. The suggested reference (Adc, Theory Simul. 3, 2000176 (2020)) has been cited as ref. 45 for the discussion of temperature uncertainty.  
+
+<|ref|>image<|/ref|><|det|>[[148, 484, 848, 636]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 642, 850, 732]]<|/det|>
+<center>Figure 5| Lifetime-based luminescence thermometry in 20% \(\mathrm{Yb^{3 + } / 1\% Er^{3 + }}\) -codoped SMO. a. Experimentally measured and exponentially fitted plots of lifetime \(\tau\) of SMO:20%Yb/1%Er at different temperatures. b. Calculated absolute sensitivity \(S_{a}\) and relative sensitivity \(S_{r}\) versus temperature based on the SMO:20%Yb/1%Er. d. Thermal dependence of temperature uncertainty for SMO:20%Yb/1%Er. </center>  
+
+<|ref|>text<|/ref|><|det|>[[148, 771, 850, 826]]<|/det|>
+Moreover, the temperature uncertainty ( \(\delta \mathrm{T}\) ) is an important parameter to assess the performance of a thermometer since it includes not only the relative sensitivity but also the error on the luminescence lifetime ( \(\delta \tau\) ) \(^{55,56}\) . \(\delta \mathrm{T}\) is calculated as follows:  
+
+<|ref|>equation<|/ref|><|det|>[[153, 833, 844, 870]]<|/det|>
+\[\delta T = \frac{1}{S_{r}}\cdot \frac{\delta\tau}{\tau} \quad (6)\]  
+
+<|ref|>text<|/ref|><|det|>[[148, 873, 850, 909]]<|/det|>
+Where \(\delta \tau /\tau\) is the uncertainty in the calculation of \(\tau\) (determined as a standard deviation in ten measurements of \(\tau\) at the same temperature), \(S_{r}\) is the relative
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 850, 400]]<|/det|>
+sensitivity of luminescence thermometer. Temperature- dependent luminescence decay curves of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors by ten measurements at the same temperature are shown in the Supplementary Figure 15 newly added. Thermal dependence of temperature uncertainty for \(\mathrm{SMO:20\%Yb / 1\%Er}\) is presented in Figure 6d. The minimum value of \(\delta \mathrm{T}\) is \(0.11\mathrm{K}\) even at \(623\mathrm{K}\) . As such, although there is a relatively low value of \(\mathbf{S}_{\mathrm{r}}\) at high- temperature range, the obviously longer luminescence lifetime \((\tau)\) can effectively reduce the value of \(\delta \tau /\tau\) , resulting in lower \(\delta T\) . It should be noted that it is possible to keep the temperature uncertainty below a threshold of \(0.33\mathrm{K}\) throughout the whole studied temperature range (298- 623 K). The \(\delta T\) threshold of \(\mathrm{SMO:20\%Yb / 1\%Er}\) for LLT is much lower than that (0.7 K) of cubic- phase \(\mathrm{LiLuF_4:18\%Yb^{3 + } / 2\%Er^{3 + }}\) nanocrystals for the conventional intensity- based thermometry. The \(\delta T\) threshold of \(\mathrm{SMO:20\%Yb / 1\%Er}\) is comparative to that (0.46) of \(\mathrm{LiLuF_4:18\%Yb^{3 + } / 2\%Er^{3 + }@SiO_2}\) (three shells) synthesized with the complicated condition. All these results indicate that \(\mathrm{SMO:Yb / Er}\) phosphor can be explored as a kind of ideal lifetime- based luminescence thermometry with high \(S\) , and low \(\delta T\) , which validate the superiority and applicability of the proposed \(\mathrm{SMO:Yb / Er}\) for luminescence thermometry.  
+
+<|ref|>image<|/ref|><|det|>[[275, 400, 717, 655]]<|/det|>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[144, 80, 850, 772]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 771, 850, 825]]<|/det|>
+<center>Supplementary Figure 15| Temperature-dependent luminescence decay curves of \(^{4}I_{132}\) excited states of \(\mathrm{Er}^{3 + }\) in \(20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) -codoped SMO at different temperatures from 298 K to 623 K. Each temperature point was measured by ten consecutive measurements. </center>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[348, 85, 686, 102]]<|/det|>
+## Reply to the Comments of Reviewer #3  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 122, 310, 139]]<|/det|>
+## General comment:  
+
+<|ref|>text<|/ref|><|det|>[[148, 149, 850, 297]]<|/det|>
+The authors present a two- dimensional negative thermal phosphor SMO:Yb/Er systematic study. They present a temperature dependent Raman, synchrotron X- ray diffraction and luminescence investigation. The thermally boosted UC and DS luminescence mechanism was investigated. The authors claim that the luminescence lifetime temperature dependency enables ratiometric thermometry with high relative sensitivity; an important topic in current research. This is a comprehensive work, very well written and with insights into the \(\mathrm{Ln}^{3 + }\) - doped phosphors for potential applications.  
+
+<|ref|>text<|/ref|><|det|>[[148, 298, 848, 334]]<|/det|>
+Although they address an important topic some issues, that I refer below, need further clarification.  
+
+<|ref|>text<|/ref|><|det|>[[148, 353, 850, 483]]<|/det|>
+Response: We greatly appreciate the reviewer for his/her positive comments on our manuscript and the efforts to improve the quality of our manuscript. We have made all the requested changes in the revised manuscript according to the reviewer's comments. Particularly, the mechanism for the NTE behaviour of SMO:Yb \(^{3 + }\) /Er \(^{3 + }\) has been clarified by means of more characterizations. Some inappropriate statement or unclear writing has been rephrased throughout the manuscript. A point- by- point response is noted below.  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 510, 266, 527]]<|/det|>
+## Comment #1:  
+
+<|ref|>text<|/ref|><|det|>[[148, 538, 848, 575]]<|/det|>
+Authors should comment, if known, what is the mechanism for the NTE behaviour in this system, whether it is structural, or electronically based for instance?  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 585, 237, 601]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 611, 850, 908]]<|/det|>
+Many thanks for this valuable suggestion. Following the reviewer's suggestion, the structural rigid unit mode has been added to explain the mechanism for the NTE behavior of SMO:Yb \(^{3 + }\) /Er \(^{3 + }\) . The NTE behavior in this system is attributed to the structural variability with temperature. To shed more light on the structure of RE \(^{3 + }\) - doped SMO, rigid unit model for RE \(^{3 + }\) - doped SMO has been given as Figure 2e- 2g (Front. Phys., 16, 53302 (2021); Rep. Prog. Phys., 79, 066503 (2016)). With the temperature increased from 420 to 623 K, the distortion of the \(\mathrm{REO}_6\) octahedron and twist of the RE- O- Mo may induce a variation in the rigid unit mode, resulting in a decrease ratio in the distance of RE- Mo along the \(a\) and \(c\) axes by \(0.26\%\) and \(0.10\%\) , as well as an increased ratio in the distance of RE- Mo along \(b\) axis by \(0.02\%\) (Figure 3d). Note that the decreased amplitudes in the distance of RE- Mo along the \(a\) and \(c\) axes are one order larger than that along the \(b\) axis. Such an alteration causes not only shrinkage of lattice but also a local distortion of RE with increasing the temperature. The angle of RE- Mo- RE nearly keeps unchanged at the temperature range of 420- 623 K, which indicates that the structure of \(\mathrm{Sc}_2\mathrm{Mo}_3\mathrm{O}_{12}\) is rigid. The above discussions have been added in lines 17- 24 of p. 5 in the revised manuscript.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[168, 90, 848, 243]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 260, 850, 410]]<|/det|>
+<center>Figure 3. e. Rigid unit mode model for \(\mathrm{Sc_2Mo_3O_{12}:20\%Yb^{3 + } / 1\%Er^{3 + }}\) extracted from the unit cell structure. f. Temperature-dependence ratio \(\delta D\) of RE-Mo distances marked in the model \(\begin{array}{r}\delta D = \frac{D_T - D_{420}}{D_{420}}\times 100\% \end{array}\) \(D_{T}\) and \(D_{420}\) stand for RE-Mo distances of the given temperature and \(420\mathrm{K}\) , respectively). g. Temperature-dependence ratio \(\delta \theta\) of RE-Mo-RE angles marked in the model \(\begin{array}{r}\delta \theta = \frac{\theta_T - \theta_{420}}{\theta_{420}}\times 100\% \end{array}\) \(\theta_{T}\) and \(\theta_{420}\) stand for RE-Mo-RE angles of the given temperature and \(420\mathrm{K}\) , respectively). </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 446, 267, 463]]<|/det|>
+## Comment #2:  
+
+<|ref|>text<|/ref|><|det|>[[148, 473, 850, 584]]<|/det|>
+What is supposed to drive the transition from PTE to NTE around \(100^{\circ}\mathrm{C}\) ? The transition from PTE to NTE occurs at ca \(100^{\circ}\mathrm{C}\) (figure 2c). How this transition impacts the luminescence properties and how this is compatible with the suggested thermally enhanced photoluminescence mechanism. It is important to clarify this issue. Some Raman modes appear to alter above \(100^{\circ}\mathrm{C}\) , can they be related with the NTE to PTE transition?  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 596, 237, 612]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 621, 850, 732]]<|/det|>
+Many thanks for this valuable suggestion. Such a transformation from PTE to NTE can be attributed to the existence of water molecules in the \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphors as revealed by the thermogravimetry (TG) analysis, in situ temperature- dependent Fourier transform infrared (FTIR) spectroscopy (Supplementary Figure 3 newly added), and temperature- dependent Raman spectra (Figure 2h).  
+
+<|ref|>text<|/ref|><|det|>[[148, 742, 850, 909]]<|/det|>
+As shown in Supplementary Figure 3a, a weight loss of about \(1\%\) for \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) was detected by heating from 298 to \(420\mathrm{K}\) . Generally, \(\mathrm{A_2Mo_3O_{12}}\) compounds with large \(\mathrm{A^{3 + }}\) cation size are highly hygroscopic and easily hydrated at ambient conditions. Therefore, \(\mathrm{Sc_2Mo_3O_{12}}\) compound with small \(\mathrm{Sc^{3 + }}\) (0.0745 nm, \(\mathrm{CN} = 6\) ) is nonhygroscopic. For \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors with the substitution of \(\mathrm{Sc^{3 + }}\) by the big \(\mathrm{Yb^{3 + }}\) (0.0868 nm) and \(\mathrm{Er^{3 + }}\) (0.089 nm, \(\mathrm{CN} = 6\) ), water molecules may be easily absorbed into the relatively large microchannels (Ref. 23, Effect of water species on the phonon modes in orthorhombic \(\mathrm{Y_2(MoO_4)_3}\) revealed by Raman spectroscopy. J. Phys. Chem. C, 112, 6577- 6581 (2008)). Accordingly, it
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 85, 849, 120]]<|/det|>
+can be deduced that weight loss of about \(1\%\) for \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) may result from the removal of water molecules.  
+
+<|ref|>text<|/ref|><|det|>[[147, 131, 850, 373]]<|/det|>
+In the FTIR spectra, the typical absorbance peak at \(3400–3500 \mathrm{cm}^{- 1}\) , corresponding to the asymmetric vibration of - OH was markedly suppressed with the temperature from 298 to \(420 \mathrm{K}\) (Supplementary Figure 3b), which further verified the existence and removal of water molecules. As such, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors with water molecules in the microchannels exhibit PTE behavior at low temperature (Ref.16, Negative thermal expansion: mechanisms and materials. Front. Phys., 16, 53302 (2021)). With the temperature rising from 298 to \(353 \mathrm{K}\) , the water molecules remain in the microchannels. With an increase in temperature to around \(353 \mathrm{K}\) , water molecules begin to escape from the microchannels. As such, the \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors without water molecules in the microchannels exhibit intrinsic NTE character. When the temperature reaches \(420 \mathrm{K}\) , water molecules are completely removed and thus \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphor may completely recover the NTE property.  
+
+<|ref|>image<|/ref|><|det|>[[147, 384, 850, 599]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 603, 850, 657]]<|/det|>
+<center>Supplementary Figure 3| Water molecules analysis. a. Thermogravimetry curves of the \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. b. Temperature-dependent Infrared spectra of \(\mathrm{SMO:20\%Yb / 1\%Er}\) phosphor. </center>  
+
+<|ref|>image<|/ref|><|det|>[[365, 667, 644, 858]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[145, 881, 844, 900]]<|/det|>
+<center>Supplementary Figure 4 c. Temperature-dependent changes of the unit cell volumes. </center>
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[333, 118, 670, 333]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[168, 337, 828, 372]]<|/det|>
+<center>Figure 2 h. Temperature-dependent in situ Raman spectra within the temperature range from 298 to 773 K. </center>  
+
+<|ref|>text<|/ref|><|det|>[[147, 400, 850, 641]]<|/det|>
+For the in situ temperature- dependent Raman spectra (Figure 2h), the Raman peaks of 330, 836, and \(945~\mathrm{cm^{- 1}}\) are characteristic of the hydrated orthorhombic structure, indicative of water species residing in the microchannels of \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) . As the temperature increases from 298 to 348 K, the intensity and position of these peaks remain unchanged, which reveals that the water molecules remain in the microchannels. As the temperature increases from 348 to 423 K, these peaks become weaker. Above 423 K, the peak at \(945~\mathrm{cm^{- 1}}\) vanished, which reveals the complete removal of water molecules. These results are in agreement with the obtained results of the TG, FTIR/spectra, as well as in situ temperature- dependent SXRD patterns (Supplementary Figures 3 and 4). Thus, it can be deduced that the temperature transition points of the onset and complete removal of water molecules from microchannels are \(\sim 348\) and \(\sim 423\) K, respectively. The above discussions have been added in lines 25- 29 of p. 5 in the revised manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[147, 668, 850, 891]]<|/det|>
+Based on these results, the photoluminescence mechanisms of \(\mathrm{SMO:Yb / Er}\) phosphors at different temperatures have been updated. For the temperature range of 298- 353 K, the \(\mathrm{SMO:Yb / Er}\) phosphors exhibit PTE due to the existence of water molecules in the microchannels. At higher temperatures above 353 K, the \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibited thermally enhanced photoluminescence behavior due to the typical NTE properties of \(\mathrm{A_2Mo_3O_{12}}\) compounds (Supplementary Figure 4c). Considering our topic is thermally boosted upconversion and downshifting luminescence in \(\mathrm{SMO:Yb / Er}\) with two- dimensional negative thermal expansion, we have displayed the temperature range (353- 773) of the thermally enhanced photoluminescence mechanism in the energy level diagram (Figure 5c). Please see also lines 2- 5 of p. 12 in the revised manuscript and Supplementary Figure 3 newly added in the Supporting Information.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[187, 92, 844, 455]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[147, 465, 850, 593]]<|/det|>
+<center>Figure 5| Thermally enhanced photoluminescence mechanism a. Temperature-dependent luminescence lifetime of \(^{2}\mathrm{F}_{5 / 2}\) excited state of \(\mathrm{Yb^{3 + }}\) in \(\mathrm{Yb^{3 + } / Er^{3 + }}\) -codoped and \(\mathrm{Yb^{3 + }}\) -doped SMO, respectively. Temperature-dependent energy transfer efficiency of \(\mathrm{Yb^{3 + }}\) -to-Er \(^{3 + }\) . b. Temperature-dependent lifetimes of \(^{2}\mathrm{H}_{11 / 2}\) (522 nm) and \(^{4}\mathrm{I}_{13 / 2}\) (1538 nm) excited states of \(\mathrm{Er^{3 + }}\) in SMO: \(20\% \mathrm{Yb} / 1\% \mathrm{Er}\) , respectively. c. Energy level diagram of green UC and NIR DS emission showing the proposed temperature dependence of electronic transition and energy-transfer processes in SMO: \(\mathrm{Yb / Er}\) with two-dimensional negative thermal expansion. </center>  
+
+<|ref|>sub_title<|/ref|><|det|>[[149, 633, 266, 649]]<|/det|>
+## Comment #3:  
+
+<|ref|>text<|/ref|><|det|>[[147, 658, 850, 787]]<|/det|>
+From lines 182- 183: "Moreover, the distance of \(\mathrm{Yb^{3 + } - Er^{3 + }}\) becomes shorter with an increase of the temperature. The ET processes between sensitizer \((\mathrm{Yb^{3 + }})\) to activator \((\mathrm{Er^{3 + }})\) are usually considered to occur through dipolar- dipolar interactions, whose ET efficiency is proportional to \(r^{- 6}\) ( \(r\) is the average donor- acceptor distance). The result indicates that the distances of \(\mathrm{Yb^{3 + } - Er^{3 + }}\) ions (13.975 \(\approx\) , at 25 \(^\circ \mathrm{C}\) ) decreased at higher temperature, which may benefit the improvement of the ET efficiency between \(\mathrm{Yb^{3 + } - Er^{3 + }}\) ."  
+
+<|ref|>text<|/ref|><|det|>[[147, 789, 850, 899]]<|/det|>
+And line 236 "Based on the above analysis, the thermally enhanced UC emission with the temperature from 25 to \(200^{\circ}\mathrm{C}\) is mainly governed by the ET from \(\mathrm{Yb^{3 + }}\) to \(\mathrm{Er^{3 + }}\) " Authors should clarify the nature for thermally enhanced UC emission within the temperature range from 25 to \(200^{\circ}\mathrm{C}\) , if this effect is mainly governed by the ET from \(\mathrm{Yb^{3 + }}\) to \(\mathrm{Er^{3 + }}\) , how it relates with the PTE of the system shown in Figure 2 observed between \(25^{\circ}\mathrm{C}\) and \(100^{\circ}\mathrm{C}\) .
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 86, 237, 102]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 112, 850, 261]]<|/det|>
+Many thanks for pointing out this problem. As indicated in the above analyses, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibit PTE behavior at the temperature below 353 K due to the existence of water molecules in the microchannels. When the temperature increases up to 353 K, water molecules begin to escape from the microchannels. As such, \(\mathrm{SMO:20\%Yb^{3 + } / 1\%Er^{3 + }}\) phosphors exhibit thermally enhanced photoluminescence behavior due to the typical NTE properties of \(\mathrm{A_2Mo_3O_{12}}\) compounds. Please see also lines 22- 29 of p. 4 in the revised manuscript and Supplementary Figure 3 newly added in the Supporting Information.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 299, 266, 315]]<|/det|>
+## Comment #4:  
+
+<|ref|>text<|/ref|><|det|>[[148, 325, 850, 399]]<|/det|>
+From lines 46- 49: "Most of such abnormal thermally enhanced UC luminescence is observed in \(\mathrm{Ln3 + }\) - doped inorganic materials with three- dimensional negative- thermal expansion characteristics (NTE)9, 10, where all the three cell parameters of the doped crystals shrink at elevated temperature."  
+
+<|ref|>text<|/ref|><|det|>[[148, 400, 850, 435]]<|/det|>
+And below, lines 58- 59 "with a unique two- dimensional NTE coefficient \((\alpha_{a} = - 8.62\times 10^{- 6} / \mathrm{K}, \alpha_{b} = 4.25\times 10^{- 6} / \mathrm{K}, \alpha_{c} = - 6.35\times 10^{- 6} / \mathrm{K})\) "  
+
+<|ref|>text<|/ref|><|det|>[[148, 437, 850, 490]]<|/det|>
+In the context of this work, is there any effective advantage in two- dimensional negative- thermal expansion materials when compared with the three- dimensional ones? The authors could also mention recent works with uniaxial NTE.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 504, 237, 520]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 530, 850, 880]]<|/det|>
+Many thanks for this valuable suggestion. The advantage of two- dimensional (2D) negative- thermal expansion materials relative to the three- dimensional (3D) ones was not explicitly elucidated perhaps because of our unclear writing in the manuscript. For the previously reported \(\mathrm{RE^{3 + }}\) doped phosphors, the existence of 3D compression at elevated temperature can promote the energy transfer between sensitizers and activators. Meanwhile, such a 3D compression may also benefit the dissipation of the excitation energy in all directions of crystal sublattice to the lattice/surface defects, which deteriorates the luminescent emission of \(\mathrm{RE^{3 + }}\) ions. In this work, we propose a new class of NTE phosphors based on \(\mathrm{SMO:Yb / Er}\) exhibiting a unique two- dimensional NTE coefficient. With increasing the temperature, the \(\mathrm{SMO:Yb / Er}\) phosphors experience an anisotropic two- dimensional shrinkage along the \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE^{3 + }}\) ions increases along the \(b\) axis. As such, the detrimental energy migration of \(\mathrm{RE^{3 + }}\) dopants can be effectively confined within the two- dimensional layered structures (e.g., (020) lattice plane) at elevated temperatures. Please see also lines 14- 16 of p. 5 in the revised manuscript. Recent works with uniaxial NTE behavior have been cited as Refs. 12- 15 (Adv. Sci. 2016, 3, 1600108; Phys. Rev. B 2020, 101, 104305; Inorg. Chem., 56, 15101- 15109 (2017); Phys. Rev. B, 96, 134113- 134120 (2017)).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 85, 266, 101]]<|/det|>
+## Comment #5:  
+
+<|ref|>text<|/ref|><|det|>[[148, 112, 849, 148]]<|/det|>
+Can the authors explain in what way the Energy migration between \(Ln^{3 + }\) dopants is confined to two dimensions?  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 160, 237, 176]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 186, 850, 410]]<|/det|>
+Thank you for pointing out this important issue. For \(\mathrm{RE}^{3 + }\) - doped inorganic phosphors, the long- distance energy migration between \(\mathrm{RE}^{3 + }\) dopants depends on the distance between lattice positions occupied by \(\mathrm{RE}^{3 + }\) ions. As such, it is possible to minimize the depletion of excitation energy by employing a crystal lattice disfavoring the long- distance energy migration (Nat. Mater. 2014, 13, 157- 162). In our work, the SMO:Yb/Er phosphors experience an anisotropic two- dimensional shrinkage along the \(a\) and \(c\) axes at elevated temperature, whereas the distance between \(\mathrm{RE}^{3 + }\) ions increases along with the \(b\) axis. Under such circumstances, the migration of excitation energy between \(\mathrm{RE}^{3 + }\) dopants may be effectively minimized long the \(b\) axis. Thus, more energy migration between \(\mathrm{RE}^{3 + }\) dopants is confined to the two- dimensional layered structures (e.g., (020) lattice plane) at elevated temperatures. Please see also lines 21- 23 of p. 2 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 446, 266, 462]]<|/det|>
+## Comment #6:  
+
+<|ref|>text<|/ref|><|det|>[[147, 472, 850, 585]]<|/det|>
+From lines 231- 233. The promoted Ar and inhibited \(A_{nr}\) in the \(Ln^{3 + }\) dopant indicate that lattice distortion enhances a crystal field with odd parity and modifies local symmetry of activator ions via the NTE effect. The authors mention the existence of lattice distortion and a change in the local symmetry of activator ions, via the NTE effect. Can these local distortions be substantiated via the temperature- dependent Raman data?  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 596, 237, 612]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 622, 850, 899]]<|/det|>
+Many thanks for this valuable suggestion. Raman spectroscopy is a valuable tool to study the phonon modes of NTE materials and evaluate the change of local structure. In situ temperature- dependent Raman spectra are shown in Figure 2h. The Raman peaks of 330, 836, and \(945~\mathrm{cm^{- 1}}\) are characteristic of the hydrated orthorhombic structure, indicative of water species residing in the microchannels of SMO: \(20\% \mathrm{Yb}^{3 + } / 1\% \mathrm{Er}^{3 + }\) . As the temperature increases from 298 to 348 K, the intensity and position of these peaks nearly remain unchanged, which reveals that the water molecules were not removed from the microchannels. As the temperature increases from 348 to \(423~\mathrm{K}\) , these peaks become weaker. Above \(423~\mathrm{K}\) , the peak at \(945~\mathrm{cm^{- 1}}\) vanished, which demonstrated that water molecules are completely removed. These results are in agreement with the obtained results of the TG, FTIR spectra, as well as in situ temperature- dependent SXRD patterns (Supplementary Figures 3 and 4). Thus, it can be deduced that the temperature transition points of the onset and complete removal of water molecules from microchannels are \(\sim 348\) and \(\sim 423~\mathrm{K}\) , respectively. In addition, as the temperature increases, a Raman peak with a frequency of \(341~\mathrm{cm^{- 1}}\)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[147, 85, 850, 288]]<|/det|>
+\(\mathrm{(v_4)}\) exhibits a blue shift (Figure 2h), suggesting that this mode is the origin of NTE in \(\mathrm{SMO:Yb / Er}\) . The median frequency at \(510~\mathrm{cm^{- 1}}\) and \(579~\mathrm{cm^{- 1}}\) is induced by the disorder of \(\mathrm{MoO_4}\) tetrahedra because of the incorporation of the \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) into the lattice, which will affect the local structure of the doped \(\mathrm{Yb^{3 + }}\) and \(\mathrm{Er^{3 + }}\) ions. Note that the temperature coefficients of high phonon frequencies (511, 579, 813, \(979~\mathrm{cm^{- 1}}\) ) are negative above \(348~\mathrm{K}\) (Figures 2i and 2j), verifying strong anharmonic stretching/bending of \(\mathrm{MoO_4}\) tetrahedra. These results indicate that high phonon frequencies contribute also to the NTE behavior of the sample. Moreover, the reduction of maximum phonon energy \((979~\mathrm{cm^{- 1}})\) may benefit the luminescence of \(\mathrm{RE^{3 + }}\) at elevated temperatures due to the suppressed nonradiative relaxation. Please see also lines 1- 17 of p. 7 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 326, 260, 343]]<|/det|>
+## Comment #7  
+
+<|ref|>text<|/ref|><|det|>[[148, 353, 850, 390]]<|/det|>
+Can this effect be witnessed by the other techniques that would perceive the negative thermal expansion at the local scale? The authors could comment on that.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 391, 237, 408]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 417, 850, 510]]<|/det|>
+The negative thermal expansion at the local scale has also been witnessed by the other techniques like atomically resolved scanning tunnel microscope (STM) topographic image. We have cited the corresponding reference about the characterization of negative thermal expansion as Ref.13 (Phys. Rev. B, 101, 104305 (2020)) in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 550, 266, 566]]<|/det|>
+## Comment #8:  
+
+<|ref|>text<|/ref|><|det|>[[148, 577, 330, 593]]<|/det|>
+Along the manuscript:  
+
+<|ref|>text<|/ref|><|det|>[[147, 594, 850, 649]]<|/det|>
+What is the degree of Sc substitution in the samples that were measured in the presented results Figure 2,3,4,5? This is only specified in the labels of the Figures presented in the Supplementary material.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 660, 237, 677]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 687, 850, 741]]<|/det|>
+Many thanks for this valuable suggestion. We have provided the degree of Sc substitution (Sc substituted by \(20\%\) Yb and \(1\%\) Er) in the samples for Figures 2- 5 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 780, 266, 796]]<|/det|>
+## Comment #9:  
+
+<|ref|>text<|/ref|><|det|>[[147, 807, 850, 842]]<|/det|>
+line 18 "decrease at the higher temperature" author should substitute "decrease for high temperatures"  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 854, 237, 870]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[147, 880, 849, 899]]<|/det|>
+"decrease at the higher temperature" has been revised as "decrease for high
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[148, 84, 849, 120]]<|/det|>
+temperatures" in the revised manuscript. Please see also line 20 of p. 1 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 159, 277, 176]]<|/det|>
+## Comment #10:  
+
+<|ref|>text<|/ref|><|det|>[[148, 187, 848, 222]]<|/det|>
+Comment #10:line 58 and 103 Where is the abbreviation "NET" defined? Do the authors mean "NET"?  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 234, 238, 250]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 261, 849, 297]]<|/det|>
+Response:Thank you for pointing out this typo. The abbreviation "NET" has been revised as "NET" in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 336, 276, 352]]<|/det|>
+## Comment #11:  
+
+<|ref|>text<|/ref|><|det|>[[148, 363, 673, 380]]<|/det|>
+Comment #11:Line 121- 122: What is the value of Xc used for the calculations?  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 392, 238, 408]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 418, 850, 491]]<|/det|>
+Response:According to the second reviewer's suggestion, the ionic distance was updated, which was determined by the crystallographic information file (CIF) of Rietveld refinement on the basis of the SXRD patterns. Therefore, Equation 1 used for the calculation has been deleted.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 530, 277, 546]]<|/det|>
+## Comment #12:  
+
+<|ref|>text<|/ref|><|det|>[[147, 556, 850, 649]]<|/det|>
+Comment #12:Line 133 - 135 "The observed mode with a frequency of \(341 \mathrm{cm}^{- 1}\) corresponds to the bending mode of MoO4 tetrahedra. When the temperature increases, this Raman peak exhibits a blue shift (Figure 2f). The median frequency B1g mode of SMO is observed at \(511 \mathrm{cm}^{- 1}\) . Is this a typo? Figure 2f is not from the \(341 \mathrm{cm}^{- 1}\) mode. The sentence should be improved.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 661, 238, 677]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 688, 849, 742]]<|/det|>
+Response:Thank you for pointing out this typo. "this Raman peak exhibits a blue shift (Figure 2f)" has been revised as "this Raman peak exhibits a blue shift (Figure 2h)". Please see also line 10 of p. 7 in the revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[148, 781, 277, 797]]<|/det|>
+## Comment #13:  
+
+<|ref|>text<|/ref|><|det|>[[148, 808, 850, 899]]<|/det|>
+Comment #13:The authors refer the studied systems as "the structures of the SMO:Yb/Er samples with different dopant concentrations", doping generally refers to a small concentrations, here the degree of ionic Sc substitution by Yb and Er cations reaches the high value of \(25\%\) or \(8\%\) . The term chemical substitution would be more appropriated.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[148, 86, 237, 102]]<|/det|>
+## Response:  
+
+<|ref|>text<|/ref|><|det|>[[148, 112, 850, 167]]<|/det|>
+Thanks. "the structures of the SMO:Yb/Er samples with different dopant concentrations" has been revised as "the structures of the SMO:Yb/Er samples with different substituted concentrations".
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 309, 100]]<|/det|>
+REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[118, 144, 404, 161]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 204, 866, 240]]<|/det|>
+The authors responded correctly to my comments. The work does not contain any factual errors. In my opinion, it can be accepted for publication in its current form  
+
+<|ref|>text<|/ref|><|det|>[[118, 314, 404, 330]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 374, 835, 409]]<|/det|>
+The authors have adequately addressed all of the reviewer's comments, thus I recommend this article for publication.  
+
+<|ref|>text<|/ref|><|det|>[[118, 484, 404, 500]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 544, 580, 589]]<|/det|>
+The paper is much improved, and it addresses the issues that concerned me. In my opinion is ready for publication.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d632a7173d2acad65ad15f0c25ed840e543186c806052efca27d2b9a9a06896a/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d632a7173d2acad65ad15f0c25ed840e543186c806052efca27d2b9a9a06896a/supplementary_0_Peer Review File__d632a7173d2acad65ad15f0c25ed840e543186c806052efca27d2b9a9a06896a.mmd

@@ -0,0 +1,315 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+A Deep Learning Approach for Complex Microstructure Inference  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  
+
+Remarks to the Author:  
+
+Overall, I find the manuscript ready for submission. The authors have taken enough care in ensuring that the domain knowledge of materials scientists is successfully ported to use with emerging DL techniques. The sample preparation steps are extensively described and the model architecture and the training techniques are discussed well. Though I am not an expert in material science and cannot judge the novelty from that point of view, I do find the work to be a good introductory application of DL to the material sciences domain. I have a few questions for the authors as listed below:  
+
+Pg. 9: "In our study, we successfully trained both random initialized networks and pre- trained networks with comparatively small data sets of approximately 50 and 30 images for LOM and SEM, respectively. This invalidates the general claim of DL being only applicable for large- scale data sets" -- Note that the validation sets in both cases are extremely small compared to DL standards. The authors validate the current work with LOO cross validation which does not reveal the true ability of the model to generalize, in that the model could be overfitting to the limited evidence. There seem to be not much distribution shift between the training and test sets.  
+
+Vanilla U- net scoring better then U- Net VGG16:  
+
+1. It is possible that this is the result of the domain of imagenet being very different than the materials. Have the authors tried finetuning more layers of the pretrained network instead of only the final layer? There is usually a tradeoff in terms of the number of layers finetuned vs the resulting performance.  
+
+2. The authors must apply the focal loss class balancing to finetune the pretrained model. Keep the loss fixed and change the models alone to verify the difference.  
+
+3. The difference between LOM/SEM cases is puzzling given that the difference in # of data points isn't by a large factor.  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+The paper addresses the problem of highlighting phases in (optical and SEM) micrographs of compound materials. The uneven distribution of grey- levels into each phase defies segmentation algorithms that rely solely on pixel intensity. The authors adopt two slightly different deep learning approaches and compare them against manually- annotated golden standard. The main objective is to show that U- Net architectures are a reliable tool to solve such complex problems with no need of fine tuning. The authors claim that this is demonstrated by the very fact that different flavors of these architectures yield similar results and that performance variations mostly depend on data manipulation (e.g., tiling and padding) rather than on architectural differences. Indeed, the results shown by the authors are quite good for the problem at hand, even more so due to the reduced size of the training set (quite common in material sciences).  
+
+This paper does not describe a novel approach to image segmentation. However, the multidisciplinary approach of this work to the segmentation problem, encompassing sample preparation, image acquisition and image segmentation, is certainly significant (and, I would say, uncommon). It points towards a direction that all future research in this field should follow. The paper thoroughly addresses several open questions about segmentation and deep- learning, including the interpretation of the inner activity of the network, that is, one of the most controversial aspects of deep learning approaches. It is also interesting in one more respect: two research groups work on the same problem to demonstrate the validity of a framework, rather than the quality of a single approach. In this sense, I think that this paper is a useful contribution to the scientific debate about the role of deep learning in industrial applications.  
+
+The paper clearly presents speculative thinking about each and every step of network training and operativity. The discussion is supported by experimental evidence and a sufficiently rigorous analysis of data. The description is reasonably detailed. A few details could be added, e.g., about the operators used in the network to process images as well as about the size of padding and overlap between image tiles. It would be interesting to compare the size of padding to a
+
+<--- Page Split --->
+
+representative measure of the size of phases.  
+
+The conclusions are consistent with experimental data. I would suggest to think a little bit more about a few statements, such as the claim that the small size of the training dataset used in experiments, compared to the quality of results, "invalidates the general claim of DL being only applicable for large- scale data sets". This is certainly true for the dataset used in this paper. However, this can well prove false for other data sets. The successive claim that the availability of reproducible, high- quality imaging would reduce the need for a large amount of training data appears a leap of logic.  
+
+At the end of page 9, the authors write that the models fail in the very same regions where human expert make mistakes during manual annotation. No further explanation is given in order to better understand the method used to locate these regions. Do the authors refer, e.g., to pixel regions for which manual annotations do not match?  
+
+Thinking about the possibility to generalize the results shown in the paper, I would appreciate a deeper discussion about how much the performance of DL methods is affected by the scale of phases in the image. Similarly, in order to accept the claim that these methods are generalizable, I would like to see a few comparative experiments with radically different datasets.  
+
+Finally, the authors do not address execution time. This is an important aspect for industrial applications, especially in manufacturing.  
+
+Reviewer #3:  
+
+Remarks to the Author:  
+
+This paper purports to lay out a holistic approach to deploying deep learning for complex microstructure inference. This is a very topical and challenging problem, and I really wanted to like this paper. However the paper as it stands does not accomplish any of the key claims. I lay out my reasons below:  
+
+1) Claim of impact: Showing segmentation on one class of microstructures is not enough (according to this reviewer). This is insufficient to illustrate their central claim of "an intuition about the required data quality and quantity and an extensive methodological DL guideline for microstructure quantification and classification are still missing". No intuition is gained given the limited number of models used, and the limited data used (see point #3).  
+
+2) There are several other real-world images (for instance, MRI images) that are as (if not more) complex than microstructure images, and a lot more critical to segment correctly. This goes right to the heart of the statement "Furthermore, microstructure recognition tasks, compared to real-world images, can be very complex regarding ..."  
+
+3) Data size: The datasize used (30-40) images makes the results of this study highly suspect. It was not clear to me how many images were finally used after augmentation and windowing, but training a U-Net from scratch is not a good idea with this few images. See for instance https://arxiv.org/pdf/2001.05566.pdf.  
+
+4) Transfer learning: As the author's comment in outlook section, using pre-training with self- or semi-supervised learning is the way to go with such small datasets. The marginal improvements of using a pre-trained U-Net is indeed indicative of this. See for instance "Zhuang F, Qi Z, Duan K, Xi D, Zhu Y, Zhu H, Xiong H, He Q. A comprehensive survey on transfer learning. Proceedings of the IEEE. 2020 Jul 7;109(1):43-76."  
+
+5) Transparent decision making: Grad-CAM has been shown to be a very poor explainability mechanism. See for instance https://arxiv.org/abs/1812.02843 where GradCAM giver poor explainability even when the model predictions are accurate.  
+
+6) Intra- and Inter-rater variability: This was an opportunity lost to discuss and provide intuition on inter- and intra- rater variability during the human annotation. That is, do changes in the fg and bg as marked by experts significantly impact results?  
+
+7) Metrics: Using accuracy is a bad metric. In some cases, the pixel count of fp to bg is 1:3 or 1:4
+
+<--- Page Split --->
+
+which always segmenting as bg will give \(75 - 80\%\) accuracy. A \(10\%\) improvement seems marginal. I encourage sticking to IoU as the key figure of merit.  
+
+8) Lack of detailed analysis: While the authors do a good job of performing ablation studies, this was an opportunity lost with showing extensive analysis that could back up (at least anecdotally) some of their claims. Specifically, what happens when depth of the U-net is changed, what happens when window size (and hence data size) changes, what happens when total data set is further reduced, how is pixel size, window size and resolution related to performance. All these would add value and generate more intuition for practitioners.  
+
+9) Data sharing: Without the availability of annotated data for the broader audience to try and evaluate, replicate and improve on these (standard) methods, I see the impact of this paper as minimal.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+Overall, I find the manuscript ready for submission. The authors have taken enough care in ensuring that the domain knowledge of materials scientists is successfully ported to use with emerging DL techniques. The sample preparation steps are extensively described and the model architecture and the training techniques are discussed well. Though I am not an expert in material science and cannot judge the novelty from that point of view, I do find the work to be a good introductory application of DL to the material sciences domain. I have a few questions for the authors as listed below:  
+
+Pg. 9: "In our study, we successfully trained both random initialized networks and pre- trained networks with comparatively small data sets of approximately 50 and 30 images for LOM and SEM, respectively. This invalidates the general claim of DL being only applicable for large- scale data sets" - - Note that the validation sets in both cases are extremely small compared to DL standards. The authors validate the current work with LOO cross validation which does not reveal the true ability of the model to generalize, in that the model could be overfitting to the limited evidence. There seem to be not much distribution shift between the training and test sets.  
+
+## Author response:  
+
+Thank you for your efforts in reviewing our work.  
+
+You are right that the validation sets are small. The datasets were acquired in a very reproducible fashion and do not contain the typical variances that are otherwise introduced in metallography through multiple operators, multiple microscopes or different etchings. A strong indication is that the applied data augmentation did not improve the performance substantially despite the small size of the dataset.  
+
+In our section on ,,Variances and generalization" on page 11 we address this:  
+
+, In instances where such material- extrinsic variance can be ensured to be insignificant, data augmentation through simple spatial (affine and even elastic) or intensity transformations can be evaded. Therefore, such models trained on comparatively small data sets are suitable for tasks with inherently small scatter, such as quality inspection, where recurring tasks and predefined workflows are set. When, for instance, etching- based contrasting methodologies are concerned, reproducibility can be difficult to attain."  
+
+Subsequently we show that the model fails when we apply it (which achieved good accuracy in the source domain) on an alternate set of micrographs where material was etched with different parameters. As we highlight in the manuscript, this can be ascribed to the low- quantity data with small material processing or imaging- induced variance. We show that if we use brightness and contrast augmentation, the generalization with respect to the otherwise etched domain improves. We want to stress that the dataset was not optimized for archiving generalization over a wide range of processing routes, materials or similar but to perform an ablation study helping material scientists to get an idea about how different approaches (hyperparameter and rescaling and so on) affects the training.  
+
+Vanilla U- net scoring better then U- Net VGG16:  
+
+1. It is possible that this is the result of the domain of imagenet being very different than the materials. Have the authors tried finetuning more layers of the pretrained network instead of only the final layer? There is usually a tradeoff in terms of the number of layers finetuned vs the resulting performance.  
+
+2. The authors must apply the focal loss class balancing to finetune the pretrained model. Keep the loss fixed and change the models alone to verify the difference.  
+
+3. The difference between LOM/SEM cases is puzzling given that the difference in # of data points isn't by a large factor.  
+
+Author response: Add 1:
+
+<--- Page Split --->
+
+That the U- Net scoring is better than the pretrained U- Net VGG16 in the optical light microscopy case was in fact an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.  
+
+## Add 2:  
+
+Thank you for your comment on fine- tuning. We did finetune over the full network in this case since the initial submission. We did add a remark on how we finetuned in the methods section.  
+
+Despite identifying this as an artifact, we did implement the identical loss function for the VGG16 network and the fine tuning training according to your remark. However, assimilating the loss functions did not affect the results significantly. We added this as a statement in the results and added more notes on the loss functions in the methods section applied.  
+
+Regarding point 3: In fact, the data amount is substantially different (not only the image amount 51. vs 36) but especially the physical image size resolution and the characterized area in \(\mu \mathrm{m}^2\) differs significantly. As the features have the same size independent of the microscopy methodology, the LOM contains a much larger number of grains. Therefore, we argue that the tendencies are well explainable. In the light optical microscopy micrographs, the effect of pre- training is smaller since more data is available, while in the SEM models the pretraining has a major impact.
+
+<--- Page Split --->
+
+Reviewer #2 (Remarks to the Author):  
+
+The paper addresses the problem of highlighting phases in (optical and SEM) micrographs of compound materials. The uneven distribution of grey- levels into each phase defies segmentation algorithms that rely solely on pixel intensity. The authors adopt two slightly different deep learning approaches and compare them against manually- annotated golden standard. The main objective is to show that U- Net architectures are a reliable tool to solve such complex problems with no need of fine tuning. The authors claim that this is demonstrated by the very fact that different flavors of these architectures yield similar results and that performance variations mostly depend on data manipulation (e.g., tiling and padding) rather than on architectural differences. Indeed, the results shown by the authors are quite good for the problem at hand, even more so due to the reduced size of the training set (quite common in material sciences).  
+
+This paper does not describe a novel approach to image segmentation. However, the multidisciplinary approach of this work to the segmentation problem, encompassing sample preparation, image acquisition and image segmentation, is certainly significant (and, I would say, uncommon). It points towards a direction that all future research in this field should follow. The paper thoroughly addresses several open questions about segmentation and deep- learning, including the interpretation of the inner activity of the network, that is, one of the most controversial aspects of deep learning approaches. It is also interesting in one more respect: two research groups work on the same problem to demonstrate the validity of a framework, rather than the quality of a single approach. In this sense, I think that this paper is a useful contribution to the scientific debate about the role of deep learning in industrial applications.  
+
+The paper clearly presents speculative thinking about each and every step of network training and operativity. The discussion is supported by experimental evidence and a sufficiently rigorous analysis of data. The description is reasonably detailed.  
+
+A few details could be added, e.g., about the operators used in the network to process images as well as about the size of padding and overlap between image tiles. It would be interesting to compare the size of padding to a representative measure of the size of phases.  
+
+The conclusions are consistent with experimental data. I would suggest to think a little bit more about a few statements, such as the claim that the small size of the training dataset used in experiments, compared to the quality of results, "invalidates the general claim of DL being only applicable for large- scale data sets". This is certainly true for the dataset used in this paper. However, this can well prove false for other data sets. The successive claim that the availability of reproducible, high- quality imaging would reduce the need for a large amount of training data appears a leap of logic.  
+
+## Author response:  
+
+Thank you for your detailed and positive review. The interdisciplinary collaboration is a fundamental necessity to advance the implementation of DL into the materials science area. Furthermore, the industrial usage requires convincing real world examples and usable frameworks. As you also pointed out, the real world of materials science and engineering works on a lot of small data cases and therefore we have targeted this as a priority.  
+
+Nevertheless, we would like to point out that fine tuning of pre- trained networks improves performance. In fact, we believe that in this low data regime, that we are in right now, pre- training and fine- tuning are of substantial relevance. In both LOM and especially SEM we show that an improvement through pretraining can be achieved. We enhanced our manuscript in the discussion to assure not to being suggestive of pretraining being unuseful.  
+
+"invalidates the general claim of DL being only applicable for large- scale data sets" \(\rightarrow\) We used "general" here to emphasize that it is a preconception that many (material) scientists have, who are not involved with data- driven techniques. We believe that in the setting of small variance data (reproducible image acquisition) fewer data is necessary to train a model for that domain. However, we observed that such a model won't be great at generalizing if trained with small processing- induced variance data. We understand this and hence the successive remark has been adjusted in the discussion.  
+
+At the end of page 9, the authors write that the models fail in the very same regions where human expert make mistakes during manual annotation. No further explanation is given in order to better understand the method used to locate these regions. Do the authors refer, e.g., to pixel regions for which manual annotations do not match?
+
+<--- Page Split --->
+
+## Author response:  
+
+You are right, we added few sentences here to render in more clear in the discussion.  
+
+Thinking about the possibility to generalize the results shown in the paper, I would appreciate a deeper discussion about how much the performance of DL methods is affected by the scale of phases in the image. Similarly, in order to accept the claim that these methods are generalizable, I would like to see a few comparative experiments with radically different datasets.  
+
+## Author response:  
+
+We added a study on relative scale of phases and image size to the manuscript. Specifically we conducted another study where we altered the size of the tiles in LOM (rather than downscaling so we could exclude information loss and receptive field based effects). This study gives more insights on how to choose the tile sizes ideally, as it correlates tile sizes with lath width and lath- bainite region size distributions. Accordingly, we extended the results (Image context and network receptive field dependency section) and discussion (Image context and network receptive field dependency section).  
+
+On the other hand, we did not claim that the model trained with this low- variance dataset is generalizable across data sets. In fact, we did apply the trained models to an otherwise etched (overetched) image and observed that the model is not so good at generalizing across domains (see section ,,Variances and generalization"). Even when adjusting and optimizing the data augmentation to lower the domain gap towards a target domain, the model does not generalize particularly well. We stress that, our objective in this paper was not to train a perfect model for cross- domain generalization but rather to highlight image processing effects (scaling, tiling...) and pretraining on DL performance as well as explainability. If generalization would have been our primary objective, we would have acquired our dataset with a more natural variance and would have applied stronger data augmentation.  
+
+Finally, the authors do not address execution time. This is an important aspect for industrial applications, especially in manufacturing.  
+
+## Author response:  
+
+We added notes on execution time both for training and for deployment in the Supplemental. We added number of trained epochs and specified hardware as well in the methods section (Deep learning segmentation approach section).  
+
+## Another important note to the reviewer:  
+
+The U- Net scoring better than the pretrained U- Net VGG16 in the light microscopy case was an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.
+
+<--- Page Split --->
+
+Reviewer #3 (Remarks to the Author):  
+
+This paper purports to lay out a holistic approach to deploying deep learning for complex microstructure inference. This is a very topical and challenging problem, and I really wanted to like this paper. However the paper as it stands does not accomplish any of the key claims. I lay out my reasons below:  
+
+1) Claim of impact: Showing segmentation on one class of microstructures is not enough (according to this reviewer). This is insufficient to illustrate their central claim of "an intuition about the required data quality and quantity and an extensive methodological DL guideline for microstructure quantification and classification are still missing". No intuition is gained given the limited number of models used, and the limited data used (see point #3).  
+
+## Author response:  
+
+Thank you for your efforts in reviewing our work.  
+
+You are right, the claim is too strong. Nevertheless, this manuscript clearly indicates that data quantities of 30- 50 images suffice to train a segmentation network for this low material- extrinsic variance domain. Please note that both data sets have quite different data amounts, since, aside from the image number, individual LOM images capture a substantially larger physical area. In addition, when it comes to image rescaling and pretraining, we firmly believe that our experiments after the corrections provide a valid indication. Namely, that increasing the context and receptive field helps when dealing with such long- range features and by how much pretraining is helping in both distinct data quantity settings (low- quantitativ datasets are very typical for materials science data sets). It is true that we do not perform a study on data quality in this manuscript. We removed this claim and softened the claim concerning data quantity. We adjusted the abstract and introduction accordingly.  
+
+2) There are several other real-world images (for instance, MRI images) that are as (if not more) complex than microstructure images, and a lot more critical to segment correctly. This goes right to the heart of the statement "Furthermore, microstructure recognition tasks, compared to real-world images, can be very complex regarding ..."  
+
+## Author response:  
+
+By real- world images we are rather refering to street scenes (Cityscapes) or Image- Net, and similar. We slightly modified the wording from ,,real- world images" to ,,natural images" and added Image- Net as an example in the introduction to cause less confusion.  
+
+3) Data size: The datasize used (30-40) images makes the results of this study highly suspect. It was not clear to me how many images were finally used after augmentation and windowing, but training a U-Net from scratch is not a good idea with this few images. See for instance https://arxiv.org/pdf/2001.05566. pdf .  
+
+## Author response:  
+
+We refer to the number of individual training tile images in the Supplemental. On page 4, we state ,,A summary of the data sets, including some characteristic metrics, can be found in the Supplemental." We adjusted the wording here to make it more clear. The amount of training tiles varies depending on the preprocessing and the exact data set. We kept it in the Supplemental rather than the manuscript since it would add large tables to an already somewhat long manuscript and since we consider the individual tile amount as not as important as the overarching raw microscope image number from which the tiles originate. For your convinience: In the non- resized LOM training, we worked with approximately 600 tiles.  
+
+In terms of augmentation we did apply online augmentation which means the number of epochs (which we now added consistently in the Methods Deep Learning section) and the dataset sizes (given in the Supplemental) directly provide the number of images which the network has seen.  
+
+We are aware that pretraining is a common and established practice in deep learning. However, we still wanted to make the comparison between ImageNet pretrained networks and random initialized ones. Finding
+
+<--- Page Split --->
+
+appropriate pretraining large-scale data sets with a small domain gap to the target task is difficult in materials science, which is why many practitioners do not apply pretraining currently.  
+
+4) Transfer learning: As the author's comment in outlook section, using pre-training with self- or semisupervised learning is the way to go with such small datasets. The marginal improvements of using a pretrained U-Net is indeed indicative of this. See for instance "Zhuang F, Qi Z, Duan K, Xi D, Zhu Y, Zhu H, Xiong H, He Q. A comprehensive survey on transfer learning. Proceedings of the IEEE. 2020 Jul 7;109(1):43-76."  
+
+## Author response:  
+
+Thank you for sharing this information. In follow- up publications we aim to include semi- supervised learning and unsupervised domain adaptation more.  
+
+5) Transparent decision making: Grad-CAM has been shown to be a very poor explainability mechanism. See for instance https://arxiv.org/abs/1812.02843 where GradCAM giver poor explainability even when the model predictions are accurate.  
+
+## Author response:  
+
+It might be true that GradCAM, as other visualization techniques, is prone to such adversarial attacks. However, GradCAM was shown to provide comparatively sensible visualizations here (→) when exposed to sanity checks.  
+
+Adebayo, Julius, et al. "Sanity checks for saliency maps." arXiv preprint arXiv:1810.03292 (2018). Sanity Checks for Saliency Maps (nips.cc)  
+
+Similarly, for our results the visualizations agree with expert annotators expectations. From our point of view, adversarial attacks are not relevant in our setting.  
+
+6) Intra- and Inter-rater variability: This was an opportunity lost to discuss and provide intuition on inter- and intra- rater variability during the human annotation. That is, do changes in the fg and bg as marked by experts significantly impact results?  
+
+## Author response:  
+
+We expect there be to substantially less rater-variability in annotation as annotation was assisted by correlative EBSD (orientation-sensitive data). We did an experiment in the beginning of our collaboration where we considered two data sets:  
+
+data set where we discarded tiles with uncertain regions data set with all tiles  
+
+We did not see a major difference in performance. This we added as a remark in the manuscript in the result section.  
+
+7) Metrics: Using accuracy is a bad metric. In some cases, the pixel count of fp to bg is 1:3 or 1:4 which always segmenting as bg will give 75-80% accuracy. A 10% improvement seems marginal. I encourage sticking to IoU as the key figure of merit.  
+
+## Author response:  
+
+Initially we wanted to provide accuracy as an additional metric because it is more intuitive than IoU. However, as you mention we have class imbalance in the light microscopy data set as it is representative for the microstructure. Therefore, omitting true negatives in IoU renders this metric more sensitive. Even though we explained these aspects in the paper for our materials science readers, we removed the accuracy values from the tables and just added a single statement in the result section that the IoUs correspond to approximately X accuracy.  
+
+Side note: We have in average 73% percent background in the light microscopy, so it is almost 20% increase over the naive baseline.  
+
+8) Lack of detailed analysis: While the authors do a good job of performing ablation studies, this was an
+
+<--- Page Split --->
+
+opportunity lost with showing extensive analysis that could back up (at least anecdotally) some of their claims. Specifically, what happens when depth of the U- net is changed, what happens when window size (and hence data size) changes, what happens when total data set is further reduced, how is pixel size, window size and resolution related to performance. All these would add value and generate more intuition for practitioners.  
+
+## Author response:  
+
+We added a study on relative scale of phases and image. To be specific, we conducted another study where we altered the size of the cropped LOM tiles (rather than downscaling, so we could exclude information loss and receptive field based effects). This addresses your window size remark. This study gives more insights on how to choose the tile sizes ideally as we compare the tile sizes with characteristic length scales (lath width or region size). Accordingly, we extended the results and discussion.  
+
+9) Data sharing: Without the availability of annotated data for the broader audience to try and evaluate, replicate and improve on these (standard) methods, I see the impact of this paper as minimal.  
+
+## Author response:  
+
+Although we would like to share the data to fuel a similar multiplication effect as observed in other domains, in this case we are limited due to industry involvement. The data is part of an ongoing study.  
+
+## Another important note to the reviewer:  
+
+The U- Net scoring better than the pretrained U- Net VGG16 in the light microscopy case was an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1: None  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+The authors addressed all the issues and concerns in my review. For that part (i.e., the point of view of a user), I think that their article has been improved with respect to the initial manuscript. If other reviewers share this thought, I think that the manuscript is ready for publication.  
+
+Just one more comment on fine- tuning that is not intended to influence the discourse in final paper. Perhaps, the authors will find it a useful thought for further investigation.  
+
+I believe that one should be very careful with tuning, at least for two reasons:  
+
+1. it often requires care, time, and competence that industrial applications cannot afford  
+
+2. there is always a risk of overtraining, especially with small datasets  
+
+Reviewer #3:  
+
+Remarks to the Author:  
+
+I thank the authors for the additional effort in responding to (some of) my questions. The manuscript has improved with the additional work, as well as the thoughtful watering down of some of the claims.  
+
+However, I do note that the watering down of several claims does bring down the overall impact of the paper. Several of my concerns (#2,#3,#7,#9) remain unanswered.  
+
+I remain unsatisfied with the key claims, especially when there are similar reports from other fields (for instance, the medical field) where complex images are successfully segmented. I also remain concerned with the small dataset used here. Finally, while understandable, the unavailability of the data makes comparative assessment impossible, further diminishing the impact.
+
+<--- Page Split --->
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors addressed all the issues and concerns in my review. For that part (i.e., the point of view of a user), I think that their article has been improved with respect to the initial manuscript. If other reviewers share this thought, I think that the manuscript is ready for publication.  
+
+Just one more comment on fine- tuning that is not intended to influence the discourse in final paper. Perhaps, the authors will find it a useful thought for further investigation. I believe that one should be very careful with tuning, at least for two reasons: 1. it often requires care, time, and competence that industrial applications cannot afford 2. there is always a risk of overtraining, especially with small datasets  
+
+Thank you for your review and remarks. We agree with your comments. In our case, fine- tuning was applied in a careful way (with optimized learning rates) and overfitting was not observed in the learning curves.  
+
+Reviewer #3 (Remarks to the Author):  
+
+I thank the authors for the additional effort in responding to (some of) my questions. The manuscript has improved with the additional work, as well as the thoughtful watering down of some of the claims.  
+
+However, I do note that the watering down of several claims does bring down the overall impact of the paper. Several of my concerns (#2,#3,#7,#9) remain unanswered.  
+
+I remain unsatisfied with the key claims, especially when there are similar reports from other fields (for instance, the medical field) where complex images are successfully segmented. I also remain concerned with the small dataset used here. Finally, while understandable, the unavailability of the data makes comparative assessment impossible, further diminishing the impact.  
+
+Thank you for your review and remarks. We agree that some works in other domains (e.g. medical) share similar observations with our work. However, we cited some of those in our manuscript, e.g.  
+
+Sabottke, C. F. & Spieler, B. M. The Effect of Image Resolution on Deep Learning in Radiography. Radiol. Artif. Intell.2, e190015, DOI: 10.1148/ryai.2019190015 (2020).  
+
+We believe that the observed scatter in the data is predominantly materials' microstructure- based, since image acquisition was performed in a very reproducible fashion. Under this assumption and considering the characteristic sizes of microstructural features (such as lath width and grain sizes), it is very likely that the imaged area is representative of the microstructural scatter. We added this in a few sentences to make clear, why we achieve good results despite the low data quantity.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d632a7173d2acad65ad15f0c25ed840e543186c806052efca27d2b9a9a06896a/supplementary_0_Peer Review File__d632a7173d2acad65ad15f0c25ed840e543186c806052efca27d2b9a9a06896a_det.mmd

@@ -0,0 +1,444 @@
+<|ref|>title<|/ref|><|det|>[[61, 41, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[70, 111, 361, 140]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[70, 155, 826, 210]]<|/det|>
+A Deep Learning Approach for Complex Microstructure Inference  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 782]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 911, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 84, 293, 97]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 113, 225, 125]]<|/det|>
+Reviewer #1:  
+
+<|ref|>text<|/ref|><|det|>[[119, 127, 300, 140]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[118, 140, 875, 238]]<|/det|>
+Overall, I find the manuscript ready for submission. The authors have taken enough care in ensuring that the domain knowledge of materials scientists is successfully ported to use with emerging DL techniques. The sample preparation steps are extensively described and the model architecture and the training techniques are discussed well. Though I am not an expert in material science and cannot judge the novelty from that point of view, I do find the work to be a good introductory application of DL to the material sciences domain. I have a few questions for the authors as listed below:  
+
+<|ref|>text<|/ref|><|det|>[[118, 252, 877, 351]]<|/det|>
+Pg. 9: "In our study, we successfully trained both random initialized networks and pre- trained networks with comparatively small data sets of approximately 50 and 30 images for LOM and SEM, respectively. This invalidates the general claim of DL being only applicable for large- scale data sets" -- Note that the validation sets in both cases are extremely small compared to DL standards. The authors validate the current work with LOO cross validation which does not reveal the true ability of the model to generalize, in that the model could be overfitting to the limited evidence. There seem to be not much distribution shift between the training and test sets.  
+
+<|ref|>text<|/ref|><|det|>[[118, 364, 483, 378]]<|/det|>
+Vanilla U- net scoring better then U- Net VGG16:  
+
+<|ref|>text<|/ref|><|det|>[[118, 378, 875, 435]]<|/det|>
+1. It is possible that this is the result of the domain of imagenet being very different than the materials. Have the authors tried finetuning more layers of the pretrained network instead of only the final layer? There is usually a tradeoff in terms of the number of layers finetuned vs the resulting performance.  
+
+<|ref|>text<|/ref|><|det|>[[116, 435, 878, 461]]<|/det|>
+2. The authors must apply the focal loss class balancing to finetune the pretrained model. Keep the loss fixed and change the models alone to verify the difference.  
+
+<|ref|>text<|/ref|><|det|>[[116, 461, 870, 490]]<|/det|>
+3. The difference between LOM/SEM cases is puzzling given that the difference in # of data points isn't by a large factor.  
+
+<|ref|>text<|/ref|><|det|>[[118, 533, 222, 546]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[118, 547, 300, 560]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[117, 560, 878, 701]]<|/det|>
+The paper addresses the problem of highlighting phases in (optical and SEM) micrographs of compound materials. The uneven distribution of grey- levels into each phase defies segmentation algorithms that rely solely on pixel intensity. The authors adopt two slightly different deep learning approaches and compare them against manually- annotated golden standard. The main objective is to show that U- Net architectures are a reliable tool to solve such complex problems with no need of fine tuning. The authors claim that this is demonstrated by the very fact that different flavors of these architectures yield similar results and that performance variations mostly depend on data manipulation (e.g., tiling and padding) rather than on architectural differences. Indeed, the results shown by the authors are quite good for the problem at hand, even more so due to the reduced size of the training set (quite common in material sciences).  
+
+<|ref|>text<|/ref|><|det|>[[117, 701, 870, 842]]<|/det|>
+This paper does not describe a novel approach to image segmentation. However, the multidisciplinary approach of this work to the segmentation problem, encompassing sample preparation, image acquisition and image segmentation, is certainly significant (and, I would say, uncommon). It points towards a direction that all future research in this field should follow. The paper thoroughly addresses several open questions about segmentation and deep- learning, including the interpretation of the inner activity of the network, that is, one of the most controversial aspects of deep learning approaches. It is also interesting in one more respect: two research groups work on the same problem to demonstrate the validity of a framework, rather than the quality of a single approach. In this sense, I think that this paper is a useful contribution to the scientific debate about the role of deep learning in industrial applications.  
+
+<|ref|>text<|/ref|><|det|>[[118, 842, 870, 910]]<|/det|>
+The paper clearly presents speculative thinking about each and every step of network training and operativity. The discussion is supported by experimental evidence and a sufficiently rigorous analysis of data. The description is reasonably detailed. A few details could be added, e.g., about the operators used in the network to process images as well as about the size of padding and overlap between image tiles. It would be interesting to compare the size of padding to a
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[120, 84, 466, 97]]<|/det|>
+representative measure of the size of phases.  
+
+<|ref|>text<|/ref|><|det|>[[118, 98, 874, 196]]<|/det|>
+The conclusions are consistent with experimental data. I would suggest to think a little bit more about a few statements, such as the claim that the small size of the training dataset used in experiments, compared to the quality of results, "invalidates the general claim of DL being only applicable for large- scale data sets". This is certainly true for the dataset used in this paper. However, this can well prove false for other data sets. The successive claim that the availability of reproducible, high- quality imaging would reduce the need for a large amount of training data appears a leap of logic.  
+
+<|ref|>text<|/ref|><|det|>[[118, 197, 877, 252]]<|/det|>
+At the end of page 9, the authors write that the models fail in the very same regions where human expert make mistakes during manual annotation. No further explanation is given in order to better understand the method used to locate these regions. Do the authors refer, e.g., to pixel regions for which manual annotations do not match?  
+
+<|ref|>text<|/ref|><|det|>[[118, 252, 878, 308]]<|/det|>
+Thinking about the possibility to generalize the results shown in the paper, I would appreciate a deeper discussion about how much the performance of DL methods is affected by the scale of phases in the image. Similarly, in order to accept the claim that these methods are generalizable, I would like to see a few comparative experiments with radically different datasets.  
+
+<|ref|>text<|/ref|><|det|>[[118, 308, 825, 336]]<|/det|>
+Finally, the authors do not address execution time. This is an important aspect for industrial applications, especially in manufacturing.  
+
+<|ref|>text<|/ref|><|det|>[[118, 378, 225, 391]]<|/det|>
+Reviewer #3:  
+
+<|ref|>text<|/ref|><|det|>[[118, 393, 300, 406]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[118, 407, 878, 462]]<|/det|>
+This paper purports to lay out a holistic approach to deploying deep learning for complex microstructure inference. This is a very topical and challenging problem, and I really wanted to like this paper. However the paper as it stands does not accomplish any of the key claims. I lay out my reasons below:  
+
+<|ref|>text<|/ref|><|det|>[[118, 476, 858, 546]]<|/det|>
+1) Claim of impact: Showing segmentation on one class of microstructures is not enough (according to this reviewer). This is insufficient to illustrate their central claim of "an intuition about the required data quality and quantity and an extensive methodological DL guideline for microstructure quantification and classification are still missing". No intuition is gained given the limited number of models used, and the limited data used (see point #3).  
+
+<|ref|>text<|/ref|><|det|>[[118, 560, 864, 616]]<|/det|>
+2) There are several other real-world images (for instance, MRI images) that are as (if not more) complex than microstructure images, and a lot more critical to segment correctly. This goes right to the heart of the statement "Furthermore, microstructure recognition tasks, compared to real-world images, can be very complex regarding ..."  
+
+<|ref|>text<|/ref|><|det|>[[118, 630, 864, 686]]<|/det|>
+3) Data size: The datasize used (30-40) images makes the results of this study highly suspect. It was not clear to me how many images were finally used after augmentation and windowing, but training a U-Net from scratch is not a good idea with this few images. See for instance https://arxiv.org/pdf/2001.05566.pdf.  
+
+<|ref|>text<|/ref|><|det|>[[118, 700, 878, 771]]<|/det|>
+4) Transfer learning: As the author's comment in outlook section, using pre-training with self- or semi-supervised learning is the way to go with such small datasets. The marginal improvements of using a pre-trained U-Net is indeed indicative of this. See for instance "Zhuang F, Qi Z, Duan K, Xi D, Zhu Y, Zhu H, Xiong H, He Q. A comprehensive survey on transfer learning. Proceedings of the IEEE. 2020 Jul 7;109(1):43-76."  
+
+<|ref|>text<|/ref|><|det|>[[118, 785, 824, 826]]<|/det|>
+5) Transparent decision making: Grad-CAM has been shown to be a very poor explainability mechanism. See for instance https://arxiv.org/abs/1812.02843 where GradCAM giver poor explainability even when the model predictions are accurate.  
+
+<|ref|>text<|/ref|><|det|>[[118, 841, 860, 882]]<|/det|>
+6) Intra- and Inter-rater variability: This was an opportunity lost to discuss and provide intuition on inter- and intra- rater variability during the human annotation. That is, do changes in the fg and bg as marked by experts significantly impact results?  
+
+<|ref|>text<|/ref|><|det|>[[115, 896, 870, 910]]<|/det|>
+7) Metrics: Using accuracy is a bad metric. In some cases, the pixel count of fp to bg is 1:3 or 1:4
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 83, 878, 111]]<|/det|>
+which always segmenting as bg will give \(75 - 80\%\) accuracy. A \(10\%\) improvement seems marginal. I encourage sticking to IoU as the key figure of merit.  
+
+<|ref|>text<|/ref|><|det|>[[118, 125, 864, 210]]<|/det|>
+8) Lack of detailed analysis: While the authors do a good job of performing ablation studies, this was an opportunity lost with showing extensive analysis that could back up (at least anecdotally) some of their claims. Specifically, what happens when depth of the U-net is changed, what happens when window size (and hence data size) changes, what happens when total data set is further reduced, how is pixel size, window size and resolution related to performance. All these would add value and generate more intuition for practitioners.  
+
+<|ref|>text<|/ref|><|det|>[[118, 223, 850, 265]]<|/det|>
+9) Data sharing: Without the availability of annotated data for the broader audience to try and evaluate, replicate and improve on these (standard) methods, I see the impact of this paper as minimal.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[93, 65, 377, 80]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[92, 96, 905, 188]]<|/det|>
+Overall, I find the manuscript ready for submission. The authors have taken enough care in ensuring that the domain knowledge of materials scientists is successfully ported to use with emerging DL techniques. The sample preparation steps are extensively described and the model architecture and the training techniques are discussed well. Though I am not an expert in material science and cannot judge the novelty from that point of view, I do find the work to be a good introductory application of DL to the material sciences domain. I have a few questions for the authors as listed below:  
+
+<|ref|>text<|/ref|><|det|>[[92, 203, 901, 310]]<|/det|>
+Pg. 9: "In our study, we successfully trained both random initialized networks and pre- trained networks with comparatively small data sets of approximately 50 and 30 images for LOM and SEM, respectively. This invalidates the general claim of DL being only applicable for large- scale data sets" - - Note that the validation sets in both cases are extremely small compared to DL standards. The authors validate the current work with LOO cross validation which does not reveal the true ability of the model to generalize, in that the model could be overfitting to the limited evidence. There seem to be not much distribution shift between the training and test sets.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 325, 220, 339]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[93, 340, 463, 355]]<|/det|>
+Thank you for your efforts in reviewing our work.  
+
+<|ref|>text<|/ref|><|det|>[[93, 368, 904, 430]]<|/det|>
+You are right that the validation sets are small. The datasets were acquired in a very reproducible fashion and do not contain the typical variances that are otherwise introduced in metallography through multiple operators, multiple microscopes or different etchings. A strong indication is that the applied data augmentation did not improve the performance substantially despite the small size of the dataset.  
+
+<|ref|>text<|/ref|><|det|>[[97, 444, 664, 460]]<|/det|>
+In our section on ,,Variances and generalization" on page 11 we address this:  
+
+<|ref|>text<|/ref|><|det|>[[92, 474, 891, 551]]<|/det|>
+, In instances where such material- extrinsic variance can be ensured to be insignificant, data augmentation through simple spatial (affine and even elastic) or intensity transformations can be evaded. Therefore, such models trained on comparatively small data sets are suitable for tasks with inherently small scatter, such as quality inspection, where recurring tasks and predefined workflows are set. When, for instance, etching- based contrasting methodologies are concerned, reproducibility can be difficult to attain."  
+
+<|ref|>text<|/ref|><|det|>[[92, 566, 895, 690]]<|/det|>
+Subsequently we show that the model fails when we apply it (which achieved good accuracy in the source domain) on an alternate set of micrographs where material was etched with different parameters. As we highlight in the manuscript, this can be ascribed to the low- quantity data with small material processing or imaging- induced variance. We show that if we use brightness and contrast augmentation, the generalization with respect to the otherwise etched domain improves. We want to stress that the dataset was not optimized for archiving generalization over a wide range of processing routes, materials or similar but to perform an ablation study helping material scientists to get an idea about how different approaches (hyperparameter and rescaling and so on) affects the training.  
+
+<|ref|>text<|/ref|><|det|>[[92, 719, 454, 734]]<|/det|>
+Vanilla U- net scoring better then U- Net VGG16:  
+
+<|ref|>text<|/ref|><|det|>[[92, 735, 896, 780]]<|/det|>
+1. It is possible that this is the result of the domain of imagenet being very different than the materials. Have the authors tried finetuning more layers of the pretrained network instead of only the final layer? There is usually a tradeoff in terms of the number of layers finetuned vs the resulting performance.  
+
+<|ref|>text<|/ref|><|det|>[[91, 795, 900, 825]]<|/det|>
+2. The authors must apply the focal loss class balancing to finetune the pretrained model. Keep the loss fixed and change the models alone to verify the difference.  
+
+<|ref|>text<|/ref|><|det|>[[90, 840, 901, 871]]<|/det|>
+3. The difference between LOM/SEM cases is puzzling given that the difference in # of data points isn't by a large factor.  
+
+<|ref|>text<|/ref|><|det|>[[92, 888, 220, 916]]<|/det|>
+Author response: Add 1:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[92, 65, 900, 186]]<|/det|>
+That the U- Net scoring is better than the pretrained U- Net VGG16 in the optical light microscopy case was in fact an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 201, 145, 214]]<|/det|>
+## Add 2:  
+
+<|ref|>text<|/ref|><|det|>[[92, 216, 870, 246]]<|/det|>
+Thank you for your comment on fine- tuning. We did finetune over the full network in this case since the initial submission. We did add a remark on how we finetuned in the methods section.  
+
+<|ref|>text<|/ref|><|det|>[[92, 261, 900, 323]]<|/det|>
+Despite identifying this as an artifact, we did implement the identical loss function for the VGG16 network and the fine tuning training according to your remark. However, assimilating the loss functions did not affect the results significantly. We added this as a statement in the results and added more notes on the loss functions in the methods section applied.  
+
+<|ref|>text<|/ref|><|det|>[[92, 337, 891, 428]]<|/det|>
+Regarding point 3: In fact, the data amount is substantially different (not only the image amount 51. vs 36) but especially the physical image size resolution and the characterized area in \(\mu \mathrm{m}^2\) differs significantly. As the features have the same size independent of the microscopy methodology, the LOM contains a much larger number of grains. Therefore, we argue that the tendencies are well explainable. In the light optical microscopy micrographs, the effect of pre- training is smaller since more data is available, while in the SEM models the pretraining has a major impact.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[93, 82, 377, 97]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[92, 113, 904, 249]]<|/det|>
+The paper addresses the problem of highlighting phases in (optical and SEM) micrographs of compound materials. The uneven distribution of grey- levels into each phase defies segmentation algorithms that rely solely on pixel intensity. The authors adopt two slightly different deep learning approaches and compare them against manually- annotated golden standard. The main objective is to show that U- Net architectures are a reliable tool to solve such complex problems with no need of fine tuning. The authors claim that this is demonstrated by the very fact that different flavors of these architectures yield similar results and that performance variations mostly depend on data manipulation (e.g., tiling and padding) rather than on architectural differences. Indeed, the results shown by the authors are quite good for the problem at hand, even more so due to the reduced size of the training set (quite common in material sciences).  
+
+<|ref|>text<|/ref|><|det|>[[92, 250, 902, 385]]<|/det|>
+This paper does not describe a novel approach to image segmentation. However, the multidisciplinary approach of this work to the segmentation problem, encompassing sample preparation, image acquisition and image segmentation, is certainly significant (and, I would say, uncommon). It points towards a direction that all future research in this field should follow. The paper thoroughly addresses several open questions about segmentation and deep- learning, including the interpretation of the inner activity of the network, that is, one of the most controversial aspects of deep learning approaches. It is also interesting in one more respect: two research groups work on the same problem to demonstrate the validity of a framework, rather than the quality of a single approach. In this sense, I think that this paper is a useful contribution to the scientific debate about the role of deep learning in industrial applications.  
+
+<|ref|>text<|/ref|><|det|>[[93, 385, 870, 429]]<|/det|>
+The paper clearly presents speculative thinking about each and every step of network training and operativity. The discussion is supported by experimental evidence and a sufficiently rigorous analysis of data. The description is reasonably detailed.  
+
+<|ref|>text<|/ref|><|det|>[[92, 429, 904, 475]]<|/det|>
+A few details could be added, e.g., about the operators used in the network to process images as well as about the size of padding and overlap between image tiles. It would be interesting to compare the size of padding to a representative measure of the size of phases.  
+
+<|ref|>text<|/ref|><|det|>[[92, 475, 900, 566]]<|/det|>
+The conclusions are consistent with experimental data. I would suggest to think a little bit more about a few statements, such as the claim that the small size of the training dataset used in experiments, compared to the quality of results, "invalidates the general claim of DL being only applicable for large- scale data sets". This is certainly true for the dataset used in this paper. However, this can well prove false for other data sets. The successive claim that the availability of reproducible, high- quality imaging would reduce the need for a large amount of training data appears a leap of logic.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 581, 220, 595]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 595, 898, 670]]<|/det|>
+Thank you for your detailed and positive review. The interdisciplinary collaboration is a fundamental necessity to advance the implementation of DL into the materials science area. Furthermore, the industrial usage requires convincing real world examples and usable frameworks. As you also pointed out, the real world of materials science and engineering works on a lot of small data cases and therefore we have targeted this as a priority.  
+
+<|ref|>text<|/ref|><|det|>[[92, 670, 890, 745]]<|/det|>
+Nevertheless, we would like to point out that fine tuning of pre- trained networks improves performance. In fact, we believe that in this low data regime, that we are in right now, pre- training and fine- tuning are of substantial relevance. In both LOM and especially SEM we show that an improvement through pretraining can be achieved. We enhanced our manuscript in the discussion to assure not to being suggestive of pretraining being unuseful.  
+
+<|ref|>text<|/ref|><|det|>[[92, 760, 904, 850]]<|/det|>
+"invalidates the general claim of DL being only applicable for large- scale data sets" \(\rightarrow\) We used "general" here to emphasize that it is a preconception that many (material) scientists have, who are not involved with data- driven techniques. We believe that in the setting of small variance data (reproducible image acquisition) fewer data is necessary to train a model for that domain. However, we observed that such a model won't be great at generalizing if trained with small processing- induced variance data. We understand this and hence the successive remark has been adjusted in the discussion.  
+
+<|ref|>text<|/ref|><|det|>[[92, 865, 900, 925]]<|/det|>
+At the end of page 9, the authors write that the models fail in the very same regions where human expert make mistakes during manual annotation. No further explanation is given in order to better understand the method used to locate these regions. Do the authors refer, e.g., to pixel regions for which manual annotations do not match?
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[93, 83, 221, 96]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[93, 97, 718, 111]]<|/det|>
+You are right, we added few sentences here to render in more clear in the discussion.  
+
+<|ref|>text<|/ref|><|det|>[[93, 126, 887, 187]]<|/det|>
+Thinking about the possibility to generalize the results shown in the paper, I would appreciate a deeper discussion about how much the performance of DL methods is affected by the scale of phases in the image. Similarly, in order to accept the claim that these methods are generalizable, I would like to see a few comparative experiments with radically different datasets.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 203, 221, 216]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 217, 891, 307]]<|/det|>
+We added a study on relative scale of phases and image size to the manuscript. Specifically we conducted another study where we altered the size of the tiles in LOM (rather than downscaling so we could exclude information loss and receptive field based effects). This study gives more insights on how to choose the tile sizes ideally, as it correlates tile sizes with lath width and lath- bainite region size distributions. Accordingly, we extended the results (Image context and network receptive field dependency section) and discussion (Image context and network receptive field dependency section).  
+
+<|ref|>text<|/ref|><|det|>[[92, 321, 897, 457]]<|/det|>
+On the other hand, we did not claim that the model trained with this low- variance dataset is generalizable across data sets. In fact, we did apply the trained models to an otherwise etched (overetched) image and observed that the model is not so good at generalizing across domains (see section ,,Variances and generalization"). Even when adjusting and optimizing the data augmentation to lower the domain gap towards a target domain, the model does not generalize particularly well. We stress that, our objective in this paper was not to train a perfect model for cross- domain generalization but rather to highlight image processing effects (scaling, tiling...) and pretraining on DL performance as well as explainability. If generalization would have been our primary objective, we would have acquired our dataset with a more natural variance and would have applied stronger data augmentation.  
+
+<|ref|>text<|/ref|><|det|>[[92, 473, 870, 504]]<|/det|>
+Finally, the authors do not address execution time. This is an important aspect for industrial applications, especially in manufacturing.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 520, 221, 533]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 534, 904, 579]]<|/det|>
+We added notes on execution time both for training and for deployment in the Supplemental. We added number of trained epochs and specified hardware as well in the methods section (Deep learning segmentation approach section).  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 611, 388, 625]]<|/det|>
+## Another important note to the reviewer:  
+
+<|ref|>text<|/ref|><|det|>[[92, 626, 904, 732]]<|/det|>
+The U- Net scoring better than the pretrained U- Net VGG16 in the light microscopy case was an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[93, 66, 377, 81]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[93, 98, 900, 144]]<|/det|>
+This paper purports to lay out a holistic approach to deploying deep learning for complex microstructure inference. This is a very topical and challenging problem, and I really wanted to like this paper. However the paper as it stands does not accomplish any of the key claims. I lay out my reasons below:  
+
+<|ref|>text<|/ref|><|det|>[[93, 159, 890, 235]]<|/det|>
+1) Claim of impact: Showing segmentation on one class of microstructures is not enough (according to this reviewer). This is insufficient to illustrate their central claim of "an intuition about the required data quality and quantity and an extensive methodological DL guideline for microstructure quantification and classification are still missing". No intuition is gained given the limited number of models used, and the limited data used (see point #3).  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 252, 220, 266]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[93, 281, 463, 295]]<|/det|>
+Thank you for your efforts in reviewing our work.  
+
+<|ref|>text<|/ref|><|det|>[[92, 312, 904, 464]]<|/det|>
+You are right, the claim is too strong. Nevertheless, this manuscript clearly indicates that data quantities of 30- 50 images suffice to train a segmentation network for this low material- extrinsic variance domain. Please note that both data sets have quite different data amounts, since, aside from the image number, individual LOM images capture a substantially larger physical area. In addition, when it comes to image rescaling and pretraining, we firmly believe that our experiments after the corrections provide a valid indication. Namely, that increasing the context and receptive field helps when dealing with such long- range features and by how much pretraining is helping in both distinct data quantity settings (low- quantitativ datasets are very typical for materials science data sets). It is true that we do not perform a study on data quality in this manuscript. We removed this claim and softened the claim concerning data quantity. We adjusted the abstract and introduction accordingly.  
+
+<|ref|>text<|/ref|><|det|>[[92, 479, 877, 540]]<|/det|>
+2) There are several other real-world images (for instance, MRI images) that are as (if not more) complex than microstructure images, and a lot more critical to segment correctly. This goes right to the heart of the statement "Furthermore, microstructure recognition tasks, compared to real-world images, can be very complex regarding ..."  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 557, 220, 570]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[93, 570, 866, 615]]<|/det|>
+By real- world images we are rather refering to street scenes (Cityscapes) or Image- Net, and similar. We slightly modified the wording from ,,real- world images" to ,,natural images" and added Image- Net as an example in the introduction to cause less confusion.  
+
+<|ref|>text<|/ref|><|det|>[[92, 631, 901, 677]]<|/det|>
+3) Data size: The datasize used (30-40) images makes the results of this study highly suspect. It was not clear to me how many images were finally used after augmentation and windowing, but training a U-Net from scratch is not a good idea with this few images. See for instance https://arxiv.org/pdf/2001.05566. pdf .  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 696, 220, 709]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 709, 904, 815]]<|/det|>
+We refer to the number of individual training tile images in the Supplemental. On page 4, we state ,,A summary of the data sets, including some characteristic metrics, can be found in the Supplemental." We adjusted the wording here to make it more clear. The amount of training tiles varies depending on the preprocessing and the exact data set. We kept it in the Supplemental rather than the manuscript since it would add large tables to an already somewhat long manuscript and since we consider the individual tile amount as not as important as the overarching raw microscope image number from which the tiles originate. For your convinience: In the non- resized LOM training, we worked with approximately 600 tiles.  
+
+<|ref|>text<|/ref|><|det|>[[92, 830, 886, 875]]<|/det|>
+In terms of augmentation we did apply online augmentation which means the number of epochs (which we now added consistently in the Methods Deep Learning section) and the dataset sizes (given in the Supplemental) directly provide the number of images which the network has seen.  
+
+<|ref|>text<|/ref|><|det|>[[92, 890, 901, 920]]<|/det|>
+We are aware that pretraining is a common and established practice in deep learning. However, we still wanted to make the comparison between ImageNet pretrained networks and random initialized ones. Finding
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[90, 66, 900, 97]]<|/det|>
+appropriate pretraining large-scale data sets with a small domain gap to the target task is difficult in materials science, which is why many practitioners do not apply pretraining currently.  
+
+<|ref|>text<|/ref|><|det|>[[92, 112, 880, 188]]<|/det|>
+4) Transfer learning: As the author's comment in outlook section, using pre-training with self- or semisupervised learning is the way to go with such small datasets. The marginal improvements of using a pretrained U-Net is indeed indicative of this. See for instance "Zhuang F, Qi Z, Duan K, Xi D, Zhu Y, Zhu H, Xiong H, He Q. A comprehensive survey on transfer learning. Proceedings of the IEEE. 2020 Jul 7;109(1):43-76."  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 206, 220, 219]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 220, 842, 250]]<|/det|>
+Thank you for sharing this information. In follow- up publications we aim to include semi- supervised learning and unsupervised domain adaptation more.  
+
+<|ref|>text<|/ref|><|det|>[[92, 265, 891, 312]]<|/det|>
+5) Transparent decision making: Grad-CAM has been shown to be a very poor explainability mechanism. See for instance https://arxiv.org/abs/1812.02843 where GradCAM giver poor explainability even when the model predictions are accurate.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 329, 220, 342]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 343, 899, 390]]<|/det|>
+It might be true that GradCAM, as other visualization techniques, is prone to such adversarial attacks. However, GradCAM was shown to provide comparatively sensible visualizations here (→) when exposed to sanity checks.  
+
+<|ref|>text<|/ref|><|det|>[[92, 389, 871, 420]]<|/det|>
+Adebayo, Julius, et al. "Sanity checks for saliency maps." arXiv preprint arXiv:1810.03292 (2018). Sanity Checks for Saliency Maps (nips.cc)  
+
+<|ref|>text<|/ref|><|det|>[[92, 420, 866, 450]]<|/det|>
+Similarly, for our results the visualizations agree with expert annotators expectations. From our point of view, adversarial attacks are not relevant in our setting.  
+
+<|ref|>text<|/ref|><|det|>[[92, 465, 902, 512]]<|/det|>
+6) Intra- and Inter-rater variability: This was an opportunity lost to discuss and provide intuition on inter- and intra- rater variability during the human annotation. That is, do changes in the fg and bg as marked by experts significantly impact results?  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 530, 220, 543]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 544, 877, 589]]<|/det|>
+We expect there be to substantially less rater-variability in annotation as annotation was assisted by correlative EBSD (orientation-sensitive data). We did an experiment in the beginning of our collaboration where we considered two data sets:  
+
+<|ref|>text<|/ref|><|det|>[[183, 590, 627, 620]]<|/det|>
+data set where we discarded tiles with uncertain regions data set with all tiles  
+
+<|ref|>text<|/ref|><|det|>[[92, 621, 891, 650]]<|/det|>
+We did not see a major difference in performance. This we added as a remark in the manuscript in the result section.  
+
+<|ref|>text<|/ref|><|det|>[[92, 683, 865, 729]]<|/det|>
+7) Metrics: Using accuracy is a bad metric. In some cases, the pixel count of fp to bg is 1:3 or 1:4 which always segmenting as bg will give 75-80% accuracy. A 10% improvement seems marginal. I encourage sticking to IoU as the key figure of merit.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 745, 220, 758]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 758, 900, 848]]<|/det|>
+Initially we wanted to provide accuracy as an additional metric because it is more intuitive than IoU. However, as you mention we have class imbalance in the light microscopy data set as it is representative for the microstructure. Therefore, omitting true negatives in IoU renders this metric more sensitive. Even though we explained these aspects in the paper for our materials science readers, we removed the accuracy values from the tables and just added a single statement in the result section that the IoUs correspond to approximately X accuracy.  
+
+<|ref|>text<|/ref|><|det|>[[92, 864, 899, 894]]<|/det|>
+Side note: We have in average 73% percent background in the light microscopy, so it is almost 20% increase over the naive baseline.  
+
+<|ref|>text<|/ref|><|det|>[[90, 910, 866, 926]]<|/det|>
+8) Lack of detailed analysis: While the authors do a good job of performing ablation studies, this was an
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[92, 65, 904, 142]]<|/det|>
+opportunity lost with showing extensive analysis that could back up (at least anecdotally) some of their claims. Specifically, what happens when depth of the U- net is changed, what happens when window size (and hence data size) changes, what happens when total data set is further reduced, how is pixel size, window size and resolution related to performance. All these would add value and generate more intuition for practitioners.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 157, 221, 171]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 171, 899, 247]]<|/det|>
+We added a study on relative scale of phases and image. To be specific, we conducted another study where we altered the size of the cropped LOM tiles (rather than downscaling, so we could exclude information loss and receptive field based effects). This addresses your window size remark. This study gives more insights on how to choose the tile sizes ideally as we compare the tile sizes with characteristic length scales (lath width or region size). Accordingly, we extended the results and discussion.  
+
+<|ref|>text<|/ref|><|det|>[[92, 262, 857, 293]]<|/det|>
+9) Data sharing: Without the availability of annotated data for the broader audience to try and evaluate, replicate and improve on these (standard) methods, I see the impact of this paper as minimal.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 310, 221, 323]]<|/det|>
+## Author response:  
+
+<|ref|>text<|/ref|><|det|>[[92, 323, 900, 355]]<|/det|>
+Although we would like to share the data to fuel a similar multiplication effect as observed in other domains, in this case we are limited due to industry involvement. The data is part of an ongoing study.  
+
+<|ref|>sub_title<|/ref|><|det|>[[93, 385, 388, 399]]<|/det|>
+## Another important note to the reviewer:  
+
+<|ref|>text<|/ref|><|det|>[[92, 399, 904, 504]]<|/det|>
+The U- Net scoring better than the pretrained U- Net VGG16 in the light microscopy case was an artifact. Since we used unpadded convolutions in the vanilla U- Net and the tiles were extracted with an overlap but train and test tiles were sampled from the same raw images we had a mixing of training and testing data. This error did not apply to the U- Net VGG16 as it used padded convolutions and center- cropping (removing the overlap region) before passing the data to the network. We resolved this issue by applying the same procedure (center- cropping and padded convolutions) in the vanilla U- Net and conducting the experiments again. As a positive side effect, this increased the comparability between the two approaches.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 84, 293, 98]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 113, 223, 140]]<|/det|>
+Reviewer #1: None  
+
+<|ref|>text<|/ref|><|det|>[[119, 154, 223, 167]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[119, 169, 300, 182]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[119, 183, 864, 225]]<|/det|>
+The authors addressed all the issues and concerns in my review. For that part (i.e., the point of view of a user), I think that their article has been improved with respect to the initial manuscript. If other reviewers share this thought, I think that the manuscript is ready for publication.  
+
+<|ref|>text<|/ref|><|det|>[[118, 238, 830, 260]]<|/det|>
+Just one more comment on fine- tuning that is not intended to influence the discourse in final paper. Perhaps, the authors will find it a useful thought for further investigation.  
+
+<|ref|>text<|/ref|><|det|>[[119, 260, 740, 274]]<|/det|>
+I believe that one should be very careful with tuning, at least for two reasons:  
+
+<|ref|>text<|/ref|><|det|>[[119, 275, 795, 289]]<|/det|>
+1. it often requires care, time, and competence that industrial applications cannot afford  
+
+<|ref|>text<|/ref|><|det|>[[119, 290, 657, 304]]<|/det|>
+2. there is always a risk of overtraining, especially with small datasets  
+
+<|ref|>text<|/ref|><|det|>[[119, 350, 223, 363]]<|/det|>
+Reviewer #3:  
+
+<|ref|>text<|/ref|><|det|>[[119, 365, 300, 377]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[119, 378, 841, 420]]<|/det|>
+I thank the authors for the additional effort in responding to (some of) my questions. The manuscript has improved with the additional work, as well as the thoughtful watering down of some of the claims.  
+
+<|ref|>text<|/ref|><|det|>[[118, 433, 876, 462]]<|/det|>
+However, I do note that the watering down of several claims does bring down the overall impact of the paper. Several of my concerns (#2,#3,#7,#9) remain unanswered.  
+
+<|ref|>text<|/ref|><|det|>[[119, 475, 878, 532]]<|/det|>
+I remain unsatisfied with the key claims, especially when there are similar reports from other fields (for instance, the medical field) where complex images are successfully segmented. I also remain concerned with the small dataset used here. Finally, while understandable, the unavailability of the data makes comparative assessment impossible, further diminishing the impact.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 84, 402, 99]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 118, 868, 168]]<|/det|>
+The authors addressed all the issues and concerns in my review. For that part (i.e., the point of view of a user), I think that their article has been improved with respect to the initial manuscript. If other reviewers share this thought, I think that the manuscript is ready for publication.  
+
+<|ref|>text<|/ref|><|det|>[[115, 187, 870, 272]]<|/det|>
+Just one more comment on fine- tuning that is not intended to influence the discourse in final paper. Perhaps, the authors will find it a useful thought for further investigation. I believe that one should be very careful with tuning, at least for two reasons: 1. it often requires care, time, and competence that industrial applications cannot afford 2. there is always a risk of overtraining, especially with small datasets  
+
+<|ref|>text<|/ref|><|det|>[[116, 299, 871, 350]]<|/det|>
+Thank you for your review and remarks. We agree with your comments. In our case, fine- tuning was applied in a careful way (with optimized learning rates) and overfitting was not observed in the learning curves.  
+
+<|ref|>text<|/ref|><|det|>[[117, 385, 402, 400]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 419, 876, 470]]<|/det|>
+I thank the authors for the additional effort in responding to (some of) my questions. The manuscript has improved with the additional work, as well as the thoughtful watering down of some of the claims.  
+
+<|ref|>text<|/ref|><|det|>[[116, 488, 866, 522]]<|/det|>
+However, I do note that the watering down of several claims does bring down the overall impact of the paper. Several of my concerns (#2,#3,#7,#9) remain unanswered.  
+
+<|ref|>text<|/ref|><|det|>[[116, 540, 865, 608]]<|/det|>
+I remain unsatisfied with the key claims, especially when there are similar reports from other fields (for instance, the medical field) where complex images are successfully segmented. I also remain concerned with the small dataset used here. Finally, while understandable, the unavailability of the data makes comparative assessment impossible, further diminishing the impact.  
+
+<|ref|>text<|/ref|><|det|>[[116, 645, 864, 679]]<|/det|>
+Thank you for your review and remarks. We agree that some works in other domains (e.g. medical) share similar observations with our work. However, we cited some of those in our manuscript, e.g.  
+
+<|ref|>text<|/ref|><|det|>[[116, 689, 736, 722]]<|/det|>
+Sabottke, C. F. & Spieler, B. M. The Effect of Image Resolution on Deep Learning in Radiography. Radiol. Artif. Intell.2, e190015, DOI: 10.1148/ryai.2019190015 (2020).  
+
+<|ref|>text<|/ref|><|det|>[[116, 733, 875, 818]]<|/det|>
+We believe that the observed scatter in the data is predominantly materials' microstructure- based, since image acquisition was performed in a very reproducible fashion. Under this assumption and considering the characteristic sizes of microstructural features (such as lath width and grain sizes), it is very likely that the imaged area is representative of the microstructural scatter. We added this in a few sentences to make clear, why we achieve good results despite the low data quantity.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d636d1552621791c70664dbc9983bc6c062889f128817d323f7d856b66a1ed53/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d636d1552621791c70664dbc9983bc6c062889f128817d323f7d856b66a1ed53/supplementary_0_Peer Review File__d636d1552621791c70664dbc9983bc6c062889f128817d323f7d856b66a1ed53.mmd

@@ -0,0 +1,324 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Transmission and Dynamics of Mother- Infant Gut Viruses during Pregnancy and Early Life  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operatin g a transparent peer review scheme. This document only contains reviewer comments and rebuttal l etters for versions considered at Nature Communications.  
+
+## REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+Garmeva et al present an analysis of a metagenomic analysis of bacteria and dsDNA virus in the stools of \(\sim 30\) mother and infant pairs followed in a longitudinal fashion. They examine the viral portion in the VLPs and via metagenomic sequencing without viral prep.  
+
+They reproduce previous observations on differences in bacterial and viral community composition between mothers and infants, on shared (maternal/infant) bacteria and bacteriophages which is important, but not novel. Relevant papers (not all): Lim et al. 2015, Zhao et al. 2017, McCann et al. 2018, Magsood et al. 2019, Liang et al. 2020, Walters et al 2023.  
+
+## Major comments:  
+
+1. Terms should be defined: metaviromes, whole metavirome, vOTU, metagenome in the beginning-it is confusing If they are referring to viruses detected in VLP prepped sample or virus in the portion of sample that did not get viral prep. This is key throughout and as written is not always clear.  
+
+2. Page 11 Line 211. This methodology assumes that any phage detected in the whole metagenome data set are prophage. This does not seem like a valid assumption, as active phage could be captured here as well.  
+
+3. Page 16 line 332. What is the rationale behind the conclusion these were through direct seeding.  
+
+4. Page 18 line 371. This is quite speculative.  
+
+5. Page 19 line 376. This is incorrect. Please see Walters et al Cell Host and Microbe, 2023 (Ref 22).  
+
+6. Page 21. Line 421-424. Three papers, Maqsood, Walters, and this one, find only a relatively small percent of shared viral component between mother and infant gut. Why invoke that small difference in percent shared is due to longitudinal sampling when the maternal virome is stable over time. Seems that the point is being missed here.  
+
+7. Page 21-22. Line 442-444 these conclusions are overstated.  
+
+8. Page 27. Line 598. Criteria 6 is concerning. Why are these necessarily thought to be viral? Were possible archaeal and bacterial sequences removed? Did the authors try blastx of these scaffolds? What did they hit by blastx? A table which outlines by which method called a virus would be helpful.  
+
+9. Page 28 Line 633. Why did authors choose this method to call temperate phage? Did this agree with other more commonly used tools for this?
+
+<--- Page Split --->
+
+10. Page 29. Line 649. It is quite surprising this high level of predicted microbial host. This is not in keeping with other studies using iPhOp.  
+
+Minor comments:  
+
+1. Page 14 line 269. Please explain what is meant by accounting for prophages detected in whole metaviromes  
+
+2. Page 26. Line 556. Seems unusual that 3 out of 4 negative controls failed. Could the authors comment more on this.  
+
+3. Page 30 line 686. Is wilxocon the proper test here?  
+
+Reviewer #2 (Remarks to the Author):  
+
+The Authors have substantially improved the manuscript by increasing the sampling size and frequency, and by improving the data analysis pipeline. While most of the concerns raised in the first round of Review have been addressed, two relevant issues still persist.  
+
+- The sequences and metadata are, differently from what stated by the authors in lines 844-845, not publicly available. The access policy associated to the provided EGA accession says "The Consortium is willing to consider applications from third-party researchers for access to the anonymized sequence data generated by the Project. [...] Access is conditional upon the availability of data and on signed agreement by the researcher to abide by the policies and conditions related to data ownership, disposal, ethical approval, confidentiality and commercialization referred herein". In fact, nor the fastq files nor the metadata can be downloaded. This is unacceptable. This kind of access restriction, entirely omitted in the manuscript, goes against the Journal Guidelines on data availability and against the FAIR data sharing principles.  
+
+- The use of "total metagenomes" vs "bacteriome" and "whole metavirome", vs "metavirome" vs "virome" is inconsistent throughput the manuscript and the meaning of these terms is not clearly described, nor in the text nor in the figures.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+Garmeva et al present an analysis of a metagenomic analysis of bacteria and dsDNA virus in the stools of \(\sim 30\) mother and infant pairs followed in a longitudinal fashion. They examine the viral portion in the VLPs and via metagenomic sequencing without viral prep.  
+
+They reproduce previous observations on differences in bacterial and viral community composition between mothers and infants, on shared (maternal/infant) bacteria and bacteriophages which is important, but not novel. Relevant papers (not all): Lim et al. 2015, Zhao et al. 2017, McCann et al. 2018, Magsood et al. 2019, Liang et al. 2020, Walters et al 2023.  
+
+Major comments:  
+
+1. Terms should be defined: metaviromes, whole metavirome, vOTU, metagenome in the beginning- it is confusing If they are referring to viruses detected in VLP prepped sample or virus in the portion of sample that did not get viral prep. This is key throughout and as written is not always clear.  
+
+We agree that the terms, as they were in the manuscript, led to confusion. For the sake of clarity we have now termed metaviromes as either MGS metaviromes or VLP metavirome. Additionally, we have created a box (Box 1) in the introduction as a glossary of the terms used in this paper. Lastly, we have changed the term microbiome to bacteriome in the text wherever applicable. We hope that our modifications increase clarity.  
+
+Hereby the text from Box 1: Bacteriome: community of all bacteria inhabiting an ecosystem Virome: community of all viruses inhabiting an ecosystem Total metagenome: the total sum of all genomes of microbes in a sample VLP metaviromes: the total sum of all genomes of viruses in a VLP-enriched sample MGS metaviromes: the total sum of all genomes of viruses in a total metagenome vOTU (virus operational taxonomic unit): species-rank (95% Average Nucleotide Identity, at 85% alignment fraction (relative to the shorter sequence)) virus groups  
+
+2. Page 11 Line 211. This methodology assumes that any phage detected in the whole metagenome data set are prophage. This does not seem like a valid assumption, as active phage could be captured here as well.
+
+<--- Page Split --->
+
+Original text: Given that the majority of temperate bacteriophages in the adult gut are in the form of prophages<sup>27</sup>, we compared the percentage of temperate bacteriophages in mothers and infants using the vOTUs detected in whole metaviromes, created by aligning the reads from whole metagenomes aligned to vOTU database reconstructed from metaviromes, thus representing the prophage content of the virome.  
+
+We agree with the reviewer that actively replicating temperate phages could be captured by this method. We have therefore rephrased our text. It now reads (lines 207- 212):  
+
+"Given that the majority of temperate bacteriophages in the adult gut are in the form of prophages<sup>27</sup>, we compared the percentage of temperate bacteriophages in mothers and infants using the vOTUs detected in MGS metaviromes, created by aligning the reads from total metagenomes aligned to the vOTU database reconstructed from VLP metaviromes, thus representing the prophage- inclusive temperate phage content of the virome."  
+
+3. Page 16 line 332. What is the rationale behind the conclusion these were through direct seeding.  
+
+Original text: "Among the 26 transmitted viruses, 21 were identified as virulent bacteriophages. Based on their predicted lifecycle, it is likely that they were transmitted through direct seeding of VLPs from the maternal to infant gut."  
+
+Since one of the criteria for inclusion of vOTUs in the strain- sharing analysis was the genome completeness (high- quality or complete genome predicted by CheckV or circularised genomes), and these 21 mentioned transmitted viruses did not have neither integrase nor site- specific recombinase genes, we concluded that they very likely exclusively undergo lytic infection cycle. For such phages, the only transmission mechanism available is the direct seeding of the viral particles. We agree with the reviewer that this might not have been clear in the original text. Hence, we have rephrased the text (lines 330- 333):  
+
+"Among the 26 transmitted viruses, 21 were identified as virulent bacteriophages. Based on their predicted lytic lifecycle, it is likely that they were directly seeded in the form of virus particles from the maternal to infant gut instead of being co- transmitted as prophages within the transmitted bacterial strains."  
+
+4. Page 18 line 371. This is quite speculative.
+
+<--- Page Split --->
+
+Original text: Gene annotation showed that, in addition to carrying an integrase gene, L37775_LS1 also carries a CAZyme (Glycosyl hydrolases family 25, Supplementary Fig. 6d), indicating that this phage might be associated with infant feeding.  
+
+We agree that this is speculative and have adapted the sentence as follows: (lines 368- 370): "Gene annotation showed that, in addition to carrying an integrase gene, L37775_LS1 also carries a CAZyme (Glycosyl hydrolases family 25, Supplementary Fig. 6d)."  
+
+5. Page 19 line 376. This is incorrect. Please see Walters et al Cell Host and Microbe, 2023 (Ref 22).  
+
+Original text: To our knowledge, this is the only study to look at the maternal virome longitudinally during pregnancy, birth and after birth. In the maternal total microbiome, we observed a notable shift in composition between the first and second trimesters of pregnancy.  
+
+We have now changed this to read (lines 373- 375): "To our knowledge, this is one of the few studies that investigates the maternal virome longitudinally during pregnancy, birth and after birth."  
+
+6. Page 21. Line 421-424. Three papers, Maqsood, Walters, and this one, find only a relatively small percent of shared viral component between mother and infant gut. Why invoke that small difference in percent shared is due to longitudinal sampling when the maternal virome is stable over time. Seems that the point is being missed here.  
+
+Original text: Our study revealed that, despite significant distinctions between the infant and maternal gut viromes, infants shared on average \(32.7\%\) of vOTUs with their mothers. This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer timeframe, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth.  
+
+We thank the reviewer for this comment. We would like to clarify why we believe that our use of longitudinal sampling is leading to the increased sharedness between maternal and infant microbiome:  
+
+1) Firstly, and most importantly, the study of Maqsood et al. used only one time point of mother and infant. This time was very close to birth. Their study of transmission from mother to infant will thus encompass mostly bacteria and viruses that are vertically transmitted during birth and not arising from the
+
+<--- Page Split --->
+
+common shared environment. Hence we expect this percentage to be much smaller than the overall shared microbiome which our study takes into account.  
+
+2) In pregnancy we indeed show that overall the maternal gut virome remains stable. However, as seen in Figure 2e, there is turnover in viral species which also contributes to the sharedness.  
+
+3) Finally, having longitudinal sampling gives us a higher chance to detect shared contigs that would be missed when looking simply at one cross-sectional sample.  
+
+We have changed the text to clarify this (lines 420- 425): "Our study revealed that, despite significant distinctions between the infant and maternal gut viromes, infants shared on average 32.7% of vOTUs with their mothers. This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+7. Page 21-22. Line 442-444 these conclusions are overstated.  
+
+Original text: We then show examples of how this occurs via direct transmission of bacteriophages and prophage induction following transmission of their bacterial hosts. We also show how infants obtain some viral strains from induction from bacteria that did not come from their mothers.  
+
+We disagree with the reviewer here. We specifically mention that we found examples of how this occurs. We have, however, reformulated the text to indicate yet again that this is our hypothesis (lines 442- 443): We then show examples of how this might occur via direct transmission of bacteriophages and prophage induction following transmission of their bacterial hosts. We also show how infants obtain some viral strains from induction from bacteria that did not come from their mothers.  
+
+8. Page 27. Line 598. Criteria 6 is concerning. Why are these necessarily thought to be viral? Were possible archaeal and bacterial sequences removed? Did the authors try blastx of these scaffolds? What did they hit by blastx? A table which outlines by which method called a virus would be helpful.  
+
+Original text: (6) being longer than 3 kbp with no hits (alignments \(>100\) nucleotides, \(90\%\) ANI, e- value of 10-10) to the nt database (release 249). 281,789 scaffolds fulfilled at least one of these six criteria.
+
+<--- Page Split --->
+
+The usage of criteria #6 is in accordance with previously published studies \(^{1,2}\) . We agree with the reviewer that criteria #6 alone is not sufficient to call a sequence putatively viral. However, we would like to stress that in our pipeline, we use a robust post- virome discovery filtration process designed using the recent studies in the metaviromics field \(^{3 - }\) \(^{6}\) . Specifically, we employ screening of putative virus scaffolds for the presence of ribosomal RNA (rRNA) genes using a BLASTn search in the SILVA 138.1 NR99 rRNA genes database along with screening for ribosomal protein genes identified using a BLASTp search (e- value threshold of \(10^{- 10}\) ) against a subset of ribosomal protein sequences from the COG database (release 2020), as mentioned in the Methods section (lines 609- 611 and lines 585- 586) and exclude these (lines 619- 621). Through this comprehensive approach, we effectively eliminate potential archaeal and bacterial sequences, ensuring that only genuine viral scaffolds are considered in the downstream analysis.  
+
+Furthermore, there is an increasing body of evidence \(^{7,8}\) that suggests that the NR might contain many prophage proteins that are currently attributed to bacteria as prophages are highly prevalent in gut- associated bacteria. We thus did not blastx these scaffolds to the NR database.  
+
+We agree with the reviewer that a table encompassing the indications based on inclusion and exclusion criteria for the virus scaffolds would be helpful. In fact, this table can be found in Figshare along with the fasta file containing all virus scaffolds used in this study:  
+
+https://figshare.com/articles/dataset/Virus scaffolds reconstructed in the study b Mot her- Infant Gut Viruses and their Bacterial Hosts Transmission Patterns and Dynamics during Pregnancy and Early Life b NEXT Pilot and their metadata /23926593 We have now added this table to the Supplementary tables (S42: Metadata for virus genomes and genome fragments reconstructed in the study).  
+
+9. Page 28 Line 633. Why did authors choose this method to call temperate phage? Did this agree with other more commonly used tools for this?  
+
+Original text: Temperate bacteriophages were identified using either the presence of integrase genes or the co- presence of recombinase genes with Cl- repressor- like protein genes from the pVOGs annotation within a vOTU representative.  
+
+The choice for the method to predict the phage lifestyle was based on well- established research in the field (Shkoporov et al. \(^{1}\) Cell Host & Microbe.2019, Howard- Varona et al. \(^{9}\) ISME Journal. 2017. Roux et al. \(^{10}\) PeerJ. 2015). These studies pinpointed integrase and site- specific recombinases as the hallmark genes for the temperate phages. Since site- specific recombinases also regulate viral gene expression during the bacterial cell
+
+<--- Page Split --->
+
+infection, our phage lifestyle prediction method included an additional condition of the presence of Cl- repressor- like protein genes (Paul.11 ISME Journal. 2008). Furthermore, the tools used in the viromics field for the lifestyle prediction like Prophage Hunter, PhiSpy, PHASTEST, BACPHLIP, and PhaTYP also rely on the presence of integrase gene within the virus sequence of interest in the marker- based parts of their pipelines (Song et al.12 Nucleic Acids Research. 2019, Akhter et al.13 Nucleic Acids Research. 2016, Arndt et al.14 Briefings in Bioinformatics,2019. Hockenberry et al.15 PeerJ. 2019. Shang et al.16 Briefings in Bioinformatics. 2022).  
+
+10. Page 29. Line 649. It is quite surprising this high level of predicted microbial host. This is not in keeping with other studies using iPhOp.  
+
+Original text: In total, the microbial host was predicted for 68,299 of 102,210 viral scaffolds (67.3%) at the genus level and for 85,135 (82.9%) of all vOTUs at the species level.  
+
+Benchmarking of the iPhOp framework in the original article (Roux et al.17, 2023) showed that the amount of successful host predictions depends on the ecosystem studied. According to the paper, more hosts could be found for predicted viruses in human- associated samples. In our study, the percentage of the viruses with predicted hosts is concordant with the benchmark: "For human- associated microbiomes, about 89% of the nonredundant high- quality genomes had a host predicted using iPhOp, including 57% with very high confidence predictions (iPhOp score \(\geq 95\) ). (Roux et al. 2023)  
+
+To the best of our knowledge, three other studies have employed the iPhOp framework. Specifically, Ter Horst et al.18 (2023) focused on the virome of wetlands, Peng et al.19 (2023) on the virome of deep- sea cold seep sediments, and Saenz et al.20 (2023) on the virome of grass silage. Given that these studies explored non- human associated ecosystems, the proportion of viruses with hosts predicted in their research is anticipated to vary from our findings.  
+
+Based on the details previously mentioned, we contend that while the percentage of the predicted host is high, it aligns with the benchmark set by the framework's authors (Roux et al., 2023).  
+
+## Minor comments:  
+
+1. Page 14 line 269. Please explain what is meant by accounting for prophages detected in whole metaviromes
+
+<--- Page Split --->
+
+Original text: As pioneer viruses in the infant gut are thought to be primarily temperate phages induced from the first gut bacterial colonisers<sup>20</sup>, we next assessed the sharedness of maternal to infant vOTUs while accounting for prophages detected in whole metaviromes.  
+
+We thank the reviewer for the comment. This means we expanded our VLP metaviromes with the temperate bacteriophages detected in MGS metaviromes and not concurrent VLP metaviromes (prophages). We have edited the sentence to make it clearer (lines 265- 268): "As pioneer viruses in the infant gut are thought to be primarily temperate phages induced from the first gut bacterial colonisers<sup>20</sup>, we next assessed the sharedness of maternal to infant vOTUs while accounting for prophages detected only in MGS metaviromes."  
+
+2. Page 26. Line 556. Seems unusual that 3 out of 4 negative controls failed. Could the authors comment more on this.  
+
+It's not uncommon for negative controls to display such behavior, in fact this is totally in line with expectations, especially, when no whole genome amplification was applied to the samples. Negative controls, by design, have minimal or no target microbial reads. As a result, during the PCR amplification within the library preparation process, there might be very little template available for amplification. Hence genomic library preparation was unsuccessful and these samples did not get sequenced. It now reads as: "Three out of four negative VLP metavirome controls failed library preparation and therefore could not be sequenced <...>".  
+
+3. Page 30 line 686. Is wilcoxon the proper test here?  
+
+As the Shannon Diversity Index is not normally distributed in our dataset, and we took only one time point per individual for this comparison into account (thus accounting for repeated measures is not necessary), the use of Wilcoxon tests is appropriate.  
+
+## Reviewer #2 (Remarks to the Author):  
+
+The Authors have substantially improved the manuscript by increasing the sampling size and frequency, and by improving the data analysis pipeline. While most of the concerns raised in the first round of Review have been addressed, two relevant issues still persist.  
+
+- The sequences and metadata are, differently from what stated by the authors in lines 844-845, not publicly available. The access policy associated to the provided EGA
+
+<--- Page Split --->
+
+accession says "The Consortium is willing to consider applications from third- party researchers for access to the anonymized sequence data generated by the Project. [...] Access is conditional upon the availability of data and on signed agreement by the researcher to abide by the policies and conditions related to data ownership, disposal, ethical approval, confidentiality and commercialization referred herein". In fact, nor the fastq files nor the metadata can be downloaded. This is unacceptable. This kind of access restriction, entirely omitted in the manuscript, goes against the Journal Guidelines on data availability and against the FAIR data sharing principles.  
+
+Thank you for this very important point. We would like to provide clarity regarding the data sharing procedure. In the previous metagenomics studies of our group, such as Chen et al. (Cell, 2021; doi: 10.1016/j.cell.2021.03.024) and Gacesa et al. (Nature, 2022; doi: 10.1038/s41586- 022- 04567- 7), we always uploaded our data to the European Genome- phenome Archive (EGA). Similarly, for the current study, our dataset — which includes sample information, all phenotypes, family structures, and quality- trimmed sequencing reads — has been securely archived in the EGA under study ID: EGAS00001005969. Our group has a history of fulfilling all data requests. We specifically use EGA for data sharing, as it is an approved repository for functional genomics data according to the Nature Data Repository Guidance. All our prior publications utilized this platform for multi- omic data sharing.  
+
+To access our dataset, interested parties must first register with the EGA. After registration, they must read the Data Access form: https://groningenmicrobiome.org/?page_id=2598. This contains a form that requires certain essential details and an acknowledgment from the requester to not use the data for commercial purposes. This form can be found here: https://forms.gle/zbJLMnojys3VVKPY8. Requests will be granted for researchers affiliated with an academic, non- profit, or government institution.  
+
+In addition to this we have specified this in the text:  
+
+Original text: Sample information, basic phenotypes, family structure and quality- trimmed sequencing reads can be found in the EGA repository (study ID: EGAS00001005969).  
+
+This has been changed to (lines 848- 851): "EGA repository (study ID: EGAS00001005969). Access to the LLNEXT Project data will be granted to qualified researchers will be governed by the provisions laid out in the LLNEXT Data Access Agreement: https://groningenmicrobiome.org/?page_id=2598"
+
+<--- Page Split --->
+
+Furthermore, all our contigs have been deposited on figshare and are freely accessible here: https://figshare.com/articles/dataset/Virus_scaffolds_reconstructed_in_the_study_b_Mother-Infant_Gut_Viruses_and_their_Bacterial_Hosts_Transmission_Patterns_and_Dynamics_during_Pregnancy_and_Early_Life_b_NEXT_Pilot_and_their_metadata_/23926593  
+
+- The use of "total metagenomes" vs "bacteriome" and "whole metavirome", vs "metavirome" vs "virome" is inconsistent throughput the manuscript and the meaning of these terms is not clearly described, nor in the text nor in the figures.  
+
+We agree with the reviewer that the use of introduced terms was inconsistent in the original text. For the sake of clarity we have now termed metaviromes as either MGS metaviromes or VLP metavirome. Additionally, we have created a box (Box 1) in the introduction as a glossary of the terms used in this paper. Lastly, we have changed the term microbiome to bacteriome in the text wherever applicable to avoid confusion.  
+
+Hereby the text from Box 1:  
+
+Bacteriome: community of all bacteria inhabiting an ecosystem  Virome: community of all viruses inhabiting an ecosystem  Total metagenome: the total sum of all genomes of microbes in a sample  VLP metaviromes: the total sum of all genomes of viruses in a VLP- enriched sample  MGS metaviromes: the total sum of all genomes of viruses in a total metagenome  vOTU (virus operational taxonomic unit): species- rank (95% Average Nucleotide Identity, at 85% alignment fraction (relative to the shorter sequence)) virus groups  
+
+## References  
+
+1. Shkoporov, A. N. et al. The Human Gut Virome Is Highly Diverse, Stable, and Individual Specific. Cell Host Microbe 26, 527-541.e5 (2019).  
+2. Shkoporov, A. N. et al. Reproducible protocols for metagenomic analysis of human faecal phageomes. Microbiome 6, 68 (2018).  
+3. Clooney, A. G. et al. Whole-Virome Analysis Sheds Light on Viral Dark Matter in Inflammatory Bowel Disease. Cell Host Microbe 26, 764-778.e5 (2019).  
+4. Garmaeva, S. et al. Stability of the human gut virome and effect of gluten-free diet. Cell Rep.
+
+<--- Page Split --->
+
+35, 109132 (2021).  
+
+5. Shkoporov, A. N. et al. The Human Gut Virome Is Highly Diverse, Stable, and Individual Specific. Cell Host Microbe 26, 527-541. e5 (2019).  
+
+6. Gulyaeva, A. et al. Diversity and Ecology of Caudoviricetes Phages with Genome Terminal Repeats in Fecal Metagenomes from Four Dutch Cohorts. Viruses 14, (2022).  
+
+7. Dahlman, S. et al. Temperate gut phages are prevalent, diverse, and predominantly inactive. bioRxiv (2023) doi:10.1101/2023.08.17.553642.  
+
+8. Maxwell Anthenelli et al. Phage and bacteria diversification through a prophage acquisition ratchet. bioRxiv 2020.04.08.028340 (2020) doi:10.1101/2020.04.08.028340.  
+
+9. Howard-Varona, C., Hargreaves, K. R., Abedon, S. T. & Sullivan, M. B. Lysogeny in nature: mechanisms, impact and ecology of temperate phages. ISME J. 11, 1511-1520 (2017).  
+
+10. Roux, S., Enault, F., Hurwitz, B. L. & Sullivan, M. B. VirSorter: mining viral signal from microbial genomic data. PeerJ 3, e985 (2015).  
+
+11. Paul, J. H. Prophages in marine bacteria: dangerous molecular time bombs or the key to survival in the seas? ISME J. 2, 579-589 (2008).  
+
+12. Song, W. et al. Prophage Hunter: an integrative hunting tool for active prophages. Nucleic Acids Res. 47, W74-W80 (2019).  
+
+13. Akhtar, M. M., Micolucci, L., Islam, M. S., Olivieri, F. & Procopio, A. D. Bioinformatics tools for microRNA dissection. Nucleic Acids Res. 44, 24-44 (2016).  
+
+14. Arndt, D., Marcu, A., Liang, Y. & Wishart, D. S. PHAST, PHASTER and PHASTEST: Tools for finding prophage in bacterial genomes. Brief. Bioinform. 20, 1560-1567 (2019).  
+
+15. Hockenberry, A. J. & Wilke, C. O. BACPHLIP: predicting bacteriophage lifestyle from conserved protein domains. PeerJ 9, e11396 (2021).  
+
+16. Shang, J., Tang, X. & Sun, Y. PhaTYP: predicting the lifestyle for bacteriophages using BERT. Brief. Bioinform. 24, bbac487 (2023).  
+
+17. Roux, S. et al. iPhOP: An integrated machine learning framework to maximize host
+
+<--- Page Split --->
+
+prediction for metagenome- derived viruses of archaea and bacteria. PLOS Biol. 21, e3002083 (2023).  
+
+18. Anneliek M. ter Horst, Jane D. Fudyma, Jacqueline L. Sones, & Joanne B. Emerson. Dispersal, habitat filtering, and eco-evolutionary dynamics as drivers of local and global wetland viral biogeography. bioRxiv 2023.04.28.538735 (2023) doi:10.1101/2023.04.28.538735.  
+
+19. Peng, Y. et al. Viruses in deep-sea cold seep sediments harbor diverse survival mechanisms and remain genetically conserved within species. ISME J. 17, 1774–1784 (2023).  
+
+20. Sáenz, J. S., Rios-Galicia, B., Rehkugler, B. & Seifert, J. Dynamic Development of Viral and Bacterial Diversity during Grass Silage Preservation. Viruses 15, (2023).
+
+<--- Page Split --->
+
+REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+Major Comments:  
+
+1. The fact that this is an analysis of only dsDNA virome should be more prominently noted such as in the abstract etc.  
+
+2. "This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+One of the other major differences between this paper and others, is the use of VLP plus metagenomic sequencing without viral prep. What is the overlap between the VLP virome and the MGS virome? This is important to share as many groups are mining metagenomic data for virome analysis.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The Authors have addressed all concerns raised in the previous round of Review.  
+
+My only recommendation is to fix the "run_sample.csv" metadata file, as part of the file is indeed comma separated, while part is separated by semi- columns, and add the corresponding column headers rather than having all the relevant metadata collapsed in the "sample_attributes" field.
+
+<--- Page Split --->
+
+## REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+Major Comments:  
+
+1. The fact that this is an analysis of only dsDNA virome should be more prominently noted such as in the abstract etc.  
+
+Original text (lines 20- 22): "We longitudinally assessed the composition of gut viruses and their bacterial hosts in 322 total metagenomes and 205 Virus Like Particle (VLP) metaviromes from 30 mothers during and after pregnancy and from their 32 infants during their first year of life."  
+
+We agree with the Reviewer and have edited the abstract accordingly (lines 4- 8): "To study the development of the infant gut virome over time and the factors that shape it, we longitudinally assess the composition of gut viruses and their bacterial hosts in 30 women during and after pregnancy and in their 32 infants during their first year of life. Using shotgun metagenomic sequencing applied to dsDNA extracted from Virus- Like Particles (VLPs) and bacteria, we generate 205 VLP metaviromes and 322 total metagenomes."  
+
+We then mention dsDNA gut virome in 205 VLP metaviromes as early as in the first sentence of Results (line 63- 66): "We profiled the gut microbiome (primarily referred to as the bacteriome) in 322 total metagenome samples and the double- strand DNA (dsDNA) gut virome in 205 VLP metavirome samples from 30 mothers and their 32 term- born infants (including 2 twin pairs) collected longitudinally from pregnancy to 12 months after birth (Fig. 1a; Supplementary Fig. 1a, b)."  
+
+2. "This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of VOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+One of the other major differences between this paper and others, is the use of VLP plus metagenomic sequencing without viral prep. What is the overlap between the VLP virome and the MGS virome? This is important to share as many groups are mining metagenomic data for virome analysis.  
+
+We agree with the Reviewer completely and now report the overlap between the VLP metaviromes and MGS metaviromes in Methods section "Generation of MGS metaviromes", lines 657- 664:  
+
+"The average alignment rate from total metagenomes to the curated virus database was \(50.6 \pm 18.2\%\) . As for VLP metaviromes, a count table was generated and transformed for MGS metaviromes. For samples where both VLP and MGS metaviromes were available (204 out of 322), the median Bray- Curtis distance between VLP and MGS metaviromes of the same faecal sample comprised \(0.86 \pm 0.20\) . This distance to the own concurrent sample was lower than to samples of unrelated individuals that averaged at \(0.99 \pm 0.05\) (p- value \(< 0.001\) , one- sided Wilcoxon rank sum test run through 1,000 permutations)."
+
+<--- Page Split --->
+
+However, it is important to bear in mind that the reported overlap is estimated using the concurrent MGS and VLP metaviromes. To truly estimate the efficacy of mining viruses solely from total metagenomes, it is necessary to run additional analyses involving VLP- independent reconstruction of virus genomes from total metagenomes and compare those to virus genomes reconstructed from concurrent VLP metaviromes.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The Authors have addressed all concerns raised in the previous round of Review. My only recommendation is to fix the "run_sample.csv" metadata file, as part of the file is indeed comma separated, while part is separated by semi- columns, and add the corresponding column headers rather than having all the relevant metadata collapsed in the "sample_attributes" field.  
+
+We agree with the Reviewer on that matter, however, the EGA web platform transforms metadata files submitted as a plain text file into SQL- like files for efficient parsing and searching within the whole archive. We have, therefore, created a script that reformats run_sample.csv back into the plain text with a tab separator and parses all relevant metadata in the respective columns. The script is accessible from the GitHub repository of the paper: https://github.com/GRONINGEN- MICROBIOME- CENTRE/LLNEXT_pilot/blob/main/Data_Access_EGA.md
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d636d1552621791c70664dbc9983bc6c062889f128817d323f7d856b66a1ed53/supplementary_0_Peer Review File__d636d1552621791c70664dbc9983bc6c062889f128817d323f7d856b66a1ed53_det.mmd

@@ -0,0 +1,448 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 508, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[66, 110, 363, 140]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[96, 152, 911, 220]]<|/det|>
+Transmission and Dynamics of Mother- Infant Gut Viruses during Pregnancy and Early Life  
+
+<|ref|>image<|/ref|><|det|>[[56, 732, 240, 782]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 733, 911, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 877, 133]]<|/det|>
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operatin g a transparent peer review scheme. This document only contains reviewer comments and rebuttal l etters for versions considered at Nature Communications.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 149, 296, 164]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[120, 179, 403, 195]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 238, 835, 292]]<|/det|>
+Garmeva et al present an analysis of a metagenomic analysis of bacteria and dsDNA virus in the stools of \(\sim 30\) mother and infant pairs followed in a longitudinal fashion. They examine the viral portion in the VLPs and via metagenomic sequencing without viral prep.  
+
+<|ref|>text<|/ref|><|det|>[[118, 336, 861, 408]]<|/det|>
+They reproduce previous observations on differences in bacterial and viral community composition between mothers and infants, on shared (maternal/infant) bacteria and bacteriophages which is important, but not novel. Relevant papers (not all): Lim et al. 2015, Zhao et al. 2017, McCann et al. 2018, Magsood et al. 2019, Liang et al. 2020, Walters et al 2023.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 453, 253, 468]]<|/det|>
+## Major comments:  
+
+<|ref|>text<|/ref|><|det|>[[117, 481, 870, 534]]<|/det|>
+1. Terms should be defined: metaviromes, whole metavirome, vOTU, metagenome in the beginning-it is confusing If they are referring to viruses detected in VLP prepped sample or virus in the portion of sample that did not get viral prep. This is key throughout and as written is not always clear.  
+
+<|ref|>text<|/ref|><|det|>[[117, 548, 875, 601]]<|/det|>
+2. Page 11 Line 211. This methodology assumes that any phage detected in the whole metagenome data set are prophage. This does not seem like a valid assumption, as active phage could be captured here as well.  
+
+<|ref|>text<|/ref|><|det|>[[115, 615, 861, 632]]<|/det|>
+3. Page 16 line 332. What is the rationale behind the conclusion these were through direct seeding.  
+
+<|ref|>text<|/ref|><|det|>[[118, 646, 451, 661]]<|/det|>
+4. Page 18 line 371. This is quite speculative.  
+
+<|ref|>text<|/ref|><|det|>[[115, 675, 857, 692]]<|/det|>
+5. Page 19 line 376. This is incorrect. Please see Walters et al Cell Host and Microbe, 2023 (Ref 22).  
+
+<|ref|>text<|/ref|><|det|>[[117, 706, 874, 777]]<|/det|>
+6. Page 21. Line 421-424. Three papers, Maqsood, Walters, and this one, find only a relatively small percent of shared viral component between mother and infant gut. Why invoke that small difference in percent shared is due to longitudinal sampling when the maternal virome is stable over time. Seems that the point is being missed here.  
+
+<|ref|>text<|/ref|><|det|>[[117, 792, 580, 807]]<|/det|>
+7. Page 21-22. Line 442-444 these conclusions are overstated.  
+
+<|ref|>text<|/ref|><|det|>[[117, 821, 870, 874]]<|/det|>
+8. Page 27. Line 598. Criteria 6 is concerning. Why are these necessarily thought to be viral? Were possible archaeal and bacterial sequences removed? Did the authors try blastx of these scaffolds? What did they hit by blastx? A table which outlines by which method called a virus would be helpful.  
+
+<|ref|>text<|/ref|><|det|>[[115, 888, 850, 922]]<|/det|>
+9. Page 28 Line 633. Why did authors choose this method to call temperate phage? Did this agree with other more commonly used tools for this?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 852, 119]]<|/det|>
+10. Page 29. Line 649. It is quite surprising this high level of predicted microbial host. This is not in keeping with other studies using iPhOp.  
+
+<|ref|>text<|/ref|><|det|>[[118, 195, 254, 209]]<|/det|>
+Minor comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 223, 844, 260]]<|/det|>
+1. Page 14 line 269. Please explain what is meant by accounting for prophages detected in whole metaviromes  
+
+<|ref|>text<|/ref|><|det|>[[118, 272, 825, 307]]<|/det|>
+2. Page 26. Line 556. Seems unusual that 3 out of 4 negative controls failed. Could the authors comment more on this.  
+
+<|ref|>text<|/ref|><|det|>[[118, 320, 518, 337]]<|/det|>
+3. Page 30 line 686. Is wilxocon the proper test here?  
+
+<|ref|>text<|/ref|><|det|>[[118, 440, 403, 456]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 500, 875, 555]]<|/det|>
+The Authors have substantially improved the manuscript by increasing the sampling size and frequency, and by improving the data analysis pipeline. While most of the concerns raised in the first round of Review have been addressed, two relevant issues still persist.  
+
+<|ref|>text<|/ref|><|det|>[[117, 598, 877, 763]]<|/det|>
+- The sequences and metadata are, differently from what stated by the authors in lines 844-845, not publicly available. The access policy associated to the provided EGA accession says "The Consortium is willing to consider applications from third-party researchers for access to the anonymized sequence data generated by the Project. [...] Access is conditional upon the availability of data and on signed agreement by the researcher to abide by the policies and conditions related to data ownership, disposal, ethical approval, confidentiality and commercialization referred herein". In fact, nor the fastq files nor the metadata can be downloaded. This is unacceptable. This kind of access restriction, entirely omitted in the manuscript, goes against the Journal Guidelines on data availability and against the FAIR data sharing principles.  
+
+<|ref|>text<|/ref|><|det|>[[118, 806, 851, 860]]<|/det|>
+- The use of "total metagenomes" vs "bacteriome" and "whole metavirome", vs "metavirome" vs "virome" is inconsistent throughput the manuscript and the meaning of these terms is not clearly described, nor in the text nor in the figures.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[116, 108, 342, 128]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 149, 466, 169]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 189, 879, 269]]<|/det|>
+Garmeva et al present an analysis of a metagenomic analysis of bacteria and dsDNA virus in the stools of \(\sim 30\) mother and infant pairs followed in a longitudinal fashion. They examine the viral portion in the VLPs and via metagenomic sequencing without viral prep.  
+
+<|ref|>text<|/ref|><|det|>[[115, 289, 861, 389]]<|/det|>
+They reproduce previous observations on differences in bacterial and viral community composition between mothers and infants, on shared (maternal/infant) bacteria and bacteriophages which is important, but not novel. Relevant papers (not all): Lim et al. 2015, Zhao et al. 2017, McCann et al. 2018, Magsood et al. 2019, Liang et al. 2020, Walters et al 2023.  
+
+<|ref|>text<|/ref|><|det|>[[115, 410, 267, 428]]<|/det|>
+Major comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 430, 859, 510]]<|/det|>
+1. Terms should be defined: metaviromes, whole metavirome, vOTU, metagenome in the beginning- it is confusing If they are referring to viruses detected in VLP prepped sample or virus in the portion of sample that did not get viral prep. This is key throughout and as written is not always clear.  
+
+<|ref|>text<|/ref|><|det|>[[115, 530, 874, 649]]<|/det|>
+We agree that the terms, as they were in the manuscript, led to confusion. For the sake of clarity we have now termed metaviromes as either MGS metaviromes or VLP metavirome. Additionally, we have created a box (Box 1) in the introduction as a glossary of the terms used in this paper. Lastly, we have changed the term microbiome to bacteriome in the text wherever applicable. We hope that our modifications increase clarity.  
+
+<|ref|>text<|/ref|><|det|>[[115, 668, 857, 827]]<|/det|>
+Hereby the text from Box 1: Bacteriome: community of all bacteria inhabiting an ecosystem Virome: community of all viruses inhabiting an ecosystem Total metagenome: the total sum of all genomes of microbes in a sample VLP metaviromes: the total sum of all genomes of viruses in a VLP-enriched sample MGS metaviromes: the total sum of all genomes of viruses in a total metagenome vOTU (virus operational taxonomic unit): species-rank (95% Average Nucleotide Identity, at 85% alignment fraction (relative to the shorter sequence)) virus groups  
+
+<|ref|>text<|/ref|><|det|>[[115, 847, 868, 907]]<|/det|>
+2. Page 11 Line 211. This methodology assumes that any phage detected in the whole metagenome data set are prophage. This does not seem like a valid assumption, as active phage could be captured here as well.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 108, 874, 208]]<|/det|>
+Original text: Given that the majority of temperate bacteriophages in the adult gut are in the form of prophages<sup>27</sup>, we compared the percentage of temperate bacteriophages in mothers and infants using the vOTUs detected in whole metaviromes, created by aligning the reads from whole metagenomes aligned to vOTU database reconstructed from metaviromes, thus representing the prophage content of the virome.  
+
+<|ref|>text<|/ref|><|det|>[[114, 228, 879, 288]]<|/det|>
+We agree with the reviewer that actively replicating temperate phages could be captured by this method. We have therefore rephrased our text. It now reads (lines 207- 212):  
+
+<|ref|>text<|/ref|><|det|>[[114, 289, 881, 409]]<|/det|>
+"Given that the majority of temperate bacteriophages in the adult gut are in the form of prophages<sup>27</sup>, we compared the percentage of temperate bacteriophages in mothers and infants using the vOTUs detected in MGS metaviromes, created by aligning the reads from total metagenomes aligned to the vOTU database reconstructed from VLP metaviromes, thus representing the prophage- inclusive temperate phage content of the virome."  
+
+<|ref|>text<|/ref|><|det|>[[114, 429, 848, 470]]<|/det|>
+3. Page 16 line 332. What is the rationale behind the conclusion these were through direct seeding.  
+
+<|ref|>text<|/ref|><|det|>[[114, 489, 872, 550]]<|/det|>
+Original text: "Among the 26 transmitted viruses, 21 were identified as virulent bacteriophages. Based on their predicted lifecycle, it is likely that they were transmitted through direct seeding of VLPs from the maternal to infant gut."  
+
+<|ref|>text<|/ref|><|det|>[[114, 569, 880, 729]]<|/det|>
+Since one of the criteria for inclusion of vOTUs in the strain- sharing analysis was the genome completeness (high- quality or complete genome predicted by CheckV or circularised genomes), and these 21 mentioned transmitted viruses did not have neither integrase nor site- specific recombinase genes, we concluded that they very likely exclusively undergo lytic infection cycle. For such phages, the only transmission mechanism available is the direct seeding of the viral particles. We agree with the reviewer that this might not have been clear in the original text. Hence, we have rephrased the text (lines 330- 333):  
+
+<|ref|>text<|/ref|><|det|>[[114, 730, 882, 810]]<|/det|>
+"Among the 26 transmitted viruses, 21 were identified as virulent bacteriophages. Based on their predicted lytic lifecycle, it is likely that they were directly seeded in the form of virus particles from the maternal to infant gut instead of being co- transmitted as prophages within the transmitted bacterial strains."  
+
+<|ref|>text<|/ref|><|det|>[[114, 830, 511, 850]]<|/det|>
+4. Page 18 line 371. This is quite speculative.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 857, 149]]<|/det|>
+Original text: Gene annotation showed that, in addition to carrying an integrase gene, L37775_LS1 also carries a CAZyme (Glycosyl hydrolases family 25, Supplementary Fig. 6d), indicating that this phage might be associated with infant feeding.  
+
+<|ref|>text<|/ref|><|det|>[[115, 168, 872, 249]]<|/det|>
+We agree that this is speculative and have adapted the sentence as follows: (lines 368- 370): "Gene annotation showed that, in addition to carrying an integrase gene, L37775_LS1 also carries a CAZyme (Glycosyl hydrolases family 25, Supplementary Fig. 6d)."  
+
+<|ref|>text<|/ref|><|det|>[[115, 268, 861, 309]]<|/det|>
+5. Page 19 line 376. This is incorrect. Please see Walters et al Cell Host and Microbe, 2023 (Ref 22).  
+
+<|ref|>text<|/ref|><|det|>[[115, 328, 860, 408]]<|/det|>
+Original text: To our knowledge, this is the only study to look at the maternal virome longitudinally during pregnancy, birth and after birth. In the maternal total microbiome, we observed a notable shift in composition between the first and second trimesters of pregnancy.  
+
+<|ref|>text<|/ref|><|det|>[[115, 428, 857, 488]]<|/det|>
+We have now changed this to read (lines 373- 375): "To our knowledge, this is one of the few studies that investigates the maternal virome longitudinally during pregnancy, birth and after birth."  
+
+<|ref|>text<|/ref|><|det|>[[115, 508, 870, 589]]<|/det|>
+6. Page 21. Line 421-424. Three papers, Maqsood, Walters, and this one, find only a relatively small percent of shared viral component between mother and infant gut. Why invoke that small difference in percent shared is due to longitudinal sampling when the maternal virome is stable over time. Seems that the point is being missed here.  
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 872, 728]]<|/det|>
+Original text: Our study revealed that, despite significant distinctions between the infant and maternal gut viromes, infants shared on average \(32.7\%\) of vOTUs with their mothers. This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer timeframe, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth.  
+
+<|ref|>text<|/ref|><|det|>[[115, 748, 852, 807]]<|/det|>
+We thank the reviewer for this comment. We would like to clarify why we believe that our use of longitudinal sampling is leading to the increased sharedness between maternal and infant microbiome:  
+
+<|ref|>text<|/ref|><|det|>[[145, 810, 840, 889]]<|/det|>
+1) Firstly, and most importantly, the study of Maqsood et al. used only one time point of mother and infant. This time was very close to birth. Their study of transmission from mother to infant will thus encompass mostly bacteria and viruses that are vertically transmitted during birth and not arising from the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[171, 89, 860, 127]]<|/det|>
+common shared environment. Hence we expect this percentage to be much smaller than the overall shared microbiome which our study takes into account.  
+
+<|ref|>text<|/ref|><|det|>[[144, 129, 875, 185]]<|/det|>
+2) In pregnancy we indeed show that overall the maternal gut virome remains stable. However, as seen in Figure 2e, there is turnover in viral species which also contributes to the sharedness.  
+
+<|ref|>text<|/ref|><|det|>[[144, 188, 877, 227]]<|/det|>
+3) Finally, having longitudinal sampling gives us a higher chance to detect shared contigs that would be missed when looking simply at one cross-sectional sample.  
+
+<|ref|>text<|/ref|><|det|>[[115, 247, 857, 389]]<|/det|>
+We have changed the text to clarify this (lines 420- 425): "Our study revealed that, despite significant distinctions between the infant and maternal gut viromes, infants shared on average 32.7% of vOTUs with their mothers. This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+<|ref|>text<|/ref|><|det|>[[115, 408, 661, 428]]<|/det|>
+7. Page 21-22. Line 442-444 these conclusions are overstated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 448, 879, 528]]<|/det|>
+Original text: We then show examples of how this occurs via direct transmission of bacteriophages and prophage induction following transmission of their bacterial hosts. We also show how infants obtain some viral strains from induction from bacteria that did not come from their mothers.  
+
+<|ref|>text<|/ref|><|det|>[[115, 548, 877, 668]]<|/det|>
+We disagree with the reviewer here. We specifically mention that we found examples of how this occurs. We have, however, reformulated the text to indicate yet again that this is our hypothesis (lines 442- 443): We then show examples of how this might occur via direct transmission of bacteriophages and prophage induction following transmission of their bacterial hosts. We also show how infants obtain some viral strains from induction from bacteria that did not come from their mothers.  
+
+<|ref|>text<|/ref|><|det|>[[115, 708, 872, 789]]<|/det|>
+8. Page 27. Line 598. Criteria 6 is concerning. Why are these necessarily thought to be viral? Were possible archaeal and bacterial sequences removed? Did the authors try blastx of these scaffolds? What did they hit by blastx? A table which outlines by which method called a virus would be helpful.  
+
+<|ref|>text<|/ref|><|det|>[[115, 809, 882, 867]]<|/det|>
+Original text: (6) being longer than 3 kbp with no hits (alignments \(>100\) nucleotides, \(90\%\) ANI, e- value of 10-10) to the nt database (release 249). 281,789 scaffolds fulfilled at least one of these six criteria.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 880, 348]]<|/det|>
+The usage of criteria #6 is in accordance with previously published studies \(^{1,2}\) . We agree with the reviewer that criteria #6 alone is not sufficient to call a sequence putatively viral. However, we would like to stress that in our pipeline, we use a robust post- virome discovery filtration process designed using the recent studies in the metaviromics field \(^{3 - }\) \(^{6}\) . Specifically, we employ screening of putative virus scaffolds for the presence of ribosomal RNA (rRNA) genes using a BLASTn search in the SILVA 138.1 NR99 rRNA genes database along with screening for ribosomal protein genes identified using a BLASTp search (e- value threshold of \(10^{- 10}\) ) against a subset of ribosomal protein sequences from the COG database (release 2020), as mentioned in the Methods section (lines 609- 611 and lines 585- 586) and exclude these (lines 619- 621). Through this comprehensive approach, we effectively eliminate potential archaeal and bacterial sequences, ensuring that only genuine viral scaffolds are considered in the downstream analysis.  
+
+<|ref|>text<|/ref|><|det|>[[114, 349, 877, 427]]<|/det|>
+Furthermore, there is an increasing body of evidence \(^{7,8}\) that suggests that the NR might contain many prophage proteins that are currently attributed to bacteria as prophages are highly prevalent in gut- associated bacteria. We thus did not blastx these scaffolds to the NR database.  
+
+<|ref|>text<|/ref|><|det|>[[114, 429, 878, 508]]<|/det|>
+We agree with the reviewer that a table encompassing the indications based on inclusion and exclusion criteria for the virus scaffolds would be helpful. In fact, this table can be found in Figshare along with the fasta file containing all virus scaffolds used in this study:  
+
+<|ref|>text<|/ref|><|det|>[[113, 509, 880, 630]]<|/det|>
+https://figshare.com/articles/dataset/Virus scaffolds reconstructed in the study b Mot her- Infant Gut Viruses and their Bacterial Hosts Transmission Patterns and Dynamics during Pregnancy and Early Life b NEXT Pilot and their metadata /23926593 We have now added this table to the Supplementary tables (S42: Metadata for virus genomes and genome fragments reconstructed in the study).  
+
+<|ref|>text<|/ref|><|det|>[[114, 650, 879, 689]]<|/det|>
+9. Page 28 Line 633. Why did authors choose this method to call temperate phage? Did this agree with other more commonly used tools for this?  
+
+<|ref|>text<|/ref|><|det|>[[114, 709, 880, 769]]<|/det|>
+Original text: Temperate bacteriophages were identified using either the presence of integrase genes or the co- presence of recombinase genes with Cl- repressor- like protein genes from the pVOGs annotation within a vOTU representative.  
+
+<|ref|>text<|/ref|><|det|>[[114, 790, 870, 889]]<|/det|>
+The choice for the method to predict the phage lifestyle was based on well- established research in the field (Shkoporov et al. \(^{1}\) Cell Host & Microbe.2019, Howard- Varona et al. \(^{9}\) ISME Journal. 2017. Roux et al. \(^{10}\) PeerJ. 2015). These studies pinpointed integrase and site- specific recombinases as the hallmark genes for the temperate phages. Since site- specific recombinases also regulate viral gene expression during the bacterial cell
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 88, 878, 250]]<|/det|>
+infection, our phage lifestyle prediction method included an additional condition of the presence of Cl- repressor- like protein genes (Paul.11 ISME Journal. 2008). Furthermore, the tools used in the viromics field for the lifestyle prediction like Prophage Hunter, PhiSpy, PHASTEST, BACPHLIP, and PhaTYP also rely on the presence of integrase gene within the virus sequence of interest in the marker- based parts of their pipelines (Song et al.12 Nucleic Acids Research. 2019, Akhter et al.13 Nucleic Acids Research. 2016, Arndt et al.14 Briefings in Bioinformatics,2019. Hockenberry et al.15 PeerJ. 2019. Shang et al.16 Briefings in Bioinformatics. 2022).  
+
+<|ref|>text<|/ref|><|det|>[[115, 269, 855, 309]]<|/det|>
+10. Page 29. Line 649. It is quite surprising this high level of predicted microbial host. This is not in keeping with other studies using iPhOp.  
+
+<|ref|>text<|/ref|><|det|>[[115, 328, 875, 388]]<|/det|>
+Original text: In total, the microbial host was predicted for 68,299 of 102,210 viral scaffolds (67.3%) at the genus level and for 85,135 (82.9%) of all vOTUs at the species level.  
+
+<|ref|>text<|/ref|><|det|>[[114, 408, 861, 580]]<|/det|>
+Benchmarking of the iPhOp framework in the original article (Roux et al.17, 2023) showed that the amount of successful host predictions depends on the ecosystem studied. According to the paper, more hosts could be found for predicted viruses in human- associated samples. In our study, the percentage of the viruses with predicted hosts is concordant with the benchmark: "For human- associated microbiomes, about 89% of the nonredundant high- quality genomes had a host predicted using iPhOp, including 57% with very high confidence predictions (iPhOp score \(\geq 95\) ). (Roux et al. 2023)  
+
+<|ref|>text<|/ref|><|det|>[[114, 599, 881, 720]]<|/det|>
+To the best of our knowledge, three other studies have employed the iPhOp framework. Specifically, Ter Horst et al.18 (2023) focused on the virome of wetlands, Peng et al.19 (2023) on the virome of deep- sea cold seep sediments, and Saenz et al.20 (2023) on the virome of grass silage. Given that these studies explored non- human associated ecosystems, the proportion of viruses with hosts predicted in their research is anticipated to vary from our findings.  
+
+<|ref|>text<|/ref|><|det|>[[115, 740, 877, 800]]<|/det|>
+Based on the details previously mentioned, we contend that while the percentage of the predicted host is high, it aligns with the benchmark set by the framework's authors (Roux et al., 2023).  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 821, 268, 839]]<|/det|>
+## Minor comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 841, 805, 880]]<|/det|>
+1. Page 14 line 269. Please explain what is meant by accounting for prophages detected in whole metaviromes
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 863, 168]]<|/det|>
+Original text: As pioneer viruses in the infant gut are thought to be primarily temperate phages induced from the first gut bacterial colonisers<sup>20</sup>, we next assessed the sharedness of maternal to infant vOTUs while accounting for prophages detected in whole metaviromes.  
+
+<|ref|>text<|/ref|><|det|>[[115, 189, 877, 328]]<|/det|>
+We thank the reviewer for the comment. This means we expanded our VLP metaviromes with the temperate bacteriophages detected in MGS metaviromes and not concurrent VLP metaviromes (prophages). We have edited the sentence to make it clearer (lines 265- 268): "As pioneer viruses in the infant gut are thought to be primarily temperate phages induced from the first gut bacterial colonisers<sup>20</sup>, we next assessed the sharedness of maternal to infant vOTUs while accounting for prophages detected only in MGS metaviromes."  
+
+<|ref|>text<|/ref|><|det|>[[115, 349, 872, 389]]<|/det|>
+2. Page 26. Line 556. Seems unusual that 3 out of 4 negative controls failed. Could the authors comment more on this.  
+
+<|ref|>text<|/ref|><|det|>[[114, 409, 872, 570]]<|/det|>
+It's not uncommon for negative controls to display such behavior, in fact this is totally in line with expectations, especially, when no whole genome amplification was applied to the samples. Negative controls, by design, have minimal or no target microbial reads. As a result, during the PCR amplification within the library preparation process, there might be very little template available for amplification. Hence genomic library preparation was unsuccessful and these samples did not get sequenced. It now reads as: "Three out of four negative VLP metavirome controls failed library preparation and therefore could not be sequenced <...>".  
+
+<|ref|>text<|/ref|><|det|>[[115, 590, 582, 609]]<|/det|>
+3. Page 30 line 686. Is wilcoxon the proper test here?  
+
+<|ref|>text<|/ref|><|det|>[[115, 611, 866, 670]]<|/det|>
+As the Shannon Diversity Index is not normally distributed in our dataset, and we took only one time point per individual for this comparison into account (thus accounting for repeated measures is not necessary), the use of Wilcoxon tests is appropriate.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 709, 466, 729]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 750, 866, 829]]<|/det|>
+The Authors have substantially improved the manuscript by increasing the sampling size and frequency, and by improving the data analysis pipeline. While most of the concerns raised in the first round of Review have been addressed, two relevant issues still persist.  
+
+<|ref|>text<|/ref|><|det|>[[115, 850, 866, 890]]<|/det|>
+- The sequences and metadata are, differently from what stated by the authors in lines 844-845, not publicly available. The access policy associated to the provided EGA
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 88, 868, 249]]<|/det|>
+accession says "The Consortium is willing to consider applications from third- party researchers for access to the anonymized sequence data generated by the Project. [...] Access is conditional upon the availability of data and on signed agreement by the researcher to abide by the policies and conditions related to data ownership, disposal, ethical approval, confidentiality and commercialization referred herein". In fact, nor the fastq files nor the metadata can be downloaded. This is unacceptable. This kind of access restriction, entirely omitted in the manuscript, goes against the Journal Guidelines on data availability and against the FAIR data sharing principles.  
+
+<|ref|>text<|/ref|><|det|>[[114, 269, 883, 489]]<|/det|>
+Thank you for this very important point. We would like to provide clarity regarding the data sharing procedure. In the previous metagenomics studies of our group, such as Chen et al. (Cell, 2021; doi: 10.1016/j.cell.2021.03.024) and Gacesa et al. (Nature, 2022; doi: 10.1038/s41586- 022- 04567- 7), we always uploaded our data to the European Genome- phenome Archive (EGA). Similarly, for the current study, our dataset — which includes sample information, all phenotypes, family structures, and quality- trimmed sequencing reads — has been securely archived in the EGA under study ID: EGAS00001005969. Our group has a history of fulfilling all data requests. We specifically use EGA for data sharing, as it is an approved repository for functional genomics data according to the Nature Data Repository Guidance. All our prior publications utilized this platform for multi- omic data sharing.  
+
+<|ref|>text<|/ref|><|det|>[[114, 509, 875, 649]]<|/det|>
+To access our dataset, interested parties must first register with the EGA. After registration, they must read the Data Access form: https://groningenmicrobiome.org/?page_id=2598. This contains a form that requires certain essential details and an acknowledgment from the requester to not use the data for commercial purposes. This form can be found here: https://forms.gle/zbJLMnojys3VVKPY8. Requests will be granted for researchers affiliated with an academic, non- profit, or government institution.  
+
+<|ref|>text<|/ref|><|det|>[[115, 670, 559, 689]]<|/det|>
+In addition to this we have specified this in the text:  
+
+<|ref|>text<|/ref|><|det|>[[115, 710, 816, 769]]<|/det|>
+Original text: Sample information, basic phenotypes, family structure and quality- trimmed sequencing reads can be found in the EGA repository (study ID: EGAS00001005969).  
+
+<|ref|>text<|/ref|><|det|>[[115, 790, 852, 870]]<|/det|>
+This has been changed to (lines 848- 851): "EGA repository (study ID: EGAS00001005969). Access to the LLNEXT Project data will be granted to qualified researchers will be governed by the provisions laid out in the LLNEXT Data Access Agreement: https://groningenmicrobiome.org/?page_id=2598"
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 881, 210]]<|/det|>
+Furthermore, all our contigs have been deposited on figshare and are freely accessible here: https://figshare.com/articles/dataset/Virus_scaffolds_reconstructed_in_the_study_b_Mother-Infant_Gut_Viruses_and_their_Bacterial_Hosts_Transmission_Patterns_and_Dynamics_during_Pregnancy_and_Early_Life_b_NEXT_Pilot_and_their_metadata_/23926593  
+
+<|ref|>text<|/ref|><|det|>[[115, 228, 875, 288]]<|/det|>
+- The use of "total metagenomes" vs "bacteriome" and "whole metavirome", vs "metavirome" vs "virome" is inconsistent throughput the manuscript and the meaning of these terms is not clearly described, nor in the text nor in the figures.  
+
+<|ref|>text<|/ref|><|det|>[[115, 308, 870, 408]]<|/det|>
+We agree with the reviewer that the use of introduced terms was inconsistent in the original text. For the sake of clarity we have now termed metaviromes as either MGS metaviromes or VLP metavirome. Additionally, we have created a box (Box 1) in the introduction as a glossary of the terms used in this paper. Lastly, we have changed the term microbiome to bacteriome in the text wherever applicable to avoid confusion.  
+
+<|ref|>text<|/ref|><|det|>[[116, 429, 355, 447]]<|/det|>
+Hereby the text from Box 1:  
+
+<|ref|>text<|/ref|><|det|>[[115, 450, 855, 589]]<|/det|>
+Bacteriome: community of all bacteria inhabiting an ecosystem  Virome: community of all viruses inhabiting an ecosystem  Total metagenome: the total sum of all genomes of microbes in a sample  VLP metaviromes: the total sum of all genomes of viruses in a VLP- enriched sample  MGS metaviromes: the total sum of all genomes of viruses in a total metagenome  vOTU (virus operational taxonomic unit): species- rank (95% Average Nucleotide Identity, at 85% alignment fraction (relative to the shorter sequence)) virus groups  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 629, 222, 647]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[113, 667, 856, 880]]<|/det|>
+1. Shkoporov, A. N. et al. The Human Gut Virome Is Highly Diverse, Stable, and Individual Specific. Cell Host Microbe 26, 527-541.e5 (2019).  
+2. Shkoporov, A. N. et al. Reproducible protocols for metagenomic analysis of human faecal phageomes. Microbiome 6, 68 (2018).  
+3. Clooney, A. G. et al. Whole-Virome Analysis Sheds Light on Viral Dark Matter in Inflammatory Bowel Disease. Cell Host Microbe 26, 764-778.e5 (2019).  
+4. Garmaeva, S. et al. Stability of the human gut virome and effect of gluten-free diet. Cell Rep.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[135, 88, 290, 107]]<|/det|>
+35, 109132 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[113, 118, 840, 172]]<|/det|>
+5. Shkoporov, A. N. et al. The Human Gut Virome Is Highly Diverse, Stable, and Individual Specific. Cell Host Microbe 26, 527-541. e5 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[113, 183, 861, 237]]<|/det|>
+6. Gulyaeva, A. et al. Diversity and Ecology of Caudoviricetes Phages with Genome Terminal Repeats in Fecal Metagenomes from Four Dutch Cohorts. Viruses 14, (2022).  
+
+<|ref|>text<|/ref|><|det|>[[113, 247, 874, 300]]<|/det|>
+7. Dahlman, S. et al. Temperate gut phages are prevalent, diverse, and predominantly inactive. bioRxiv (2023) doi:10.1101/2023.08.17.553642.  
+
+<|ref|>text<|/ref|><|det|>[[113, 311, 861, 364]]<|/det|>
+8. Maxwell Anthenelli et al. Phage and bacteria diversification through a prophage acquisition ratchet. bioRxiv 2020.04.08.028340 (2020) doi:10.1101/2020.04.08.028340.  
+
+<|ref|>text<|/ref|><|det|>[[113, 375, 866, 428]]<|/det|>
+9. Howard-Varona, C., Hargreaves, K. R., Abedon, S. T. & Sullivan, M. B. Lysogeny in nature: mechanisms, impact and ecology of temperate phages. ISME J. 11, 1511-1520 (2017).  
+
+<|ref|>text<|/ref|><|det|>[[113, 439, 860, 492]]<|/det|>
+10. Roux, S., Enault, F., Hurwitz, B. L. & Sullivan, M. B. VirSorter: mining viral signal from microbial genomic data. PeerJ 3, e985 (2015).  
+
+<|ref|>text<|/ref|><|det|>[[113, 503, 878, 556]]<|/det|>
+11. Paul, J. H. Prophages in marine bacteria: dangerous molecular time bombs or the key to survival in the seas? ISME J. 2, 579-589 (2008).  
+
+<|ref|>text<|/ref|><|det|>[[113, 567, 825, 620]]<|/det|>
+12. Song, W. et al. Prophage Hunter: an integrative hunting tool for active prophages. Nucleic Acids Res. 47, W74-W80 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[113, 631, 841, 684]]<|/det|>
+13. Akhtar, M. M., Micolucci, L., Islam, M. S., Olivieri, F. & Procopio, A. D. Bioinformatics tools for microRNA dissection. Nucleic Acids Res. 44, 24-44 (2016).  
+
+<|ref|>text<|/ref|><|det|>[[113, 695, 850, 748]]<|/det|>
+14. Arndt, D., Marcu, A., Liang, Y. & Wishart, D. S. PHAST, PHASTER and PHASTEST: Tools for finding prophage in bacterial genomes. Brief. Bioinform. 20, 1560-1567 (2019).  
+
+<|ref|>text<|/ref|><|det|>[[113, 758, 850, 812]]<|/det|>
+15. Hockenberry, A. J. & Wilke, C. O. BACPHLIP: predicting bacteriophage lifestyle from conserved protein domains. PeerJ 9, e11396 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[113, 822, 867, 876]]<|/det|>
+16. Shang, J., Tang, X. & Sun, Y. PhaTYP: predicting the lifestyle for bacteriophages using BERT. Brief. Bioinform. 24, bbac487 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[113, 886, 835, 908]]<|/det|>
+17. Roux, S. et al. iPhOP: An integrated machine learning framework to maximize host
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[135, 88, 803, 140]]<|/det|>
+prediction for metagenome- derived viruses of archaea and bacteria. PLOS Biol. 21, e3002083 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[115, 152, 848, 268]]<|/det|>
+18. Anneliek M. ter Horst, Jane D. Fudyma, Jacqueline L. Sones, & Joanne B. Emerson. Dispersal, habitat filtering, and eco-evolutionary dynamics as drivers of local and global wetland viral biogeography. bioRxiv 2023.04.28.538735 (2023) doi:10.1101/2023.04.28.538735.  
+
+<|ref|>text<|/ref|><|det|>[[115, 280, 830, 363]]<|/det|>
+19. Peng, Y. et al. Viruses in deep-sea cold seep sediments harbor diverse survival mechanisms and remain genetically conserved within species. ISME J. 17, 1774–1784 (2023).  
+
+<|ref|>text<|/ref|><|det|>[[115, 375, 872, 428]]<|/det|>
+20. Sáenz, J. S., Rios-Galicia, B., Rehkugler, B. & Seifert, J. Dynamic Development of Viral and Bacterial Diversity during Grass Silage Preservation. Viruses 15, (2023).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 308, 100]]<|/det|>
+REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[118, 145, 404, 161]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 205, 256, 220]]<|/det|>
+Major Comments:  
+
+<|ref|>text<|/ref|><|det|>[[117, 265, 878, 300]]<|/det|>
+1. The fact that this is an analysis of only dsDNA virome should be more prominently noted such as in the abstract etc.  
+
+<|ref|>text<|/ref|><|det|>[[118, 343, 866, 416]]<|/det|>
+2. "This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of vOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+<|ref|>text<|/ref|><|det|>[[118, 458, 858, 531]]<|/det|>
+One of the other major differences between this paper and others, is the use of VLP plus metagenomic sequencing without viral prep. What is the overlap between the VLP virome and the MGS virome? This is important to share as many groups are mining metagenomic data for virome analysis.  
+
+<|ref|>text<|/ref|><|det|>[[118, 636, 404, 652]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 695, 721, 712]]<|/det|>
+The Authors have addressed all concerns raised in the previous round of Review.  
+
+<|ref|>text<|/ref|><|det|>[[118, 726, 850, 780]]<|/det|>
+My only recommendation is to fix the "run_sample.csv" metadata file, as part of the file is indeed comma separated, while part is separated by semi- columns, and add the corresponding column headers rather than having all the relevant metadata collapsed in the "sample_attributes" field.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[119, 84, 369, 101]]<|/det|>
+## REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[119, 117, 457, 134]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[119, 149, 279, 166]]<|/det|>
+Major Comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 182, 878, 216]]<|/det|>
+1. The fact that this is an analysis of only dsDNA virome should be more prominently noted such as in the abstract etc.  
+
+<|ref|>text<|/ref|><|det|>[[119, 231, 879, 298]]<|/det|>
+Original text (lines 20- 22): "We longitudinally assessed the composition of gut viruses and their bacterial hosts in 322 total metagenomes and 205 Virus Like Particle (VLP) metaviromes from 30 mothers during and after pregnancy and from their 32 infants during their first year of life."  
+
+<|ref|>text<|/ref|><|det|>[[118, 313, 880, 428]]<|/det|>
+We agree with the Reviewer and have edited the abstract accordingly (lines 4- 8): "To study the development of the infant gut virome over time and the factors that shape it, we longitudinally assess the composition of gut viruses and their bacterial hosts in 30 women during and after pregnancy and in their 32 infants during their first year of life. Using shotgun metagenomic sequencing applied to dsDNA extracted from Virus- Like Particles (VLPs) and bacteria, we generate 205 VLP metaviromes and 322 total metagenomes."  
+
+<|ref|>text<|/ref|><|det|>[[118, 428, 880, 528]]<|/det|>
+We then mention dsDNA gut virome in 205 VLP metaviromes as early as in the first sentence of Results (line 63- 66): "We profiled the gut microbiome (primarily referred to as the bacteriome) in 322 total metagenome samples and the double- strand DNA (dsDNA) gut virome in 205 VLP metavirome samples from 30 mothers and their 32 term- born infants (including 2 twin pairs) collected longitudinally from pregnancy to 12 months after birth (Fig. 1a; Supplementary Fig. 1a, b)."  
+
+<|ref|>text<|/ref|><|det|>[[118, 543, 879, 625]]<|/det|>
+2. "This difference in sharedness may be attributed to the fact that our study encompassed longitudinal samples from both mother and infant for a longer time frame after birth, whereas Maqsood et al. looked at the sharing of VOTUs between mother and infant cross-sectionally around birth which would mostly represent vertical transmission viruses during birth".  
+
+<|ref|>text<|/ref|><|det|>[[118, 640, 879, 707]]<|/det|>
+One of the other major differences between this paper and others, is the use of VLP plus metagenomic sequencing without viral prep. What is the overlap between the VLP virome and the MGS virome? This is important to share as many groups are mining metagenomic data for virome analysis.  
+
+<|ref|>text<|/ref|><|det|>[[118, 722, 879, 771]]<|/det|>
+We agree with the Reviewer completely and now report the overlap between the VLP metaviromes and MGS metaviromes in Methods section "Generation of MGS metaviromes", lines 657- 664:  
+
+<|ref|>text<|/ref|><|det|>[[118, 772, 879, 904]]<|/det|>
+"The average alignment rate from total metagenomes to the curated virus database was \(50.6 \pm 18.2\%\) . As for VLP metaviromes, a count table was generated and transformed for MGS metaviromes. For samples where both VLP and MGS metaviromes were available (204 out of 322), the median Bray- Curtis distance between VLP and MGS metaviromes of the same faecal sample comprised \(0.86 \pm 0.20\) . This distance to the own concurrent sample was lower than to samples of unrelated individuals that averaged at \(0.99 \pm 0.05\) (p- value \(< 0.001\) , one- sided Wilcoxon rank sum test run through 1,000 permutations)."
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 880, 183]]<|/det|>
+However, it is important to bear in mind that the reported overlap is estimated using the concurrent MGS and VLP metaviromes. To truly estimate the efficacy of mining viruses solely from total metagenomes, it is necessary to run additional analyses involving VLP- independent reconstruction of virus genomes from total metagenomes and compare those to virus genomes reconstructed from concurrent VLP metaviromes.  
+
+<|ref|>text<|/ref|><|det|>[[118, 215, 457, 232]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 248, 880, 330]]<|/det|>
+The Authors have addressed all concerns raised in the previous round of Review. My only recommendation is to fix the "run_sample.csv" metadata file, as part of the file is indeed comma separated, while part is separated by semi- columns, and add the corresponding column headers rather than having all the relevant metadata collapsed in the "sample_attributes" field.  
+
+<|ref|>text<|/ref|><|det|>[[118, 345, 880, 461]]<|/det|>
+We agree with the Reviewer on that matter, however, the EGA web platform transforms metadata files submitted as a plain text file into SQL- like files for efficient parsing and searching within the whole archive. We have, therefore, created a script that reformats run_sample.csv back into the plain text with a tab separator and parses all relevant metadata in the respective columns. The script is accessible from the GitHub repository of the paper: https://github.com/GRONINGEN- MICROBIOME- CENTRE/LLNEXT_pilot/blob/main/Data_Access_EGA.md
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d653731047956cc0b103cf02a2f64e949f0a906e6db2d343242ba37fff97238a/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d653731047956cc0b103cf02a2f64e949f0a906e6db2d343242ba37fff97238a/supplementary_0_Peer Review File__d653731047956cc0b103cf02a2f64e949f0a906e6db2d343242ba37fff97238a.mmd

@@ -0,0 +1,17 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Bioelectric stimulation controls tissue shape and size  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operating a transparent peer review scheme. This document only contains reviewer comments and rebuttal letters for versions considered at Nature Communications.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d653731047956cc0b103cf02a2f64e949f0a906e6db2d343242ba37fff97238a/supplementary_0_Peer Review File__d653731047956cc0b103cf02a2f64e949f0a906e6db2d343242ba37fff97238a_det.mmd

@@ -0,0 +1,19 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[69, 110, 362, 140]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[70, 155, 794, 180]]<|/det|>
+Bioelectric stimulation controls tissue shape and size  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 783]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 912, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 879, 134]]<|/det|>
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operating a transparent peer review scheme. This document only contains reviewer comments and rebuttal letters for versions considered at Nature Communications.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6a4e80f7d8d1230ce2e0fe001d740ad7fa02a6213d5a611cfa2992ed9c0d451/images_list.json

@@ -0,0 +1,25 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_2c.jpg",
+    "caption": "Supplementary Figure 2c. Starch at 84 h normalized to volumetric biomass in WT cells.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  },
+  {
+    "type": "image",
+    "img_path": "images/Supplementary_Figure_3b.jpg",
+    "caption": "Supplementary Fig. 3b. Plot of Principal Components (Dim) 4 and 8 which distinguish strains.",
+    "footnote": [],
+    "bbox": [
+      [
+        262,
+        365,
+        730,
+        586
+      ]
+    ],
+    "page_idx": 6
+  }
+]

peer_reviews/supplementary_0_Peer Review File__d6a4e80f7d8d1230ce2e0fe001d740ad7fa02a6213d5a611cfa2992ed9c0d451/supplementary_0_Peer Review File__d6a4e80f7d8d1230ce2e0fe001d740ad7fa02a6213d5a611cfa2992ed9c0d451.mmd

@@ -0,0 +1,249 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis  
+
+![](images/Supplementary_Figure_2c.jpg)
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+In the manuscript entitled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis" by Jeffers et al., the authors conducted an extensive investigation into the intricate interplay between glucose, HXK1, and Fe in regulating the metabolism of the green alga C. zofingiensis. The team untangled the roles of these three factors in shifting the trophic modes of this green alga. Through proteomics analysis, the authors shed new light on the variations in protein abundances across distinct trophic conditions. This study tackles timely research questions and provides compelling evidence to support its conclusions. Certain aspects of the study demand refinement, particularly some methods employed for data analysis. Notably, specific datasets are absent for particular experimental conditions, see details in the following comments:  
+
+Figure 1: The plots of normalized chlorophyll fluorescence and delta O2 nmol are provided for only two conditions, while the remaining panels encompass all four conditions, which raises questions. To enhance transparency, it's recommended that data for - Fe - Glc and +Fe +Glc conditions be added to panels b and d. This data indicates that replete glucose affected (elevated) growth rate more than supplementation of Fe. Were the glucose consumption rates measured for these experiments? Please provide this information.  
+
+Throughout the manuscript, volumetric biomass emerges as a primary factor for defining growth and normalizing various characteristics such as O2 evolution. However, an inconsistency arises in the choice of normalizing factors. For instance, "Total chlorophyll" is normalized by the number of cells in one instance and by volumetric biomass in another. While the explanation provided in lines 212- 213 is sound, it would be advantageous to substantiate this approach with either new or previously published data. By looking at the cell images in Fig. 2, it looks like cells in - Fe - Glc may contain a similar amount of starch as cells in +Fe +Glc. This is very critical data for the main outcomes of the manuscript. There is also some inconsistency with other parts that need clarification. For example WT culture: Photosynthetic efficiency under - Fe +Glc is almost negligible compared to +Fe +Glc condition in Fig 1 and Fig 4; the same trend is shown for Total chlorophyll in Fig. 4. However, Total chlorophyll under +Fe +Glc is as low as in - Fe +Glc condition - this is not in agreement with one of the main conclusions of this work.  
+
+Proteomics data: the Principal Component Analysis (PCA) reveals well- separated treatment groups; however, a notable lack of distinction between WT and mutant groups raises questions. Providing an explanation for this observation is imperative. On a broader note, the downstream analysis of the proteomics data appears to rely heavily on fitting to linear model equations. It is
+
+<--- Page Split --->
+
+difficult or almost impossible without further validation to assess if the data is overfitted or generated unrealistic abundances of proteins. It simply can be seen in differences between model- predicted/normalized protein abundances ranges in each subsets of the data. The authors might want to ensure, using statistical methods, that conclusions are not a result of overfitting.  
+
+Figure 4a, please correct the abbreviation from "Wt" to "WT".  
+
+Reviewer #2 (Remarks to the Author):  
+
+Here the authors have performed a rigorous and comprehensive set of experiments designed to tease out the intricate relationship between glucose, regulator gene HXK1, and iron in governing photosynthesis vs respiration in gene Alga. The work revealed that multiple effectors have a stronger impact on organism response than individual effectors operating alone (an important consideration for things like drug design). The study also demonstrates that these protein response mechanisms are likely conserved across many plants and alga. The heterotrophic conditions studied enabled the authors to tease apart protein membership in various important biological functions like photosynthesis and the accumulation of TAG. The authors fit a linear model to their proteomics data from these heterotrophic condition studies to determine which conditions exercised the greatest control over protein abundance.  
+
+Overall, the results are quite compelling. The study and model are able to tease apart different groups of proteins that behave and respond differentially to the tested conditions. The linear model is extremely simple, yet also very effective at binning the proteins based on their distinctive response patterns. Generally the studies are comprehensive, rigorous, well documented, and well described. The writing is excellent. The work is exciting. The authors have established an approach for studying protein function genome- wide using mixotrophic growth that could be replicated in many other systems, and in the process of doing so, they learned a great deal about the roles all the proteins in this particular species play in photosynthesis, energy storage, and respiration. The role iron resource sharing between competing pathways that both have iron requirements plays in dictating how the cell selects its energy strategy is deeply enlightening. This reinforces a growing view that resource allocation is a major driving design consideration in cell regulation. Essentially, this unveils a fundamental design principle for life.  
+
+I have only a few suggestions, which I leave to the authors' discretion:
+
+<--- Page Split --->
+
+1. ) The protein expression heatmaps in figure 3e are tough to interpret. That said, I don't know that there is a better way of doing this. The current representation is very compact, concise, and information rich.  
+
+2. ) I'd like to have a little more information on how much data agreed with expectations vs being surprising. In the discussion, the authors raise FNR1 as surprising for not changing its expression profile despite being in subsystems that general did change in the tested conditions. How many other proteins were like this? It would be interesting to see a break down of how many genes stayed constant despite being involved in one of the differentially expressed subsystems, how many genes change expression despite being in static subsystems, and so on. This speaks to a potential boon from this work in the form of improved annotations.  
+
+3. ) Proteomics data not only provides protein abundance information, but also information on many post-translational modifications. Is there evidence of shifts in PTMs across the studied conditions? Particularly for proteins with abundances that stayed constant?  
+
+Reviewer #3 (Remarks to the Author):  
+
+In their paper titled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis", Jeffers and colleagues explore the response of photosynthesis in the green alga Chromochloris zofingiensis to the presence of glucose, and its dependence on iron supplementation. This work included the characterization of the physiological response of these algae to a combination of iron and glucose availability treatments, and was also complemented with studies on mutants of Hexokinase- 1 encoding gene. The main dataset generated in this work is the proteome response of WT and HXK1 mutants (one of them might still demonstrate some expression) to these combined treatments, confirming the previously reported physiological strategy of Chromochloris zofingiensis, including dnFAS under - Fe+Glc conditions. These proteomes are further studied for novel genes associated with photosynthesis which are downregulated, and for better understanding of iron transport at the subcellular level, under these conditions. Finally, Jeffers and colleagues discuss means by which targeting FDX and FAD encoding genes in C. zofingiensis might be beneficial for biofuel- directed synthetic approaches.  
+
+I found this work interesting, though the role of iron in the carbon response has been characterized in other model green algae. In addition, there are some issues which require clarification or I think must be clarified, some of which are major and may limit the interoperation that can be made based on this work, including the choice of (rich, peptone- based) medium, extremely low levels of iron in the medium, and the dual role of HXK1 which cannot be fully dissected.
+
+<--- Page Split --->
+
+Detailed points:  
+
+It is unclear to this reviewer why the authors chose a rich medium for an experiment testing effects of glucose addition on PS activity. The rich medium provides a background 'noise' for the heterotrophic conditions elicited by Glc addition. This may be related to the fact that no positive net PS rates are reported in figure 1d (maybe they appear in the - glc controls but they are not presented). Incidentally, in my eyes net rather than gross PS rates should be the values presented for various reasons (the nonlinearity of respiration in dark and light conditions is one) both in general, and especially in the question at hand.  
+
+Line 110- 112 - mixotrophy is highly unlikely to explain the 4- fold increase in biomass in the \(+Fe + Glc\) in view of the negative slopes of net PS as reported in Figure 1d. Moreover, such large increase implies that the cells are Fe- starved, and not just limited, which may limit any interpretation of resource allocation between respiration and photosynthesis, as stress response, protein aggregation and longer reversal time of activity may play a role (see e.g. Devadaus et al 2021).  
+
+Lines 119- 120 - the second part of the sentence may be completely wrong. I agree that - Fe is the primary nutritional limitation in the experiments performed by the authors, but this on its own doesn't mean that 'Fe is uniquely required for Glc- mediated repression of photosynthesis'. To argue that, the authors must first show that they generate the same limitation for all other nutrients (e.g. as shown in Figure 1e for Fe) and then supplement the cells with each of them separately.  
+
+Lines 125- 127 - to make such claims, the authors need to provide more than 3 images per treatment in Supp_dataset 2. What more, I am not convinced that 'starch was more prevalent inside +Glc plastids' even in the images provided, but this actually support the authors claims regarding the metabolic importance of Fe.  
+
+Line 142 - HXK1 mutants cannot fully allow 'distinguish general +Glc response.' as some of the 'general' response is not fully HXK1- independent and is also mediated through its (main) role as a metabolic/glycolytic and not just signaling enzyme. I don't think it's only a terminology issue, but also an important point to qualify the limitations of the analysis (e.g. if a specific signaling associated downstream of HXK1 would have been known and targeted, this could have provided the distinction declared here), which is also supported the PCA where strain identity was only separated in PC4 and PC8.  
+
+Line 215- 216 - any suggestion for the smaller investment in antennae? Less cost- effective under mixotrophy? More light penetration per biomass with these cells?  
+
+Line 340 - I find the terminology "highest/lowest- in- TAG" potentially inaccurate and at a certain point confusing. Can the authors guarantee that TAG accumulation is the only/main process defining the +Glc treatments?
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+In the manuscript entitled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis" by Jeffers et al., the authors conducted an extensive investigation into the intricate interplay between glucose, HXK1, and Fe in regulating the metabolism of the green alga C. zofingiensis. The team untangled the roles of these three factors in shifting the trophic modes of this green alga. Through proteomics analysis, the authors shed new light on the variations in protein abundances across distinct trophic conditions. This study tackles timely research questions and provides compelling evidence to support its conclusions. Certain aspects of the study demand refinement, particularly some methods employed for data analysis. Notably, specific datasets are absent for particular experimental conditions, see details in the following comments:  
+
+Response: We thank Reviewer #1 for positive comments on our research questions, experimental design and conclusions. We appreciate your suggestions and recommendations, and we have added the mentioned datasets.  
+
+Figure 1: The plots of normalized chlorophyll fluorescence and delta O2 nmol are provided for only two conditions, while the remaining panels encompass all four conditions, which raises questions. To enhance transparency, it's recommended that data for - Fe - Glc and +Fe +Glc conditions be added to panels b and d.  
+
+Response: Thank you for your suggestion. We updated Figure 1 with all treatments showing raw data for chlorophyll fluorescence. Because of space, we moved delta O2 data to the supplement and show data for WT and mutants (Supplementary Fig. 6d).  
+
+This data indicates that replete glucose affected (elevated) growth rate more than supplementation of Fe.  
+
+Response: Thank you for highlighting this point. We agree and clarified this in the manuscript in lines 112- 113.  
+
+Were the glucose consumption rates measured for these experiments? Please provide this information.  
+
+Response: Thank you for the suggestion. We have added glucose consumption data to the manuscript in lines 110- 111 and to the methods in 579- 582.  
+
+Throughout the manuscript, volumetric biomass emerges as a primary factor for defining growth and normalizing various characteristics such as O2 evolution. However, an inconsistency arises in the choice of normalizing factors. For instance, "Total chlorophyll" is normalized by the number of cells in one instance and by volumetric biomass in another. While the explanation provided in lines 212- 213 is sound, it would be advantageous to substantiate this approach with either new or previously published data.  
+
+Response: Thank you for your detailed points on complexities in these data. Because C. zofingiensis cell volume increases dramatically both from adding glucose and between - Fe+Glc
+
+<--- Page Split --->
+
+and \(+Fe + Glc\) (Supplementary Fig. 2a), the normalization is a challenge. We find it logical to look at O2 evolution normalized by biomass so that one is comparing the same amount of cellular material, and this is what was done in previous publications (Roth et al. 2019 Plant Cell, Roth et al. 2019 Comms Biol). We have updated Fig. 4 to include both total chlorophyll normalizations (Fig. 4e,f) and discussed these data on lines 222- 225.  
+
+By looking at the cell images in Fig. 2, it looks like cells in - Fe - Glc may contain a similar amount of starch as cells in \(+e + Glc\) . This is very critical data for the main outcomes of the manuscript. Response: Thank you for the suggestion. We added starch concentration measurements to the manuscript in lines 130- 132 and Supplementary Fig. 2c as well as the methods in In 571- 578.  
+
+![](images/Supplementary_Figure_3b.jpg)
+
+<center>Supplementary Figure 2c. Starch at 84 h normalized to volumetric biomass in WT cells. </center>  
+
+There is also some inconsistency with other parts that need clarification. For example WT culture: Photosynthetic efficiency under - Fe +Glc is almost negligible compared to \(+Fe + Glc\) condition in Fig 1 and Fig 4; the same trend is shown for Total chlorophyll in Fig. 4. However, Total chlorophyll under \(+Fe + Glc\) is as low as in - Fe +Glc condition - this is not in agreement with one of the main conclusions of this work.  
+
+Response: We appreciate your careful attention and have clarified that total chlorophyll per cell is larger \(+Fe + Glc\) than - Fe+Glc (Fig. 4e) while total chlorophyll per biomass (Fig. 4f) is similar due to the large increase in cell biomass in \(+Fe + Glc\) , likely due to the buildup of starch and lipids (Fig. 2 and Supplementary Fig. 2c). The differences in photosynthesis between WT+Fe+Glc and WT- Fe+Glc, as evidenced by photosynthetic efficiency and oxygen evolution (Fig. 1 & 4; Supplementary Figure 6), can be explained by the difference in the make- up of chlorophyll. A higher ratio of chl \(a / chl b\) in WT+Fe+Glc than WT- Fe+Glc implicates more core photosystem complexes that evolve oxygen vs. antenna complexes (Supplementary Fig. 6), and the proteomics analysis shows significant upregulation of photosynthetic subunits in WT+Fe+Glc vs. WT- Fe+Glc (Fig. 7a). We also emphasize that oxygen evolution normalized by chlorophyll concentration, a common standardization to compare photosynthetic efficiency (Yang et al. 2020), is \(\sim 3x\) higher in WT+Fe+Glc than in any other photosynthetic condition (Supplementary Fig. 6c) in lines 230- 232.
+
+<--- Page Split --->
+
+We have modified and added text in lines 223- 234 which now references all normalizations in Fig 4 and Supplementary Fig. 6.  
+
+Proteomics data: the Principal Component Analysis (PCA) reveals well- separated treatment groups; however, a notable lack of distinction between WT and mutant groups raises questions. Providing an explanation for this observation is imperative.  
+
+Response: Thank you for highlighting this point. Two key pieces of evidence underly this result: Evidence 1) hxk- independent signaling mechanisms (Aguilera- Alvarado and Sanchez- Nieto, 2017) which include increased cell size in our dataset (Supplementary Fig. 2a)—we have clarified this in lines 202- 203. Evidence 2), the finding that these mutants were HXK1 knockdown rather than knockout mutants may imply that residual HXK1 activates some +Glc responses—we have clarified this in the manuscript lines 203- 206. Additionally, we have added the strain- dependent PCAs in the manuscript Supplementary Fig. 3b and clarified this in the manuscript in lines 163- 164.  
+
+![PLACEHOLDER_7_0]
+
+<center>Supplementary Fig. 3b. Plot of Principal Components (Dim) 4 and 8 which distinguish strains. </center>  
+
+Based on these findings, we focused on the lack of photosynthetic repression or TAG- accumulation in hxk1 strains as controls for gene discovery rather on elucidating the molecular mechanisms of HXK1 in this paper. We have clarified this in manuscript in lines 206- 209.  
+
+On a broader note, the downstream analysis of the proteomics data appears to rely heavily on fitting to linear model equations. It is difficult or almost impossible without further validation to assess if the data is overfitted or generated unrealistic abundances of proteins. It simply can be seen in differences between model- predicted/normalized protein abundances ranges in each subsets of the data. The authors might want to ensure, using statistical methods, that conclusions are not a result of overfitting.  
+
+Response: Thank you for these comments. We have added the data analyses listed below:  
+
+To prevent overfitting, we created additional model simplification steps written in the Methods section ("Model of the effects of Fe, Glc, and strain on protein abundance" starting on line 606). Importantly, we relied on cutoffs based on both relative size of coefficient (i.e. slope of each linear term) and not just solely based on meeting a significant \(p\) - value. This prevents overfitting
+
+<--- Page Split --->
+
+by not allowing small but "technically" significant values to inform the protein response by linear model categorization.  
+
+We also looked at the difference between the predicted abundance and the actual abundance (i.e. residuals) and found that for residuals of all models that we derived in Fig 3e and Supplementary Data 3, \(\sim 89\%\) of the residuals are within a \(0.5 \log_2\) difference while \(99\%\) have a \(\log_2\) difference \(< 1\) . This shows the majority of models accurately predict the abundance. We have incorporated this information in lines 650- 660 and lines 178- 179.  
+
+While the results of the full modeling pipeline show accuracy, we additionally conducted a cross- validation to test that our approach does not overfit. As our full model predicts that most protein abundance effects are strain- independent, we tested if our pipeline models are overfit by running the modelling pipeline on protein abundances from one strain (training strain) and observing how well they predict protein abundance in the same conditions as the other strain (test strain). With WT as the training strain, we reran our pipeline testing for the significance of a "Fe + Glc + Fe:Glc" linear model, removing the strain variables as the model is only run on WT. Model simplification steps then proceeded as mentioned in the Methods section, STEPS1- 3. For proteins that were found to be strain- independent in the original full modelling pipeline, the distribution of residuals is similar for the training strain (WT) vs. the test train (hxk1- 2, histogram left, Supplementary 4c). This follow- up shows that our pipeline does not overfit under relevant strain- independent datasets, as the test data has similar residuals to the training data. However, proteins found to be strain- dependent tend to have more extreme residuals in hxk1- 2 (histogram, right side) while they predict WT data well. This shows that for these proteins the model suffers from the exclusion of strain- dependent effects. The cross- validation shows the replicability of the model simplification when strain effects have insignificant contributions to protein abundance, confirming the data is not overfit, but also show need to incorporate the full model with strain effects to accurately capture strain- dependent responses.  
+
+![PLACEHOLDER_8_0]
+  
+
+Supplementary Figure 4c. Histogram of the distribution of the residual (difference in \(\log_2\) abundance) of a reduced cross- validation model (Methods). Models were determined for the equation "Fe + Glc + Fe:Glc" in WT and simplified as in the full- model. Residuals are compared in both WT actual data and hxk1- 2 strain data, which was not used to develop the model. Separate facets of the graphs indicate whether the full modeling pipeline determined a given's protein abundance was independent (left) or dependent (right) on strain.
+
+<--- Page Split --->
+
+Figure 4a, please correct the abbreviation from "Wt" to "WT". Response: Corrected, thank you for catching this.  
+
+Reviewer #2 (Remarks to the Author):  
+
+Here the authors have performed a rigorous and comprehensive set of experiments designed to tease out the intricate relationship between glucose, regulator gene HXK1, and iron in governing photosynthesis vs respiration in gene Alga. The work revealed that multiple effectors have a stronger impact on organism response than individual effectors operating alone (an important consideration for things like drug design). The study also demonstrates that these protein response mechanisms are likely conserved across many plants and alga. The heterotrophic conditions studied enabled the authors to tease apart protein membership in various important biological functions like photosynthesis and the accumulation of TAG. The authors fit a linear model to their proteomics data from these heterotrophic condition studies to determine which conditions exercised the greatest control over protein abundance. Overall, the results are quite compelling. The study and model are able to tease apart different groups of proteins that behave and respond differentially to the tested conditions. The linear model is extremely simple, yet also very effective at binning the proteins based on their distinctive response patterns. Generally the studies are comprehensive, rigorous, well documented, and well described. The writing is excellent. The work is exciting. The authors have established an approach for studying protein function genome- wide using mixotrophic growth that could be replicated in many other systems, and in the process of doing so, they learned a great deal about the roles all the proteins in this particular species play in photosynthesis, energy storage, and respiration. The role iron resource sharing between competing pathways that both have iron requirements plays in dictating how the cell selects its energy strategy is deeply enlightening. This reinforces a growing view that resource allocation is a major driving design consideration in cell regulation. Essentially, this unveils a fundamental design principle for life.  
+
+Response: We are incredibly grateful for these enthusiastic comments about our work!  
+
+I have only a few suggestions, which I leave to the authors' discretion:  
+
+1. ) The protein expression heatmaps in figure 3e are tough to interpret. That said, I don't know that there is a better way of doing this. The current representation is very compact, concise, and information rich.  
+
+Response: Thank you for acknowledging the complexity of these data and encouraging comments on our representation of these data. Based on previous multifactorial studies (e.g. Swift et al. 2019 Fig. 1d), we decided heatmap plotting was the most concise way to simultaneously capture the number of proteins assigned to each equation- term groups alongside expression differences. While there is a lot of information in one figure, we wanted to have these data available so that a researcher interested in a topic outside the focus of this paper could observe regulatory patterns relevant to their research.  
+
+2. ) I'd like to have a little more information on how much data agreed with expectations vs
+
+<--- Page Split --->
+
+being surprising. In the discussion, the authors raise FNR1 as surprising for not changing its expression profile despite being in subsystems that general did change in the tested conditions. How many other proteins were like this? It would be interesting to see a break down of how many genes stayed constant despite being involved in one of the differentially expressed subsystems, how many genes change expression despite being in static subsystems, and so on. This speaks to a potential boon from this work in the form of improved annotations.  
+
+Response: Thank you, this was a very thoughtful comment, and it inspired us to provide a systematic investigation of surprising protein expression deviations from expected function. We used automated bioinformatic software that predicts protein domains (Pfam) and functions based on homology. Then we explored their absence in process related enrichment lists (e.g. lowest- in- heterotrophy) or unexpected linear models to narrow down a list of surprising responses (Supplementary Data 5).  
+
+As an example, we investigated all 34 detected proteins with a predicted "Chlorophyll A- B binding" domain. From this Pfam annotation, a researcher may assume these proteins would be linked to photosynthetic light harvesting. 20/34 proteins were indeed lowest- in- heterotrophy, consistent with a role in photosynthesis. However, 14/34 had protein regulatory patterns that did not place them in this list. Interestingly observing the linear models of these 14 that have subterms like - Fe (higher in low Fe) or +Glc (higher when Glc is added) show examples where Chlorophyll A- B binding proteins have higher expression in conditions that normally downregulate chlorophyll abundance. LHC13 tends to be highest in - Fe+Glc conditions and ELIP4 (early light inducible protein) tends to be upregulated in - Fe conditions (Supplementary Data 5). While other LHCs and ELIPs are lowest in heterotrophy (14 LHCs and 4 ELIPs), suggesting their role in light harvesting, the remaining LHC and ELIPs might play more of a role in thylakoid stability, stress tolerance, and storage of non- photosynthetic chlorophyll when photosynthesis is downregulated. We have added this analysis in lines 476- 481, made the results available in Supplementary Data 5, and hope this type of research may be used to improve functional annotation in other studies, as the reviewer mentions.  
+
+3. Proteomics data not only provides protein abundance information, but also information on many post-translational modifications. Is there evidence of shifts in PTMs across the studied conditions? Particularly for proteins with abundances that stayed constant? Response: Thank you for bringing up this important aspect. While bioinformatics techniques exist to predict PTM from total protein extracts, we believe it is more appropriate to investigate PTMs from samples that have been enriched for a given PTM because PTMs occur at sparse levels. For this study, we ran mass-spectrometry of whole cell extracts with the specific goal of measuring relative protein abundance. We are currently investigating specific PTMs and hope to include these data in future publications.  
+
+Reviewer #3 (Remarks to the Author):  
+
+In their paper titled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis", Jeffers and colleagues explore the response of photosynthesis in the green alga Chromochloris zofingiensis to the presence of glucose, and its dependence on iron supplementation. This work included the characterization
+
+<--- Page Split --->
+
+of the physiological response of these algae to a combination of iron and glucose availability treatments, and was also complemented with studies on mutants of Hexokinase- 1 encoding gene. The main dataset generated in this work is the proteome response of WT and HXK1 mutants (one of them might still demonstrate some expression) to these combined treatments, confirming the previously reported physiological strategy of Chromochloris zofingiensis, including dnFAS under - Fe+Glic conditions. These proteomes are further studied for novel genes associated with photosynthesis which are downregulated, and for better understanding of iron transport at the sub- cellular level, under these conditions. Finally, Jeffers and colleagues discuss means by which targeting FDX and FAD encoding genes in C. zofingiensis might be beneficial for biofuel- directed synthetic approaches.  
+
+I found this work interesting, though the role of iron in the carbon response has been characterized in other model green algae. In addition, there are some issues which require clarification or I think must be clarified, some of which are major and may limit the interoperation that can be made based on this work, including the choice of (rich, peptone- based) medium, extremely low levels of iron in the medium, and the dual role of HXK1 which cannot be fully dissected.  
+
+Response: We are happy you found our working interesting and thank you for pointing out areas to clarify; our manuscript will be improved by your suggestions.  
+
+Detailed points:  
+
+It is unclear to this reviewer why the authors chose a rich medium for an experiment testing effects of glucose addition on PS activity.  
+
+Response: Thank you for bringing attention to the medium. We chose this medium because it was what this organism was grown in when we ordered this strain from the SAG culture collection and what was recommended for its cultivation. The algae grew well in the medium and in fact increasing the iron levels did not change the growth rate or biomass accumulation in - Glc conditions (Fig. 1c,d, Supplementary Fig. 2b). We used this medium in our original publication on this organism for its genome, transcriptome and carotenoid mutants (Roth et al 2017) and then our former studies of its transcriptomics (Roth et al 2019a) and genetics of hxk1 strains (Roth et al 2019b). Because of the role of nutrient deficiency in algal metabolism, we started experimenting with iron and others nutrients and discovered the role of iron in rescuing +Glc- mediated repression of photosynthesis (this manuscript). As the physiological, morphological, and molecular responses are easily distinguished between +Glc and - Glc in Proteose medium and iron in Proteose medium is sufficient to rescue photosynthesis, we conducted this studies' multifactorial experiment in a Proteose growth medium. We have clarified this information in the manuscript in lines 508- 510. We strongly believe that this is an important study to bring to the scientific community and that these proteomes are a valuable resource.  
+
+The rich medium provides a background 'noise' for the heterotrophic conditions elicited by Glc addition. This may be related to the fact that no positive net PS rates are reported in figure 1d (maybe they appear in the - glc controls but they are not presented). Incidentally, in my eyes net rather than gross PS rates should be the values presented for various reasons (the nonlinearity of respiration in dark and light conditions is one) both in general, and especially in the question at hand.
+
+<--- Page Split --->
+
+Response: Thank you for these comments. We have added net oxygen evolution in the light for WT and mutants (Fig. 4c) and oxygen consumption data in the dark for WT and mutants (Fig. 4d). We also included a representation of raw oxygen measurements in the light and dark for WT and both mutants (Supplemental Fig. 6d). Fig. 4b- c and Supplementary Fig. 6d show minimal if any background 'noise' from the - Glc protease medium.  
+
+Line 110- 112 - mixotrophy is highly unlikely to explain the 4- fold increase in biomass in the +Fe+Gic in view of the negative slopes of net PS as reported in Figure 1d. Moreover, such large increase implies that the cells are Fe- starved, and not just limited, which may limit any interpretation of resource allocation between respiration and photosynthesis, as stress response, protein aggregation and longer reversal time of activity may play a role (see e.g. Devadaus et al 2021).  
+
+Response: Thank you for this comment. We agree that mixotrophy is highly unlikely to cause the 4- fold increase in biomass, which is stated in Lines 113- 115. In extreme iron starvation, as in the study by Devadaus et al. 2021, cells have a dramatic reduction in growth (50% in one generation), and the shape of the cell changes to reflect the stress. In contrast, neither of these changes were observed in C. zofingiensis. In - Fe- Glc and - Fe+Gic, the cells grow at a rate similar or more (respectively) than cells in +Fe- Glc (Fig. 1c,d), and the cells do not change shape dramatically (Fig. 2). We have added the growth curves for biomass (in \(\log_{10}\) normalized form) and added this to the manuscript with Supplementary Figure 2b and in lines 111- 113).  
+
+![PLACEHOLDER_12_0]
+  
+
+Supplementary Figure 2b. Experimental replicates of WT biomass growth curves after iron and glucose treatment.  
+
+Lines 119- 120 - the second part of the sentence may be completely wrong. I agree that - Fe is the primary nutritional limitation in the experiments performed by the authors, but this on its own doesn't mean that '- Fe is uniquely required for Glc- mediated repression of photosynthesis'. To argue that, the authors must first show that they generate the same
+
+<--- Page Split --->
+
+limitation for all other nutrients (e.g. as shown in Figure 1e for Fe) and then supplement the cells with each of them separately.  
+
+Response: Thank you for bringing up this specific point. We have corrected this and made our language more precise in the manuscript in lines 122- 123.  
+
+Lines 125- 127 - to make such claims, the authors need to provide more than 3 images per treatment in Supp_dataset 2. What more, I am not convinced that 'starch was more prevalent inside +Glc plastids' even in the images provided, but this actually support the authors claims regarding the metabolic importance of Fe.  
+
+Response: Thank you for this comment. We have added quantified starch data to the manuscript in lines 130- 132 and Supplementary Fig. 2c. Additionally, we have made TEM datasets available (link: https://osf.io/r8dbe/), and this has been noted in the Figure 2 legend.  
+
+Line 142 - HXK1 mutants cannot fully allow 'distinguish general +Glc response..' as some of the 'general' response is not fully HXK1- independent and is also mediated through its (main) role as a metabolic/glycolytic and not just signaling enzyme. I don't think it's only a terminology issue, but also an important point to qualify the limitations of the analysis (e.g. if a specific signaling associated downstream of HXK1 would have been known and targeted, this could have provided the distinction declared here), which is also supported the PCA where strain identity was only separated in PC4 and PC8.  
+
+Response: Thank you for this comment. We agree that untangling HXK glycolytic and signaling roles is challenging, and we acknowledge there are limitations in these data. Therefore, we use the hxk1 mutants as controls to improve our predictions of players of photosynthesis and lipid accumulation rather than to elucidate the mechanism of HXK1. We have revised/clarified this in the manuscript in lines 206- 209.  
+
+Line 215- 216 - any suggestion for the smaller investment in antennae? Less cost- effective under mixotrophy? More light penetration per biomass with these cells?  
+
+Response: Thank you for these interesting comments. We interpret the cause as due to regulatory competition: In +Fe+Glc, the reaction centers contain iron cofactors and therefore chlorophyll a may be thus induced by +Fe at a stronger level than the periphery complexes. +Glc may have the reverse effect of inhibiting periphery complexes more than reaction center complexes in +Fe conditions. We have revised this in the manuscript in lines 228- 230. The reviewer's suggestion on increase light penetration is interesting as +Glc causes cell size increase and astaxanthin accumulation, two features that may reduce light penetration to the plastid.  
+
+Line 340 - I find the terminology "highest/lowest- in- TAG" potentially inaccurate and at a certain point confusing. Can the authors guarantee that TAG accumulation is the only/main process defining the +Glc treatments?  
+
+Response: Thank you for pointing this out. To make our results clearer, we have changed our terminology in the manuscript to "highest/lowest- during- TAG" to avoid applying lipid accumulation is the only upregulated pathway in these cells.
+
+<--- Page Split --->
+
+REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors have diligently addressed all previous comments. I do not have any major comments.  
+
+This manuscript provides an honest representation of data related to nutrient- mediated trophic transitions of a microalgae. This study contributes to our understanding of different nutrient cycles in the environment.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors responded to my limited comments with unexpected vigor and, as a result, added interesting new content to the paper. This was wonderful to see. The response concerning PTMs was very reasonable, and I look forward to seeing future work that digs into this critical aspect of the biology of this system. Overall, I am satisfied this work is ready for publication.  
+
+Reviewer #3 (Remarks to the Author):  
+
+After going over the revised manuscript, I find that the authors have fully adressed my and other reviewers' comments, and I support its publication.  
+
+The only small additional comment I have is that I would briefly mention next to the new text in lines 508- 510 that iron did not alter growth /biomass accumulation rates in - Glc conditions, pointing to the relevant date displays.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6a4e80f7d8d1230ce2e0fe001d740ad7fa02a6213d5a611cfa2992ed9c0d451/supplementary_0_Peer Review File__d6a4e80f7d8d1230ce2e0fe001d740ad7fa02a6213d5a611cfa2992ed9c0d451_det.mmd

@@ -0,0 +1,332 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 508, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[68, 110, 362, 140]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[70, 155, 866, 241]]<|/det|>
+Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 912, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 147, 402, 163]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[114, 203, 880, 386]]<|/det|>
+In the manuscript entitled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis" by Jeffers et al., the authors conducted an extensive investigation into the intricate interplay between glucose, HXK1, and Fe in regulating the metabolism of the green alga C. zofingiensis. The team untangled the roles of these three factors in shifting the trophic modes of this green alga. Through proteomics analysis, the authors shed new light on the variations in protein abundances across distinct trophic conditions. This study tackles timely research questions and provides compelling evidence to support its conclusions. Certain aspects of the study demand refinement, particularly some methods employed for data analysis. Notably, specific datasets are absent for particular experimental conditions, see details in the following comments:  
+
+<|ref|>text<|/ref|><|det|>[[114, 423, 875, 533]]<|/det|>
+Figure 1: The plots of normalized chlorophyll fluorescence and delta O2 nmol are provided for only two conditions, while the remaining panels encompass all four conditions, which raises questions. To enhance transparency, it's recommended that data for - Fe - Glc and +Fe +Glc conditions be added to panels b and d. This data indicates that replete glucose affected (elevated) growth rate more than supplementation of Fe. Were the glucose consumption rates measured for these experiments? Please provide this information.  
+
+<|ref|>text<|/ref|><|det|>[[114, 572, 877, 791]]<|/det|>
+Throughout the manuscript, volumetric biomass emerges as a primary factor for defining growth and normalizing various characteristics such as O2 evolution. However, an inconsistency arises in the choice of normalizing factors. For instance, "Total chlorophyll" is normalized by the number of cells in one instance and by volumetric biomass in another. While the explanation provided in lines 212- 213 is sound, it would be advantageous to substantiate this approach with either new or previously published data. By looking at the cell images in Fig. 2, it looks like cells in - Fe - Glc may contain a similar amount of starch as cells in +Fe +Glc. This is very critical data for the main outcomes of the manuscript. There is also some inconsistency with other parts that need clarification. For example WT culture: Photosynthetic efficiency under - Fe +Glc is almost negligible compared to +Fe +Glc condition in Fig 1 and Fig 4; the same trend is shown for Total chlorophyll in Fig. 4. However, Total chlorophyll under +Fe +Glc is as low as in - Fe +Glc condition - this is not in agreement with one of the main conclusions of this work.  
+
+<|ref|>text<|/ref|><|det|>[[115, 829, 850, 902]]<|/det|>
+Proteomics data: the Principal Component Analysis (PCA) reveals well- separated treatment groups; however, a notable lack of distinction between WT and mutant groups raises questions. Providing an explanation for this observation is imperative. On a broader note, the downstream analysis of the proteomics data appears to rely heavily on fitting to linear model equations. It is
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 877, 162]]<|/det|>
+difficult or almost impossible without further validation to assess if the data is overfitted or generated unrealistic abundances of proteins. It simply can be seen in differences between model- predicted/normalized protein abundances ranges in each subsets of the data. The authors might want to ensure, using statistical methods, that conclusions are not a result of overfitting.  
+
+<|ref|>text<|/ref|><|det|>[[115, 202, 580, 219]]<|/det|>
+Figure 4a, please correct the abbreviation from "Wt" to "WT".  
+
+<|ref|>text<|/ref|><|det|>[[115, 316, 402, 332]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[114, 371, 877, 553]]<|/det|>
+Here the authors have performed a rigorous and comprehensive set of experiments designed to tease out the intricate relationship between glucose, regulator gene HXK1, and iron in governing photosynthesis vs respiration in gene Alga. The work revealed that multiple effectors have a stronger impact on organism response than individual effectors operating alone (an important consideration for things like drug design). The study also demonstrates that these protein response mechanisms are likely conserved across many plants and alga. The heterotrophic conditions studied enabled the authors to tease apart protein membership in various important biological functions like photosynthesis and the accumulation of TAG. The authors fit a linear model to their proteomics data from these heterotrophic condition studies to determine which conditions exercised the greatest control over protein abundance.  
+
+<|ref|>text<|/ref|><|det|>[[114, 564, 880, 783]]<|/det|>
+Overall, the results are quite compelling. The study and model are able to tease apart different groups of proteins that behave and respond differentially to the tested conditions. The linear model is extremely simple, yet also very effective at binning the proteins based on their distinctive response patterns. Generally the studies are comprehensive, rigorous, well documented, and well described. The writing is excellent. The work is exciting. The authors have established an approach for studying protein function genome- wide using mixotrophic growth that could be replicated in many other systems, and in the process of doing so, they learned a great deal about the roles all the proteins in this particular species play in photosynthesis, energy storage, and respiration. The role iron resource sharing between competing pathways that both have iron requirements plays in dictating how the cell selects its energy strategy is deeply enlightening. This reinforces a growing view that resource allocation is a major driving design consideration in cell regulation. Essentially, this unveils a fundamental design principle for life.  
+
+<|ref|>text<|/ref|><|det|>[[115, 823, 648, 840]]<|/det|>
+I have only a few suggestions, which I leave to the authors' discretion:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 867, 143]]<|/det|>
+1. ) The protein expression heatmaps in figure 3e are tough to interpret. That said, I don't know that there is a better way of doing this. The current representation is very compact, concise, and information rich.  
+
+<|ref|>text<|/ref|><|det|>[[114, 182, 880, 310]]<|/det|>
+2. ) I'd like to have a little more information on how much data agreed with expectations vs being surprising. In the discussion, the authors raise FNR1 as surprising for not changing its expression profile despite being in subsystems that general did change in the tested conditions. How many other proteins were like this? It would be interesting to see a break down of how many genes stayed constant despite being involved in one of the differentially expressed subsystems, how many genes change expression despite being in static subsystems, and so on. This speaks to a potential boon from this work in the form of improved annotations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 349, 883, 403]]<|/det|>
+3. ) Proteomics data not only provides protein abundance information, but also information on many post-translational modifications. Is there evidence of shifts in PTMs across the studied conditions? Particularly for proteins with abundances that stayed constant?  
+
+<|ref|>text<|/ref|><|det|>[[115, 471, 402, 487]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[114, 526, 879, 782]]<|/det|>
+In their paper titled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis", Jeffers and colleagues explore the response of photosynthesis in the green alga Chromochloris zofingiensis to the presence of glucose, and its dependence on iron supplementation. This work included the characterization of the physiological response of these algae to a combination of iron and glucose availability treatments, and was also complemented with studies on mutants of Hexokinase- 1 encoding gene. The main dataset generated in this work is the proteome response of WT and HXK1 mutants (one of them might still demonstrate some expression) to these combined treatments, confirming the previously reported physiological strategy of Chromochloris zofingiensis, including dnFAS under - Fe+Glc conditions. These proteomes are further studied for novel genes associated with photosynthesis which are downregulated, and for better understanding of iron transport at the subcellular level, under these conditions. Finally, Jeffers and colleagues discuss means by which targeting FDX and FAD encoding genes in C. zofingiensis might be beneficial for biofuel- directed synthetic approaches.  
+
+<|ref|>text<|/ref|><|det|>[[114, 792, 874, 884]]<|/det|>
+I found this work interesting, though the role of iron in the carbon response has been characterized in other model green algae. In addition, there are some issues which require clarification or I think must be clarified, some of which are major and may limit the interoperation that can be made based on this work, including the choice of (rich, peptone- based) medium, extremely low levels of iron in the medium, and the dual role of HXK1 which cannot be fully dissected.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 91, 240, 106]]<|/det|>
+Detailed points:  
+
+<|ref|>text<|/ref|><|det|>[[114, 118, 881, 245]]<|/det|>
+It is unclear to this reviewer why the authors chose a rich medium for an experiment testing effects of glucose addition on PS activity. The rich medium provides a background 'noise' for the heterotrophic conditions elicited by Glc addition. This may be related to the fact that no positive net PS rates are reported in figure 1d (maybe they appear in the - glc controls but they are not presented). Incidentally, in my eyes net rather than gross PS rates should be the values presented for various reasons (the nonlinearity of respiration in dark and light conditions is one) both in general, and especially in the question at hand.  
+
+<|ref|>text<|/ref|><|det|>[[114, 255, 877, 346]]<|/det|>
+Line 110- 112 - mixotrophy is highly unlikely to explain the 4- fold increase in biomass in the \(+Fe + Glc\) in view of the negative slopes of net PS as reported in Figure 1d. Moreover, such large increase implies that the cells are Fe- starved, and not just limited, which may limit any interpretation of resource allocation between respiration and photosynthesis, as stress response, protein aggregation and longer reversal time of activity may play a role (see e.g. Devadaus et al 2021).  
+
+<|ref|>text<|/ref|><|det|>[[114, 357, 878, 448]]<|/det|>
+Lines 119- 120 - the second part of the sentence may be completely wrong. I agree that - Fe is the primary nutritional limitation in the experiments performed by the authors, but this on its own doesn't mean that 'Fe is uniquely required for Glc- mediated repression of photosynthesis'. To argue that, the authors must first show that they generate the same limitation for all other nutrients (e.g. as shown in Figure 1e for Fe) and then supplement the cells with each of them separately.  
+
+<|ref|>text<|/ref|><|det|>[[114, 458, 883, 531]]<|/det|>
+Lines 125- 127 - to make such claims, the authors need to provide more than 3 images per treatment in Supp_dataset 2. What more, I am not convinced that 'starch was more prevalent inside +Glc plastids' even in the images provided, but this actually support the authors claims regarding the metabolic importance of Fe.  
+
+<|ref|>text<|/ref|><|det|>[[114, 541, 863, 668]]<|/det|>
+Line 142 - HXK1 mutants cannot fully allow 'distinguish general +Glc response.' as some of the 'general' response is not fully HXK1- independent and is also mediated through its (main) role as a metabolic/glycolytic and not just signaling enzyme. I don't think it's only a terminology issue, but also an important point to qualify the limitations of the analysis (e.g. if a specific signaling associated downstream of HXK1 would have been known and targeted, this could have provided the distinction declared here), which is also supported the PCA where strain identity was only separated in PC4 and PC8.  
+
+<|ref|>text<|/ref|><|det|>[[114, 680, 857, 716]]<|/det|>
+Line 215- 216 - any suggestion for the smaller investment in antennae? Less cost- effective under mixotrophy? More light penetration per biomass with these cells?  
+
+<|ref|>text<|/ref|><|det|>[[114, 727, 844, 781]]<|/det|>
+Line 340 - I find the terminology "highest/lowest- in- TAG" potentially inaccurate and at a certain point confusing. Can the authors guarantee that TAG accumulation is the only/main process defining the +Glc treatments?
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[116, 91, 306, 108]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[116, 127, 418, 144]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[113, 163, 882, 350]]<|/det|>
+In the manuscript entitled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis" by Jeffers et al., the authors conducted an extensive investigation into the intricate interplay between glucose, HXK1, and Fe in regulating the metabolism of the green alga C. zofingiensis. The team untangled the roles of these three factors in shifting the trophic modes of this green alga. Through proteomics analysis, the authors shed new light on the variations in protein abundances across distinct trophic conditions. This study tackles timely research questions and provides compelling evidence to support its conclusions. Certain aspects of the study demand refinement, particularly some methods employed for data analysis. Notably, specific datasets are absent for particular experimental conditions, see details in the following comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 350, 882, 404]]<|/det|>
+Response: We thank Reviewer #1 for positive comments on our research questions, experimental design and conclusions. We appreciate your suggestions and recommendations, and we have added the mentioned datasets.  
+
+<|ref|>text<|/ref|><|det|>[[115, 422, 872, 495]]<|/det|>
+Figure 1: The plots of normalized chlorophyll fluorescence and delta O2 nmol are provided for only two conditions, while the remaining panels encompass all four conditions, which raises questions. To enhance transparency, it's recommended that data for - Fe - Glc and +Fe +Glc conditions be added to panels b and d.  
+
+<|ref|>text<|/ref|><|det|>[[115, 497, 880, 552]]<|/det|>
+Response: Thank you for your suggestion. We updated Figure 1 with all treatments showing raw data for chlorophyll fluorescence. Because of space, we moved delta O2 data to the supplement and show data for WT and mutants (Supplementary Fig. 6d).  
+
+<|ref|>text<|/ref|><|det|>[[115, 570, 772, 606]]<|/det|>
+This data indicates that replete glucose affected (elevated) growth rate more than supplementation of Fe.  
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 882, 643]]<|/det|>
+Response: Thank you for highlighting this point. We agree and clarified this in the manuscript in lines 112- 113.  
+
+<|ref|>text<|/ref|><|det|>[[115, 662, 840, 698]]<|/det|>
+Were the glucose consumption rates measured for these experiments? Please provide this information.  
+
+<|ref|>text<|/ref|><|det|>[[115, 700, 882, 736]]<|/det|>
+Response: Thank you for the suggestion. We have added glucose consumption data to the manuscript in lines 110- 111 and to the methods in 579- 582.  
+
+<|ref|>text<|/ref|><|det|>[[115, 755, 882, 865]]<|/det|>
+Throughout the manuscript, volumetric biomass emerges as a primary factor for defining growth and normalizing various characteristics such as O2 evolution. However, an inconsistency arises in the choice of normalizing factors. For instance, "Total chlorophyll" is normalized by the number of cells in one instance and by volumetric biomass in another. While the explanation provided in lines 212- 213 is sound, it would be advantageous to substantiate this approach with either new or previously published data.  
+
+<|ref|>text<|/ref|><|det|>[[115, 867, 882, 903]]<|/det|>
+Response: Thank you for your detailed points on complexities in these data. Because C. zofingiensis cell volume increases dramatically both from adding glucose and between - Fe+Glc
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 883, 183]]<|/det|>
+and \(+Fe + Glc\) (Supplementary Fig. 2a), the normalization is a challenge. We find it logical to look at O2 evolution normalized by biomass so that one is comparing the same amount of cellular material, and this is what was done in previous publications (Roth et al. 2019 Plant Cell, Roth et al. 2019 Comms Biol). We have updated Fig. 4 to include both total chlorophyll normalizations (Fig. 4e,f) and discussed these data on lines 222- 225.  
+
+<|ref|>text<|/ref|><|det|>[[114, 201, 875, 275]]<|/det|>
+By looking at the cell images in Fig. 2, it looks like cells in - Fe - Glc may contain a similar amount of starch as cells in \(+e + Glc\) . This is very critical data for the main outcomes of the manuscript. Response: Thank you for the suggestion. We added starch concentration measurements to the manuscript in lines 130- 132 and Supplementary Fig. 2c as well as the methods in In 571- 578.  
+
+<|ref|>image<|/ref|><|det|>[[346, 294, 648, 525]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[114, 531, 687, 546]]<|/det|>
+<center>Supplementary Figure 2c. Starch at 84 h normalized to volumetric biomass in WT cells. </center>  
+
+<|ref|>text<|/ref|><|det|>[[114, 565, 880, 656]]<|/det|>
+There is also some inconsistency with other parts that need clarification. For example WT culture: Photosynthetic efficiency under - Fe +Glc is almost negligible compared to \(+Fe + Glc\) condition in Fig 1 and Fig 4; the same trend is shown for Total chlorophyll in Fig. 4. However, Total chlorophyll under \(+Fe + Glc\) is as low as in - Fe +Glc condition - this is not in agreement with one of the main conclusions of this work.  
+
+<|ref|>text<|/ref|><|det|>[[114, 657, 884, 897]]<|/det|>
+Response: We appreciate your careful attention and have clarified that total chlorophyll per cell is larger \(+Fe + Glc\) than - Fe+Glc (Fig. 4e) while total chlorophyll per biomass (Fig. 4f) is similar due to the large increase in cell biomass in \(+Fe + Glc\) , likely due to the buildup of starch and lipids (Fig. 2 and Supplementary Fig. 2c). The differences in photosynthesis between WT+Fe+Glc and WT- Fe+Glc, as evidenced by photosynthetic efficiency and oxygen evolution (Fig. 1 & 4; Supplementary Figure 6), can be explained by the difference in the make- up of chlorophyll. A higher ratio of chl \(a / chl b\) in WT+Fe+Glc than WT- Fe+Glc implicates more core photosystem complexes that evolve oxygen vs. antenna complexes (Supplementary Fig. 6), and the proteomics analysis shows significant upregulation of photosynthetic subunits in WT+Fe+Glc vs. WT- Fe+Glc (Fig. 7a). We also emphasize that oxygen evolution normalized by chlorophyll concentration, a common standardization to compare photosynthetic efficiency (Yang et al. 2020), is \(\sim 3x\) higher in WT+Fe+Glc than in any other photosynthetic condition (Supplementary Fig. 6c) in lines 230- 232.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 883, 127]]<|/det|>
+We have modified and added text in lines 223- 234 which now references all normalizations in Fig 4 and Supplementary Fig. 6.  
+
+<|ref|>text<|/ref|><|det|>[[115, 145, 881, 201]]<|/det|>
+Proteomics data: the Principal Component Analysis (PCA) reveals well- separated treatment groups; however, a notable lack of distinction between WT and mutant groups raises questions. Providing an explanation for this observation is imperative.  
+
+<|ref|>text<|/ref|><|det|>[[114, 202, 883, 348]]<|/det|>
+Response: Thank you for highlighting this point. Two key pieces of evidence underly this result: Evidence 1) hxk- independent signaling mechanisms (Aguilera- Alvarado and Sanchez- Nieto, 2017) which include increased cell size in our dataset (Supplementary Fig. 2a)—we have clarified this in lines 202- 203. Evidence 2), the finding that these mutants were HXK1 knockdown rather than knockout mutants may imply that residual HXK1 activates some +Glc responses—we have clarified this in the manuscript lines 203- 206. Additionally, we have added the strain- dependent PCAs in the manuscript Supplementary Fig. 3b and clarified this in the manuscript in lines 163- 164.  
+
+<|ref|>image<|/ref|><|det|>[[262, 365, 730, 586]]<|/det|>
+<|ref|>image_caption<|/ref|><|det|>[[115, 592, 850, 612]]<|/det|>
+<center>Supplementary Fig. 3b. Plot of Principal Components (Dim) 4 and 8 which distinguish strains. </center>  
+
+<|ref|>text<|/ref|><|det|>[[115, 630, 883, 685]]<|/det|>
+Based on these findings, we focused on the lack of photosynthetic repression or TAG- accumulation in hxk1 strains as controls for gene discovery rather on elucidating the molecular mechanisms of HXK1 in this paper. We have clarified this in manuscript in lines 206- 209.  
+
+<|ref|>text<|/ref|><|det|>[[115, 703, 880, 815]]<|/det|>
+On a broader note, the downstream analysis of the proteomics data appears to rely heavily on fitting to linear model equations. It is difficult or almost impossible without further validation to assess if the data is overfitted or generated unrealistic abundances of proteins. It simply can be seen in differences between model- predicted/normalized protein abundances ranges in each subsets of the data. The authors might want to ensure, using statistical methods, that conclusions are not a result of overfitting.  
+
+<|ref|>text<|/ref|><|det|>[[115, 815, 835, 833]]<|/det|>
+Response: Thank you for these comments. We have added the data analyses listed below:  
+
+<|ref|>text<|/ref|><|det|>[[115, 834, 883, 907]]<|/det|>
+To prevent overfitting, we created additional model simplification steps written in the Methods section ("Model of the effects of Fe, Glc, and strain on protein abundance" starting on line 606). Importantly, we relied on cutoffs based on both relative size of coefficient (i.e. slope of each linear term) and not just solely based on meeting a significant \(p\) - value. This prevents overfitting
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 881, 127]]<|/det|>
+by not allowing small but "technically" significant values to inform the protein response by linear model categorization.  
+
+<|ref|>text<|/ref|><|det|>[[115, 127, 883, 220]]<|/det|>
+We also looked at the difference between the predicted abundance and the actual abundance (i.e. residuals) and found that for residuals of all models that we derived in Fig 3e and Supplementary Data 3, \(\sim 89\%\) of the residuals are within a \(0.5 \log_2\) difference while \(99\%\) have a \(\log_2\) difference \(< 1\) . This shows the majority of models accurately predict the abundance. We have incorporated this information in lines 650- 660 and lines 178- 179.  
+
+<|ref|>text<|/ref|><|det|>[[114, 220, 883, 552]]<|/det|>
+While the results of the full modeling pipeline show accuracy, we additionally conducted a cross- validation to test that our approach does not overfit. As our full model predicts that most protein abundance effects are strain- independent, we tested if our pipeline models are overfit by running the modelling pipeline on protein abundances from one strain (training strain) and observing how well they predict protein abundance in the same conditions as the other strain (test strain). With WT as the training strain, we reran our pipeline testing for the significance of a "Fe + Glc + Fe:Glc" linear model, removing the strain variables as the model is only run on WT. Model simplification steps then proceeded as mentioned in the Methods section, STEPS1- 3. For proteins that were found to be strain- independent in the original full modelling pipeline, the distribution of residuals is similar for the training strain (WT) vs. the test train (hxk1- 2, histogram left, Supplementary 4c). This follow- up shows that our pipeline does not overfit under relevant strain- independent datasets, as the test data has similar residuals to the training data. However, proteins found to be strain- dependent tend to have more extreme residuals in hxk1- 2 (histogram, right side) while they predict WT data well. This shows that for these proteins the model suffers from the exclusion of strain- dependent effects. The cross- validation shows the replicability of the model simplification when strain effects have insignificant contributions to protein abundance, confirming the data is not overfit, but also show need to incorporate the full model with strain effects to accurately capture strain- dependent responses.  
+
+<|ref|>image<|/ref|><|det|>[[323, 570, 675, 733]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[114, 740, 883, 851]]<|/det|>
+Supplementary Figure 4c. Histogram of the distribution of the residual (difference in \(\log_2\) abundance) of a reduced cross- validation model (Methods). Models were determined for the equation "Fe + Glc + Fe:Glc" in WT and simplified as in the full- model. Residuals are compared in both WT actual data and hxk1- 2 strain data, which was not used to develop the model. Separate facets of the graphs indicate whether the full modeling pipeline determined a given's protein abundance was independent (left) or dependent (right) on strain.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 610, 127]]<|/det|>
+Figure 4a, please correct the abbreviation from "Wt" to "WT". Response: Corrected, thank you for catching this.  
+
+<|ref|>text<|/ref|><|det|>[[116, 145, 419, 163]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[112, 180, 880, 608]]<|/det|>
+Here the authors have performed a rigorous and comprehensive set of experiments designed to tease out the intricate relationship between glucose, regulator gene HXK1, and iron in governing photosynthesis vs respiration in gene Alga. The work revealed that multiple effectors have a stronger impact on organism response than individual effectors operating alone (an important consideration for things like drug design). The study also demonstrates that these protein response mechanisms are likely conserved across many plants and alga. The heterotrophic conditions studied enabled the authors to tease apart protein membership in various important biological functions like photosynthesis and the accumulation of TAG. The authors fit a linear model to their proteomics data from these heterotrophic condition studies to determine which conditions exercised the greatest control over protein abundance. Overall, the results are quite compelling. The study and model are able to tease apart different groups of proteins that behave and respond differentially to the tested conditions. The linear model is extremely simple, yet also very effective at binning the proteins based on their distinctive response patterns. Generally the studies are comprehensive, rigorous, well documented, and well described. The writing is excellent. The work is exciting. The authors have established an approach for studying protein function genome- wide using mixotrophic growth that could be replicated in many other systems, and in the process of doing so, they learned a great deal about the roles all the proteins in this particular species play in photosynthesis, energy storage, and respiration. The role iron resource sharing between competing pathways that both have iron requirements plays in dictating how the cell selects its energy strategy is deeply enlightening. This reinforces a growing view that resource allocation is a major driving design consideration in cell regulation. Essentially, this unveils a fundamental design principle for life.  
+
+<|ref|>text<|/ref|><|det|>[[115, 607, 810, 626]]<|/det|>
+Response: We are incredibly grateful for these enthusiastic comments about our work!  
+
+<|ref|>text<|/ref|><|det|>[[116, 644, 673, 663]]<|/det|>
+I have only a few suggestions, which I leave to the authors' discretion:  
+
+<|ref|>text<|/ref|><|det|>[[115, 681, 877, 736]]<|/det|>
+1. ) The protein expression heatmaps in figure 3e are tough to interpret. That said, I don't know that there is a better way of doing this. The current representation is very compact, concise, and information rich.  
+
+<|ref|>text<|/ref|><|det|>[[115, 737, 884, 866]]<|/det|>
+Response: Thank you for acknowledging the complexity of these data and encouraging comments on our representation of these data. Based on previous multifactorial studies (e.g. Swift et al. 2019 Fig. 1d), we decided heatmap plotting was the most concise way to simultaneously capture the number of proteins assigned to each equation- term groups alongside expression differences. While there is a lot of information in one figure, we wanted to have these data available so that a researcher interested in a topic outside the focus of this paper could observe regulatory patterns relevant to their research.  
+
+<|ref|>text<|/ref|><|det|>[[112, 884, 839, 903]]<|/det|>
+2. ) I'd like to have a little more information on how much data agreed with expectations vs
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 880, 201]]<|/det|>
+being surprising. In the discussion, the authors raise FNR1 as surprising for not changing its expression profile despite being in subsystems that general did change in the tested conditions. How many other proteins were like this? It would be interesting to see a break down of how many genes stayed constant despite being involved in one of the differentially expressed subsystems, how many genes change expression despite being in static subsystems, and so on. This speaks to a potential boon from this work in the form of improved annotations.  
+
+<|ref|>text<|/ref|><|det|>[[114, 202, 883, 311]]<|/det|>
+Response: Thank you, this was a very thoughtful comment, and it inspired us to provide a systematic investigation of surprising protein expression deviations from expected function. We used automated bioinformatic software that predicts protein domains (Pfam) and functions based on homology. Then we explored their absence in process related enrichment lists (e.g. lowest- in- heterotrophy) or unexpected linear models to narrow down a list of surprising responses (Supplementary Data 5).  
+
+<|ref|>text<|/ref|><|det|>[[114, 312, 883, 590]]<|/det|>
+As an example, we investigated all 34 detected proteins with a predicted "Chlorophyll A- B binding" domain. From this Pfam annotation, a researcher may assume these proteins would be linked to photosynthetic light harvesting. 20/34 proteins were indeed lowest- in- heterotrophy, consistent with a role in photosynthesis. However, 14/34 had protein regulatory patterns that did not place them in this list. Interestingly observing the linear models of these 14 that have subterms like - Fe (higher in low Fe) or +Glc (higher when Glc is added) show examples where Chlorophyll A- B binding proteins have higher expression in conditions that normally downregulate chlorophyll abundance. LHC13 tends to be highest in - Fe+Glc conditions and ELIP4 (early light inducible protein) tends to be upregulated in - Fe conditions (Supplementary Data 5). While other LHCs and ELIPs are lowest in heterotrophy (14 LHCs and 4 ELIPs), suggesting their role in light harvesting, the remaining LHC and ELIPs might play more of a role in thylakoid stability, stress tolerance, and storage of non- photosynthetic chlorophyll when photosynthesis is downregulated. We have added this analysis in lines 476- 481, made the results available in Supplementary Data 5, and hope this type of research may be used to improve functional annotation in other studies, as the reviewer mentions.  
+
+<|ref|>text<|/ref|><|det|>[[114, 608, 883, 774]]<|/det|>
+3. Proteomics data not only provides protein abundance information, but also information on many post-translational modifications. Is there evidence of shifts in PTMs across the studied conditions? Particularly for proteins with abundances that stayed constant? Response: Thank you for bringing up this important aspect. While bioinformatics techniques exist to predict PTM from total protein extracts, we believe it is more appropriate to investigate PTMs from samples that have been enriched for a given PTM because PTMs occur at sparse levels. For this study, we ran mass-spectrometry of whole cell extracts with the specific goal of measuring relative protein abundance. We are currently investigating specific PTMs and hope to include these data in future publications.  
+
+<|ref|>text<|/ref|><|det|>[[116, 792, 418, 810]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 829, 870, 904]]<|/det|>
+In their paper titled "Iron rescues glucose- mediated photosynthesis repression during lipid accumulation in the green alga Chromochloris zofingiensis", Jeffers and colleagues explore the response of photosynthesis in the green alga Chromochloris zofingiensis to the presence of glucose, and its dependence on iron supplementation. This work included the characterization
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 881, 274]]<|/det|>
+of the physiological response of these algae to a combination of iron and glucose availability treatments, and was also complemented with studies on mutants of Hexokinase- 1 encoding gene. The main dataset generated in this work is the proteome response of WT and HXK1 mutants (one of them might still demonstrate some expression) to these combined treatments, confirming the previously reported physiological strategy of Chromochloris zofingiensis, including dnFAS under - Fe+Glic conditions. These proteomes are further studied for novel genes associated with photosynthesis which are downregulated, and for better understanding of iron transport at the sub- cellular level, under these conditions. Finally, Jeffers and colleagues discuss means by which targeting FDX and FAD encoding genes in C. zofingiensis might be beneficial for biofuel- directed synthetic approaches.  
+
+<|ref|>text<|/ref|><|det|>[[114, 276, 860, 385]]<|/det|>
+I found this work interesting, though the role of iron in the carbon response has been characterized in other model green algae. In addition, there are some issues which require clarification or I think must be clarified, some of which are major and may limit the interoperation that can be made based on this work, including the choice of (rich, peptone- based) medium, extremely low levels of iron in the medium, and the dual role of HXK1 which cannot be fully dissected.  
+
+<|ref|>text<|/ref|><|det|>[[114, 387, 881, 423]]<|/det|>
+Response: We are happy you found our working interesting and thank you for pointing out areas to clarify; our manuscript will be improved by your suggestions.  
+
+<|ref|>text<|/ref|><|det|>[[115, 443, 243, 459]]<|/det|>
+Detailed points:  
+
+<|ref|>text<|/ref|><|det|>[[115, 461, 852, 497]]<|/det|>
+It is unclear to this reviewer why the authors chose a rich medium for an experiment testing effects of glucose addition on PS activity.  
+
+<|ref|>text<|/ref|><|det|>[[114, 498, 883, 774]]<|/det|>
+Response: Thank you for bringing attention to the medium. We chose this medium because it was what this organism was grown in when we ordered this strain from the SAG culture collection and what was recommended for its cultivation. The algae grew well in the medium and in fact increasing the iron levels did not change the growth rate or biomass accumulation in - Glc conditions (Fig. 1c,d, Supplementary Fig. 2b). We used this medium in our original publication on this organism for its genome, transcriptome and carotenoid mutants (Roth et al 2017) and then our former studies of its transcriptomics (Roth et al 2019a) and genetics of hxk1 strains (Roth et al 2019b). Because of the role of nutrient deficiency in algal metabolism, we started experimenting with iron and others nutrients and discovered the role of iron in rescuing +Glc- mediated repression of photosynthesis (this manuscript). As the physiological, morphological, and molecular responses are easily distinguished between +Glc and - Glc in Proteose medium and iron in Proteose medium is sufficient to rescue photosynthesis, we conducted this studies' multifactorial experiment in a Proteose growth medium. We have clarified this information in the manuscript in lines 508- 510. We strongly believe that this is an important study to bring to the scientific community and that these proteomes are a valuable resource.  
+
+<|ref|>text<|/ref|><|det|>[[114, 792, 882, 903]]<|/det|>
+The rich medium provides a background 'noise' for the heterotrophic conditions elicited by Glc addition. This may be related to the fact that no positive net PS rates are reported in figure 1d (maybe they appear in the - glc controls but they are not presented). Incidentally, in my eyes net rather than gross PS rates should be the values presented for various reasons (the nonlinearity of respiration in dark and light conditions is one) both in general, and especially in the question at hand.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[114, 90, 883, 183]]<|/det|>
+Response: Thank you for these comments. We have added net oxygen evolution in the light for WT and mutants (Fig. 4c) and oxygen consumption data in the dark for WT and mutants (Fig. 4d). We also included a representation of raw oxygen measurements in the light and dark for WT and both mutants (Supplemental Fig. 6d). Fig. 4b- c and Supplementary Fig. 6d show minimal if any background 'noise' from the - Glc protease medium.  
+
+<|ref|>text<|/ref|><|det|>[[114, 201, 874, 312]]<|/det|>
+Line 110- 112 - mixotrophy is highly unlikely to explain the 4- fold increase in biomass in the +Fe+Gic in view of the negative slopes of net PS as reported in Figure 1d. Moreover, such large increase implies that the cells are Fe- starved, and not just limited, which may limit any interpretation of resource allocation between respiration and photosynthesis, as stress response, protein aggregation and longer reversal time of activity may play a role (see e.g. Devadaus et al 2021).  
+
+<|ref|>text<|/ref|><|det|>[[114, 312, 884, 461]]<|/det|>
+Response: Thank you for this comment. We agree that mixotrophy is highly unlikely to cause the 4- fold increase in biomass, which is stated in Lines 113- 115. In extreme iron starvation, as in the study by Devadaus et al. 2021, cells have a dramatic reduction in growth (50% in one generation), and the shape of the cell changes to reflect the stress. In contrast, neither of these changes were observed in C. zofingiensis. In - Fe- Glc and - Fe+Gic, the cells grow at a rate similar or more (respectively) than cells in +Fe- Glc (Fig. 1c,d), and the cells do not change shape dramatically (Fig. 2). We have added the growth curves for biomass (in \(\log_{10}\) normalized form) and added this to the manuscript with Supplementary Figure 2b and in lines 111- 113).  
+
+<|ref|>image<|/ref|><|det|>[[285, 464, 707, 723]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[114, 746, 870, 783]]<|/det|>
+Supplementary Figure 2b. Experimental replicates of WT biomass growth curves after iron and glucose treatment.  
+
+<|ref|>text<|/ref|><|det|>[[114, 802, 866, 876]]<|/det|>
+Lines 119- 120 - the second part of the sentence may be completely wrong. I agree that - Fe is the primary nutritional limitation in the experiments performed by the authors, but this on its own doesn't mean that '- Fe is uniquely required for Glc- mediated repression of photosynthesis'. To argue that, the authors must first show that they generate the same
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 857, 127]]<|/det|>
+limitation for all other nutrients (e.g. as shown in Figure 1e for Fe) and then supplement the cells with each of them separately.  
+
+<|ref|>text<|/ref|><|det|>[[115, 127, 864, 164]]<|/det|>
+Response: Thank you for bringing up this specific point. We have corrected this and made our language more precise in the manuscript in lines 122- 123.  
+
+<|ref|>text<|/ref|><|det|>[[115, 182, 870, 256]]<|/det|>
+Lines 125- 127 - to make such claims, the authors need to provide more than 3 images per treatment in Supp_dataset 2. What more, I am not convinced that 'starch was more prevalent inside +Glc plastids' even in the images provided, but this actually support the authors claims regarding the metabolic importance of Fe.  
+
+<|ref|>text<|/ref|><|det|>[[115, 256, 882, 312]]<|/det|>
+Response: Thank you for this comment. We have added quantified starch data to the manuscript in lines 130- 132 and Supplementary Fig. 2c. Additionally, we have made TEM datasets available (link: https://osf.io/r8dbe/), and this has been noted in the Figure 2 legend.  
+
+<|ref|>text<|/ref|><|det|>[[115, 330, 880, 459]]<|/det|>
+Line 142 - HXK1 mutants cannot fully allow 'distinguish general +Glc response..' as some of the 'general' response is not fully HXK1- independent and is also mediated through its (main) role as a metabolic/glycolytic and not just signaling enzyme. I don't think it's only a terminology issue, but also an important point to qualify the limitations of the analysis (e.g. if a specific signaling associated downstream of HXK1 would have been known and targeted, this could have provided the distinction declared here), which is also supported the PCA where strain identity was only separated in PC4 and PC8.  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 883, 552]]<|/det|>
+Response: Thank you for this comment. We agree that untangling HXK glycolytic and signaling roles is challenging, and we acknowledge there are limitations in these data. Therefore, we use the hxk1 mutants as controls to improve our predictions of players of photosynthesis and lipid accumulation rather than to elucidate the mechanism of HXK1. We have revised/clarified this in the manuscript in lines 206- 209.  
+
+<|ref|>text<|/ref|><|det|>[[115, 570, 882, 607]]<|/det|>
+Line 215- 216 - any suggestion for the smaller investment in antennae? Less cost- effective under mixotrophy? More light penetration per biomass with these cells?  
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 883, 737]]<|/det|>
+Response: Thank you for these interesting comments. We interpret the cause as due to regulatory competition: In +Fe+Glc, the reaction centers contain iron cofactors and therefore chlorophyll a may be thus induced by +Fe at a stronger level than the periphery complexes. +Glc may have the reverse effect of inhibiting periphery complexes more than reaction center complexes in +Fe conditions. We have revised this in the manuscript in lines 228- 230. The reviewer's suggestion on increase light penetration is interesting as +Glc causes cell size increase and astaxanthin accumulation, two features that may reduce light penetration to the plastid.  
+
+<|ref|>text<|/ref|><|det|>[[115, 755, 880, 810]]<|/det|>
+Line 340 - I find the terminology "highest/lowest- in- TAG" potentially inaccurate and at a certain point confusing. Can the authors guarantee that TAG accumulation is the only/main process defining the +Glc treatments?  
+
+<|ref|>text<|/ref|><|det|>[[115, 811, 882, 866]]<|/det|>
+Response: Thank you for pointing this out. To make our results clearer, we have changed our terminology in the manuscript to "highest/lowest- during- TAG" to avoid applying lipid accumulation is the only upregulated pathway in these cells.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 313, 106]]<|/det|>
+REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 147, 403, 163]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 203, 866, 220]]<|/det|>
+The authors have diligently addressed all previous comments. I do not have any major comments.  
+
+<|ref|>text<|/ref|><|det|>[[115, 260, 872, 313]]<|/det|>
+This manuscript provides an honest representation of data related to nutrient- mediated trophic transitions of a microalgae. This study contributes to our understanding of different nutrient cycles in the environment.  
+
+<|ref|>text<|/ref|><|det|>[[115, 411, 403, 427]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 467, 857, 539]]<|/det|>
+The authors responded to my limited comments with unexpected vigor and, as a result, added interesting new content to the paper. This was wonderful to see. The response concerning PTMs was very reasonable, and I look forward to seeing future work that digs into this critical aspect of the biology of this system. Overall, I am satisfied this work is ready for publication.  
+
+<|ref|>text<|/ref|><|det|>[[115, 608, 403, 623]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 664, 850, 699]]<|/det|>
+After going over the revised manuscript, I find that the authors have fully adressed my and other reviewers' comments, and I support its publication.  
+
+<|ref|>text<|/ref|><|det|>[[115, 710, 879, 763]]<|/det|>
+The only small additional comment I have is that I would briefly mention next to the new text in lines 508- 510 that iron did not alter growth /biomass accumulation rates in - Glc conditions, pointing to the relevant date displays.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6a69bed05a64141bd0130e6ebb30dedcb1c1aa61c6ada15b1de153ae15ddc10/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d6a69bed05a64141bd0130e6ebb30dedcb1c1aa61c6ada15b1de153ae15ddc10/supplementary_0_Peer Review File__d6a69bed05a64141bd0130e6ebb30dedcb1c1aa61c6ada15b1de153ae15ddc10.mmd

@@ -0,0 +1,151 @@
+
+# nature portfolio  
+
+# Peer Review File  
+
+Theoretical and experimental analysis of circularly polarized luminescence spectrophotometers for artifact- free measurements using a single CCD camera  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+The manuscript reports a CPL instrument designed on a novel concept. It uses a single CCD camera and two independent channels. The authors show experimentally and theoretically that errors and artifacts can be minimized, by different ways of separating the two polarizations. In general, this work is very well carried out, and there is a huge interest nowadays in CPL and how to measure it. I believe that all efforts leading to cheaper (with respect to commercial CPL spectrofluoropolarimeters) but reliable alternatives to measure CPL are welcome and have the potentials to expand the field even more. For these reasons, I would like to recommend the manuscript for publication after a few points are taken into account.  
+
+1) The authors are aware that a general problem of CPL, no matter how it is measured, is due to linear anisotropies of emission. Indeed, in the present set-up, such artifact comes from the coupling between S1 and QWP non-ideal retardance. The authors may elaborate on the fact that in solution, these contributions become significant where emission lifetime and molecular tumbling rate occur on a similar time scale, and therefore a depolarization due to orientational averaging can not occur completely (according to Perrin equation). When using a \(90^{\circ}\) excitation geometry, an effective way to cancel this contribution is by linearly polarizing the excitation beam on the plane of the detection. In this way only molecules that can not contribute to linear anisotropy are excited, but the cost is a relevant decrease of the signal. I suggest the authors perform the experiment by using an achiral fluorescent system giving a significant linear contribution and try to eliminate the artifacts as described above. This is important to assess, because such situation is very common when working with solid samples and rather common with organic fluorophores in solution.  
+
+2) When the polarizations are time-separated, the stability of the light source becomes critical, as the authors describe. I wonder if the instabilities observed by the authors are due to a constant drift (up or downwards) or just to random fluctuation. If the excitation lamp signal were continuously acquired as well, would it be possible to normalize constantly the signal by the excitation intensity?  
+
+3) It is not very clear to me the origin of the artifact due to wavelength mismatch and why it is expected to be proportional to the derivative of the total luminescence. In particular, in the experiments shows in Fig. 7, is this artifact expected to be symmetric with respect to the tilt angle?  
+
+4) I appreciate that with the instruments proposed by the authors g_lum as low as \(10^{\wedge} - 4\) can be measured. This is a lower bound, anyway, because the actual limit depends on the efficiency of the emitter, as well. Even higher g factors become difficult to measure if the output photons are simply not enough. Can the authors quantify, the limit of their instrumentation in terms of circularly polarization brightness (B_CPL)? I think that most instruments can easily measure compounds with a B_CPL around 1 M-1. cm-1.  
+
+5) A rough estimate of the cost of the proposed apparatus could be given to put it into context with commercial instrumentation.  
+
+6) When mixing the two enantiomers of the Eu complex to generate Fig. 8, the authors write that "Figure 8 clearly shows this gradual racemization effect". Actually, there is no racemization effect, as this would imply some chemical process, the authors are simply preparing scalemic solutions. Please rephrase this sentence.  
+
+7) Concerning NIR-CPL, the authors may be interested in two recent papers in which measurements of Er CPL around 1500 nm with simple optics are reported (10.1021/jacs.2c01134, 10.1002/anie.202208326)
+
+<--- Page Split --->
+
+8) The language of the manuscript should be revised. For example (the list is non-comprehensive): "analyxis", "cosinus", "and CPL measurements [18, 19] and ourselves [20, 21]", "by saving one spectrophotometer, as regard CPL setups based on two of theme", etc.  
+
+## Reviewer #2 (Remarks to the Author):  
+
+The manuscript describes the construction of a CPL spectrometer with an unconventional design. The instrument allows rapid acquisition of CPL spectra. A paper describing an alternative design of instrument for rapid CPL detection has already been published in Nat COMM https://www.nature.com/articles/s41467- 020- 15469- 5.  
+
+I'm not sure the the design outlined in the current manuscript is demonstrably superior to that reported earlier.  
+
+This is a personal opinion, but I'm not sure that instrument development manuscripts should be published in a general science journal unless the new instrument is used to provide insight into a novel scientific problem, of reasonably broad interest. I do not believe that this manuscript meets this criteria. Consequently, I would not recommend publication.
+
+<--- Page Split --->
+
+## Response to Reviewers  
+
+Response to ReviewersWe would like to the reviewers for having spent their time and efforts to provide a feedback on our manuscript entitled "Theoretical and experimental analysis of circularly polarized luminescence spectrophotometers for artifact- free measurements using a single CCD camera". We believe that the reviewers comments helped us to improve the quality of the manuscript. From the reviewers comments we noticed that some aspects of the manuscript are not well understood and we did our best to make them clearer. We have taken into account all the remarks and issues of the referees. As asked by the reviewer 1, we have added some new experimental results to estimate the contribution of the fluorescence linear anisotropy on the CPL artifacts.  
+
+Please find in the following the reviewers comments in blue and our point- by- point responses.  
+
+## Reviewer 1:  
+
+The manuscript reports a CPL instrument designed on a novel concept. It uses a single CCD camera and two independent channels. The authors show experimentally and theoretically that errors and artifacts can be minimized, by different ways of separating the two polarizations. In general, this work is very well carried out, and there is a huge interest nowadays in CPL and how to measure it. I believe that all efforts leading to cheaper (with respect to commercial CPL spectrofluoropolarimeters) but reliable alternatives to measure CPL are welcome and have the potentials to expand the field even more. For these reasons, I would like to recommend the manuscript for publication after a few points are taken into account.  
+
++ We appreciate the appreciation of Reviewer 1 about our work.  
+
+1) The authors are aware that a general problem of CPL, no matter how it is measured, is due to linear anisotropies of emission. Indeed, in the present set-up, such artifact comes from the coupling between S1 and QWP non-ideal retardance. The authors may elaborate on the fact that in solution, these contributions become significant where emission lifetime and molecular tumbling rate occur on a similar time scale, and therefore a depolarization due to orientational averaging can not occur completely (according to Perrin equation). When using a \(90^{\circ}\) excitation geometry, an effective way to cancel this contribution is by linearly polarizing the excitation beam on the plane of the detection. In this way only molecules that can not contribute to linear anisotropy are excited, but the cost is a relevant decrease of the signal. I suggest the authors perform the experiment by using an achiral fluorescent system giving a significant linear contribution and try to eliminate the artifacts as described above. This is important to assess, because such situation is very common when working with solid samples and rather common with organic fluorophores in solution.  
+
++This comment highlight a very important point. In the first version of the article, we worked only with isotropic solutions without polarization photoselection. We agree that linear anisotropies may lead to undesired measurement artifacts when the isotropy of the medium can not be ensured. We therefore determined the possible contribution of the linear anisotropies to the CPL signal from both a theoretical and an experimental point of view:  
+
+1. We have added an experimental study on fluorescein molecules as a function of the viscosity of the solution - Section 5. It shows that indeed a CPL signals can be measured for highly viscous solutions even if fluorescein is achiral! This founding (anticipated by Reviewer 1) makes us go deeper into the analysis of our set-up.  
+
+2. From a theoretical point of view, we took into account circular and \(+45^{\circ}\) dichroism in the modeling of the PBS (lines 145-150). We have included these parameters the SI, equation 2 and table 2 which gathers all the results. Finally, these additional parameters complicate the maths but explain very well the experimental results.  
+
+3. The contributions of the LD' and the CD of the PBS to the measured CPL are now quantified (table 2). We choose not to add the consideration of the linear anisotropies for the description of the space separation set-up while in any event it is not the best to perform the CPL measurement. For the time and the time+space set-up, only the residual CD in the PBS induces a \(CDS_{1}^{e}\) signal that can not be differentiated from the \(S_{3}^{e}\) one's. It is added in Eq.4 and 6 and discuss accordingly.  
+
+4. In parallel, we have recorded the \(+45\) and CD of the polarizing beam splitter with our CD/LD spectrometer (SI Section 3). It reveals an unexpected few percents of signals for the circular and \(+45^{\circ}\) linear dichroism. This value of PBS circular dichroism is in very good agreement with the theory and experimental result on the fluorescein.  
+
+Thanks again to Reviewer 1, for pointing out what was missing for a complete analysis.
+
+<--- Page Split --->
+
+2) When the polarizations are time-separated, the stability of the light source becomes critical, as the authors describe. I wonder if the instabilities observed by the authors are due to a constant drift (up or downwards) or just to random fluctuation. If the excitation lamp signal were continuously acquired as well, would it be possible to normalize constantly the signal by the excitation intensity?  
+
+Within our integration time the random fluctuation is very small and the signal difference comes mainly from the drift. In Sec. 3.2 (line 303), we estimate this drift around \(1\%\) . Theoretically we agree Reviewer 1 : if we are able to continuously measure the pump power, we can normalize the CPL signal. However, from an experimental point of view, we need a measurement set- up with an accuracy \(\sim \frac{1\%}{10} = 0.1\%\) . Standard photodiode sensors (thorlabs- Photodiode Power Sensors) which combine enough speed and accuracy are proposed with a linearity of \(\pm 0.5\%\) and an accuracy of \(\pm 5\%\) which make them useless. Trying to normalize the signal with the recorded pump power will therefore report the issue from the stability of the pump to the stability of the measurement set- up and require a rather sophisticated detection apparatus with thermally controlled electronic. We like better rely on the two measurements set- up which reveals to be very efficient.  
+
+To explain this point in the paper we have added line 303: These intensity variations come from the slowly varying drift of the pump source. To monitor this drift in order to scale the fluorescence intensity would require detection set- up with an accuracy of \(\approx 0.1\%\) . This can not be performed with standard photodiode power sensors which usually present a few percent accuracy.  
+
+3) It is not very clear to me the origin of the artifact due to wavelength mismatch and why it is expected to be proportional to the derivative of the total luminescence. In particular, in the experiments shows in Fig. 7, is this artifact expected to be symmetric with respect to the tilt angle?  
+
+We now explain this in the paper line 201 after "Less intuitive, the wavelength mismatch of the two recorded spectra induces a CPL signal proportional to the luminescence derivative \(\left(\frac{\partial S_0}{\partial \lambda}\right)\) : " Indeed the difference between two identical spectra recorded with a wavelength mismatch \(\Delta \lambda\) writes: \(\Delta S = S(\lambda + \Delta \lambda) - S(\lambda)\) . Because \(\Delta \lambda \ll \lambda\) , we may use the Taylor- series expansion: \(S(\lambda + \Delta \lambda) \approx S(\lambda) + \Delta \lambda \frac{dS}{d\lambda} (\lambda)\) . Therefore, the recorded CPL signal, being a difference between two unmatched signals, contains unwanted terms proportional to the signal derivative."  
+
+In the experiment shown in Fig. 7, we tilted the camera in order to experimentally add a wavelength mismatch between the top and bottom spectra recorded on the camera. The goal is to qualitatively check the theory not to do quantification. Indeed, the exact wavelength mismatch is wavelength dependent and is quite complicated to model because it depends on the position of the rotation center, the exact dispersion formula and the optical aberrations at the high aperture angles.  
+
+However as Reviewer 1 mentioned, by inverting the tilt angle, the mismatch is inverted and the artifact also (false \(\mathrm{CPL}\propto \Delta \lambda \frac{\partial S_0}{\partial \lambda}\) ). We now describe this in the new version (line 328): "These artifacts are especially intense in the \(620 \mathrm{nm}\) region where the fluorescence derivative is the highest. At a maximum camera tilt angle of \(- 1.2^{\circ}\) , the artifacts are nearly proportional to the derivative (black dashed curve). By inverting the angle to \(+1.2^{\circ}\) , the artifacts change in sign but not in shape as expected by the theory (Tab. 2 line 3, false \(\mathrm{CPL}\propto \Delta \lambda \frac{\partial S_0}{\partial \lambda}\) )."  
+
+I appreciate that with the instruments proposed by the authors \(g_{lum}\) as low as \(10^{- 4}\) can be measured. This is a lower bound, anyway, because the actual limit depends on the efficiency of the emitter, as well. Even higher g factors become difficult to measure if the output photons are simply not enough. Can the authors quantify, the limit of their instrumentation in terms of circularly polarization brightness \(\mathrm{(B_{CPL})}\) ? I think that most instruments can easily measure compounds with a \(\mathrm{B_{CPL}}\) around \(1 \mathrm{M}^{- 1} \mathrm{cm}^{- 1}\) .  
+
+We agree with the remark of Reviewer 1, it is better to consider the \(\mathrm{B_{CPL}}\) in order to compare the results. We discuss now of the capability of our set- up considering the \(\mathrm{B_{CPL}}\) (emitter parameter) and the integration time required to get "good" spectra (signal to noise higher than 10). It was written in the last two paragraphs of the introduction, in the signal to noise ratio discussion (section 3.5) and in the conclusion.  
+
+5) A rough estimate of the cost of the proposed apparatus could be given to put it into context with commercial instrumentation. The cost of this kind of apparatus is about \(40\mathrm{k}\mathrm{\textregistered}\) mainly driven by the CCD camera (about \(30\mathrm{k}\mathrm{\textregistered}\) ). As requested, we have included this information in the conclusion.  
+
+6) When mixing the two enantiomers of the Eu complex to generate Fig. 8, the authors write that Figure 8 clearly shows this gradual racemization effect. Actually, there is no racemization effect, as this would imply some chemical process, the authors are simply preparing scalemic solutions. Please rephrase this sentence.  
+
+We, agree totally, as requested, the sentence as been rephrased.  
+
+7) Concerning NIR-CPL, the authors may be interested in two recent papers in which measurements of Er CPL around \(1500 \mathrm{nm}\) with simple optics are reported (10.1021/jacs.2c01134, 10.1002/anie.202208326)  
+
+We thank Reviewer 1 for the information. These references have been added in the introduction.  
+
+8) The language of the manuscript should be revised. For example (the list is non-comprehensive): analyzis, cosinus, and CPL measurements [18, 19] and ourselves [20, 21], by saving one spectrophotometer, as regard CPL setups based on two of theme, etc.
+
+<--- Page Split --->
+
+As requested, the text was thoroughly read and we made our best to rephrase or correct the language.
+
+<--- Page Split --->
+
+## Reviewer 2  
+
+The manuscript describes the construction of a CPL spectrometer with an unconventional design. The instrument allows rapid acquisition of CPL spectra. A paper describing an alternative design of instrument for rapid CPL detection has already been published in Nat Comm https://www.nature.com/articles/s41467- 020- 15469- 5.  
+
+I'm not sure the the design outlined in the current manuscript is demonstrably superior to that reported earlier.  
+
+We are aware of this paper and it is cited in the article. Only bright spectra of CPL Eu3+ complexes have been recorded using this setup. However, we are sorry to say that this paper presents fundamental misunderstandings. First, the calibration procedure is based on a division between two signals. While it sounds mathematically possible, it is very difficult to perform properly when the denominator signal tend to zero. They rightly use for these two signals the Eu3+ fluorescence to be sure that the beams follow the same paths. Therefore, the denominator of the calibration function is an Eu3+ spectra which has low and zero intensity values for wavelength outside the emission bands. It leads therefore to pure mathematical drawbacks: when the denominator goes to zero, the noise of the calibration function will dramatic enhance. Surprisingly, it is not visible in the presented spectra. In the SI, the authors explain that they use some signal treatment to calculate a common baseline. It makes sense for the non signal spectral zones but what about the bottom of the bands where the signal tends to zero?  
+
+We have finally decided to comment the limitation of this approach in the introduction of this manuscript by including the following sentence: ...This is almost impossible to achieve over the entire wavelength range and a calibration routine must be set up. A first strategy consists in measuring the calibration function between the two arms[1]. It is based on the division of one spectrum recorded on one of polarization encoding arm by the same spectrum recorded on the other polarization encoding arm. This mathematical procedure leads to a dramatic noise enhancement of the calibrating function when the intensity of the denominator is close to the background noise and requires complex baseline removal in order to avoid "division by zero" errors. Nevertheless, it allows fast CPL measurements on high \(g_{\mathrm{lum}} \geq 0.1\) molecules but no validation on low \(g_{\mathrm{lum}}\) molecules has been described.  
+
+We also give more details in the manuscript when speaking about the spatial separation of the polarization: It is described by MacKenzie et al in [1] who divide the two signals coming from the two arms after inverting the role of one from left to right circular polarizer. Theoretically it eliminates the sources of artifacts proportional to \(\Delta T\) . However, as it is based on the division of two signals, it induces high noise in the foot bands where the signal is close to the noise and special mathematical treatments are needed to avoid "division by zero" errors. Moreover, the derivative artifacts are enhanced as they appear both in the division and the subtraction of non perfectly \(\lambda\) matched spectra and can explained the remaining artifacts discussed in this work.  
+
+Another very important error is the use of rotating polarizers in front of the spectrometers. Indeed spectrometers with their grating are highly polarization dependent devices. This has two consequences on the alignment/calibration procedure:  
+
+1. The described alignment procedure of the relative position of the polarizer and the QWP is based on the rotation of the polarizers and the comparison with a CPL spectrum (SI section 1) measured on a commercial instrument. It combines the polarizer/QWP and polarizer/grating orientations effects! Their procedure is useless and probably ends up with a compromise between right polarization orientation and arms balances.Relative orientation between QWP and polarizer can be easily done with two cross polarizers with an accuracy of about \(0.05^{\circ}\)  
+
+2. The calibration function to balance the arms (SI Section 2) uses two orientations of one polarizer on one arm to compute the calibration factor. It is false because by changing the orientation of the polarizer the response of the arms is also modified via the monochromator response.  
+
+This last point is added at the end of the theoretical study: A final important point is that the arm inversion must be done without changing the optical property of the arms except for the polarization encoding. It is therefore imperative not to change the polarization orientation just before the monochromator because the gratings are very polarization sensitive and with a complicated wavelength dependence. It is why we use a rotating waveplate followed by a fixed PBS: in this configuration each detector has the same response whatever the type of polarization encoding and the auto-compensation is valid. Inverted the polarizing elements (fixed QWP and rotating polarizers) results in erroneous calibration and measurement stages.  
+
+This is a personal opinion, but I'm not sure that instrument development manuscripts should be published in a general science journal unless the new instrument is used to provide insight into a novel scientific problem, of reasonably broad interest. I do not believe that this manuscript meets this criteria. Consequently, I would not recommend publication.
+
+<--- Page Split --->
+
+We think that our set- up could allow the development of a new apparatus for measuring fast CPL measurement which is a topic in exponentially growth. All the ways to get an easier detection of CPL could have a high impact from chemical monitoring in pharmacology to in- situ biological measurements. All the "simple" CPL set- ups published earlier deal with high \(\mathrm{B_{CPL}}\geq 0.5\mathrm{M^{- 1}cm^{- 1}}\) . Here, we show that it can be also used for low \(\mathrm{B_{CPL}}\) molecules as the camphorequinone providing the auto- compensation procedure. It opens the use of CPL detection to wide range of molecules. This point is added in the introduction (line 63): All these results have been obtained with emitters having CPL brightness factors \(B_{CPL}\geq 0.5M^{- 1}cm^{- 1}\) [2]. In this study, we show that a single scan CPL setup can be extended for the measurement of CPL spectra of chiral compounds whatever their CPL brightness.  
+
+## References  
+
+[1] Lewis E. MacKenzie, Lars- Olof Pålsson, David Parker, Andrew Beeby, and Robert Pal. Rapid time- resolved circular polarization luminescence (cpl) emission spectroscopy. Nature Communications, 11(1):1676, April 2020. [2] Lorenzo Arrico, Lorenzo Di Bari, and Francesco Zinna. Quantifying the overall efficiency of circularly polarized emitters. Chemistry - A European Journal, 27(9):2920- 2934, 2021.
+
+<--- Page Split --->
+
+## REVIEWERS' COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+I thank the authors for their careful revisions. They have thoroughly revised the manuscripts by adding new important data and addressing all my points. I think that now the work is complete and of excellent quality. I therefore recommend publication as is.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6a69bed05a64141bd0130e6ebb30dedcb1c1aa61c6ada15b1de153ae15ddc10/supplementary_0_Peer Review File__d6a69bed05a64141bd0130e6ebb30dedcb1c1aa61c6ada15b1de153ae15ddc10_det.mmd

@@ -0,0 +1,206 @@
+<|ref|>title<|/ref|><|det|>[[61, 41, 508, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>title<|/ref|><|det|>[[66, 111, 360, 140]]<|/det|>
+# Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[88, 154, 910, 241]]<|/det|>
+Theoretical and experimental analysis of circularly polarized luminescence spectrophotometers for artifact- free measurements using a single CCD camera  
+
+<|ref|>image<|/ref|><|det|>[[56, 732, 240, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 911, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[119, 85, 357, 101]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 137, 448, 152]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 165, 863, 278]]<|/det|>
+The manuscript reports a CPL instrument designed on a novel concept. It uses a single CCD camera and two independent channels. The authors show experimentally and theoretically that errors and artifacts can be minimized, by different ways of separating the two polarizations. In general, this work is very well carried out, and there is a huge interest nowadays in CPL and how to measure it. I believe that all efforts leading to cheaper (with respect to commercial CPL spectrofluoropolarimeters) but reliable alternatives to measure CPL are welcome and have the potentials to expand the field even more. For these reasons, I would like to recommend the manuscript for publication after a few points are taken into account.  
+
+<|ref|>text<|/ref|><|det|>[[117, 291, 880, 475]]<|/det|>
+1) The authors are aware that a general problem of CPL, no matter how it is measured, is due to linear anisotropies of emission. Indeed, in the present set-up, such artifact comes from the coupling between S1 and QWP non-ideal retardance. The authors may elaborate on the fact that in solution, these contributions become significant where emission lifetime and molecular tumbling rate occur on a similar time scale, and therefore a depolarization due to orientational averaging can not occur completely (according to Perrin equation). When using a \(90^{\circ}\) excitation geometry, an effective way to cancel this contribution is by linearly polarizing the excitation beam on the plane of the detection. In this way only molecules that can not contribute to linear anisotropy are excited, but the cost is a relevant decrease of the signal. I suggest the authors perform the experiment by using an achiral fluorescent system giving a significant linear contribution and try to eliminate the artifacts as described above. This is important to assess, because such situation is very common when working with solid samples and rather common with organic fluorophores in solution.  
+
+<|ref|>text<|/ref|><|det|>[[118, 488, 868, 559]]<|/det|>
+2) When the polarizations are time-separated, the stability of the light source becomes critical, as the authors describe. I wonder if the instabilities observed by the authors are due to a constant drift (up or downwards) or just to random fluctuation. If the excitation lamp signal were continuously acquired as well, would it be possible to normalize constantly the signal by the excitation intensity?  
+
+<|ref|>text<|/ref|><|det|>[[118, 572, 850, 629]]<|/det|>
+3) It is not very clear to me the origin of the artifact due to wavelength mismatch and why it is expected to be proportional to the derivative of the total luminescence. In particular, in the experiments shows in Fig. 7, is this artifact expected to be symmetric with respect to the tilt angle?  
+
+<|ref|>text<|/ref|><|det|>[[118, 642, 880, 727]]<|/det|>
+4) I appreciate that with the instruments proposed by the authors g_lum as low as \(10^{\wedge} - 4\) can be measured. This is a lower bound, anyway, because the actual limit depends on the efficiency of the emitter, as well. Even higher g factors become difficult to measure if the output photons are simply not enough. Can the authors quantify, the limit of their instrumentation in terms of circularly polarization brightness (B_CPL)? I think that most instruments can easily measure compounds with a B_CPL around 1 M-1. cm-1.  
+
+<|ref|>text<|/ref|><|det|>[[118, 740, 844, 769]]<|/det|>
+5) A rough estimate of the cost of the proposed apparatus could be given to put it into context with commercial instrumentation.  
+
+<|ref|>text<|/ref|><|det|>[[118, 782, 868, 839]]<|/det|>
+6) When mixing the two enantiomers of the Eu complex to generate Fig. 8, the authors write that "Figure 8 clearly shows this gradual racemization effect". Actually, there is no racemization effect, as this would imply some chemical process, the authors are simply preparing scalemic solutions. Please rephrase this sentence.  
+
+<|ref|>text<|/ref|><|det|>[[118, 852, 870, 895]]<|/det|>
+7) Concerning NIR-CPL, the authors may be interested in two recent papers in which measurements of Er CPL around 1500 nm with simple optics are reported (10.1021/jacs.2c01134, 10.1002/anie.202208326)
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 84, 868, 126]]<|/det|>
+8) The language of the manuscript should be revised. For example (the list is non-comprehensive): "analyxis", "cosinus", "and CPL measurements [18, 19] and ourselves [20, 21]", "by saving one spectrophotometer, as regard CPL setups based on two of theme", etc.  
+
+<|ref|>sub_title<|/ref|><|det|>[[120, 215, 448, 229]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 243, 872, 298]]<|/det|>
+The manuscript describes the construction of a CPL spectrometer with an unconventional design. The instrument allows rapid acquisition of CPL spectra. A paper describing an alternative design of instrument for rapid CPL detection has already been published in Nat COMM https://www.nature.com/articles/s41467- 020- 15469- 5.  
+
+<|ref|>text<|/ref|><|det|>[[118, 312, 845, 340]]<|/det|>
+I'm not sure the the design outlined in the current manuscript is demonstrably superior to that reported earlier.  
+
+<|ref|>text<|/ref|><|det|>[[118, 354, 867, 410]]<|/det|>
+This is a personal opinion, but I'm not sure that instrument development manuscripts should be published in a general science journal unless the new instrument is used to provide insight into a novel scientific problem, of reasonably broad interest. I do not believe that this manuscript meets this criteria. Consequently, I would not recommend publication.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[333, 71, 662, 92]]<|/det|>
+## Response to Reviewers  
+
+<|ref|>text<|/ref|><|det|>[[105, 97, 893, 197]]<|/det|>
+Response to ReviewersWe would like to the reviewers for having spent their time and efforts to provide a feedback on our manuscript entitled "Theoretical and experimental analysis of circularly polarized luminescence spectrophotometers for artifact- free measurements using a single CCD camera". We believe that the reviewers comments helped us to improve the quality of the manuscript. From the reviewers comments we noticed that some aspects of the manuscript are not well understood and we did our best to make them clearer. We have taken into account all the remarks and issues of the referees. As asked by the reviewer 1, we have added some new experimental results to estimate the contribution of the fluorescence linear anisotropy on the CPL artifacts.  
+
+<|ref|>text<|/ref|><|det|>[[125, 197, 816, 211]]<|/det|>
+Please find in the following the reviewers comments in blue and our point- by- point responses.  
+
+<|ref|>sub_title<|/ref|><|det|>[[105, 242, 205, 255]]<|/det|>
+## Reviewer 1:  
+
+<|ref|>text<|/ref|><|det|>[[105, 263, 893, 364]]<|/det|>
+The manuscript reports a CPL instrument designed on a novel concept. It uses a single CCD camera and two independent channels. The authors show experimentally and theoretically that errors and artifacts can be minimized, by different ways of separating the two polarizations. In general, this work is very well carried out, and there is a huge interest nowadays in CPL and how to measure it. I believe that all efforts leading to cheaper (with respect to commercial CPL spectrofluoropolarimeters) but reliable alternatives to measure CPL are welcome and have the potentials to expand the field even more. For these reasons, I would like to recommend the manuscript for publication after a few points are taken into account.  
+
+<|ref|>text<|/ref|><|det|>[[130, 364, 608, 377]]<|/det|>
++ We appreciate the appreciation of Reviewer 1 about our work.  
+
+<|ref|>text<|/ref|><|det|>[[104, 378, 893, 533]]<|/det|>
+1) The authors are aware that a general problem of CPL, no matter how it is measured, is due to linear anisotropies of emission. Indeed, in the present set-up, such artifact comes from the coupling between S1 and QWP non-ideal retardance. The authors may elaborate on the fact that in solution, these contributions become significant where emission lifetime and molecular tumbling rate occur on a similar time scale, and therefore a depolarization due to orientational averaging can not occur completely (according to Perrin equation). When using a \(90^{\circ}\) excitation geometry, an effective way to cancel this contribution is by linearly polarizing the excitation beam on the plane of the detection. In this way only molecules that can not contribute to linear anisotropy are excited, but the cost is a relevant decrease of the signal. I suggest the authors perform the experiment by using an achiral fluorescent system giving a significant linear contribution and try to eliminate the artifacts as described above. This is important to assess, because such situation is very common when working with solid samples and rather common with organic fluorophores in solution.  
+
+<|ref|>text<|/ref|><|det|>[[105, 533, 893, 603]]<|/det|>
++This comment highlight a very important point. In the first version of the article, we worked only with isotropic solutions without polarization photoselection. We agree that linear anisotropies may lead to undesired measurement artifacts when the isotropy of the medium can not be ensured. We therefore determined the possible contribution of the linear anisotropies to the CPL signal from both a theoretical and an experimental point of view:  
+
+<|ref|>text<|/ref|><|det|>[[125, 612, 893, 670]]<|/det|>
+1. We have added an experimental study on fluorescein molecules as a function of the viscosity of the solution - Section 5. It shows that indeed a CPL signals can be measured for highly viscous solutions even if fluorescein is achiral! This founding (anticipated by Reviewer 1) makes us go deeper into the analysis of our set-up.  
+
+<|ref|>text<|/ref|><|det|>[[125, 679, 893, 735]]<|/det|>
+2. From a theoretical point of view, we took into account circular and \(+45^{\circ}\) dichroism in the modeling of the PBS (lines 145-150). We have included these parameters the SI, equation 2 and table 2 which gathers all the results. Finally, these additional parameters complicate the maths but explain very well the experimental results.  
+
+<|ref|>text<|/ref|><|det|>[[125, 745, 893, 817]]<|/det|>
+3. The contributions of the LD' and the CD of the PBS to the measured CPL are now quantified (table 2). We choose not to add the consideration of the linear anisotropies for the description of the space separation set-up while in any event it is not the best to perform the CPL measurement. For the time and the time+space set-up, only the residual CD in the PBS induces a \(CDS_{1}^{e}\) signal that can not be differentiated from the \(S_{3}^{e}\) one's. It is added in Eq.4 and 6 and discuss accordingly.  
+
+<|ref|>text<|/ref|><|det|>[[125, 826, 893, 882]]<|/det|>
+4. In parallel, we have recorded the \(+45\) and CD of the polarizing beam splitter with our CD/LD spectrometer (SI Section 3). It reveals an unexpected few percents of signals for the circular and \(+45^{\circ}\) linear dichroism. This value of PBS circular dichroism is in very good agreement with the theory and experimental result on the fluorescein.  
+
+<|ref|>text<|/ref|><|det|>[[130, 892, 774, 907]]<|/det|>
+Thanks again to Reviewer 1, for pointing out what was missing for a complete analysis.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[105, 74, 893, 131]]<|/det|>
+2) When the polarizations are time-separated, the stability of the light source becomes critical, as the authors describe. I wonder if the instabilities observed by the authors are due to a constant drift (up or downwards) or just to random fluctuation. If the excitation lamp signal were continuously acquired as well, would it be possible to normalize constantly the signal by the excitation intensity?  
+
+<|ref|>text<|/ref|><|det|>[[105, 131, 893, 262]]<|/det|>
+Within our integration time the random fluctuation is very small and the signal difference comes mainly from the drift. In Sec. 3.2 (line 303), we estimate this drift around \(1\%\) . Theoretically we agree Reviewer 1 : if we are able to continuously measure the pump power, we can normalize the CPL signal. However, from an experimental point of view, we need a measurement set- up with an accuracy \(\sim \frac{1\%}{10} = 0.1\%\) . Standard photodiode sensors (thorlabs- Photodiode Power Sensors) which combine enough speed and accuracy are proposed with a linearity of \(\pm 0.5\%\) and an accuracy of \(\pm 5\%\) which make them useless. Trying to normalize the signal with the recorded pump power will therefore report the issue from the stability of the pump to the stability of the measurement set- up and require a rather sophisticated detection apparatus with thermally controlled electronic. We like better rely on the two measurements set- up which reveals to be very efficient.  
+
+<|ref|>text<|/ref|><|det|>[[105, 260, 893, 316]]<|/det|>
+To explain this point in the paper we have added line 303: These intensity variations come from the slowly varying drift of the pump source. To monitor this drift in order to scale the fluorescence intensity would require detection set- up with an accuracy of \(\approx 0.1\%\) . This can not be performed with standard photodiode power sensors which usually present a few percent accuracy.  
+
+<|ref|>text<|/ref|><|det|>[[105, 316, 893, 359]]<|/det|>
+3) It is not very clear to me the origin of the artifact due to wavelength mismatch and why it is expected to be proportional to the derivative of the total luminescence. In particular, in the experiments shows in Fig. 7, is this artifact expected to be symmetric with respect to the tilt angle?  
+
+<|ref|>text<|/ref|><|det|>[[105, 359, 893, 444]]<|/det|>
+We now explain this in the paper line 201 after "Less intuitive, the wavelength mismatch of the two recorded spectra induces a CPL signal proportional to the luminescence derivative \(\left(\frac{\partial S_0}{\partial \lambda}\right)\) : " Indeed the difference between two identical spectra recorded with a wavelength mismatch \(\Delta \lambda\) writes: \(\Delta S = S(\lambda + \Delta \lambda) - S(\lambda)\) . Because \(\Delta \lambda \ll \lambda\) , we may use the Taylor- series expansion: \(S(\lambda + \Delta \lambda) \approx S(\lambda) + \Delta \lambda \frac{dS}{d\lambda} (\lambda)\) . Therefore, the recorded CPL signal, being a difference between two unmatched signals, contains unwanted terms proportional to the signal derivative."  
+
+<|ref|>text<|/ref|><|det|>[[105, 444, 893, 516]]<|/det|>
+In the experiment shown in Fig. 7, we tilted the camera in order to experimentally add a wavelength mismatch between the top and bottom spectra recorded on the camera. The goal is to qualitatively check the theory not to do quantification. Indeed, the exact wavelength mismatch is wavelength dependent and is quite complicated to model because it depends on the position of the rotation center, the exact dispersion formula and the optical aberrations at the high aperture angles.  
+
+<|ref|>text<|/ref|><|det|>[[105, 516, 893, 601]]<|/det|>
+However as Reviewer 1 mentioned, by inverting the tilt angle, the mismatch is inverted and the artifact also (false \(\mathrm{CPL}\propto \Delta \lambda \frac{\partial S_0}{\partial \lambda}\) ). We now describe this in the new version (line 328): "These artifacts are especially intense in the \(620 \mathrm{nm}\) region where the fluorescence derivative is the highest. At a maximum camera tilt angle of \(- 1.2^{\circ}\) , the artifacts are nearly proportional to the derivative (black dashed curve). By inverting the angle to \(+1.2^{\circ}\) , the artifacts change in sign but not in shape as expected by the theory (Tab. 2 line 3, false \(\mathrm{CPL}\propto \Delta \lambda \frac{\partial S_0}{\partial \lambda}\) )."  
+
+<|ref|>text<|/ref|><|det|>[[105, 608, 893, 681]]<|/det|>
+I appreciate that with the instruments proposed by the authors \(g_{lum}\) as low as \(10^{- 4}\) can be measured. This is a lower bound, anyway, because the actual limit depends on the efficiency of the emitter, as well. Even higher g factors become difficult to measure if the output photons are simply not enough. Can the authors quantify, the limit of their instrumentation in terms of circularly polarization brightness \(\mathrm{(B_{CPL})}\) ? I think that most instruments can easily measure compounds with a \(\mathrm{B_{CPL}}\) around \(1 \mathrm{M}^{- 1} \mathrm{cm}^{- 1}\) .  
+
+<|ref|>text<|/ref|><|det|>[[105, 681, 893, 739]]<|/det|>
+We agree with the remark of Reviewer 1, it is better to consider the \(\mathrm{B_{CPL}}\) in order to compare the results. We discuss now of the capability of our set- up considering the \(\mathrm{B_{CPL}}\) (emitter parameter) and the integration time required to get "good" spectra (signal to noise higher than 10). It was written in the last two paragraphs of the introduction, in the signal to noise ratio discussion (section 3.5) and in the conclusion.  
+
+<|ref|>text<|/ref|><|det|>[[105, 739, 893, 781]]<|/det|>
+5) A rough estimate of the cost of the proposed apparatus could be given to put it into context with commercial instrumentation. The cost of this kind of apparatus is about \(40\mathrm{k}\mathrm{\textregistered}\) mainly driven by the CCD camera (about \(30\mathrm{k}\mathrm{\textregistered}\) ). As requested, we have included this information in the conclusion.  
+
+<|ref|>text<|/ref|><|det|>[[105, 781, 893, 823]]<|/det|>
+6) When mixing the two enantiomers of the Eu complex to generate Fig. 8, the authors write that Figure 8 clearly shows this gradual racemization effect. Actually, there is no racemization effect, as this would imply some chemical process, the authors are simply preparing scalemic solutions. Please rephrase this sentence.  
+
+<|ref|>text<|/ref|><|det|>[[125, 823, 598, 838]]<|/det|>
+We, agree totally, as requested, the sentence as been rephrased.  
+
+<|ref|>text<|/ref|><|det|>[[105, 838, 893, 866]]<|/det|>
+7) Concerning NIR-CPL, the authors may be interested in two recent papers in which measurements of Er CPL around \(1500 \mathrm{nm}\) with simple optics are reported (10.1021/jacs.2c01134, 10.1002/anie.202208326)  
+
+<|ref|>text<|/ref|><|det|>[[120, 866, 840, 880]]<|/det|>
+We thank Reviewer 1 for the information. These references have been added in the introduction.  
+
+<|ref|>text<|/ref|><|det|>[[105, 881, 893, 922]]<|/det|>
+8) The language of the manuscript should be revised. For example (the list is non-comprehensive): analyzis, cosinus, and CPL measurements [18, 19] and ourselves [20, 21], by saving one spectrophotometer, as regard CPL setups based on two of theme, etc.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[128, 74, 883, 90]]<|/det|>
+As requested, the text was thoroughly read and we made our best to rephrase or correct the language.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[104, 75, 201, 88]]<|/det|>
+## Reviewer 2  
+
+<|ref|>text<|/ref|><|det|>[[104, 96, 893, 154]]<|/det|>
+The manuscript describes the construction of a CPL spectrometer with an unconventional design. The instrument allows rapid acquisition of CPL spectra. A paper describing an alternative design of instrument for rapid CPL detection has already been published in Nat Comm https://www.nature.com/articles/s41467- 020- 15469- 5.  
+
+<|ref|>text<|/ref|><|det|>[[103, 154, 891, 182]]<|/det|>
+I'm not sure the the design outlined in the current manuscript is demonstrably superior to that reported earlier.  
+
+<|ref|>text<|/ref|><|det|>[[104, 183, 893, 337]]<|/det|>
+We are aware of this paper and it is cited in the article. Only bright spectra of CPL Eu3+ complexes have been recorded using this setup. However, we are sorry to say that this paper presents fundamental misunderstandings. First, the calibration procedure is based on a division between two signals. While it sounds mathematically possible, it is very difficult to perform properly when the denominator signal tend to zero. They rightly use for these two signals the Eu3+ fluorescence to be sure that the beams follow the same paths. Therefore, the denominator of the calibration function is an Eu3+ spectra which has low and zero intensity values for wavelength outside the emission bands. It leads therefore to pure mathematical drawbacks: when the denominator goes to zero, the noise of the calibration function will dramatic enhance. Surprisingly, it is not visible in the presented spectra. In the SI, the authors explain that they use some signal treatment to calculate a common baseline. It makes sense for the non signal spectral zones but what about the bottom of the bands where the signal tends to zero?  
+
+<|ref|>text<|/ref|><|det|>[[104, 337, 893, 466]]<|/det|>
+We have finally decided to comment the limitation of this approach in the introduction of this manuscript by including the following sentence: ...This is almost impossible to achieve over the entire wavelength range and a calibration routine must be set up. A first strategy consists in measuring the calibration function between the two arms[1]. It is based on the division of one spectrum recorded on one of polarization encoding arm by the same spectrum recorded on the other polarization encoding arm. This mathematical procedure leads to a dramatic noise enhancement of the calibrating function when the intensity of the denominator is close to the background noise and requires complex baseline removal in order to avoid "division by zero" errors. Nevertheless, it allows fast CPL measurements on high \(g_{\mathrm{lum}} \geq 0.1\) molecules but no validation on low \(g_{\mathrm{lum}}\) molecules has been described.  
+
+<|ref|>text<|/ref|><|det|>[[104, 466, 894, 580]]<|/det|>
+We also give more details in the manuscript when speaking about the spatial separation of the polarization: It is described by MacKenzie et al in [1] who divide the two signals coming from the two arms after inverting the role of one from left to right circular polarizer. Theoretically it eliminates the sources of artifacts proportional to \(\Delta T\) . However, as it is based on the division of two signals, it induces high noise in the foot bands where the signal is close to the noise and special mathematical treatments are needed to avoid "division by zero" errors. Moreover, the derivative artifacts are enhanced as they appear both in the division and the subtraction of non perfectly \(\lambda\) matched spectra and can explained the remaining artifacts discussed in this work.  
+
+<|ref|>text<|/ref|><|det|>[[105, 580, 893, 622]]<|/det|>
+Another very important error is the use of rotating polarizers in front of the spectrometers. Indeed spectrometers with their grating are highly polarization dependent devices. This has two consequences on the alignment/calibration procedure:  
+
+<|ref|>text<|/ref|><|det|>[[124, 627, 893, 714]]<|/det|>
+1. The described alignment procedure of the relative position of the polarizer and the QWP is based on the rotation of the polarizers and the comparison with a CPL spectrum (SI section 1) measured on a commercial instrument. It combines the polarizer/QWP and polarizer/grating orientations effects! Their procedure is useless and probably ends up with a compromise between right polarization orientation and arms balances.Relative orientation between QWP and polarizer can be easily done with two cross polarizers with an accuracy of about \(0.05^{\circ}\)  
+
+<|ref|>text<|/ref|><|det|>[[124, 720, 893, 763]]<|/det|>
+2. The calibration function to balance the arms (SI Section 2) uses two orientations of one polarizer on one arm to compute the calibration factor. It is false because by changing the orientation of the polarizer the response of the arms is also modified via the monochromator response.  
+
+<|ref|>text<|/ref|><|det|>[[104, 768, 893, 869]]<|/det|>
+This last point is added at the end of the theoretical study: A final important point is that the arm inversion must be done without changing the optical property of the arms except for the polarization encoding. It is therefore imperative not to change the polarization orientation just before the monochromator because the gratings are very polarization sensitive and with a complicated wavelength dependence. It is why we use a rotating waveplate followed by a fixed PBS: in this configuration each detector has the same response whatever the type of polarization encoding and the auto-compensation is valid. Inverted the polarizing elements (fixed QWP and rotating polarizers) results in erroneous calibration and measurement stages.  
+
+<|ref|>text<|/ref|><|det|>[[104, 869, 893, 925]]<|/det|>
+This is a personal opinion, but I'm not sure that instrument development manuscripts should be published in a general science journal unless the new instrument is used to provide insight into a novel scientific problem, of reasonably broad interest. I do not believe that this manuscript meets this criteria. Consequently, I would not recommend publication.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[104, 74, 893, 204]]<|/det|>
+We think that our set- up could allow the development of a new apparatus for measuring fast CPL measurement which is a topic in exponentially growth. All the ways to get an easier detection of CPL could have a high impact from chemical monitoring in pharmacology to in- situ biological measurements. All the "simple" CPL set- ups published earlier deal with high \(\mathrm{B_{CPL}}\geq 0.5\mathrm{M^{- 1}cm^{- 1}}\) . Here, we show that it can be also used for low \(\mathrm{B_{CPL}}\) molecules as the camphorequinone providing the auto- compensation procedure. It opens the use of CPL detection to wide range of molecules. This point is added in the introduction (line 63): All these results have been obtained with emitters having CPL brightness factors \(B_{CPL}\geq 0.5M^{- 1}cm^{- 1}\) [2]. In this study, we show that a single scan CPL setup can be extended for the measurement of CPL spectra of chiral compounds whatever their CPL brightness.  
+
+<|ref|>sub_title<|/ref|><|det|>[[105, 224, 232, 242]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[104, 253, 893, 333]]<|/det|>
+[1] Lewis E. MacKenzie, Lars- Olof Pålsson, David Parker, Andrew Beeby, and Robert Pal. Rapid time- resolved circular polarization luminescence (cpl) emission spectroscopy. Nature Communications, 11(1):1676, April 2020. [2] Lorenzo Arrico, Lorenzo Di Bari, and Francesco Zinna. Quantifying the overall efficiency of circularly polarized emitters. Chemistry - A European Journal, 27(9):2920- 2934, 2021.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[119, 85, 376, 101]]<|/det|>
+## REVIEWERS' COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 138, 448, 153]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 166, 853, 209]]<|/det|>
+I thank the authors for their careful revisions. They have thoroughly revised the manuscripts by adding new important data and addressing all my points. I think that now the work is complete and of excellent quality. I therefore recommend publication as is.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6ae4391acd52f73e2a8b74ef0407d018f726bf116b37a0148dd127ca664269f/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d6ae4391acd52f73e2a8b74ef0407d018f726bf116b37a0148dd127ca664269f/supplementary_0_Peer Review File__d6ae4391acd52f73e2a8b74ef0407d018f726bf116b37a0148dd127ca664269f.mmd

@@ -0,0 +1,27 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Decoding the olfactory map through targeted transcriptomics links murine olfactory receptors to glomeruli  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operating a transparent peer review scheme. This document only contains reviewer comments and rebuttal letters for versions considered at Nature Communications. Mentions of the other journal have been redacted.  
+
+## REVIEWERS' COMMENTS  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors have further clarified their findings and edited the text accordingly to address the concerns raised. In my opinion, the revised manuscript is ready for publication. Dinu Albeanu  
+
+Reviewer #3 (Remarks to the Author):  
+
+I was enthusiastic in supporting the publication of this paper in [redacted]. The authors appeared to have modified the paper to address some of the questions I raised in that review. I am enthusiastic in supporting the publication of this paper in Nature Communications.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6ae4391acd52f73e2a8b74ef0407d018f726bf116b37a0148dd127ca664269f/supplementary_0_Peer Review File__d6ae4391acd52f73e2a8b74ef0407d018f726bf116b37a0148dd127ca664269f_det.mmd

@@ -0,0 +1,34 @@
+<|ref|>title<|/ref|><|det|>[[60, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[68, 110, 361, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[75, 155, 920, 211]]<|/det|>
+Decoding the olfactory map through targeted transcriptomics links murine olfactory receptors to glomeruli  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 781]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 912, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 90, 866, 153]]<|/det|>
+Editorial Note: This manuscript has been previously reviewed at another journal that is not operating a transparent peer review scheme. This document only contains reviewer comments and rebuttal letters for versions considered at Nature Communications. Mentions of the other journal have been redacted.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 166, 330, 181]]<|/det|>
+## REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[119, 193, 415, 207]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[119, 218, 840, 259]]<|/det|>
+The authors have further clarified their findings and edited the text accordingly to address the concerns raised. In my opinion, the revised manuscript is ready for publication. Dinu Albeanu  
+
+<|ref|>text<|/ref|><|det|>[[119, 283, 415, 297]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 309, 865, 350]]<|/det|>
+I was enthusiastic in supporting the publication of this paper in [redacted]. The authors appeared to have modified the paper to address some of the questions I raised in that review. I am enthusiastic in supporting the publication of this paper in Nature Communications.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6b1a321937aeb718b93604db1067eb406c84e4efe4fff5fa0e724a8088d0b6e/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d6b1a321937aeb718b93604db1067eb406c84e4efe4fff5fa0e724a8088d0b6e/supplementary_0_Peer Review File__d6b1a321937aeb718b93604db1067eb406c84e4efe4fff5fa0e724a8088d0b6e.mmd

@@ -0,0 +1,413 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Phylogenomic analyses highlight innovation and introgression in the continental radiations of Fagaceae across the Northern Hemisphere  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  
+
+Remarks to the Author:  
+
+This exceptionally well- written manuscript deals with a number of fundamental evolutionary questions in a novel way. Specifically, Biao- Feng Zhou, Shuai Yuan, Andrew Crowl and co- workers demonstrate using an impressive amount of data how unrelated chloroplast and nuclear genomes both contribute to a holistic understanding of the evolutionary history of a major north temperate group of woody angiosperms, the Fagaceae family. Using time calibrated phylogenies for both genomes and the conflicting signal in these genomes, the authors show that chloroplast genomes encapsulate the shared biogeographic histories of (partly) distantly related groups, whereas the nuclear genome reflects species and higher taxa relationships.  
+
+They further show that clade ages from dated phylogenies and diversification rates correlate with major biological innovations possibly triggered by global tectonic- climatic changes. For example, the shift to wind- pollination in oaks, whereas chestnuts and other castaneoids are mainly insect- pollinated, or the shift to hypogeous germination in the HS clade that is often correlated with larger seeds, facilitated major radiations of members of the HS clade under globally cooling conditions. Methodology and analytical approaches are all cutting- edge and the figures accompanying the manuscript are very informative. Figure 2 will most likely be reproduced in text books as it impressively shows how different the chloroplast and nuclear genomes function.  
+
+I have only very few comments:  
+
+Line 71: Although it is impossible to cite all literature relevant to the evolutionary history of Fagaceae main groups, I think two papers should be cited here: Sadowski, E.- M., J. U. Hammel, and T. Denk. 2018. Synchrotron X- ray imaging of a dichasium cupule of Castanopsis from Eocene Baltic amber. American Journal of Botany 105: 2025- 2036. Naryshkina, N. N., and T. A. Evstigneeva. 2020. Fagaceae in the Eocene palynoflora of the South of Primorskii Region: New data on taxonomy and morphology. Paleontological Journal 54: 429- 439.  
+
+Both papers report Eocene findings of Cyclobalanopsis from the Northern Hemisphere and partly contradict a hypothesis put forward in another paper by Wilf et al. 2019 (cited in the manuscript).  
+
+Line 117: I would say fruits instead of fruit.  
+
+Line 329: How would you then explain the presence of sect. Quercus pollen in Eocene strata of Hainan, S China? I think the NW origin may be ok, but I was wondering if you have any ideas.  
+
+Line 335: Northern Hemisphere instead of northern hemisphere.  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+This study by Zhou et al. is a multifaceted examination of the evolutionary history of Fagaceae aimed at understanding the extent and consequences of hybridization during the early diversification of the family and the oak genus (Quercus). Overall, I think this is a high- quality study and a strong candidate for publication in Nature Communications. The dataset is impressive, the analyses conducted were extensive and rigorous, and the presentation generally does a nice job of synthesizing the results in an interesting and compelling fashion. The economic and ecological importance of the family and particularly the oak genus should also make the study of interest to a broad audience.  
+
+Below I outline my major and minor suggestions for revision.  
+
+MAJOR COMMENTS
+
+<--- Page Split --->
+
+1. At times I think the manuscript is over-written. And by this, I mean (a) the importance of various attributes is exaggerated or based on speculation and (b) attempts at making grandiose statements often results in convoluted, vague, or sometimes not entirely coherent sentences. Below are some examples and comments. I strongly encourage the authors to revise such passages so that overall the presentation is more modest, concrete, and readable.  
+
+"Innovations related to seed and pollen dispersal are implicated in triggering waves of continental radiations, while fungal symbioses fortified a competitive edge underground" This is highly speculative especially since the paper does not include any analyses of trait- dependent diversification.  
+
+"This resulted in adaptive introgression, further amplifying global proliferation" Some analyses were done that suggest positive selection/maintenance of introgression haplotypes, but the leap to "further amplifying global proliferation" seems to much  
+
+"Northern Hemisphere forests and shrublands are now dominated by species comprising temperate and subtropical lineages, marking one of the greatest floristic transitions in the vegetation history of the Cenozoic"  
+
+NH forests also included many representatives of 'temperate/subtropical' lineages in the early Cenozoic. Undoubtedly there was a major floristic shift after the Eocene, but this sentence reads as vague and empty to me.  
+
+"With a minimum age of ca. 80 million years ago (Ma) and precise aging of new fossilized pollen and macrofossils assigned to some modern groups by 50 Ma, the evolution of major lineages appears to be unusually rapid for forest tree species"  
+
+One, this doesn't seem that rapid given the ages provided, and two, there are plenty of examples of rapidly radiated woody groups.  
+
+The sentence at lines 83- 86 seems overly convoluted, as does the sentence from lines 89 to 94.  
+
+102: "...the first complete family- wide phylogenetic context"  
+
+What does 'complete' mean? This study includes a small fraction of the species diversity. Is it the first nuclear phylogenomic study to include all genera? All genera and all sections of oaks? This seems like another example of trying to stretch the significance when it would be better to just be straightforward and clear.  
+
+191 onwards. "Shift to wind- pollination alone did not increase the diversification rate of oak species immediately, but instead served as a predisposed neutral change that later facilitated rapid radiation of this genus during the expansion of seasonal climates (Fig. 1)."  
+
+This is speculative. Fine to suggest this as a hypothesis but it is unreasonable to declare it so without any sort of direct tests or evidence. I also don't think the next sentence follows very clearly. I think this could be reframed as a hypothesis with compelling contextual or anecdotal evidence.  
+
+231- 232: "However, gene flow between modern genera is without precedent"  
+
+As you noted previously, phylogenetic studies based on the plastome have suggested gene flow between Quercus and related genera (given the geographic structure of the plastid tree). So, I don't think this statement is accurate. But a comprehensive investigation of this using both the nuclear and plastid genomes is novel.  
+
+2. I have a few concerns about the dating analyses. One, dating analyses based on a large number of genes can introduce a lot of heterogeneity into analyses that is hard to accommodate. It is now becoming more common to filter your genes for those that (a) match the species-tree topology and (b) are more 'clock-like' (less root-to-tip variation). Since you have such a large number of loci, this time of approach could be beneficial. I understand that many downstream analyses are based on the
+
+<--- Page Split --->
+
+dated tree, and so asking to have the dating analyses re- done is a big ask. But this alternative could be tried and compared with the original results to see if there are major differences. Concerning the dated chloroplast tree, you might clarify that the 'ML tree' used as the reference was the CP ML tree (not nuclear). Upon first reading I thought you meant you constrained the plastid analysis to the nuclear topology when dating (which would be problematic) but looking at the Suppl I realized this was not the case. So I would just clarify. I also wonder if it would be worthwhile to do a nuclear dating analysis with only the two calibrations used for the CP analysis. Using different calibration schemes for these datasets seems like it would result in a biased comparison between the two.  
+
+## MINOR COMMENTS  
+
+57: "often repeated seasonal biomes" what does often repeated mean?  
+
+69 onward: "Fossil analogs" is perhaps not the best way to put this since 'analog' often refers to independently evolved structures. I would just say "Fossils of Fagaceae are well represented in the Northern Hemisphere..."  
+
+106, 553 (and perhaps elsewhere): "coalescent analyses using ASTRAL- III and SVDquartets" Both ASTRAL and SVDquartets are summary methods—they do not model the coalescent, or in other words, they are not based on the coalescent—and so calling them "coalescent- based" approaches is inaccurate  
+
+107 change to "for all but a few branches"?  
+
+110: "early- diverging lineages" There are many variations on this throughout the manuscript ('early diverging', 'early branching'). Two branches from a node are equal in age so it doesn't make sense to call one lineage "earlier" with respect to the other(s). This is basically the same issue as calling something "basal". I would encourage the authors to revise this language (change to things like "successively sister" etc). This sentence (109 onward) is also generally convoluted and would benefit from revision.  
+
+117 should be "a' single rounded fruit"?  
+
+119: "studies based on sequences derived from RAD- seq datasets and nuclear loci" – this syntax doesn't really make sense  
+
+120- 122: "Despite phylogenetic congruence across methods, high levels of gene- tree conflict within the nuclear genome were observed, likely due to incomplete lineage sorting (ILS; Supplementary Figs. 2, 3 and 4)." This seems a bit hand- wavy given that the goal of the paper is to integrate deep conflicts to understand ancient hybridization  
+
+126: early diverging used again here—revise  
+
+134- 135: "These events closely follow the Cretaceous- Paleogene (K- Pg) boundary dated at 66  
+
+Ma" – 15 million years seems like a broad window to say this "closely followed" the KPg  
+
+Lines around 156: Do you think any seed traits in this clade might have also conferred greater survivorship following the ecological devastation of the KPg?
+
+<--- Page Split --->
+
+175- 178: A little confusing whether fagaceae or fungi is being referred to with respect to increased speciation  
+
+184: 'coincident' might be a better word here?  
+
+291 to 294: Does this also require information on population sizes and, if so, how did you determine those?  
+
+323: I don't know if it makes sense for the genomes to be "contributing unique inferences" since the researcher is doing the inferring, "...each contributing unique insights on the complex combination of divergent..." instead?  
+
+333: "this level of scrutiny" to "our detection"?  
+
+335- 336: This paragraph ender seems a little overblown and vague.  
+
+419: remove 's' after 'SCG': 'SGC regions'  
+
+Reviewer #3:  
+
+Remarks to the Author:  
+
+This is an important and exceptionally well- written paper which describes the evolutionary history of the oak family Fagaceae (comprising oak, chestnut, stone oak and beech species) that dominates temperate forests in the northern hemisphere. The impressive, detailed analysis of plastomes and nuclear genomes of representative species across the family builds on previous results obtained for Quercus by Manos and others. The robust dated nuclear phylogeny generated tells us when lineages (genera, sub- genera, species) originated and how this ties into ecological innovations (hypogeal from epigeal seed germination; wind from insect pollination) and associations (with seed dispersers and mycorrhiza). Most importantly, a comparison of plastome with nuclear phylogenies plus in- depth analysis of gene flow between species and higher lineages provides information on chloroplast capture, historical phylogeography, and adaptive introgression in the family. In conclusion, the authors emphasise the importance of historical introgression to the success of the family.  
+
+It is pointed out in the text that many closely related oak species occur in sympatry and hybridise. Consequently, I did wonder how confident we might be that all samples analysed represented the species they were said to be. For example, Q. robur and Q. petraea are two such species that hybridise to the point that it is very difficult to find plants in the wild that are not hybrids. Are the authors confident that the samples used for these species, and of other species that similarly hybridise extensively, were in fact non- hybrid. If not, how might this affect their findings?  
+
+Minor points  
+
+lines 497- 504. Peculiar. Is there a possible biological reason for the occurrence of these divergent Q. ilex plastomes or are they likely artifacts.  
+
+676. 'One individual gene trees'?  
+
+716. significantly  
+
+722. Suggest modify to 'time that gene flow'  
+
+723. Suggest modify to 'time that gene flow'  
+
+732. Change to 'block'
+
+<--- Page Split --->
+
+## Response to Reviewer #1  
+
+This exceptionally well- written manuscript deals with a number of fundamental evolutionary questions in a novel way. Specifically, Biao- Feng Zhou, Shuai Yuan, Andrew Crowl and co- workers demonstrate using an impressive amount of data how unrelated chloroplast and nuclear genomes both contribute to a holistic understanding of the evolutionary history of a major north temperate group of woody angiosperms, the Fagaceae family. Using time calibrated phylogenies for both genomes and the conflicting signal in these genomes, the authors show that chloroplast genomes encapsulate the shared biogeographic histories of (partly) distantly related groups, whereas the nuclear genome reflects species and higher taxa relationships.  
+
+They further show that clade ages from dated phylogenies and diversification rates correlate with major biological innovations possibly triggered by global tectonic- climatic changes. For example, the shift to wind- pollination in oaks, whereas chestnuts and other castaneoids are mainly insect- pollinated, or the shift to hypogeous germination in the HS clade that is often correlated with larger seeds, facilitated major radiations of members of the HS clade under globally cooling conditions.  
+
+Methodology and analytical approaches are all cutting- edge and the figures accompanying the manuscript are very informative. Figure 2 will most likely be reproduced in text books as it impressively shows how different the chloroplast and nuclear genomes function.  
+
+I have only very few comments:  
+
+Line 71: Although it is impossible to cite all literature relevant to the evolutionary history of Fagaceae main groups, I think two papers should be cited here:  
+
+Sadowski, E.- M., J. U. Hammel, and T. Denk. 2018. Synchrotron \(X\) - ray imaging of a dichasium cupule of Castanopsis from Eocene Baltic amber. American Journal of Botany 105: 2025- 2036.  
+
+Naryshkina, N. N., and T. A. Evstigneeva. 2020. Fagaceae in the Eocene palynoflora of the South of Primorskii Region: New data on taxonomy and morphology. Paleontological Journal 54: 429- 439.  
+
+Both papers report Eocene findings of Cyclobalanopsis from the Northern Hemisphere and partly contradict a hypothesis put forward in another paper by Wilf et al. 2019 (cited in the manuscript).  
+
+---Thank you. We have added these two literatures, although addressing the point on the Wilf et al. (2019) paper would require a separate review on the biogeographic history of Fagaceae. Lines 70, 874- 879.  
+
+Line 117: I would say fruits instead of fruit.  
+
+---We have changed this to read "a single rounded fruit". Line 117.  
+
+Line 329: How would you then explain the presence of sect. Quercus pollen in Eocene strata of Hainan, S China? I think the NW origin may be ok, but I was wondering if you have any ideas.  
+
+---Interesting challenge. We made some minor changes to the text to reflect the potential for an early presence of section Quercus at tropical latitudes in Asia based on
+
+<--- Page Split --->
+
+studies of Hofmann et al data (2019) from Eocene strata in Hainan. Our divergence time estimates based on nuclear data for NW/OW white oak split is roughly as early as 26 Ma (mid- Oligocene), so a late Eocene presence of section Quercus is reasonable, except for the surprisingly tropical extension of its distribution in Hainan. Lines 325, 366- 368.  
+
+The earlier presence of section could be explained by broadly distributed ancestral section Quercus or Pacific lineage that spanned the OW/NW through Beringia at a time when northwestern North America and northeastern Asia were more or less continuous areas with a shared flora. A similar pattern of expansion and contraction has been invoked to explain the fossil history of Fagus, which is now extinct from northwestern North America, but extant in east Asia (Denk & Grimm 2009). The paleo distribution of Fagaceae in and around the high latitudes associated with the Beringian Land Bridge and into more southward amphi- Pacific areas is clearly at odds with the modern distribution of taxa.  
+
+As Hainan became progressively more tropical (extant Fagaceae include Lithocarpus, Castanopsis, Trigonobalanus, sections Cyclobalanopsis, Cerris, and Ilex), subtropical and temperate Fagaceae like section Quercus may have gone extinct (Zhu 2016). A second expansion of section Quercus derived from a NW temperate ancestor shared with sect Quercus subsect Albae could explain a more recent dispersion of Eurasian white oaks across Eurasia during the expansion of seasonal climates in the Oligocene.  
+
+Hofmann CC, Kodrul TM, Liu XY, Jin JH. 2019. Scanning electron microscopy investigations of middle to late Eocene pollen from the Changchang Basin (Hainan island, south China) – Insights into the paleobiogeography and fossil history of Juglans, Fagus, Lagerstroemia, Mortoniodendron, Cornus, Nyssa, Symplocos and some Icacinaceae in SE Asia. Review of Palaeobotany and Palynology 265: 41–61.  
+
+Denk T, Grimm GW. 2009. The biogeographic history of beech trees. Review of Palaeobotany and Palynology 158: 83–100.  
+
+Zhu H. 2016. Biogeographical evidences help revealing the origin of Hainan Island. PLoS ONE 11: e0151941.  
+
+Line 335: Northern Hemisphere instead of northern hemisphere. This change has been made. Line 334.  
+
+Reviewer #2 (Remarks to the Author):  
+
+This study by Zhou et al. is a multifaceted examination of the evolutionary history of Fagaceae aimed at understanding the extent and consequences of hybridization during the early diversification of the family and the oak genus (Quercus). Overall, I think this is a high- quality study and a strong candidate for publication in Nature Communications. The dataset is impressive, the analyses conducted were extensive and rigorous, and the presentation generally does a nice job of synthesizing the results in an interesting and compelling fashion. The economic and ecological importance of the family and particularly the oak genus should also make the study of interest to a broad audience.
+
+<--- Page Split --->
+
+Below I outline my major and minor suggestions for revision.  
+
+## MAJOR COMMENTS  
+
+I. At times I think the manuscript is over-written. And by this, I mean (a) the importance of various attributes is exaggerated or based on speculation and (b) attempts at making grandiose statements often results in convoluted, vague, or sometimes not entirely coherent sentences. Below are some examples and comments. I strongly encourage the authors to revise such passages so that overall the presentation is more modest, concrete, and readable.  
+
+"Innovations related to seed and pollen dispersal are implicated in triggering waves of continental radiations, while fungal symbioses fortified a competitive edge underground" This is highly speculative especially since the paper does not include any analyses of trait-dependent diversification.  
+
+---We have edited this statement to focus only on seed dispersal, which was found to be associated with a spike in diversification in our study. We removed mention of pollen dispersal and fungal symbioses. Lines 41- 44.  
+
+"This resulted in adaptive introgression, further amplifying global proliferation" Some analyses were done that suggest positive selection/maintenance of introgression haplotypes, but the leap to "further amplifying global proliferation" seems to much. ---We have removed this problematic phrase and have rewritten this sentence to focus specifically on the Eurasian white oak clade. Diversification analyses identified this clade as having an increase in net diversification. We feel the link between positive selection and increased diversification in this clade is sufficient to make such a statement. Lines 48- 49.  
+
+"Northern Hemisphere forests and shrublands are now dominated by species comprising temperate and subtropical lineages, marking one of the greatest floristic transitions in the vegetation history of the Cenozoic"  
+
+NH forests also included many representatives of 'temperate/subtropical' lineages in the early Cenozoic. Undoubtedly there was a major floristic shift after the Eocene, but this sentence reads as vague and empty to me.  
+
+---We have toned this statement down to read "one of the major floristic transitions". Line 53.  
+
+"With a minimum age of ca. 80 million years ago (Ma) and precise aging of new fossilized pollen and macrofossils assigned to some modern groups by 50 Ma, the evolution of major lineages appears to be unusually rapid for forest tree species" One, this doesn't seem that rapid given the ages provided, and two, there are plenty of examples of rapidly radiated woody groups.  
+
+---We appreciate this comment and have revised the text to instead focus on the timing of the diversification of the HS clade of Fagaceae, which we found to be associated with an increase in net diversification rate (Lines 72- 76). Fossil taxa ascribed to those genera fall within a much narrower window of time. We are using the rich fossil record of
+
+<--- Page Split --->
+
+Fagaceae to suggest that macroevolutionary change (pollen, cupule/fruit) was rapid in deep time, in contrast to the general findings that woody lineages evolve more slowly than herbaceous lineages using relative rate analysis across sister clades. Life history correlates such as generation time also have figured into this broader question (see Smith & Beaulieu 2009)  
+
+This is an interesting topic that intersects with ideas on the role of life history, climate niche evolution, and correlates with net diversification rates. Following the publication of Smith and Donoghue (2008, cited here) and Smith & Beaulieu (2009), some literature suggests that secondary woodiness is a prerequisite to higher speciation rates in isolated niche space, such as islands and continental sky islands. While there is some evidence for rapid diversification in woody clades such as temperate Quercus (Hipp et al. 2020, cited here and detected in this study as well) and tropical Inga (Fabaceae, Richardson et al. 2001), slower rates of evolution as a function life history, limited sequence divergence or net diversification rates remain associated with tree lineages. More recent analyses suggest that plant size, another proxy for arborescence, also is correlated with lower diversification rates across the angiosperms (Boucher et al. 2017).  
+
+Boucher FC, Verboom GA, Musker S, Ellis AG (2017) Plant size: a key determinant of diversification? New Phytologist 216: 24- 31.  
+
+Nürk NM, Atchison GW, Hughes CE (2019) Island woodiness underpins accelerated disparification in plant radiations. New Phytologist 224: 518- 531.  
+
+Richardson JE, Pennington RT, Pennington TD, Hollingsworth PM (2001) Rapid diversification of a species- rich genus of Neotropical rain forest trees. Science 293: 2242- 2245  
+
+Smith SA and Beaulieu JM (2009) Life- history influences rates of climatic niche evolution in flowering plants. Proceedings of the Royal Society B 276: 4345- 4352  
+
+The sentence at lines 83- 86 seems overly convoluted, as does the sentence from lines 89 to 94.  
+
+---We have attempted to clarify and reduce the length of these sentences. Lines 84- 86, 92.  
+
+102: "...the first complete family- wide phylogenetic context" What does 'complete' mean? This study includes a small fraction of the species diversity. Is it the first nuclear phylogenomic study to include all genera? All genera and all sections of oaks? This seems like another example of trying to stretch the significance when it would be better to just be straightforward and clear. ---We have changed this to simply read "within a broad phylogenetic context." Line 102.  
+
+191 onwards. "Shift to wind- pollination alone did not increase the diversification rate of oak species immediately, but instead served as a predisposed neutral change that later
+
+<--- Page Split --->
+
+facilitated rapid radiation of this genus during the expansion of seasonal climates (Fig. 1)."  
+
+This is speculative. Fine to suggest this as a hypothesis but it is unreasonable to declare it so without any sort of direct tests or evidence. I also don't think the next sentence follows very clearly. I think this could be reframed as a hypothesis with compelling contextual or anecdotal evidence.  
+
+---We agree and have rephrased this as a hypothesis. Lines 190, 193- 194.  
+
+231- 232: "However, gene flow between modern genera is without precedent" As you noted previously, phylogenetic studies based on the plastome have suggested gene flow between Quercus and related genera (given the geographic structure of the plastid tree). So, I don't think this statement is accurate. But a comprehensive investigation of this using both the nuclear and plastid genomes is novel.  
+
+---We have clarified this to read, "inference of gene flow between modern genera has been based solely on plastome data.". Lines 230- 231.  
+
+2. I have a few concerns about the dating analyses. One, dating analyses based on a large number of genes can introduce a lot of heterogeneity into analyses that is hard to accommodate. It is now becoming more common to filter your genes for those that (a) match the species-tree topology and (b) are more 'clock-like' (less root-to-tip variation). Since you have such a large number of loci, this time of approach could be beneficial. I understand that many downstream analyses are based on the dated tree, and so asking to have the dating analyses re-done is a big ask. But this alternative could be tried and compared with the original results to see if there are major differences.  
+
+--- Thank you for raising this important point. We have applied two "gene- shopping" methods to identify genes with the best information for dating. First, we used SortaData (Smith et al. 2018) to filter 212 (top \(90^{\text{th}}\) percentile) most clock- like loci by considering clock- like as the primary criterion, followed by tree- like and tree length. Second, we calculated the Robinson- Foulds (RF) distance between gene trees and the reference tree following Johns et al. (2018), and retained 212 loci with the least RF distance and greater concordant phylogenetic signals. By using three different topologies as reference trees (see above), we generated six reduced datasets. The divergence time estimated on the 2124 genes were almost identical to the six reduced data sets (Pearson's correlation coefficient \(= 0.991 - 0.995\) , \(P < 2\mathrm{e}^{- 16}\) ; Supplementary Fig. 15). These new approaches have been described in the revised Methods (Lines 566- 576), a new Supplementary Table 11 and a new Supplementary Fig. S15.  
+
+Concerning the dated chloroplast tree, you might clarify that the 'ML tree' used as the reference was the CP ML tree (not nuclear). Upon first reading I thought you meant you constrained the plastid analysis to the nuclear topology when dating (which would be problematic) but looking at the Suppl I realized this was not the case. So I would just clarify.  
+
+--- We have clarified this to "plastome ML tree". Line 578.
+
+<--- Page Split --->
+
+I also wonder if it would be worthwhile to do a nuclear dating analysis with only the two calibrations used for the CP analysis. Using different calibration schemes for these datasets seems like it would result in a biased comparison between the two.  
+
+---We have dated the nuclear tree by using the same two calibrations for plastome tree (Lines 585- 587). This result was presented in updated Supplementary Fig. 5g- i.  
+
+## MINOR COMMENTS  
+
+57: "often repeated seasonal biomes" what does often repeated mean? ---This was intended to plant the seed that similar biomes present the opportunity for ecological convergence in Fagaceae. The text has been simplified. Lines 56- 57.  
+
+69 onward: "Fossil analogs" is perhaps not the best way to put this since 'analog' often refers to independently evolved structures. I would just say "Fossils of Fagaceae are well represented in the Northern Hemisphere..."  
+
+---Good point. This change has been made. Line 69.  
+
+106, 553 (and perhaps elsewhere): "coalescent analyses using ASTRAL- III and SVDquartets" Both ASTRAL and SVDquartets are summary methods—they do not model the coalescent, or in other words, they are not based on the coalescent—and so calling them "coalescent- based" approaches is inaccurate  
+
+to read "species- tree analyses" throughout the manuscript. Lines 106, 530 and 563.  
+
+107 change to "for all but a few branches"?  
+
+---This sentence has been revised. Lines 107.  
+
+110: "early- diverging lineages" There are many variations on this throughout the manuscript ('early diverging', 'early branching'). Two branches from a node are equal in age so it doesn't make sense to call one lineage "earlier" with respect to the other(s). This is basically the same issue as calling something "basal". I would encourage the authors to revise this language (change to things like "successively sister" etc). This sentence (109 onward) is also generally convoluted and would benefit from revision.  
+
+---All mention of "early- diverging lineages" has been removed or replaced with "successively sister lineages". Lines 110, 126, 149, 160 and 204.  
+
+117 should be "a' single rounded fruit"?  
+
+---This change has been made. Lines 117.  
+
+119: "studies based on sequences derived from RAD- seq datasets and nuclear loci" — this syntax doesn't really make sense  
+
+---This sentence has been amended to read, "...previous studies based on RAD- seq data." Lines 119.  
+
+120- 122: "Despite phylogenetic congruence across methods, high levels of gene- tree conflict within the nuclear genome were observed, likely due to incomplete lineage
+
+<--- Page Split --->
+
+sorting (ILS; Supplementary Figs. 2, 3 and 4)." This seems a bit hand- wavy given that the goal of the paper is to integrate deep conflicts to understand ancient hybridization This observation refers to the relationships among Quercus, Notholithocarpus, Lithocarpus, and Chrysolepis which we found to be in conflict across nuclear gene trees. Not surprisingly, ILS and hybridization present challenges for estimating the Fagaceae tree of life. For this particular set of relationships and narrow time frame, a stochastic pattern of gene tree resolution is consistent with ILS. These results were detailed in Supplementary Figs. 2, 3 and 4, and briefly referred in main text (lines 119- 122)  
+
+126: early diverging used again here—revise  
+
+This sentence has been revised. Line 126.  
+
+134- 135: "These events closely follow the Cretaceous- Paleogene (K- Pg) boundary dated at 66 Ma" - 15 million years seems like a broad window to say this "closely followed" the KPg  
+
+closely' has been removed from this sentence to state that these events simply followed the KPg. Line 133.  
+
+Lines around 156: Do you think any seed traits in this clade might have also conferred greater survivorship following the ecological devastation of the KPg?  
+
+---Interesting question, and much appreciated. But difficult to add anything concrete on the subject. All seeds produced by Fagaceae are known to have a short life span in the soil. As there is no seed bank, the seeds produced in one season are the only source of recruitment. Recent ecological work suggests that seed mating is the main driver for seedling recruitment in forest trees, highlighting the strong selection pressure of seed predators and episodic nature of successful recruitment. While the shift to hypogeous cotyledons appears to be associated with diversifying potential dispersers, its evolution Bhas been linked to increased survivorship over short time scales. Regarding other traits, anecdotal evidence suggests that selection in fruit/cupule evolution in Castanea + Castanopsis was driven to protect seeds through development whereas enhanced dispersal capacity is a general theme in Lithocarpus and Quercus (see Fig 1 images of spiny cupules vs scaly cupules surrounding or subtending nuts), with secondarily evolved exceptions scattered in both genera to protect the seed, and one notable origin of the acorn fruit type in Castanopsis.  
+
+175- 178: A little confusing whether fagaceae or fungi is being referred to with respect to increased speciation ---We have clarified that it is species radiations of Fagaceae. Lines 176- 177.  
+
+184: 'coincident' might be a better word here?  
+
+---This change has been made. Line 183.  
+
+291 to 294: Does this also require information on population sizes and, if so, how did you determine those?  
+
+---In the method developed by Huerta- Sanchez et al. (2014), the probability of maintaining selectively neutral haplotypes of a given length in both oak sections after
+
+<--- Page Split --->
+
+introgression was determined by the recombination rate, the time that gene flow occurred and generation time. The effective population size was not included in the function. See details in Methods (Lines 724- 733).  
+
+323: I don't know if it makes sense for the genomes to be "contributing unique inferences" since the researcher is doing the inferring. "...each contributing unique insights on the complex combination of divergent..." instead?  
+
+---Agreed. This change has been made. Line 322.  
+
+333: "this level of scrutiny" to "our detection"?  
+
+---This change has been made. Lines 331- 332.  
+
+335- 336: This paragraph ender seems a little overblown and vague.  
+
+---We have edited this to simply read, "The ecological implications of these biotic exchanges of keystone lineages await future study." Lines 334- 335.  
+
+419: remove 's' after 'SCG': 'SGC regions'  
+
+---This change has been made. Line 417.  
+
+## Reviewer #3 (Remarks to the Author):  
+
+This is an important and exceptionally well- written paper which describes the evolutionary history of the oak family Fagaceae (comprising oak, chestnut, stone oak and beech species) that dominates temperate forests in the northern hemisphere. The impressive, detailed analysis of plastomes and nuclear genomes of representative species across the family builds on previous results obtained for Quercus by Manos and others. The robust dated nuclear phylogeny generated tells us when lineages (genera, sub- genera, species) originated and how this ties into ecological innovations (hypogeal from epigeal seed germination; wind from insect pollination) and associations (with seed dispersers and mycorrhiza). Most importantly, a comparison of plastome with nuclear phylogenies plus in- depth analysis of gene flow between species and higher lineages provides information on chloroplast capture, historical phylogeography, and adaptive introgression in the family. In conclusion, the authors emphasise the importance of historical introgression to the success of the family.  
+
+It is pointed out in the text that many closely related oak species occur in sympatry and hybridise. Consequently, I did wonder how confident we might be that all samples analysed represented the species they were said to be. For example, Q. robur and Q. petraea are two such species that hybridise to the point that it is very difficult to find plants in the wild that are not hybrids. Are the authors confident that the samples used for these species, and of other species that similarly hybridise extensively, were in fact non- hybrid. If not, how might this affect their findings?  
+
+--- Good point. The plant material we used in the study was derived from a combination of collections made from natural populations and cultivated plants grown from wild- collected seed. We have updated Supplementary Table 1 to provide data on our voucher specimens.
+
+<--- Page Split --->
+
+Where possible, our joint expertise in Fagaceae taxonomy guided the selection of species, in addition to generally including species that are taxonomically well- understood and easily identified. However, it is true that later generation hybrids may express phenotypes that often converge on one parental species. While we do present evidence of more recent hybridization between several morphologically typical species (see Figure 3A), we believe the impact on our findings at deeper phylogenetic levels is minimal. Our tests basically support the mosaic nature of closely related oak genomes, e.g., sympatric white oaks, initially suggested by analyses using RAD- seq data (Hipp et al. 2020, cited here). And despite gene flow involving the tips of the tree, we detected a high degree of coalescence for the gene tree histories that support the common ancestry of the genera and sections of Quercus. From our analyses, it seems that ILS (Incomplete Lineage Sorting) is the main challenge for estimating the species tree in the crown clade of Fagaceae.  
+
+Minor points  
+
+lines 497- 504. Peculiar. Is there a possible biological reason for the occurrence of these divergent \(Q\) . ilex plastomes or are they likely artifacts.  
+
+One potential biological explanation for the two highly divergent \(Q\) . ilex plastomes could be ancient haplotypes retained in this species. However, previous analyses with extensive sampling spanning the geographic distribution of \(Q\) . ilex placed this species within a clade formed by Eurasian oaks and genera Castanea and Castanopsis based on plastid data (Simeone et al. 2016). Therefore, we concluded that the two \(Q\) . ilex plastomes in question are most likely artifacts, although more sampling is required to increase our confidence of that assertion.  
+
+676. 'One individual gene trees'? This has been changed to read "One individual gene tree". Line 686.  
+
+716. significantly This change has been made. Line 726.  
+
+722. Suggest modify to 'time that gene flow' This change has been made. Line 732.  
+
+723. Suggest modify to 'time that gene flow' This change has been made. Line 733.  
+
+732. Change to 'block' This change has been made. Line 742.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1: Remarks to the Author: The manuscript has further improved after addressing the reviewers comments and I recommend accepting it for publication.  
+
+Reviewer #2: Remarks to the Author:  
+
+This revision by Zhou et al. does an excellent job of revising and responding to the reviewer comments (including my own). I am very satisfied with how they responding to my comments regarding different aspects of the text, and also appreciate the thorough additional dating analyses conducted in responses to my suggestion. I have one comment/critique related to the revised sentence starting on Line 72 (described below), that the authors might consider for a slight revision. But otherwise I think the manuscript is in excellent shape and now suitable for publication in Nature Communications.  
+
+Line 72 onward. I appreciate the reply to my original comment. But I think the sentence still reads a bit awkwardly and I think it still has issues. The information presented here does not itself suggest anything about rapid morphological change. Why not say more specifically that although the family/stem is ca. 80 million years old, divergence of crown groups/the HS clade appears to have occurred rapidly in the early Cenozoic, suggesting rapid morphological evolution?  
+
+I understand you position and agree that both evidence and expectations suggest generally slower evolutionary rates in long- lived lineages such as trees. But the 30- million- year window you present in the sentence is not a particularly narrow window of time, even for lineages we would typically expect to be slowly evolving. But it seems that you could revise to more clearly and accurately support your position—the diversification of the specific clade(s) in question occurred in a narrower window.  
+
+In this context, you might also cite this recent paper suggesting a general link between phylogenomic conflict (as a signature of rapid diversification) and rapid phenotypic evolution, which might be relevant in your case:  
+
+Parins- Fukuchi, Caroline, Gregory W. Stull, and Stephen A. Smith. "Phylogenomic conflict coincides with rapid morphological innovation." Proceedings of the National Academy of Sciences 118, no. 19 (2021).
+
+<--- Page Split --->
+
+## Response to Reviewer #1  
+
+Reviewer #1 (Remarks to the Author): The manuscript has further improved after addressing the reviewers comments and I recommend accepting it for publication. ---Many thanks for your insightful comments on a previous draft.  
+
+## Response to Reviewer #2  
+
+Reviewer #2 (Remarks to the Author):  
+
+This revision by Zhou et al. does an excellent job of revising and responding to the reviewer comments (including my own). I am very satisfied with how they responding to my comments regarding different aspects of the text, and also appreciate the thorough additional dating analyses conducted in responses to my suggestion. I have one comment/critique related to the revised sentence starting on Line 72 (described below), that the authors might consider for a slight revision. But otherwise I think the manuscript is in excellent shape and now suitable for publication in Nature Communications.  
+
+Line 72 onward. I appreciate the reply to my original comment. But I think the sentence still reads a bit awkwardly and I think it still has issues. The information presented here does not itself suggest anything about rapid morphological change. Why not say more specifically that although the family/stem is ca. 80 million years old, divergence of crown groups/the HS clade appears to have occurred rapidly in the early Cenozoic, suggesting rapid morphological evolution?  
+
+I understand you position and agree that both evidence and expectations suggest generally slower evolutionary rates in long- lived lineages such as trees. But the 30- million- year window you present in the sentence is not a particularly narrow window of time, even for lineages we would typically expect to be slowly evolving. But it seems that you could revise to more clearly and accurately support your position—the diversification of the specific clade(s) in question occurred in a narrower window.  
+
+In this context, you might also cite this recent paper suggesting a general link between phylogenomic conflict (as a signature of rapid diversification) and rapid phenotypic evolution, which might be relevant in your case:  
+
+Parins- Fukuchi, Caroline, Gregory W. Stull, and Stephen A. Smith. "Phylogenomic conflict coincides with rapid morphological innovation." Proceedings of the National Academy of Sciences 118, no. 19 (2021).  
+
+---Thank you for raising this point. We have revised the statement as follows: "While a minimum divergence age of ca. 80 million years ago (Ma) is estimated for the family, divergence of crown groups appears to have occurred rapidly in the early Cenozoic, suggesting the potential for rapid morphological change in forest tree species \(^{44- 47}\) ". The suggested reference has been cited. Lines 78- 81 and 945- 947.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6b1a321937aeb718b93604db1067eb406c84e4efe4fff5fa0e724a8088d0b6e/supplementary_0_Peer Review File__d6b1a321937aeb718b93604db1067eb406c84e4efe4fff5fa0e724a8088d0b6e_det.mmd

@@ -0,0 +1,585 @@
+<|ref|>title<|/ref|><|det|>[[100, 40, 508, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[107, 110, 373, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[107, 154, 883, 241]]<|/det|>
+Phylogenomic analyses highlight innovation and introgression in the continental radiations of Fagaceae across the Northern Hemisphere  
+
+<|ref|>image<|/ref|><|det|>[[94, 732, 262, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[271, 732, 880, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[116, 89, 286, 102]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[116, 119, 216, 133]]<|/det|>
+Reviewer #1:  
+
+<|ref|>text<|/ref|><|det|>[[116, 135, 291, 148]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 149, 875, 270]]<|/det|>
+This exceptionally well- written manuscript deals with a number of fundamental evolutionary questions in a novel way. Specifically, Biao- Feng Zhou, Shuai Yuan, Andrew Crowl and co- workers demonstrate using an impressive amount of data how unrelated chloroplast and nuclear genomes both contribute to a holistic understanding of the evolutionary history of a major north temperate group of woody angiosperms, the Fagaceae family. Using time calibrated phylogenies for both genomes and the conflicting signal in these genomes, the authors show that chloroplast genomes encapsulate the shared biogeographic histories of (partly) distantly related groups, whereas the nuclear genome reflects species and higher taxa relationships.  
+
+<|ref|>text<|/ref|><|det|>[[115, 269, 865, 388]]<|/det|>
+They further show that clade ages from dated phylogenies and diversification rates correlate with major biological innovations possibly triggered by global tectonic- climatic changes. For example, the shift to wind- pollination in oaks, whereas chestnuts and other castaneoids are mainly insect- pollinated, or the shift to hypogeous germination in the HS clade that is often correlated with larger seeds, facilitated major radiations of members of the HS clade under globally cooling conditions. Methodology and analytical approaches are all cutting- edge and the figures accompanying the manuscript are very informative. Figure 2 will most likely be reproduced in text books as it impressively shows how different the chloroplast and nuclear genomes function.  
+
+<|ref|>text<|/ref|><|det|>[[116, 403, 357, 417]]<|/det|>
+I have only very few comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 418, 872, 507]]<|/det|>
+Line 71: Although it is impossible to cite all literature relevant to the evolutionary history of Fagaceae main groups, I think two papers should be cited here: Sadowski, E.- M., J. U. Hammel, and T. Denk. 2018. Synchrotron X- ray imaging of a dichasium cupule of Castanopsis from Eocene Baltic amber. American Journal of Botany 105: 2025- 2036. Naryshkina, N. N., and T. A. Evstigneeva. 2020. Fagaceae in the Eocene palynoflora of the South of Primorskii Region: New data on taxonomy and morphology. Paleontological Journal 54: 429- 439.  
+
+<|ref|>text<|/ref|><|det|>[[115, 521, 844, 552]]<|/det|>
+Both papers report Eocene findings of Cyclobalanopsis from the Northern Hemisphere and partly contradict a hypothesis put forward in another paper by Wilf et al. 2019 (cited in the manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[116, 566, 440, 581]]<|/det|>
+Line 117: I would say fruits instead of fruit.  
+
+<|ref|>text<|/ref|><|det|>[[115, 596, 824, 627]]<|/det|>
+Line 329: How would you then explain the presence of sect. Quercus pollen in Eocene strata of Hainan, S China? I think the NW origin may be ok, but I was wondering if you have any ideas.  
+
+<|ref|>text<|/ref|><|det|>[[115, 641, 594, 656]]<|/det|>
+Line 335: Northern Hemisphere instead of northern hemisphere.  
+
+<|ref|>text<|/ref|><|det|>[[116, 702, 216, 715]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[116, 717, 291, 730]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 731, 877, 833]]<|/det|>
+This study by Zhou et al. is a multifaceted examination of the evolutionary history of Fagaceae aimed at understanding the extent and consequences of hybridization during the early diversification of the family and the oak genus (Quercus). Overall, I think this is a high- quality study and a strong candidate for publication in Nature Communications. The dataset is impressive, the analyses conducted were extensive and rigorous, and the presentation generally does a nice job of synthesizing the results in an interesting and compelling fashion. The economic and ecological importance of the family and particularly the oak genus should also make the study of interest to a broad audience.  
+
+<|ref|>text<|/ref|><|det|>[[115, 848, 570, 863]]<|/det|>
+Below I outline my major and minor suggestions for revision.  
+
+<|ref|>text<|/ref|><|det|>[[116, 877, 259, 892]]<|/det|>
+MAJOR COMMENTS
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 878, 164]]<|/det|>
+1. At times I think the manuscript is over-written. And by this, I mean (a) the importance of various attributes is exaggerated or based on speculation and (b) attempts at making grandiose statements often results in convoluted, vague, or sometimes not entirely coherent sentences. Below are some examples and comments. I strongly encourage the authors to revise such passages so that overall the presentation is more modest, concrete, and readable.  
+
+<|ref|>text<|/ref|><|det|>[[115, 179, 867, 239]]<|/det|>
+"Innovations related to seed and pollen dispersal are implicated in triggering waves of continental radiations, while fungal symbioses fortified a competitive edge underground" This is highly speculative especially since the paper does not include any analyses of trait- dependent diversification.  
+
+<|ref|>text<|/ref|><|det|>[[115, 253, 855, 297]]<|/det|>
+"This resulted in adaptive introgression, further amplifying global proliferation" Some analyses were done that suggest positive selection/maintenance of introgression haplotypes, but the leap to "further amplifying global proliferation" seems to much  
+
+<|ref|>text<|/ref|><|det|>[[115, 312, 864, 355]]<|/det|>
+"Northern Hemisphere forests and shrublands are now dominated by species comprising temperate and subtropical lineages, marking one of the greatest floristic transitions in the vegetation history of the Cenozoic"  
+
+<|ref|>text<|/ref|><|det|>[[115, 356, 860, 400]]<|/det|>
+NH forests also included many representatives of 'temperate/subtropical' lineages in the early Cenozoic. Undoubtedly there was a major floristic shift after the Eocene, but this sentence reads as vague and empty to me.  
+
+<|ref|>text<|/ref|><|det|>[[115, 415, 881, 459]]<|/det|>
+"With a minimum age of ca. 80 million years ago (Ma) and precise aging of new fossilized pollen and macrofossils assigned to some modern groups by 50 Ma, the evolution of major lineages appears to be unusually rapid for forest tree species"  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 866, 490]]<|/det|>
+One, this doesn't seem that rapid given the ages provided, and two, there are plenty of examples of rapidly radiated woody groups.  
+
+<|ref|>text<|/ref|><|det|>[[115, 505, 830, 520]]<|/det|>
+The sentence at lines 83- 86 seems overly convoluted, as does the sentence from lines 89 to 94.  
+
+<|ref|>text<|/ref|><|det|>[[115, 536, 563, 550]]<|/det|>
+102: "...the first complete family- wide phylogenetic context"  
+
+<|ref|>text<|/ref|><|det|>[[115, 551, 881, 609]]<|/det|>
+What does 'complete' mean? This study includes a small fraction of the species diversity. Is it the first nuclear phylogenomic study to include all genera? All genera and all sections of oaks? This seems like another example of trying to stretch the significance when it would be better to just be straightforward and clear.  
+
+<|ref|>text<|/ref|><|det|>[[115, 624, 870, 668]]<|/det|>
+191 onwards. "Shift to wind- pollination alone did not increase the diversification rate of oak species immediately, but instead served as a predisposed neutral change that later facilitated rapid radiation of this genus during the expansion of seasonal climates (Fig. 1)."  
+
+<|ref|>text<|/ref|><|det|>[[115, 669, 870, 713]]<|/det|>
+This is speculative. Fine to suggest this as a hypothesis but it is unreasonable to declare it so without any sort of direct tests or evidence. I also don't think the next sentence follows very clearly. I think this could be reframed as a hypothesis with compelling contextual or anecdotal evidence.  
+
+<|ref|>text<|/ref|><|det|>[[115, 729, 695, 744]]<|/det|>
+231- 232: "However, gene flow between modern genera is without precedent"  
+
+<|ref|>text<|/ref|><|det|>[[115, 744, 872, 803]]<|/det|>
+As you noted previously, phylogenetic studies based on the plastome have suggested gene flow between Quercus and related genera (given the geographic structure of the plastid tree). So, I don't think this statement is accurate. But a comprehensive investigation of this using both the nuclear and plastid genomes is novel.  
+
+<|ref|>text<|/ref|><|det|>[[115, 818, 872, 893]]<|/det|>
+2. I have a few concerns about the dating analyses. One, dating analyses based on a large number of genes can introduce a lot of heterogeneity into analyses that is hard to accommodate. It is now becoming more common to filter your genes for those that (a) match the species-tree topology and (b) are more 'clock-like' (less root-to-tip variation). Since you have such a large number of loci, this time of approach could be beneficial. I understand that many downstream analyses are based on the
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 88, 870, 222]]<|/det|>
+dated tree, and so asking to have the dating analyses re- done is a big ask. But this alternative could be tried and compared with the original results to see if there are major differences. Concerning the dated chloroplast tree, you might clarify that the 'ML tree' used as the reference was the CP ML tree (not nuclear). Upon first reading I thought you meant you constrained the plastid analysis to the nuclear topology when dating (which would be problematic) but looking at the Suppl I realized this was not the case. So I would just clarify. I also wonder if it would be worthwhile to do a nuclear dating analysis with only the two calibrations used for the CP analysis. Using different calibration schemes for these datasets seems like it would result in a biased comparison between the two.  
+
+<|ref|>sub_title<|/ref|><|det|>[[115, 253, 260, 267]]<|/det|>
+## MINOR COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 268, 646, 283]]<|/det|>
+57: "often repeated seasonal biomes" what does often repeated mean?  
+
+<|ref|>text<|/ref|><|det|>[[115, 298, 857, 343]]<|/det|>
+69 onward: "Fossil analogs" is perhaps not the best way to put this since 'analog' often refers to independently evolved structures. I would just say "Fossils of Fagaceae are well represented in the Northern Hemisphere..."  
+
+<|ref|>text<|/ref|><|det|>[[115, 386, 869, 446]]<|/det|>
+106, 553 (and perhaps elsewhere): "coalescent analyses using ASTRAL- III and SVDquartets" Both ASTRAL and SVDquartets are summary methods—they do not model the coalescent, or in other words, they are not based on the coalescent—and so calling them "coalescent- based" approaches is inaccurate  
+
+<|ref|>text<|/ref|><|det|>[[116, 462, 444, 477]]<|/det|>
+107 change to "for all but a few branches"?  
+
+<|ref|>text<|/ref|><|det|>[[115, 491, 872, 581]]<|/det|>
+110: "early- diverging lineages" There are many variations on this throughout the manuscript ('early diverging', 'early branching'). Two branches from a node are equal in age so it doesn't make sense to call one lineage "earlier" with respect to the other(s). This is basically the same issue as calling something "basal". I would encourage the authors to revise this language (change to things like "successively sister" etc). This sentence (109 onward) is also generally convoluted and would benefit from revision.  
+
+<|ref|>text<|/ref|><|det|>[[116, 596, 415, 611]]<|/det|>
+117 should be "a' single rounded fruit"?  
+
+<|ref|>text<|/ref|><|det|>[[115, 625, 839, 655]]<|/det|>
+119: "studies based on sequences derived from RAD- seq datasets and nuclear loci" – this syntax doesn't really make sense  
+
+<|ref|>text<|/ref|><|det|>[[115, 685, 880, 745]]<|/det|>
+120- 122: "Despite phylogenetic congruence across methods, high levels of gene- tree conflict within the nuclear genome were observed, likely due to incomplete lineage sorting (ILS; Supplementary Figs. 2, 3 and 4)." This seems a bit hand- wavy given that the goal of the paper is to integrate deep conflicts to understand ancient hybridization  
+
+<|ref|>text<|/ref|><|det|>[[116, 760, 450, 775]]<|/det|>
+126: early diverging used again here—revise  
+
+<|ref|>text<|/ref|><|det|>[[115, 789, 819, 805]]<|/det|>
+134- 135: "These events closely follow the Cretaceous- Paleogene (K- Pg) boundary dated at 66  
+
+<|ref|>text<|/ref|><|det|>[[115, 819, 763, 835]]<|/det|>
+Ma" – 15 million years seems like a broad window to say this "closely followed" the KPg  
+
+<|ref|>text<|/ref|><|det|>[[115, 849, 820, 879]]<|/det|>
+Lines around 156: Do you think any seed traits in this clade might have also conferred greater survivorship following the ecological devastation of the KPg?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 856, 120]]<|/det|>
+175- 178: A little confusing whether fagaceae or fungi is being referred to with respect to increased speciation  
+
+<|ref|>text<|/ref|><|det|>[[116, 133, 463, 149]]<|/det|>
+184: 'coincident' might be a better word here?  
+
+<|ref|>text<|/ref|><|det|>[[115, 163, 865, 194]]<|/det|>
+291 to 294: Does this also require information on population sizes and, if so, how did you determine those?  
+
+<|ref|>text<|/ref|><|det|>[[116, 208, 866, 253]]<|/det|>
+323: I don't know if it makes sense for the genomes to be "contributing unique inferences" since the researcher is doing the inferring, "...each contributing unique insights on the complex combination of divergent..." instead?  
+
+<|ref|>text<|/ref|><|det|>[[116, 267, 470, 282]]<|/det|>
+333: "this level of scrutiny" to "our detection"?  
+
+<|ref|>text<|/ref|><|det|>[[116, 297, 622, 313]]<|/det|>
+335- 336: This paragraph ender seems a little overblown and vague.  
+
+<|ref|>text<|/ref|><|det|>[[116, 327, 430, 342]]<|/det|>
+419: remove 's' after 'SCG': 'SGC regions'  
+
+<|ref|>text<|/ref|><|det|>[[116, 387, 216, 400]]<|/det|>
+Reviewer #3:  
+
+<|ref|>text<|/ref|><|det|>[[116, 402, 291, 416]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 417, 872, 582]]<|/det|>
+This is an important and exceptionally well- written paper which describes the evolutionary history of the oak family Fagaceae (comprising oak, chestnut, stone oak and beech species) that dominates temperate forests in the northern hemisphere. The impressive, detailed analysis of plastomes and nuclear genomes of representative species across the family builds on previous results obtained for Quercus by Manos and others. The robust dated nuclear phylogeny generated tells us when lineages (genera, sub- genera, species) originated and how this ties into ecological innovations (hypogeal from epigeal seed germination; wind from insect pollination) and associations (with seed dispersers and mycorrhiza). Most importantly, a comparison of plastome with nuclear phylogenies plus in- depth analysis of gene flow between species and higher lineages provides information on chloroplast capture, historical phylogeography, and adaptive introgression in the family. In conclusion, the authors emphasise the importance of historical introgression to the success of the family.  
+
+<|ref|>text<|/ref|><|det|>[[115, 595, 880, 686]]<|/det|>
+It is pointed out in the text that many closely related oak species occur in sympatry and hybridise. Consequently, I did wonder how confident we might be that all samples analysed represented the species they were said to be. For example, Q. robur and Q. petraea are two such species that hybridise to the point that it is very difficult to find plants in the wild that are not hybrids. Are the authors confident that the samples used for these species, and of other species that similarly hybridise extensively, were in fact non- hybrid. If not, how might this affect their findings?  
+
+<|ref|>text<|/ref|><|det|>[[116, 701, 208, 715]]<|/det|>
+Minor points  
+
+<|ref|>text<|/ref|><|det|>[[115, 730, 866, 760]]<|/det|>
+lines 497- 504. Peculiar. Is there a possible biological reason for the occurrence of these divergent Q. ilex plastomes or are they likely artifacts.  
+
+<|ref|>text<|/ref|><|det|>[[116, 760, 366, 774]]<|/det|>
+676. 'One individual gene trees'?  
+
+<|ref|>text<|/ref|><|det|>[[116, 775, 244, 789]]<|/det|>
+716. significantly  
+
+<|ref|>text<|/ref|><|det|>[[116, 790, 450, 805]]<|/det|>
+722. Suggest modify to 'time that gene flow'  
+
+<|ref|>text<|/ref|><|det|>[[116, 806, 450, 820]]<|/det|>
+723. Suggest modify to 'time that gene flow'  
+
+<|ref|>text<|/ref|><|det|>[[116, 821, 283, 834]]<|/det|>
+732. Change to 'block'
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[146, 90, 353, 108]]<|/det|>
+## Response to Reviewer #1  
+
+<|ref|>text<|/ref|><|det|>[[145, 124, 856, 282]]<|/det|>
+This exceptionally well- written manuscript deals with a number of fundamental evolutionary questions in a novel way. Specifically, Biao- Feng Zhou, Shuai Yuan, Andrew Crowl and co- workers demonstrate using an impressive amount of data how unrelated chloroplast and nuclear genomes both contribute to a holistic understanding of the evolutionary history of a major north temperate group of woody angiosperms, the Fagaceae family. Using time calibrated phylogenies for both genomes and the conflicting signal in these genomes, the authors show that chloroplast genomes encapsulate the shared biogeographic histories of (partly) distantly related groups, whereas the nuclear genome reflects species and higher taxa relationships.  
+
+<|ref|>text<|/ref|><|det|>[[145, 282, 855, 386]]<|/det|>
+They further show that clade ages from dated phylogenies and diversification rates correlate with major biological innovations possibly triggered by global tectonic- climatic changes. For example, the shift to wind- pollination in oaks, whereas chestnuts and other castaneoids are mainly insect- pollinated, or the shift to hypogeous germination in the HS clade that is often correlated with larger seeds, facilitated major radiations of members of the HS clade under globally cooling conditions.  
+
+<|ref|>text<|/ref|><|det|>[[145, 386, 855, 456]]<|/det|>
+Methodology and analytical approaches are all cutting- edge and the figures accompanying the manuscript are very informative. Figure 2 will most likely be reproduced in text books as it impressively shows how different the chloroplast and nuclear genomes function.  
+
+<|ref|>text<|/ref|><|det|>[[145, 474, 395, 491]]<|/det|>
+I have only very few comments:  
+
+<|ref|>text<|/ref|><|det|>[[145, 491, 855, 526]]<|/det|>
+Line 71: Although it is impossible to cite all literature relevant to the evolutionary history of Fagaceae main groups, I think two papers should be cited here:  
+
+<|ref|>text<|/ref|><|det|>[[144, 530, 856, 582]]<|/det|>
+Sadowski, E.- M., J. U. Hammel, and T. Denk. 2018. Synchrotron \(X\) - ray imaging of a dichasium cupule of Castanopsis from Eocene Baltic amber. American Journal of Botany 105: 2025- 2036.  
+
+<|ref|>text<|/ref|><|det|>[[144, 584, 856, 636]]<|/det|>
+Naryshkina, N. N., and T. A. Evstigneeva. 2020. Fagaceae in the Eocene palynoflora of the South of Primorskii Region: New data on taxonomy and morphology. Paleontological Journal 54: 429- 439.  
+
+<|ref|>text<|/ref|><|det|>[[144, 637, 856, 689]]<|/det|>
+Both papers report Eocene findings of Cyclobalanopsis from the Northern Hemisphere and partly contradict a hypothesis put forward in another paper by Wilf et al. 2019 (cited in the manuscript).  
+
+<|ref|>text<|/ref|><|det|>[[145, 690, 855, 741]]<|/det|>
+---Thank you. We have added these two literatures, although addressing the point on the Wilf et al. (2019) paper would require a separate review on the biogeographic history of Fagaceae. Lines 70, 874- 879.  
+
+<|ref|>text<|/ref|><|det|>[[145, 759, 490, 776]]<|/det|>
+Line 117: I would say fruits instead of fruit.  
+
+<|ref|>text<|/ref|><|det|>[[145, 777, 680, 794]]<|/det|>
+---We have changed this to read "a single rounded fruit". Line 117.  
+
+<|ref|>text<|/ref|><|det|>[[145, 811, 855, 863]]<|/det|>
+Line 329: How would you then explain the presence of sect. Quercus pollen in Eocene strata of Hainan, S China? I think the NW origin may be ok, but I was wondering if you have any ideas.  
+
+<|ref|>text<|/ref|><|det|>[[144, 864, 855, 899]]<|/det|>
+---Interesting challenge. We made some minor changes to the text to reflect the potential for an early presence of section Quercus at tropical latitudes in Asia based on
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 90, 854, 160]]<|/det|>
+studies of Hofmann et al data (2019) from Eocene strata in Hainan. Our divergence time estimates based on nuclear data for NW/OW white oak split is roughly as early as 26 Ma (mid- Oligocene), so a late Eocene presence of section Quercus is reasonable, except for the surprisingly tropical extension of its distribution in Hainan. Lines 325, 366- 368.  
+
+<|ref|>text<|/ref|><|det|>[[145, 161, 854, 317]]<|/det|>
+The earlier presence of section could be explained by broadly distributed ancestral section Quercus or Pacific lineage that spanned the OW/NW through Beringia at a time when northwestern North America and northeastern Asia were more or less continuous areas with a shared flora. A similar pattern of expansion and contraction has been invoked to explain the fossil history of Fagus, which is now extinct from northwestern North America, but extant in east Asia (Denk & Grimm 2009). The paleo distribution of Fagaceae in and around the high latitudes associated with the Beringian Land Bridge and into more southward amphi- Pacific areas is clearly at odds with the modern distribution of taxa.  
+
+<|ref|>text<|/ref|><|det|>[[145, 318, 854, 440]]<|/det|>
+As Hainan became progressively more tropical (extant Fagaceae include Lithocarpus, Castanopsis, Trigonobalanus, sections Cyclobalanopsis, Cerris, and Ilex), subtropical and temperate Fagaceae like section Quercus may have gone extinct (Zhu 2016). A second expansion of section Quercus derived from a NW temperate ancestor shared with sect Quercus subsect Albae could explain a more recent dispersion of Eurasian white oaks across Eurasia during the expansion of seasonal climates in the Oligocene.  
+
+<|ref|>text<|/ref|><|det|>[[145, 456, 840, 545]]<|/det|>
+Hofmann CC, Kodrul TM, Liu XY, Jin JH. 2019. Scanning electron microscopy investigations of middle to late Eocene pollen from the Changchang Basin (Hainan island, south China) – Insights into the paleobiogeography and fossil history of Juglans, Fagus, Lagerstroemia, Mortoniodendron, Cornus, Nyssa, Symplocos and some Icacinaceae in SE Asia. Review of Palaeobotany and Palynology 265: 41–61.  
+
+<|ref|>text<|/ref|><|det|>[[145, 561, 792, 596]]<|/det|>
+Denk T, Grimm GW. 2009. The biogeographic history of beech trees. Review of Palaeobotany and Palynology 158: 83–100.  
+
+<|ref|>text<|/ref|><|det|>[[145, 613, 805, 648]]<|/det|>
+Zhu H. 2016. Biogeographical evidences help revealing the origin of Hainan Island. PLoS ONE 11: e0151941.  
+
+<|ref|>text<|/ref|><|det|>[[145, 666, 654, 701]]<|/det|>
+Line 335: Northern Hemisphere instead of northern hemisphere. This change has been made. Line 334.  
+
+<|ref|>text<|/ref|><|det|>[[146, 718, 450, 735]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[145, 753, 856, 892]]<|/det|>
+This study by Zhou et al. is a multifaceted examination of the evolutionary history of Fagaceae aimed at understanding the extent and consequences of hybridization during the early diversification of the family and the oak genus (Quercus). Overall, I think this is a high- quality study and a strong candidate for publication in Nature Communications. The dataset is impressive, the analyses conducted were extensive and rigorous, and the presentation generally does a nice job of synthesizing the results in an interesting and compelling fashion. The economic and ecological importance of the family and particularly the oak genus should also make the study of interest to a broad audience.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 108, 627, 125]]<|/det|>
+Below I outline my major and minor suggestions for revision.  
+
+<|ref|>sub_title<|/ref|><|det|>[[146, 142, 323, 159]]<|/det|>
+## MAJOR COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[145, 160, 855, 265]]<|/det|>
+I. At times I think the manuscript is over-written. And by this, I mean (a) the importance of various attributes is exaggerated or based on speculation and (b) attempts at making grandiose statements often results in convoluted, vague, or sometimes not entirely coherent sentences. Below are some examples and comments. I strongly encourage the authors to revise such passages so that overall the presentation is more modest, concrete, and readable.  
+
+<|ref|>text<|/ref|><|det|>[[145, 282, 856, 352]]<|/det|>
+"Innovations related to seed and pollen dispersal are implicated in triggering waves of continental radiations, while fungal symbioses fortified a competitive edge underground" This is highly speculative especially since the paper does not include any analyses of trait-dependent diversification.  
+
+<|ref|>text<|/ref|><|det|>[[145, 352, 855, 404]]<|/det|>
+---We have edited this statement to focus only on seed dispersal, which was found to be associated with a spike in diversification in our study. We removed mention of pollen dispersal and fungal symbioses. Lines 41- 44.  
+
+<|ref|>text<|/ref|><|det|>[[145, 420, 856, 560]]<|/det|>
+"This resulted in adaptive introgression, further amplifying global proliferation" Some analyses were done that suggest positive selection/maintenance of introgression haplotypes, but the leap to "further amplifying global proliferation" seems to much. ---We have removed this problematic phrase and have rewritten this sentence to focus specifically on the Eurasian white oak clade. Diversification analyses identified this clade as having an increase in net diversification. We feel the link between positive selection and increased diversification in this clade is sufficient to make such a statement. Lines 48- 49.  
+
+<|ref|>text<|/ref|><|det|>[[145, 577, 855, 630]]<|/det|>
+"Northern Hemisphere forests and shrublands are now dominated by species comprising temperate and subtropical lineages, marking one of the greatest floristic transitions in the vegetation history of the Cenozoic"  
+
+<|ref|>text<|/ref|><|det|>[[145, 630, 855, 682]]<|/det|>
+NH forests also included many representatives of 'temperate/subtropical' lineages in the early Cenozoic. Undoubtedly there was a major floristic shift after the Eocene, but this sentence reads as vague and empty to me.  
+
+<|ref|>text<|/ref|><|det|>[[144, 682, 852, 716]]<|/det|>
+---We have toned this statement down to read "one of the major floristic transitions". Line 53.  
+
+<|ref|>text<|/ref|><|det|>[[144, 734, 856, 805]]<|/det|>
+"With a minimum age of ca. 80 million years ago (Ma) and precise aging of new fossilized pollen and macrofossils assigned to some modern groups by 50 Ma, the evolution of major lineages appears to be unusually rapid for forest tree species" One, this doesn't seem that rapid given the ages provided, and two, there are plenty of examples of rapidly radiated woody groups.  
+
+<|ref|>text<|/ref|><|det|>[[145, 825, 855, 892]]<|/det|>
+---We appreciate this comment and have revised the text to instead focus on the timing of the diversification of the HS clade of Fagaceae, which we found to be associated with an increase in net diversification rate (Lines 72- 76). Fossil taxa ascribed to those genera fall within a much narrower window of time. We are using the rich fossil record of
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 90, 854, 177]]<|/det|>
+Fagaceae to suggest that macroevolutionary change (pollen, cupule/fruit) was rapid in deep time, in contrast to the general findings that woody lineages evolve more slowly than herbaceous lineages using relative rate analysis across sister clades. Life history correlates such as generation time also have figured into this broader question (see Smith & Beaulieu 2009)  
+
+<|ref|>text<|/ref|><|det|>[[145, 178, 854, 370]]<|/det|>
+This is an interesting topic that intersects with ideas on the role of life history, climate niche evolution, and correlates with net diversification rates. Following the publication of Smith and Donoghue (2008, cited here) and Smith & Beaulieu (2009), some literature suggests that secondary woodiness is a prerequisite to higher speciation rates in isolated niche space, such as islands and continental sky islands. While there is some evidence for rapid diversification in woody clades such as temperate Quercus (Hipp et al. 2020, cited here and detected in this study as well) and tropical Inga (Fabaceae, Richardson et al. 2001), slower rates of evolution as a function life history, limited sequence divergence or net diversification rates remain associated with tree lineages. More recent analyses suggest that plant size, another proxy for arborescence, also is correlated with lower diversification rates across the angiosperms (Boucher et al. 2017).  
+
+<|ref|>text<|/ref|><|det|>[[145, 385, 839, 421]]<|/det|>
+Boucher FC, Verboom GA, Musker S, Ellis AG (2017) Plant size: a key determinant of diversification? New Phytologist 216: 24- 31.  
+
+<|ref|>text<|/ref|><|det|>[[145, 437, 816, 474]]<|/det|>
+Nürk NM, Atchison GW, Hughes CE (2019) Island woodiness underpins accelerated disparification in plant radiations. New Phytologist 224: 518- 531.  
+
+<|ref|>text<|/ref|><|det|>[[145, 490, 852, 543]]<|/det|>
+Richardson JE, Pennington RT, Pennington TD, Hollingsworth PM (2001) Rapid diversification of a species- rich genus of Neotropical rain forest trees. Science 293: 2242- 2245  
+
+<|ref|>text<|/ref|><|det|>[[145, 560, 794, 595]]<|/det|>
+Smith SA and Beaulieu JM (2009) Life- history influences rates of climatic niche evolution in flowering plants. Proceedings of the Royal Society B 276: 4345- 4352  
+
+<|ref|>text<|/ref|><|det|>[[145, 629, 854, 664]]<|/det|>
+The sentence at lines 83- 86 seems overly convoluted, as does the sentence from lines 89 to 94.  
+
+<|ref|>text<|/ref|><|det|>[[145, 665, 854, 700]]<|/det|>
+---We have attempted to clarify and reduce the length of these sentences. Lines 84- 86, 92.  
+
+<|ref|>text<|/ref|><|det|>[[145, 717, 855, 820]]<|/det|>
+102: "...the first complete family- wide phylogenetic context" What does 'complete' mean? This study includes a small fraction of the species diversity. Is it the first nuclear phylogenomic study to include all genera? All genera and all sections of oaks? This seems like another example of trying to stretch the significance when it would be better to just be straightforward and clear. ---We have changed this to simply read "within a broad phylogenetic context." Line 102.  
+
+<|ref|>text<|/ref|><|det|>[[145, 838, 857, 873]]<|/det|>
+191 onwards. "Shift to wind- pollination alone did not increase the diversification rate of oak species immediately, but instead served as a predisposed neutral change that later
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 90, 853, 125]]<|/det|>
+facilitated rapid radiation of this genus during the expansion of seasonal climates (Fig. 1)."  
+
+<|ref|>text<|/ref|><|det|>[[144, 125, 854, 195]]<|/det|>
+This is speculative. Fine to suggest this as a hypothesis but it is unreasonable to declare it so without any sort of direct tests or evidence. I also don't think the next sentence follows very clearly. I think this could be reframed as a hypothesis with compelling contextual or anecdotal evidence.  
+
+<|ref|>text<|/ref|><|det|>[[144, 195, 728, 213]]<|/det|>
+---We agree and have rephrased this as a hypothesis. Lines 190, 193- 194.  
+
+<|ref|>text<|/ref|><|det|>[[144, 230, 857, 317]]<|/det|>
+231- 232: "However, gene flow between modern genera is without precedent" As you noted previously, phylogenetic studies based on the plastome have suggested gene flow between Quercus and related genera (given the geographic structure of the plastid tree). So, I don't think this statement is accurate. But a comprehensive investigation of this using both the nuclear and plastid genomes is novel.  
+
+<|ref|>text<|/ref|><|det|>[[144, 317, 854, 352]]<|/det|>
+---We have clarified this to read, "inference of gene flow between modern genera has been based solely on plastome data.". Lines 230- 231.  
+
+<|ref|>text<|/ref|><|det|>[[144, 368, 855, 510]]<|/det|>
+2. I have a few concerns about the dating analyses. One, dating analyses based on a large number of genes can introduce a lot of heterogeneity into analyses that is hard to accommodate. It is now becoming more common to filter your genes for those that (a) match the species-tree topology and (b) are more 'clock-like' (less root-to-tip variation). Since you have such a large number of loci, this time of approach could be beneficial. I understand that many downstream analyses are based on the dated tree, and so asking to have the dating analyses re-done is a big ask. But this alternative could be tried and compared with the original results to see if there are major differences.  
+
+<|ref|>text<|/ref|><|det|>[[144, 510, 855, 719]]<|/det|>
+--- Thank you for raising this important point. We have applied two "gene- shopping" methods to identify genes with the best information for dating. First, we used SortaData (Smith et al. 2018) to filter 212 (top \(90^{\text{th}}\) percentile) most clock- like loci by considering clock- like as the primary criterion, followed by tree- like and tree length. Second, we calculated the Robinson- Foulds (RF) distance between gene trees and the reference tree following Johns et al. (2018), and retained 212 loci with the least RF distance and greater concordant phylogenetic signals. By using three different topologies as reference trees (see above), we generated six reduced datasets. The divergence time estimated on the 2124 genes were almost identical to the six reduced data sets (Pearson's correlation coefficient \(= 0.991 - 0.995\) , \(P < 2\mathrm{e}^{- 16}\) ; Supplementary Fig. 15). These new approaches have been described in the revised Methods (Lines 566- 576), a new Supplementary Table 11 and a new Supplementary Fig. S15.  
+
+<|ref|>text<|/ref|><|det|>[[144, 735, 855, 822]]<|/det|>
+Concerning the dated chloroplast tree, you might clarify that the 'ML tree' used as the reference was the CP ML tree (not nuclear). Upon first reading I thought you meant you constrained the plastid analysis to the nuclear topology when dating (which would be problematic) but looking at the Suppl I realized this was not the case. So I would just clarify.  
+
+<|ref|>text<|/ref|><|det|>[[144, 823, 617, 840]]<|/det|>
+--- We have clarified this to "plastome ML tree". Line 578.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 90, 854, 142]]<|/det|>
+I also wonder if it would be worthwhile to do a nuclear dating analysis with only the two calibrations used for the CP analysis. Using different calibration schemes for these datasets seems like it would result in a biased comparison between the two.  
+
+<|ref|>text<|/ref|><|det|>[[145, 143, 854, 179]]<|/det|>
+---We have dated the nuclear tree by using the same two calibrations for plastome tree (Lines 585- 587). This result was presented in updated Supplementary Fig. 5g- i.  
+
+<|ref|>sub_title<|/ref|><|det|>[[145, 195, 322, 211]]<|/det|>
+## MINOR COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[145, 212, 854, 265]]<|/det|>
+57: "often repeated seasonal biomes" what does often repeated mean? ---This was intended to plant the seed that similar biomes present the opportunity for ecological convergence in Fagaceae. The text has been simplified. Lines 56- 57.  
+
+<|ref|>text<|/ref|><|det|>[[145, 282, 854, 334]]<|/det|>
+69 onward: "Fossil analogs" is perhaps not the best way to put this since 'analog' often refers to independently evolved structures. I would just say "Fossils of Fagaceae are well represented in the Northern Hemisphere..."  
+
+<|ref|>text<|/ref|><|det|>[[145, 335, 562, 352]]<|/det|>
+---Good point. This change has been made. Line 69.  
+
+<|ref|>text<|/ref|><|det|>[[145, 369, 855, 456]]<|/det|>
+106, 553 (and perhaps elsewhere): "coalescent analyses using ASTRAL- III and SVDquartets" Both ASTRAL and SVDquartets are summary methods—they do not model the coalescent, or in other words, they are not based on the coalescent—and so calling them "coalescent- based" approaches is inaccurate  
+
+<|ref|>text<|/ref|><|det|>[[145, 440, 855, 475]]<|/det|>
+to read "species- tree analyses" throughout the manuscript. Lines 106, 530 and 563.  
+
+<|ref|>text<|/ref|><|det|>[[145, 491, 507, 508]]<|/det|>
+107 change to "for all but a few branches"?  
+
+<|ref|>text<|/ref|><|det|>[[145, 509, 507, 525]]<|/det|>
+---This sentence has been revised. Lines 107.  
+
+<|ref|>text<|/ref|><|det|>[[145, 543, 855, 647]]<|/det|>
+110: "early- diverging lineages" There are many variations on this throughout the manuscript ('early diverging', 'early branching'). Two branches from a node are equal in age so it doesn't make sense to call one lineage "earlier" with respect to the other(s). This is basically the same issue as calling something "basal". I would encourage the authors to revise this language (change to things like "successively sister" etc). This sentence (109 onward) is also generally convoluted and would benefit from revision.  
+
+<|ref|>text<|/ref|><|det|>[[145, 648, 854, 682]]<|/det|>
+---All mention of "early- diverging lineages" has been removed or replaced with "successively sister lineages". Lines 110, 126, 149, 160 and 204.  
+
+<|ref|>text<|/ref|><|det|>[[145, 700, 480, 716]]<|/det|>
+117 should be "a' single rounded fruit"?  
+
+<|ref|>text<|/ref|><|det|>[[145, 718, 480, 734]]<|/det|>
+---This change has been made. Lines 117.  
+
+<|ref|>text<|/ref|><|det|>[[145, 752, 855, 785]]<|/det|>
+119: "studies based on sequences derived from RAD- seq datasets and nuclear loci" — this syntax doesn't really make sense  
+
+<|ref|>text<|/ref|><|det|>[[145, 787, 854, 820]]<|/det|>
+---This sentence has been amended to read, "...previous studies based on RAD- seq data." Lines 119.  
+
+<|ref|>text<|/ref|><|det|>[[145, 838, 855, 873]]<|/det|>
+120- 122: "Despite phylogenetic congruence across methods, high levels of gene- tree conflict within the nuclear genome were observed, likely due to incomplete lineage
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 90, 856, 230]]<|/det|>
+sorting (ILS; Supplementary Figs. 2, 3 and 4)." This seems a bit hand- wavy given that the goal of the paper is to integrate deep conflicts to understand ancient hybridization This observation refers to the relationships among Quercus, Notholithocarpus, Lithocarpus, and Chrysolepis which we found to be in conflict across nuclear gene trees. Not surprisingly, ILS and hybridization present challenges for estimating the Fagaceae tree of life. For this particular set of relationships and narrow time frame, a stochastic pattern of gene tree resolution is consistent with ILS. These results were detailed in Supplementary Figs. 2, 3 and 4, and briefly referred in main text (lines 119- 122)  
+
+<|ref|>text<|/ref|><|det|>[[145, 247, 503, 263]]<|/det|>
+126: early diverging used again here—revise  
+
+<|ref|>text<|/ref|><|det|>[[145, 265, 501, 281]]<|/det|>
+This sentence has been revised. Line 126.  
+
+<|ref|>text<|/ref|><|det|>[[145, 299, 856, 352]]<|/det|>
+134- 135: "These events closely follow the Cretaceous- Paleogene (K- Pg) boundary dated at 66 Ma" - 15 million years seems like a broad window to say this "closely followed" the KPg  
+
+<|ref|>text<|/ref|><|det|>[[145, 352, 855, 386]]<|/det|>
+closely' has been removed from this sentence to state that these events simply followed the KPg. Line 133.  
+
+<|ref|>text<|/ref|><|det|>[[145, 404, 856, 439]]<|/det|>
+Lines around 156: Do you think any seed traits in this clade might have also conferred greater survivorship following the ecological devastation of the KPg?  
+
+<|ref|>text<|/ref|><|det|>[[144, 439, 856, 683]]<|/det|>
+---Interesting question, and much appreciated. But difficult to add anything concrete on the subject. All seeds produced by Fagaceae are known to have a short life span in the soil. As there is no seed bank, the seeds produced in one season are the only source of recruitment. Recent ecological work suggests that seed mating is the main driver for seedling recruitment in forest trees, highlighting the strong selection pressure of seed predators and episodic nature of successful recruitment. While the shift to hypogeous cotyledons appears to be associated with diversifying potential dispersers, its evolution Bhas been linked to increased survivorship over short time scales. Regarding other traits, anecdotal evidence suggests that selection in fruit/cupule evolution in Castanea + Castanopsis was driven to protect seeds through development whereas enhanced dispersal capacity is a general theme in Lithocarpus and Quercus (see Fig 1 images of spiny cupules vs scaly cupules surrounding or subtending nuts), with secondarily evolved exceptions scattered in both genera to protect the seed, and one notable origin of the acorn fruit type in Castanopsis.  
+
+<|ref|>text<|/ref|><|det|>[[144, 700, 855, 752]]<|/det|>
+175- 178: A little confusing whether fagaceae or fungi is being referred to with respect to increased speciation ---We have clarified that it is species radiations of Fagaceae. Lines 176- 177.  
+
+<|ref|>text<|/ref|><|det|>[[145, 770, 517, 786]]<|/det|>
+184: 'coincident' might be a better word here?  
+
+<|ref|>text<|/ref|><|det|>[[145, 788, 473, 804]]<|/det|>
+---This change has been made. Line 183.  
+
+<|ref|>text<|/ref|><|det|>[[144, 821, 856, 856]]<|/det|>
+291 to 294: Does this also require information on population sizes and, if so, how did you determine those?  
+
+<|ref|>text<|/ref|><|det|>[[144, 857, 856, 891]]<|/det|>
+---In the method developed by Huerta- Sanchez et al. (2014), the probability of maintaining selectively neutral haplotypes of a given length in both oak sections after
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 90, 854, 142]]<|/det|>
+introgression was determined by the recombination rate, the time that gene flow occurred and generation time. The effective population size was not included in the function. See details in Methods (Lines 724- 733).  
+
+<|ref|>text<|/ref|><|det|>[[145, 160, 854, 210]]<|/det|>
+323: I don't know if it makes sense for the genomes to be "contributing unique inferences" since the researcher is doing the inferring. "...each contributing unique insights on the complex combination of divergent..." instead?  
+
+<|ref|>text<|/ref|><|det|>[[145, 211, 541, 228]]<|/det|>
+---Agreed. This change has been made. Line 322.  
+
+<|ref|>text<|/ref|><|det|>[[145, 247, 534, 263]]<|/det|>
+333: "this level of scrutiny" to "our detection"?  
+
+<|ref|>text<|/ref|><|det|>[[145, 265, 516, 281]]<|/det|>
+---This change has been made. Lines 331- 332.  
+
+<|ref|>text<|/ref|><|det|>[[145, 300, 681, 316]]<|/det|>
+335- 336: This paragraph ender seems a little overblown and vague.  
+
+<|ref|>text<|/ref|><|det|>[[145, 318, 854, 352]]<|/det|>
+---We have edited this to simply read, "The ecological implications of these biotic exchanges of keystone lineages await future study." Lines 334- 335.  
+
+<|ref|>text<|/ref|><|det|>[[145, 369, 492, 386]]<|/det|>
+419: remove 's' after 'SCG': 'SGC regions'  
+
+<|ref|>text<|/ref|><|det|>[[145, 388, 475, 404]]<|/det|>
+---This change has been made. Line 417.  
+
+<|ref|>sub_title<|/ref|><|det|>[[145, 421, 468, 439]]<|/det|>
+## Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[144, 456, 854, 683]]<|/det|>
+This is an important and exceptionally well- written paper which describes the evolutionary history of the oak family Fagaceae (comprising oak, chestnut, stone oak and beech species) that dominates temperate forests in the northern hemisphere. The impressive, detailed analysis of plastomes and nuclear genomes of representative species across the family builds on previous results obtained for Quercus by Manos and others. The robust dated nuclear phylogeny generated tells us when lineages (genera, sub- genera, species) originated and how this ties into ecological innovations (hypogeal from epigeal seed germination; wind from insect pollination) and associations (with seed dispersers and mycorrhiza). Most importantly, a comparison of plastome with nuclear phylogenies plus in- depth analysis of gene flow between species and higher lineages provides information on chloroplast capture, historical phylogeography, and adaptive introgression in the family. In conclusion, the authors emphasise the importance of historical introgression to the success of the family.  
+
+<|ref|>text<|/ref|><|det|>[[144, 700, 854, 821]]<|/det|>
+It is pointed out in the text that many closely related oak species occur in sympatry and hybridise. Consequently, I did wonder how confident we might be that all samples analysed represented the species they were said to be. For example, Q. robur and Q. petraea are two such species that hybridise to the point that it is very difficult to find plants in the wild that are not hybrids. Are the authors confident that the samples used for these species, and of other species that similarly hybridise extensively, were in fact non- hybrid. If not, how might this affect their findings?  
+
+<|ref|>text<|/ref|><|det|>[[145, 822, 854, 891]]<|/det|>
+--- Good point. The plant material we used in the study was derived from a combination of collections made from natural populations and cultivated plants grown from wild- collected seed. We have updated Supplementary Table 1 to provide data on our voucher specimens.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[145, 90, 856, 317]]<|/det|>
+Where possible, our joint expertise in Fagaceae taxonomy guided the selection of species, in addition to generally including species that are taxonomically well- understood and easily identified. However, it is true that later generation hybrids may express phenotypes that often converge on one parental species. While we do present evidence of more recent hybridization between several morphologically typical species (see Figure 3A), we believe the impact on our findings at deeper phylogenetic levels is minimal. Our tests basically support the mosaic nature of closely related oak genomes, e.g., sympatric white oaks, initially suggested by analyses using RAD- seq data (Hipp et al. 2020, cited here). And despite gene flow involving the tips of the tree, we detected a high degree of coalescence for the gene tree histories that support the common ancestry of the genera and sections of Quercus. From our analyses, it seems that ILS (Incomplete Lineage Sorting) is the main challenge for estimating the species tree in the crown clade of Fagaceae.  
+
+<|ref|>text<|/ref|><|det|>[[146, 335, 250, 352]]<|/det|>
+Minor points  
+
+<|ref|>text<|/ref|><|det|>[[145, 369, 855, 404]]<|/det|>
+lines 497- 504. Peculiar. Is there a possible biological reason for the occurrence of these divergent \(Q\) . ilex plastomes or are they likely artifacts.  
+
+<|ref|>text<|/ref|><|det|>[[145, 404, 855, 527]]<|/det|>
+One potential biological explanation for the two highly divergent \(Q\) . ilex plastomes could be ancient haplotypes retained in this species. However, previous analyses with extensive sampling spanning the geographic distribution of \(Q\) . ilex placed this species within a clade formed by Eurasian oaks and genera Castanea and Castanopsis based on plastid data (Simeone et al. 2016). Therefore, we concluded that the two \(Q\) . ilex plastomes in question are most likely artifacts, although more sampling is required to increase our confidence of that assertion.  
+
+<|ref|>text<|/ref|><|det|>[[145, 544, 710, 578]]<|/det|>
+676. 'One individual gene trees'? This has been changed to read "One individual gene tree". Line 686.  
+
+<|ref|>text<|/ref|><|det|>[[145, 596, 473, 630]]<|/det|>
+716. significantly This change has been made. Line 726.  
+
+<|ref|>text<|/ref|><|det|>[[145, 648, 492, 682]]<|/det|>
+722. Suggest modify to 'time that gene flow' This change has been made. Line 732.  
+
+<|ref|>text<|/ref|><|det|>[[145, 700, 495, 734]]<|/det|>
+723. Suggest modify to 'time that gene flow' This change has been made. Line 733.  
+
+<|ref|>text<|/ref|><|det|>[[145, 752, 473, 786]]<|/det|>
+732. Change to 'block' This change has been made. Line 742.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 286, 104]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 120, 842, 180]]<|/det|>
+Reviewer #1: Remarks to the Author: The manuscript has further improved after addressing the reviewers comments and I recommend accepting it for publication.  
+
+<|ref|>text<|/ref|><|det|>[[115, 225, 291, 253]]<|/det|>
+Reviewer #2: Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 254, 861, 358]]<|/det|>
+This revision by Zhou et al. does an excellent job of revising and responding to the reviewer comments (including my own). I am very satisfied with how they responding to my comments regarding different aspects of the text, and also appreciate the thorough additional dating analyses conducted in responses to my suggestion. I have one comment/critique related to the revised sentence starting on Line 72 (described below), that the authors might consider for a slight revision. But otherwise I think the manuscript is in excellent shape and now suitable for publication in Nature Communications.  
+
+<|ref|>text<|/ref|><|det|>[[115, 372, 865, 447]]<|/det|>
+Line 72 onward. I appreciate the reply to my original comment. But I think the sentence still reads a bit awkwardly and I think it still has issues. The information presented here does not itself suggest anything about rapid morphological change. Why not say more specifically that although the family/stem is ca. 80 million years old, divergence of crown groups/the HS clade appears to have occurred rapidly in the early Cenozoic, suggesting rapid morphological evolution?  
+
+<|ref|>text<|/ref|><|det|>[[115, 462, 872, 537]]<|/det|>
+I understand you position and agree that both evidence and expectations suggest generally slower evolutionary rates in long- lived lineages such as trees. But the 30- million- year window you present in the sentence is not a particularly narrow window of time, even for lineages we would typically expect to be slowly evolving. But it seems that you could revise to more clearly and accurately support your position—the diversification of the specific clade(s) in question occurred in a narrower window.  
+
+<|ref|>text<|/ref|><|det|>[[115, 551, 872, 596]]<|/det|>
+In this context, you might also cite this recent paper suggesting a general link between phylogenomic conflict (as a signature of rapid diversification) and rapid phenotypic evolution, which might be relevant in your case:  
+
+<|ref|>text<|/ref|><|det|>[[115, 610, 857, 656]]<|/det|>
+Parins- Fukuchi, Caroline, Gregory W. Stull, and Stephen A. Smith. "Phylogenomic conflict coincides with rapid morphological innovation." Proceedings of the National Academy of Sciences 118, no. 19 (2021).
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[145, 91, 372, 108]]<|/det|>
+## Response to Reviewer #1  
+
+<|ref|>text<|/ref|><|det|>[[144, 125, 815, 196]]<|/det|>
+Reviewer #1 (Remarks to the Author): The manuscript has further improved after addressing the reviewers comments and I recommend accepting it for publication. ---Many thanks for your insightful comments on a previous draft.  
+
+<|ref|>sub_title<|/ref|><|det|>[[145, 230, 373, 248]]<|/det|>
+## Response to Reviewer #2  
+
+<|ref|>text<|/ref|><|det|>[[145, 265, 448, 282]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[144, 283, 848, 405]]<|/det|>
+This revision by Zhou et al. does an excellent job of revising and responding to the reviewer comments (including my own). I am very satisfied with how they responding to my comments regarding different aspects of the text, and also appreciate the thorough additional dating analyses conducted in responses to my suggestion. I have one comment/critique related to the revised sentence starting on Line 72 (described below), that the authors might consider for a slight revision. But otherwise I think the manuscript is in excellent shape and now suitable for publication in Nature Communications.  
+
+<|ref|>text<|/ref|><|det|>[[144, 421, 850, 526]]<|/det|>
+Line 72 onward. I appreciate the reply to my original comment. But I think the sentence still reads a bit awkwardly and I think it still has issues. The information presented here does not itself suggest anything about rapid morphological change. Why not say more specifically that although the family/stem is ca. 80 million years old, divergence of crown groups/the HS clade appears to have occurred rapidly in the early Cenozoic, suggesting rapid morphological evolution?  
+
+<|ref|>text<|/ref|><|det|>[[144, 543, 844, 648]]<|/det|>
+I understand you position and agree that both evidence and expectations suggest generally slower evolutionary rates in long- lived lineages such as trees. But the 30- million- year window you present in the sentence is not a particularly narrow window of time, even for lineages we would typically expect to be slowly evolving. But it seems that you could revise to more clearly and accurately support your position—the diversification of the specific clade(s) in question occurred in a narrower window.  
+
+<|ref|>text<|/ref|><|det|>[[144, 665, 827, 717]]<|/det|>
+In this context, you might also cite this recent paper suggesting a general link between phylogenomic conflict (as a signature of rapid diversification) and rapid phenotypic evolution, which might be relevant in your case:  
+
+<|ref|>text<|/ref|><|det|>[[144, 734, 820, 787]]<|/det|>
+Parins- Fukuchi, Caroline, Gregory W. Stull, and Stephen A. Smith. "Phylogenomic conflict coincides with rapid morphological innovation." Proceedings of the National Academy of Sciences 118, no. 19 (2021).  
+
+<|ref|>text<|/ref|><|det|>[[144, 804, 833, 891]]<|/det|>
+---Thank you for raising this point. We have revised the statement as follows: "While a minimum divergence age of ca. 80 million years ago (Ma) is estimated for the family, divergence of crown groups appears to have occurred rapidly in the early Cenozoic, suggesting the potential for rapid morphological change in forest tree species \(^{44- 47}\) ". The suggested reference has been cited. Lines 78- 81 and 945- 947.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d6b8258bbfb33cf543845b32afbb9cd16a9f1a68ee0104bb9f461c53aca7b284/images_list.json

@@ -0,0 +1,10 @@
+[
+  {
+    "type": "image",
+    "img_path": "images/Figure_4.jpg",
+    "caption": "Figure 4. Rain-bearing moisture back-trajectories for dry season (DJFM, top) and wet season (JJAS, bottom) for Cueva de la Puente, analyzed with HYSPLIT (Stein et al., 2015) and HySPLIT (Warner, 2018) software. These trajectories show that the Caribbean Sea is the main source of moisture to CP, and that the vast majority of rain falls in the summer. Calculated with GDAS 0.5-degree meteorological data. See Appendix Figure A1 for trajectories in southern Mexico and Belize.",
+    "footnote": [],
+    "bbox": [],
+    "page_idx": 0
+  }
+]

peer_reviews/supplementary_0_Peer Review File__d71bdbebab246bc381157702d44cee9063a8735d6941e27b5a7c48699b0cf6a8/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d71bdbebab246bc381157702d44cee9063a8735d6941e27b5a7c48699b0cf6a8/supplementary_0_Peer Review File__d71bdbebab246bc381157702d44cee9063a8735d6941e27b5a7c48699b0cf6a8.mmd

@@ -0,0 +1,217 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Selective PROTAC- mediated degradation of SMARCA2 is efficacious in SMARCA4 mutant cancers  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1: Remarks to the Author:  
+
+In the manuscript the authors report a new moderately specific SMARCA2 degrader molecule and show this has efficacy in vitro and in vivo on SMARCA4 mutant cells. Several non- selective inhibitors of the highly homologous SMARCA2/4 proteins exist, but this papers shows the designed PROTAC strategy yields some specificity for SMARCA2. The mechanism that controls this specificity is relatively unexplored, but nevertheless the paper describes a useful molecule that will help investigation of SMARCA4 mutant cancers. Overall, the experiments are well- described, cautiously interpreted, and the conclusions support the main point of the paper.  
+
+I have only minor comments on experimental presentation and request for clarifications. Notably, I am not a chemical synthesis expert and have not evaluated if the information on synthesis of the PROTAC is sufficient.  
+
+1. In figure 3C, I would encourage inclusion of the control cell lines in the main figure. It was difficult to compare G1 cells without seeing what the controls look like.  
+
+2. Figure 4a, and extended data. These plots were difficult to read with the multiple axis. I'd suggest breaking this into independent figures.  
+
+3. In several places in the discussion there are results presented as data not shown. These statements should be removed or the data should be presented. Otherwise, future work cannot accurately build on the results.  
+
+4. I did not see an accession number for the RNAseq data presented in Figure 3d  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+The authors have generated a proteolysis targeting chimera (PROTAC) degrader using a molecule that binds the bromodomains of SMARCA2, SMARCA4 and PBRM1 to the E3 ubiquitin ligase VHL. This molecule (A947) seems to exhibit a potent and, to some extent, selective degradation of SMARCA2. There has been at least another previous publication on the generation of a different PROTAC targeting SMARCA2 for degradation, which is referred to by the authors in their manuscript.  
+
+Therapies targeting SMARCA4- mutant tumors are of great interest, given the relatively high frequency of these alterations in many solid tumors, including lung cancer, among other highly aggressive cancers. Therefore, the work is valuable in terms of clinical purposes.  
+
+## Comments:  
+
+1. - The authors use a panel of lung cancer cell lines that are mutant for SMARCA4 and a control panel of wild-type SMARCA4 cells. It is known that, for still unknown reasons, some SMARCA4- mutant cells show very low or complete absence of SMARCA2 protein, by western-blot. They provide the information on the genetic alterations in SMARCA4 in these cells (extended Table 3) and on the mRNA levels (RNA-seq of the databases) of SMARCA4 and SMARCA2. However, the authors should not rely on the mRNA to determine the presence or absence of SMARCA2 in each of the cell lines. Instead, they should perform a western-blot to determine the total protein levels of SMARCA4 and SMARCA2 of every one of the cell lines included in the study. The results should be included in a main figure.  
+
+2. - To provide more robust data, the authors need to show the reduction in SMARCA2 protein levels with A947 administration and determine if this reduction parallels effects on cell growth inhibition in a wide panel of lung cancer cell lines (among those used to determine the effects of the drug on the growth inhibition cell), not just a single cell line as shown in the present work. For this, it would be appropriate to carry out a western blot of the SMARCA2, SMARCA4 (for SMARCA4- wt cells) and PBRM1 proteins after the administration of the drug.
+
+<--- Page Split --->
+
+3. - Some immunostainings of SMARCA2 and SMARCA4 (or at least of SMARCA2) should be performed in the animal models to more clearly evidenced the reduction in the levels of these proteins, as compared to the controls. It is intriguing that, in their work, the authors detect SMARCA4 protein by wb in most of the SMARCA4-mutant cells, since most truncated SMARCA4 proteins are usually degraded by the cells (non-sense mediated decay mechanism).  
+
+4. - Statistics are lacking in the figure 3, fig 5c, and in ext. data fig 6.  
+
+5. - Is the data in figure 4g from a WB? Please indicate it and show the blot. In fig 5b, shouldn't some SMARCA4wt cells been included for comparison?  
+
+6. - There are only 5 main figures but 10 extended data figures. The authors could move some of the figures from the extended data, into main figures. Extended data 1 and 5a could be added to figure 1. Extended data fig 6b and c could be moved to figure 3. Extended fig. 10 moved to Fig 4.7. - To make the reading easier, please include the name of the cell line used in each of the figures.  
+
+Reviewer #3: Remarks to the Author:  
+
+This is an excellent degrader development story detailing the discovery and detailed characterization of a heterobivalent degrader of SMARCA2 with selectivity versus SMARCA4. The conclusions are well supported by the data provided and detailed control studies are performed. A detailed experimental section is included and the chemistry is well described. The authors also do an excellent job characterizing the compound using in vivo tumor models which only a few degrader papers to- date have done. The paper is also nicely written, easy to read and the conclusion section is good. The figures are nice and easy to follow. I think the paper deserves to be published without further modification.  
+
+Specific Review Questions:  
+
+- What are the noteworthy results?  
+
+Discovery of SMARCA2 selective degraders, corroboration of selective cytotoxicity in cells harboring loss of function in SMARCA4, demonstration of in vivo efficacy.  
+
+- Will the work be of significance to the field and related fields? How does it compare to the established literature? If the work is not original, please provide relevant references.  
+
+Helps advance the field of small molecule degraders, establishing differentiation relative to inhibitors, demonstrates ability to get paralog selectivity.  
+
+- Does the work support the conclusions and claims, or is additional evidence needed?  
+
+Conclusions are well supported by the provided data.  
+
+- Are there any flaws in the data analysis, interpretation and conclusions? Do these prohibit publication or require revision?  
+
+None noted.  
+
+- Is the methodology sound? Does the work meet the expected standards in your field? Yes  
+
+- Is there enough detail provided in the methods for the work to be reproduced?  
+
+Yes
+
+<--- Page Split --->
+
+For ease in reviewing our changes to the manuscript, we additionally uploaded a marked- up word version of the main body and methods section of the manuscript to more readily allow the reviewers to see the exact changes we made. Figures are embedded within the final manuscript. We look forward to your review of this revised manuscript.  
+
+## RESPONSE TO REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+In the manuscript the authors report a new moderately specific SMARCA2 degrader molecule and show this has efficacy in vitro and in vivo on SMARCA4 mutant cells. Several non- selective inhibitors of the highly homologous SMARCA2/4 proteins exist, but this papers shows the designed PROTAC strategy yields some specificity for SMARCA2. The mechanism that controls this specificity is relatively unexplored, but nevertheless the paper describes a useful molecule that will help investigation of SMARCA4 mutant cancers. Overall, the experiments are well- described, cautiously interpreted, and the conclusions support the main point of the paper.  
+
+We thank the reviewer for their commentary.  
+
+I have only minor comments on experimental presentation and request for clarifications. Notably, I am not a chemical synthesis expert and have not evaluated if the information on synthesis of the PROTAC is sufficient.  
+
+1. In figure 3C, I would encourage inclusion of the control cell lines in the main figure. It was difficult to compare G1 cells without seeing what the controls look like. We have added the control lines to the main figure (revised Fig. 3D)  
+
+2. Figure 4a, and extended data. These plots were difficult to read with the multiple axis. I'd suggest breaking this into independent figures.  
+
+We have separated out the PK data from the tumor PD data. (Fig 4a and revised Extended Data Figure 5a)  
+
+3. In several places in the discussion there are results presented as data not shown. These statements should be removed or the data should be presented. Otherwise, future work cannot accurately build on the results. We have made the suggested edits, removing any statements referencing data not shown.  
+
+4. I did not see an accession number for the RNAseq data presented in Figure 3d Our RNAseq data has been accessed under GEO205542. We have updated the methods accordingly.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The authors have generated a proteolysis targeting chimera (PROTAC) degrader using a molecule that binds the bromodomains of SMARCA2, SMARCA4 and PBRM1 to the E3 ubiquitin ligase VHL. This molecule (A947) seems to exhibit a potent and, to some extent, selective degradation of SMARCA2. There has been at least another previous publication on the generation of a different PROTAC targeting SMARCA2 for degradation, which is referred to by the authors in their manuscript.  
+
+Therapies targeting SMARCA4- mutant tumors are of great interest, given the relatively high frequency of these alterations in many solid tumors, including lung cancer, among other highly aggressive cancers. Therefore, the work is valuable in terms of clinical purposes.  
+
+We thank the reviewer for their commentary.  
+
+Comments:
+
+<--- Page Split --->
+
+1. - The authors use a panel of lung cancer cell lines that are mutant for SMARCA4 and a control panel of wild-type SMARCA4 cells. it is known that, for still unknown reasons, some SMARCA4-mutant cells show very low or complete absence of SMARCA2 protein, by western-blot. They provide the information on the genetic alterations in SMARCA4 in these cells (extended Table 3) and on the mRNA levels (RNA-seq of the databases) of SMARCA4 and SMARCA2. However, the authors should not rely on the mRNA to determine the presence or absence of SMARCA2 in each of the cell lines. Instead, they should perform a western-blot to determine the total protein levels of SMARCA4 and SMARCA2 of every one of the cell lines included in the study. The results should be included in a main figure.  
+
+We have incorporated a new figure (revised Fig. 3b) evaluating SMARCA2, SMARCA4, PBRM1 and a loading control across the full panel of cell line models utilized in the manuscript. Given the number of cell lines, we were unable to run all cell lines in one immunoblot. Hence, we separated into 2 immunoblots and purposely included models representative of both SMARCA4 mutant and wild-type into each blot; in order to provide the reader better perspective on protein levels between these states.  
+
+2. - To provide more robust data, the authors need to show the reduction in SMARCA2 protein levels with A947 administration and determine if this reduction parallels effects on cell growth inhibition in a wide panel of lung cancer cell lines (among those used to determine the effects of the drug on the growth inhibition cell), not just a single cell line as shown in the present work. For this, it would be appropriate to carry out a western blot of the SMARCA2, SMARCA4 (for SMARCA4-wt cells) and PBRM1 proteins after the administration of the drug.  
+
+This is an important point that we apologize that we missed in the initial submission. We addressed the reviewers concern 2 different ways.  
+
+(1) We evaluated a dose response of A947 across a panel of 7 SMARCA4 mutant and 7 SMARCA4 wild-type models by immunofluorescence and monitored degradation of SMARCA2 and SMARCA4. The calculated \(\mathrm{DC}_{50}\) and \(95\%\) confidence interval values are incorporated into Supplemental Table 3 (columns L/M and N/O). Although not included in the manuscript, we have pasted the dose response curves from which these \(\mathrm{DC}_{50}\) 's were calculated at the end of this letter for full transparency. Despite the differences in cellular growth inhibition between SMARCA4 mutant and WT cell line models, this was not solely due to differences in the ability of A947 to degrade SMARCA2 (Extended Data Fig. 4b). A947 was equally potent on SMARCA2 between these models. Also, although there was a range of \(\mathrm{IC}_{50}\) 's for growth inhibition within SMARCA4 mutant and WT models, we did not observe a direct correlation with growth inhibition \((\mathrm{IC}_{50})\) and degradation of SMARCA2 within the SMARCA4 mutant or within the SMARCA4 WT models. The differences in growth inhibition within the SMARCA4 WT models could also not be accounted for due to differences in SMARCA4 degradation (Extended Data Fig. 4d). The plots to highlight a lack of correlation are incorporated as Extended Data Figures 4b and d and we additionally modified the text communicate this point.  
+
+(2) We additionally evaluated a dose response of A947 across a smaller panel of 4 SMARCA4 mutant and 4 SMARCA4 wild-type models and monitored not only degradation of SMARCA2 and SMARCA4, but PBRM1 by Western blotting given that we did not have a robust immunofluorescence assay. We selected cell lines that represented the largest differences in growth inhibition (ie. the \(\mathrm{IC}_{50}\) outliers). This data is incorporated in Extended Data Figure 4c. Differences in growth inhibition could also not be accounted for by differences in degradation of PBRM1 between or within SMARCA4<sup>mut</sup> or SMARCA4<sup>WT</sup> models.  
+
+3. - Some immunostainings of SMARCA2 and SMARCA4 (or at least of SMARCA2) should be performed in the animal models to more clearly evidenced the reduction in the levels of these proteins, as compared to the controls. It is intriguing that, in their work, the authors detect SMARCA4 protein by wb in most of the SMARCA4-mutant cells, since most truncated SMARCA4 proteins are usually degraded by the cells (non-sense mediated decay mechanism).  
+
+We have evaluated SMARCA2 levels by immunohistochemistry to provide orthogonal evidence demonstrating the reduction in SMARCA2 in vivo (revised Extended Data Fig. 5b). Regarding the commentary on SMARCA4, we apologize if this was misunderstood. Indeed, SMARCA4 transcript undergoes non-sense mediated decay and we observe little evidence for strong SMARCA4 protein levels in cell lines with SMARCA4 truncating mutations (see revised Figure 3b) (note: cell lines with SMARCA4
+
+<--- Page Split --->
+
+missense mutations do still express SMARCA4 protein) However, we do observe a Smarca4 band by Western blotting from the in vivo xenografts of human SMARCA4 mutant cell lines due to antibody cross- reactivity with murine Smarca4 (Extended Data Figure 8). This is further highlighted by the immunohistochemistry evaluating SMARCA4 levels in xenografts (Extended Data Fig. 5b), whereby a lack of signal is observed within the tumor epithelium, however expressed within the stromal infiltrate.  
+
+4. - Statistics are lacking in the figure 3, fig 5c, and in ext. data fig 6. We have added the respective statistics.  
+
+5. - Is the data in figure 4g from a WB? Please indicate it and show the blot. In fig 5b, shouldn't some SMARCA4wt cells been included for comparison?  
+
+Yes, the data from Figure 4g is derived from Western blot. We modified the legend to indicate this and included the Western blot data in revised Extended Data Fig. 8a to provide direct comparison with the SMARCA4<sup>mut</sup> models in Extended Data Fig. 8b and 8c.  
+
+We have included 2 SMARCA4<sup>WT</sup> models in Extended Data Fig. 9 to demonstrate a lack of synergy between SMARCA2 degradation and MCL1 inhibition.  
+
+6. - There are only 5 main figures but 10 extended data figures. The authors could move some of the figures from the extended data, into main figures. Extended data 1 and 5a could be added to figure 1. Extended data fig 6b and c could be moved to figure 3. Extended fig. 10 moved to Fig 4...  
+
+We have moved Extended Data 1 to Main Figure 1d.  
+
+We have moved Extended Data 5a to Main Figure 1g.  
+
+Extended Data Figure 6c was moved to Main Figure 3d  
+
+Due to size constraints for the revised Figures 3 and 4, we were unable to move Extended Data Figures 6b and 10, respectively. However, we did re- arrange the Extended Data Figures and combine a few to make the extended data easier for the reader to visualize. As such, we have shortened the Extended Data figures from 11 to 9. A summary of the changes made to the original Extended Data Figures are:  
+
+Ext Fig 1.  
+
+- Moved to Main body Fig. 1d  
+
+Ext Fig 2.  
+
+- Not modified, revised Extended Data Fig. 1  
+
+Ext Fig 3.  
+
+- Not modified, revised Extended Data Fig. 2  
+
+Ext Fig 4.  
+
+- Not modified, but combined with Ext Data Fig. 4d to make revised Extended Data Fig. 3  
+
+Ext Fig 5.  
+
+- Moved 5a to Main body Fig. 1g  
+
+- Combined 5b with Ext Data Fig. 4d to make revised Extended Data Fig. 3  
+
+Ext Fig 6.  
+
+- Moved 6c to Main body Fig. 3d  
+
+- 6a, 6b, revised Extended Data Fig. 4a, 4e, respectively  
+
+Ext Fig 7.  
+
+- Split into 2 separate graphs as per Reviewer #1, revised Extended Data Fig. 5a  
+
+Ext. Fig 8- 11.  
+
+- Not modified, revised Extended Data Figures 6-9.  
+
+7. - To make the reading easier, please include the name of the cell line used in each of the figures. We have added cell line names to the figures.  
+
+Reviewer #3 (Remarks to the Author):
+
+<--- Page Split --->
+
+This is an excellent degrader development story detailing the discovery and detailed characterization of a heterobivalent degrader of SMARCA2 with selectivity versus SMARCA4. The conclusions are well supported by the data provided and detailed control studies are performed. A detailed experimental section is included and the chemistry is well described. The authors also do an excellent job characterizing the compound using in vivo tumor models which only a few degrader papers to- date have done. The paper is also nicely written, easy to read and the conclusion section is good. The figures are nice and easy to follow. I think the paper deserves to be published without further modification. We thank the reviewer for their commentary.
+
+<--- Page Split --->
+![PLACEHOLDER_7_0]
+
+
+<--- Page Split --->
+
+# 1.1.1.1.1.1.1.1.1.1.1
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  Remarks to the Author:  The authors have addressed all my concerns. I believe this should be published.  
+
+Reviewer #2:  Remarks to the Author:  The authors have adequately addressed all the comments and concerns raised.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d71bdbebab246bc381157702d44cee9063a8735d6941e27b5a7c48699b0cf6a8/supplementary_0_Peer Review File__d71bdbebab246bc381157702d44cee9063a8735d6941e27b5a7c48699b0cf6a8_det.mmd

@@ -0,0 +1,302 @@
+<|ref|>title<|/ref|><|det|>[[61, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[68, 110, 362, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[70, 154, 839, 210]]<|/det|>
+Selective PROTAC- mediated degradation of SMARCA2 is efficacious in SMARCA4 mutant cancers  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 911, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 84, 293, 97]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 113, 300, 140]]<|/det|>
+Reviewer #1: Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[119, 141, 877, 238]]<|/det|>
+In the manuscript the authors report a new moderately specific SMARCA2 degrader molecule and show this has efficacy in vitro and in vivo on SMARCA4 mutant cells. Several non- selective inhibitors of the highly homologous SMARCA2/4 proteins exist, but this papers shows the designed PROTAC strategy yields some specificity for SMARCA2. The mechanism that controls this specificity is relatively unexplored, but nevertheless the paper describes a useful molecule that will help investigation of SMARCA4 mutant cancers. Overall, the experiments are well- described, cautiously interpreted, and the conclusions support the main point of the paper.  
+
+<|ref|>text<|/ref|><|det|>[[119, 251, 877, 294]]<|/det|>
+I have only minor comments on experimental presentation and request for clarifications. Notably, I am not a chemical synthesis expert and have not evaluated if the information on synthesis of the PROTAC is sufficient.  
+
+<|ref|>text<|/ref|><|det|>[[119, 308, 839, 336]]<|/det|>
+1. In figure 3C, I would encourage inclusion of the control cell lines in the main figure. It was difficult to compare G1 cells without seeing what the controls look like.  
+
+<|ref|>text<|/ref|><|det|>[[119, 350, 828, 378]]<|/det|>
+2. Figure 4a, and extended data. These plots were difficult to read with the multiple axis. I'd suggest breaking this into independent figures.  
+
+<|ref|>text<|/ref|><|det|>[[119, 392, 850, 433]]<|/det|>
+3. In several places in the discussion there are results presented as data not shown. These statements should be removed or the data should be presented. Otherwise, future work cannot accurately build on the results.  
+
+<|ref|>text<|/ref|><|det|>[[119, 448, 738, 462]]<|/det|>
+4. I did not see an accession number for the RNAseq data presented in Figure 3d  
+
+<|ref|>text<|/ref|><|det|>[[119, 504, 222, 516]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[119, 519, 300, 531]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[119, 532, 864, 616]]<|/det|>
+The authors have generated a proteolysis targeting chimera (PROTAC) degrader using a molecule that binds the bromodomains of SMARCA2, SMARCA4 and PBRM1 to the E3 ubiquitin ligase VHL. This molecule (A947) seems to exhibit a potent and, to some extent, selective degradation of SMARCA2. There has been at least another previous publication on the generation of a different PROTAC targeting SMARCA2 for degradation, which is referred to by the authors in their manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[119, 617, 848, 659]]<|/det|>
+Therapies targeting SMARCA4- mutant tumors are of great interest, given the relatively high frequency of these alterations in many solid tumors, including lung cancer, among other highly aggressive cancers. Therefore, the work is valuable in terms of clinical purposes.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 674, 207, 687]]<|/det|>
+## Comments:  
+
+<|ref|>text<|/ref|><|det|>[[119, 701, 877, 825]]<|/det|>
+1. - The authors use a panel of lung cancer cell lines that are mutant for SMARCA4 and a control panel of wild-type SMARCA4 cells. It is known that, for still unknown reasons, some SMARCA4- mutant cells show very low or complete absence of SMARCA2 protein, by western-blot. They provide the information on the genetic alterations in SMARCA4 in these cells (extended Table 3) and on the mRNA levels (RNA-seq of the databases) of SMARCA4 and SMARCA2. However, the authors should not rely on the mRNA to determine the presence or absence of SMARCA2 in each of the cell lines. Instead, they should perform a western-blot to determine the total protein levels of SMARCA4 and SMARCA2 of every one of the cell lines included in the study. The results should be included in a main figure.  
+
+<|ref|>text<|/ref|><|det|>[[119, 827, 877, 910]]<|/det|>
+2. - To provide more robust data, the authors need to show the reduction in SMARCA2 protein levels with A947 administration and determine if this reduction parallels effects on cell growth inhibition in a wide panel of lung cancer cell lines (among those used to determine the effects of the drug on the growth inhibition cell), not just a single cell line as shown in the present work. For this, it would be appropriate to carry out a western blot of the SMARCA2, SMARCA4 (for SMARCA4- wt cells) and PBRM1 proteins after the administration of the drug.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 83, 850, 150]]<|/det|>
+3. - Some immunostainings of SMARCA2 and SMARCA4 (or at least of SMARCA2) should be performed in the animal models to more clearly evidenced the reduction in the levels of these proteins, as compared to the controls. It is intriguing that, in their work, the authors detect SMARCA4 protein by wb in most of the SMARCA4-mutant cells, since most truncated SMARCA4 proteins are usually degraded by the cells (non-sense mediated decay mechanism).  
+
+<|ref|>text<|/ref|><|det|>[[118, 152, 648, 166]]<|/det|>
+4. - Statistics are lacking in the figure 3, fig 5c, and in ext. data fig 6.  
+
+<|ref|>text<|/ref|><|det|>[[118, 167, 852, 195]]<|/det|>
+5. - Is the data in figure 4g from a WB? Please indicate it and show the blot. In fig 5b, shouldn't some SMARCA4wt cells been included for comparison?  
+
+<|ref|>text<|/ref|><|det|>[[118, 196, 872, 252]]<|/det|>
+6. - There are only 5 main figures but 10 extended data figures. The authors could move some of the figures from the extended data, into main figures. Extended data 1 and 5a could be added to figure 1. Extended data fig 6b and c could be moved to figure 3. Extended fig. 10 moved to Fig 4.7. - To make the reading easier, please include the name of the cell line used in each of the figures.  
+
+<|ref|>text<|/ref|><|det|>[[118, 295, 300, 324]]<|/det|>
+Reviewer #3: Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[118, 364, 870, 476]]<|/det|>
+This is an excellent degrader development story detailing the discovery and detailed characterization of a heterobivalent degrader of SMARCA2 with selectivity versus SMARCA4. The conclusions are well supported by the data provided and detailed control studies are performed. A detailed experimental section is included and the chemistry is well described. The authors also do an excellent job characterizing the compound using in vivo tumor models which only a few degrader papers to- date have done. The paper is also nicely written, easy to read and the conclusion section is good. The figures are nice and easy to follow. I think the paper deserves to be published without further modification.  
+
+<|ref|>text<|/ref|><|det|>[[118, 476, 323, 490]]<|/det|>
+Specific Review Questions:  
+
+<|ref|>text<|/ref|><|det|>[[118, 491, 390, 504]]<|/det|>
+- What are the noteworthy results?  
+
+<|ref|>text<|/ref|><|det|>[[118, 504, 802, 532]]<|/det|>
+Discovery of SMARCA2 selective degraders, corroboration of selective cytotoxicity in cells harboring loss of function in SMARCA4, demonstration of in vivo efficacy.  
+
+<|ref|>text<|/ref|><|det|>[[118, 533, 820, 560]]<|/det|>
+- Will the work be of significance to the field and related fields? How does it compare to the established literature? If the work is not original, please provide relevant references.  
+
+<|ref|>text<|/ref|><|det|>[[118, 561, 815, 588]]<|/det|>
+Helps advance the field of small molecule degraders, establishing differentiation relative to inhibitors, demonstrates ability to get paralog selectivity.  
+
+<|ref|>text<|/ref|><|det|>[[118, 589, 780, 602]]<|/det|>
+- Does the work support the conclusions and claims, or is additional evidence needed?  
+
+<|ref|>text<|/ref|><|det|>[[118, 603, 530, 616]]<|/det|>
+Conclusions are well supported by the provided data.  
+
+<|ref|>text<|/ref|><|det|>[[118, 617, 820, 644]]<|/det|>
+- Are there any flaws in the data analysis, interpretation and conclusions? Do these prohibit publication or require revision?  
+
+<|ref|>text<|/ref|><|det|>[[118, 645, 210, 658]]<|/det|>
+None noted.  
+
+<|ref|>text<|/ref|><|det|>[[118, 659, 787, 688]]<|/det|>
+- Is the methodology sound? Does the work meet the expected standards in your field? Yes  
+
+<|ref|>text<|/ref|><|det|>[[118, 688, 736, 701]]<|/det|>
+- Is there enough detail provided in the methods for the work to be reproduced?  
+
+<|ref|>text<|/ref|><|det|>[[118, 702, 147, 714]]<|/det|>
+Yes
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 91, 837, 160]]<|/det|>
+For ease in reviewing our changes to the manuscript, we additionally uploaded a marked- up word version of the main body and methods section of the manuscript to more readily allow the reviewers to see the exact changes we made. Figures are embedded within the final manuscript. We look forward to your review of this revised manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[144, 212, 458, 228]]<|/det|>
+## RESPONSE TO REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[145, 241, 398, 256]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[144, 270, 829, 375]]<|/det|>
+In the manuscript the authors report a new moderately specific SMARCA2 degrader molecule and show this has efficacy in vitro and in vivo on SMARCA4 mutant cells. Several non- selective inhibitors of the highly homologous SMARCA2/4 proteins exist, but this papers shows the designed PROTAC strategy yields some specificity for SMARCA2. The mechanism that controls this specificity is relatively unexplored, but nevertheless the paper describes a useful molecule that will help investigation of SMARCA4 mutant cancers. Overall, the experiments are well- described, cautiously interpreted, and the conclusions support the main point of the paper.  
+
+<|ref|>text<|/ref|><|det|>[[145, 374, 441, 388]]<|/det|>
+We thank the reviewer for their commentary.  
+
+<|ref|>text<|/ref|><|det|>[[144, 401, 844, 445]]<|/det|>
+I have only minor comments on experimental presentation and request for clarifications. Notably, I am not a chemical synthesis expert and have not evaluated if the information on synthesis of the PROTAC is sufficient.  
+
+<|ref|>text<|/ref|><|det|>[[144, 459, 839, 503]]<|/det|>
+1. In figure 3C, I would encourage inclusion of the control cell lines in the main figure. It was difficult to compare G1 cells without seeing what the controls look like. We have added the control lines to the main figure (revised Fig. 3D)  
+
+<|ref|>text<|/ref|><|det|>[[144, 516, 803, 546]]<|/det|>
+2. Figure 4a, and extended data. These plots were difficult to read with the multiple axis. I'd suggest breaking this into independent figures.  
+
+<|ref|>text<|/ref|><|det|>[[144, 546, 840, 562]]<|/det|>
+We have separated out the PK data from the tumor PD data. (Fig 4a and revised Extended Data Figure 5a)  
+
+<|ref|>text<|/ref|><|det|>[[144, 575, 848, 620]]<|/det|>
+3. In several places in the discussion there are results presented as data not shown. These statements should be removed or the data should be presented. Otherwise, future work cannot accurately build on the results. We have made the suggested edits, removing any statements referencing data not shown.  
+
+<|ref|>text<|/ref|><|det|>[[144, 633, 808, 664]]<|/det|>
+4. I did not see an accession number for the RNAseq data presented in Figure 3d Our RNAseq data has been accessed under GEO205542. We have updated the methods accordingly.  
+
+<|ref|>text<|/ref|><|det|>[[145, 706, 398, 721]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[144, 735, 848, 808]]<|/det|>
+The authors have generated a proteolysis targeting chimera (PROTAC) degrader using a molecule that binds the bromodomains of SMARCA2, SMARCA4 and PBRM1 to the E3 ubiquitin ligase VHL. This molecule (A947) seems to exhibit a potent and, to some extent, selective degradation of SMARCA2. There has been at least another previous publication on the generation of a different PROTAC targeting SMARCA2 for degradation, which is referred to by the authors in their manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[144, 808, 842, 852]]<|/det|>
+Therapies targeting SMARCA4- mutant tumors are of great interest, given the relatively high frequency of these alterations in many solid tumors, including lung cancer, among other highly aggressive cancers. Therefore, the work is valuable in terms of clinical purposes.  
+
+<|ref|>text<|/ref|><|det|>[[145, 852, 441, 866]]<|/det|>
+We thank the reviewer for their commentary.  
+
+<|ref|>text<|/ref|><|det|>[[144, 880, 221, 894]]<|/det|>
+Comments:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 102, 850, 220]]<|/det|>
+1. - The authors use a panel of lung cancer cell lines that are mutant for SMARCA4 and a control panel of wild-type SMARCA4 cells. it is known that, for still unknown reasons, some SMARCA4-mutant cells show very low or complete absence of SMARCA2 protein, by western-blot. They provide the information on the genetic alterations in SMARCA4 in these cells (extended Table 3) and on the mRNA levels (RNA-seq of the databases) of SMARCA4 and SMARCA2. However, the authors should not rely on the mRNA to determine the presence or absence of SMARCA2 in each of the cell lines. Instead, they should perform a western-blot to determine the total protein levels of SMARCA4 and SMARCA2 of every one of the cell lines included in the study. The results should be included in a main figure.  
+
+<|ref|>text<|/ref|><|det|>[[144, 220, 850, 293]]<|/det|>
+We have incorporated a new figure (revised Fig. 3b) evaluating SMARCA2, SMARCA4, PBRM1 and a loading control across the full panel of cell line models utilized in the manuscript. Given the number of cell lines, we were unable to run all cell lines in one immunoblot. Hence, we separated into 2 immunoblots and purposely included models representative of both SMARCA4 mutant and wild-type into each blot; in order to provide the reader better perspective on protein levels between these states.  
+
+<|ref|>text<|/ref|><|det|>[[144, 306, 844, 395]]<|/det|>
+2. - To provide more robust data, the authors need to show the reduction in SMARCA2 protein levels with A947 administration and determine if this reduction parallels effects on cell growth inhibition in a wide panel of lung cancer cell lines (among those used to determine the effects of the drug on the growth inhibition cell), not just a single cell line as shown in the present work. For this, it would be appropriate to carry out a western blot of the SMARCA2, SMARCA4 (for SMARCA4-wt cells) and PBRM1 proteins after the administration of the drug.  
+
+<|ref|>text<|/ref|><|det|>[[144, 395, 825, 424]]<|/det|>
+This is an important point that we apologize that we missed in the initial submission. We addressed the reviewers concern 2 different ways.  
+
+<|ref|>text<|/ref|><|det|>[[173, 423, 852, 640]]<|/det|>
+(1) We evaluated a dose response of A947 across a panel of 7 SMARCA4 mutant and 7 SMARCA4 wild-type models by immunofluorescence and monitored degradation of SMARCA2 and SMARCA4. The calculated \(\mathrm{DC}_{50}\) and \(95\%\) confidence interval values are incorporated into Supplemental Table 3 (columns L/M and N/O). Although not included in the manuscript, we have pasted the dose response curves from which these \(\mathrm{DC}_{50}\) 's were calculated at the end of this letter for full transparency. Despite the differences in cellular growth inhibition between SMARCA4 mutant and WT cell line models, this was not solely due to differences in the ability of A947 to degrade SMARCA2 (Extended Data Fig. 4b). A947 was equally potent on SMARCA2 between these models. Also, although there was a range of \(\mathrm{IC}_{50}\) 's for growth inhibition within SMARCA4 mutant and WT models, we did not observe a direct correlation with growth inhibition \((\mathrm{IC}_{50})\) and degradation of SMARCA2 within the SMARCA4 mutant or within the SMARCA4 WT models. The differences in growth inhibition within the SMARCA4 WT models could also not be accounted for due to differences in SMARCA4 degradation (Extended Data Fig. 4d). The plots to highlight a lack of correlation are incorporated as Extended Data Figures 4b and d and we additionally modified the text communicate this point.  
+
+<|ref|>text<|/ref|><|det|>[[173, 640, 850, 743]]<|/det|>
+(2) We additionally evaluated a dose response of A947 across a smaller panel of 4 SMARCA4 mutant and 4 SMARCA4 wild-type models and monitored not only degradation of SMARCA2 and SMARCA4, but PBRM1 by Western blotting given that we did not have a robust immunofluorescence assay. We selected cell lines that represented the largest differences in growth inhibition (ie. the \(\mathrm{IC}_{50}\) outliers). This data is incorporated in Extended Data Figure 4c. Differences in growth inhibition could also not be accounted for by differences in degradation of PBRM1 between or within SMARCA4<sup>mut</sup> or SMARCA4<sup>WT</sup> models.  
+
+<|ref|>text<|/ref|><|det|>[[144, 756, 850, 830]]<|/det|>
+3. - Some immunostainings of SMARCA2 and SMARCA4 (or at least of SMARCA2) should be performed in the animal models to more clearly evidenced the reduction in the levels of these proteins, as compared to the controls. It is intriguing that, in their work, the authors detect SMARCA4 protein by wb in most of the SMARCA4-mutant cells, since most truncated SMARCA4 proteins are usually degraded by the cells (non-sense mediated decay mechanism).  
+
+<|ref|>text<|/ref|><|det|>[[144, 830, 850, 903]]<|/det|>
+We have evaluated SMARCA2 levels by immunohistochemistry to provide orthogonal evidence demonstrating the reduction in SMARCA2 in vivo (revised Extended Data Fig. 5b). Regarding the commentary on SMARCA4, we apologize if this was misunderstood. Indeed, SMARCA4 transcript undergoes non-sense mediated decay and we observe little evidence for strong SMARCA4 protein levels in cell lines with SMARCA4 truncating mutations (see revised Figure 3b) (note: cell lines with SMARCA4
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 90, 852, 163]]<|/det|>
+missense mutations do still express SMARCA4 protein) However, we do observe a Smarca4 band by Western blotting from the in vivo xenografts of human SMARCA4 mutant cell lines due to antibody cross- reactivity with murine Smarca4 (Extended Data Figure 8). This is further highlighted by the immunohistochemistry evaluating SMARCA4 levels in xenografts (Extended Data Fig. 5b), whereby a lack of signal is observed within the tumor epithelium, however expressed within the stromal infiltrate.  
+
+<|ref|>text<|/ref|><|det|>[[144, 177, 592, 207]]<|/det|>
+4. - Statistics are lacking in the figure 3, fig 5c, and in ext. data fig 6. We have added the respective statistics.  
+
+<|ref|>text<|/ref|><|det|>[[144, 220, 810, 250]]<|/det|>
+5. - Is the data in figure 4g from a WB? Please indicate it and show the blot. In fig 5b, shouldn't some SMARCA4wt cells been included for comparison?  
+
+<|ref|>text<|/ref|><|det|>[[144, 250, 825, 294]]<|/det|>
+Yes, the data from Figure 4g is derived from Western blot. We modified the legend to indicate this and included the Western blot data in revised Extended Data Fig. 8a to provide direct comparison with the SMARCA4<sup>mut</sup> models in Extended Data Fig. 8b and 8c.  
+
+<|ref|>text<|/ref|><|det|>[[144, 307, 803, 337]]<|/det|>
+We have included 2 SMARCA4<sup>WT</sup> models in Extended Data Fig. 9 to demonstrate a lack of synergy between SMARCA2 degradation and MCL1 inhibition.  
+
+<|ref|>text<|/ref|><|det|>[[144, 350, 849, 394]]<|/det|>
+6. - There are only 5 main figures but 10 extended data figures. The authors could move some of the figures from the extended data, into main figures. Extended data 1 and 5a could be added to figure 1. Extended data fig 6b and c could be moved to figure 3. Extended fig. 10 moved to Fig 4...  
+
+<|ref|>text<|/ref|><|det|>[[144, 395, 494, 409]]<|/det|>
+We have moved Extended Data 1 to Main Figure 1d.  
+
+<|ref|>text<|/ref|><|det|>[[144, 410, 504, 424]]<|/det|>
+We have moved Extended Data 5a to Main Figure 1g.  
+
+<|ref|>text<|/ref|><|det|>[[144, 425, 510, 439]]<|/det|>
+Extended Data Figure 6c was moved to Main Figure 3d  
+
+<|ref|>text<|/ref|><|det|>[[144, 452, 844, 511]]<|/det|>
+Due to size constraints for the revised Figures 3 and 4, we were unable to move Extended Data Figures 6b and 10, respectively. However, we did re- arrange the Extended Data Figures and combine a few to make the extended data easier for the reader to visualize. As such, we have shortened the Extended Data figures from 11 to 9. A summary of the changes made to the original Extended Data Figures are:  
+
+<|ref|>text<|/ref|><|det|>[[173, 512, 240, 525]]<|/det|>
+Ext Fig 1.  
+
+<|ref|>text<|/ref|><|det|>[[203, 527, 425, 541]]<|/det|>
+- Moved to Main body Fig. 1d  
+
+<|ref|>text<|/ref|><|det|>[[173, 543, 240, 555]]<|/det|>
+Ext Fig 2.  
+
+<|ref|>text<|/ref|><|det|>[[203, 557, 520, 570]]<|/det|>
+- Not modified, revised Extended Data Fig. 1  
+
+<|ref|>text<|/ref|><|det|>[[173, 572, 240, 585]]<|/det|>
+Ext Fig 3.  
+
+<|ref|>text<|/ref|><|det|>[[203, 587, 520, 600]]<|/det|>
+- Not modified, revised Extended Data Fig. 2  
+
+<|ref|>text<|/ref|><|det|>[[173, 602, 240, 615]]<|/det|>
+Ext Fig 4.  
+
+<|ref|>text<|/ref|><|det|>[[203, 617, 815, 631]]<|/det|>
+- Not modified, but combined with Ext Data Fig. 4d to make revised Extended Data Fig. 3  
+
+<|ref|>text<|/ref|><|det|>[[173, 632, 240, 645]]<|/det|>
+Ext Fig 5.  
+
+<|ref|>text<|/ref|><|det|>[[203, 647, 444, 660]]<|/det|>
+- Moved 5a to Main body Fig. 1g  
+
+<|ref|>text<|/ref|><|det|>[[203, 662, 719, 676]]<|/det|>
+- Combined 5b with Ext Data Fig. 4d to make revised Extended Data Fig. 3  
+
+<|ref|>text<|/ref|><|det|>[[173, 677, 240, 690]]<|/det|>
+Ext Fig 6.  
+
+<|ref|>text<|/ref|><|det|>[[203, 692, 444, 705]]<|/det|>
+- Moved 6c to Main body Fig. 3d  
+
+<|ref|>text<|/ref|><|det|>[[203, 707, 590, 721]]<|/det|>
+- 6a, 6b, revised Extended Data Fig. 4a, 4e, respectively  
+
+<|ref|>text<|/ref|><|det|>[[173, 722, 240, 736]]<|/det|>
+Ext Fig 7.  
+
+<|ref|>text<|/ref|><|det|>[[203, 737, 748, 752]]<|/det|>
+- Split into 2 separate graphs as per Reviewer #1, revised Extended Data Fig. 5a  
+
+<|ref|>text<|/ref|><|det|>[[173, 752, 265, 765]]<|/det|>
+Ext. Fig 8- 11.  
+
+<|ref|>text<|/ref|><|det|>[[203, 767, 562, 781]]<|/det|>
+- Not modified, revised Extended Data Figures 6-9.  
+
+<|ref|>text<|/ref|><|det|>[[144, 810, 784, 839]]<|/det|>
+7. - To make the reading easier, please include the name of the cell line used in each of the figures. We have added cell line names to the figures.  
+
+<|ref|>text<|/ref|><|det|>[[144, 882, 398, 897]]<|/det|>
+Reviewer #3 (Remarks to the Author):
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[144, 146, 850, 255]]<|/det|>
+This is an excellent degrader development story detailing the discovery and detailed characterization of a heterobivalent degrader of SMARCA2 with selectivity versus SMARCA4. The conclusions are well supported by the data provided and detailed control studies are performed. A detailed experimental section is included and the chemistry is well described. The authors also do an excellent job characterizing the compound using in vivo tumor models which only a few degrader papers to- date have done. The paper is also nicely written, easy to read and the conclusion section is good. The figures are nice and easy to follow. I think the paper deserves to be published without further modification. We thank the reviewer for their commentary.
+
+<--- Page Split --->
+<|ref|>image<|/ref|><|det|>[[170, 99, 525, 899]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[0, 0, 997, 997]]<|/det|>
+# 1.1.1.1.1.1.1.1.1.1.1
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[119, 84, 294, 98]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[118, 113, 732, 155]]<|/det|>
+Reviewer #1:  Remarks to the Author:  The authors have addressed all my concerns. I believe this should be published.  
+
+<|ref|>text<|/ref|><|det|>[[118, 196, 726, 239]]<|/det|>
+Reviewer #2:  Remarks to the Author:  The authors have adequately addressed all the comments and concerns raised.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d748b2742d4524306a919a7ae9c28e44f8c142a9dcd214c3fe54b9afb3ac4c59/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d748b2742d4524306a919a7ae9c28e44f8c142a9dcd214c3fe54b9afb3ac4c59/supplementary_0_Peer Review File__d748b2742d4524306a919a7ae9c28e44f8c142a9dcd214c3fe54b9afb3ac4c59.mmd

@@ -0,0 +1,139 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+A Mechanistic Reinterpretation of Fast Inactivation in Voltage- Gated Na+ Channels  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  
+
+Remarks to the Author:  
+
+This is an important paper. The cryo- EM structures of vertebrate voltage- dependent sodium channels have shown that the long- held model of channel inactivation as physical occlusion of the pore by the IFM motif in the III- IV intracellular linker is not correct, because the IFM motif was not in the pore but rather buried in a pocket fairly distant from the pore. However, the structures do not give much insight into exactly how binding of the IFM motif in this pocket produces a closed, inactivated channel. Here, the authors use the structures of inactivated channels to recognize two rings of hydrophobic residues that form a "double- seal" near the inner moth of the pore. Using this modeled structure, they then show that mutating a residue in each ring results in an inactivated channel that is no longer fully closed but instead is leaky, thus offering strong support for their proposed mechanism. This is probably the single most remarkable result from mutagenesis of an ion channel that I know, because it would have been impossible to come up with this combination of mutations without thinking deeply about the structure. Even more remarkably, the authors find that the leaky channel formed by these mutations has an altered ionic selectivity, with reduced selectivity for Na over K. This implies that there is some sort of coupling between this region near the inner pore and the selectivity filter of the channel located near the outer pore.  
+
+Besides presenting fascinating and very important results, the paper is a technical tour- de- force of channel biophysics, using well- designed measurements of gating kinetics, gating current, and ionic selectivity in the mutated channels to lead the authors to their conclusions and new model. The presentation of the motivation for the experiments and the discussion of their interpretation are done with clarity and concision. I can find only minor points of wording that might be improved.  
+
+Minor  
+
+13 "The hinged- lid model is long accepted as the canonical model for fast inactivation" would be accurate as "13 The hinged- lid model was long accepted....". It has now been 6 years that the cryo- EM structures showed that the model is incorrect.  
+
+44 "positively charge" should be "positively charged"  
+
+57 "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif, though docking into a hydrophobic pocket, resided far away from the pore in the putative inactivated state to block the permeation path, in contrast to the predictions of the canonical "hinged- lid" model."  
+
+I found this sentence a little difficult to interpret. Maybe the meaning would be clearer as "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif docked into a hydrophobic pocket far away from the pore rather than physically occluding the pore, in contrast to the predictions of the canonical "hinged- lid" model."  
+
+104 "Given the fact that both structures are determined at 0mV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore resemble a fast inactivated state."  
+
+Since channels in cryoEM have been at 0 mV for a long time, I would think the structure is more likely to be a slow inactivated state rather than fast inactivated state. It does not seem necessary or useful to get into a discussion of this in the present paper, but the sentence might be slightly more accurate as "Given the fact that both structures are determined at 0mV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore is in an inactivated state."  
+
+439 "In DIII and DIV, none of the single alanine mutation yielded..." should be "mutations"  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+Liu et al investigate whether hydrophobic residues lining the intracellular pore are part of the inactivation gate of Nav channels. Given that Nav channel inactivation is essential to the function of excitable cells, insight into this mechanism is of high interest. Previous work suggests that an IFM motif located within the intracellular DIII- DIV linker of Nav channels directly blocks the channel pore to induce inactivation. However, recent Nav channel structures show that the binding of the IFM motif
+
+<--- Page Split --->
+
+is far from the pore, even though it is required for inactivation. The authors demonstrate that inactivation of the channel is impaired by mutation of several intracellular pore residues on the DIII and DIV S6 segments, preventing inactivation even though IFM is still intact, suggesting that these residues are part of the inactivation mechanism.  
+
+The manuscript is well- written and delivers a hypothesis for a new mechanism for the fast inactivation voltage- gated sodium channels. The experimental results shown provide evidence for the connection of intracellular S6 helices to an inactivation gate.  
+
+Major Comments:  
+
+- Voltage clamp fluorometry is likely to provide higher resolution data on whether the VSDs are being affected, including changes in activation and deactivation kinetics. It is surprising that these experiments were not included.- The addition of the IQM to DIIIIAA was helpful in assessing how these different channel components interact. Are the results with IQM added to DIVAA similar?- The model proposed in Figure 8 is very DIV centric even though the data presented in this work and previous work by this group and others have shown a significant role for DIII.- The authors provide a comparison between different species to demonstrate the conservation of the amino acids observed in this study. It would be valuable to compare across Nav channel isoforms that inactivate if the mechanism is thought to be universal, as suggested.- Some caveats should be included. Without identifying the coupling pathway, it is possible that the mutation of the S6 residues interrupts the normal inactivation pathway even though these positions might not directly participate. Further work is needed to convincingly make definitive statements about the participation of these positions in the inactivation gate. As noted in the manuscript and shown in refs 40 and 42, DI mutations have been shown to have similar effects on inactivation.  
+
+Minor Comments  
+
+- The different treatment of showing DIA and DIIAA data in figure 7 is confusing. Why not plot DIIAA GV and inactivation curves along with DIA?- It would be helpful to include hypotheses relating to the coupling of S6 helices to IFM binding that would cause inactivation via the intracellular S6 residues that were identified.
+
+<--- Page Split --->
+
+## RESPONSE TO THE REVIEWERS' COMMENTS  
+
+We thank the reviewers for their insightful comments. The modifications they proposed have made the manuscript much better. All the changes in the text are marked in yellow and they include the changes as a result of the reviewers' comments, the changes required by the Editor, in addition to other small corrections. We also include an extra Supplementary File with changes in yellow and the file is labeled MARK.  
+
+Reviewer #1 (Remarks to the Author):  
+
+This is an important paper. The cryo- EM structures of vertebrate voltage- dependent sodium channels have shown that the long- held model of channel inactivation as physical occlusion of the pore by the IFM motif in the III- IV intracellular linker is not correct, because the IFM motif was not in the pore but rather buried in a pocket fairly distant from the pore. However, the structures do not give much insight into exactly how binding of the IFM motif in this pocket produces a closed, inactivated channel. Here, the authors use the structures of inactivated channels to recognize two rings of hydrophobic residues that form a "double- seal" near the inner moth of the pore. Using this modeled structure, they then show that mutating a residue in each ring results in an inactivated channel that is no longer fully closed but instead is leaky, thus offering strong support for their proposed mechanism. This is probably the single most remarkable result from mutagenesis of an ion channel that I know, because it would have been impossible to come up with this combination of mutations without thinking deeply about the structure. Even more remarkably, the authors find that the leaky channel formed by these mutations has an altered ionic selectivity, with reduced selectivity for Na over K. This implies that there is some sort of coupling between this region near the inner pore and the selectivity filter of the channel located near the outer pore.  
+
+Besides presenting fascinating and very important results, the paper is a technical tour- de- force of channel biophysics, using well- designed measurements of gating kinetics, gating current, and ionic selectivity in the mutated channels to lead the authors to their conclusions and new model. The presentation of the motivation for the experiments and the discussion of their interpretation are done with clarity and concision. I can find only minor points of wording that might be improved.  
+
+We appreciate the reviewer positive feedback and the recognition of the importance of our findings.  
+
+Minor  
+
+13 "The hinged- lid model is long accepted as the canonical model for fast inactivation" would be accurate as "13 The hinged- lid model was long accepted....". It has now been 6 years that the cryo- EM structures showed that the model is incorrect.  
+
+We thank the reviewer for spotting this typo and we have changed it in the manuscript.  
+
+44 "positively charge" should be "positively charged"  
+
+We thank the reviewer for spotting this and we have changed it in the manuscript.  
+
+57 "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif, though docking into a hydrophobic pocket, resided far away from the pore in the putative inactivated state to block the permeation path, in contrast to the predictions of the canonical "hinged- lid" model." I found this sentence a little difficult to interpret. Maybe the meaning would be clearer as "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif docked into a hydrophobic pocket far away from the pore rather than physically occluding the pore, in contrast to the predictions of the canonical "hinged- lid" model."
+
+<--- Page Split --->
+
+We agree with the reviewer that by stating that particular sentence in the suggested way makes it clearer to understand. We have rephrased this particular sentence in the manuscript. 104 "Given the fact that both structures are determined at OmV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore resemble a fast inactivated state." Since channels in cryoEM have been at \(0\mathrm{mV}\) for a long time, I would think the structure is more likely to be a slow inactivated state rather than fast inactivated state. It does not seem necessary or useful to get into a discussion of this in the present paper, but the sentence might be slightly more accurate as "Given the fact that both structures are determined at OmV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore is in an inactivated state." We agree with the reviewer that it is not necessary to discuss the exact state the structures were captured in. We have rephrased that particular sentence as suggested. 439 "In DIII and DIV, none of the single alanine mutation yielded..." should be "mutations" We thank the reviewer for spotting this typo and we have changed it in the manuscript.  
+
+Reviewer #2 (Remarks to the Author):  
+
+Liu et al investigate whether hydrophobic residues lining the intracellular pore are part of the inactivation gate of Nav channels. Given that Nav channel inactivation is essential to the function of excitable cells, insight into this mechanism is of high interest. Previous work suggests that an IFM motif located within the intracellular DIII- DIV linker of Nav channels directly blocks the channel pore to induce inactivation. However, recent Nav channel structures show that the binding of the IFM motif is far from the pore, even though it is required for inactivation. The authors demonstrate that inactivation of the channel is impaired by mutation of several intracellular pore residues on the DIII and DIV S6 segments, preventing inactivation even though IFM is still intact, suggesting that these residues are part of the inactivation mechanism.  
+
+The manuscript is well- written and delivers a hypothesis for a new mechanism for the fast inactivation voltage- gated sodium channels. The experimental results shown provide evidence for the connection of intracellular S6 helices to an inactivation gate.  
+
+We thank the reviewer for the thorough evaluation of our manuscript.  
+
+## Major Comments:  
+
+- Voltage clamp fluorometry is likely to provide higher resolution data on whether the VSDs are being affected, including changes in activation and deactivation kinetics. It is surprising that these experiments were not included.  
+
+We believe gating current measurement is more suitable for this particular work because it allows us to directly assess charge immobilization, which is a critical aspect of inactivation. Additionally, by obtaining robust gating current recordings in all mutants with a high signal-to-noise ratio (which is not the case with fluorometry), we can accurately analyze the kinetic properties. While site-directed voltage clamp fluorometry (presently ongoing) may provide valuable information on individual VSD movements, our focus in this manuscript is primarily on understanding the relationship between the IFM motif and the inactivation gate. As such, measuring gating currents aligns with the traditional approach for characterizing charge immobilization and enables a comprehensive analysis of the inactivation mechanism.
+
+<--- Page Split --->
+
+- The addition of the IQM to DIIAA was helpful in assessing how these different channel components interact. Are the results with IQM added to DIVAA similar?  
+
+Interesting point raised by this reviewer. We did not try IQM_DIVAA because, unfortunately, DIVAA express much less than DIIAA and IQM also decreases expression significantly.  
+
+- The model proposed in Figure 8 is very DIV centric even though the data presented in this work and previous work by this group and others have shown a significant role for DIII.  
+
+The reviewer is correct to point the involvement of DIII on inactivation. The model is clearly a simplification but, to correct this oversight, we have added a note in the figure legend indicating that DIII is also participating in inactivation.  
+
+- The authors provide a comparison between different species to demonstrate the conservation of the amino acids observed in this study. It would be valuable to compare across Nav channel isoforms that inactivate if the mechanism is thought to be universal, as suggested.  
+
+We thank the reviewer for pointing this out. We agree that while comparing amino acid sequence across different species could provide evolutionary insights, it still remains valuable to include a sequence comparison across different isoforms of Nav channels to demonstrate similarities and special adaptations within them. We have included another supplementary figure (SuppleFig. 2B). It is true that not all the identified residues are conserved across all isoforms. However, the property of the amino acids at those positions are relatively conserved.  
+
+- Some caveats should be included. Without identifying the coupling pathway, it is possible that the mutation of the S6 residues interrupts the normal inactivation pathway even though these positions might not directly participate. Further work is needed to convincingly make definitive statements about the participation of these positions in the inactivation gate. As noted in the manuscript and shown in refs 40 and 42, DI mutations have been shown to have similar effects on inactivation.  
+
+We would like to respectfully argue against this comment. Indeed, mutations in DI have been shown to disrupt fast inactivation. L437C together with A438W (CW) in particular has been shown to remove fast inactivation almost completely. Considering the location of those residues at S6, it is tempting to conclude that those residues might form the fast inactivation gate. However, we have recorded the charge immobilization profile of CW (unpublished results\*) and we discovered that it has been significantly altered which is not case for the DIIAA and DIVAA mutants described here.  
+
+We agree that without the complete knowledge of the entire coupling pathway, there exists a possibility that the identified residues might exert their effects by interrupting the pathway. However, we believe that the existence of a leaky inactivated state provides strong evidence that support our interpretation that the S6 residues identified are the fast inactivation gate and we are not actually interrupting the inactivation pathway (as IQM or CW might be acting).  
+
+## Minor Comments  
+
+- The different treatment of showing DIA and DIIAA data in figure 7 is confusing. Why not plot DIIAA GV and inactivation curves along with DIA?  
+
+It is challenging to acquire an accurate GV curve for DIIAA. The mixture of gating current and ionic current at the beginning of the trace makes it unfeasible to reliably obtain the peak for the ionic current. While it is possible to record the gating current together with the ionic current first then apply external TTX to record only the gating current subsequently, this method is highly sensitive to small fluctuation within the system (rundown, changes in capacitance, temperature etc.,). Since, the main conclusion of this work rely mostly on results on DIIAA and DIVAA, we decided to describe the effects of DIIAA qualitatively and leave room for future investigations.  
+
+- It would be helpful to include hypotheses relating to the coupling of S6 helices to IFM binding that would cause inactivation via the intracellular S6 residues that were identified.
+
+<--- Page Split --->
+
+We thank the reviewer for this comment. Indeed, we think that there is a chain of amino acids that couple the IFM binding to the closing of the inactivation gate. We have hypothesized the residues that form this chain, and it is the subject of our future research, but as we do not have results yet, we believe that to explicitly state it is premature.  
+
+\*unpublished results have been communicated as an abstract. Yichen Liu & Francisco Bezanilla "Charge immobilization in two different fast- inactivation- removed Nav1.4 sodium channel mutants", Biophysical Journal, Volume 121, Issue 3, Supplement 1, 2022, Pages 95a- 96a, ISSN 0006- 3495, https://doi.org/10.1016/j.bpj.2021.11.2252.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d748b2742d4524306a919a7ae9c28e44f8c142a9dcd214c3fe54b9afb3ac4c59/supplementary_0_Peer Review File__d748b2742d4524306a919a7ae9c28e44f8c142a9dcd214c3fe54b9afb3ac4c59_det.mmd

@@ -0,0 +1,192 @@
+<|ref|>title<|/ref|><|det|>[[99, 40, 508, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[106, 110, 373, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[106, 155, 889, 211]]<|/det|>
+A Mechanistic Reinterpretation of Fast Inactivation in Voltage- Gated Na+ Channels  
+
+<|ref|>image<|/ref|><|det|>[[95, 732, 262, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[271, 732, 880, 785]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[116, 90, 286, 103]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[116, 118, 216, 131]]<|/det|>
+Reviewer #1:  
+
+<|ref|>text<|/ref|><|det|>[[116, 134, 291, 147]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 148, 879, 371]]<|/det|>
+This is an important paper. The cryo- EM structures of vertebrate voltage- dependent sodium channels have shown that the long- held model of channel inactivation as physical occlusion of the pore by the IFM motif in the III- IV intracellular linker is not correct, because the IFM motif was not in the pore but rather buried in a pocket fairly distant from the pore. However, the structures do not give much insight into exactly how binding of the IFM motif in this pocket produces a closed, inactivated channel. Here, the authors use the structures of inactivated channels to recognize two rings of hydrophobic residues that form a "double- seal" near the inner moth of the pore. Using this modeled structure, they then show that mutating a residue in each ring results in an inactivated channel that is no longer fully closed but instead is leaky, thus offering strong support for their proposed mechanism. This is probably the single most remarkable result from mutagenesis of an ion channel that I know, because it would have been impossible to come up with this combination of mutations without thinking deeply about the structure. Even more remarkably, the authors find that the leaky channel formed by these mutations has an altered ionic selectivity, with reduced selectivity for Na over K. This implies that there is some sort of coupling between this region near the inner pore and the selectivity filter of the channel located near the outer pore.  
+
+<|ref|>text<|/ref|><|det|>[[115, 371, 872, 447]]<|/det|>
+Besides presenting fascinating and very important results, the paper is a technical tour- de- force of channel biophysics, using well- designed measurements of gating kinetics, gating current, and ionic selectivity in the mutated channels to lead the authors to their conclusions and new model. The presentation of the motivation for the experiments and the discussion of their interpretation are done with clarity and concision. I can find only minor points of wording that might be improved.  
+
+<|ref|>text<|/ref|><|det|>[[116, 448, 160, 460]]<|/det|>
+Minor  
+
+<|ref|>text<|/ref|><|det|>[[115, 461, 864, 504]]<|/det|>
+13 "The hinged- lid model is long accepted as the canonical model for fast inactivation" would be accurate as "13 The hinged- lid model was long accepted....". It has now been 6 years that the cryo- EM structures showed that the model is incorrect.  
+
+<|ref|>text<|/ref|><|det|>[[115, 504, 511, 519]]<|/det|>
+44 "positively charge" should be "positively charged"  
+
+<|ref|>text<|/ref|><|det|>[[115, 520, 872, 562]]<|/det|>
+57 "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif, though docking into a hydrophobic pocket, resided far away from the pore in the putative inactivated state to block the permeation path, in contrast to the predictions of the canonical "hinged- lid" model."  
+
+<|ref|>text<|/ref|><|det|>[[115, 562, 861, 623]]<|/det|>
+I found this sentence a little difficult to interpret. Maybe the meaning would be clearer as "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif docked into a hydrophobic pocket far away from the pore rather than physically occluding the pore, in contrast to the predictions of the canonical "hinged- lid" model."  
+
+<|ref|>text<|/ref|><|det|>[[115, 624, 825, 653]]<|/det|>
+104 "Given the fact that both structures are determined at 0mV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore resemble a fast inactivated state."  
+
+<|ref|>text<|/ref|><|det|>[[115, 654, 875, 727]]<|/det|>
+Since channels in cryoEM have been at 0 mV for a long time, I would think the structure is more likely to be a slow inactivated state rather than fast inactivated state. It does not seem necessary or useful to get into a discussion of this in the present paper, but the sentence might be slightly more accurate as "Given the fact that both structures are determined at 0mV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore is in an inactivated state."  
+
+<|ref|>text<|/ref|><|det|>[[115, 728, 792, 743]]<|/det|>
+439 "In DIII and DIV, none of the single alanine mutation yielded..." should be "mutations"  
+
+<|ref|>text<|/ref|><|det|>[[116, 789, 216, 802]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[116, 805, 291, 818]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 819, 875, 892]]<|/det|>
+Liu et al investigate whether hydrophobic residues lining the intracellular pore are part of the inactivation gate of Nav channels. Given that Nav channel inactivation is essential to the function of excitable cells, insight into this mechanism is of high interest. Previous work suggests that an IFM motif located within the intracellular DIII- DIV linker of Nav channels directly blocks the channel pore to induce inactivation. However, recent Nav channel structures show that the binding of the IFM motif
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 860, 150]]<|/det|>
+is far from the pore, even though it is required for inactivation. The authors demonstrate that inactivation of the channel is impaired by mutation of several intracellular pore residues on the DIII and DIV S6 segments, preventing inactivation even though IFM is still intact, suggesting that these residues are part of the inactivation mechanism.  
+
+<|ref|>text<|/ref|><|det|>[[115, 164, 875, 209]]<|/det|>
+The manuscript is well- written and delivers a hypothesis for a new mechanism for the fast inactivation voltage- gated sodium channels. The experimental results shown provide evidence for the connection of intracellular S6 helices to an inactivation gate.  
+
+<|ref|>text<|/ref|><|det|>[[115, 225, 248, 238]]<|/det|>
+Major Comments:  
+
+<|ref|>text<|/ref|><|det|>[[111, 238, 876, 463]]<|/det|>
+- Voltage clamp fluorometry is likely to provide higher resolution data on whether the VSDs are being affected, including changes in activation and deactivation kinetics. It is surprising that these experiments were not included.- The addition of the IQM to DIIIIAA was helpful in assessing how these different channel components interact. Are the results with IQM added to DIVAA similar?- The model proposed in Figure 8 is very DIV centric even though the data presented in this work and previous work by this group and others have shown a significant role for DIII.- The authors provide a comparison between different species to demonstrate the conservation of the amino acids observed in this study. It would be valuable to compare across Nav channel isoforms that inactivate if the mechanism is thought to be universal, as suggested.- Some caveats should be included. Without identifying the coupling pathway, it is possible that the mutation of the S6 residues interrupts the normal inactivation pathway even though these positions might not directly participate. Further work is needed to convincingly make definitive statements about the participation of these positions in the inactivation gate. As noted in the manuscript and shown in refs 40 and 42, DI mutations have been shown to have similar effects on inactivation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 476, 243, 490]]<|/det|>
+Minor Comments  
+
+<|ref|>text<|/ref|><|det|>[[115, 505, 866, 566]]<|/det|>
+- The different treatment of showing DIA and DIIAA data in figure 7 is confusing. Why not plot DIIAA GV and inactivation curves along with DIA?- It would be helpful to include hypotheses relating to the coupling of S6 helices to IFM binding that would cause inactivation via the intracellular S6 residues that were identified.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[115, 90, 444, 107]]<|/det|>
+## RESPONSE TO THE REVIEWERS' COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[115, 123, 872, 210]]<|/det|>
+We thank the reviewers for their insightful comments. The modifications they proposed have made the manuscript much better. All the changes in the text are marked in yellow and they include the changes as a result of the reviewers' comments, the changes required by the Editor, in addition to other small corrections. We also include an extra Supplementary File with changes in yellow and the file is labeled MARK.  
+
+<|ref|>text<|/ref|><|det|>[[116, 226, 393, 243]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[114, 258, 876, 513]]<|/det|>
+This is an important paper. The cryo- EM structures of vertebrate voltage- dependent sodium channels have shown that the long- held model of channel inactivation as physical occlusion of the pore by the IFM motif in the III- IV intracellular linker is not correct, because the IFM motif was not in the pore but rather buried in a pocket fairly distant from the pore. However, the structures do not give much insight into exactly how binding of the IFM motif in this pocket produces a closed, inactivated channel. Here, the authors use the structures of inactivated channels to recognize two rings of hydrophobic residues that form a "double- seal" near the inner moth of the pore. Using this modeled structure, they then show that mutating a residue in each ring results in an inactivated channel that is no longer fully closed but instead is leaky, thus offering strong support for their proposed mechanism. This is probably the single most remarkable result from mutagenesis of an ion channel that I know, because it would have been impossible to come up with this combination of mutations without thinking deeply about the structure. Even more remarkably, the authors find that the leaky channel formed by these mutations has an altered ionic selectivity, with reduced selectivity for Na over K. This implies that there is some sort of coupling between this region near the inner pore and the selectivity filter of the channel located near the outer pore.  
+
+<|ref|>text<|/ref|><|det|>[[115, 512, 867, 599]]<|/det|>
+Besides presenting fascinating and very important results, the paper is a technical tour- de- force of channel biophysics, using well- designed measurements of gating kinetics, gating current, and ionic selectivity in the mutated channels to lead the authors to their conclusions and new model. The presentation of the motivation for the experiments and the discussion of their interpretation are done with clarity and concision. I can find only minor points of wording that might be improved.  
+
+<|ref|>text<|/ref|><|det|>[[115, 615, 844, 652]]<|/det|>
+We appreciate the reviewer positive feedback and the recognition of the importance of our findings.  
+
+<|ref|>text<|/ref|><|det|>[[115, 670, 163, 684]]<|/det|>
+Minor  
+
+<|ref|>text<|/ref|><|det|>[[115, 686, 866, 736]]<|/det|>
+13 "The hinged- lid model is long accepted as the canonical model for fast inactivation" would be accurate as "13 The hinged- lid model was long accepted....". It has now been 6 years that the cryo- EM structures showed that the model is incorrect.  
+
+<|ref|>text<|/ref|><|det|>[[115, 737, 755, 754]]<|/det|>
+We thank the reviewer for spotting this typo and we have changed it in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 755, 504, 770]]<|/det|>
+44 "positively charge" should be "positively charged"  
+
+<|ref|>text<|/ref|><|det|>[[115, 771, 717, 787]]<|/det|>
+We thank the reviewer for spotting this and we have changed it in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 788, 870, 908]]<|/det|>
+57 "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif, though docking into a hydrophobic pocket, resided far away from the pore in the putative inactivated state to block the permeation path, in contrast to the predictions of the canonical "hinged- lid" model." I found this sentence a little difficult to interpret. Maybe the meaning would be clearer as "Upon the first mammalian Nav channel structure (22) it became clear that the IFM motif docked into a hydrophobic pocket far away from the pore rather than physically occluding the pore, in contrast to the predictions of the canonical "hinged- lid" model."
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 884, 310]]<|/det|>
+We agree with the reviewer that by stating that particular sentence in the suggested way makes it clearer to understand. We have rephrased this particular sentence in the manuscript. 104 "Given the fact that both structures are determined at OmV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore resemble a fast inactivated state." Since channels in cryoEM have been at \(0\mathrm{mV}\) for a long time, I would think the structure is more likely to be a slow inactivated state rather than fast inactivated state. It does not seem necessary or useful to get into a discussion of this in the present paper, but the sentence might be slightly more accurate as "Given the fact that both structures are determined at OmV and the DIV VSD is in the "up" conformation, it is reasonable to assume that the pore is in an inactivated state." We agree with the reviewer that it is not necessary to discuss the exact state the structures were captured in. We have rephrased that particular sentence as suggested. 439 "In DIII and DIV, none of the single alanine mutation yielded..." should be "mutations" We thank the reviewer for spotting this typo and we have changed it in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[116, 328, 393, 344]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[114, 361, 878, 514]]<|/det|>
+Liu et al investigate whether hydrophobic residues lining the intracellular pore are part of the inactivation gate of Nav channels. Given that Nav channel inactivation is essential to the function of excitable cells, insight into this mechanism is of high interest. Previous work suggests that an IFM motif located within the intracellular DIII- DIV linker of Nav channels directly blocks the channel pore to induce inactivation. However, recent Nav channel structures show that the binding of the IFM motif is far from the pore, even though it is required for inactivation. The authors demonstrate that inactivation of the channel is impaired by mutation of several intracellular pore residues on the DIII and DIV S6 segments, preventing inactivation even though IFM is still intact, suggesting that these residues are part of the inactivation mechanism.  
+
+<|ref|>text<|/ref|><|det|>[[115, 530, 876, 582]]<|/det|>
+The manuscript is well- written and delivers a hypothesis for a new mechanism for the fast inactivation voltage- gated sodium channels. The experimental results shown provide evidence for the connection of intracellular S6 helices to an inactivation gate.  
+
+<|ref|>text<|/ref|><|det|>[[116, 599, 673, 616]]<|/det|>
+We thank the reviewer for the thorough evaluation of our manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 635, 249, 650]]<|/det|>
+## Major Comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 652, 880, 701]]<|/det|>
+- Voltage clamp fluorometry is likely to provide higher resolution data on whether the VSDs are being affected, including changes in activation and deactivation kinetics. It is surprising that these experiments were not included.  
+
+<|ref|>text<|/ref|><|det|>[[114, 702, 877, 855]]<|/det|>
+We believe gating current measurement is more suitable for this particular work because it allows us to directly assess charge immobilization, which is a critical aspect of inactivation. Additionally, by obtaining robust gating current recordings in all mutants with a high signal-to-noise ratio (which is not the case with fluorometry), we can accurately analyze the kinetic properties. While site-directed voltage clamp fluorometry (presently ongoing) may provide valuable information on individual VSD movements, our focus in this manuscript is primarily on understanding the relationship between the IFM motif and the inactivation gate. As such, measuring gating currents aligns with the traditional approach for characterizing charge immobilization and enables a comprehensive analysis of the inactivation mechanism.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[112, 90, 864, 125]]<|/det|>
+- The addition of the IQM to DIIAA was helpful in assessing how these different channel components interact. Are the results with IQM added to DIVAA similar?  
+
+<|ref|>text<|/ref|><|det|>[[115, 125, 840, 159]]<|/det|>
+Interesting point raised by this reviewer. We did not try IQM_DIVAA because, unfortunately, DIVAA express much less than DIIAA and IQM also decreases expression significantly.  
+
+<|ref|>text<|/ref|><|det|>[[115, 159, 864, 193]]<|/det|>
+- The model proposed in Figure 8 is very DIV centric even though the data presented in this work and previous work by this group and others have shown a significant role for DIII.  
+
+<|ref|>text<|/ref|><|det|>[[115, 194, 879, 244]]<|/det|>
+The reviewer is correct to point the involvement of DIII on inactivation. The model is clearly a simplification but, to correct this oversight, we have added a note in the figure legend indicating that DIII is also participating in inactivation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 244, 875, 294]]<|/det|>
+- The authors provide a comparison between different species to demonstrate the conservation of the amino acids observed in this study. It would be valuable to compare across Nav channel isoforms that inactivate if the mechanism is thought to be universal, as suggested.  
+
+<|ref|>text<|/ref|><|det|>[[115, 294, 880, 396]]<|/det|>
+We thank the reviewer for pointing this out. We agree that while comparing amino acid sequence across different species could provide evolutionary insights, it still remains valuable to include a sequence comparison across different isoforms of Nav channels to demonstrate similarities and special adaptations within them. We have included another supplementary figure (SuppleFig. 2B). It is true that not all the identified residues are conserved across all isoforms. However, the property of the amino acids at those positions are relatively conserved.  
+
+<|ref|>text<|/ref|><|det|>[[115, 396, 876, 481]]<|/det|>
+- Some caveats should be included. Without identifying the coupling pathway, it is possible that the mutation of the S6 residues interrupts the normal inactivation pathway even though these positions might not directly participate. Further work is needed to convincingly make definitive statements about the participation of these positions in the inactivation gate. As noted in the manuscript and shown in refs 40 and 42, DI mutations have been shown to have similar effects on inactivation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 481, 875, 583]]<|/det|>
+We would like to respectfully argue against this comment. Indeed, mutations in DI have been shown to disrupt fast inactivation. L437C together with A438W (CW) in particular has been shown to remove fast inactivation almost completely. Considering the location of those residues at S6, it is tempting to conclude that those residues might form the fast inactivation gate. However, we have recorded the charge immobilization profile of CW (unpublished results\*) and we discovered that it has been significantly altered which is not case for the DIIAA and DIVAA mutants described here.  
+
+<|ref|>text<|/ref|><|det|>[[115, 583, 880, 667]]<|/det|>
+We agree that without the complete knowledge of the entire coupling pathway, there exists a possibility that the identified residues might exert their effects by interrupting the pathway. However, we believe that the existence of a leaky inactivated state provides strong evidence that support our interpretation that the S6 residues identified are the fast inactivation gate and we are not actually interrupting the inactivation pathway (as IQM or CW might be acting).  
+
+<|ref|>sub_title<|/ref|><|det|>[[116, 685, 245, 700]]<|/det|>
+## Minor Comments  
+
+<|ref|>text<|/ref|><|det|>[[115, 716, 875, 750]]<|/det|>
+- The different treatment of showing DIA and DIIAA data in figure 7 is confusing. Why not plot DIIAA GV and inactivation curves along with DIA?  
+
+<|ref|>text<|/ref|><|det|>[[115, 751, 880, 870]]<|/det|>
+It is challenging to acquire an accurate GV curve for DIIAA. The mixture of gating current and ionic current at the beginning of the trace makes it unfeasible to reliably obtain the peak for the ionic current. While it is possible to record the gating current together with the ionic current first then apply external TTX to record only the gating current subsequently, this method is highly sensitive to small fluctuation within the system (rundown, changes in capacitance, temperature etc.,). Since, the main conclusion of this work rely mostly on results on DIIAA and DIVAA, we decided to describe the effects of DIIAA qualitatively and leave room for future investigations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 870, 847, 904]]<|/det|>
+- It would be helpful to include hypotheses relating to the coupling of S6 helices to IFM binding that would cause inactivation via the intracellular S6 residues that were identified.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 878, 158]]<|/det|>
+We thank the reviewer for this comment. Indeed, we think that there is a chain of amino acids that couple the IFM binding to the closing of the inactivation gate. We have hypothesized the residues that form this chain, and it is the subject of our future research, but as we do not have results yet, we believe that to explicitly state it is premature.  
+
+<|ref|>text<|/ref|><|det|>[[135, 308, 872, 460]]<|/det|>
+\*unpublished results have been communicated as an abstract. Yichen Liu & Francisco Bezanilla "Charge immobilization in two different fast- inactivation- removed Nav1.4 sodium channel mutants", Biophysical Journal, Volume 121, Issue 3, Supplement 1, 2022, Pages 95a- 96a, ISSN 0006- 3495, https://doi.org/10.1016/j.bpj.2021.11.2252.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d751de784711fe985e21fc71a4089fbbdcb890f642dcc7aa3028ff9a5fdfca83/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d751de784711fe985e21fc71a4089fbbdcb890f642dcc7aa3028ff9a5fdfca83/supplementary_0_Peer Review File__d751de784711fe985e21fc71a4089fbbdcb890f642dcc7aa3028ff9a5fdfca83.mmd

@@ -0,0 +1,284 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Introducing dorsoventral patterning into adult regenerating lizard tails with gene- edited embryonic neural stem cells.  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  
+
+Remarks to the Author:  
+
+In this manuscript Lozito et al determine the molecular signals that inhibit ependymal tube regeneration in an adult lizard. They use CRISPR technology to knock- out the signal inhibitory to correct dorsoventral patterning of the ependymal tube in cultured neurospheres and through transplantation experiments so this suffices to induce ependymal tube regeneration after tail amputation in an adult lizard.  
+
+Overall it is a very interesting manuscript that gives novel insights into the roadblocks to regeneration in the lizard and shows how one of these can be overcome but does not give any information on other pathways that may also limit the obstacle of improving neural regeneration at the expense of cartilage regeneration. The manuscript focuses heavily on shh signalling but does not consider the pathways like Fgf, Wnt and BMP that will established from in vivo work in a range of organisms to play important roles in patterning the neural tube and directing neuronal differentiation.  
+
+The manuscript is very dense and in places difficult to read, many figure need better labelling as there are many panels under one label.  
+
+Specific points:  
+
+1. Figure shows that markers of dorsoventral patterning are expressed during embryogenesis but not during tail regeneration. Pax6 is a marker of lateral tissue, here is appears to be also expressed in the dorsal domain (fig.1 F). Is there overlap in the cells that express Pax6 and Pax7 in the embryos? 
+2. In fig.1 H, J adult ET shh, a ventral marker appears to be expressed in the dorsal region? Are these panels inverted? Are these cells sox2 positive? 
+3. Are cells in the ependymal tube in the adult or in the regenerating tail slowly dividing. I would be interesting to look at EdU incorporation in these cells. 
+4. Where are the neuronal cell bodies located in the adult, are there cells which express NeuN and Sox2? NeuN is often found expressed in neural stems that have begun the program of differentiation. 
+5. Are all cells in the adult or regeneration ependymal tube GFAP and sox2 positive? Are other markers of astrocytes expressed in these cells? Are markers of glial scar formation expressed in the regenerating ependymal tube cells after injury? 
+6. Do the cells when cultured default to a certain identify because they are missing secreted factors like wnts ? 
+7. Fig.3 & 4 could be combined into one figure and the text describing them could be shortened. 
+8. Figure 5 is a nice technique but this could be moved to supplementary data, it seems odd here and disrupts the flow of the manuscript. It is unclear why the authors did not continue to use this technology to label cells for further experiments, it is better than DiI labelling neurospheres. 
+9. In the neurosphere cultures that authors report they do not see expression of the neuronal marker Tuj1, in the adult derived neurospheres this is a late marker. It is important to determine if the cells begin the neuronal differentiation program but cannot terminally differentiate. 
+10. The authors use nice clever modern technology to knockout Smo and generate gene edited neurospheres which they implant into the adult tail and assay the outcome of regeneration. The limited data shown suggests that this allows better patterning of the regenerated ependymal tube and differentiation of neurons. What percentage of the regenerated neural tube is derived from the implanted cells? What percentage of cells become neurons? How many animals was this tested on? Do the animals regenerate the same length of tail that was amputated ?  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+In this study the authors have performed a detailed analysis of the role of Hedgehog signalling during lizard tail regeneration. They have compared this to developmental roles including a detailed analysis
+
+<--- Page Split --->
+
+of neural stem cells. Remarkably they have been able use this knowledge to improve upon tail regeneration by restoring aspects of dorsal/ventral patterning that are normally absent in regenerated tails. The study is also impressive in its development of many seemingly novel techniques for the analysis of regenerative biology in lizards. Over all I would strongly support publication. Here are my comments:  
+
+1) Figure 1 HJK - Isn't the expression of Hedgehog more widespread than discussed in the results?  
+
+2) Figure 5 - Aren't there more RFP+ cells than just in the dorsal root ganglion and neural tube? If these are derived from sox2-cre expressing cells shouldn't they also be neurons?  
+
+3) More quantification is needed. The quantification in the supplemental data is impressive, but more numbers should be provided for each figure. For example \(N = 6\) for figure 5 does not provide enough information.  
+
+4) Some of the figures are very complicated and difficult to follow. There are useful diagrams to show the method, please also add summary diagrams to illustrate your interpretation/model. Doing this for Figure 5 to 8 would be good.  
+
+5) Please go over the text to improve the logical flow: you should avoid using the word "Next we" to introduce an experiment. Recap and introduce your hypothesis before describing the next experiment. Th authors have done this very well for fig 8/pg 10.  
+
+6) Consider capitalising "Hedgehog" when referring to the pathway.  
+
+7) Isn't there also a neural crest contribution to dorsal root ganglion? Shouldn't this be considered in the interpretation of the results?  
+
+8) It should be pointed out that neural mesodermal progenitors (NMPs) have been identified in other models and these cells are thought to be embryonic tail stem cells capable of generating both neural and mesodermal derivatives. NMPs co-express tbx genes with sox2. It would be good to consider the role of NMPs by including tbx gene analysis in future studies.  
+
+Henry Roehl  
+
+Reviewer #3:  
+
+Remarks to the Author:  
+
+This is an exciting manuscript by Lozito et al showing that a loss of dorsal- ventral patterning during tail regeneration in lizards can be re- activated through CRISPR/Cas9 gene editing. This would seem first glance as a niche finding, but it actually has broad implications for regenerative medicine. It is an exciting example of enhancing an in vivo regenerative response in an otherwise limited regenerative capacity. This should have broad interest to researchers working on spinal cord regeneration, but also people investigating non neural regeneration that may use similar strategies to promote regeneration in other organs. It is one of very few examples showing enhancement of regeneration in vivo.  
+
+Noteworthy results include the de novo dorsal pattern in an amniote that would otherwise not generate this pattern during spinal cord regeneration. It is of interest that shh expression is the default state of the regenerating tail, which can even overcome embryonic dorsal neural tube cells. I found it particularly striking that the dorsal identity of smoothened KO NSCs generated what looked to be very large DRG structures. To me, this is a significant feat to generate a DRG that otherwise would not be generated. It will be important in the future (not here) to understand whether this DRG innervates the appropriate targets and sends processes back to the spinal cord.
+
+<--- Page Split --->
+
+There are some small, unfortunate findings that limit the impact (although very high) such as the observation that the smo KO NSCs do not generate CNS neurons and that a double neural tube is generated after engraftment. Regardless, the manuscript clearly shows that using gene editing can make a NSC competent to participate in proper patterning for spinal cord regeneration, which is an important feet for future therapeutics.  
+
+If I was to have one critique on the methodology, the manuscript relies heavily on a single staining modality, IHC instead of incorporating qPCR or FISH as second forms of validation. My concern for this is dampened because the proteins the authors are assaying are known to have specific, robust expression patterns in the neural tube and the staining throughout the manuscript is very specific. Therefore, for this particular case the single staining approach is sufficient.  
+
+Overall, this manuscript was a logical and clear. Although the final outcome would have been more impactful if a single neural tube was formed and CNS differentiation occurred after stem cell transplant, this is indeed an exciting advance using clever techniques for tissue grafting and lineage tracing.  
+
+Below are some minor comments and mistakes I found in the manuscript:  
+
+Page 3, line 13 - Change NCS to NSC  
+
+Page 10, line 9- 17: Is there any indication why the embryo- derived NSCs/cyclopamine- treated cells don't incorporate into the existing ependymal tube? I would think that it would just generate a dorsal portion of the existing ET. This is quite interesting and may be worth adding a sentence in the results section as it read like an incomplete explanation for what is going on here. Is it because the entire animal is treated with cyclopamine?  
+
+In Figure 8H- O, why are most of the GFP+ cells sox2- ? Does removing Shh signaling make them differentiate?  
+
+Page 12, line 3: I think the phrase "exhibited Col2+ cartilage rods dorsal to ETs" was meant to say ventral instead of dorsal  
+
+James Monaghan  Department of Biology  Institute for Chemical Imaging of Living Systems  Northeastern University  Boston, MA 02115
+
+<--- Page Split --->
+
+We thank the reviewers for their insightful and thoughtful feedback. This document addresses the concerns and comments that were brought up by the reviewers.  
+
+Reviewer #1 (Remarks to the Author):  
+
+1. Figure shows that markers of dorsoventral patterning are expressed during embryogenesis but not during tail regeneration. Pax6 is a marker of lateral tissue, here is appears to be also expressed in the dorsal domain (fig.1 F). Is there overlap in the cells that express Pax6 and Pax7 in the embryos?  
+
+Thank you for this observation, which is spot on. There appears to be some overlap between \(\mathrm{Pax7^{+}}\) and \(\mathrm{Pax6^{+}}\) cells in lizard tail neural tubes, an observation that has been added to the results section of the revised manuscript. We suspect this to be indicative of differences in cell origins and patterning of primary vs secondary neural tubes. Indeed, in embryonic lizard trunks, primary neural tubes (which originate from neural plates) exhibit more distinct lateral \(\mathrm{Pax6}\) expression. In contrast, secondary neural tubes of embryonic lizard tails (which originate from tail bud neural mesodermal progenitors) exhibit a dorsal shift in \(\mathrm{Pax6}\) expression, as observed in Fig. 1F. We have begun comparing primary vs secondary lizard neural tubes by single cell RNA sequencing to further investigate differences in cell identity and patterning that might affect divergent healing capabilities. This will be the subject of a future manuscript.  
+
+2. In fig.1 H, J adult ET shh, a ventral marker appears to be expressed in the dorsal region? Are these panels inverted? Are these cells sox2 positive?  
+
+Thank you for your observation. These images are oriented correctly and depict one of the most important findings of this study; adult original and regenerated tail ependyma/NSC populations exhibit floor plate markers such as Shh completely around central canal circumferences, including dorsal regions. Some transverse sections of adult original tail ependyma, such as those depicted in Fig. 1 H, J, exhibit patchier Shh expression than regenerated tail sections, possibly due to the presence of Shh+ Tuj+ Sox2 neurons. These points have been added to the manuscript text in the revision.  
+
+3. Are cells in the ependymal tube in the adult or in the regenerating tail slowly dividing. I would be interesting to look at EdU incorporation in these cells.  
+
+We thank the reviewer for their interest, and these EdU/proliferation studies have been previously published by our group (Sun et al. PNAS 2018, Figure S6). For convenience, we included a simplified version of the published figure below. 5- ethynyl- 2'- deoxyuridine (EdU) incorporation/staining assays were used to visualize proliferating \(\mathrm{Sox2 + }\) NSC populations in original and regenerating (14, 28, and 56 DPA) lizard tails along their lengths (proximal, middle, and distal). 14- DPA samples, which correspond to the blastema stage of regeneration, exhibited the most numbers of proliferative NSCs. 28- DPA samples, which correspond to lengthening tails, exhibited higher numbers of proliferative NSCs in distal regions. 56- DPA samples, which correspond to fully regenerated tails, resembled original tails and exhibited low levels of proliferating cells.  
+
+![PLACEHOLDER_4_0]
+
+
+<--- Page Split --->
+
+## 4. Where are the neuronal cell bodies located in the adult, are there cells which express NeuN and Sox2? NeuN is often found expressed in neural stems that have begun the program of differentiation.  
+
+Thank you for the suggestion. To address this comment, we analyzed cross sections of embryonic, adult, and regenerated lizard tails with NeuN, Sox2, and TuJ1 immunofluorescence. NeuN signal was detected in nuclei of TuJ1+ neurons in interior regions of original adult spinal cords, but not in Sox2+ ependymal cells. NeuN signal was not detected in regenerated tail spinal cord, corroborating our observations that no new neurons are formed during tail regrowth. Sox2/NeuN co- expression was only observed in tail neural tube cells, indicating spinal cord neurogenesis was restricted to embryonic stages. This new data is presented in new Supplementary Fig.4.  
+
+![PLACEHOLDER_5_0]
+  
+
+## 5. Are all cells in the adult or regeneration ependymal tube GFAP and sox2 positive? Are other markers of astrocytes expressed in these cells? Are markers of glial scar formation expressed in the regenerating ependymal tube cells after injury?  
+
+To address this comment, we quantified the percentages of GFAP+ ependymal cells in original and regenerated tails that expressed Sox2. Cross sections of regenerated tail ependymal tubes were immunostained for GFAP and Sox2, and individual cells were identified with DAPI staining. Keyence microscope software was used to quantify the percentages of ependymal tube cells that exhibited GFAP and/or Sox2 signal. Ten separate tails/animals were assayed, and average percentages \(+ / -\) standard deviations were calculated. \(94.3 + / - 3.8\%\) of GFAP+ original tail ependymal cells were positive for GFAP, while \(95.2 + / - 4.1\%\) of GFAP+ ET cells were positive for Sox2. This data is presented in new Supplementary Fig. 9.  
+
+To determine other markers expressed by regenerated tail NSCs, we referenced our data sets from single cell RNA sequencing data analyses of lizard ependymal tubes and spinal cords and neurospheres, included in the figure below. Regenerated tail GFAP+ Sox2+ NSC clusters co- expressed radial glial markers FABP7 and Sox9. Original tail spinal cord data sets exhibited distinct clusters of GLUL+ astrocytes and Sox10+ MBP+ CLDN11+ oligodendrocytes. These clusters were not represented in ependymal tube or neurosphere data sets, and regenerated tail NSCs did not express other astrocyte and oligodendrocyte markers, confirming their radial glia/ependymal identity. We also did not observe genes commonly associated with mammalian glial scar formation (CSPGs, laminin, tenascin, fibronectin, etc.) in ependymal tube cell transcriptomes. Similarly, we did not observe pro- inflammatory cytokines (CCL2, CCL5, CXCL8 etc.) in regenerated tail spinal cord pericytes. Future work will be aimed at comparing scar- associated genes in regenerating lizard spinal cord cell populations with those expressed during other models of spinal cord injury/healing. We suspect the regenerated tail environment somehow inhibits glial and pericyte scaring processes characteristic of trunk spinal cord injury.  
+
+![PLACEHOLDER_5_1]
+
+
+<--- Page Split --->
+
+## 6. Do the cells when cultured default to a certain identify because they are missing secreted factors like wnts ?  
+
+We thank the reviewer for their interest and suggestion. To answer this question, we referenced our single cell RNA sequencing data sets to compare growth factor expression in original tail spinal cord, regenerated tail ependymal tube, and cultured neurosphere NSCs. General trends indicated higher Wnt expression (Wnt1, Wnt5a, Wnt11) in ependymal tubes than in original spinal cords and neurospheres. Interestingly, cultured neurospheres expressed high levels of Wnt5b. Conversely, BMP expression (BMP2, BMP3) was higher in original spinal cord and undetectable in ependymal tubes and neurospheres. Finally, we treated cultured neurospheres with Wnt1 and Wnt5a during differentiation assays and did not detect roof plate or neuronal differentiation. In summary, we concluded that environmental factors such as growth factor signaling do not prevent adult lizard tail NSCs from undergoing neurogenesis during tail regrowth. Instead, perhaps NSCs undergo epigenetic changes between embryonic development and adult tail regeneration that inhibit neural differentiation, and future work will explore this topic. We are currently preparing our single cell RNA sequencing results for a separate publication.  
+
+## 7. Fig.3 & 4 could be combined into one figure and the text describing them could be shortened.  
+
+Figures 3 and 4 of the original manuscript have been combined into a single figure (Revised Figure 3), and the text has been adjusted and shortened accordingly.  
+
+## 8. Figure 5 is a nice technique but this could be moved to supplementary data, it seems odd here and disrupts the flow of the manuscript. It is unclear why the authors did not continue to use this technology to label cells for further experiments, it is better than Dil labelling neurospheres.  
+
+We appreciate the reviewer's suggestions and have moved old Figure 5 to Supplementary Data (Supplementary Fig. 8)  
+
+Our choice to use Dil labelling rather than the lentivirus system to trace cell lineages in neurosphere implantation studies was due to the time and supply costs associated with lentivirus production and use coupled with the large number of NSCs needed for implantation. For example, experiments comparing embryonic and adult NSC behavior post- implantation, such as those presented in revised Figure 6), required increased numbers of neurospheres due to limitations in adult NSC propagation in vivo. Unlike embryonic NSCs, adult NSCs exhibited diminished proliferative potential beyond passage 5, meaning that more lizards had to be used to generate the neurospheres required for these experiments. Furthermore, we didn't use the lentiviral system to label Control- CRISPR cells in gene editing experiments in order to provide the most appropriate controls for Smo- CRISPR cells without confounding effects caused by genomic insertion of lentiviral constructs. We had previously reported on Dil labeling as an effective and cost- efficient means for labeling large numbers of lizard neurospheres with minimal effects on NSC behavior and genetics, filling the requirements for the studies included in this manuscript.  
+
+## 9. In the neurosphere cultures that authors report they do not see expression of the neuronal marker Tuj1, in the adult derived neurospheres this is a late marker. It is important to determine if the cells begin the neuronal differentiation program but cannot terminally differentiate.  
+
+We thank the reviewer for their insight and suggestion. To address these comments, we analyzed embryonic and adult NSC- derived neurospheres for expression of doublecortin (DCX) and NeuroD2 (early neuronal differentiation markers) by real- time RT PCR. The results of these experiments are presented in revised Supplemental Fig. 11 and are included below for easy referencing. mRNA was collected from either undifferentiated neurospheres or neurospheres cultured under differentiated conditions for 14 days treated with either vehicle control, cyclopamine, or SAG and assayed for Sox2, DCX, NeuroD2, GFAP, and Tuj1 expression. Each experimental condition involved pooled mRNA from 10 neurospheres and was repeated 3 times. Results are presented as fold changes compared to undifferentiated conditions. Results showed that differentiation caused significant increases in DCX, neuroD1, and Tuj1 in embryonic tail, but not adult tail, NSC- derived neurospheres. These increases were unaffected by cyclopamine treatment, but SAG treatment caused significant decreases in DCX, NeuroD2, and Tuj1 expression. In contrast, adult lizard NSC expression of Sox2, GFAP, DCX, NeuroD2, and Tuj1 expression were unaffected by either differentiation or Hedgehog signaling
+
+<--- Page Split --->
+
+modulation. Future work will be aimed at forcing expression of pro-neurogenesis factors like NeuroD2 to attempt initiating neurogenesis in adult lizard NSCs.  
+
+![PLACEHOLDER_7_0]
+  
+
+10. The authors use nice clever modern technology to knockout Smo and generate gene edited neurospheres which they implant into the adult tail and assay the outcome of regeneration. The limited data shown suggests that this allows better patterning of the regenerated ependymal tube and differentiation of neurons. What percentage of the regenerated neural tube is derived from the implanted cells? What percentage of cells become neurons? How many animals was this tested on? Do the animals regenerate the same length of tail that was amputated?  
+
+We thank the reviewer for their inquiries. First, we quantified the contribution of implanted embryonic NSCs to adult regenerated tail dorsal and ventral ET regions. These results are presented in the revised manuscript in Supplementary Fig. 16 and are included below for reference. Smo KO NSCs labeled with GFP and control NSCs pre- labeled with Dil were engrafted to dorsal ependyma of amputated tail spinal cords. Following 28 days of regrowth, regenerated tails were collected and processed for histology. ET cross sections were analyzed by histology/IF/fluorescence microscopy for GFP/Dil signal and Sox2 expression every 1 mm along entire tail lengths. Horizontal lines drawn through the centers of ETs bisected tube images into dorsal and ventral regions. Quantification of dorsal and ventral Sox2 and GFP/Dil signal areas were performed for each cross- section along tail lengths. Percentages of Sox2+ IF signal areas co- expressing GFP/Dil were calculated for dorsal and ventral ET regions and are presented in the histograms below. Dorsal ET regions exhibited significantly higher GFP/Dil signal from labeled exogenous NSC populations than ventral regions, indicating that starting NSC spatial arrangements were maintained along regenerated ET lengths.  
+
+![PLACEHOLDER_7_1]
+
+
+<--- Page Split --->
+
+Next, we quantified the percentages of exogenous embryonic NCS undergoing neurogenesis following incorporation in adult regenerating tails. These results are presented in the revised manuscript in Supplementary Fig. 17 and are included below for reference. GFP- labeled Smo KO NSCs and control NSCs pre- labeled with Dil engrafted to dorsal spinal cord ependyma cell populations. Following 28 days of regrowth, regenerated tails were collected and processed for histology. ET cross sections were analyzed by IF/fluorescence microscopy for GFP/Dil signal and Tuj1 expression every 1 mm along entire tail lengths. 10 different animals/tails were analyzed for each condition. Tuj1 and GFP/Dil signal areas were quantified for each cross- section, and percentages of Tuj1+ IF signal areas co- expressing GFP/Dil are presented in the histograms below. Tuj1+ neural cells co- expressed GFP in lizards treated with Smo KO NSCs, indicating significantly higher levels of neurogenesis in Smo KO NSCs than control cells. \*, p<0.001 compared to corresponding Control NSC condition for each measurement.  
+
+![PLACEHOLDER_8_0]
+  
+
+Finally, we compared the tail lengths regenerated by lizards treated with Smo KO NSCs vs Control NSCs vs untreated controls. Both groups of tails regrew to approximately 1.2 cm 28 DPA. Differences in tail lengths were insignificant between conditions, indicating that incorporation of exogenous embryonic NSCs did not affect regenerated tail lengths. The revised text has been updated with these observations. Furthermore, we emphasized the numbers of animals used in each experiment in the revised text and figure legends.  
+
+  
+
+## Reviewer #2 (Remarks to the Author):  
+
+In this study the authors have performed a detailed analysis of the role of Hedgehog signalling during lizard tail regeneration. They have compared this to developmental roles including a detailed analysis of neural stem cells. Remarkably they have been able use this knowledge to improve upon tail regeneration by restoring aspects of dorsal/ventral patterning that are normally absent in regenerated tails. The study is also impressive in its development of many seemingly novel techniques for the analysis of regenerative biology in lizards. Over all I would strongly support publication. Here are my comments:  
+
+1) Figure 1 HJK - Isn't the expression of Hedgehog more widespread than discussed in the results?  
+
+We appreciate this note and have also noted that Shh protein is detected among spinal cord nerves surrounding original tail ependyma. The revised results section has been updated to reflect these observations.  
+
+2) Figure 5 - Aren't there more RFP+ cells than just in the dorsal root ganglion and neural tube? If these are derived from sox2-cre expressing cells shouldn't they also be neurons?  
+
+This is a good question and, to answer it, we re- analyzed the images from our lineage tracing experiments with Keyence microscope software to quantify the percentage overlap of RFP signal with Sox2, GFAP, and TUJ1 signals. These results are presented in revised Supplementary Fig. 9. Overall, they show that the vast majority of RFP+ cells (>98%) co- localize with Sox2, GFAP, or TUJ1 regions within either neural tubes, developing and
+
+<--- Page Split --->
+
+adult spinal cords or DRG, indicating that most labeled embryonic NSCs differentiate into tail ependymal cells, glia, or neurons. For example, in embryonic samples, the majority of \(\mathrm{Tuj1 + }\) signal was observed in peri-neural tube regions rather than neural tubes themselves. We hypothesized that these regions were the sites populated by early neurons that would later form tail spinal cord. Interestingly, adult peripheral nerves were only sporadically labeled with RFP using this system, perhaps indicating additional cell sources of tail peripheral nerves. The revised text has been updated with these observations.  
+
+3) More quantification is needed. The quantification in the supplemental data is impressive, but more numbers should be provided for each figure. For example \(N = 6\) for figure 5 does not provide enough information.  
+
+We have taken this suggestion to heart and have quantified immunofluorescence findings presented in Figures 5, 6, 7, 8, and the results are depicted in the revised Figures 4, 6, 8 and in newly added Supplemental Fig. 9, 11, 16, and 17. Specifically, Keyence microscope software was used to quantify percentages of \(\mathrm{RFP + }\) or \(\mathrm{GFP + }\) /Dil+ regions that co- express Sox2, GFAP, TUJ1, Shh, and Pax7. We also provided better descriptions of the animal and sample numbers associated with each experiment.  
+
+4) Some of the figures are very complicated and difficult to follow. There are useful diagrams to show the method, please also add summary diagrams to illustrate your interpretation/model. Doing this for Figure 5 to 8 would be good.  
+
+Summary diagrams have been added to revised Figures 5 and 7 to illustrate our interpretation of results from lineage tracing and gene- edited NSC transplantation, respectively.  
+
+5) Please go over the text to improve the logical flow: you should avoid using the word "Next we" to introduce an experiment. Recap and introduce your hypothesis before describing the next experiment. The authors have done this very well for fig 8/pg 10.  
+
+The text has been edited to improve logical flow and to highlight hypothesis- driven experiments.  
+
+6) Consider capitalising "Hedgehog" when referring to the pathway.  
+
+We have capitalized "Hedgehog" when referring to the pathway throughout the revised text.  
+
+7) Isn't there also a neural crest contribution to dorsal root ganglion? Shouldn't this be considered in the interpretation of the results?  
+
+We thank the reviewer for this note as it touches upon several interesting aspects of our study considering the unique case of the lizard tail, the only amniote tail that contains both spinal cord and DRG. In the trunks of amniote embryos, folding of \(\mathrm{Sox2 + }\) neural plates gives rise to lateral and floor domains of primary neural tubes, while neural plate borders pinch off and contribute to neural tube roof plate domains and \(\mathrm{Sox2 - }\) neural crest. Trunk neural crest cells migrate mediolaterally and begin to express \(\mathrm{Sox2}\) before forming DRG. The situation is very different in amniote tails, and even more exceptional in lizard tails. Tails are formed from tail buds, which do not have neural plates or neural folds, and hence, no neural crest cells. Instead, \(\mathrm{Sox2 + }\) neural mesodermal progenitors (NMPs) derived from \(\mathrm{Sox - }\) tail bud cells directly differentiate into secondary neural tubes. Mammalian secondary neural tubes degenerate before DRG formation during tail development, and adult mammal tails contain neither DRG nor spinal cord. In contrast, lizard secondary neural tubes survive tail development, and lizards are the only amniotes that exhibit both spinal cords and DRG in adult tails. The origins of amniote tail DRG has not been specifically studied, but lineage tracing data presented here suggest that lizard tail DRG derive from \(\mathrm{Sox2 + }\) tail bud cells that pass through a neural tube intermediate. These observations have been added to the Discussion section of the manuscript, and hopefully future work will expand on this topic.  
+
+8) It should be pointed out that neural mesodermal progenitors (NMPs) have been identified in other models and these cells are thought to be embryonic tail stem cells capable of generating both neural and mesodermal derivatives. NMPs co-express tbx genes with sox2. It would be good to consider the role of NMPs by including tbx gene analysis in future studies.
+
+<--- Page Split --->
+
+We thank the reviewer for this wonderful observation, and we certainly plan on looking into the relationship among embryonic NMPs, regenerated tail NSCs, and our gene- edited NSCs. In fact, since our initial submission, we have compared these three populations by single cell RNA seq and observed differential TBX18 expression biased to more embryonic cell origins. However, we did not observe the abilities of either native NSCs or transplanted embryonic NSCs to contribute to mesodermal lineages in regenerated lizard tails that has been reported in salamanders. This may mark another difference between salamanders and lizards and may be responsible for their divergent regenerative potentials and will be the subject of future work.  
+
+## Henry Roehl  
+
+- -  
+
+Reviewer #3 (Remarks to the Author):  
+
+This is an exciting manuscript by Lozito et al showing that a loss of dorsal- ventral patterning during tail regeneration in lizards can be re- activated through CRISPR/Cas9 gene editing. This would seem first glance as a niche finding, but it actually has broad implications for regenerative medicine. It is an exciting example of enhancing an in vivo regenerative response in an otherwise limited regenerative capacity. This should have broad interest to researchers working on spinal cord regeneration, but also people investigating non neural regeneration that may use similar strategies to promote regeneration in other organs. It is one of very few examples showing enhancement of regeneration in vivo.  
+
+Noteworthy results include the de novo dorsal pattern in an amniote that would otherwise not generate this pattern during spinal cord regeneration. It is of interest that shh expression is the default state of the regenerating tail, which can even overcome embryonic dorsal neural tube cells. I found it particularly striking that the dorsal identity of smoothened KO NSCs generated what looked to be very large DRG structures. To me, this is a significant feat to generate a DRG that otherwise would not be generated. It will be important in the future (not here) to understand whether this DRG innervates the appropriate targets and sends processes back to the spinal cord.  
+
+There are some small, unfortunate findings that limit the impact (although very high) such as the observation that the smo KO NSCs do not generate CNS neurons and that a double neural tube is generated after engraftment. Regardless, the manuscript clearly shows that using gene editing can make a NSC competent to participate in proper patterning for spinal cord regeneration, which is an important feet for future therapeutics.  
+
+If I was to have one critique on the methodology, the manuscript relies heavily on a single staining modality, IHC instead of incorporating qPCR or FISH as second forms of validation. My concern for this is dampened because the proteins the authors are assaying are known to have specific, robust expression patterns in the neural tube and the staining throughout the manuscript is very specific. Therefore, for this particular case the single staining approach is sufficient.  
+
+Overall, this manuscript was a logical and clear. Although the final outcome would have been more impactful if a single neural tube was formed and CNS differentiation occurred after stem cell transplant, this is indeed an exciting advance using clever techniques for tissue grafting and lineage tracing.  
+
+Below are some minor comments and mistakes I found in the manuscript:  
+
+Page 3, line 13 - Change NCS to NSC  
+
+This mistake has been corrected, thank you for catching it.  
+
+Page 10, line 9- 17: Is there any indication why the embryo- derived NSCs/cyclopamine- treated cells don't incorporate into the existing ependymal tube? I would think that it would just generate a dorsal portion of the existing ET. This is quite interesting and may be worth adding a sentence in the results
+
+<--- Page Split --->
+
+## section as it read like an incomplete explanation for what is going on here. Is it because the entire animal is treated with cyclopamine?  
+
+We thank the reviewer for identifying this interesting point. We believe that these results indicate a resistance of embryonic NSCs to fully integrate with adult NSC structures when protected from endogenous Hedgehog signaling (yes, because the entire animal is treated with cyclopamine). This explanation has been added to the text of the manuscript. Specifically, we suspect some differences in hedgehog- sensitive cadherin expression between embryonic and adult NSC populations result in their resistance to intermingle. We are currently investigating these possibilities.  
+
+## In Figure 8H-O, why are most of the \(\mathsf{GFP + }\) cells sox2-? Does removing Shh signaling make them differentiate?  
+
+We thank the reviewer for their observation. We have quantified the percentages of \(\mathsf{GFP + }\) exogenous gene- . edited NSCs that co- express Sox2 and other markers following incorporation into regenerated tail ETs, and these results are presented in Figure 7 Z and Supplementary Fig. 16 and 17 in the revised manuscript. These quantifications showed that both GFP- labeled Sno KO NSCs and Dil- labeled Control NSCs persisted among Sox2+ ependyma following incorporation in regenerated tail ETs. However, only Sno KO NSCs were able to differentiate into Sox2- TuJ1+ neurons, supporting the conclusion that removing Shh signaling allows exogenous embryonic NSCs to undergo neurogenesis within adult regenerated tail structures. These points have been added to the text.  
+
+## Page 12, line 3: I think the phrase "exhibited Col2+ cartilage rods dorsal to ETs" was meant to say ventral instead of dorsal  
+
+This mistake has been corrected, thank you for catching it.  
+
+James Monaghan  Department of Biology  Institute for Chemical Imaging of Living Systems  Northeastern University  Boston, MA 02115
+
+<--- Page Split --->
+
+Reviewers' Comments:  
+
+Reviewer #1:  
+
+Remarks to the Author:  
+
+The authors have addressed all reviewers comments and in doing so have improved the manuscript overall.  
+
+Reviewer #2:  
+
+Remarks to the Author:  
+
+The authors have made many revisions to their initial submission and I am happy that my concerns have been addressed. I would recommend publication of this latest version.  
+
+Reviewer #3:  
+
+Remarks to the Author:  
+
+The authors have adequately responded to the concerns raised by me and the other reviewers. I do not have any more concerns with the manuscript.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d751de784711fe985e21fc71a4089fbbdcb890f642dcc7aa3028ff9a5fdfca83/supplementary_0_Peer Review File__d751de784711fe985e21fc71a4089fbbdcb890f642dcc7aa3028ff9a5fdfca83_det.mmd

@@ -0,0 +1,377 @@
+<|ref|>title<|/ref|><|det|>[[100, 40, 507, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[106, 110, 373, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[108, 155, 857, 210]]<|/det|>
+Introducing dorsoventral patterning into adult regenerating lizard tails with gene- edited embryonic neural stem cells.  
+
+<|ref|>image<|/ref|><|det|>[[95, 732, 262, 780]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[271, 732, 880, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[116, 90, 286, 103]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[116, 120, 218, 133]]<|/det|>
+Reviewer #1:  
+
+<|ref|>text<|/ref|><|det|>[[116, 135, 291, 148]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[116, 150, 838, 223]]<|/det|>
+In this manuscript Lozito et al determine the molecular signals that inhibit ependymal tube regeneration in an adult lizard. They use CRISPR technology to knock- out the signal inhibitory to correct dorsoventral patterning of the ependymal tube in cultured neurospheres and through transplantation experiments so this suffices to induce ependymal tube regeneration after tail amputation in an adult lizard.  
+
+<|ref|>text<|/ref|><|det|>[[115, 223, 881, 313]]<|/det|>
+Overall it is a very interesting manuscript that gives novel insights into the roadblocks to regeneration in the lizard and shows how one of these can be overcome but does not give any information on other pathways that may also limit the obstacle of improving neural regeneration at the expense of cartilage regeneration. The manuscript focuses heavily on shh signalling but does not consider the pathways like Fgf, Wnt and BMP that will established from in vivo work in a range of organisms to play important roles in patterning the neural tube and directing neuronal differentiation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 313, 876, 342]]<|/det|>
+The manuscript is very dense and in places difficult to read, many figure need better labelling as there are many panels under one label.  
+
+<|ref|>text<|/ref|><|det|>[[116, 357, 228, 372]]<|/det|>
+Specific points:  
+
+<|ref|>text<|/ref|><|det|>[[111, 385, 880, 789]]<|/det|>
+1. Figure shows that markers of dorsoventral patterning are expressed during embryogenesis but not during tail regeneration. Pax6 is a marker of lateral tissue, here is appears to be also expressed in the dorsal domain (fig.1 F). Is there overlap in the cells that express Pax6 and Pax7 in the embryos? 
+2. In fig.1 H, J adult ET shh, a ventral marker appears to be expressed in the dorsal region? Are these panels inverted? Are these cells sox2 positive? 
+3. Are cells in the ependymal tube in the adult or in the regenerating tail slowly dividing. I would be interesting to look at EdU incorporation in these cells. 
+4. Where are the neuronal cell bodies located in the adult, are there cells which express NeuN and Sox2? NeuN is often found expressed in neural stems that have begun the program of differentiation. 
+5. Are all cells in the adult or regeneration ependymal tube GFAP and sox2 positive? Are other markers of astrocytes expressed in these cells? Are markers of glial scar formation expressed in the regenerating ependymal tube cells after injury? 
+6. Do the cells when cultured default to a certain identify because they are missing secreted factors like wnts ? 
+7. Fig.3 & 4 could be combined into one figure and the text describing them could be shortened. 
+8. Figure 5 is a nice technique but this could be moved to supplementary data, it seems odd here and disrupts the flow of the manuscript. It is unclear why the authors did not continue to use this technology to label cells for further experiments, it is better than DiI labelling neurospheres. 
+9. In the neurosphere cultures that authors report they do not see expression of the neuronal marker Tuj1, in the adult derived neurospheres this is a late marker. It is important to determine if the cells begin the neuronal differentiation program but cannot terminally differentiate. 
+10. The authors use nice clever modern technology to knockout Smo and generate gene edited neurospheres which they implant into the adult tail and assay the outcome of regeneration. The limited data shown suggests that this allows better patterning of the regenerated ependymal tube and differentiation of neurons. What percentage of the regenerated neural tube is derived from the implanted cells? What percentage of cells become neurons? How many animals was this tested on? Do the animals regenerate the same length of tail that was amputated ?  
+
+<|ref|>text<|/ref|><|det|>[[116, 835, 216, 848]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[116, 851, 291, 864]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[113, 865, 872, 895]]<|/det|>
+In this study the authors have performed a detailed analysis of the role of Hedgehog signalling during lizard tail regeneration. They have compared this to developmental roles including a detailed analysis
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 89, 877, 164]]<|/det|>
+of neural stem cells. Remarkably they have been able use this knowledge to improve upon tail regeneration by restoring aspects of dorsal/ventral patterning that are normally absent in regenerated tails. The study is also impressive in its development of many seemingly novel techniques for the analysis of regenerative biology in lizards. Over all I would strongly support publication. Here are my comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 178, 852, 195]]<|/det|>
+1) Figure 1 HJK - Isn't the expression of Hedgehog more widespread than discussed in the results?  
+
+<|ref|>text<|/ref|><|det|>[[115, 208, 852, 238]]<|/det|>
+2) Figure 5 - Aren't there more RFP+ cells than just in the dorsal root ganglion and neural tube? If these are derived from sox2-cre expressing cells shouldn't they also be neurons?  
+
+<|ref|>text<|/ref|><|det|>[[115, 253, 870, 298]]<|/det|>
+3) More quantification is needed. The quantification in the supplemental data is impressive, but more numbers should be provided for each figure. For example \(N = 6\) for figure 5 does not provide enough information.  
+
+<|ref|>text<|/ref|><|det|>[[115, 313, 872, 358]]<|/det|>
+4) Some of the figures are very complicated and difficult to follow. There are useful diagrams to show the method, please also add summary diagrams to illustrate your interpretation/model. Doing this for Figure 5 to 8 would be good.  
+
+<|ref|>text<|/ref|><|det|>[[115, 372, 875, 417]]<|/det|>
+5) Please go over the text to improve the logical flow: you should avoid using the word "Next we" to introduce an experiment. Recap and introduce your hypothesis before describing the next experiment. Th authors have done this very well for fig 8/pg 10.  
+
+<|ref|>text<|/ref|><|det|>[[115, 431, 618, 447]]<|/det|>
+6) Consider capitalising "Hedgehog" when referring to the pathway.  
+
+<|ref|>text<|/ref|><|det|>[[115, 462, 866, 491]]<|/det|>
+7) Isn't there also a neural crest contribution to dorsal root ganglion? Shouldn't this be considered in the interpretation of the results?  
+
+<|ref|>text<|/ref|><|det|>[[115, 506, 868, 566]]<|/det|>
+8) It should be pointed out that neural mesodermal progenitors (NMPs) have been identified in other models and these cells are thought to be embryonic tail stem cells capable of generating both neural and mesodermal derivatives. NMPs co-express tbx genes with sox2. It would be good to consider the role of NMPs by including tbx gene analysis in future studies.  
+
+<|ref|>text<|/ref|><|det|>[[115, 581, 208, 595]]<|/det|>
+Henry Roehl  
+
+<|ref|>text<|/ref|><|det|>[[115, 640, 217, 654]]<|/det|>
+Reviewer #3:  
+
+<|ref|>text<|/ref|><|det|>[[115, 657, 291, 670]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 670, 877, 775]]<|/det|>
+This is an exciting manuscript by Lozito et al showing that a loss of dorsal- ventral patterning during tail regeneration in lizards can be re- activated through CRISPR/Cas9 gene editing. This would seem first glance as a niche finding, but it actually has broad implications for regenerative medicine. It is an exciting example of enhancing an in vivo regenerative response in an otherwise limited regenerative capacity. This should have broad interest to researchers working on spinal cord regeneration, but also people investigating non neural regeneration that may use similar strategies to promote regeneration in other organs. It is one of very few examples showing enhancement of regeneration in vivo.  
+
+<|ref|>text<|/ref|><|det|>[[115, 789, 879, 895]]<|/det|>
+Noteworthy results include the de novo dorsal pattern in an amniote that would otherwise not generate this pattern during spinal cord regeneration. It is of interest that shh expression is the default state of the regenerating tail, which can even overcome embryonic dorsal neural tube cells. I found it particularly striking that the dorsal identity of smoothened KO NSCs generated what looked to be very large DRG structures. To me, this is a significant feat to generate a DRG that otherwise would not be generated. It will be important in the future (not here) to understand whether this DRG innervates the appropriate targets and sends processes back to the spinal cord.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 103, 855, 179]]<|/det|>
+There are some small, unfortunate findings that limit the impact (although very high) such as the observation that the smo KO NSCs do not generate CNS neurons and that a double neural tube is generated after engraftment. Regardless, the manuscript clearly shows that using gene editing can make a NSC competent to participate in proper patterning for spinal cord regeneration, which is an important feet for future therapeutics.  
+
+<|ref|>text<|/ref|><|det|>[[115, 194, 878, 269]]<|/det|>
+If I was to have one critique on the methodology, the manuscript relies heavily on a single staining modality, IHC instead of incorporating qPCR or FISH as second forms of validation. My concern for this is dampened because the proteins the authors are assaying are known to have specific, robust expression patterns in the neural tube and the staining throughout the manuscript is very specific. Therefore, for this particular case the single staining approach is sufficient.  
+
+<|ref|>text<|/ref|><|det|>[[115, 282, 860, 343]]<|/det|>
+Overall, this manuscript was a logical and clear. Although the final outcome would have been more impactful if a single neural tube was formed and CNS differentiation occurred after stem cell transplant, this is indeed an exciting advance using clever techniques for tissue grafting and lineage tracing.  
+
+<|ref|>text<|/ref|><|det|>[[115, 357, 666, 372]]<|/det|>
+Below are some minor comments and mistakes I found in the manuscript:  
+
+<|ref|>text<|/ref|><|det|>[[116, 386, 397, 401]]<|/det|>
+Page 3, line 13 - Change NCS to NSC  
+
+<|ref|>text<|/ref|><|det|>[[115, 416, 870, 491]]<|/det|>
+Page 10, line 9- 17: Is there any indication why the embryo- derived NSCs/cyclopamine- treated cells don't incorporate into the existing ependymal tube? I would think that it would just generate a dorsal portion of the existing ET. This is quite interesting and may be worth adding a sentence in the results section as it read like an incomplete explanation for what is going on here. Is it because the entire animal is treated with cyclopamine?  
+
+<|ref|>text<|/ref|><|det|>[[115, 505, 839, 535]]<|/det|>
+In Figure 8H- O, why are most of the GFP+ cells sox2- ? Does removing Shh signaling make them differentiate?  
+
+<|ref|>text<|/ref|><|det|>[[115, 549, 852, 579]]<|/det|>
+Page 12, line 3: I think the phrase "exhibited Col2+ cartilage rods dorsal to ETs" was meant to say ventral instead of dorsal  
+
+<|ref|>text<|/ref|><|det|>[[115, 595, 480, 670]]<|/det|>
+James Monaghan  Department of Biology  Institute for Chemical Imaging of Living Systems  Northeastern University  Boston, MA 02115
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[56, 39, 940, 72]]<|/det|>
+We thank the reviewers for their insightful and thoughtful feedback. This document addresses the concerns and comments that were brought up by the reviewers.  
+
+<|ref|>text<|/ref|><|det|>[[58, 87, 378, 103]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[56, 104, 930, 152]]<|/det|>
+1. Figure shows that markers of dorsoventral patterning are expressed during embryogenesis but not during tail regeneration. Pax6 is a marker of lateral tissue, here is appears to be also expressed in the dorsal domain (fig.1 F). Is there overlap in the cells that express Pax6 and Pax7 in the embryos?  
+
+<|ref|>text<|/ref|><|det|>[[56, 167, 942, 310]]<|/det|>
+Thank you for this observation, which is spot on. There appears to be some overlap between \(\mathrm{Pax7^{+}}\) and \(\mathrm{Pax6^{+}}\) cells in lizard tail neural tubes, an observation that has been added to the results section of the revised manuscript. We suspect this to be indicative of differences in cell origins and patterning of primary vs secondary neural tubes. Indeed, in embryonic lizard trunks, primary neural tubes (which originate from neural plates) exhibit more distinct lateral \(\mathrm{Pax6}\) expression. In contrast, secondary neural tubes of embryonic lizard tails (which originate from tail bud neural mesodermal progenitors) exhibit a dorsal shift in \(\mathrm{Pax6}\) expression, as observed in Fig. 1F. We have begun comparing primary vs secondary lizard neural tubes by single cell RNA sequencing to further investigate differences in cell identity and patterning that might affect divergent healing capabilities. This will be the subject of a future manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[56, 326, 940, 359]]<|/det|>
+2. In fig.1 H, J adult ET shh, a ventral marker appears to be expressed in the dorsal region? Are these panels inverted? Are these cells sox2 positive?  
+
+<|ref|>text<|/ref|><|det|>[[56, 374, 942, 470]]<|/det|>
+Thank you for your observation. These images are oriented correctly and depict one of the most important findings of this study; adult original and regenerated tail ependyma/NSC populations exhibit floor plate markers such as Shh completely around central canal circumferences, including dorsal regions. Some transverse sections of adult original tail ependyma, such as those depicted in Fig. 1 H, J, exhibit patchier Shh expression than regenerated tail sections, possibly due to the presence of Shh+ Tuj+ Sox2 neurons. These points have been added to the manuscript text in the revision.  
+
+<|ref|>text<|/ref|><|det|>[[56, 486, 940, 518]]<|/det|>
+3. Are cells in the ependymal tube in the adult or in the regenerating tail slowly dividing. I would be interesting to look at EdU incorporation in these cells.  
+
+<|ref|>text<|/ref|><|det|>[[56, 534, 942, 663]]<|/det|>
+We thank the reviewer for their interest, and these EdU/proliferation studies have been previously published by our group (Sun et al. PNAS 2018, Figure S6). For convenience, we included a simplified version of the published figure below. 5- ethynyl- 2'- deoxyuridine (EdU) incorporation/staining assays were used to visualize proliferating \(\mathrm{Sox2 + }\) NSC populations in original and regenerating (14, 28, and 56 DPA) lizard tails along their lengths (proximal, middle, and distal). 14- DPA samples, which correspond to the blastema stage of regeneration, exhibited the most numbers of proliferative NSCs. 28- DPA samples, which correspond to lengthening tails, exhibited higher numbers of proliferative NSCs in distal regions. 56- DPA samples, which correspond to fully regenerated tails, resembled original tails and exhibited low levels of proliferating cells.  
+
+<|ref|>image<|/ref|><|det|>[[378, 678, 616, 952]]<|/det|>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[56, 39, 940, 74]]<|/det|>
+## 4. Where are the neuronal cell bodies located in the adult, are there cells which express NeuN and Sox2? NeuN is often found expressed in neural stems that have begun the program of differentiation.  
+
+<|ref|>text<|/ref|><|det|>[[55, 87, 943, 185]]<|/det|>
+Thank you for the suggestion. To address this comment, we analyzed cross sections of embryonic, adult, and regenerated lizard tails with NeuN, Sox2, and TuJ1 immunofluorescence. NeuN signal was detected in nuclei of TuJ1+ neurons in interior regions of original adult spinal cords, but not in Sox2+ ependymal cells. NeuN signal was not detected in regenerated tail spinal cord, corroborating our observations that no new neurons are formed during tail regrowth. Sox2/NeuN co- expression was only observed in tail neural tube cells, indicating spinal cord neurogenesis was restricted to embryonic stages. This new data is presented in new Supplementary Fig.4.  
+
+<|ref|>image<|/ref|><|det|>[[223, 198, 773, 318]]<|/det|>  
+
+<|ref|>sub_title<|/ref|><|det|>[[55, 342, 940, 393]]<|/det|>
+## 5. Are all cells in the adult or regeneration ependymal tube GFAP and sox2 positive? Are other markers of astrocytes expressed in these cells? Are markers of glial scar formation expressed in the regenerating ependymal tube cells after injury?  
+
+<|ref|>text<|/ref|><|det|>[[55, 405, 943, 519]]<|/det|>
+To address this comment, we quantified the percentages of GFAP+ ependymal cells in original and regenerated tails that expressed Sox2. Cross sections of regenerated tail ependymal tubes were immunostained for GFAP and Sox2, and individual cells were identified with DAPI staining. Keyence microscope software was used to quantify the percentages of ependymal tube cells that exhibited GFAP and/or Sox2 signal. Ten separate tails/animals were assayed, and average percentages \(+ / -\) standard deviations were calculated. \(94.3 + / - 3.8\%\) of GFAP+ original tail ependymal cells were positive for GFAP, while \(95.2 + / - 4.1\%\) of GFAP+ ET cells were positive for Sox2. This data is presented in new Supplementary Fig. 9.  
+
+<|ref|>text<|/ref|><|det|>[[55, 518, 943, 712]]<|/det|>
+To determine other markers expressed by regenerated tail NSCs, we referenced our data sets from single cell RNA sequencing data analyses of lizard ependymal tubes and spinal cords and neurospheres, included in the figure below. Regenerated tail GFAP+ Sox2+ NSC clusters co- expressed radial glial markers FABP7 and Sox9. Original tail spinal cord data sets exhibited distinct clusters of GLUL+ astrocytes and Sox10+ MBP+ CLDN11+ oligodendrocytes. These clusters were not represented in ependymal tube or neurosphere data sets, and regenerated tail NSCs did not express other astrocyte and oligodendrocyte markers, confirming their radial glia/ependymal identity. We also did not observe genes commonly associated with mammalian glial scar formation (CSPGs, laminin, tenascin, fibronectin, etc.) in ependymal tube cell transcriptomes. Similarly, we did not observe pro- inflammatory cytokines (CCL2, CCL5, CXCL8 etc.) in regenerated tail spinal cord pericytes. Future work will be aimed at comparing scar- associated genes in regenerating lizard spinal cord cell populations with those expressed during other models of spinal cord injury/healing. We suspect the regenerated tail environment somehow inhibits glial and pericyte scaring processes characteristic of trunk spinal cord injury.  
+
+<|ref|>image<|/ref|><|det|>[[245, 725, 750, 912]]<|/det|>
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[56, 40, 908, 72]]<|/det|>
+## 6. Do the cells when cultured default to a certain identify because they are missing secreted factors like wnts ?  
+
+<|ref|>text<|/ref|><|det|>[[55, 87, 942, 280]]<|/det|>
+We thank the reviewer for their interest and suggestion. To answer this question, we referenced our single cell RNA sequencing data sets to compare growth factor expression in original tail spinal cord, regenerated tail ependymal tube, and cultured neurosphere NSCs. General trends indicated higher Wnt expression (Wnt1, Wnt5a, Wnt11) in ependymal tubes than in original spinal cords and neurospheres. Interestingly, cultured neurospheres expressed high levels of Wnt5b. Conversely, BMP expression (BMP2, BMP3) was higher in original spinal cord and undetectable in ependymal tubes and neurospheres. Finally, we treated cultured neurospheres with Wnt1 and Wnt5a during differentiation assays and did not detect roof plate or neuronal differentiation. In summary, we concluded that environmental factors such as growth factor signaling do not prevent adult lizard tail NSCs from undergoing neurogenesis during tail regrowth. Instead, perhaps NSCs undergo epigenetic changes between embryonic development and adult tail regeneration that inhibit neural differentiation, and future work will explore this topic. We are currently preparing our single cell RNA sequencing results for a separate publication.  
+
+<|ref|>sub_title<|/ref|><|det|>[[60, 294, 875, 312]]<|/det|>
+## 7. Fig.3 & 4 could be combined into one figure and the text describing them could be shortened.  
+
+<|ref|>text<|/ref|><|det|>[[56, 326, 940, 360]]<|/det|>
+Figures 3 and 4 of the original manuscript have been combined into a single figure (Revised Figure 3), and the text has been adjusted and shortened accordingly.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 374, 925, 423]]<|/det|>
+## 8. Figure 5 is a nice technique but this could be moved to supplementary data, it seems odd here and disrupts the flow of the manuscript. It is unclear why the authors did not continue to use this technology to label cells for further experiments, it is better than Dil labelling neurospheres.  
+
+<|ref|>text<|/ref|><|det|>[[57, 437, 933, 470]]<|/det|>
+We appreciate the reviewer's suggestions and have moved old Figure 5 to Supplementary Data (Supplementary Fig. 8)  
+
+<|ref|>text<|/ref|><|det|>[[55, 470, 942, 662]]<|/det|>
+Our choice to use Dil labelling rather than the lentivirus system to trace cell lineages in neurosphere implantation studies was due to the time and supply costs associated with lentivirus production and use coupled with the large number of NSCs needed for implantation. For example, experiments comparing embryonic and adult NSC behavior post- implantation, such as those presented in revised Figure 6), required increased numbers of neurospheres due to limitations in adult NSC propagation in vivo. Unlike embryonic NSCs, adult NSCs exhibited diminished proliferative potential beyond passage 5, meaning that more lizards had to be used to generate the neurospheres required for these experiments. Furthermore, we didn't use the lentiviral system to label Control- CRISPR cells in gene editing experiments in order to provide the most appropriate controls for Smo- CRISPR cells without confounding effects caused by genomic insertion of lentiviral constructs. We had previously reported on Dil labeling as an effective and cost- efficient means for labeling large numbers of lizard neurospheres with minimal effects on NSC behavior and genetics, filling the requirements for the studies included in this manuscript.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 677, 919, 726]]<|/det|>
+## 9. In the neurosphere cultures that authors report they do not see expression of the neuronal marker Tuj1, in the adult derived neurospheres this is a late marker. It is important to determine if the cells begin the neuronal differentiation program but cannot terminally differentiate.  
+
+<|ref|>text<|/ref|><|det|>[[55, 741, 942, 934]]<|/det|>
+We thank the reviewer for their insight and suggestion. To address these comments, we analyzed embryonic and adult NSC- derived neurospheres for expression of doublecortin (DCX) and NeuroD2 (early neuronal differentiation markers) by real- time RT PCR. The results of these experiments are presented in revised Supplemental Fig. 11 and are included below for easy referencing. mRNA was collected from either undifferentiated neurospheres or neurospheres cultured under differentiated conditions for 14 days treated with either vehicle control, cyclopamine, or SAG and assayed for Sox2, DCX, NeuroD2, GFAP, and Tuj1 expression. Each experimental condition involved pooled mRNA from 10 neurospheres and was repeated 3 times. Results are presented as fold changes compared to undifferentiated conditions. Results showed that differentiation caused significant increases in DCX, neuroD1, and Tuj1 in embryonic tail, but not adult tail, NSC- derived neurospheres. These increases were unaffected by cyclopamine treatment, but SAG treatment caused significant decreases in DCX, NeuroD2, and Tuj1 expression. In contrast, adult lizard NSC expression of Sox2, GFAP, DCX, NeuroD2, and Tuj1 expression were unaffected by either differentiation or Hedgehog signaling
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 38, 940, 72]]<|/det|>
+modulation. Future work will be aimed at forcing expression of pro-neurogenesis factors like NeuroD2 to attempt initiating neurogenesis in adult lizard NSCs.  
+
+<|ref|>image<|/ref|><|det|>[[174, 85, 821, 315]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[56, 348, 937, 448]]<|/det|>
+10. The authors use nice clever modern technology to knockout Smo and generate gene edited neurospheres which they implant into the adult tail and assay the outcome of regeneration. The limited data shown suggests that this allows better patterning of the regenerated ependymal tube and differentiation of neurons. What percentage of the regenerated neural tube is derived from the implanted cells? What percentage of cells become neurons? How many animals was this tested on? Do the animals regenerate the same length of tail that was amputated?  
+
+<|ref|>text<|/ref|><|det|>[[55, 460, 943, 653]]<|/det|>
+We thank the reviewer for their inquiries. First, we quantified the contribution of implanted embryonic NSCs to adult regenerated tail dorsal and ventral ET regions. These results are presented in the revised manuscript in Supplementary Fig. 16 and are included below for reference. Smo KO NSCs labeled with GFP and control NSCs pre- labeled with Dil were engrafted to dorsal ependyma of amputated tail spinal cords. Following 28 days of regrowth, regenerated tails were collected and processed for histology. ET cross sections were analyzed by histology/IF/fluorescence microscopy for GFP/Dil signal and Sox2 expression every 1 mm along entire tail lengths. Horizontal lines drawn through the centers of ETs bisected tube images into dorsal and ventral regions. Quantification of dorsal and ventral Sox2 and GFP/Dil signal areas were performed for each cross- section along tail lengths. Percentages of Sox2+ IF signal areas co- expressing GFP/Dil were calculated for dorsal and ventral ET regions and are presented in the histograms below. Dorsal ET regions exhibited significantly higher GFP/Dil signal from labeled exogenous NSC populations than ventral regions, indicating that starting NSC spatial arrangements were maintained along regenerated ET lengths.  
+
+<|ref|>image<|/ref|><|det|>[[222, 666, 780, 927]]<|/det|>
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 55, 943, 232]]<|/det|>
+Next, we quantified the percentages of exogenous embryonic NCS undergoing neurogenesis following incorporation in adult regenerating tails. These results are presented in the revised manuscript in Supplementary Fig. 17 and are included below for reference. GFP- labeled Smo KO NSCs and control NSCs pre- labeled with Dil engrafted to dorsal spinal cord ependyma cell populations. Following 28 days of regrowth, regenerated tails were collected and processed for histology. ET cross sections were analyzed by IF/fluorescence microscopy for GFP/Dil signal and Tuj1 expression every 1 mm along entire tail lengths. 10 different animals/tails were analyzed for each condition. Tuj1 and GFP/Dil signal areas were quantified for each cross- section, and percentages of Tuj1+ IF signal areas co- expressing GFP/Dil are presented in the histograms below. Tuj1+ neural cells co- expressed GFP in lizards treated with Smo KO NSCs, indicating significantly higher levels of neurogenesis in Smo KO NSCs than control cells. \*, p<0.001 compared to corresponding Control NSC condition for each measurement.  
+
+<|ref|>image<|/ref|><|det|>[[171, 250, 830, 495]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[55, 507, 919, 590]]<|/det|>
+Finally, we compared the tail lengths regenerated by lizards treated with Smo KO NSCs vs Control NSCs vs untreated controls. Both groups of tails regrew to approximately 1.2 cm 28 DPA. Differences in tail lengths were insignificant between conditions, indicating that incorporation of exogenous embryonic NSCs did not affect regenerated tail lengths. The revised text has been updated with these observations. Furthermore, we emphasized the numbers of animals used in each experiment in the revised text and figure legends.  
+
+<|ref|>text<|/ref|><|det|>[[56, 592, 80, 600]]<|/det|>  
+
+<|ref|>sub_title<|/ref|><|det|>[[58, 619, 378, 637]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[55, 651, 943, 750]]<|/det|>
+In this study the authors have performed a detailed analysis of the role of Hedgehog signalling during lizard tail regeneration. They have compared this to developmental roles including a detailed analysis of neural stem cells. Remarkably they have been able use this knowledge to improve upon tail regeneration by restoring aspects of dorsal/ventral patterning that are normally absent in regenerated tails. The study is also impressive in its development of many seemingly novel techniques for the analysis of regenerative biology in lizards. Over all I would strongly support publication. Here are my comments:  
+
+<|ref|>text<|/ref|><|det|>[[63, 763, 936, 782]]<|/det|>
+1) Figure 1 HJK - Isn't the expression of Hedgehog more widespread than discussed in the results?  
+
+<|ref|>text<|/ref|><|det|>[[58, 795, 937, 829]]<|/det|>
+We appreciate this note and have also noted that Shh protein is detected among spinal cord nerves surrounding original tail ependyma. The revised results section has been updated to reflect these observations.  
+
+<|ref|>text<|/ref|><|det|>[[58, 842, 936, 876]]<|/det|>
+2) Figure 5 - Aren't there more RFP+ cells than just in the dorsal root ganglion and neural tube? If these are derived from sox2-cre expressing cells shouldn't they also be neurons?  
+
+<|ref|>text<|/ref|><|det|>[[56, 890, 941, 956]]<|/det|>
+This is a good question and, to answer it, we re- analyzed the images from our lineage tracing experiments with Keyence microscope software to quantify the percentage overlap of RFP signal with Sox2, GFAP, and TUJ1 signals. These results are presented in revised Supplementary Fig. 9. Overall, they show that the vast majority of RFP+ cells (>98%) co- localize with Sox2, GFAP, or TUJ1 regions within either neural tubes, developing and
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[56, 39, 941, 136]]<|/det|>
+adult spinal cords or DRG, indicating that most labeled embryonic NSCs differentiate into tail ependymal cells, glia, or neurons. For example, in embryonic samples, the majority of \(\mathrm{Tuj1 + }\) signal was observed in peri-neural tube regions rather than neural tubes themselves. We hypothesized that these regions were the sites populated by early neurons that would later form tail spinal cord. Interestingly, adult peripheral nerves were only sporadically labeled with RFP using this system, perhaps indicating additional cell sources of tail peripheral nerves. The revised text has been updated with these observations.  
+
+<|ref|>text<|/ref|><|det|>[[56, 151, 941, 201]]<|/det|>
+3) More quantification is needed. The quantification in the supplemental data is impressive, but more numbers should be provided for each figure. For example \(N = 6\) for figure 5 does not provide enough information.  
+
+<|ref|>text<|/ref|><|det|>[[56, 215, 941, 296]]<|/det|>
+We have taken this suggestion to heart and have quantified immunofluorescence findings presented in Figures 5, 6, 7, 8, and the results are depicted in the revised Figures 4, 6, 8 and in newly added Supplemental Fig. 9, 11, 16, and 17. Specifically, Keyence microscope software was used to quantify percentages of \(\mathrm{RFP + }\) or \(\mathrm{GFP + }\) /Dil+ regions that co- express Sox2, GFAP, TUJ1, Shh, and Pax7. We also provided better descriptions of the animal and sample numbers associated with each experiment.  
+
+<|ref|>text<|/ref|><|det|>[[56, 311, 941, 360]]<|/det|>
+4) Some of the figures are very complicated and difficult to follow. There are useful diagrams to show the method, please also add summary diagrams to illustrate your interpretation/model. Doing this for Figure 5 to 8 would be good.  
+
+<|ref|>text<|/ref|><|det|>[[56, 375, 940, 408]]<|/det|>
+Summary diagrams have been added to revised Figures 5 and 7 to illustrate our interpretation of results from lineage tracing and gene- edited NSC transplantation, respectively.  
+
+<|ref|>text<|/ref|><|det|>[[56, 422, 940, 472]]<|/det|>
+5) Please go over the text to improve the logical flow: you should avoid using the word "Next we" to introduce an experiment. Recap and introduce your hypothesis before describing the next experiment. The authors have done this very well for fig 8/pg 10.  
+
+<|ref|>text<|/ref|><|det|>[[58, 486, 818, 503]]<|/det|>
+The text has been edited to improve logical flow and to highlight hypothesis- driven experiments.  
+
+<|ref|>text<|/ref|><|det|>[[56, 517, 636, 535]]<|/det|>
+6) Consider capitalising "Hedgehog" when referring to the pathway.  
+
+<|ref|>text<|/ref|><|det|>[[58, 550, 783, 567]]<|/det|>
+We have capitalized "Hedgehog" when referring to the pathway throughout the revised text.  
+
+<|ref|>text<|/ref|><|det|>[[56, 581, 940, 615]]<|/det|>
+7) Isn't there also a neural crest contribution to dorsal root ganglion? Shouldn't this be considered in the interpretation of the results?  
+
+<|ref|>text<|/ref|><|det|>[[56, 630, 941, 855]]<|/det|>
+We thank the reviewer for this note as it touches upon several interesting aspects of our study considering the unique case of the lizard tail, the only amniote tail that contains both spinal cord and DRG. In the trunks of amniote embryos, folding of \(\mathrm{Sox2 + }\) neural plates gives rise to lateral and floor domains of primary neural tubes, while neural plate borders pinch off and contribute to neural tube roof plate domains and \(\mathrm{Sox2 - }\) neural crest. Trunk neural crest cells migrate mediolaterally and begin to express \(\mathrm{Sox2}\) before forming DRG. The situation is very different in amniote tails, and even more exceptional in lizard tails. Tails are formed from tail buds, which do not have neural plates or neural folds, and hence, no neural crest cells. Instead, \(\mathrm{Sox2 + }\) neural mesodermal progenitors (NMPs) derived from \(\mathrm{Sox - }\) tail bud cells directly differentiate into secondary neural tubes. Mammalian secondary neural tubes degenerate before DRG formation during tail development, and adult mammal tails contain neither DRG nor spinal cord. In contrast, lizard secondary neural tubes survive tail development, and lizards are the only amniotes that exhibit both spinal cords and DRG in adult tails. The origins of amniote tail DRG has not been specifically studied, but lineage tracing data presented here suggest that lizard tail DRG derive from \(\mathrm{Sox2 + }\) tail bud cells that pass through a neural tube intermediate. These observations have been added to the Discussion section of the manuscript, and hopefully future work will expand on this topic.  
+
+<|ref|>text<|/ref|><|det|>[[56, 870, 941, 935]]<|/det|>
+8) It should be pointed out that neural mesodermal progenitors (NMPs) have been identified in other models and these cells are thought to be embryonic tail stem cells capable of generating both neural and mesodermal derivatives. NMPs co-express tbx genes with sox2. It would be good to consider the role of NMPs by including tbx gene analysis in future studies.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[55, 38, 943, 152]]<|/det|>
+We thank the reviewer for this wonderful observation, and we certainly plan on looking into the relationship among embryonic NMPs, regenerated tail NSCs, and our gene- edited NSCs. In fact, since our initial submission, we have compared these three populations by single cell RNA seq and observed differential TBX18 expression biased to more embryonic cell origins. However, we did not observe the abilities of either native NSCs or transplanted embryonic NSCs to contribute to mesodermal lineages in regenerated lizard tails that has been reported in salamanders. This may mark another difference between salamanders and lizards and may be responsible for their divergent regenerative potentials and will be the subject of future work.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 183, 163, 200]]<|/det|>
+## Henry Roehl  
+
+<|ref|>text<|/ref|><|det|>[[56, 220, 80, 230]]<|/det|>
+- -  
+
+<|ref|>text<|/ref|><|det|>[[57, 232, 378, 249]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[55, 263, 936, 375]]<|/det|>
+This is an exciting manuscript by Lozito et al showing that a loss of dorsal- ventral patterning during tail regeneration in lizards can be re- activated through CRISPR/Cas9 gene editing. This would seem first glance as a niche finding, but it actually has broad implications for regenerative medicine. It is an exciting example of enhancing an in vivo regenerative response in an otherwise limited regenerative capacity. This should have broad interest to researchers working on spinal cord regeneration, but also people investigating non neural regeneration that may use similar strategies to promote regeneration in other organs. It is one of very few examples showing enhancement of regeneration in vivo.  
+
+<|ref|>text<|/ref|><|det|>[[55, 389, 936, 504]]<|/det|>
+Noteworthy results include the de novo dorsal pattern in an amniote that would otherwise not generate this pattern during spinal cord regeneration. It is of interest that shh expression is the default state of the regenerating tail, which can even overcome embryonic dorsal neural tube cells. I found it particularly striking that the dorsal identity of smoothened KO NSCs generated what looked to be very large DRG structures. To me, this is a significant feat to generate a DRG that otherwise would not be generated. It will be important in the future (not here) to understand whether this DRG innervates the appropriate targets and sends processes back to the spinal cord.  
+
+<|ref|>text<|/ref|><|det|>[[56, 517, 907, 599]]<|/det|>
+There are some small, unfortunate findings that limit the impact (although very high) such as the observation that the smo KO NSCs do not generate CNS neurons and that a double neural tube is generated after engraftment. Regardless, the manuscript clearly shows that using gene editing can make a NSC competent to participate in proper patterning for spinal cord regeneration, which is an important feet for future therapeutics.  
+
+<|ref|>text<|/ref|><|det|>[[56, 613, 907, 694]]<|/det|>
+If I was to have one critique on the methodology, the manuscript relies heavily on a single staining modality, IHC instead of incorporating qPCR or FISH as second forms of validation. My concern for this is dampened because the proteins the authors are assaying are known to have specific, robust expression patterns in the neural tube and the staining throughout the manuscript is very specific. Therefore, for this particular case the single staining approach is sufficient.  
+
+<|ref|>text<|/ref|><|det|>[[56, 709, 916, 775]]<|/det|>
+Overall, this manuscript was a logical and clear. Although the final outcome would have been more impactful if a single neural tube was formed and CNS differentiation occurred after stem cell transplant, this is indeed an exciting advance using clever techniques for tissue grafting and lineage tracing.  
+
+<|ref|>text<|/ref|><|det|>[[57, 789, 689, 807]]<|/det|>
+Below are some minor comments and mistakes I found in the manuscript:  
+
+<|ref|>text<|/ref|><|det|>[[57, 821, 377, 838]]<|/det|>
+Page 3, line 13 - Change NCS to NSC  
+
+<|ref|>text<|/ref|><|det|>[[57, 853, 522, 870]]<|/det|>
+This mistake has been corrected, thank you for catching it.  
+
+<|ref|>text<|/ref|><|det|>[[56, 885, 923, 935]]<|/det|>
+Page 10, line 9- 17: Is there any indication why the embryo- derived NSCs/cyclopamine- treated cells don't incorporate into the existing ependymal tube? I would think that it would just generate a dorsal portion of the existing ET. This is quite interesting and may be worth adding a sentence in the results
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[57, 39, 895, 73]]<|/det|>
+## section as it read like an incomplete explanation for what is going on here. Is it because the entire animal is treated with cyclopamine?  
+
+<|ref|>text<|/ref|><|det|>[[56, 87, 942, 185]]<|/det|>
+We thank the reviewer for identifying this interesting point. We believe that these results indicate a resistance of embryonic NSCs to fully integrate with adult NSC structures when protected from endogenous Hedgehog signaling (yes, because the entire animal is treated with cyclopamine). This explanation has been added to the text of the manuscript. Specifically, we suspect some differences in hedgehog- sensitive cadherin expression between embryonic and adult NSC populations result in their resistance to intermingle. We are currently investigating these possibilities.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 199, 940, 232]]<|/det|>
+## In Figure 8H-O, why are most of the \(\mathsf{GFP + }\) cells sox2-? Does removing Shh signaling make them differentiate?  
+
+<|ref|>text<|/ref|><|det|>[[56, 247, 942, 377]]<|/det|>
+We thank the reviewer for their observation. We have quantified the percentages of \(\mathsf{GFP + }\) exogenous gene- . edited NSCs that co- express Sox2 and other markers following incorporation into regenerated tail ETs, and these results are presented in Figure 7 Z and Supplementary Fig. 16 and 17 in the revised manuscript. These quantifications showed that both GFP- labeled Sno KO NSCs and Dil- labeled Control NSCs persisted among Sox2+ ependyma following incorporation in regenerated tail ETs. However, only Sno KO NSCs were able to differentiate into Sox2- TuJ1+ neurons, supporting the conclusion that removing Shh signaling allows exogenous embryonic NSCs to undergo neurogenesis within adult regenerated tail structures. These points have been added to the text.  
+
+<|ref|>sub_title<|/ref|><|det|>[[56, 391, 940, 423]]<|/det|>
+## Page 12, line 3: I think the phrase "exhibited Col2+ cartilage rods dorsal to ETs" was meant to say ventral instead of dorsal  
+
+<|ref|>text<|/ref|><|det|>[[57, 439, 523, 456]]<|/det|>
+This mistake has been corrected, thank you for catching it.  
+
+<|ref|>text<|/ref|><|det|>[[56, 472, 475, 551]]<|/det|>
+James Monaghan  Department of Biology  Institute for Chemical Imaging of Living Systems  Northeastern University  Boston, MA 02115
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 90, 286, 104]]<|/det|>
+Reviewers' Comments:  
+
+<|ref|>text<|/ref|><|det|>[[115, 120, 220, 134]]<|/det|>
+Reviewer #1:  
+
+<|ref|>text<|/ref|><|det|>[[115, 136, 291, 149]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 150, 860, 180]]<|/det|>
+The authors have addressed all reviewers comments and in doing so have improved the manuscript overall.  
+
+<|ref|>text<|/ref|><|det|>[[115, 225, 218, 238]]<|/det|>
+Reviewer #2:  
+
+<|ref|>text<|/ref|><|det|>[[115, 240, 291, 253]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 254, 860, 284]]<|/det|>
+The authors have made many revisions to their initial submission and I am happy that my concerns have been addressed. I would recommend publication of this latest version.  
+
+<|ref|>text<|/ref|><|det|>[[115, 328, 218, 342]]<|/det|>
+Reviewer #3:  
+
+<|ref|>text<|/ref|><|det|>[[115, 344, 291, 357]]<|/det|>
+Remarks to the Author:  
+
+<|ref|>text<|/ref|><|det|>[[115, 358, 860, 388]]<|/det|>
+The authors have adequately responded to the concerns raised by me and the other reviewers. I do not have any more concerns with the manuscript.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d7666663fe77730e88d0a0083aacf89a7b35973d18830a7da0a9d9465bb23514/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d7666663fe77730e88d0a0083aacf89a7b35973d18830a7da0a9d9465bb23514/supplementary_0_Peer Review File__d7666663fe77730e88d0a0083aacf89a7b35973d18830a7da0a9d9465bb23514.mmd

@@ -0,0 +1,545 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+Reviewer #1 (Remarks to the Author):Key results: The authors describe a strategy to transiently phosphorylate histidine, with the goal of mimicking the transient activation of enzymes in cells. Monoamidophosphate (MAP) is the main phosphorylating agent, and water is the dephosphorylating agent. Histidine catalyses the hydrolysis of MAP, getting phosphorylated/de- phosphorylated in the process. When histidine is embedded in a tripeptide, de- phosphorylation proves to be too slow for the reaction cycle to be considered dynamic, and the authors deploy pyridine to accelerate it. In the presence of MAP and pyridine, the half- life of the phosphorylated peptide is reduced to a few hours. Experimental data and kinetic modeling are used to depict the flow of phosphate in the system. The authors show that when the reaction is carried out in the presence of a polycation, supramolecular assemblies (coacervates) form and disappear following the dynamic phosphorylation of histidine.  
+
+Validity & Significance: The experimental data in the manuscript is complete and the kinetic model provides a good fit. However, I believe the concept and strategy described in this manuscript are not significantly novel, and would not have a sufficient impact in the community to grant publication in Nature Communications. Transient compartmentalization controlled by phosphorylation has been shown previously (DOI: 10.1038/nchem.2414, 10.1039/c7sm01897e). It is true that the manuscript provides a non- enzymatic version, but still the applicability is limited by the fact that the reaction only works with the pair MAP/histidine (tables in Figure 1). Recent literature on prebiotic phosphorylation tends to look for agents with a broader spectrum (e.g. DOI 10.1021/jacs.3c08539, 10.1038/s41467- 021- 25555- x); or in the case of histidine itself, for functional reactions with greater resemblance to natural catalytic cycles (see alternative with imidazole phosphate in Maguire ChemRxiv 2023). MAP, as the authors point out themselves, is not the most prebiotically plausible phosphorylating agent; and the use of pyridine as a substrate (that is mostly phosphorylated directly and competes with histidine) is somewhat confusing and undermines, in my opinion, its potential impact in the prebiotic chemistry field. The results on the formation of compartments are also weakened by the use of PEG 8k (at 2.5% concentration), a crowding agent known to promote and interfere with phase separation –
+
+<--- Page Split --->
+
+more than a simple turbidity enhancer. Combined with the fact that most micrographs show only one droplet, or one microreactor, it indicates that the supramolecular results may be not robust or reproducible, and that the impact on the protocell/dynamic assemblies community would be very limited.  
+
+I include below some specific minor and major concerns about the manuscript.  
+
+Clarity and context:  
+
+1. The authors use terms such as "active droplets" in the title, "self-sustaining complex coacervate droplets", but do not justify it in the text or in the results. Active droplets would imply a non-equilibrium behavior, which is not clear to me in the data (the droplets emerge, coalesce, and dissolve). Self-sustaining would imply that the droplets facilitate the phosphorylation reaction that leads to their own formation, which was not probed in the manuscript.  
+
+2. The description of the reaction as a cycle, or as a network (lines 64, 129), may be misleading, with the deactivation agent being the reaction's solvent (or pyridine), and the fuel MAP not being regenerated.  
+
+3. I am missing the explanation as to how short of a half-life is desirable in a biological or chemical context, and why the choice to rely on water or pyridine as de-activating agents, as opposed of some prebiotically relevant substrates. The authors could include a more specific discussion on histidyl kinases, instead of the general introduction on ATP.  
+
+4. Figures: scheme 1 would be more appropriate as a graphical abstract, it is taking up quite a lot of space but not adding as much information. Figure 1 could be reorganized to take up less space (the table format may not be necessary). Figure 2 contains some redundant data, panels F and G could be in supplementary information. Instead, a table for half-life and conversion% would be helpful. In Figure 3, what does Y'Dam mean?  
+
+5. The clarity of the manuscript could be increased by changing the acronyms. Instead of 1- pH, His1-P for example, as pH already has its own meaning.  
+
+6. Line 27: The sentence starting with "Does not occur..." is incomplete. The text from line 27 up to 39 needs to be restructured, it is confusing and not conveying its message – which I think is that biological cycles benefit from using ATP as a fuel due to it being thermodynamically unstable but kinetically stable.
+
+<--- Page Split --->
+
+7. Line 187: there is a typo, in "Either it is hydrolyzed directly or through phosphorylating histidine (unwanted)" it should be pyridine, not histidine. In line 192 there is a repetition "activation phosphorylation". This paragraph could be made clearer and the quantitative data mentioned could be included in a table elsewhere for clarity.  
+
+Data and methodology:  
+
+8. Both experimental and modeling kinetic data are well presented, error bars are shown but it is not mentioned if the mean is taken over multiple experiments or different measurements or time points. I appreciate how the methodology was extensively described. 
+9. Overall, some experiments could be better motivated in the text: for example, it is not clear to me that such a complete kinetic modeling is necessary, or how it can contribute to the understanding of the system. Another example is why the authors choose to conduct microscopy in microfluidically-produced micro-reactors, and not on a typical microscopy slide/chamber - what advantages to the paper's goals does it bring? The ITC and DNA partitioning data is included but not well integrated within the text - how is this data relevant to the manuscript? 
+10. The microscopy images do not immediately point to coacervate droplets (if you compare to other peptide/peptide complex coacervates in related literature), although I understand that FRAP supports it. Perhaps resolution and visibility would be increased without the micro-reactors. And instead of adding PEG, I would consider reducing buffer concentration. 
+11. Scale bars are often missing.  
+
+## Reviewer #2 (Remarks to the Author):  
+
+Referee report  
+
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+Authors: Simone Poprawa,1 Michele Stasi,1 Monika Wenisch,1 Brigitte A. K. Kriebisch,1 Judit Sastre,1 Job Boekhoven1  
+
+Reviewer: The article presents an elegant system for the construction of active droplets using dynamic phosphorylation in an enzyme- free manner, and hence furnishes a platform that others can readily built upon. This work will likely be picked up rapidly by several communities (for instance systems chemistry, active matter, and origins of life) and can be
+
+<--- Page Split --->
+
+expected to become a highly cited paper. As such, I am overall favorable to its publication in Nature Communications. Nevertheless, in terms of writing, presentation and contextualization, the present version feels unfinished and the scholarly quality of the work can be favorably improved.  
+
+In chronological order:  
+
+## Abstract  
+
+Reviewer: The abstract provides hardly more information than the title about the content of the paper and remains vague in instances where being precise would hardly consume more space, for instance it may be pointed out that the system is specific to histidine, and what phosphorylating agent is used.  
+
+Authors: Enzyme- free, synthetic systems that use dynamic phosphorylation to regulate supramolecular processes do not exist  
+
+Reviewer: Do not exist is a confusing phrasing. The authors presumably mean that such systems 'have not yet been reported, to the best of our knowledge'.  
+
+Authors: Our new reaction cycle will be valuable as a model for biological phosphorylation but can also offer insights into protocell formation.  
+
+Reviewer: Given the long history of MAP and DAP chemistry and renewed recent interest (which should be cited), the possessive phrasing 'our new reaction cycle' feels unnecessary here. The article does not clarify in what sense(s) the reaction cycle may be expected to be deliver on these promises. Since the phosphorylation is specific to histidine, it is not a priori obvious that it provides a model for biological phosphorylation.  
+
+## Introduction  
+
+Reviewer: The introduction explains at length certain aspects that are widely known, while assuming intimate familiarity of the reader with communities and highly specific branches of literature that are not cited. In particular, pertinent recent literature involving phosphates in systems chemistry and MAP/DAP in prebiotic chemistry appears to be overlooked. With respect to the former, notably:
+
+<--- Page Split --->
+
+Spontaneous and Selective Peptide Elongation in Water Driven by Aminoacyl Phosphate Esters and Phase Changes, Kun Dai, Mahesh D. Pol, Lenard Saile, Arti Sharma, Bin Liu, Ralf Thomann, Johanna L. Trefs, Danye Qiu, Sandra Moser, Stefan Wiesler, Bizan N. Balzer, Thorsten Hugel, Henning J. Jessen, and Charalampos G. Pappas\*, J. Am. Chem. Soc. 2023, 145, 48, 26086- 26094, https://doi.org/10.1021/jacs.3c07918A Ribonucleotide <-> Phosphoramidate Reaction Network Optimized by Computer- Aided Design. Andreas Englert, Julian F. Vogel, Tim Bergner, Jessica Loske, and Max von Delius. J. Am. Chem. Soc. 2022, 144, 33, 15266- 15274  
+
+Authors: Does not occur on a reasonable time scale; Reviewer: This phrase is grammatically incorrect.  
+
+Authors: Their need for highly evolved enzymes also offers limited insight into phosphorylation before life existed  
+
+Reviewer: This statement is suggestive and lacks nuance. It suggests an implicit historical claim that phosphorylation is a process that predates life. This is a popular - though unproven - conjecture in a niche of origins of life, where it is more broadly believed that modern biomolecules coincide with those present at abiogenesis. Since much of the prospective readership is not directly from the origins community and may not be aware of niche conjectures, any reference to them needs to be contextualized. If the authors choose to do so, they may also want to cite appropriate literature that lays out these conjectures and viewpoints thereon. A recent work on conjectured roles of phosphate groups in prebiotic chemistry is for instance: Ring Opening of Glycerol Cyclic Phosphates Leads to a Diverse Array of Potentially Prebiotic Phospholipids, Maiia Aleksandrova, Fidan Rahmatova, David A. Russell, and Claudia Bonfio. J. Am. Chem. Soc. 2023, 145, 47, 25614- 25620, https://doi.org/10.1021/jacs.3c07319  
+
+It should be stressed that biochemistry- first as a conjecture has been questioned many times over, even by its proponents within the origins literature, a relatively recent source being:  
+
+Life's Biological Chemistry: A Destiny or Destination Starting from Prebiotic Chemistry? R. Krishnamurthy, Chem. Eur. J. 2018, 24, 16708.  
+
+A more nuanced phrasing would be '.. into the roles phosphorylation may potentially have
+
+<--- Page Split --->
+
+played before life existed'.  
+
+Authors: ...prebiotically plausible phosphorylating agentsReviewer: Prebiotic plausibility constitutes a scientific claim that needs to be motivated and whose premises need to be stipulated. In the present version of the article we find no demonstration of this claim, nor references to the literature where such a claim is voiced. Although the rigor behind this terminology has undergone considerable erosion in the last 50 years, members of the community have started to voice concerns (see: Benner SA. Prebiotic plausibility and networks of paradox- resolving independent models. Nat. Commun. 2018 Dec 12;9(1):5173. doi: 10.1038/s41467- 018- 07274- y). One should not freely call a compound 'prebiotically plausible', it needs to be motivated.In the case of DAP and MAP, there is indeed a recent body of literature that explores and argues for these species to be prebiotically plausible phosphorylating agents, for instance: Diamidophosphate (DAP) – A Plausible Prebiotic Phosphorylating Reagent with a Chem to BioChem Potential? Osumah, A.; Krishnamurthy, R. ChemBioChem, 2021, 22, 3001- 3009. Geochemical Sources and Availability of Amidophosphates on the Early Earth. Gibard, C.; Jiménez, E. I.; Gorrell, I. B.; Kee, T. P.; Pasek, M. A.; Krishnamurthy, R. Angew. Chemie Int. Ed. 2019, 58, 8151- 8155.  
+
+Authors: The reaction cycle is enzyme- free and produces a phosphorylated amino acid or peptide that hydrolyzes without requiring enzymes with a tunable half- life.Reviewer: This phrase can be improved, for instance by adding 'and' in front of 'with a tunable half- life'.  
+
+Results and discussionExploration of the chemical reaction network.Reviewer: Some minor mistakes.Prebiotically -> prebiotically  
+
+The kinetic model is unusually predictive for a system of this level of complexity, especially for the species 3- pH. Nevertheless, there is unexplained variation, i.e. qualitative and quantitative deviations for the other species. It would be instructive if the authors could include a short speculation on the possible sources of this variation.
+
+<--- Page Split --->
+
+The labels C, D, F are missing from Fig.3.  
+
+## Conclusions  
+
+Reviewer: The conclusion is rather short, mentions few results and is not very specific. I recommend expanding the conclusion to be more reflective of the content of the paper.  
+
+## References  
+
+Reviewer: There is an error in the author list for reference 7.  
+
+Sup. Matt.  
+
+Reviewer: Among the most valuable products of this research is the high- quality kinetic data, involving multiple species, many timepoints, and repetitions. If this was not already being done, I encourage the authors to make it as easy as possible for their colleagues to access the raw data, e.g. by storing it in online repositories.  
+
+## Reviewer #3 (Remarks to the Author):  
+
+This manuscript describes the use of histidine phosphorylation by amidophosphates as a new reaction network for nonequilibrium systems chemistry. The authors establish the fundamental chemistry by optimizing the conditions and demonstrate its applicability through the formation of transient coacervate droplets.  
+
+The design of this system is simple and very elegant, and in my view is likely to be of broad interest to the field, especially considering some relevance to prebiotic chemistry. The authors have been thorough in exploring the parameter space for this reaction. I support publication in Nature Communications subject to a few comments.  
+
+1. The authors make use of a large, complex kinetic model. My major concern is the possibility that some of the parameters determined by fitting may be correlated (i.e., the model cannot measure them accurately because simultaneous changes in two or more cancel out in the concentration vs time simulations). If this is the case, the reported parameters (Table S1) may not be all that meaningful. There are a few things I noticed that
+
+<--- Page Split --->
+
+seem like red flags:  
+
+- In some cases (e.g., k3 for Ac-GHG-OH vs acY'DHDDAM) there seem to be very large discrepancies in the k values for what would seem to be similar reactions. 
+- In other cases, the errors determined (I assume) from the regression appear to be very large relative to the k's themselves (e.g., k-5, k7). 
+- In one case k-4 << k-1, which seems chemically unreasonable.  
+
+This issue should be discussed in the SI. If it cannot be adequately addressed, I would suggest removing explicit discussion of any parameters that are not determined by the data. It may only be possible to state that the mechanism is consistent with the data but that specific parameters cannot be accurately determined. To my mind this would not substantially affect interest in this work.  
+
+2. On a related note, how were the errors determined on the parameters in Tables S1 and S2? Is this through multiple replicates, or are they errors from the fits? This should be specified.  
+
+3. Some authors, notably Astumian, Aprahamian, and Goldup, have objected to how some terminology is used in this field and how some basic concepts are frequently described. I don't personally take as extreme a view, but I do think there is merit to these arguments. The manuscript describes unstable fuel molecules decomposing to waste, which I believe is one of the explanations that is often objected to. Really the issue is not specific unstable molecules, but rather that the reaction network must be driven by a reaction where the reactant and product concentrations are out of equilibrium; the distinction between fuel and waste is essentially arbitrary. That is, ADP can be a fuel for a reaction network if its concentration is above its equilibrium concentration. I would suggest making the language used in the introduction of this manuscript a bit more rigorous.  
+
+4. I'm a bit confused by how scientific notation is being used throughout the SI. In Table S1, there are frequent examples of numbers and uncertainties with both parts expressed in scientific notation (e.g., 1.40e-3 ± 7.58e-5). In Table S2, however, it seems like the exponent
+
+<--- Page Split --->
+
+of the uncertainty is supposed to apply to both parts (e.g., \(2.34 \pm 0.11e - 1 = 2.34e - 1 \pm 0.11e - 1?\) ). I'm therefore not sure what convention some numbers are following (e.g., \(2.17 \pm 0.50e - 3\) in Table S1). This should be clarified. I would personally suggest making it very explicit throughout, such as \("(2.34 \pm 0.11) \times 10^{- 1}"\) .  
+
+5. There are some incorrect table numberings in the SI. For example, there are frequent references to Table S7 that I think are supposed to be to Table S1. These should all be checked.  
+
+6. The notation is a bit confusing. For example, "3-pHis" (text) and "3-pH" (figures) appear to be used interchangeably.  
+
+7. I may be misunderstanding, but I believe the discussion on lines 113–115 is saying that the half life for loss of phosphate from 1,3-bpHis is the same or slightly longer than that for 1-pHis. This seems unlikely? If there are correlated parameters as noted above, I am concerned about quantitative assessments like this (depending on what's correlated with what, of course).  
+
+8. A bunch of the labels (C, D, F) are missing in Figure 3.  
+
+9. I noticed that the authors have tended to use near-stoichiometric or sub-stoichiometric ratios of MAP to peptide. Is there a reason for this? Especially considering that the system is not tremendously efficient (Figure 2F/G), it would seem like excess fuel might be useful.
+
+<--- Page Split --->
+
+Response to the reviewers for: NCOMMS- 23- 62971  
+
+Reviewer #1 (Remarks to the Author):  
+
+Key results: The authors describe a strategy to transiently phosphorylate histidine, with the goal of mimicking the transient activation of enzymes in cells. Monoamidophosphate (MAP) is the main phosphorylating agent, and water is the dephosphorylating agent. Histidine catalyses the hydrolysis of MAP, getting phosphorylated/de- phosphorylated in the process. When histidine is embedded in a tripeptide, de- phosphorylation proves to be too slow for the reaction cycle to be considered dynamic, and the authors deploy pyridine to accelerate it. In the presence of MAP and pyridine, the half- life of the phosphorylated peptide is reduced to a few hours. Experimental data and kinetic modeling are used to depict the flow of phosphate in the system. The authors show that when the reaction is carried out in the presence of a polycation, supramolecular assemblies (coacervates) form and disappear following the dynamic phosphorylation of histidine.  
+
+Authors: We thank the reviewer for the time invested in the manuscript. We addressed the comments below.  
+
+Validity & Significance: The experimental data in the manuscript is complete and the kinetic model provides a good fit. However, I believe the concept and strategy described in this manuscript are not significantly novel, and would not have a sufficient impact in the community to grant publication in Nature Communications. Transient compartmentalization controlled by phosphorylation has been shown previously (DOI: 10.1038/nchem.2414, 10.1039/c7sm01897e). It is true that the manuscript provides a non- enzymatic version, but still the applicability is limited by the fact that the reaction only works with the pair MAP/histidine (tables in Figure 1). We have now added the reference of Keating and Huck's work. Moreover we also cited Spruijt's work describing ATP based active droplets. Recent literature on prebiotic phosphorylation tends to look for agents with a broader spectrum (e.g. DOI 10.1021/jacs.3c08539, 10.1038/s41467- 021- 25555- x); or in the case of histidine itself, for functional reactions with greater resemblance to natural catalytic cycles (see alternative with imidazole phosphate in Maguire ChemRXiv 2023). For the reaction cycle to operate like in biology, both activation and deactivation occur on similar timescales without side reactions. In this work, we focused on a quantitative understanding of a chemical reaction cycler rather than its breadth. While we are convinced there are other substrate and fuel combinations, besides the ones that we tried, that would also work, we
+
+<--- Page Split --->
+
+decided to focus on the combination of histidine and its peptide derivatives with MAP as a fuel.  
+
+MAP, as the authors point out themselves, is not the most prebiotically plausible phosphorylating agent; and the use of pyridine as a substrate (that is mostly phosphorylated directly and competes with histidine) is somewhat confusing and undermines, in my opinion, its potential impact in the prebiotic chemistry field.  
+
+We appreciate the input, and we toned down the claim regarding prebiotic relevance. Moreover, in line with reviewer two, we cited recent work on the prebiotic relevance of MAP. The results on the formation of compartments are also weakened by the use of PEG 8k (at \(2.5\%\) concentration), a crowding agent known to promote and interfere with phase separation - more than a simple turbidity enhancer. Combined with the fact that most micrographs show only one droplet, or one microreactor, it indicates that the supramolecular results may be not robust or reproducible, and that the impact on the protocell/dynamic assemblies community would be very limited.  
+
+The way the data was presented indeed only showed one droplet per micrograph. We now added further evidence of the droplet formation. On the one hand, we show the formation of multiple droplets within one microreactor without the need of PEG. On the other hand, we show multiple microreactors containing droplets within one micrograph. Finally, the evolution of the droplet volume over time was performed in triplicate demonstrating that it is reproducible.  
+
+I include below some specific minor and major concerns about the manuscript.  
+
+Clarity and context:  
+
+1. The authors use terms such as "active droplets" in the title, "self-sustaining complex coacervate droplets", but do not justify it in the text or in the results. Active droplets would imply a non-equilibrium behavior, which is not clear to me in the data (the droplets emerge, coalesce, and dissolve). Self-sustaining would imply that the droplets facilitate the phosphorylation reaction that leads to their own formation, which was not probed in the manuscript.  
+
+We agree that self- sustaining would imply a more complex system that can, for example, facilitate its own formation. To clarify and not confuse our readers, we've changed from self- sustaining to active, with active meaning with activated precursor/product present (highlighted in yellow).
+
+<--- Page Split --->
+
+Our work demonstrates the emergence, the coalescence, and the dissolution of droplets, which can only happen out of equilibrium.  
+
+2. The description of the reaction as a cycle, or as a network (lines 64, 129), may be misleading, with the deactivation agent being the reaction's solvent (or pyridine), and the fuel MAP not being regenerated.  
+
+This was not meant to be misleading. We picture it as a cycle from the histidine's point of view. The histidine in our cycle undergoes cyclical phosphorylation drive by the hydrolysis of phosphorylating agent. Conceptually, this is similar to how proteins can undergo cyclical phosphorylation and dephosphorylation driven by the hydrolysis of ATP. We've now further clarified in the main text (highlighted in yellow).  
+
+3. I am missing the explanation as to how short of a half-life is desirable in a biological or chemical context, and why the choice to rely on water or pyridine as de-activating agents, as opposed of some prebiotically relevant substrates. The authors could include a more specific discussion on histidyl kinases, instead of the general introduction on ATP. We added a part on the histidine kinases to the introduction (highlighted in yellow). We were interested in what influences the half-lives.  
+
+4. Figures: scheme 1 would be more appropriate as a graphical abstract, it is taking up quite a lot of space but not adding as much information. Figure 1 could be reorganized to take up less space (the table format may not be necessary).  
+
+We agree scheme 1 is more like a graphical abstract, and we reduced the size and compressed it. Nevertheless, it includes all the information needed (histidine-based precursor as 5membered ring, phosphorylation coupled to droplet formation with information about the constitution of the droplets). Besides nature communications does not work with graphical abstract, which is why we use Scheme 1 to graphically demonstrate the idea of the work.  
+
+In our opinion, the table format in Fig. 1 is in this case useful, nicely aligned and gives a good overview.  
+
+Figure 2 contains some redundant data, panels F and G could be in supplementary information. Instead, a table for half-life and conversion% would be helpful. In our opinion, the panels (bar graphs) are clearer than tables. In a table the conversion% is less obvious than the bar graph with colorfully highlighted %ratio, as well as the change of half-lives under different conditions. But if you want to directly compare the numbers, we added the half-lives to Table S1 and S3 in the SI.  
+
+In Figure 3, what does Y'Dam mean?
+
+<--- Page Split --->
+
+We've changed it to peptide 1.  
+
+5. The clarity of the manuscript could be increased by changing the acronyms. Instead of 1- pH, His1-P for example, as pH already has its own meaning.  
+
+We appreciate the feedback. It is also in agreement with the literature, so we changed it to x- pHis.  
+
+6. Line 27: The sentence starting with "Does not occur..." is incomplete. The text from line 27 up to 39 needs to be restructured, it is confusing and not conveying its message – which I think is that biological cycles benefit from using ATP as a fuel due to it being thermodynamically unstable but kinetically stable.  
+
+We rephrased and clarified the sentences.  
+
+7. Line 187: there is a typo, in "Either it is hydrolyzed directly or through phosphorylating histidine (unwanted)" it should be pyridine, not histidine.  
+
+In line 192 there is a repetition "activation phosphorylation". This paragraph could be made clearer and the quantitative data mentioned could be included in a table elsewhere for clarity. Thank you, we rephrased the sentence.  
+
+Data and methodology:  
+
+8. Both experimental and modeling kinetic data are well presented, error bars are shown but it is not mentioned if the mean is taken over multiple experiments or different measurements or time points. I appreciate how the methodology was extensively described.  
+
+We appreciate the comment. In line with nature communication policy, we updated the presentation of the experimental data. We now show each data point of each experiment. The only data with mean error bars is the droplet volume/microreactor volume (droplets analyzed per micrograph \(>10\) ), but the triplicates are plotted individually.  
+
+9. Overall, some experiments could be better motivated in the text: for example, it is not clear to me that such a complete kinetic modeling is necessary, or how it can contribute to the understanding of the system.  
+
+We performed the kinetic modeling to predict the concentration evolution of the different species as precisely as possible. We also want to help others start a project based on this system. We have clarified in the main text.  
+
+Another example is why the authors choose to conduct microscopy in microfluidically- produced micro- reactors, and not on a typical microscopy slide/chamber - what advantages to the paper's goals does it bring?  
+
+We address this comment in point 10.  
+
+The ITC and DNA partitioning data is included but not well integrated within the text - how is this data relevant to the manuscript?
+
+<--- Page Split --->
+
+We integrated it better in the text.  
+
+10. The microscopy images do not immediately point to coacervate droplets (if you compare to other peptide/peptide complex coacervates in related literature), although I understand that FRAP supports it. Perhaps resolution and visibility would be increased without the microreactors. And instead of adding PEG, I would consider reducing buffer concentration. Thank you for your comment. We have now added data that shows that the droplets fuse. Besides, we have clarified that we have used the dye sulforhodamine B, a dye that is water soluble demonstrating an aqueous interior. We used microreactors to analyze the droplets because the droplets do not adhere to the microreactor walls. In contrast, they quickly adhere to glass slides. We clarified this in the main text (highlighted in yellow). We use PEG instead of reducing the buffer concentration to keep the ionic strength the same as in the microreactors. 11. Scale bars are often missing. Thank you for your comment we added them, where needed.
+
+<--- Page Split --->
+
+Reviewer #2 (Remarks to the Author):  
+
+## Referee report  
+
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+Authors: Simone Poprawa,1 Michele Stasi,1 Monika Wenisch,1 Brigitte A. K. Kriebisch,1 Judit Sastre,1 Job Boekhoven1  
+
+Reviewer: The article presents an elegant system for the construction of active droplets using dynamic phosphorylation in an enzyme- free manner, and hence furnishes a platform that others can readily built upon. This work will likely be picked up rapidly by several communities (for instance systems chemistry, active matter, and origins of life) and can be expected to become a highly cited paper. As such, I am overall favorable to its publication in Nature Communications. Nevertheless, in terms of writing, presentation and contextualization, the present version feels unfinished and the scholarly quality of the work can be favorably improved.  
+
+Thank you for your time invested in the manuscript. We have addressed your comments below.  
+
+In chronological order:  
+
+## Abstract  
+
+Reviewer: The abstract provides hardly more information than the title about the content of the paper and remains vague in instances where being precise would hardly consume more space, for instance it may be pointed out that the system is specific to histidine, and what phosphorylating agent is used.  
+
+We rephrased the sentences.  
+
+Authors: Enzyme- free, synthetic systems that use dynamic phosphorylation to regulate supramolecular processes do not exist  
+
+Reviewer: Do not exist is a confusing phrasing. The authors presumably mean that such systems 'have not yet been reported, to the best of our knowledge'.  
+
+Thank you, we rephrased the sentences.  
+
+Authors: Our new reaction cycle will be valuable as a model for biological phosphorylation but can also offer insights into protocell formation.  
+
+Reviewer: Given the long history of MAP and DAP chemistry and renewed recent interest (which should be cited), the possessive phrasing 'our new reaction cycle' feels unnecessary here. The article does not clarify in what sense(s) the reaction cycle may be expected to be deliver on these promises. Since the phosphorylation is specific to histidine, it is not a priori obvious that it provides a model for biological phosphorylation.
+
+<--- Page Split --->
+
+Indeed, "new" is too possessive. We added a part of histidine phosphorylation in the introduction (highlighted in yellow).  
+
+Introduction  
+
+Reviewer: The introduction explains at length certain aspects that are widely known, while assuming intimate familiarity of the reader with communities and highly specific branches of literature that are not cited. In particular, pertinent recent literature involving phosphates in systems chemistry and MAP/DAP in prebiotic chemistry appears to be overlooked.  
+
+With respect to the former, notably:  
+
+Spontaneous and Selective Peptide Elongation in Water Driven by Aminoacyl Phosphate Esters and Phase Changes, Kun Dai, Mahesh D. Pol, Lenard Saile, Arti Sharma, Bin Liu, Ralf Thomann, Johanna L. Trefs, Danye Qiu, Sandra Moser, Stefan Wiesler, Bizan N. Balzer, Thorsten Hugel, Henning J. Jessen, and Charalampos G. Pappas\*, J. Am. Chem. Soc. 2023, 145, 48, 26086- 26094, https://doi.org/10.1021/jacs.3c07918  
+
+A Ribonucleotide \(\leq - >\) Phosphoramidate Reaction Network Optimized by Computer- Aided Design. Andreas Englert, Julian F. Vogel, Tim Bergner, Jessica Loske, and Max von Delius. J. Am. Chem. Soc. 2022, 144, 33, 15266- 15274  
+
+We added the references.  
+
+Authors: Does not occur on a reasonable time scale; Reviewer: This phrase is grammatically incorrect.  
+
+We rephrased the sentence.  
+
+Authors: Their need for highly evolved enzymes also offers limited insight into phosphorylation before life existed  
+
+Reviewer: This statement is suggestive and lacks nuance. It suggests an implicit historical claim that phosphorylation is a process that predates life. This is a popular - though unproven - conjecture in a niche of origins of life, where it is more broadly believed that modern biomolecules coincide with those present at abiogenesis. Since much of the prospective readership is not directly from the origins community and may not be aware of niche conjectures, any reference to them needs to be contextualized. If the authors choose to do so, they may also want to cite appropriate literature that lays out these conjectures and viewpoints thereon. A recent work on conjectured roles of phosphate groups in prebiotic chemistry is for instance: Ring Opening of Glycerol Cyclic Phosphates Leads to a Diverse Array of Potentially Prebiotic Phospholipids, Maia Aleksandrova, Fidan Rahmatova, David A. Russell, and Claudia Bonfio. J. Am. Chem. Soc. 2023, 145, 47, 25614- 25620, https://doi.org/10.1021/jacs.3c07319  
+
+It should be stressed that biochemistry- first as a conjecture has been questioned many times over, even by its proponents within the origins literature, a relatively recent source being:
+
+<--- Page Split --->
+
+Life's Biological Chemistry: A Destiny or Destination Starting from Prebiotic Chemistry? R. Krishnamurthy, Chem. Eur. J. 2018, 24, 16708.  
+
+A more nuanced phrasing would be '.. into the roles phosphorylation may potentially have played before life existed'.  
+
+We totally agree and appreciate the thoughtful feedback. We nuanced the statement and added supporting citations.  
+
+Authors: ...prebiotically plausible phosphorylating agents  
+
+Reviewer: Prebiotic plausibility constitutes a scientific claim that needs to be motivated and whose premises need to be stipulated. In the present version of the article we find no demonstration of this claim, nor references to the literature where such a claim is voiced. Although the rigor behind this terminology has undergone considerable erosion in the last 50 years, members of the community have started to voice concerns (see: Benner SA. Prebiotic plausibility and networks of paradox- resolving independent models. Nat. Commun. 2018 Dec 12;9(1):5173. doi: 10.1038/s41467- 018- 07274- y). One should not freely call a compound 'prebiotically plausible', it needs to be motivated.  
+
+In the case of DAP and MAP, there is indeed a recent body of literature that explores and argues for these species to be prebiotically plausible phosphorylating agents, for instance: Diamidophosphate (DAP) - A Plausible Prebiotic Phosphorylating Reagent with a Chem to BioChem Potential? Osumah, A.; Krishnamurthy, R. ChemBioChem, 2021, 22, 3001- 3009. Geochemical Sources and Availability of Amidophosphates on the Early Earth. Gibard, C.; Jiménez, E. I.; Gorrell, I. B.; Kee, T. P.; Pasek, M. A.; Krishnamurthy, R. Angew. Chemie Int. Ed. 2019, 58, 8151- 8155.  
+
+Thank you for reminding us, of course these are very important publications and need to be cited.  
+
+Authors: The reaction cycle is enzyme- free and produces a phosphorylated amino acid or peptide that hydrolyzes without requiring enzymes with a tunable half- life. Reviewer: This phrase can be improved, for instance by adding 'and' in front of 'with a tunable half- life'.  
+
+We changed the sentence.  
+
+Results and discussionExploration of the chemical reaction network.Reviewer: Some minor mistakes.Prebiotically -> prebiotically
+
+<--- Page Split --->
+
+We corrected the typo.  
+
+The kinetic model is unusually predictive for a system of this level of complexity, especially for the species 3- pH. Nevertheless, there is unexplained variation, i.e. qualitative and quantitative deviations for the other species. It would be instructive if the authors could include a short speculation on the possible sources of this variation.  
+
+The reaction rate constant of the 3- pHs was determined empirically when no other species or fuel were present anymore. Therefore, the predicted and the experimentally measured evolution fit nicely. The hydrolysis of MAP was also calculated empirically, but the other constants (up to 17) were fitted and thus show higher deviations (Table S1, error of the fit) as there are many combinations possible, also with correlated parameters. Despite that, the fits align very well with the experimental data.  
+
+As you and reviewer 3 mentioned, the kinetic model is very good for predicting the evolution of a single species and can potentially help to design a self- assembling system. However, the reaction rate constants, which are not determined by the data, should not be compared or addressed directly. We have removed any references to half- lives or rate constants that were determined by the kinetic model.  
+
+Besides, we have added the comment on the value and the limitation of the kinetic model in the SI.  
+
+The labels C, D, F are missing from Fig.3.  
+
+We changed the presentation of our data and also added the labels.  
+
+Conclusions  
+
+Reviewer: The conclusion is rather short, mentions few results and is not very specific. I recommend expanding the conclusion to be more reflective of the content of the paper. We have addressed your comments on the conclusion.  
+
+References  
+
+Reviewer: There is an error in the author list for reference 7. We updated the references.  
+
+Sup. Matt.  
+
+Reviewer: Among the most valuable products of this research is the high- quality kinetic data, involving multiple species, many timepoints, and repetitions. If this was not already being done, I encourage the authors to make it as easy as possible for their colleagues to access the raw data, e.g. by storing it in online repositories.  
+
+Thank you, we appreciate that! We added the Excel sheet with the raw data of the kinetic experiments and the COPASI sample file.
+
+<--- Page Split --->
+
+Reviewer #3 (Remarks to the Author):  
+
+This manuscript describes the use of histidine phosphorylation by amidophosphates as a new reaction network for nonequilibrium systems chemistry. The authors establish the fundamental chemistry by optimizing the conditions and demonstrate its applicability through the formation of transient coacervate droplets.  
+
+The design of this system is simple and very elegant, and in my view is likely to be of broad interest to the field, especially considering some relevance to prebiotic chemistry. The authors have been thorough in exploring the parameter space for this reaction. I support publication in Nature Communications subject to a few comments.  
+
+Thank you for your positive feedback, we appreciate your time invested in the manuscript. 1. The authors make use of a large, complex kinetic model. My major concern is the possibility that some of the parameters determined by fitting may be correlated (i.e., the model cannot measure them accurately because simultaneous changes in two or more cancel out in the concentration vs time simulations). If this is the case, the reported parameters (Table S1) may not be all that meaningful. There are a few things I noticed that seem like red flags:  
+
+- In some cases (e.g., k3 for Ac-GHG-OH vs acY'DHDDam) there seem to be very large discrepancies in the k values for what would seem to be similar reactions. 
+- In other cases, the errors determined (I assume) from the regression appear to be very large relative to the k's themselves (e.g., k-5, k7). 
+- In one case k-4 << k-1, which seems chemically unreasonable.  
+
+This issue should be discussed in the SI. If it cannot be adequately addressed, I would suggest removing explicit discussion of any parameters that are not determined by the data. It may only be possible to state that the mechanism is consistent with the data but that specific parameters cannot be accurately determined. To my mind this would not substantially affect interest in this work.  
+
+Thank you for your detailed comment. We have distanced ourselves to compare the fitted values directly, but use them as a guideline and compare empirically determined parameters. Please see the answer to reviewer 2.  
+
+2. On a related note, how were the errors determined on the parameters in Tables S1 and S2? Is this through multiple replicates, or are they errors from the fits? This should be
+
+<--- Page Split --->
+
+specified.  
+
+We added the explanation to the description of the tables.  
+
+3. Some authors, notably Astumian, Aprahamian, and Goldup, have objected to how some terminology is used in this field and how some basic concepts are frequently described. I don't personally take as extreme a view, but I do think there is merit to these arguments. The manuscript describes unstable fuel molecules decomposing to waste, which I believe is one of the explanations that is often objected to. Really the issue is not specific unstable molecules, but rather that the reaction network must be driven by a reaction where the reactant and product concentrations are out of equilibrium; the distinction between fuel and waste is essentially arbitrary. That is, ADP can be a fuel for a reaction network if its concentration is above its equilibrium concentration. I would suggest making the language used in the introduction of this manuscript a bit more rigorous.  
+
+We have adjusted the introduction to define fuel better.  
+
+4. I'm a bit confused by how scientific notation is being used throughout the SI. In Table S1, there are frequent examples of numbers and uncertainties with both parts expressed in scientific notation (e.g., \(1.40\mathrm{e - 3} \pm 7.58\mathrm{e - 5}\) ). In Table S2, however, it seems like the exponent of the uncertainty is supposed to apply to both parts (e.g., \(2.34 \pm 0.11\mathrm{e - 1} = 2.34\mathrm{e - 1} \pm 0.11\mathrm{e - 1}\) ?). I'm therefore not sure what convention some numbers are following (e.g., \(2.17 \pm 0.50\mathrm{e - 3}\) in Table S1). This should be clarified. I would personally suggest making it very explicit throughout, such as \((2.34 \pm 0.11) \times 10^{-1}\) .  
+
+Thank you, we changed it.  
+
+5. There are some incorrect table numberings in the SI. For example, there are frequent references to Table S7 that I think are supposed to be to Table S1. These should all be checked.  
+
+We updated the numbering.  
+
+6. The notation is a bit confusing. For example, "3-pHis" (text) and "3-pH" (figures) appear to be used interchangeably.  
+
+To be consistent, also with literature, we changed to x-pHis.  
+
+7. I may be misunderstanding, but I believe the discussion on lines 113-115 is saying that the half life for loss of phosphate from 1,3-bpHis is the same or slightly longer than that for 1-pHis. This seems unlikely? If there are correlated parameters as noted above, I am concerned about quantitative assessments like this (depending on what's correlated with what, of course).  
+
+See the answer to your comment 1.  
+
+8. A bunch of the labels (C, D, F) are missing in Figure 3. We changed presentation of our data and also added the labels.
+
+<--- Page Split --->
+
+9. I noticed that the authors have tended to use near-stoichiometric or sub-stoichiometric ratios of MAP to peptide. Is there a reason for this? Especially considering that the system is not tremendously efficient (Figure 2F/G), it would seem like excess fuel might be useful. We agree that the yield would be higher with a higher amount of fuel. However, an excess of fuel also elongates the total experimental time further. To keep the experiment on a reasonable timescale, we usually work with stoichiometric amounts. Moreover, to potentially observe the bisphosphorylated species, we used nearly stoichiometric and high amount of peptide to fuel.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+The authors addressed some of my concerns, but my major concerns regarding the publication of this manuscript in Nature Communications remain. That said, at least the writing needs further improvement in order for me to be comfortable supporting publication. I really think the manuscript will benefit from a more rigorous writing, and that time put on these final changes will be well spent.  
+
+I refer to the numbering in my previous report:  
+
+1. Partially addressed. In line 229, I disagree from the definition of active droplets or active protocells (also comparing to David Zwicker's group definition): In this context, active means that the droplets are sustained as long as a certain amount of product is present. The way it is written at least, it is not a remarkable behaviour – that the droplets exist while their building blocks exist. Reaction-controlled coacervates are not an active per se; it helps to make your case if you compare these dynamic droplets, sustained by the reaction, to droplets formed from building blocks that don't form or decay in situ.  
+
+2. Despite the explanation in the reply to my comments, the introduction text still lacks specificity. As a reader, I am still not convinced that this is a cycle, or a controlled cycle: surely, while there is fuel, there is phosphorylation; and surely when you stop adding fuel, hydrolysis dominates, but that does not mean control. A key point of a cycle would be to demonstrate that His phosphorylation (and coacervates) can be recovered by a fresh addition of MAP. Can you show this?  
+
+a. Line 27 is an example of what I mean by lack of specificity in the writing: That way, the equilibrium position for the fuel-to-waste equilibrium lies to the right such that most fuel will eventually be converted into waste. (and btw line 33 can start a new paragraph) 
+b. In line 287, you write "reaction cycle that phosphorylates histidine as an amino acid or in a peptide at the expense of phosphorylating agents". There is no need to say "at the expense of phosphorylating agents", this is the case in any chemical reaction. This and other similar repetitions take space from a more meaningful discussion/conclusion.
+
+<--- Page Split --->
+
+3. Addressed.  
+
+4. Since you chose to keep the table in Figure 1, I find it that there needs to be some discussion in the text as to why the cycle only works with histidine, or why, if we know from the literature that MAP phosphorylates imidazole groups, other residues were also tested. What was the hypothesis?  
+
+5. Addressed. 
+6. Addressed. 
+7. Addressed. 
+8. Addressed.  
+
+9. Partially addressed. The use of microreactors still confuses me, as there are many ways to passivate glass slides to prevent droplet adhesion. Can you at least include in SI micrographs of droplets outside these microreactors? This would show that the microreactor itself is not crucial to the droplets' properties.  
+
+10. In SI Figure 10, I see you now added the whole field of view, showing all microreactors. Can you include the emission channel as well, which is what you mostly show in the main text? The caption says it shows the emergence, growth and dissolution of the droplets, but I can only see emergence and dissolution.  
+
+11. Addressed.  
+
+New comments:  
+
+12. Line 161: consider reorganizing the long list of references. The comparison to the literature is barely made, yet there are 11 references listed. I prefer it if you elaborate on this discussion rather than remove references, as I think it will make the manuscript stronger.  
+
+13. The conclusions are still vague and imprecise despite another reviewer's comment. Most statements are based on this system being a dynamic cycle and the droplets being active, so
+
+<--- Page Split --->
+
+I think resolving these conceptual problems in the main text will help better support the conclusions.  
+
+## Reviewer #2 (Remarks to the Author):  
+
+All reviewer points are addressed well.  
+
+## Reviewer #3 (Remarks to the Author):  
+
+In my opinion, the authors have addressed the points I raised in my original review. I believe that the manuscript is now suitable for Nature Communications.
+
+<--- Page Split --->
+
+## REVIEWER COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+The authors addressed some of my concerns, but my major concerns regarding the publication of this manuscript in Nature Communications remain. That said, at least the writing needs further improvement in order for me to be comfortable supporting publication. I really think the manuscript will benefit from a more rigorous writing, and that time put on these final changes will be well spent.  
+
+I refer to the numbering in my previous report:  
+
+1. Partially addressed. In line 229, I disagree from the definition of active droplets or active protocells (also comparing to David Zwicker's group definition): In this context, active means that the droplets are sustained as long as a certain amount of product is present. The way it is written at least, it is not a remarkable behaviour – that the droplets exist while their building blocks exist. Reaction-controlled coacervates are not an active per se; it helps to make your case if you compare these dynamic droplets, sustained by the reaction, to droplets formed from building blocks that don't form or decay in situ.  
+
+We agree and changed it for:  
+
+In this context, active means that the droplet material is controlled by two chemical reactions: activation and deactivation.  
+
+We used the nomenclature as introduced by Zwicker: „Droplets can become chemically active if the material of the droplet is produced and destroyed by chemical reactions. An example that resembles a simple protocell is shown schematically in Fig. 1a. The droplet is formed by a droplet material D that is generated inside the droplet from a high-energy precursor N, which plays the role of a nutrient. Droplet material can degrade into a lower energy component, W, that plays the role of a waste, which leaves the droplet by diffusion. The droplet can survive if N is continuously supplied and W is continuously removed. This can be achieved by recycling N using an external energy source such as a fuel or radiation.“1  
+
+(1) Zwicker, D.; Seyboldt, R.; Weber, C. A.; Hyman, A. A.; Jülicher, F. Growth and division of active droplets provides a model for protocells. Nature Physics 2017, 13 (4), 408-413. DOI: 10.1038/nphys3984.  
+
+2. Despite the explanation in the reply to my comments, the introduction text still lacks specificity. As a reader, I am still not convinced that this is a cycle, or a controlled cycle: surely, while there is fuel, there is phosphorylation; and surely when you stop adding fuel, hydrolysis dominates, but that does not mean control. A key point of a cycle would be to demonstrate that His phosphorylation (and coacervates) can be recovered by a fresh addition of MAP. Can you show this?  
+
+We tested refueling with three cycles on the amino acid His on the NMR, to demonstrate that we can recover the phosphorylation by fresh addition of MAP. Besides, we also refueled the peptide with droplets two times on the plate reader. The graphs can be found in the SI.
+
+<--- Page Split --->
+
+a. Line 27 is an example of what I mean by lack of specificity in the writing: That way, the equilibrium position for the fuel-to-waste equilibrium lies to the right such that most fuel will eventually be converted into waste. (and btw line 33 can start a new paragraph)  
+
+We rephrased to: That way, the equilibrium position for the fuel-to-waste equilibrium lies on the waste side such that most fuel will eventually be converted into waste. This fuel-to-waste conversion should be slow so that a catalyst can accelerate it. Put differently, the fuel should be thermodynamically unstable, such as kinetically inert. In the catalytic reaction cycle, the catalyst accelerates the fuel-to-waste conversion. In doing so, the catalyst is temporarily activated by reacting with the fuel, after which it spontaneously deactivates. In other words, the catalyst can undergo numerous activation-deactivation cycles, and the fuel is converted fast due to the presence of the catalyst.  
+
+Highlighted in pink.  
+
+b. In line 287, you write "reaction cycle that phosphorylates histidine as an amino acid or in a peptide at the expense of phosphorylating agents". There is no need to say "at the expense of phosphorylating agents", this is the case in any chemical reaction. This and other similar repetitions take space from a more meaningful discussion/conclusion.  
+
+We agree that it is not needed. We removed it where we thought it was not required. Highlighted in pink.  
+
+## 3. Addressed.  
+
+4. Since you chose to keep the table in Figure 1, I find it that there needs to be some discussion in the text as to why the cycle only works with histidine, or why, if we know from the literature that MAP phosphorylates imidazole groups, other residues were also tested. What was the hypothesis?  
+
+We focused on amino acids and their peptides as they are powerful building blocks in peptide self-assembly, including liquid-liquid phase separation, which we show later in this paper. We emphasized our reasoning in the text and highlighted it in pink.  
+
+5. Addressed. 
+6. Addressed. 
+7. Addressed. 
+8. Addressed.  
+
+9. Partially addressed. The use of microreactors still confuses me, as there are many ways to passivate glass slides to prevent droplet adhesion. Can you at least include in SI micrographs of droplets outside these microreactors? This would show that the microreactor itself is not crucial to the droplets' properties.  
+
+To clarify, the microreactor helps us to study droplet formation without the downsides of flow in the sample, droplets settling on the glass, and droplets escaping out of the focal plane during imaging. We also understand your concern, so we added micrographs of droplets outside the microreactors to the SI.
+
+<--- Page Split --->
+
+10. In SI Figure 10, I see you now added the whole field of view, showing all microreactors. Can you include the emission channel as well, which is what you mostly show in the main text? The caption says it shows the emergence, growth and dissolution of the droplets, but I can only see emergence and dissolution. We added the emission channel of sulforhodamine B and the micrographs after 9 and 25 h. Fig. 3 f shows the growth more clearly as the percentage of droplet material per reactor correlates with the droplet's size.  
+
+## 11. Addressed.  
+
+New comments:  
+
+12. Line 161: consider reorganizing the long list of references. The comparison to the literature is barely made, yet there are 11 references listed. I prefer it if you elaborate on this discussion rather than remove references, as I think it will make the manuscript stronger.  
+
+Indeed, 11 references were a lot, and some of them fitted better to other statements that have been made before. Changes are highlighted in pink.  
+
+13. The conclusions are still vague and imprecise despite another reviewer's comment. Most statements are based on this system being a dynamic cycle and the droplets being active, so I think resolving these conceptual problems in the main text will help better support the conclusions.  
+
+We have now clarified what we mean by active droplets (following your earlier statement) and addressed the fact that it is a cycle (following your proposed experiments). So, the conceptual problems should now be resolved.
+
+<--- Page Split --->
+
+## REVIEWERS' COMMENTS  
+
+## Reviewer #1 (Remarks to the Author):  
+
+The authors have now addressed all remarks and included new experiments. I believe the manuscript's clarity and scientific rigor improved greatly during the revision process, and appreciate their effort in doing so. I support publication of the most recent version.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d7666663fe77730e88d0a0083aacf89a7b35973d18830a7da0a9d9465bb23514/supplementary_0_Peer Review File__d7666663fe77730e88d0a0083aacf89a7b35973d18830a7da0a9d9465bb23514_det.mmd

@@ -0,0 +1,750 @@
+<|ref|>title<|/ref|><|det|>[[60, 40, 506, 90]]<|/det|>
+# nature portfolio  
+
+<|ref|>text<|/ref|><|det|>[[66, 110, 360, 139]]<|/det|>
+Peer Review File  
+
+<|ref|>text<|/ref|><|det|>[[178, 155, 817, 210]]<|/det|>
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+<|ref|>image<|/ref|><|det|>[[57, 732, 240, 781]]<|/det|>  
+
+<|ref|>text<|/ref|><|det|>[[250, 732, 911, 784]]<|/det|>
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 85, 315, 101]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 136, 437, 153]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 161, 876, 443]]<|/det|>
+Reviewer #1 (Remarks to the Author):Key results: The authors describe a strategy to transiently phosphorylate histidine, with the goal of mimicking the transient activation of enzymes in cells. Monoamidophosphate (MAP) is the main phosphorylating agent, and water is the dephosphorylating agent. Histidine catalyses the hydrolysis of MAP, getting phosphorylated/de- phosphorylated in the process. When histidine is embedded in a tripeptide, de- phosphorylation proves to be too slow for the reaction cycle to be considered dynamic, and the authors deploy pyridine to accelerate it. In the presence of MAP and pyridine, the half- life of the phosphorylated peptide is reduced to a few hours. Experimental data and kinetic modeling are used to depict the flow of phosphate in the system. The authors show that when the reaction is carried out in the presence of a polycation, supramolecular assemblies (coacervates) form and disappear following the dynamic phosphorylation of histidine.  
+
+<|ref|>text<|/ref|><|det|>[[115, 475, 880, 912]]<|/det|>
+Validity & Significance: The experimental data in the manuscript is complete and the kinetic model provides a good fit. However, I believe the concept and strategy described in this manuscript are not significantly novel, and would not have a sufficient impact in the community to grant publication in Nature Communications. Transient compartmentalization controlled by phosphorylation has been shown previously (DOI: 10.1038/nchem.2414, 10.1039/c7sm01897e). It is true that the manuscript provides a non- enzymatic version, but still the applicability is limited by the fact that the reaction only works with the pair MAP/histidine (tables in Figure 1). Recent literature on prebiotic phosphorylation tends to look for agents with a broader spectrum (e.g. DOI 10.1021/jacs.3c08539, 10.1038/s41467- 021- 25555- x); or in the case of histidine itself, for functional reactions with greater resemblance to natural catalytic cycles (see alternative with imidazole phosphate in Maguire ChemRxiv 2023). MAP, as the authors point out themselves, is not the most prebiotically plausible phosphorylating agent; and the use of pyridine as a substrate (that is mostly phosphorylated directly and competes with histidine) is somewhat confusing and undermines, in my opinion, its potential impact in the prebiotic chemistry field. The results on the formation of compartments are also weakened by the use of PEG 8k (at 2.5% concentration), a crowding agent known to promote and interfere with phase separation –
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 84, 879, 181]]<|/det|>
+more than a simple turbidity enhancer. Combined with the fact that most micrographs show only one droplet, or one microreactor, it indicates that the supramolecular results may be not robust or reproducible, and that the impact on the protocell/dynamic assemblies community would be very limited.  
+
+<|ref|>text<|/ref|><|det|>[[118, 214, 766, 233]]<|/det|>
+I include below some specific minor and major concerns about the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[118, 266, 281, 283]]<|/det|>
+Clarity and context:  
+
+<|ref|>text<|/ref|><|det|>[[116, 291, 877, 442]]<|/det|>
+1. The authors use terms such as "active droplets" in the title, "self-sustaining complex coacervate droplets", but do not justify it in the text or in the results. Active droplets would imply a non-equilibrium behavior, which is not clear to me in the data (the droplets emerge, coalesce, and dissolve). Self-sustaining would imply that the droplets facilitate the phosphorylation reaction that leads to their own formation, which was not probed in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[116, 450, 858, 520]]<|/det|>
+2. The description of the reaction as a cycle, or as a network (lines 64, 129), may be misleading, with the deactivation agent being the reaction's solvent (or pyridine), and the fuel MAP not being regenerated.  
+
+<|ref|>text<|/ref|><|det|>[[116, 528, 880, 624]]<|/det|>
+3. I am missing the explanation as to how short of a half-life is desirable in a biological or chemical context, and why the choice to rely on water or pyridine as de-activating agents, as opposed of some prebiotically relevant substrates. The authors could include a more specific discussion on histidyl kinases, instead of the general introduction on ATP.  
+
+<|ref|>text<|/ref|><|det|>[[116, 632, 876, 756]]<|/det|>
+4. Figures: scheme 1 would be more appropriate as a graphical abstract, it is taking up quite a lot of space but not adding as much information. Figure 1 could be reorganized to take up less space (the table format may not be necessary). Figure 2 contains some redundant data, panels F and G could be in supplementary information. Instead, a table for half-life and conversion% would be helpful. In Figure 3, what does Y'Dam mean?  
+
+<|ref|>text<|/ref|><|det|>[[116, 763, 868, 808]]<|/det|>
+5. The clarity of the manuscript could be increased by changing the acronyms. Instead of 1- pH, His1-P for example, as pH already has its own meaning.  
+
+<|ref|>text<|/ref|><|det|>[[116, 815, 877, 911]]<|/det|>
+6. Line 27: The sentence starting with "Does not occur..." is incomplete. The text from line 27 up to 39 needs to be restructured, it is confusing and not conveying its message – which I think is that biological cycles benefit from using ATP as a fuel due to it being thermodynamically unstable but kinetically stable.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 852, 181]]<|/det|>
+7. Line 187: there is a typo, in "Either it is hydrolyzed directly or through phosphorylating histidine (unwanted)" it should be pyridine, not histidine. In line 192 there is a repetition "activation phosphorylation". This paragraph could be made clearer and the quantitative data mentioned could be included in a table elsewhere for clarity.  
+
+<|ref|>text<|/ref|><|det|>[[118, 216, 313, 233]]<|/det|>
+Data and methodology:  
+
+<|ref|>text<|/ref|><|det|>[[115, 240, 880, 610]]<|/det|>
+8. Both experimental and modeling kinetic data are well presented, error bars are shown but it is not mentioned if the mean is taken over multiple experiments or different measurements or time points. I appreciate how the methodology was extensively described. 
+9. Overall, some experiments could be better motivated in the text: for example, it is not clear to me that such a complete kinetic modeling is necessary, or how it can contribute to the understanding of the system. Another example is why the authors choose to conduct microscopy in microfluidically-produced micro-reactors, and not on a typical microscopy slide/chamber - what advantages to the paper's goals does it bring? The ITC and DNA partitioning data is included but not well integrated within the text - how is this data relevant to the manuscript? 
+10. The microscopy images do not immediately point to coacervate droplets (if you compare to other peptide/peptide complex coacervates in related literature), although I understand that FRAP supports it. Perhaps resolution and visibility would be increased without the micro-reactors. And instead of adding PEG, I would consider reducing buffer concentration. 
+11. Scale bars are often missing.  
+
+<|ref|>sub_title<|/ref|><|det|>[[119, 685, 438, 702]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 712, 243, 728]]<|/det|>
+Referee report  
+
+<|ref|>text<|/ref|><|det|>[[118, 737, 641, 755]]<|/det|>
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+<|ref|>text<|/ref|><|det|>[[118, 762, 880, 806]]<|/det|>
+Authors: Simone Poprawa,1 Michele Stasi,1 Monika Wenisch,1 Brigitte A. K. Kriebisch,1 Judit Sastre,1 Job Boekhoven1  
+
+<|ref|>text<|/ref|><|det|>[[117, 815, 866, 912]]<|/det|>
+Reviewer: The article presents an elegant system for the construction of active droplets using dynamic phosphorylation in an enzyme- free manner, and hence furnishes a platform that others can readily built upon. This work will likely be picked up rapidly by several communities (for instance systems chemistry, active matter, and origins of life) and can be
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 84, 877, 181]]<|/det|>
+expected to become a highly cited paper. As such, I am overall favorable to its publication in Nature Communications. Nevertheless, in terms of writing, presentation and contextualization, the present version feels unfinished and the scholarly quality of the work can be favorably improved.  
+
+<|ref|>text<|/ref|><|det|>[[118, 216, 306, 232]]<|/det|>
+In chronological order:  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 267, 191, 283]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[117, 292, 879, 390]]<|/det|>
+Reviewer: The abstract provides hardly more information than the title about the content of the paper and remains vague in instances where being precise would hardly consume more space, for instance it may be pointed out that the system is specific to histidine, and what phosphorylating agent is used.  
+
+<|ref|>text<|/ref|><|det|>[[117, 423, 835, 468]]<|/det|>
+Authors: Enzyme- free, synthetic systems that use dynamic phosphorylation to regulate supramolecular processes do not exist  
+
+<|ref|>text<|/ref|><|det|>[[118, 476, 838, 520]]<|/det|>
+Reviewer: Do not exist is a confusing phrasing. The authors presumably mean that such systems 'have not yet been reported, to the best of our knowledge'.  
+
+<|ref|>text<|/ref|><|det|>[[117, 554, 864, 599]]<|/det|>
+Authors: Our new reaction cycle will be valuable as a model for biological phosphorylation but can also offer insights into protocell formation.  
+
+<|ref|>text<|/ref|><|det|>[[117, 606, 870, 730]]<|/det|>
+Reviewer: Given the long history of MAP and DAP chemistry and renewed recent interest (which should be cited), the possessive phrasing 'our new reaction cycle' feels unnecessary here. The article does not clarify in what sense(s) the reaction cycle may be expected to be deliver on these promises. Since the phosphorylation is specific to histidine, it is not a priori obvious that it provides a model for biological phosphorylation.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 764, 224, 780]]<|/det|>
+## Introduction  
+
+<|ref|>text<|/ref|><|det|>[[117, 789, 883, 910]]<|/det|>
+Reviewer: The introduction explains at length certain aspects that are widely known, while assuming intimate familiarity of the reader with communities and highly specific branches of literature that are not cited. In particular, pertinent recent literature involving phosphates in systems chemistry and MAP/DAP in prebiotic chemistry appears to be overlooked. With respect to the former, notably:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 858, 285]]<|/det|>
+Spontaneous and Selective Peptide Elongation in Water Driven by Aminoacyl Phosphate Esters and Phase Changes, Kun Dai, Mahesh D. Pol, Lenard Saile, Arti Sharma, Bin Liu, Ralf Thomann, Johanna L. Trefs, Danye Qiu, Sandra Moser, Stefan Wiesler, Bizan N. Balzer, Thorsten Hugel, Henning J. Jessen, and Charalampos G. Pappas\*, J. Am. Chem. Soc. 2023, 145, 48, 26086- 26094, https://doi.org/10.1021/jacs.3c07918A Ribonucleotide <-> Phosphoramidate Reaction Network Optimized by Computer- Aided Design. Andreas Englert, Julian F. Vogel, Tim Bergner, Jessica Loske, and Max von Delius. J. Am. Chem. Soc. 2022, 144, 33, 15266- 15274  
+
+<|ref|>text<|/ref|><|det|>[[117, 319, 550, 363]]<|/det|>
+Authors: Does not occur on a reasonable time scale; Reviewer: This phrase is grammatically incorrect.  
+
+<|ref|>text<|/ref|><|det|>[[117, 396, 760, 441]]<|/det|>
+Authors: Their need for highly evolved enzymes also offers limited insight into phosphorylation before life existed  
+
+<|ref|>text<|/ref|><|det|>[[115, 448, 874, 755]]<|/det|>
+Reviewer: This statement is suggestive and lacks nuance. It suggests an implicit historical claim that phosphorylation is a process that predates life. This is a popular - though unproven - conjecture in a niche of origins of life, where it is more broadly believed that modern biomolecules coincide with those present at abiogenesis. Since much of the prospective readership is not directly from the origins community and may not be aware of niche conjectures, any reference to them needs to be contextualized. If the authors choose to do so, they may also want to cite appropriate literature that lays out these conjectures and viewpoints thereon. A recent work on conjectured roles of phosphate groups in prebiotic chemistry is for instance: Ring Opening of Glycerol Cyclic Phosphates Leads to a Diverse Array of Potentially Prebiotic Phospholipids, Maiia Aleksandrova, Fidan Rahmatova, David A. Russell, and Claudia Bonfio. J. Am. Chem. Soc. 2023, 145, 47, 25614- 25620, https://doi.org/10.1021/jacs.3c07319  
+
+<|ref|>text<|/ref|><|det|>[[117, 761, 855, 834]]<|/det|>
+It should be stressed that biochemistry- first as a conjecture has been questioned many times over, even by its proponents within the origins literature, a relatively recent source being:  
+
+<|ref|>text<|/ref|><|det|>[[117, 841, 852, 886]]<|/det|>
+Life's Biological Chemistry: A Destiny or Destination Starting from Prebiotic Chemistry? R. Krishnamurthy, Chem. Eur. J. 2018, 24, 16708.  
+
+<|ref|>text<|/ref|><|det|>[[115, 893, 861, 912]]<|/det|>
+A more nuanced phrasing would be '.. into the roles phosphorylation may potentially have
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 338, 101]]<|/det|>
+played before life existed'.  
+
+<|ref|>text<|/ref|><|det|>[[117, 135, 876, 545]]<|/det|>
+Authors: ...prebiotically plausible phosphorylating agentsReviewer: Prebiotic plausibility constitutes a scientific claim that needs to be motivated and whose premises need to be stipulated. In the present version of the article we find no demonstration of this claim, nor references to the literature where such a claim is voiced. Although the rigor behind this terminology has undergone considerable erosion in the last 50 years, members of the community have started to voice concerns (see: Benner SA. Prebiotic plausibility and networks of paradox- resolving independent models. Nat. Commun. 2018 Dec 12;9(1):5173. doi: 10.1038/s41467- 018- 07274- y). One should not freely call a compound 'prebiotically plausible', it needs to be motivated.In the case of DAP and MAP, there is indeed a recent body of literature that explores and argues for these species to be prebiotically plausible phosphorylating agents, for instance: Diamidophosphate (DAP) – A Plausible Prebiotic Phosphorylating Reagent with a Chem to BioChem Potential? Osumah, A.; Krishnamurthy, R. ChemBioChem, 2021, 22, 3001- 3009. Geochemical Sources and Availability of Amidophosphates on the Early Earth. Gibard, C.; Jiménez, E. I.; Gorrell, I. B.; Kee, T. P.; Pasek, M. A.; Krishnamurthy, R. Angew. Chemie Int. Ed. 2019, 58, 8151- 8155.  
+
+<|ref|>text<|/ref|><|det|>[[118, 579, 852, 675]]<|/det|>
+Authors: The reaction cycle is enzyme- free and produces a phosphorylated amino acid or peptide that hydrolyzes without requiring enzymes with a tunable half- life.Reviewer: This phrase can be improved, for instance by adding 'and' in front of 'with a tunable half- life'.  
+
+<|ref|>text<|/ref|><|det|>[[118, 711, 500, 806]]<|/det|>
+Results and discussionExploration of the chemical reaction network.Reviewer: Some minor mistakes.Prebiotically -> prebiotically  
+
+<|ref|>text<|/ref|><|det|>[[118, 814, 864, 911]]<|/det|>
+The kinetic model is unusually predictive for a system of this level of complexity, especially for the species 3- pH. Nevertheless, there is unexplained variation, i.e. qualitative and quantitative deviations for the other species. It would be instructive if the authors could include a short speculation on the possible sources of this variation.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 448, 102]]<|/det|>
+The labels C, D, F are missing from Fig.3.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 138, 220, 153]]<|/det|>
+## Conclusions  
+
+<|ref|>text<|/ref|><|det|>[[118, 164, 844, 206]]<|/det|>
+Reviewer: The conclusion is rather short, mentions few results and is not very specific. I recommend expanding the conclusion to be more reflective of the content of the paper.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 242, 213, 257]]<|/det|>
+## References  
+
+<|ref|>text<|/ref|><|det|>[[118, 267, 615, 283]]<|/det|>
+Reviewer: There is an error in the author list for reference 7.  
+
+<|ref|>text<|/ref|><|det|>[[118, 320, 206, 335]]<|/det|>
+Sup. Matt.  
+
+<|ref|>text<|/ref|><|det|>[[117, 345, 852, 441]]<|/det|>
+Reviewer: Among the most valuable products of this research is the high- quality kinetic data, involving multiple species, many timepoints, and repetitions. If this was not already being done, I encourage the authors to make it as easy as possible for their colleagues to access the raw data, e.g. by storing it in online repositories.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 502, 437, 519]]<|/det|>
+## Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[117, 527, 848, 624]]<|/det|>
+This manuscript describes the use of histidine phosphorylation by amidophosphates as a new reaction network for nonequilibrium systems chemistry. The authors establish the fundamental chemistry by optimizing the conditions and demonstrate its applicability through the formation of transient coacervate droplets.  
+
+<|ref|>text<|/ref|><|det|>[[117, 658, 863, 754]]<|/det|>
+The design of this system is simple and very elegant, and in my view is likely to be of broad interest to the field, especially considering some relevance to prebiotic chemistry. The authors have been thorough in exploring the parameter space for this reaction. I support publication in Nature Communications subject to a few comments.  
+
+<|ref|>text<|/ref|><|det|>[[117, 788, 866, 911]]<|/det|>
+1. The authors make use of a large, complex kinetic model. My major concern is the possibility that some of the parameters determined by fitting may be correlated (i.e., the model cannot measure them accurately because simultaneous changes in two or more cancel out in the concentration vs time simulations). If this is the case, the reported parameters (Table S1) may not be all that meaningful. There are a few things I noticed that
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 277, 101]]<|/det|>
+seem like red flags:  
+
+<|ref|>text<|/ref|><|det|>[[117, 136, 844, 258]]<|/det|>
+- In some cases (e.g., k3 for Ac-GHG-OH vs acY'DHDDAM) there seem to be very large discrepancies in the k values for what would seem to be similar reactions. 
+- In other cases, the errors determined (I assume) from the regression appear to be very large relative to the k's themselves (e.g., k-5, k7). 
+- In one case k-4 << k-1, which seems chemically unreasonable.  
+
+<|ref|>text<|/ref|><|det|>[[117, 291, 877, 415]]<|/det|>
+This issue should be discussed in the SI. If it cannot be adequately addressed, I would suggest removing explicit discussion of any parameters that are not determined by the data. It may only be possible to state that the mechanism is consistent with the data but that specific parameters cannot be accurately determined. To my mind this would not substantially affect interest in this work.  
+
+<|ref|>text<|/ref|><|det|>[[117, 448, 860, 519]]<|/det|>
+2. On a related note, how were the errors determined on the parameters in Tables S1 and S2? Is this through multiple replicates, or are they errors from the fits? This should be specified.  
+
+<|ref|>text<|/ref|><|det|>[[116, 553, 870, 806]]<|/det|>
+3. Some authors, notably Astumian, Aprahamian, and Goldup, have objected to how some terminology is used in this field and how some basic concepts are frequently described. I don't personally take as extreme a view, but I do think there is merit to these arguments. The manuscript describes unstable fuel molecules decomposing to waste, which I believe is one of the explanations that is often objected to. Really the issue is not specific unstable molecules, but rather that the reaction network must be driven by a reaction where the reactant and product concentrations are out of equilibrium; the distinction between fuel and waste is essentially arbitrary. That is, ADP can be a fuel for a reaction network if its concentration is above its equilibrium concentration. I would suggest making the language used in the introduction of this manuscript a bit more rigorous.  
+
+<|ref|>text<|/ref|><|det|>[[117, 840, 875, 910]]<|/det|>
+4. I'm a bit confused by how scientific notation is being used throughout the SI. In Table S1, there are frequent examples of numbers and uncertainties with both parts expressed in scientific notation (e.g., 1.40e-3 ± 7.58e-5). In Table S2, however, it seems like the exponent
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 876, 182]]<|/det|>
+of the uncertainty is supposed to apply to both parts (e.g., \(2.34 \pm 0.11e - 1 = 2.34e - 1 \pm 0.11e - 1?\) ). I'm therefore not sure what convention some numbers are following (e.g., \(2.17 \pm 0.50e - 3\) in Table S1). This should be clarified. I would personally suggest making it very explicit throughout, such as \("(2.34 \pm 0.11) \times 10^{- 1}"\) .  
+
+<|ref|>text<|/ref|><|det|>[[117, 214, 840, 285]]<|/det|>
+5. There are some incorrect table numberings in the SI. For example, there are frequent references to Table S7 that I think are supposed to be to Table S1. These should all be checked.  
+
+<|ref|>text<|/ref|><|det|>[[117, 318, 878, 364]]<|/det|>
+6. The notation is a bit confusing. For example, "3-pHis" (text) and "3-pH" (figures) appear to be used interchangeably.  
+
+<|ref|>text<|/ref|><|det|>[[117, 396, 870, 521]]<|/det|>
+7. I may be misunderstanding, but I believe the discussion on lines 113–115 is saying that the half life for loss of phosphate from 1,3-bpHis is the same or slightly longer than that for 1-pHis. This seems unlikely? If there are correlated parameters as noted above, I am concerned about quantitative assessments like this (depending on what's correlated with what, of course).  
+
+<|ref|>text<|/ref|><|det|>[[117, 553, 574, 572]]<|/det|>
+8. A bunch of the labels (C, D, F) are missing in Figure 3.  
+
+<|ref|>text<|/ref|><|det|>[[117, 605, 876, 678]]<|/det|>
+9. I noticed that the authors have tended to use near-stoichiometric or sub-stoichiometric ratios of MAP to peptide. Is there a reason for this? Especially considering that the system is not tremendously efficient (Figure 2F/G), it would seem like excess fuel might be useful.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 543, 99]]<|/det|>
+Response to the reviewers for: NCOMMS- 23- 62971  
+
+<|ref|>text<|/ref|><|det|>[[118, 130, 430, 146]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[115, 175, 876, 430]]<|/det|>
+Key results: The authors describe a strategy to transiently phosphorylate histidine, with the goal of mimicking the transient activation of enzymes in cells. Monoamidophosphate (MAP) is the main phosphorylating agent, and water is the dephosphorylating agent. Histidine catalyses the hydrolysis of MAP, getting phosphorylated/de- phosphorylated in the process. When histidine is embedded in a tripeptide, de- phosphorylation proves to be too slow for the reaction cycle to be considered dynamic, and the authors deploy pyridine to accelerate it. In the presence of MAP and pyridine, the half- life of the phosphorylated peptide is reduced to a few hours. Experimental data and kinetic modeling are used to depict the flow of phosphate in the system. The authors show that when the reaction is carried out in the presence of a polycation, supramolecular assemblies (coacervates) form and disappear following the dynamic phosphorylation of histidine.  
+
+<|ref|>text<|/ref|><|det|>[[117, 435, 858, 477]]<|/det|>
+Authors: We thank the reviewer for the time invested in the manuscript. We addressed the comments below.  
+
+<|ref|>text<|/ref|><|det|>[[113, 504, 872, 902]]<|/det|>
+Validity & Significance: The experimental data in the manuscript is complete and the kinetic model provides a good fit. However, I believe the concept and strategy described in this manuscript are not significantly novel, and would not have a sufficient impact in the community to grant publication in Nature Communications. Transient compartmentalization controlled by phosphorylation has been shown previously (DOI: 10.1038/nchem.2414, 10.1039/c7sm01897e). It is true that the manuscript provides a non- enzymatic version, but still the applicability is limited by the fact that the reaction only works with the pair MAP/histidine (tables in Figure 1). We have now added the reference of Keating and Huck's work. Moreover we also cited Spruijt's work describing ATP based active droplets. Recent literature on prebiotic phosphorylation tends to look for agents with a broader spectrum (e.g. DOI 10.1021/jacs.3c08539, 10.1038/s41467- 021- 25555- x); or in the case of histidine itself, for functional reactions with greater resemblance to natural catalytic cycles (see alternative with imidazole phosphate in Maguire ChemRXiv 2023). For the reaction cycle to operate like in biology, both activation and deactivation occur on similar timescales without side reactions. In this work, we focused on a quantitative understanding of a chemical reaction cycler rather than its breadth. While we are convinced there are other substrate and fuel combinations, besides the ones that we tried, that would also work, we
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 83, 855, 124]]<|/det|>
+decided to focus on the combination of histidine and its peptide derivatives with MAP as a fuel.  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 875, 243]]<|/det|>
+MAP, as the authors point out themselves, is not the most prebiotically plausible phosphorylating agent; and the use of pyridine as a substrate (that is mostly phosphorylated directly and competes with histidine) is somewhat confusing and undermines, in my opinion, its potential impact in the prebiotic chemistry field.  
+
+<|ref|>text<|/ref|><|det|>[[115, 247, 875, 430]]<|/det|>
+We appreciate the input, and we toned down the claim regarding prebiotic relevance. Moreover, in line with reviewer two, we cited recent work on the prebiotic relevance of MAP. The results on the formation of compartments are also weakened by the use of PEG 8k (at \(2.5\%\) concentration), a crowding agent known to promote and interfere with phase separation - more than a simple turbidity enhancer. Combined with the fact that most micrographs show only one droplet, or one microreactor, it indicates that the supramolecular results may be not robust or reproducible, and that the impact on the protocell/dynamic assemblies community would be very limited.  
+
+<|ref|>text<|/ref|><|det|>[[115, 435, 876, 570]]<|/det|>
+The way the data was presented indeed only showed one droplet per micrograph. We now added further evidence of the droplet formation. On the one hand, we show the formation of multiple droplets within one microreactor without the need of PEG. On the other hand, we show multiple microreactors containing droplets within one micrograph. Finally, the evolution of the droplet volume over time was performed in triplicate demonstrating that it is reproducible.  
+
+<|ref|>text<|/ref|><|det|>[[115, 599, 759, 617]]<|/det|>
+I include below some specific minor and major concerns about the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[117, 648, 278, 664]]<|/det|>
+Clarity and context:  
+
+<|ref|>text<|/ref|><|det|>[[115, 670, 864, 807]]<|/det|>
+1. The authors use terms such as "active droplets" in the title, "self-sustaining complex coacervate droplets", but do not justify it in the text or in the results. Active droplets would imply a non-equilibrium behavior, which is not clear to me in the data (the droplets emerge, coalesce, and dissolve). Self-sustaining would imply that the droplets facilitate the phosphorylation reaction that leads to their own formation, which was not probed in the manuscript.  
+
+<|ref|>text<|/ref|><|det|>[[115, 812, 870, 900]]<|/det|>
+We agree that self- sustaining would imply a more complex system that can, for example, facilitate its own formation. To clarify and not confuse our readers, we've changed from self- sustaining to active, with active meaning with activated precursor/product present (highlighted in yellow).
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[116, 83, 845, 125]]<|/det|>
+Our work demonstrates the emergence, the coalescence, and the dissolution of droplets, which can only happen out of equilibrium.  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 878, 220]]<|/det|>
+2. The description of the reaction as a cycle, or as a network (lines 64, 129), may be misleading, with the deactivation agent being the reaction's solvent (or pyridine), and the fuel MAP not being regenerated.  
+
+<|ref|>text<|/ref|><|det|>[[115, 224, 875, 336]]<|/det|>
+This was not meant to be misleading. We picture it as a cycle from the histidine's point of view. The histidine in our cycle undergoes cyclical phosphorylation drive by the hydrolysis of phosphorylating agent. Conceptually, this is similar to how proteins can undergo cyclical phosphorylation and dephosphorylation driven by the hydrolysis of ATP. We've now further clarified in the main text (highlighted in yellow).  
+
+<|ref|>text<|/ref|><|det|>[[115, 364, 878, 500]]<|/det|>
+3. I am missing the explanation as to how short of a half-life is desirable in a biological or chemical context, and why the choice to rely on water or pyridine as de-activating agents, as opposed of some prebiotically relevant substrates. The authors could include a more specific discussion on histidyl kinases, instead of the general introduction on ATP. We added a part on the histidine kinases to the introduction (highlighted in yellow). We were interested in what influences the half-lives.  
+
+<|ref|>text<|/ref|><|det|>[[115, 506, 875, 571]]<|/det|>
+4. Figures: scheme 1 would be more appropriate as a graphical abstract, it is taking up quite a lot of space but not adding as much information. Figure 1 could be reorganized to take up less space (the table format may not be necessary).  
+
+<|ref|>text<|/ref|><|det|>[[115, 576, 876, 711]]<|/det|>
+We agree scheme 1 is more like a graphical abstract, and we reduced the size and compressed it. Nevertheless, it includes all the information needed (histidine-based precursor as 5membered ring, phosphorylation coupled to droplet formation with information about the constitution of the droplets). Besides nature communications does not work with graphical abstract, which is why we use Scheme 1 to graphically demonstrate the idea of the work.  
+
+<|ref|>text<|/ref|><|det|>[[115, 718, 880, 758]]<|/det|>
+In our opinion, the table format in Fig. 1 is in this case useful, nicely aligned and gives a good overview.  
+
+<|ref|>text<|/ref|><|det|>[[115, 765, 875, 907]]<|/det|>
+Figure 2 contains some redundant data, panels F and G could be in supplementary information. Instead, a table for half-life and conversion% would be helpful. In our opinion, the panels (bar graphs) are clearer than tables. In a table the conversion% is less obvious than the bar graph with colorfully highlighted %ratio, as well as the change of half-lives under different conditions. But if you want to directly compare the numbers, we added the half-lives to Table S1 and S3 in the SI.  
+
+<|ref|>text<|/ref|><|det|>[[117, 906, 425, 923]]<|/det|>
+In Figure 3, what does Y'Dam mean?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 367, 99]]<|/det|>
+We've changed it to peptide 1.  
+
+<|ref|>text<|/ref|><|det|>[[115, 106, 872, 147]]<|/det|>
+5. The clarity of the manuscript could be increased by changing the acronyms. Instead of 1- pH, His1-P for example, as pH already has its own meaning.  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 877, 194]]<|/det|>
+We appreciate the feedback. It is also in agreement with the literature, so we changed it to x- pHis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 200, 880, 290]]<|/det|>
+6. Line 27: The sentence starting with "Does not occur..." is incomplete. The text from line 27 up to 39 needs to be restructured, it is confusing and not conveying its message – which I think is that biological cycles benefit from using ATP as a fuel due to it being thermodynamically unstable but kinetically stable.  
+
+<|ref|>text<|/ref|><|det|>[[117, 295, 465, 312]]<|/det|>
+We rephrased and clarified the sentences.  
+
+<|ref|>text<|/ref|><|det|>[[115, 319, 844, 360]]<|/det|>
+7. Line 187: there is a typo, in "Either it is hydrolyzed directly or through phosphorylating histidine (unwanted)" it should be pyridine, not histidine.  
+
+<|ref|>text<|/ref|><|det|>[[115, 365, 880, 430]]<|/det|>
+In line 192 there is a repetition "activation phosphorylation". This paragraph could be made clearer and the quantitative data mentioned could be included in a table elsewhere for clarity. Thank you, we rephrased the sentence.  
+
+<|ref|>text<|/ref|><|det|>[[117, 436, 311, 453]]<|/det|>
+Data and methodology:  
+
+<|ref|>text<|/ref|><|det|>[[115, 459, 875, 525]]<|/det|>
+8. Both experimental and modeling kinetic data are well presented, error bars are shown but it is not mentioned if the mean is taken over multiple experiments or different measurements or time points. I appreciate how the methodology was extensively described.  
+
+<|ref|>text<|/ref|><|det|>[[115, 530, 850, 618]]<|/det|>
+We appreciate the comment. In line with nature communication policy, we updated the presentation of the experimental data. We now show each data point of each experiment. The only data with mean error bars is the droplet volume/microreactor volume (droplets analyzed per micrograph \(>10\) ), but the triplicates are plotted individually.  
+
+<|ref|>text<|/ref|><|det|>[[115, 623, 880, 688]]<|/det|>
+9. Overall, some experiments could be better motivated in the text: for example, it is not clear to me that such a complete kinetic modeling is necessary, or how it can contribute to the understanding of the system.  
+
+<|ref|>text<|/ref|><|det|>[[115, 694, 857, 758]]<|/det|>
+We performed the kinetic modeling to predict the concentration evolution of the different species as precisely as possible. We also want to help others start a project based on this system. We have clarified in the main text.  
+
+<|ref|>text<|/ref|><|det|>[[115, 765, 868, 829]]<|/det|>
+Another example is why the authors choose to conduct microscopy in microfluidically- produced micro- reactors, and not on a typical microscopy slide/chamber - what advantages to the paper's goals does it bring?  
+
+<|ref|>text<|/ref|><|det|>[[117, 836, 429, 852]]<|/det|>
+We address this comment in point 10.  
+
+<|ref|>text<|/ref|><|det|>[[115, 859, 870, 900]]<|/det|>
+The ITC and DNA partitioning data is included but not well integrated within the text - how is this data relevant to the manuscript?
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 390, 99]]<|/det|>
+We integrated it better in the text.  
+
+<|ref|>text<|/ref|><|det|>[[113, 105, 880, 430]]<|/det|>
+10. The microscopy images do not immediately point to coacervate droplets (if you compare to other peptide/peptide complex coacervates in related literature), although I understand that FRAP supports it. Perhaps resolution and visibility would be increased without the microreactors. And instead of adding PEG, I would consider reducing buffer concentration. Thank you for your comment. We have now added data that shows that the droplets fuse. Besides, we have clarified that we have used the dye sulforhodamine B, a dye that is water soluble demonstrating an aqueous interior. We used microreactors to analyze the droplets because the droplets do not adhere to the microreactor walls. In contrast, they quickly adhere to glass slides. We clarified this in the main text (highlighted in yellow). We use PEG instead of reducing the buffer concentration to keep the ionic strength the same as in the microreactors. 11. Scale bars are often missing. Thank you for your comment we added them, where needed.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 430, 100]]<|/det|>
+Reviewer #2 (Remarks to the Author):  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 130, 240, 146]]<|/det|>
+## Referee report  
+
+<|ref|>text<|/ref|><|det|>[[117, 153, 633, 171]]<|/det|>
+Active droplets through enzyme- free, dynamic phosphorylation  
+
+<|ref|>text<|/ref|><|det|>[[117, 177, 840, 217]]<|/det|>
+Authors: Simone Poprawa,1 Michele Stasi,1 Monika Wenisch,1 Brigitte A. K. Kriebisch,1 Judit Sastre,1 Job Boekhoven1  
+
+<|ref|>text<|/ref|><|det|>[[115, 223, 884, 405]]<|/det|>
+Reviewer: The article presents an elegant system for the construction of active droplets using dynamic phosphorylation in an enzyme- free manner, and hence furnishes a platform that others can readily built upon. This work will likely be picked up rapidly by several communities (for instance systems chemistry, active matter, and origins of life) and can be expected to become a highly cited paper. As such, I am overall favorable to its publication in Nature Communications. Nevertheless, in terms of writing, presentation and contextualization, the present version feels unfinished and the scholarly quality of the work can be favorably improved.  
+
+<|ref|>text<|/ref|><|det|>[[117, 412, 841, 452]]<|/det|>
+Thank you for your time invested in the manuscript. We have addressed your comments below.  
+
+<|ref|>text<|/ref|><|det|>[[117, 459, 303, 476]]<|/det|>
+In chronological order:  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 506, 189, 522]]<|/det|>
+## Abstract  
+
+<|ref|>text<|/ref|><|det|>[[115, 529, 872, 616]]<|/det|>
+Reviewer: The abstract provides hardly more information than the title about the content of the paper and remains vague in instances where being precise would hardly consume more space, for instance it may be pointed out that the system is specific to histidine, and what phosphorylating agent is used.  
+
+<|ref|>text<|/ref|><|det|>[[117, 623, 360, 639]]<|/det|>
+We rephrased the sentences.  
+
+<|ref|>text<|/ref|><|det|>[[117, 647, 830, 687]]<|/det|>
+Authors: Enzyme- free, synthetic systems that use dynamic phosphorylation to regulate supramolecular processes do not exist  
+
+<|ref|>text<|/ref|><|det|>[[117, 694, 842, 735]]<|/det|>
+Reviewer: Do not exist is a confusing phrasing. The authors presumably mean that such systems 'have not yet been reported, to the best of our knowledge'.  
+
+<|ref|>text<|/ref|><|det|>[[117, 741, 454, 757]]<|/det|>
+Thank you, we rephrased the sentences.  
+
+<|ref|>text<|/ref|><|det|>[[117, 764, 856, 805]]<|/det|>
+Authors: Our new reaction cycle will be valuable as a model for biological phosphorylation but can also offer insights into protocell formation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 810, 870, 924]]<|/det|>
+Reviewer: Given the long history of MAP and DAP chemistry and renewed recent interest (which should be cited), the possessive phrasing 'our new reaction cycle' feels unnecessary here. The article does not clarify in what sense(s) the reaction cycle may be expected to be deliver on these promises. Since the phosphorylation is specific to histidine, it is not a priori obvious that it provides a model for biological phosphorylation.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 805, 124]]<|/det|>
+Indeed, "new" is too possessive. We added a part of histidine phosphorylation in the introduction (highlighted in yellow).  
+
+<|ref|>text<|/ref|><|det|>[[117, 130, 217, 146]]<|/det|>
+Introduction  
+
+<|ref|>text<|/ref|><|det|>[[115, 153, 870, 243]]<|/det|>
+Reviewer: The introduction explains at length certain aspects that are widely known, while assuming intimate familiarity of the reader with communities and highly specific branches of literature that are not cited. In particular, pertinent recent literature involving phosphates in systems chemistry and MAP/DAP in prebiotic chemistry appears to be overlooked.  
+
+<|ref|>text<|/ref|><|det|>[[117, 248, 408, 265]]<|/det|>
+With respect to the former, notably:  
+
+<|ref|>text<|/ref|><|det|>[[115, 271, 861, 384]]<|/det|>
+Spontaneous and Selective Peptide Elongation in Water Driven by Aminoacyl Phosphate Esters and Phase Changes, Kun Dai, Mahesh D. Pol, Lenard Saile, Arti Sharma, Bin Liu, Ralf Thomann, Johanna L. Trefs, Danye Qiu, Sandra Moser, Stefan Wiesler, Bizan N. Balzer, Thorsten Hugel, Henning J. Jessen, and Charalampos G. Pappas\*, J. Am. Chem. Soc. 2023, 145, 48, 26086- 26094, https://doi.org/10.1021/jacs.3c07918  
+
+<|ref|>text<|/ref|><|det|>[[115, 389, 868, 455]]<|/det|>
+A Ribonucleotide \(\leq - >\) Phosphoramidate Reaction Network Optimized by Computer- Aided Design. Andreas Englert, Julian F. Vogel, Tim Bergner, Jessica Loske, and Max von Delius. J. Am. Chem. Soc. 2022, 144, 33, 15266- 15274  
+
+<|ref|>text<|/ref|><|det|>[[117, 461, 330, 477]]<|/det|>
+We added the references.  
+
+<|ref|>text<|/ref|><|det|>[[115, 484, 550, 525]]<|/det|>
+Authors: Does not occur on a reasonable time scale; Reviewer: This phrase is grammatically incorrect.  
+
+<|ref|>text<|/ref|><|det|>[[117, 532, 350, 547]]<|/det|>
+We rephrased the sentence.  
+
+<|ref|>text<|/ref|><|det|>[[115, 554, 755, 595]]<|/det|>
+Authors: Their need for highly evolved enzymes also offers limited insight into phosphorylation before life existed  
+
+<|ref|>text<|/ref|><|det|>[[115, 600, 878, 870]]<|/det|>
+Reviewer: This statement is suggestive and lacks nuance. It suggests an implicit historical claim that phosphorylation is a process that predates life. This is a popular - though unproven - conjecture in a niche of origins of life, where it is more broadly believed that modern biomolecules coincide with those present at abiogenesis. Since much of the prospective readership is not directly from the origins community and may not be aware of niche conjectures, any reference to them needs to be contextualized. If the authors choose to do so, they may also want to cite appropriate literature that lays out these conjectures and viewpoints thereon. A recent work on conjectured roles of phosphate groups in prebiotic chemistry is for instance: Ring Opening of Glycerol Cyclic Phosphates Leads to a Diverse Array of Potentially Prebiotic Phospholipids, Maia Aleksandrova, Fidan Rahmatova, David A. Russell, and Claudia Bonfio. J. Am. Chem. Soc. 2023, 145, 47, 25614- 25620, https://doi.org/10.1021/jacs.3c07319  
+
+<|ref|>text<|/ref|><|det|>[[115, 884, 878, 925]]<|/det|>
+It should be stressed that biochemistry- first as a conjecture has been questioned many times over, even by its proponents within the origins literature, a relatively recent source being:
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 82, 855, 123]]<|/det|>
+Life's Biological Chemistry: A Destiny or Destination Starting from Prebiotic Chemistry? R. Krishnamurthy, Chem. Eur. J. 2018, 24, 16708.  
+
+<|ref|>text<|/ref|><|det|>[[115, 129, 852, 170]]<|/det|>
+A more nuanced phrasing would be '.. into the roles phosphorylation may potentially have played before life existed'.  
+
+<|ref|>text<|/ref|><|det|>[[115, 177, 850, 218]]<|/det|>
+We totally agree and appreciate the thoughtful feedback. We nuanced the statement and added supporting citations.  
+
+<|ref|>text<|/ref|><|det|>[[115, 247, 585, 265]]<|/det|>
+Authors: ...prebiotically plausible phosphorylating agents  
+
+<|ref|>text<|/ref|><|det|>[[114, 271, 880, 453]]<|/det|>
+Reviewer: Prebiotic plausibility constitutes a scientific claim that needs to be motivated and whose premises need to be stipulated. In the present version of the article we find no demonstration of this claim, nor references to the literature where such a claim is voiced. Although the rigor behind this terminology has undergone considerable erosion in the last 50 years, members of the community have started to voice concerns (see: Benner SA. Prebiotic plausibility and networks of paradox- resolving independent models. Nat. Commun. 2018 Dec 12;9(1):5173. doi: 10.1038/s41467- 018- 07274- y). One should not freely call a compound 'prebiotically plausible', it needs to be motivated.  
+
+<|ref|>text<|/ref|><|det|>[[114, 459, 872, 618]]<|/det|>
+In the case of DAP and MAP, there is indeed a recent body of literature that explores and argues for these species to be prebiotically plausible phosphorylating agents, for instance: Diamidophosphate (DAP) - A Plausible Prebiotic Phosphorylating Reagent with a Chem to BioChem Potential? Osumah, A.; Krishnamurthy, R. ChemBioChem, 2021, 22, 3001- 3009. Geochemical Sources and Availability of Amidophosphates on the Early Earth. Gibard, C.; Jiménez, E. I.; Gorrell, I. B.; Kee, T. P.; Pasek, M. A.; Krishnamurthy, R. Angew. Chemie Int. Ed. 2019, 58, 8151- 8155.  
+
+<|ref|>text<|/ref|><|det|>[[115, 624, 867, 665]]<|/det|>
+Thank you for reminding us, of course these are very important publications and need to be cited.  
+
+<|ref|>text<|/ref|><|det|>[[115, 694, 853, 784]]<|/det|>
+Authors: The reaction cycle is enzyme- free and produces a phosphorylated amino acid or peptide that hydrolyzes without requiring enzymes with a tunable half- life. Reviewer: This phrase can be improved, for instance by adding 'and' in front of 'with a tunable half- life'.  
+
+<|ref|>text<|/ref|><|det|>[[117, 790, 339, 806]]<|/det|>
+We changed the sentence.  
+
+<|ref|>text<|/ref|><|det|>[[115, 836, 488, 925]]<|/det|>
+Results and discussionExploration of the chemical reaction network.Reviewer: Some minor mistakes.Prebiotically -> prebiotically
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 304, 99]]<|/det|>
+We corrected the typo.  
+
+<|ref|>text<|/ref|><|det|>[[115, 105, 864, 196]]<|/det|>
+The kinetic model is unusually predictive for a system of this level of complexity, especially for the species 3- pH. Nevertheless, there is unexplained variation, i.e. qualitative and quantitative deviations for the other species. It would be instructive if the authors could include a short speculation on the possible sources of this variation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 201, 878, 336]]<|/det|>
+The reaction rate constant of the 3- pHs was determined empirically when no other species or fuel were present anymore. Therefore, the predicted and the experimentally measured evolution fit nicely. The hydrolysis of MAP was also calculated empirically, but the other constants (up to 17) were fitted and thus show higher deviations (Table S1, error of the fit) as there are many combinations possible, also with correlated parameters. Despite that, the fits align very well with the experimental data.  
+
+<|ref|>text<|/ref|><|det|>[[115, 341, 872, 455]]<|/det|>
+As you and reviewer 3 mentioned, the kinetic model is very good for predicting the evolution of a single species and can potentially help to design a self- assembling system. However, the reaction rate constants, which are not determined by the data, should not be compared or addressed directly. We have removed any references to half- lives or rate constants that were determined by the kinetic model.  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 869, 501]]<|/det|>
+Besides, we have added the comment on the value and the limitation of the kinetic model in the SI.  
+
+<|ref|>text<|/ref|><|det|>[[117, 530, 460, 547]]<|/det|>
+The labels C, D, F are missing from Fig.3.  
+
+<|ref|>text<|/ref|><|det|>[[117, 553, 674, 570]]<|/det|>
+We changed the presentation of our data and also added the labels.  
+
+<|ref|>text<|/ref|><|det|>[[117, 577, 222, 593]]<|/det|>
+Conclusions  
+
+<|ref|>text<|/ref|><|det|>[[115, 600, 835, 666]]<|/det|>
+Reviewer: The conclusion is rather short, mentions few results and is not very specific. I recommend expanding the conclusion to be more reflective of the content of the paper. We have addressed your comments on the conclusion.  
+
+<|ref|>text<|/ref|><|det|>[[117, 672, 215, 688]]<|/det|>
+References  
+
+<|ref|>text<|/ref|><|det|>[[115, 695, 610, 736]]<|/det|>
+Reviewer: There is an error in the author list for reference 7. We updated the references.  
+
+<|ref|>text<|/ref|><|det|>[[117, 765, 203, 782]]<|/det|>
+Sup. Matt.  
+
+<|ref|>text<|/ref|><|det|>[[115, 788, 875, 877]]<|/det|>
+Reviewer: Among the most valuable products of this research is the high- quality kinetic data, involving multiple species, many timepoints, and repetitions. If this was not already being done, I encourage the authors to make it as easy as possible for their colleagues to access the raw data, e.g. by storing it in online repositories.  
+
+<|ref|>text<|/ref|><|det|>[[117, 883, 850, 924]]<|/det|>
+Thank you, we appreciate that! We added the Excel sheet with the raw data of the kinetic experiments and the COPASI sample file.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 430, 100]]<|/det|>
+Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 128, 875, 220]]<|/det|>
+This manuscript describes the use of histidine phosphorylation by amidophosphates as a new reaction network for nonequilibrium systems chemistry. The authors establish the fundamental chemistry by optimizing the conditions and demonstrate its applicability through the formation of transient coacervate droplets.  
+
+<|ref|>text<|/ref|><|det|>[[116, 246, 866, 337]]<|/det|>
+The design of this system is simple and very elegant, and in my view is likely to be of broad interest to the field, especially considering some relevance to prebiotic chemistry. The authors have been thorough in exploring the parameter space for this reaction. I support publication in Nature Communications subject to a few comments.  
+
+<|ref|>text<|/ref|><|det|>[[116, 342, 861, 501]]<|/det|>
+Thank you for your positive feedback, we appreciate your time invested in the manuscript. 1. The authors make use of a large, complex kinetic model. My major concern is the possibility that some of the parameters determined by fitting may be correlated (i.e., the model cannot measure them accurately because simultaneous changes in two or more cancel out in the concentration vs time simulations). If this is the case, the reported parameters (Table S1) may not be all that meaningful. There are a few things I noticed that seem like red flags:  
+
+<|ref|>text<|/ref|><|det|>[[116, 529, 841, 642]]<|/det|>
+- In some cases (e.g., k3 for Ac-GHG-OH vs acY'DHDDam) there seem to be very large discrepancies in the k values for what would seem to be similar reactions. 
+- In other cases, the errors determined (I assume) from the regression appear to be very large relative to the k's themselves (e.g., k-5, k7). 
+- In one case k-4 << k-1, which seems chemically unreasonable.  
+
+<|ref|>text<|/ref|><|det|>[[116, 670, 872, 784]]<|/det|>
+This issue should be discussed in the SI. If it cannot be adequately addressed, I would suggest removing explicit discussion of any parameters that are not determined by the data. It may only be possible to state that the mechanism is consistent with the data but that specific parameters cannot be accurately determined. To my mind this would not substantially affect interest in this work.  
+
+<|ref|>text<|/ref|><|det|>[[116, 789, 878, 854]]<|/det|>
+Thank you for your detailed comment. We have distanced ourselves to compare the fitted values directly, but use them as a guideline and compare empirically determined parameters. Please see the answer to reviewer 2.  
+
+<|ref|>text<|/ref|><|det|>[[116, 882, 856, 924]]<|/det|>
+2. On a related note, how were the errors determined on the parameters in Tables S1 and S2? Is this through multiple replicates, or are they errors from the fits? This should be
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[117, 83, 197, 99]]<|/det|>
+specified.  
+
+<|ref|>text<|/ref|><|det|>[[117, 106, 590, 124]]<|/det|>
+We added the explanation to the description of the tables.  
+
+<|ref|>text<|/ref|><|det|>[[115, 129, 878, 360]]<|/det|>
+3. Some authors, notably Astumian, Aprahamian, and Goldup, have objected to how some terminology is used in this field and how some basic concepts are frequently described. I don't personally take as extreme a view, but I do think there is merit to these arguments. The manuscript describes unstable fuel molecules decomposing to waste, which I believe is one of the explanations that is often objected to. Really the issue is not specific unstable molecules, but rather that the reaction network must be driven by a reaction where the reactant and product concentrations are out of equilibrium; the distinction between fuel and waste is essentially arbitrary. That is, ADP can be a fuel for a reaction network if its concentration is above its equilibrium concentration. I would suggest making the language used in the introduction of this manuscript a bit more rigorous.  
+
+<|ref|>text<|/ref|><|det|>[[117, 365, 567, 382]]<|/det|>
+We have adjusted the introduction to define fuel better.  
+
+<|ref|>text<|/ref|><|det|>[[115, 388, 881, 548]]<|/det|>
+4. I'm a bit confused by how scientific notation is being used throughout the SI. In Table S1, there are frequent examples of numbers and uncertainties with both parts expressed in scientific notation (e.g., \(1.40\mathrm{e - 3} \pm 7.58\mathrm{e - 5}\) ). In Table S2, however, it seems like the exponent of the uncertainty is supposed to apply to both parts (e.g., \(2.34 \pm 0.11\mathrm{e - 1} = 2.34\mathrm{e - 1} \pm 0.11\mathrm{e - 1}\) ?). I'm therefore not sure what convention some numbers are following (e.g., \(2.17 \pm 0.50\mathrm{e - 3}\) in Table S1). This should be clarified. I would personally suggest making it very explicit throughout, such as \((2.34 \pm 0.11) \times 10^{-1}\) .  
+
+<|ref|>text<|/ref|><|det|>[[118, 554, 335, 570]]<|/det|>
+Thank you, we changed it.  
+
+<|ref|>text<|/ref|><|det|>[[115, 577, 840, 641]]<|/det|>
+5. There are some incorrect table numberings in the SI. For example, there are frequent references to Table S7 that I think are supposed to be to Table S1. These should all be checked.  
+
+<|ref|>text<|/ref|><|det|>[[117, 648, 347, 664]]<|/det|>
+We updated the numbering.  
+
+<|ref|>text<|/ref|><|det|>[[115, 670, 875, 712]]<|/det|>
+6. The notation is a bit confusing. For example, "3-pHis" (text) and "3-pH" (figures) appear to be used interchangeably.  
+
+<|ref|>text<|/ref|><|det|>[[117, 718, 607, 735]]<|/det|>
+To be consistent, also with literature, we changed to x-pHis.  
+
+<|ref|>text<|/ref|><|det|>[[115, 740, 878, 855]]<|/det|>
+7. I may be misunderstanding, but I believe the discussion on lines 113-115 is saying that the half life for loss of phosphate from 1,3-bpHis is the same or slightly longer than that for 1-pHis. This seems unlikely? If there are correlated parameters as noted above, I am concerned about quantitative assessments like this (depending on what's correlated with what, of course).  
+
+<|ref|>text<|/ref|><|det|>[[117, 860, 410, 876]]<|/det|>
+See the answer to your comment 1.  
+
+<|ref|>text<|/ref|><|det|>[[115, 883, 644, 923]]<|/det|>
+8. A bunch of the labels (C, D, F) are missing in Figure 3. We changed presentation of our data and also added the labels.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[113, 82, 875, 266]]<|/det|>
+9. I noticed that the authors have tended to use near-stoichiometric or sub-stoichiometric ratios of MAP to peptide. Is there a reason for this? Especially considering that the system is not tremendously efficient (Figure 2F/G), it would seem like excess fuel might be useful. We agree that the yield would be higher with a higher amount of fuel. However, an excess of fuel also elongates the total experimental time further. To keep the experiment on a reasonable timescale, we usually work with stoichiometric amounts. Moreover, to potentially observe the bisphosphorylated species, we used nearly stoichiometric and high amount of peptide to fuel.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 85, 315, 101]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 137, 437, 154]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[117, 162, 872, 285]]<|/det|>
+The authors addressed some of my concerns, but my major concerns regarding the publication of this manuscript in Nature Communications remain. That said, at least the writing needs further improvement in order for me to be comfortable supporting publication. I really think the manuscript will benefit from a more rigorous writing, and that time put on these final changes will be well spent.  
+
+<|ref|>text<|/ref|><|det|>[[117, 319, 505, 336]]<|/det|>
+I refer to the numbering in my previous report:  
+
+<|ref|>text<|/ref|><|det|>[[116, 370, 880, 546]]<|/det|>
+1. Partially addressed. In line 229, I disagree from the definition of active droplets or active protocells (also comparing to David Zwicker's group definition): In this context, active means that the droplets are sustained as long as a certain amount of product is present. The way it is written at least, it is not a remarkable behaviour – that the droplets exist while their building blocks exist. Reaction-controlled coacervates are not an active per se; it helps to make your case if you compare these dynamic droplets, sustained by the reaction, to droplets formed from building blocks that don't form or decay in situ.  
+
+<|ref|>text<|/ref|><|det|>[[116, 580, 860, 728]]<|/det|>
+2. Despite the explanation in the reply to my comments, the introduction text still lacks specificity. As a reader, I am still not convinced that this is a cycle, or a controlled cycle: surely, while there is fuel, there is phosphorylation; and surely when you stop adding fuel, hydrolysis dominates, but that does not mean control. A key point of a cycle would be to demonstrate that His phosphorylation (and coacervates) can be recovered by a fresh addition of MAP. Can you show this?  
+
+<|ref|>text<|/ref|><|det|>[[115, 737, 876, 910]]<|/det|>
+a. Line 27 is an example of what I mean by lack of specificity in the writing: That way, the equilibrium position for the fuel-to-waste equilibrium lies to the right such that most fuel will eventually be converted into waste. (and btw line 33 can start a new paragraph) 
+b. In line 287, you write "reaction cycle that phosphorylates histidine as an amino acid or in a peptide at the expense of phosphorylating agents". There is no need to say "at the expense of phosphorylating agents", this is the case in any chemical reaction. This and other similar repetitions take space from a more meaningful discussion/conclusion.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 231, 101]]<|/det|>
+3. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[117, 135, 875, 234]]<|/det|>
+4. Since you chose to keep the table in Figure 1, I find it that there needs to be some discussion in the text as to why the cycle only works with histidine, or why, if we know from the literature that MAP phosphorylates imidazole groups, other residues were also tested. What was the hypothesis?  
+
+<|ref|>text<|/ref|><|det|>[[117, 266, 232, 361]]<|/det|>
+5. Addressed. 
+6. Addressed. 
+7. Addressed. 
+8. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[117, 396, 878, 494]]<|/det|>
+9. Partially addressed. The use of microreactors still confuses me, as there are many ways to passivate glass slides to prevent droplet adhesion. Can you at least include in SI micrographs of droplets outside these microreactors? This would show that the microreactor itself is not crucial to the droplets' properties.  
+
+<|ref|>text<|/ref|><|det|>[[117, 526, 875, 624]]<|/det|>
+10. In SI Figure 10, I see you now added the whole field of view, showing all microreactors. Can you include the emission channel as well, which is what you mostly show in the main text? The caption says it shows the emergence, growth and dissolution of the droplets, but I can only see emergence and dissolution.  
+
+<|ref|>text<|/ref|><|det|>[[118, 658, 241, 675]]<|/det|>
+11. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[118, 710, 255, 726]]<|/det|>
+New comments:  
+
+<|ref|>text<|/ref|><|det|>[[117, 735, 853, 833]]<|/det|>
+12. Line 161: consider reorganizing the long list of references. The comparison to the literature is barely made, yet there are 11 references listed. I prefer it if you elaborate on this discussion rather than remove references, as I think it will make the manuscript stronger.  
+
+<|ref|>text<|/ref|><|det|>[[118, 866, 880, 911]]<|/det|>
+13. The conclusions are still vague and imprecise despite another reviewer's comment. Most statements are based on this system being a dynamic cycle and the droplets being active, so
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[118, 85, 844, 128]]<|/det|>
+I think resolving these conceptual problems in the main text will help better support the conclusions.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 189, 438, 205]]<|/det|>
+## Reviewer #2 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 216, 438, 231]]<|/det|>
+All reviewer points are addressed well.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 294, 438, 310]]<|/det|>
+## Reviewer #3 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 320, 877, 363]]<|/det|>
+In my opinion, the authors have addressed the points I raised in my original review. I believe that the manuscript is now suitable for Nature Communications.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[117, 83, 350, 101]]<|/det|>
+## REVIEWER COMMENTS  
+
+<|ref|>text<|/ref|><|det|>[[117, 117, 457, 135]]<|/det|>
+Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[116, 151, 875, 238]]<|/det|>
+The authors addressed some of my concerns, but my major concerns regarding the publication of this manuscript in Nature Communications remain. That said, at least the writing needs further improvement in order for me to be comfortable supporting publication. I really think the manuscript will benefit from a more rigorous writing, and that time put on these final changes will be well spent.  
+
+<|ref|>text<|/ref|><|det|>[[116, 254, 528, 272]]<|/det|>
+I refer to the numbering in my previous report:  
+
+<|ref|>text<|/ref|><|det|>[[115, 288, 875, 426]]<|/det|>
+1. Partially addressed. In line 229, I disagree from the definition of active droplets or active protocells (also comparing to David Zwicker's group definition): In this context, active means that the droplets are sustained as long as a certain amount of product is present. The way it is written at least, it is not a remarkable behaviour – that the droplets exist while their building blocks exist. Reaction-controlled coacervates are not an active per se; it helps to make your case if you compare these dynamic droplets, sustained by the reaction, to droplets formed from building blocks that don't form or decay in situ.  
+
+<|ref|>text<|/ref|><|det|>[[117, 444, 378, 460]]<|/det|>
+We agree and changed it for:  
+
+<|ref|>text<|/ref|><|det|>[[115, 460, 808, 495]]<|/det|>
+In this context, active means that the droplet material is controlled by two chemical reactions: activation and deactivation.  
+
+<|ref|>text<|/ref|><|det|>[[115, 495, 876, 649]]<|/det|>
+We used the nomenclature as introduced by Zwicker: „Droplets can become chemically active if the material of the droplet is produced and destroyed by chemical reactions. An example that resembles a simple protocell is shown schematically in Fig. 1a. The droplet is formed by a droplet material D that is generated inside the droplet from a high-energy precursor N, which plays the role of a nutrient. Droplet material can degrade into a lower energy component, W, that plays the role of a waste, which leaves the droplet by diffusion. The droplet can survive if N is continuously supplied and W is continuously removed. This can be achieved by recycling N using an external energy source such as a fuel or radiation.“1  
+
+<|ref|>text<|/ref|><|det|>[[116, 665, 861, 717]]<|/det|>
+(1) Zwicker, D.; Seyboldt, R.; Weber, C. A.; Hyman, A. A.; Jülicher, F. Growth and division of active droplets provides a model for protocells. Nature Physics 2017, 13 (4), 408-413. DOI: 10.1038/nphys3984.  
+
+<|ref|>text<|/ref|><|det|>[[115, 750, 875, 850]]<|/det|>
+2. Despite the explanation in the reply to my comments, the introduction text still lacks specificity. As a reader, I am still not convinced that this is a cycle, or a controlled cycle: surely, while there is fuel, there is phosphorylation; and surely when you stop adding fuel, hydrolysis dominates, but that does not mean control. A key point of a cycle would be to demonstrate that His phosphorylation (and coacervates) can be recovered by a fresh addition of MAP. Can you show this?  
+
+<|ref|>text<|/ref|><|det|>[[116, 851, 844, 920]]<|/det|>
+We tested refueling with three cycles on the amino acid His on the NMR, to demonstrate that we can recover the phosphorylation by fresh addition of MAP. Besides, we also refueled the peptide with droplets two times on the plate reader. The graphs can be found in the SI.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 100, 877, 170]]<|/det|>
+a. Line 27 is an example of what I mean by lack of specificity in the writing: That way, the equilibrium position for the fuel-to-waste equilibrium lies to the right such that most fuel will eventually be converted into waste. (and btw line 33 can start a new paragraph)  
+
+<|ref|>text<|/ref|><|det|>[[115, 170, 880, 325]]<|/det|>
+We rephrased to: That way, the equilibrium position for the fuel-to-waste equilibrium lies on the waste side such that most fuel will eventually be converted into waste. This fuel-to-waste conversion should be slow so that a catalyst can accelerate it. Put differently, the fuel should be thermodynamically unstable, such as kinetically inert. In the catalytic reaction cycle, the catalyst accelerates the fuel-to-waste conversion. In doing so, the catalyst is temporarily activated by reacting with the fuel, after which it spontaneously deactivates. In other words, the catalyst can undergo numerous activation-deactivation cycles, and the fuel is converted fast due to the presence of the catalyst.  
+
+<|ref|>text<|/ref|><|det|>[[117, 325, 286, 340]]<|/det|>
+Highlighted in pink.  
+
+<|ref|>text<|/ref|><|det|>[[115, 357, 860, 444]]<|/det|>
+b. In line 287, you write "reaction cycle that phosphorylates histidine as an amino acid or in a peptide at the expense of phosphorylating agents". There is no need to say "at the expense of phosphorylating agents", this is the case in any chemical reaction. This and other similar repetitions take space from a more meaningful discussion/conclusion.  
+
+<|ref|>text<|/ref|><|det|>[[115, 444, 870, 478]]<|/det|>
+We agree that it is not needed. We removed it where we thought it was not required. Highlighted in pink.  
+
+<|ref|>sub_title<|/ref|><|det|>[[117, 511, 241, 528]]<|/det|>
+## 3. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 545, 881, 614]]<|/det|>
+4. Since you chose to keep the table in Figure 1, I find it that there needs to be some discussion in the text as to why the cycle only works with histidine, or why, if we know from the literature that MAP phosphorylates imidazole groups, other residues were also tested. What was the hypothesis?  
+
+<|ref|>text<|/ref|><|det|>[[115, 615, 880, 666]]<|/det|>
+We focused on amino acids and their peptides as they are powerful building blocks in peptide self-assembly, including liquid-liquid phase separation, which we show later in this paper. We emphasized our reasoning in the text and highlighted it in pink.  
+
+<|ref|>text<|/ref|><|det|>[[115, 683, 242, 750]]<|/det|>
+5. Addressed. 
+6. Addressed. 
+7. Addressed. 
+8. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[115, 767, 877, 836]]<|/det|>
+9. Partially addressed. The use of microreactors still confuses me, as there are many ways to passivate glass slides to prevent droplet adhesion. Can you at least include in SI micrographs of droplets outside these microreactors? This would show that the microreactor itself is not crucial to the droplets' properties.  
+
+<|ref|>text<|/ref|><|det|>[[115, 837, 875, 905]]<|/det|>
+To clarify, the microreactor helps us to study droplet formation without the downsides of flow in the sample, droplets settling on the glass, and droplets escaping out of the focal plane during imaging. We also understand your concern, so we added micrographs of droplets outside the microreactors to the SI.
+
+<--- Page Split --->
+<|ref|>text<|/ref|><|det|>[[115, 116, 875, 238]]<|/det|>
+10. In SI Figure 10, I see you now added the whole field of view, showing all microreactors. Can you include the emission channel as well, which is what you mostly show in the main text? The caption says it shows the emergence, growth and dissolution of the droplets, but I can only see emergence and dissolution. We added the emission channel of sulforhodamine B and the micrographs after 9 and 25 h. Fig. 3 f shows the growth more clearly as the percentage of droplet material per reactor correlates with the droplet's size.  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 255, 253, 272]]<|/det|>
+## 11. Addressed.  
+
+<|ref|>text<|/ref|><|det|>[[117, 290, 262, 306]]<|/det|>
+New comments:  
+
+<|ref|>text<|/ref|><|det|>[[117, 307, 878, 375]]<|/det|>
+12. Line 161: consider reorganizing the long list of references. The comparison to the literature is barely made, yet there are 11 references listed. I prefer it if you elaborate on this discussion rather than remove references, as I think it will make the manuscript stronger.  
+
+<|ref|>text<|/ref|><|det|>[[117, 375, 868, 410]]<|/det|>
+Indeed, 11 references were a lot, and some of them fitted better to other statements that have been made before. Changes are highlighted in pink.  
+
+<|ref|>text<|/ref|><|det|>[[117, 443, 872, 512]]<|/det|>
+13. The conclusions are still vague and imprecise despite another reviewer's comment. Most statements are based on this system being a dynamic cycle and the droplets being active, so I think resolving these conceptual problems in the main text will help better support the conclusions.  
+
+<|ref|>text<|/ref|><|det|>[[117, 529, 812, 580]]<|/det|>
+We have now clarified what we mean by active droplets (following your earlier statement) and addressed the fact that it is a cycle (following your proposed experiments). So, the conceptual problems should now be resolved.
+
+<--- Page Split --->
+<|ref|>sub_title<|/ref|><|det|>[[118, 85, 330, 101]]<|/det|>
+## REVIEWERS' COMMENTS  
+
+<|ref|>sub_title<|/ref|><|det|>[[118, 137, 437, 154]]<|/det|>
+## Reviewer #1 (Remarks to the Author):  
+
+<|ref|>text<|/ref|><|det|>[[118, 163, 856, 234]]<|/det|>
+The authors have now addressed all remarks and included new experiments. I believe the manuscript's clarity and scientific rigor improved greatly during the revision process, and appreciate their effort in doing so. I support publication of the most recent version.
+
+<--- Page Split --->

peer_reviews/supplementary_0_Peer Review File__d76e4ddfc072109f1cb55abf879617c13e06aa86bd4b76bb284d09932e20ab1c/images_list.json

@@ -0,0 +1 @@
+[]

peer_reviews/supplementary_0_Peer Review File__d76e4ddfc072109f1cb55abf879617c13e06aa86bd4b76bb284d09932e20ab1c/supplementary_0_Peer Review File__d76e4ddfc072109f1cb55abf879617c13e06aa86bd4b76bb284d09932e20ab1c.mmd

@@ -0,0 +1,458 @@
+
+# nature portfolio  
+
+Peer Review File  
+
+Cyclic AMP induces reversible EPAC1 condensates that regulate histone transcription  
+
+![PLACEHOLDER_0_0]
+  
+
+Open Access This file is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. In the cases where the authors are anonymous, such as is the case for the reports of anonymous peer reviewers, author attribution should be to 'Anonymous Referee' followed by a clear attribution to the source work. The images or other third party material in this file are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
+
+<--- Page Split --->
+
+Reviewer #1 (Remarks to the Author):  
+
+The manuscript by Lannucci et al investigated and reports an unexpected mechanism underlying cAMP- induced nuclear condensation of EPAC1 and identification of the primary functional target regulated by nuclear EPAC1 condensates. Using various in vitro approaches, the authors found specific mechanism regulating the entry of EPAC1 into the nucleus, dependent of intrinsically disordered regions present at its amino- terminus. They demonstrate that nuclear EPAC1 condensates assemble at genomic loci on chromosome 6 and promote the transcription of a histone gene cluster. Overall, the experiments were well designed and executed. The results are important given the phenomenon of cAMP- induced nuclear condensation of EPAC was reported. Especially the results about the mechanism underlying EPAC1 entering into nucleus are convincing and support two distinct regions for engaging the nuclear pore and entering. However, I have three major concerning raised by the information reported in a new publication (Yang et al, Science Advances April 20, 2022. DOI: 10.1126/sciadv.abm2960) about the functional roles of cAMP- induced EPAC1 nuclear condensates in cell models of HUVEC and HEK.  
+
+1. Using high throughput whole transcriptome RNA-sequencing, the authors were focusing on identification and characterization the functional role of nuclear EPAC1 condensates on histone gene cluster. Yang et al reported Activation of Epac1 by intracellular cAMP triggers phase separation and the formation of nuclear condensates containing Epac1 and general components of the SUMOylation machinery to promote cellular sumoylation (DOI: 10.1126/sciadv.abm2960). Given that histone sumoylation is associated with regulation on transcriptions (Shii O Y, et al, PNAS 2003;100:13225https://doi.org/10.1073/pnas.1735528100), the authors need to address and discuss the potential role of histone sumoylation in EPAC1 nuclear condensate-regulated transcription.  
+
+2. Now, the authors should address or discuss the differences in all presented scientific data between this manuscript and the Yang et al reported one, demonstrating the novelty and deeper insight into similar topic.  
+
+3. In mammals, the EPAC protein family contains two members: EPAC1 and EPAC2. Both EPAC isoforms function by responding to increased intracellular cAMP levels in a PKA-independent manner and act on the same immediate downstream effectors, the small G proteins Rap1 and Rap2. EPAC1 is the major isoform in HUVECs, while both are detected in HEK293 cells. The author should discuss their claimed unexpected information in the context of EPAC2, in particular within their cell model using HEK293.  
+
+Minor comment:  
+
+1. Line 21, after stating "accumulating evidence", there is only one citation of reference 6.  
+
+2. Line 34, please extend description by more information about "the classic model".  
+
+3. Line 43, correct the typo "that PKA it not..."  
+
+4. Line 51, please correct grammar error "which effectively enters..."
+
+<--- Page Split --->
+
+5. Line 70 "we overexpressed EPAC1-YFP in EPAC1-deficient HEK cells". There is critical technical information missing about "EPAC1-deficient HEK cells". Do the authors use knock-out or knock-down technology?  
+
+6. Line 398, why was there such big range 4 to 10 to select image fields between different experiments.  
+
+7. Line 400, please describe in detail the standard to define as "selected nuclear ROIs".  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript by Iannucci et al addresses the role of EPAC1 in response to cyclic AMP and describe a previously uncharacterized function in the regulation of histone transcription. This regulation involves the ability of EPAC1 to undergo liquid- liquid phase separation. Ehese experiments have the potential to provide new insights on how cAMP might regulate transcription. However, despite this potential impact, this manuscript suffers of several shortcomings.  
+
+1. The claim that EPAC can phase separate remains preliminary. The experiments performed by the authors have often been used in the literature as evidence of phase separation but they are very limited and only show consistency. Several claims that the authors made should be quantified. For example, it is not immediately clear from the images that the condensates are indeed spherical. Similarly, the paper could benefit from experiments showing that fusing condensate regain sphericity quickly. While a definite demonstration that condensate indeed form by LLPS is a difficult task, this paper could be strengthened by at the very least some quantifications. At this point, the claim is not strongly supported.  
+
+2. The overlap between the EPAC and histone locus and the histone locus body is not easy to interpret mechanistically. The EPAC condensate appears to colocalize but not perfectly with the histone locus on chromosome 6. The colocalization in Figure 4 with NPAT appears stronger. However, there is only one example shown and not quantification. EPAC condensate are not passive followers of HLBs, as in the absence of cAMP stimulation there are no condensate but HLBs are still present. Moreover, the EPAC condensates do not always overlap with the HLBs. It would be important to understand these relationship in quantitative term and their possible hierarchy. For example, it is unclear if HLBs are required for EPAC condensate to overlap with the histone locus. Experiments to understand the relationship between the two biocondensates would be crucial.  
+
+## Minor point:  
+
+Recent work has shown that the HLB forms by phase separation (Hur et al Dev Cell 2020) and has elucidated ways to perturb this process. Addressing the effects of some of these perturbations on EPAC biocondense could be useful.
+
+<--- Page Split --->
+
+Reviewer #3 (Remarks to the Author):  
+
+\*\*Summary\*\*  
+
+The overarching goal of this work by Iannucci et al. is to uncover the mechanism of PKA- independent nuclear cAMP signaling. To this end, the authors focus on the cAMP- activated GEF EPAC1 that is know to shuttle from the cytoplasm into the nucleus upon cAMP production. This work nicely combines mutagenesis and chemical biology to show that nuclear EPAC1 localizes to cAMP- induced and - dependent foci. The authors furthermore intriguingly link EPAC1 foci localization to differences in gene expression, most notably of histone gene clusters on chromosome 6. The authors conclude that EPAC1 forms nuclear condensates in response to PKA- independent cAMP signaling and that this cAMP/EPAC1 condensate axis represents a novel signaling paradigm. Overall, this work is interesting and thought- provoking, and in principle a nice fit for Nature Communications. However, the authors over- interpret their results throughout the manuscript, raising doubts that EPAC1 is indeed a central hub in PKA- independent cAMP signaling. Thus, this work requires major revisions.  
+
+\*\*Major comments\*\*  
+
+Specifically, the burden of proof has not been met for these key statements:  
+
+1. ) The observed co-localization of EPAC1Δ179-208-YFP and RANBP2 in Extended Data Figure 1D is not sufficient to draw the conclusion that they directly bind each other, especially given that: (a) this is a novel EPAC1 mutation, (b) nuclear localization signals (NLSs) typically bind to nuclear transport receptors (NTRs) and not nucleoporins like RANBP2, and (c) there is prior biochemical evidence for a direct, NLS/NTR-independent tethering of EPAC1 to RANBP2 (Gloerich et al. 2011). Thus, to truly identify the specific mechanism of EPAC1 nuclear import (as claimed in lines 68-69), the authors must directly test binding of wild-type and mutant EPAC1 to NTRs and nucleoporins, for example by immunoprecipitation assays (IPs) or binding assays with purified components.  
+
+2. ) The data presented is not sufficient to definitively conclude that EPAC1 forms condensates, as opposed to being merely recruited to condensates. In fact, the authors show evidence for the latter in the form of (a) the (incomplete) co-localization of EPAC1 foci, PML bodies and HLBs, and (b) changes in histone gene cluster transcription between wild-type and mutant EPAC1. To show that EPAC1 is indeed the core scaffold component of a novel type of condensates (that unequivocally can be called EPAC1 condensates), the authors must reconstitute the phase separation of EPAC1 in vitro.
+
+<--- Page Split --->
+
+3. ) The author's efforts to map the domains driving the potential phase separation of EPAC1 are cursory, leaving more doubts than providing understanding for three main reasons. First, IDRs are by no means cardinal to phase separation as claimed in line 126. What is cardinal to phase separation is multivalency, which can be achieved through both folded and disordered domains. Indeed, the prevalent view in the field is that IDR typically work in concert with oligomerization domains, and that IDRs often merely tune phase separation behavior (see e.g. Martin and Holehouse, 2020). Second, EPAC1 appears to only have one very short IDR (approx. residues 1-66) based on DisoPred3, NetSurfP2 and AlphaFold2 predictions. The other IDRs that Iannucci et al. highlighted merely seem to be classic disordered linkers connecting folded domains. Third, it appears that EPAC1 contains folded domains like the DEP domain that are know to mediate protein-protein interactions in signaling pathways. In this regard, the author note that truncating the DEP domain (by virtue of the EPAC1A48-148 mutant) interferes with condensate localization. Together, the domain mapping appears biased towards pushing the preconceived model that EPAC1 undergoes IDR-dependent liquid-liquid phase separation, whereas both the data presented in this manuscript as well as the literature cast doubts on this model. Thus, the authors must devise more comprehensive and unbiased experiments to distinguish between the EPAC1 phase separation and condensate recruitment models (and identify the domains that drive this), or significantly revise and tone down their conclusions.  
+
+\*\*Minor comments\*\*  
+
+- How were the EPAC1-deficient HEK cells generated? This information is missing in the manuscript, despite the importance of this resource for this study.  
+
+- Lines 115-118: Imprecise argument. Hexanediol sensitivity is not diagnostic of LLPS, but merely suggests that hydrophobic interactions may be relevant. Much stronger evidence for the model that EPAC1 behaves as if in a viscous liquid phase comes from the fusion and FRAP experiments.  
+
+- It would greatly help if the authors would consolidate the EPAC1 domain architecture cartoons in Figure 3A and Extended Data Figure 1A into one scheme.  
+
+- Structuring the manuscript into sections with subheadings would greatly help the flow of the paper.
+
+<--- Page Split --->
+
+Reviewer #4 (Remarks to the Author):  
+
+In this manuscript, Iannucci et al. explored the nuclear functions of cAMP effector, EPAC1. They found that upon cAMP elevation, EPAC1 enters into the nucleus where it forms reversible biomolecular condenstates through liquid- liquid phase separation. This phenomenon is independent on the canonical effector of cAMP, PKA. Moreover, they found that EPAC1 condensates assemble on chromosome 6 with colocalization with histone gene regulator NPAT to promote histone gene transcription. Overall, this is an interesting story as it uncovered a mechanism through which cAMP contributes to nuclear spatial compartmentalization and promotes the transcription of specific genes. However, lacking of proper controls and highly rely on cell imaging make some conclusions not convincing. The following are my concerns that need addressed.  
+
+1. My big concern is whether and how the EPAC1 condensate regulates histone gene expression. It is well-known that histone gene expression is tightly controlled by cell cycle. It is unknown whether treatment with cAMP agonist will affect cell cycle. The authors should try to explain how EPAC1 condensates regulate histone gene expression. What about the EPAC1 mutants? Does the EPAC1 732-764 mutant and EPAC1179-208 mutant that cannot enter into the nucleus regulate histone gene expression independent on cell cycle?  
+
+2. Fig. 4: The co-localization of NPAT with EPAC1 condensates is not that clear. Moreover, it is recommended to use other techniques, such as Co-IP to confirm the interaction between NPAT and EPAC1. ChIP EPAC1 and NPAT at histone locus is also a better way to demonstrate that point.  
+
+3. To make the story more convincing, the authors are recommended to express and purify the recombinant EPAC1 to demonstrate it can form liquid-liquid phase separation in vitro.  
+
+4. The probe signal for chromosome 21 is very weak. It is too preliminary to conclude that there is no colocalization between chromosome 21 and EPAC1 condensates given the so weak signals.  
+
+5. Extended data Fig. 1: The authors should examine whether these two domains affect the protein expression of EPAC1. For extended data 1c and 1d, WT EPAC1 should be added as a positive control.  
+
+Other comments:  
+
+6. Fig. 1c: what is time points used for calculation?
+
+<--- Page Split --->
+
+7. Fig. 1d and 1g: The DAPI should be added to indicate the nucleus.  
+
+8. Fig. 5b: the EPAC1 1-148 mutant should be included to show its effect on histone gene expression.  
+
+9. The language of this manuscript needs to be polished. There are a lot of grammar errors and mistakes.
+
+<--- Page Split --->
+
+Cyclic AMP- induced reversible EPAC1 condensates regulate histone transcription Iannucci et. al.  
+
+## Point-by-point response to the reviewers' comments  
+
+Firstly, we would like to thank the reviewers for taking the time to review our manuscript and their insightful comments. In the present version we added more experiments solving the concerns expressed. Hereafter a point- by- point response is presented.  
+
+Reviewer #1 (Remarks to the Author):  
+
+The manuscript by Iannucci et al investigated and reports an unexpected mechanism underlying cAMP- induced nuclear condensation of EPAC1 and identification of the primary functional target regulated by nuclear EPAC1 condensates. Using various in vitro approaches, the authors found specific mechanism regulating the entry of EPAC1 into the nucleus, dependent of intrinsically disordered regions present at its amino- terminus. They demonstrate that nuclear EPAC1 condensates assemble at genomic loci on chromosome 6 and promote the transcription of a histone gene cluster. Overall, the experiments were well designed and executed. The results are important given the phenomenon of cAMP- induced nuclear condensation of EPAC was reported. Especially the results about the mechanism underlying EPAC1 entering into nucleus are convincing and support two distinct regions for engaging the nuclear pore and entering. However, I have three major concerning raised by the information reported in a new publication (Yang et al, Science Advances April 20, 2022. DOI: 10.1126/sciadv.abm2960) about the functional roles of cAMP- induced EPAC1 nuclear condensates in cell models of HUVEC and HEK.  
+
+## Major points  
+
+1. Using high throughput whole transcriptome RNA-sequencing, the authors were focusing on identification and characterization the functional role of nuclear EPAC1 condensates on histone gene cluster. Yang et al reported Activation of Epac1 by intracellular cAMP triggers phase separation and the formation of nuclear condensates containing Epac1 and general components of the SUMOylation machinery to promote cellular sumoylation (DOI: 10.1126/sciadv.abm2960). Given that histone sumoylation is associated with regulation on transcriptions (Shiio Y, et al, PNAS 2003; 100:13225https://doi.org/10.1073/pnas. 1735528100), the authors need to address and discuss the potential role of histone sumoylation in EPAC1 nuclear condensate-regulated transcription.  
+
+We agree with the reviewer that the involvement of EPAC1 condensates in SUMOylation raises the question of whether this process is at the basis of the transcriptional effects we observe on the histone locus 1. As demonstrated by the manuscript suggested by the reviewer<sup>1</sup> and other literature<sup>2</sup>, the process of SUMOylation is mostly connected to transcriptional repression, while on the contrary, we observe increased transcription of histone genes in response to nEPAC1 condensate formation. Nevertheless, to address the reviewer's concern, we performed ad hoc experiments.
+
+<--- Page Split --->
+
+- We tested the ability of nEPAC1 condensates to activate transcription of histones in the presence or absence of a SUMOylation inhibitor. As shown in the new Extended Data Figure 7, a SUMOylation inhibitor that was fully able to completely reverse the effect of EPAC1 activation on SUMOylation, had no effect on histone transcription assessed by real time experiments on a representative histone gene.  
+
+- In addition, we performed immunofluorescence experiments simultaneously labelling histone locus bodies (NPAT) and SUMO2/3 in cells expressing EPAC1-YFP. Interestingly, we observed colocalization of EPAC1 with NPAT and SUMO2/3 but we never found a complex containing EPAC1 NPAT and SUMO2/3 Extended Data Figure 7a,b. In the new version of the manuscript, we present the new data and discuss the implications of EPAC1 condensates on SUMOylation and transcription.  
+
+2. Now, the authors should address or discuss the differences in all presented scientific data between this manuscript and the Yang et al reported one, demonstrating the novelty and deeper insight into similar topic.  
+
+The manuscript from Yang et al. was published while our manuscript was in revision, and presents some similarities with our work, albeit not extensive. As requested by the reviewer we performed experiments that excluded the possible mechanistic overlap between our results and those of Yang et al. (please see answer to previous point), these data were also further elaborated in the discussion. Other than experimentally demonstrating no mechanistic overlap, our manuscript offers a more detailed analysis of the mechanism through which EPAC1 enters the nucleus (Extended Data Figure 1) not at all addressed by Yang et al. Moreover, we present a more detailed map of the sequences necessary for the phase separation of EPAC1 and identify their importance in the recruitment of EPAC1 in condensates (Figure 3 and Extended Data Figure 4 & 5). Finally, we also offer a more detailed view of the modalities through which nEPAC1 condensates interact with different nuclear membraneless organelles, which revealed to be surprisingly organelle specific (Figure 4f). The more detailed analysis should be added to the completely unrelated and novel finding that EPAC1 condensates regulate histone transcription. Based on these considerations, in our opinion, Yang et al. does not affect the novelty of our findings. On the contrary, by reporting an unrelated and distinct function of EPAC1 condensates, Yang et al reinforces the need of more studies to better define how the cAMP/EPAC1 axis in the nucleus achieves its specificity of action and contributes to cell physiology.  
+
+3. In mammals, the EPAC protein family contains two members: EPAC1 and EPAC2. Both EPAC isoforms function by responding to increased intracellular cAMP levels in a PKA-independent manner and act on the same immediate downstream effectors, the small G proteins Rap1 and Rap2. EPAC1 is the major isoform in HUVECs, while both are detected in HEK293 cells. The author should discuss their claimed unexpected information in the context of EPAC2, in particular within their cell model using HEK293.  
+
+Contrary to EPAC1, that was shown to localize to the nuclear envelope, EPAC2 presents no nuclear or perinuclear localization and is mainly found (in a isoform-dependent manner) in the cytosol or plasma membrane<sup>3</sup>. In addition, the cell permeable cAMP agonist used for the
+
+<--- Page Split --->
+
+expression experiments was shown to be a better activator for EPAC1 as compared to EPAC2<sup>4</sup> (suggesting that the transcriptional effects were most likely due to the activation of the former). Nevertheless, we agree with the reviewer that the structural similarity of the two proteins and the existence of common pathways between the two, dictate further investigation. For this reason, in the present version of the manuscript, we performed experiments addressing the subcellular localization of EPAC1 and its behavior in response to cAMP- dependent activation. In the new Extended Data Figure 3a, Western Blotting experiments demonstrate that EPAC2 was not found in the nuclear fractions of HEK cells. Moreover, in Extended Data Figure 3b- c, imaging experiments demonstrate that exogenously EPAC2 does not enter the nucleus and does not participate in the formation of any structures in response to the activation of the cAMP signalling cascade.  
+
+## Minor points:  
+
+1. Line 21, after stating "accumulating evidence", there is only one citation of reference 6.  
+
+In the Nature Communications format, it is not allowed to add references in the abstract therefore we had to take away reference 6. Nevertheless, the point on the evidence suggesting that PKA is not the only effector of cAMP in the nucleus is further discussed in the new introduction section and 3 references (12- 15) have been added.  
+
+2. Line 34, please extend description by more information about "the classic model".  
+
+In the new version we further elaborate on the classic model of the modalities through which PKA can be activated in the nucleus.  
+
+3. Line 43, correct the typo "that PKA it not..."  
+
+## Done  
+
+4. Line 51, please correct grammar error "which effectively enters..."  
+
+This phrase has been replaced in the new version of the manuscript.  
+
+5. Line 70 "we overexpressed EPAC1-YFP in EPAC1-deficient HEK cells". There is critical technical information missing about "EPAC1-deficient HEK cells". Do the authors use knockout or knock-down technology?  
+
+The HEK cell line is well documented to be naturally EPAC1 deficient<sup>5</sup>, likely due to hypermethylation. The lack of EPAC is demonstrated in the manuscript in figure 1a by Western Blotting. In the new version of the manuscript, we make a better point on the deficiency of EPAC1 in HEK cells.  
+
+6. Line 398, why was there such big range 4 to 10 to select image fields between different experiments.
+
+<--- Page Split --->
+
+The lower number of cells depended on the low availability of the probe for the Chromosome 21. In the new version of the manuscript, we repeated these experiments and applied new probes as well (Figure 7).  
+
+7. Line 400, please describe in detail the standard to define as "selected nuclear ROIs".  
+
+Done  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript by Iannucci et al addresses the role of EPAC1 in response to cyclic AMP and describe a previously uncharacterized function in the regulation of histone transcription. This regulation involves the ability of EPAC1 to undergo liquid- liquid phase separation. These experiments have the potential to provide new insights on how cAMP might regulate transcription. However, despite this potential impact, this manuscript suffers of several shortcomings.  
+
+## Major points  
+
+1. The claim that EPAC can phase separate remains preliminary. The experiments performed by the authors have often been used in the literature as evidence of phase separation but they are very limited and only show consistency. Several claims that the authors made should be quantified. For example, it is not immediately clear from the images that the condensates are indeed spherical. Similarly, the paper could benefit from experiments showing that fusing condensate regain sphericity quickly. While a definite demonstration that condensate indeed form by LLPS is a difficult task, this paper could be strengthened by at the very least some quantifications. At this point, the claim is not strongly supported.  
+
+We thank the reviewer for the comments and suggestions. In the new version of the manuscript, we performed both the requested analysis and new experiments to further consolidate our claims on EPAC1 phase separation. In particular:  
+
+- As suggested by the reviewer we calculated the circularity index of EPAC1 condensates presented now in Figure 2b. In addition, using superresolution Airyscan microscopy we reconstructed condensates and both a rendering and a movie are presented (Figure 2b and Extended Data Movie 3).- As suggested by the reviewer we calculated the circularity index of condensates before, during and after fusion and confirmed that after the expected momentaneous loss of circularity during fusion, the resulting condensate rapidly becomes spherical (Figure 2d).- In the new version of the manuscript, we also tested the ability of purified EPAC1 to constitute condensates. We found that purified EPAC1 forms condensates in the presence of cAMP, while the mutant \(\Delta 2 - 148\) was unable to form condensates independently of the cAMP levels. These data are now presented in Figure 3c and Extended figure 4b-d.
+
+<--- Page Split --->
+
+- In line with these data, we also tested the ability of the EPAC1 mutants \(\Delta 2 - 148\) , \(\Delta 2 - 24\) and \(\Delta 48 - 148\) to participate to condensates formed by EPAC1 wild type. We found that \(\Delta 2 - 148\) was unable to participate, however both other mutants were able to participate in the growth of EPAC1 condensates albeit unable to trigger condensate formation by their own. These data are now presented in Extended figure 5.  
+
+2. The overlap between the EPAC and histone locus and the histone locus body is not easy to interpret mechanistically. The EPAC condensate appears to colocalize but not perfectly with the histone locus on chromosome 6. The colocalization in Figure 4 with NPAT appears stronger. However, there is only one example shown and not quantification. EPAC condensate are not passive followers of HLBs, as in the absence of cAMP stimulation there are no condensate but HLBs are still present. Moreover, the EPAC condensates do not always overlap with the HLBs. It would be important to understand these relationship in quantitative term and their possible hierarchy. For example, it is unclear if HLBs are required for EPAC condensate to overlap with the histone locus. Experiments to understand the relationship between the two biocondensates would be crucial.  
+
+In the new version of the manuscript, we performed both, analysis and new experiments that better define the relation of EPAC1 condensates and HLBs.  
+
+- We performed quantification analysis of the overlap between EPAC1 condensates and other nuclear condensates (including HLBs). These data are now present in Figure 4e and demonstrate that EPAC1 and NPAT colocalize.- In figure 4f we also present a 3D rendering of the interactions between nEPAC1 condensates and HLBs, PML and Cajal bodies which evidence clear differences on how the interaction between these organelles occurs.- We further studied the relationship of EPAC1 condensates to NPAT and SUMO2/3. These experiments evidenced that EPAC1 can interact with NPAT or SUMO2/3 individually, however, the interaction between the three proteins was not detected, excluding SUMOylation as the mechanistic link between the two condensates. Extended Data Figure 7a,b.  
+
+Together our data suggest that EPAC1 condensates can interact with other nuclear condensates, however with different modalities. The most striking difference is that between PML and NPAT. As can be seen from the superresolution images and analysis, PML and EPAC1 constitute hybrid condensates with the two proteins distributed within the same structure (Figure 4f). On the other hand, NPAT and EPAC1 condensates partially overlap without though constituting a single structure and maintaining their condensate "identity". These data would suggest that EPAC1 could have different regulatory activities in the cell, for instance by interaction with PML could affect SUMOylation (as suggested by Yang et al, Science Advances April 20, 2022. DOI: 10.1126/sciadv.amb2960), while on the other hand, the interaction with HLBs allows EPAC1 to impinge in histone transcription.  
+
+## Minor point:  
+
+Recent work has shown that the HLB forms by phase separation (Hur et al Dev Cell 2020) and has elucidated ways to perturb this process. Addressing the effects of some of these perturbations on EPAC biocondense could be useful.
+
+<--- Page Split --->
+
+We thank the reviewer for the suggestion. In this manuscript Hur and colleagues study the modalities of HLBs formation in \*Drosophila\* embryos. The authors combine bioinformatic in silico approaches with genetics on embryos, that while very interesting and informative would be too difficult and too time- consuming to apply to our manuscript. In addition, the authors also propose two pharmacological approaches that apparently blocked the recruitment of mxc (the \*Drosophila\* orthologue of NPAT) to the HLBs (alpha- amanitin and SNS- 032), resulting in smaller HLBs. In our opinion modification of the HLB size is out of the focus of the present study.  
+
+## Reviewer #3 (Remarks to the Author):  
+
+The overarching goal of this work by Iannucci et al. is to uncover the mechanism of PKA- independent nuclear cAMP signaling. To this end, the authors focus on the cAMP- activated GEF EPAC1 that is know to shuttle from the cytoplasm into the nucleus upon cAMP production. This work nicely combines mutagenesis and chemical biology to show that nuclear EPAC1 localizes to cAMP- induced and - dependent foci. The authors furthermore intriguingly link EPAC1 foci localization to differences in gene expression, most notably of histone gene clusters on chromosome 6. The authors conclude that EPAC1 forms nuclear condensates in response to PKA- independent cAMP signaling and that this cAMP/EPAC1 condensate axis represents a novel signaling paradigm. Overall, this work is interesting and thought- provoking, and in principle a nice fit for Nature Communications. However, the authors over- interpret their results throughout the manuscript, raising doubts that EPAC1 is indeed a central hub inPKA- independent cAMP signaling. Thus, this work requires major revisions. Specifically, the burden of proof has not been met for thee key statements:  
+
+## Major points  
+
+1. ) The observed co-localization of EPAC1Δ179-208-YFP and RANBP2 in Extended Data Figure 1D is not sufficient to draw the conclusion that they directly bind each other, especially given that: (a) this is a novel EPAC1 mutation, (b) nuclear localization signals (NLSs) typically bind to nuclear transport receptors (NTRs) and not nucleoporins like RANBP2, and (c) there is prior biochemical evidence for a direct, NLS/NTR-independent tethering of EPAC1 to RANBP2 (Gloerich et al. 2011). Thus, to truly identify the specific mechanism of EPAC1 nuclear import (as claimed in lines 68-69), the authors must directly test binding of wild-type and mutant EPAC1 to NTRs and nucleoporins, for example by immunoprecipitation assays (IPs) or binding assays with purified components.  
+
+While we agree that the identification of the partners through which EPAC1 enters the nucleus would increase our molecular understanding on how this protein transits the nuclear envelope, we feel that this level of mechanistic detail is beyond the scope of the present study. In the field of EPAC1, at least to our knowledge, there is no unequivocal evidence to confirm that EPAC1 enters the nucleus and most literature assumes the nuclear presence of EPAC1 based on its binding to the nuclear envelope (in particular its binding to RANBP2<sup>6</sup> and Nup98<sup>7</sup>). We reasoned that a protein as big and complex as EPAC1 would require a precise molecular mechanism to enter the nucleus, and even more importantly, the existence of a
+
+<--- Page Split --->
+
+conserved mechanism would strongly suggest a functional role of EPAC1 in that compartment.  
+
+As requested by the reviewer we performed more experiments to substantiate the data suggesting that two sequences are both important for EPAC1 entry and apparently have different roles. In particular:  
+
+1. We repeated both the Western Blotting and imaging experiments for all constructs (EPAC1-YFP, \(\Delta 179 - 208\) and \(\Delta 732 - 764)\) and confirmed our previous results Extended figure 1b-e. 
+2. As requested by the reviewer we also performed pull down experiments that demonstrate the importance of residues 732-764 in the binding of RANBP2 (Extended figure 1f).  
+
+2. ) The data presented is not sufficient to definitively conclude that EPAC1 forms condensates, as opposed to being merely recruited to condensates. In fact, the authors show evidence for the latter in the form of (a) the (incomplete) co-localization of EPAC1 foci, PML bodies and HLBs, and (b) changes in histone gene cluster transcription between wild-type and mutant EPAC1. To show that EPAC1 is indeed the core scaffold component of a novel type of condensates (that unequivocally can be called EPAC1 condensates), the authors must reconstitute the phase separation of EPAC1 in vitro.  
+
+We agree with the reviewer that reconstitution of EPAC1 condensates in vitro would further confirm the ability of EPAC1 to undergo phase separation. In a recent manuscript published while ours was under revision, Yang et al, Science Advances April 20, 2022. DOI: 10.1126/sciadv.abm2960 presented evidence for EPAC1 phase separation in vitro, however no information on EPAC1 mutants was available. For this reason, we purified EPAC1-WT and EPAC1 \(\Delta 2 - 148\) and performed experiments to reconstitute phase separation in vitro. We found that EPAC1WT undergoes phase separation in the presence of cAMP (in salt concentrations around 150mM which are similar to physiological ones). On the other hand, EPAC1 \(\Delta 2 - 148\) was insensible to cAMP variations and did not phase separate. These data are presented in figure 3c,d and Extended figure 4b-d.  
+
+3. ) The author's efforts to map the domains driving the potential phase separation of EPAC1 are cursory, leaving more doubts than providing understanding for three main reasons. First, IDRs are by no means cardinal to phase separation as claimed in line 126. What is cardinal to phase separation is multivalency, which can be achieved through both folded and disordered domains. Indeed, the prevalent view in the field is that IDR typically work in concert with oligomerization domains, and that IDRs often merely tune phase separation behavior (see e.g. Martin and Holehouse, 2020). Second, EPAC1 appears to only have one very short IDR (approx. residues 1-66) based on DisoPred3, NetSurfP2 and AlphaFold2 predictions. The other IDRs that Iannucci et al. highlighted merely seem to be classic disordered linkers connecting folded domains. Third, it appears that EPAC1 contains folded domains like the DEP domain that are know to mediate protein-protein interactions in signaling pathways. In this regard, the author note that truncating the DEP domain (by virtue of the EPAC1 \(\Delta 48 - 148\) mutant) interferes with condensate localization. Together, the domain mapping appears biased towards pushing the preconceived model that EPAC1 undergoes IDR-dependent liquid-liquid phase separation, whereas both the data presented in this manuscript as well as the literature cast doubts on this model. Thus, the authors must
+
+<--- Page Split --->
+
+devise more comprehensive and unbiased experiments to distinguish between the EPAC1 phase separation and condensate recruitment models (and identify the domains that drive this), or significantly revise and tone down their conclusions.  
+
+As suggested by the reviewer we did both, revised and toned down our conclusions and devised new experiments that allowed us to better define the role of the amino terminus IDRs of EPAC1. The new evidence presented are:  
+
+1. In vitro experiments demonstrated that the EPAC1 amino terminus is necessary for phase separation figure 3c,d (please see also previous point)  
+2. We performed reconstitution experiments using the EPAC1 mutants and found that \(\Delta 2 - 148\) was unable to form condensates in response to cAMP but also to participate on the condensates generated by EPAC1-WT. On the other hand, the mutants \(\Delta 2 - 24\) and \(\Delta 48 - 148\) that did not form condensates in response to cAMP, promptly participated in the formation of EPAC1-WT condensates suggesting a role of these domains in the processes that trigger condensation and pointing to the importance of this region both for the condensation and recruitment processes (Extended figure 5).  
+
+## Minor points:  
+
+1. How were the EPAC1-deficient HEK cells generated? This information is missing in the manuscript, despite the importance of this resource for this study.  
+
+The HEK cell line is well documented to be naturally EPAC1 deficient<sup>5</sup>. The lack of EPAC is demonstrated in the manuscript in figure 1a by Western Blotting. In the new version of the manuscript, we make a better point on the deficiency of EPAC1 in HEK cells.  
+
+2. Lines 115-118: Imprecise argument. Hexanediol sensitivity is not diagnostic of LLPS, but merely suggests that hydrophobic interactions may be relevant. Much stronger evidence for the model that EPAC1 behaves as if in a viscous liquid phase comes from the fusion and FRAP experiments.  
+
+We thank the reviewer for the suggestion, in the new version of the manuscript we re- wrote that part accordingly.  
+
+3. It would greatly help if the authors would consolidate the EPAC1 domain architecture cartoons in Figure 3A and Extended Data Figure 1A into one scheme.  
+
+As suggested by the reviewer we have integrated the domain architecture of EPAC1 in figure 3a.  
+
+4. Structuring the manuscript into sections with subheadings would greatly help the flow of the paper.  
+
+The new version of the manuscript is conformed to the Nature Communications format that requires sections defined by subheadings.  
+
+Reviewer #4 (Remarks to the Author):
+
+<--- Page Split --->
+
+In this manuscript, Iannucci et al. explored the nuclear functions of cAMP effector, EPAC1. They found that upon cAMP elevation, EPAC1 enters into the nucleus where it forms reversible biomolecular condenstates through liquid- liquid phase separation. This phenomenon is independent on the canonical effector of cAMP, PKA. Moreover, they found that EPAC1 condensates assemble on chromosome 6 with colocalization with histone gene regulator NPAT to promote histone gene transcription. Overall, this is an interesting story as it uncovered a mechanism through which cAMP contributes to nuclear spatial compartmentalization and promotes the transcription of specific genes. However, lacking of proper controls and highly rely on cell imaging make some conclusions not convincing. The following are my concerns that need addressed.  
+
+## Major points  
+
+1. My big concern is whether and how the EPAC1 condensate regulates histone gene expression. It is well-known that histone gene expression is tightly controlled by cell cycle. It is unknown whether treatment with cAMP agonist will affect cell cycle. The authors should try to explain how EPAC1 condensates regulate histone gene expression. What about the EPAC1 mutants? Does the EPAC1 732-764 mutant and EPAC1 179-208 mutant that cannot enter into the nucleus regulate histone gene expression independent on cell cycle?  
+
+In the first version of the manuscript, we did not address the cell cycle- dependence of histone transcription because we reasoned that our treatment (40 minutes of the cell permanent EPAC- specific cAMP analog 8CPT- cAMP) was unlikely to dramatically affect the cell cycle progression. As requested by the reviewer we used Fluorescence- activated Cell Sorting (FACS) to monitor the variation of cell cycle progression in cells expressing EPAC1 treated with 8CPT- cAMP. As shown in Extended figure 6d, treatment with 8CPT- cAMP had no effect on cell cycle progression as compared to the vehicle control and thus is unlikely that the effects on histone transcription depend on cell cycle variations.  
+
+In regards of the effects that the different mutants may have on transcription, we feel that the results of the original RNAseq experiments, demonstrating that the \(\Delta 2 - 148\) mutant (which enters the nucleus, contains a fully functional catalytic domain, and can respond to cAMP but is unable to form condensates) has no transcriptional effects, clearly suggests that EPAC1 condensates are the means through which histone transcription is activated in response to cAMP. For this reason, we feel that testing the ability of EPAC1 mutants that do not enter the nucleus to alter transcription would be less informative than the \(\Delta 2 - 148\) mutant.  
+
+2. Fig. 4: The co-localization of NPAT with EPAC1 condensates is not that clear. Moreover, it is recommended to use other techniques, such as Co-IP to confirm the interaction between NPAT and EPAC1. ChIP EPAC1 and NPAT at histone locus is also a better way to demonstrate that point.  
+
+In the new version of the manuscript, we present new data that confirm the co- localization of NPAT and EPAC1 condensates (new Extended figure 7a,b). As suggested from our imaging experiments, NPAT and EPAC1 appear to be part of partially overlapping condensates
+
+<--- Page Split --->
+
+(contrary to EPAC1 and PML that are clearly part of the same structures (see also response to major point 2 Reviewer #2). As requested by the reviewer we performed co- IP experiments, but we were unable to detect interaction of EPAC1 and NPAT, which is not surprising considering the low number of interactions between HLBs and EPAC1 condensates observed by immunofluorescence (Figure 4f).  
+
+3. To make the story more convincing, the authors are recommended to express and purify the recombinant EPAC1 to demonstrate it can form liquid-liquid phase separation in vitro.  
+
+We now present data demonstrating the reconstruction of EPAC1 condensates in vitro (please see answer to major point 2 Reviewer #3).  
+
+4. The probe signal for chromosome 21 is very weak. It is too preliminary to conclude that there is no co-localization between chromosome 21 and EPAC1 condensates given the so weak signals.  
+
+As requested by the reviewer we repeated these experiments and obtained the same results. To further substantiate our data, we also used two new probes for the centromeric regions of chromosome 15 and 12 representative images are now presented and quantified in the new figure 6.  
+
+5. Extended data Fig. 1: The authors should examine whether these two domains affect the protein expression of EPAC1. For extended data 1c and 1d, WT EPAC1 should be added as a positive control.  
+
+We agree with the reviewer that the original Western Blotting data presented in Extended figure 1 indicated a significant difference in the expression levels of the two mutants as compared to the WT protein. We repeated the Western Blotting experiments and found that the differences were exacerbated (probably due to exposure differences). In fact, by loading all the samples in the same gel we found that the two mutants are only slightly less expressed than the WT protein. Nevertheless, we believe that this slight difference in expression does not affect the conclusions of our experiments as both mutants are clearly not detectable in the nucleus. In addition, in the new version of Extended figure 1 we added EPAC1WT as control, as requested by the reviewer.  
+
+## Minor points  
+
+1. Fig. 1c: what is time points used for calculation?  
+
+The treatment we used for all experiments presented is between 40 and 60 minutes. 2. Fig. 1d and 1g: The DAPI should be added to indicate the nucleus.
+
+<--- Page Split --->
+
+We added the DAPI labelling of the nucleus in all images except of the experiment presented in Extended figure 6d- e to facilitate the interpretation of the data (colocalization of three fluorescent proteins).  
+
+3. Fig. 5b: the EPAC1 1-148 mutant should be included to show its effect on histone gene expression.  
+
+In Figure 5c are depicted only the statistically significant DEGs in the comparison between cells expressing the EPAC1- 148 mutant and treated with vehicle (DMSO) or 8CPT- cAMP. In  
+
+![PLACEHOLDER_17_0]
+  
+
+this comparison no histone gene varied significantly. As requested by the reviewer we generated the heat map of the expression of the histone genes in the three RNAseq replicates of the EPAC1- 148 mutant and found no significant changes in expression. We enclose here the heat map for the reviewer's information, but we think not necessary to include this figure in the manuscript. If, however, the reviewer thinks it is necessary to include these data in the manuscript we have no objection.  
+
+4. The language of this manuscript needs to be polished. There are a lot of grammar errors and mistakes.  
+
+The manuscript has been carefully revised and formatted according to the format requested by Nature Communications.  
+
+## References  
+
+1. Shiio, Y. & Eisenman, R. N. Histone sumoylation is associated with transcriptional repression. Proc. Natl. Acad. Sci. U. S. A. 100, 13225-13230 (2003).  
+2. Duronio, R. J. & Marzluff, W. F. Coordinating cell cycle-regulated histone gene expression through assembly and function of the Histone Locus Body. RNA Biol. 14, 726 (2017).  
+3. Niimura, M. et al. Critical role of the N-terminal cyclic AMP-binding domain of Epac2 in its subcellular localization and function. J. Cell. Physiol. 219, 652-658 (2009).  
+4. Schwede, F. et al. Structure-guided design of selective Epac1 and Epac2 agonists. PLoS Biol. 13, (2015).  
+5. Zhu, Y., Mei, F., Luo, P. & Cheng, X. A cell-based, quantitative and isoform-specific assay for exchange proteins directly activated by cAMP. Sci. Rep. 7, (2017).  
+6. Gloerich, M. et al. The nucleoporin RanBP2 tethers the cAMP effector Epac1 and inhibits its catalytic activity. J. Cell Biol. 193, 1009-1020 (2011).  
+7. Liu, C. et al. The Interaction of Epac1 and Ran Promotes Rap1 Activation at the
+
+<--- Page Split --->
+
+Nuclear Envelope. Mol. Cell. Biol. 30, 3956 (2010).
+
+<--- Page Split --->
+
+REVIEWERS' COMMENTS  
+
+Reviewer #1 (Remarks to the Author):  
+
+My concerns or suggestions have been taken into account and resolved in the updated version.  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript is improved with the new revisions and I think that the authors claims are better supported now.  
+
+Reviewer #4 (Remarks to the Author):  
+
+The authors have properly answered all my concerns and the manuscript has been improved after revision. No further concerns.  
+
+Reviewer #5 (Remarks to the Author):  
+
+## Overview:  
+
+In this revised manuscript by Iannucci et al, the authors described that phase separation of nuclear EPAC1 - controlled by the second messenger cAMP- is as a bona fide signaling event that impacts transcription and potentially the function of nuclear membraneless organelles. It was found that these nuclear EPAC1- containing condensates can regulate SUMOylation, a crucial process for nuclear function, as well as the expression of a specific histone gene cluster at chromosome 6. Interestingly, the effects of nuclear EPAC1 condensates on transcription were found to be independent of SUMOylation, suggesting that they are regulated by distinct mechanisms. The study suggests that the nuclear cAMP/EPAC1- condensate axis may represent a novel molecular mechanism that could influence the physiology of cells.
+
+<--- Page Split --->
+
+## Comment #1:  
+
+Comment #1:The authors responded to Comment #1 by claiming that investigating the molecular mechanism through which EPAC1 transits the nuclear envelope is beyond the scope of the study due to its inherent complexity. However, the authors substantiated the data by repeating all the western blotting and IF experiments for EPAC1- YFP, \(\Delta 179 - 208\) and \(\Delta 732 - 764\) . The experimental results showed a subtle decrease in nuclear EPAC1 upon deletion of either region, while \(\Delta 732 - 764\) further abolished the colocalization with RANBP2. These findings are further supported by the additional pulldown experiment (Extended Figure 1F), where the interaction between RANBP2 and EPAC1 is observed with the WT and \(\Delta 179 - 208\) , but not with \(\Delta 732 - 764\) . It should be noted, however, that the band quality of the immunoprecipitation is sub- optimal, possibly due to the over- expression of EPAC1- YFP. To address this, an alternative approach involving the pulldown of endogenous RANBP2 and subsequent blotting for EPAC1- YFP could be considered. Notably, in the figure (Extended Figure 1F), there are two lanes labeled EPAC1- YFP, one of which does not exhibit an interaction with RANBP2 (the rightmost lane). The authors should provide further clarification regarding this particular result.  
+
+Also, there is a typo in line 96 (neither should be neither).  
+
+## Comment #2:  
+
+Comment #2:In addressing Comment #2, the authors purified recombinant EPAC1- WT and \(\Delta 2 - 148\) variants, followed by conducting in vitro phase separation assays (Figures 3C- D, Extended Figures 4B- D). The findings revealed that at room temperature conditions, in the presence of cAMP, and at a salt concentration of \(150 \text{mM}\) EPAC1- WT undergoes phase separation. However, EPAC1 \(\Delta 2 - 148\) failed to do so and remained insensitive to changes in cAMP levels. These results demonstrate that the presence of cAMP modulates the ability of EPAC1 to phase separate in vitro, with the N- terminus of EPAC1 playing a crucial role in facilitating this process.  
+
+## Comment #3:  
+
+Comment #3:In response to Comment #3, the authors revised their conclusions by presenting two potential models: either condensate initiation or recruitment to existing condensates. To further support their argument, the authors conducted experiments involving EPAC1 \(\Delta 2 - 24\) and \(\Delta 48 - 148\) variants, which demonstrated their ability to participate in the formation of condensates alongside untagged WT EPAC1 (Extended Figure 5). In contrast, \(\Delta 2 - 148\) exhibited an inability to engage in condensate formation (Extended Figures 5C- D). The authors also successfully eliminated the dominant negative effect by co- expressing both EPA1- YFP WT and \(\Delta 2 - 148\) , wherein EPA1- YFP WT retained its capacity for phase separation in the presence of the mutant. These additional findings, combined with the results obtained from the in vitro phase separation assays, provide additional evidence supporting the assertion that EPAC1 undergoes IDR- dependent phase separation.
+
+<--- Page Split --->
+
+Minor Points:  
+
+Point #1: The authors have supported their claim that HEK cells are EPAC1- deficient at the protein level by showing western blotting data (Figure 1A) and citing existing literature.  
+
+Point #2: The authors have corrected the function of 1,6- hexanediol in the manuscript and showed fusion and FRAP results to support LLPS of EPAC1 (lines 169- 179) (Figure 2).  
+
+Point #3: The authors have modified the figures accordingly (Figure 3A).  
+
+Point #4: The authors have added subheadings to improve the flow of the paper.  
+
+Overall, this reviewer thinks the authors addressed most (not all) of the raised concerns in the original submission.
+
+<--- Page Split --->
+
+Cyclic AMP- induced reversible EPAC1 condensates regulate histone transcription Iannucci et. al.  
+
+## Point-by-point response to the reviewers' comments  
+
+We would like to thank the reviewers for taking the time to review the revisited version of our manuscript. Hereafter a point- by- point response is presented.  
+
+Reviewer #1 (Remarks to the Author):  
+
+My concerns or suggestions have been taken into account and resolved in the updated version.  
+
+Thank you  
+
+Reviewer #2 (Remarks to the Author):  
+
+The manuscript is improved with the new revisions and I think that the authors claims are better supported now.  
+
+Thank you  
+
+Reviewer #4 (Remarks to the Author):  
+
+The authors have properly answered all my concerns and the manuscript has been improved after revision. No further concerns.  
+
+Thank you  
+
+Reviewer #5 (Remarks to the Author):  
+
+Overview:  
+
+In this revised manuscript by Iannucci et al, the authors described that phase separation of nuclear EPAC1- controlled by the second messenger cAMP- is as a bona fide signaling event that impacts transcription and potentially the function of nuclear membraneless organelles. It was found that these nuclear EPAC1- containing condensates can regulate SUMOylation, a crucial process for nuclear function, as well as the expression of a specific histone gene cluster at chromosome 6. Interestingly, the effects of nuclear EPAC1 condensates on transcription were found to be independent of SUMOylation, suggesting that they are regulated by distinct mechanisms. The study suggests that the nuclear cAMP/EPAC1- condensate axis may represent a novel molecular mechanism that could influence the physiology of cells.  
+
+Comment #1:  
+
+The authors responded to Comment #1 by claiming that investigating the molecular mechanism through which EPAC1 transits the nuclear envelope is beyond the scope of the study due to its inherent complexity. However, the authors substantiated the data by
+
+<--- Page Split --->
+
+repeating all the western blotting and IF experiments for EPAC1- YFP, \(\Delta 179 - 208\) and \(\Delta 732 - 764\) . The experimental results showed a subtle decrease in nuclear EPAC1 upon deletion of either region, while \(\Delta 732 - 764\) further abolished the co- localization with RANBP2. These findings are further supported by the additional pulldown experiment (Extended Figure 1F), where the interaction between RANBP2 and EPAC1 is observed with the WT and \(\Delta 179 - 208\) , but not with \(\Delta 732 - 764\) . It should be noted, however, that the band quality of the immunoprecipitation is sub- optimal, possibly due to the over- expression of EPAC1- YFP. To address this, an alternative approach involving the pulldown of endogenous RANBP2 and subsequent blotting for EPAC1- YFP could be considered. Notably, in the figure (Extended Figure 1F), there are two lanes labeled EPAC1- YFP, one of which does not exhibit an interaction with RANBP2 (the rightmost lane). The authors should provide further clarification regarding this particular result.  
+
+We agree with the reviewer that the RANBP2 bands in the Western Blotting experiment presented in supplementary figure 1F was sub- optimal. Unfortunately, RANBP2 is a large protein of approximately 350kD and the quality of the available antibodies is not optimal. To repeat the experiments would require the acquisition of new antibodies which will take more than the 2- weeks provided by the editor for the final revisions of the manuscript. Moreover, we reasoned that supplementary figure 1f was used to confirm that the C- terminus of EPAC1 contains the sequences necessary for engaging RANBP2. However, this is well accepted and has been demonstrated previously. For these reasons, in the final version of the manuscript we decide NOT to include the panel f of Supplementary figure 1. We therefore removed the description of this panel from the main manuscript and added the references demonstrating that EPAC1 can interact with RANBP2.  
+
+Also, there is a typo in line 96 (neither should be neither).  
+
+## Corrected  
+
+## Comment #2:  
+
+In addressing Comment #2, the authors purified recombinant EPAC1- WT and \(\Delta 2 - 148\) variants, followed by conducting in vitro phase separation assays (Figures 3C- D, Extended Figures 4B- D). The findings revealed that at room temperature conditions, in the presence of cAMP, and at a salt concentration of 150 mM EPAC1- WT undergoes phase separation. However, EPAC1 \(\Delta 2 - 148\) failed to do so and remained insensitive to changes in cAMP levels. These results demonstrate that the presence of cAMP modulates the ability of EPAC1 to phase separate in vitro, with the N- terminus of EPAC1 playing a crucial role in facilitating this process.  
+
+## Comment #3:  
+
+In response to Comment #3, the authors revised their conclusions by presenting two potential models: either condensate initiation or recruitment to existing condensates. To further support their argument, the authors conducted experiments involving EPAC1 \(\Delta 2 - 24\) and \(\Delta 48 - 148\) variants, which demonstrated their ability to participate in the formation of condensates alongside untagged WT EPAC1 (Extended Figure 5). In contrast, \(\Delta 2 - 148\) exhibited an inability to engage in condensate formation (Extended Figures 5C- D). The authors also successfully eliminated the dominant negative effect by co- expressing both EPA1- YFP WT and \(\Delta 2 - 148\) , wherein EPA1- YFP WT retained its capacity for phase separation in the presence of the mutant. These additional findings, combined with the results obtained from the in vitro phase
+
+<--- Page Split --->
+
+separation assays, provide additional evidence supporting the assertion that EPAC1 undergoes IDR- dependent phase separation.  
+
+Minor Points:  
+
+Point #1: The authors have supported their claim that HEK cells are EPAC1- deficient at the protein level by showing western blotting data (Figure 1A) and citing existing literature.  
+
+Point #2: The authors have corrected the function of 1,6- hexanediol in the manuscript and showed fusion and FRAP results to support LLPS of EPAC1 (lines 169- 179) (Figure 2).  
+
+Point #3: The authors have modified the figures accordingly (Figure 3A).  
+
+Point #4: The authors have added subheadings to improve the flow of the paper.  
+
+Overall, this reviewer thinks the authors addressed most (not all) of the raised concerns in the original submission.
+
+<--- Page Split --->