Reviewer #3 (Public review):

Summary:

This study investigates how Rhino, a chromatin-associated HP1-family protein essential for germline piRNA biogenesis in Drosophila, is initially recruited to specific genomic loci. Although canonical dual-strand piRNA clusters such as 42AB, 38C, 80F, and 102F produce the majority of germline piRNAs, the mechanisms guiding Rhino to these regions remain poorly understood. To explore the earliest steps of Rhino loading, the authors use a doxycycline-inducible Rhino transgene in OSC cells, a system that expresses only the primary Piwi pathway and therefore provides an experimentally accessible, epigenetically naïve context distinct from the endogenous germline environment. Through a combination of inducible Rhino expression, knockdown of selected Drosophila PRMTs (DARTs), ChIP-seq, small RNA sequencing, and imaging, the authors propose that asymmetric arginine-methylated histones, particularly those deposited by DART4, contribute to defining initial sites of Rhino association. They identify a subset of Rhino-bound loci, termed DART4-dependent piRNA source loci (piSL), which lose Rhino, Kipferl, and piRNA production upon DART4 depletion and may represent nascent or transitional piRNA clusters. Overall, the study provides intriguing evidence for a link between ADMA histone marks and de novo Rhino recruitment, particularly in the simplified OSC context, and offers new candidate loci for further exploration of early piRNA-cluster chromatin dynamics.

Strengths:

This study offers important insights into how asymmetric dimethylarginine (ADMA) histone marks contribute to the initial recruitment of Rhino, a Drosophila HP1-family protein essential for dual-strand piRNA cluster specification. Using an integrative approach that includes ectopic expression of a Rhino transgene in OSC cells, germline knockdown of DART4 in Drosophila ovaries, ChIP-seq, small RNA-seq, and imaging, the authors show that ADMA marks particularly H3R17me2a and H4R3me2acorrelate with Rhino binding at the boundaries of canonical piRNA clusters and at DART4-dependent piRNA source loci (piSL). These piSL may represent nascent or transitional piRNA-generating regions. Overall, the dataset presented here provides a valuable resource for understanding the chromatin features associated with the emergence and maturation of piRNA clusters.

Weaknesses:

Despite the strengths of the study, several important limitations remain. Although Rhino binding correlates with ADMA-enriched boundaries, the data do not directly demonstrate that these histone marks are required for Rhino spreading, leaving the mechanistic relationship correlative rather than causal. The DART4-dependent piRNA source loci identified here produce only low levels of piRNAs, and their functional contribution remains uncertain. In addition, redundancy among DART family methyltransferases remains unresolved: only DART4 was tested in the germline, and effective knockdown of DART1 or other DARTs could not be achieved, limiting the ability to evaluate whether ADMA-histones more broadly regulate Rhino recruitment at canonical clusters. Consequently, the current dataset primarily supports DART4-dependent effects at a small subset of evolutionarily young loci, and both the model and the title may overstate the generality of this mechanism across the full repertoire of dual-strand piRNA clusters.

In conclusion, this study is carefully executed and puts forward compelling hypotheses regarding the early chromatin environment that may underlie piRNA cluster formation. The findings will be relevant to researchers interested in genome regulation, small RNA biology, and chromatin-mediated transposon control.

Author response:

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

Reviewer #1(Public review):

Summary:

In this study, the authors aim to understand how Rhino, a chromatin protein essential for small RNA production in fruit flies, is initially recruited to specific regions of the genome. They propose that asymmetric arginine methylation of histones, particularly mediated by the enzyme DART4, plays a key role in defining the first genomic sites of Rhino localization. Using a combination of inducible expression systems, chromatin immunoprecipitation, and genetic knockdowns, the authors identify a new class of Rhinobound loci, termed DART4 clusters, that may represent nascent or transitional piRNA clusters.

Strengths:

One of the main strengths of this work lies in its comprehensive use of genomic data to reveal a correlation between ADMA histones and Rhino enrichment at the border of known piRNA clusters. The use of both cultured cells and ovaries adds robustness to this observation. The knockdown of DART4 supports a role for H3R17me2a in shaping Rhino binding at a subset of genomic regions.

Weaknesses:

However, Rhino binding at, and piRNA production from, canonical piRNA clusters appears largely unaffected by DART4 depletion, and spreading of Rhino from ADMArich boundaries was not directly demonstrated. Therefore, while the correlation is clearly documented, further investigation would be needed to determine the functional requirement of these histone marks in piRNA cluster specification.

The study identify piRNA cluster-like regions called DART4 clusters. While the model proposes that DART4 clusters represent evolutionary precursors of mature piRNA clusters, the functional output of these clusters remains limited. Additional experiments could help clarify whether low-level piRNA production from these loci is sufficient to guide Piwi-dependent silencing.

In summary, the authors present a well-executed study that raises intriguing hypotheses about the early chromatin context of piRNA cluster formation. The work will be of interest to researchers studying genome regulation, small RNA pathways, and the chromatin mechanisms of transposon control. It provides useful resources and new candidate loci for follow-up studies, while also highlighting the need for further functional validation to fully support the proposed model.

We sincerely thank Reviewer #1 for the thoughtful and constructive summary of our work. We appreciate the reviewer’s recognition that our study provides a comprehensive analysis of the relationship between ADMA-histones and Rhino localization, and that it raises intriguing hypotheses about the early chromatin context of piRNA cluster formation.

We fully agree with the reviewer that our data primarily demonstrate correlation between ADMA-histones and Rhino localization, rather than direct causation. In response, we have carefully revised the text throughout the manuscript to avoid overstatements implying causality (details provided below).

We also acknowledge the reviewer’s important point that the functional requirement of ADMA-histones for piRNA clusters specification remains to be further established. We have now added the discussion about our experimental limitations (page 18).

Overall, we have revised the manuscript to present our findings more cautiously and transparently, emphasizing that our data reveal a correlation between ADMA-histone marks and the initial localization of Rhino, rather than proving a direct mechanistic requirement. We thank the reviewer again for highlighting these important distinctions.

Reviewer #2 (Public review):

This study seeks to understand how the Rhino factor knows how to localize to specific transposon loci and to specific piRNA clusters to direct the correct formation of specialized heterochromatin that promotes piRNA biogenesis in the fly germline. In particular, these dual-strand piRNA clusters with names like 42AB, 38C, 80F, and 102F generate the bulk of ovarian piRNAs in the nurse cells of the fly ovary, but the evolutionary significance of these dual-strand piRNA clusters remains mysterious since triple null mutants of these dual-strand piRNA clusters still allows fly ovaries to develop and remain fertile. Nevertheless, mutants of Rhino and its interactors Deadlock, Cutoff, Kipferl and Moonshiner, etc, causes more piRNA loss beyond these dual-strand clusters and exhibit the phenotype of major female infertility, so the impact of proper assembly of Rhino, the RDC, Kipferl etc onto proper piRNA chromatin is an important and interesting biological question that is not fully understood.

This study tries to first test ectopic expression of Rhino via engineering a Dox-inducible Rhino transgene in the OSC line that only expresses the primary Piwi pathway that reflects the natural single pathway expression the follicle cells and is quite distinct from the nurse cell germline piRNA pathway that is promoted by Rhino, Moonshiner, etc. The authors present some compelling evidence that this ectopic Rhino expression in OSCs may reveal how Rhino can initiate de novo binding via ADMA histone marks, a feat that would be much more challenging to demonstrate in the germline where this epigenetic naïve state cannot be modeled since germ cell collapse would likely ensue. In the OSC, the authors have tested the knockdown of four of the 11 known Drosophila PRMTs (DARTs), and comparing to ectopic Rhino foci that they observe in HP1a knockdown (KD), they conclude DART1 and DART4 are the prime factors to study further in looking for disruption of ADMA histone marks. The authors also test KD of DART8 and CG17726 in OSCs, but in the fly, the authors only test Germ Line KD of DART4 only, they do not explain why these other DARTs are not tested in GLKD, the UAS-RNAi resources in Drosophila strain repositories should be very complete and have reagents for these knockdowns to be accessible.

The authors only characterize some particular ADMA marks of H3R17me2a as showing strong decrease after DART4 GLKD, and then they see some small subset of piRNA clusters go down in piRNA production as shown in Figure 6B and Figure 6F and Supplementary Figure 7. This small subset of DART4-dependent piRNA clusters does lose Rhino and Kipferl recruitment, which is an interesting result.

However, the biggest issue with this study is the mystery that the set of the most prominent dual-strand piRNA clusters. 42AB, 38C, 80F, and 102F, are the prime genomic loci subjected to Rhino regulation, and they do not show any change in piRNA production in the GLKD of DART4. The authors bury this surprising negative result in Supplementary Figure 5E, but this is also evident in no decrease (actually an n.s. increase) in Rhino association in Figure 5D. Since these main piRNA clusters involve the RDC, Kipferl, Moonshiner, etc, and it does not change in ADMA status and piRNA loss after DART4 GLKD, this poses a problem with the model in Figure 7C. In this study, there is only a GLKD of DART4 and no GLKD of the other DARTs in fly ovaries.

One way the authors rationalize this peculiar exception is the argument that DART4 is only acting on evolutionarily "young" piRNA clusters like the bx, CG14629, and CG31612, but the lack of any change on the majority of other piRNA clusters in Figure 6F leaves upon the unsatisfying concern that there is much functional redundancy remaining with other DARTs not being tested by GLKD in the fly that would have a bigger impact on the other main dual-strand piRNA clusters being regulated by Rhino and ADMA-histone marks.

Also, the current data does not provide convincing enough support for the model Figure 7C and the paper title of ADMA-histones being the key determinant in the fly ovary for Rhino recognition of the dual-strand piRNA clusters. Although much of this study's data is well constructed and presented, there remains a large gap that no other DARTs were tested in GLKD that would show a big loss of piRNAs from the main dual-strand piRNA clusters of 42AB, 38C, 80F, and 102F, where Rhino has prominent spreading in these regions.

As the manuscript currently stands, I do not think the authors present enough data to conclude that "ADMA-histones [As a Major new histone mark class] does play a crucial role in the initial recognition of dual-strand piRNA cluster regions by Rhino" because the data here mainly just show a small subset of evolutionarily young piRNA clusters have a strong effect from GLKD of DART4. The authors could extensively revise the study to be much more specific in the title and conclusion that they have uncovered this very unique niche of a small subset of DART4-dependent piRNA clusters, but this niche finding may dampen the impact and significance of this study since other major dual-strand piRNA clusters do not change during DART4 GLKD, and the authors do not show data GLKD of any other DARTs. The niche finding of just a small subset of DART-4-dependent piRNA clusters might make another specialized genetics forum a more appropriate venue.

We are deeply grateful to Reviewer #2 for the detailed and insightful review that carefully situates our study in the broader context of Rhino-mediated piRNA cluster regulation. We appreciate the reviewer’s recognition that our inducible Rhino expression system in OSCs provides a valuable model to explore de novo Rhino recruitment under a simplified chromatin environment.

At the same time, we agree that the current data mainly support a role for DART4 in regulating a subset of evolutionarily young piRNA clusters, and do not demonstrate a requirement for ADMA-histones at the major dual-strand piRNA clusters such as 42AB or 38C. We have therefore revised the title and main conclusions to more accurately reflect the scope of our findings.

We agree with the reviewer that functional redundancy among DARTs may explain why major dual-strand piRNA clusters are unaffected by DART4 GLKD. Indeed, we have tried DART1 GLKD in the germline, which shows collapse of Rhino foci in OSCs.For DART1 GLKD, two approaches were possible:

(1) Crossing the BDSC UAS-RNAi line (ID: 36891) with nos-GAL4.

(2) Crossing the VDRC UAS-RNAi line (ID: 110391) with nos-GAL4 and UAS-Dcr2.

The first approach was not feasible because the UAS-RNAi line always arrived as dead on arrival (DOA) and could not be maintained in our laboratory. The second approach did not yield effective and stable knockdown (as follows).

DART8 and CG17726 did not alter Rhino foci in OSC knockdown experiments; therefore, we did not attempt germline knockdown (GLKD) of these DARTs in the ovary.  We agree with the reviewer’s opinion that there are piRNA source loci where Rhino localization depends on DART1, and that simultaneous depletion of multiple DARTs may indeed reveal additional positive results because ADMA-histones such as H3R8me2a may be completely eliminated by the knockdown of multiple DARTs. At the same time, we note that many evolutionarily conserved piRNA clusters show a loss of ADMA accumulation compared with evolutionarily young piRNA clusters, with levels that are comparable to the background input in ChIP-seq reads. Therefore, conserved clusters such as 42AB and 38C may no longer be regulated by ADMA. Even if multiple DARTs function redundantly to regulate ADMA, it may be difficult to disrupt Rhino localization at such conserved piRNA clusters by depletion of DARTs. While disruption of Rhino localization at conserved clusters like 42AB and 38C may be challenging, we cannot exclude the possibility that DART depletion affects Rhino binding at less conserved piRNA clusters, where ADMA modification remains detectable. We added clarifications in the Discussion to acknowledge the potential redundancy with other DARTs and to note that further knockdown experiments in the germline will be necessary to test this model comprehensively (page 18).

We appreciate the reviewer’s critical feedback, which has helped us refine the message and strengthen the interpretative balance of the paper.

Reviewer #1 (Recommendations for the authors):

In multiple places, the link between ADMA histones and Rhino recruitment is presented in terms that imply causality. Please revise these statements to reflect that, in most cases, the evidence supports correlation rather than direct functional necessity. Similarly, statements suggesting that ADMA histones promote Rhino spreading should be revised unless supported by direct evidence.

We sincerely thank the reviewer for the insightful comments. We recognize that these suggestions are crucial for improving the manuscript, and we have revised it accordingly to address the concerns. The specific revisions we made are detailed below.

(1) Page 1, line 14: The original sentence “in establishing the sites” was changed to “may establish the potential sites.”

(2) Page 4, lines 11-12: The original sentence “genomic regions where Rhino binds at the ends and propagates in the areas in a DART4-dependent manner, but not stably anchored” was changed to “genomic regions that have ADMA-histones at their ends and exhibit broad Rhino spreading across their internal regions in a DART4dependent manner”

(3) Page4, lines 12-15: The original sentence “Kipferl is present at the regions but not sufficient to stabilize Rhino-genomic binding after Rhino propagates.” was changed to “In contrast to authentic piRNA clusters, Kipferl was lost together with Rhino upon DART4 depletion in these regions, suggesting that Kipferl by itself is not sufficient to stabilize Rhino binding; rather, their localization depends on DART4.”

(4) Page4, lines17-18: The original sentence “are considered to be primitive clusters” was changed to “might be nascent dual-strand piRNA source loci”.

(5) Page 8, line 7: The original sentence “Involvement of ADMA-histones in the genomic localization of Rhino was implicated.” was changed to “Correlation of ADMA-histones in the genomic localization of Rhino was implicated.”

(6) Page 8, lines 19-21: The original sentence “These results suggest that ADMAhistones, together with H3K9me3, contribute significantly and specifically to the recruitment of Rhino to the ends of dual-strand clusters in OSCs.” was changed to “These results raise the possibility that ADMA-histones, together with H3K9me3, may contribute specifically to the recruitment of Rhino to the ends of dual-strand clusters in OSCs.”

(7) Page 10, lines 11-13: The original sentence “These results suggest that DART1 and DART4 are involved in Rhino recruitment at distinct genomic sites through the decreases in ADMA-histones in each of their KD conditions (H4R3me2a and H3R17me2a, respectively).” was changed to ”These results suggest that DART1 and DART4 could contribute to Rhino recruitment at distinct genomic sites through the decreases in ADMA-histones in each of their KD conditions (H4R3me2a and H3R17me2a, respectively).”

(8) Page 13, line 2: The original sentence “Genomic regions where Rhino spreads in a DART4-dependent manner, but not stably anchored, produce some piRNAs“ was changed to “Genomic regions where Rhino binds broadly in a DART4-dependent manner, but not stably anchored, produce some piRNAs”

(9) Page 13, lines 21-22: The original sentence “These results support the hypothesis that ADMA-histones are involved in the genomic binding of Rhino both before and after Rhino spreading, resulting in stable genome binding.” was changed to “These results raise the possibility that a subset of Rhino localized to genomic regions correlating with ADMA-histones may serve as origins of spreading.”

(10) Page 16, lines 6-8: The original sentence “In this study, we took advantage of cultured OSCs for our analysis and found that chromatin marks (i.e., ADMA-histones) play a crucial role in the loading of Rhino onto the genome.” was changed to “In this study, we took advantage of cultured OSCs for our analysis and found that chromatin marks (i.e., bivalent nucleosomes containing H3K9me3 and ADMA-histones) appear to contribute to the initial loading of Rhino onto the genome.”

(11) Page16, line 12: The original sentence “We propose that the process of piRNA cluster formation begins with the initial loading of Rhino onto bivalent nucleosomes containing H3K9me3 and ADMA-histones (Fig. 7C). In OSCs, the absence of Kipferl and other necessary factors means that Rhino loading into the genome does not proceed to the next step.” was removed.

Major points

(1)  Clarify the limited colocalization between Rhino and H3K9me3 in OSCs. The observation that FLAG-Rhino foci show minimal overlap with H3K9me3 in OSCs appears inconsistent with the proposed model by the authors in the discussion, in which Rhino is initially recruited to bivalent nucleosomes bearing both H3K9me3 and ADMA marks. This discrepancy should be addressed. 

We thank the reviewer’s insightful comments. Indeed, ChIP-seq shows that Rhino partially overlaps with H3K9me3 (Fig. 1F), but immunofluorescence did not reveal any detectable overlap (Fig. 1A). We interpret this discrepancy as arising from the fact that immunofluorescence primarily visualizes H3K9me3 foci that are localized as broad domains in the genome, such as those at centromeres, pericentromeres, or telomeres (named chromocenters), whereas the sharp and interspersed H3K9me3 signals along chromosome arms are difficult to detect by immunofluorescence. We now have these explanations in the revised text (page 6).

(2)  Please indicate whether the FLAG-Rhino used in OSCs has been tested for functionality in vivo-for example, by rescuing Rhino mutant phenotypes. This is particularly relevant given that no spreading is observed with this construct.

We thank the reviewer for raising this important point. We have not directly tested the functionality of FLAG-Rhino construct used in OSCs in living Drosophila fly; i.e., it has not been used to rescue Rhino mutant phenotypes in flies. We acknowledge that FLAGRhino has not previously been expressed in OSCs, and that its localization pattern in OSCs differs from that observed in ovaries, where Rhino is endogenously expressed. However, several lines of evidence suggest that the addition of the N-terminal FLAG tag is unlikely to compromise Rhino function

(1) In previous studies, N-terminally tagged Rhino (e.g., 3xFLAG-V5-Precision-GFPRhino) was expressed in a living Drosophila ovary and was shown to localize properly to piRNA clusters, indicating that the tag does not prevent Rhino from binding its genomic targets (Baumgartner et al., 2022; eLife. Fig. 3 supplement 1G).

(2) In Drosophila S2 cells, FLAG-tagged tandem Rhino chromodomains construct was shown to bind H3K9me3/H3K27me3 bivalent chromatin, demonstrating that the FLAG tag does not impair this fundamental chromatin interaction (Akkouche et al., 2025; Nat Struct Mol Biol. Fig. 4b).

(3) GFP-tagged Rhino has been demonstrated to rescue the transposon derepression phenotype of Rhino mutant flies, further supporting that the addition of tags does not abolish its in vivo function. (Parhad et al., 2017; Dev Cell. Fig.1D).

Therefore, we interpret the partial localization of FLAG-Rhino in OSCs as reflecting the specific chromatin environment and regulatory context of OSCs rather than functional impairment due to the FLAG tag.

(3) Given the low levels of piRNA production and the absence of measurable effects on transposon expression or fertility upon DART4 knockdown, the rationale for classifying these regions as piRNA clusters should be clearly stated. Additional experiments could help clarify whether low-level piRNA production from these loci is sufficient to guide Piwidependent silencing. The authors should also consider and discuss the possibility that some of these differences may reflect background-specific genomic variation rather than DART4-dependent regulation per see.

We thank the reviewer for the insightful comments. As noted, DART4 knockdown did not measurably affect transposon expression or fertility. piRNAs generated from DART4associated clusters associate with Piwi but are insufficient for target repression. Although loss of DART4 largely eliminated piRNAs from these clusters, the cluster-derived transcripts themselves were unchanged. To clarify this point, we now refer to these regions as DART4-dependent piRNA-source loci (DART4 piSLs) in the revised text. We also acknowledge that some observed differences may reflect strain-specific genomic variation and have added this caveat on page 16.

(4)  The authors should describe the genomic context of DART4 clusters in more detail. Specifically, it would be helpful to indicate whether these regions overlap with known transposable elements, gene bodies, or intergenic regions, and to report the typical size range of the clusters. Are any of the piRNAs produced from these clusters predicted to target known transcripts? 

We thank the reviewer’s insightful comments. The overlap of DART4 piSL with transposable elements, gene bodies, and intergenic regions is shown in the right panel of Supplementary Fig. 6E (denoted as “Rhino reduced regions in DART4 GLKD” in the figure). The typical size range of these clusters is presented in Supplementary Fig. 6G. The annotation of piRNA reads derived from these piSL is shown in the right panel of Supplementary Fig. 6F, indicating that most of them appear to target host genes. The specific genes and transposons matched by the piRNAs produced from DART4 piSL are listed in Supplementary Table 8.

(5)  While correlations between Rhino and ADMA histone marks (especially H3R8me2a,H3R17me2a, H4R3me2a) are robust, many ADMA-enriched regions do not recruit Rhino. Please discuss this observation and consider the possible involvement of additional factors.

We thank the reviewer’s insightful comments. As pointed out, not all ADMA-enriched regions recruit Rhino; rather, Rhino is recruited only at sites where ADMAs overlap with H3K9me3. Furthermore, the combination of H3K9me3 and ADMAs alone does not fully account for the specificity of Rhino recruitment, suggesting the involvement of additional co-factors (for example, other ADMA marks such as H3R42me2a, or chromatininteracting proteins). In addition, since histone modifications—including arginine methylation—have the possibility that they are secondary consequences of modifications on other proteins rather than primary regulatory events, it is possible that DART1/4 contribute to Rhino recruitment not only through histone methylation but also via arginine methylation of non-histone chromatin-interacting factors. However, methylation of HP1a does not appear to be involved (Supplementary Fig. 3G). We have added new sentences about these points in the Discussion section (page 18).

(6) The manuscript states that Kipferl is present at DART4 clusters but does not stabilize Rhino binding. Please specify which experimental results support this conclusion and explain.

We apologize for the lack of clarity regarding Kipferl data. Supplementary Fig. 7A and 7B show that Kipferl localizes at major DART4 piSL. This Kipferl localization is lost together with Rhino upon DART4 GLKD, indicating that Rhino localization at DART4 piSL depends on DART4 rather than on Kipferl. From these results, we infer that, unlike at authentic piRNA clusters, Kipferl may not be sufficient to stabilize the association of Rhino with the genome at DART4 piSL. We have added this interpretation on page 14.

Minor points

(1) Figure 1D: Please specify which piRNA clusters are included in the metaplot - all clusters, or only the major producers? 

We thank the reviewer for the question. The metaplot was not generated from a predefined list of “all” piRNA clusters or only the “major producers.” Instead, it was constructed from Rhino ChIP–seq peaks (“Rhino domains”) that are ≥1.5 kb in length.These Rhino domains mainly correspond to the subregions within major dual-strand clusters (e.g., 42AB, 38C) as well as additional clusters such as 80F, 102F, and eyeless, among others. We have provided the full list of domains and their corresponding piRNA clusters (with genomic coordinates) in Supplementary Table 9 and added the additional explanation in Fig. 1d legend.

(2) Supplemental Figure 5E is referred to as 5D in the main text.

We corrected the figure citations on pages 11-12: the reference to Supplementary Fig. 5E has been changed to 5D, and the reference to Supplementary Fig. 5F has been changed to 5E.

(3) Supplemental Figure 7C: The color legend does not match the pie chart, which may confuse readers.

We thank the reviewer for the helpful comment. We are afraid we were not entirely sure what specific aspect of the legend was confusing, but to avoid any possible misunderstanding, we revised Supplemental Fig. 7C so that the color boxes in the legend now exactly match the corresponding colors in the pie chart. We hope this modification improves clarity.

(4) Since the manuscript focuses on the roles of DART1 and DART4, including their expression profiles in OSCs and ovaries would help contextualize the observed phenotypes. Please consider adding this information if available.

We thank the reviewer for the suggestion. We have now included a scatter plot comparing RNA-seq expression in OSCs and ovaries (Supplementary Fig. 3H). In these datasets, DART1 is strongly expressed in both tissues, whereas DART4 shows no detectable reads. Notably, ref. 28 reports strong expression of both DART1 and DART4 in ovaries by western blot and northern blot. In our own qPCR analysis in OSCs, DART4 expression is about 3% of DART1, which, although low, may still be sufficient for functional roles such as modification of H3R17me2a (Fig. 3C, Supplementary Fig. 3F and 3I). We have added these new data and additional explanation in the revised manuscript (page 11).

(5) Several of the genome browser snapshots, particularly scale and genome coordinates, are difficult to read. 

We apologize for the difficulty in reading several of the genome browser snapshots in the original submission. We have re-generated the relevant figures using IGV, which provides clearer visualization of scale and genome coordinates. The previous images have been replaced with the improved versions in the revised manuscript.

Reviewer #2 (Recommendations for the authors):

(1) The authors need to elaborate on what this sentence means, as it is very unclear what they are describing about Rhino residency: "The results show that Rhino in OSCs tends to reside in the genome where Rhino binds locally in the ovary (Fig. 1C)." 

We apologize for the lack of clarity in the original sentence. The text has been revised as follows:

”Rhino expressed in OSCs bound predominantly to genomic sites exhibiting sharp and interspersed Rhino localization patterns in the ovary, while showing little localization within broad Rhino domains, including major piRNA clusters.”

In addition, to clarify the behavior of Rhino at broad domains, we have added the phrase “the terminal regions of broad domains, such as major piRNA clusters” to the subsequent sentence.

(2) The red correlation line is very confusing in Figure 5F. What sort of line does this mean in this scatter plot? 

We apologize for the lack of clarity regarding the red line in Fig. 5F. The red line represents the least-squares linear regression fit to the data points, calculated using the lm() function in R, and was added with abline() to illustrate the correlation between ctrl GLKD and DART4 GLKD values. In the revised figure, we have clarified this in the legend by specifying that it is a regression line.

(3) There is no confirmation of the successful knockdown of the various DARTs in the OSCs.

We thank the reviewer for the comment. The knockdown efficiency of the various DARTs in OSCs was confirmed by RT–qPCR. The data are now shown in Supplementary Fig. 3J. 

(4) What is the purpose of an unnumbered "Method Figure" in the supplementary data file? Why not just give it a number and mention it properly in the text? 

We thank the reviewer for the suggestion. We have now assigned a number to the previously unnumbered "Method Figure" and have included it as Supplementary Fig. 9.

The figure is now properly cited in the Methods section.

(5) For Figure 5A, those fly strain numbers in the labels are better reserved in the Methods, and a more appropriate label is to describe the GAL4 driver and the UAS-RNAi construct by their conventional names.

We thank the reviewer for the suggestion. The labels in Fig. 5A have been updated to use the conventional names of the GAL4 drivers and UAS-RNAi constructs. Specifically, they now read Ctrl GLKD (nos-GAL4 > UAS-emp) and DART4 GLKD (nos-GAL4 > UASDART4). The original fly strain numbers are listed in the Methods section.

The reverse halo effect shows the negative judgement due to individuals’ appearance can distort other perceptions of them, leading to unfair treatment. It may lead to some tragedies or pities in the broader society.

How can we train ourselves to correct the halo effect when making critical decisions?

Why does society allow more economic gain for the ones with better appearance, isn't this related one' ability to work? Can it be changed or it is too deeply rooted to change?

The halo effect is not always positive. While attractive people are often idealized, they can also be stereotyped as manipulative or morally untrustworthy. These opposite views based on the appearance of individuals are quite shallow sometimes.

This directly connects to The Wretched and the Beautiful, where those seen as “beautiful” are trusted , respected , and valued while the “wretched” are dehumanized. The story exaggerates the halo effect to critique how beauty becomes a justification for cruelty.

Halo effect is interesting. It shows how quickly humans make judgments based on appearance rather than characterization or actions.

This is interesting because the halo effect was not only expressed in literature but also deliberately used in other areas like art.

Reality is actually different from appearance but it doesn't matter that much, it's all about what is seen instead of what is really done.

This is interesting because appearance can bring tangible benefits like more money as well as how appearance affects the thoughts and judgements.

Why are teachers paying less attention to students' appearance in online courses?

This is what happened in the text. The second group of aliens changed their appearance to appear as attractive and manipulated humans on Earth to make it easier for them to achieve their goal. However, humans in the text believed in their words without doubt.

What specific characteristics are correlated with each other?

What specific characteristics are correlated with each other?

This is interesting, I wonder if this happens around me. This also connects to the. other article about 'civilised' group of people, in social media, people doesn't always have an complete view on other communities, so stereotypes eventually appears. Or maybe they have only encountered negative things about another group of people.

This is interesting as it also connect with the two groups of aliens. While the first group of alien is harmless and shows peace, the second group of aliens aim to manipulate people using their appearance. Q: Can we assume that people often use appearance as tool for controlling other's opinion and use it as a way to overturn public opinions?

This is interesting because it mean that in most of the companies, its employees or workers aren’t always having deserved position based on their actual power or knowledge. Therefore, even when they believe that they have competence, the reality isn't always true.

Does this mean that most of the explanation from people about their racial judgements, exclusion, and superiority is mostly based on people's looks or simple characteristics? (e.g. "The Wretched and the Beautiful", The White Man's Burden)

The halo effect has mutually influenced people

it's undeniable that people's decision was largely influenced by one's appearance.

It is the same phenomenon that appeared in "The Wretched And Ths Beautiful", which people treated differently with two groups of aliens just based on their appearance.

eLife Assessment

This useful study presents the potentially interesting idea that LRRK2 regulates cellular BMP levels and their release via extracellular vesicles, with GCase activity further modulating this process in mutant LRRK2-expressing cells. However, some of the evidence supporting these conclusions remains incomplete, and additional work is suggested under certain conditions. Overall, the study will be of interest to cell biologists working on Parkinson's disease.

Reviewer #1 (Public review):

Summary:

Even though mutations in LRRK2 and GBA1 (which encodes the protein GCase) increase the risk of developing Parkinson's disease (PD), the specific mechanisms driving neurodegeneration remain unclear. Given their known roles in lysosomal function, the authors investigate how LRRK2 and GCase activity influence the exocytosis of the lysosomal lipid BMP via extracellular vesicles (EVs). They use fibroblasts carrying the PD-associated LRRK2-R1441G mutation and pharmacologically modulate LRRK2 and GCase activity.

Strengths:

The authors examine both proteins at endogenous levels, using MEFs instead of cancer cells. The study's scope is potentially interesting and could yield relevant insights into PD disease mechanisms.

Weaknesses:

Many of the authors' conclusions are overstated and not sufficiently supported by the data. Several statistical errors undermine their claims. Pharmacological treatment is very long, leading to potential off target effects. Additionally, the authors should be more rigorous when using EV markers.

Comments on revisions:

The authors have not addressed most of my concerns. For example, instead of trying with a 1-2 hour MLi2 treatment, they cited all the papers that use extremely long time points for LRRK2 inhibition; the fact that other groups do it does not mean it is biologically correct. They also refused to quantify their western blots in a proper manner, without the "hyper-normalization" claiming that it is an accepted way to quantify western blots. Again, it is statistically incorrect and biologically impossible. They also do not have a satisfactory explanation as to why the R1441G cells (which increase LRRK2 kinase activity) have no effect on EV release, but they still claim it is LRRK2 kinase activity dependent.

Overall, I am very confused by the model proposed by the authors. They only see increased EV release in the G2019S expressing cells, but not the R1441G cells, yet they claim that the increase of EV release is LRRK2 kinase activity dependent. Then, they claim that the presence of BMP (unchanged in R1441G vs CTL) in EVs is also LRRK2 kinase activity dependent. Finally, they perform TIRF with pHluorin-CD63 construct and observed an increase in G2019S cells vs CTL "further confirming that BMP release is associated with EV secretion". First, I could not see the increase in BMP release in G2019S cells (if I missed it, I apologize). And second, why didn't they do this experiment in R1441G cells? As, the R1441G cells have not displayed an increase in EV release compared to CTL cells, it could also be possible that the BMP release might be more abundant through lysosomal exocytosis (which could explain the pHluorin results) than EVs. Overall, the authors nicely demonstrate that the R1441G cells have more BMP species, likely due to increase CLN5 expression, but the release of the BMP is still not clear to this reviewer.

Reviewer #2 (Public review):

Summary:

In this paper, authors used MEFs expressing the R1441G mutant of leucine-rich repeat kinase 2 (LRRK2), a mutant associated with the early onset of Parkinson's disease. They report that in these cells LAMP2 fluorescence is higher but BMP fluorescence is lower, MVE size is reduced and that MVEs contain less ILVs. They also report that LAMP2-positive EVs are increased in mutant cells in a process sensitive to LRRK2 kinase inhibition but are further increased by glucocerebrosidase (GCase) inhibition, and that total di-22:6-BMP and total di-18:1-BMP are increased in mutant LRRK2 MEFs compared to WT cells by mass spectrometry. They also report that LRRK2 kinase inhibition partially restores cellular BMP levels, and that GCase inhibition further increased BMP levels, and that in EVs from the LRRK2 mutant, LRRK2 inhibition decreases BMP while GCase inhibition has the opposite effect. Moreover, they report that BMP increase is not due to increased BMP synthesis, although authors observe that CLN5 is increased in LRRK2 mutant cells. Finally, they report that GW4869 decreases EV release and exosomal BMP, while bafilomycin A1 increases EV release. They conclude that LRRK2 regulates BMP levels (in cells) and release (via EVs). They also conclude that the process is modulated by GCase in LRRK2 mutant cells, and that these studies may contribute to the use of BMP-positive EVs as a biomarker for Parkinson's disease and associated treatments.

Strengths:

This is a potentially interesting paper,. However, I had comments that authors needed to address to clarify some aspects of their study.

Weaknesses:

(1) The authors seem to have missed the point in their reply to my first comment. They mention the paper by Stuffers et al., who reports that endosome biogenesis continues without ESCRT. This is a nice paper, but it is irrelevant to the subject at hand. In my initial comment, I drew the author's attention to an apparent contradiction: higher LAMP2 staining in R1441G LRRK2 knock-in MEFs and yet smaller MVEs with a reduced surface area. LAMP2 being one of the major glycoproteins of MVE's limiting membrane, one would have expected lower LAMP2 staining if cells contain fewer and smaller MVEs. Authors now state that elevated LAMP2 expression in cells expressing R1441G reflects a cell type-specific effect (differential penetrance of LRRK2 signaling on lysosomal biogenesis), because amounts of LAMP1 and CD63 are similar in cells from LRRK2 G2019S PD patients and control cells (new Fig 7A-F). However, authors still conclude that LRRK2 modulates the lysosomal network, including LAMP2 and CLN5. Does it?

Similarly, the mass spec analysis of BMP (Fig S1H) does not support the data in Fig 1. Does this Table include all major isoforms found in these cells? If so, the dominant isoform is by far the di-18:1 isoform in wt and R1441G cells (at least 10X more abundant than other isoforms). Now, di-18:1-BMP is roughly 4X more abundant in R1441G cells when compared to wt cells, while BMP is reduced by half in R1441G cells (light microscopy in Fig 1). Authors argue that light microscopy may only detects a so-called antibody accessible pool. What is this? And why would this pool decrease in R1441G cells when LAMP2 is higher? Alternatively, they argue that the anti-BMP antibody may be less specific and detect other analytes. As I had already mentioned, this makes no sense, since the observed signal is lower and not higher. If authors do not trust their light microscopy analysis, why show the data?

(2) Cells contain 3 LAMP2 isoforms. Which one is upregulated and/or secreted in exosomes?

(3) The new Fig S4A is far from convincing. How were cells fractionated and what are the gradients (not described in Methods)? CD63 (presumably endolysosomes) is spread over fractions 8 - 13. LRRK2 (fractions 8-9) does not copurify with CD63. The bulk of LRRK2 is at the bottom (presumably cytosol if this is a floatation gradient), and a minor fraction moves into the gradient. CLN5 is even less clear since the bulk is also at the bottom with a tiny fraction only between LRRK2 and CD63. Also, why do authors conclude that a considerable pool of newly synthesized CLN5 did not reach its final destination at the endolysosome and may instead be retained in the ER? Where is the ER on the gradient?

(4) Fig S4B shows blots of whole cell lysates from CTRL and LRRK2 mutant-derived fibroblasts: 6 lanes are shown but without captions, containing varying amounts of calnexin and CD63. In addition, the blots look very dirty. Where is CD63? Is it the minor band at ≈37 kD (as in Fig S4A)? Or the major band below the 50kD marker? What are the other bands on these blots? As a result, the quantification shown in the bar graph does not mean much.

(5) The cell content of 18.1-BMP is increased approx. 5X by BafA1 (Fig 6C) but amounts of 18.1-BMP secreted in EVs hardly changes (Fig 6E). Since BMP is mostly present as 18.1 isoform (22:6-BMP being only a minor species, Fig S1H), does it mean that BafA1 does not increase BMP secretion and/or only a minor fraction of total cellular BMP is secreted in exosomes?

Comments on revisions:

How come 0.2 mmol/L of 22:6 and 18:1 fatty acid both correspond to 65 µg/mL (Fig 4A)?

It is stated in the Legend of Fig4 that long (B-C) and short (D) chase time points are shown as fold change. There is no panel D in the figure.

Author response:

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

eLife Assessment

This useful study presents the potentially interesting concept that LRRK2 regulates cellular BMP levels and their release via extracellular vesicles, with GCase activity further modulating this process in mutant LRRK2-expressing cells. However, the evidence supporting the conclusions remains incomplete, and certain statistical analyses are inadequate. This work would be of interest to cell biologists working on Parkinson's disease.

Reviewer #1 (Public review):

Summary:

Even though mutations in LRRK2 and GBA1 (which encodes the protein GCase) increase the risk of developing Parkinson's disease (PD), the specific mechanisms driving neurodegeneration remain unclear. Given their known roles in lysosomal function, the authors investigate how LRRK2 and GCase activity influence the exocytosis of the lysosomal lipid BMP via extracellular vesicles (EVs). They use fibroblasts carrying the PDassociated LRRK2-R1441G mutation and pharmacologically modulate LRRK2 and GCase activity.

Strengths:

The authors examine both proteins at endogenous levels, using MEFs instead of cancer cells. The study's scope is potentially interesting and could yield relevant insights into PD disease mechanisms.

Weaknesses:

Many of the authors' conclusions are overstated and not sufficiently supported by the data. Several statistical errors undermine their claims. Pharmacological treatment is very long, leading to potential off-target effects. Additionally, the authors should be more rigorous when using EV markers.

We thank the reviewer for these valuable observations. In the revised manuscript, we have addressed each of these points as follows:

(1) Conclusions and data support – We carefully revised our text throughout the manuscript to ensure that all conclusions are better supported by the presented data. For instance, we now explicitly state that while pharmacological modulation supports the regulatory role of LRRK2 activity in EV-mediated BMP release, we have softened our conclusions concerning the contribution of GCase in this model (see revised Results and Discussion sections).

(2) Statistical analyses – We reanalyzed experiments involving more than two groups and replaced simple t-tests with non-parametric Kruskal-Wallis tests followed by Dunn’s post hoc comparisons. This approach, described in the updated figure legends (e.g., Figure 2D-F and H-J), provides a more rigorous statistical framework that accounts for small sample sizes and variability typical of EV quantifications.

(3) Pharmacological treatment duration – Prolonged MLi-2 treatments have been extensively used in the field without evidence of significant off-target effects. Several studies, including Fell et al. (2015, J Pharmacol Exp Ther 355:397-409), De Wit et al. (2019, Mol Neurobiol 56:5273-5286), Ho et al. (2022, NPJ Parkinson’s Dis 8:115),Tengberg et al. (2024, Neurobiol Dis 202:106728), and Jaimon et al. (2025, Sci Signal 18:eads5761), have applied long-term (24-48 h) MLi-2 treatments at comparable concentrations without detecting toxicity or off-target alterations, including in MEFs (Ho et al., 2022; Dhekne et al., 2018, eLife 7:e40202).  In our study, 48-hour incubations were necessary to sustain full LRRK2 inhibition throughout the extracellular vesicle (EV) collection period. EV biogenesis, BMP biosynthesis, and packaging into EVs are timedependent processes; therefore, extended incubation and collection periods (48 h) were required to allow downstream effects of LRRK2 inhibition on BMP production and release to manifest, and to obtain sufficient EV material for biochemical and lipidomic analyses. This experimental design also reflects our and others’ previous observations in humans and non-human primates, where urinary BMP changes are associated with chronic or subchronic LRRK2 inhibitor treatment (Baptista MAS, Merchant K, et al. Sci Transl Med. 2020, 12:eaav0820; Jennings D, et al. Sci Transl Med. 2022, 14:eabj2658; Maloney MT, et al. Mol Neurodegener. 2025, 20:89). Importantly, under these conditions, we did not observe significant changes in cell viability or morphology, supporting that the treatment was well tolerated.  We have clarified this rationale in the revised Methods section to emphasize that the prolonged incubation reflects the experimental design for EV isolation rather than a requirement for achieving LRRK2 inhibition.

(4) EV markers – We and others have reported enrichment of Flotillin-1 and LAMP proteins in isolated small EV fractions (Kowal et al., 2016; Lu et al., 2018; Mathieu et al., 2021; Ferreira et al., 2022). Moreover, LAMP proteins have been reported to be more enriched in EVs of endolysosomal origin (Mathieu et al., 2021). To further strengthen this point, we performed new experiments using a CD63-pHluorin sensor combined with TIRF microscopy, which allowed real-time visualization of CD63-positive exosome release. These new data (now presented in Figure 7, Panels G-I; Videos 1 and 2) confirm increased CD63-positive EV release in LRRK2 mutant fibroblasts, which was reversed by LRRK2 inhibition with MLi-2. The CD63-positive compartment was also largely BMPpositive (new Figure 7D, F, G), reinforcing our conclusions and providing additional rigor in EV marker validation.

Reviewer #2 (Public review):

Summary:

In this paper, the authors used MEFs expressing the R1441G mutant of leucine-rich repeat kinase 2 (LRRK2), a mutant associated with the early onset of Parkinson's disease. They report that in these cells LAMP2 fluorescence is higher but BMP fluorescence is lower, MVE size is reduced, and that MVEs contain less ILVs. They also report that LAMP2-positive EVs are increased in mutant cells in a process sensitive to LRRK2 kinase inhibition but are further increased by glucocerebrosidase (GCase) inhibition, and that total di-22:6-BMP and total di-18:1-BMP are increased in mutant LRRK2 MEFs compared to WT cells by mass spectrometry. They also report that LRRK2 kinase inhibition partially restores cellular BMP levels, and that GCase inhibition further increases BMP levels, and that in EVs from the LRRK2 mutant, LRRK2 inhibition decreases BMP while GCase inhibition has the opposite effect. Moreover, they report that the BMP increase is not due to increased BMP synthesis, although the authors observe that CLN5 is increased in LRRK2 mutant cells. Finally, they report that GW4869 decreases EV release and exosomal BMP, while bafilomycin A1 increases EV release. They conclude that LRRK2 regulates BMP levels (in cells) and release (via EVs). They also conclude that the process is modulated by GCase in LRRK2 mutant cells, and that these studies may contribute to the use of BMP-positive EVs as a biomarker for Parkinson's disease and associated treatments.

Strengths:

This is an interesting paper, which provides novel insights into the biogenesis of exosomes with exciting biomedical potential. However, I have comments that authors need to address to clarify some aspects of their study.

Weaknesses:

(1) The intensity of LAMP2 staining is increased significantly in cells expressing the R1441G mutant of LRRK2 when compared to WT cells (Figure 1C). Yet mutant cells contain significantly smaller MVEs with fewer ILVs, and the MVE surface area is reduced (Figure 1D-F). This is quite surprising since LAMP2 is a major component of the limiting membrane of late endosomes. Are other proteins of endo-lysosomes (eg, LAMP1, CD63, RAB7) or markers (lysotracker) also decreased (see also below)?

As referenced in our original manuscript, several previous studies have reported endolysosomal morphological and homeostatic defects in cells harboring pathogenic LRRK2 mutations. LAMP2 can be upregulated as part of a lysosomal biogenesis or stress response (e.g., via MiT/TFE transcription factors such as TFEB; Sardiello et al., Science 2009, 325:473-477), whereas ILV biogenesis is primarily controlled by ESCRT- and SMPD3-dependent pathways that are regulated independently of MiT/TFE-driven transcriptional programs. Indeed, Stuffers et al. (Traffic 2009, 10:925-937) demonstrated that depletion of key ESCRT subunits markedly inhibited ILV formation while concomitantly increasing LAMP2 expression, highlighting the mechanistic dissociation between LAMP2 abundance and ILV number. In our study, we observed a similar pattern in R1441G LRRK2 MEFs, in which elevated LAMP2 staining and protein levels occurred despite a reduction in MVE size and ILV number. We interpret this as a compensatory lysosomal biogenesis response.

Our revised manuscript now includes new immunofluorescence data for BMP, LAMP1 and CD63 (New Figure 7, Panels A-F) together with biochemical analysis of CD63 protein levels (New Supplemental Figure 4, Panel B) in human skin fibroblasts derived from healthy donors and LRRK2 G2019S PD patients. Quantitative analysis of these experiments revealed no statistically significant differences in total cellular levels of either LAMP1 or CD63 between groups. However, we observed a consistent decrease in BMP immunostaining intensity (New Figure 7, Panel A and B), in agreement with our findings in mouse fibroblasts. We therefore propose that the elevated LAMP2 expression observed in the engineered MEF clone expressing R1441G may reflect a cell type-specific effect, potentially linked to differential penetrance of LRRK2 signaling on the lysosomal biogenesis response. We have updated the Results and Discussion section of the manuscript to incorporate and clarify these findings.

(2) LRRK2 has been reported to interact with endolysosomal membranes. Does the R1441G mutant bind LAMP2- and/or BMP-positive membranes? 

We agree that LRRK2 has been reported to associate dynamically with endolysosomal membranes, particularly under conditions of endolysosomal stress or damage (Eguchi T, et al. PNAS 2018, 115:E9115-E9124; Bonet-Ponce L, et al. Sci Adv. 2020, 6:eabb2454; Wang X, et al. Elife. 2023, 12:e87255).

Nevertheless, to explore whether LRRK2 associates with BMP-positive endolysosomes, we performed subcellular fractionation followed by biochemical analysis of endolysosomal fractions, since our available LRRK2 antibodies did not provide reliable immunofluorescence signals. These experiments were carried out using human skin fibroblasts derived from both healthy controls and Parkinson’s disease patients carrying the LRRK2-G2019S mutation. In both control and mutant fibroblasts, a pool of LRRK2 was detected in fractions positive for the BMP synthase CLN5 and the endolysosomal marker CD63 (New Supplementary Figure 4, Panel A), supporting the localization of LRRK2 to endolysosomal membranes that are likely BMP-enriched. Our manuscript’s Results and Methods sections have been updated accordingly.

Does the mutant affect endolysosomes?

As referenced in our original manuscript, several studies have reported that pathogenic LRRK2 mutations can lead to endolysosomal defects. Consistent with these reports, we also observed morphological alterations in endolysosomes of cells expressing mutant LRRK2, including reduced MVE size and fewer ILVs, as shown in Figure 1D–F. These observations are in agreement with previously described phenotypes associated with pathogenic LRRK2 variants. Furthermore, in mutant LRRK2 MEFs, and now in humanderived fibroblasts (see new Figure 7, Panel A and B), we observed a decrease in BMP immunostaining signal.

(3) Immunofluorescence data indicate that BMP is decreased in mutant LRRK2expressing cells compared to WT (Figure 1A-B), but mass spec data indicate that di-22:6BMP and di-18:1-BMP are increased (Figure 3). Authors conclude that the BMP pool detected by mass spec in mutant cells is less antibody-accessible than that present in wt cells, or that the anti-BMP antibody is less specific and that it detects other analytes. This is an awkward conclusion, since the IF signal with the antibody is lower (not higher): why would the antibody be less specific? Could it be that the antibody does not see all BMP isoforms equally well? Moreover, the observations that mutant cells contain smaller MVEs (Figure 1D-F) with fewer ILVs are consistent with the IF data and reduced BMP amounts. This needs to be clarified.

As previously reported by us (Lu et al., J Cell Biol 2022;221:e202105060) and others (Berg AL, et al. Cancer Lett. 2023, 557:216090), discrepancies can occur between BMP levels detected by immunofluorescence and those quantified by mass spectrometry. This is because immunostaining reflects the pool of antibody-accessible BMP, whereas lipidomics measures the total cellular content of all BMP molecular species, irrespective of their distribution or accessibility.

We agree that the anti-BMP antibody may not detect all BMP isoforms equally well. Differences in acyl chain composition (such as the degree of saturation or chain length) can alter the stereochemistry of BMP and, consequently, epitope accessibility to antibody binding.

In addition, in a personal communication with Monther Abu-Remaileh (Stanford University), we were informed that the antibody may also cross-react with other lipid species in endolysosomes. Nevertheless, since there is no formal evidence supporting this, we have removed the sentence in the Discussion section stating “Alternatively, the antibody may also detect non-BMP analytes” to avoid any potential misinterpretations. In its place, we have added a short statement noting that “not all BMP isoforms may be detected equally well”.

Mass spectrometry data are only shown for two BMP species (di-22:6, di-18:1). What are the major BMP isoforms in WT cells? The authors should show the complete analysis for all BMP species if they wish to draw quantitative conclusions about the amounts of BMP in wt and mutant cells. Finally, BMP and PG are isobaric lipids. Fragmentation of BMPs or PGs results in characteristic fingerprints, but the presence of each daughter ion is not absolutely specific for either lipid. This should be clarified, e.g., were BMP and PG separated before mass spec analysis? Was PG affected? The authors should also compare the BMP data with mass spec data obtained with a control lipid, e.g., PC.

Regarding BMP isoforms, our targeted UPLC-MS/MS analyses revealed that 2,2′-di-22:6-BMP (sn2/sn2′) and 2,2′-di-18:1-BMP (sn2/sn2′) are the predominant BMP isoforms in MEF cells, consistent with previous reports showing docosahexaenoyl (22:6; DHA) and oleoyl (18:1) BMP as the most abundant isoforms. Across diverse mammalian cells and tissues, BMP typically exhibits a fatty acid composition dominated by oleoyl, with polyunsaturated fatty acids (particularly DHA) also contributing substantially. Enrichment of DHA-containing BMP species has been observed in multiple systems, including rat uterine stromal cells, PC12 cells, THP-1 and RAW macrophages, as well as in rat and human liver. This consistent presence of oleoyl- and docosahexaenoyl-containing BMP species across tissues indicates that these acyl chains are conserved features influencing the lipid’s structural and functional characteristics (Kobayashi et al. J Biol Chem, 2002; Hullin-Matsuda et al. Prostaglandins Leukotriens Essent Fatty Acids, 2009; Thompson et al. Int J Toxicol. 2012; Delton-Vandenbroucke et al. J Lipid Res, 2019).

Nevertheless, we have included a Table (Panel H in updated Supplemental Figure 1) showing other BMP species that were also detected in our lipidomics analysis. Overall, dioleoyl (18:1)- and di-docosahexaenoyl (22:6)-BMP species were the most abundant in MEF cells, whereas di-arachidonoyl (20:4)- and di-linoleoyl (18:2)-BMP isoforms were present at lower levels. Consistently, R1441G LRRK2 MEFs displayed higher levels of dioleoyl- and di-docosahexaenoyl-BMP compared with WT cells, and these elevations were reduced following LRRK2 kinase inhibition with MLi-2. Data from three independent representative experiments are shown, and the manuscript has been revised accordingly to include these results.

Regarding the separation of BMP and PG species, we confirm that BMP and PG were chromatographically resolved prior to MS/MS detection using a validated UPLC-MS/MS method developed by Nextcea, Inc. PG exhibits a substantially longer LC retention time than BMP, ensuring complete baseline separation. This approach (established by Nextcea nearly two decades ago and later validated through a multi-year collaboration with the U.S. FDA to clinically qualify di-22:6-BMP as a biomarker) prevents any ambiguity arising from the isobaric nature of BMP and PG species. No changes in PG levels were detected under any experimental conditions.

Finally, we employed isotope-labeled BMP as an internal standard to ensure robust normalization across samples. These additional details and references cited above have been included in the revised Methods and References sections to further clarify the analytical rigor of our lipidomics workflow.

(4) It is quite surprising that the amounts of labeled BMP continue to increase for up to 24h after a short 25min pulse with heavy BMP precursors (Figure 4B).

In these isotope-labeling experiments, it is important to note (as described in our original manuscript) that two distinct pools of metabolically labeled BMP species were detected: semi-labeled BMP (with only one heavy isotope-labeled fatty acyl chain) and fully-labeled BMP (with both fatty acyl chains labeled). We consider the fully-labeled BMP pool to provide the most reliable readout for BMP turnover, as it showed a rapid decline after a 1h chase (decreasing by more than 50% within 8 h in all conditions), reaching its lowest levels at the end of the 48-h chase period.

The apparent increase in semi-labeled BMP species over time may be explained by continued incorporation of labeled precursors following the initial pulse. Specifically, once existing semi-labeled and fully-labeled BMP molecules are degraded by PLA2G15 (Nyame K, et al. Nature 2025, 642:474-483), the resulting isotope-labeled lysophosphatidylglycerol (LPG) and fatty acids could be recycled and re-enter a new round of BMP biosynthesis, leading to a gradual accumulation of semi-labeled BMP such as di-18:1-BMP. Why would this reasoning not also apply to the fully-labeled species? Once the pulse is completed, newly incorporated non-labeled fatty acyl chains present in the cellular pool can compete with labeled ones during subsequent rounds of lipid remodeling or synthesis. As a result, the probability of generating semi-labeled BMP molecules becomes higher than that of forming fully-labeled species. Consistent with this, our data show an increase in only semi-labeled BMP species (but not in fully-labeled ones) up to 24 hours after the pulse. We have added a clarification regarding this point in the revised manuscript.

(5) It is argued that upregulation of CLN5 may be due to an overall upregulation of lysosomal enzymes, as LAMP2 levels were also increased (Figure 2A, C, E). Again, this is not consistent with the observed decrease in MVE size and number (Figure 1D-F). As mentioned above, other independent markers of endo-lysosomes should be analyzed (eg, LAMP1, CD63, RAB7), and/or other lysosomal enzymes (e.g. cathepsin. D).

Our revised manuscript now includes new immunofluorescence data for BMP, LAMP1 and CD63 (New Figure 7, Panels A-F) together with biochemical analysis of CD63 protein levels (New Supplemental Figure 4, Panel B) in human skin fibroblasts derived from healthy controls and LRRK2 G2019S PD patients. Quantitative analysis of these experiments revealed no statistically significant differences in total cellular levels of either LAMP1 or CD63 between groups. However, our results consistently show increased CLN5 protein levels in both mouse and human fibroblast cell lines harboring pathogenic LRRK2 mutations. Upregulation of CLN5 may reflect a compensatory effect from loss of BMP via EV exocytosis. As discussed above, the elevated LAMP2 signal observed in the engineered MEF clone expressing R1441G could represent a cell type-specific effect, potentially linked to differential penetrance of LRRK2 signaling on the lysosomal biogenesis response. Our Results and Discussion sections have been updated accordingly.

(6) The authors report that the increase in BMP is not due to an increase in BMP synthesis (Figure 4), although they observe a significant increase in CLN5 (Figure 5A) in LRRK2 mutant cells. Some clarification is needed.

In our original manuscript, we proposed that although CLN5 protein levels are increased in R1441G LRRK2 MEFs, the absence of significant changes in BMP synthesis rates (Figure 4B, C) may reflect either limited substrate availability or that CLN5 is already operating near its maximal enzymatic capacity. Our new subcellular fractionation data (new Figure 7, Panel A) further indicate that, despite a relative increase in total CLN5 levels in G2019S LRRK2 human fibroblasts, the amount of CLN5 associated with endolysosomes remains comparable between mutant LRRK2 and control cells. This suggests that a considerable fraction of upregulated CLN5 may not localize to endolysosomes, potentially accumulating in the endoplasmic reticulum due to enhanced translation or impaired trafficking. Unfortunately, the available anti-CLN5 antibody did not yield reliable immunofluorescence signals, preventing us from directly confirming this possibility. Nevertheless, in light of our new data (new Supplemental Figure 4A), we have included a clarification in the revised manuscript discussing this possibility as well.

(7) Authors observe that both LAMP2 and BMP are decreased in EVs by GW4869 and increased by bafilomycin (Figure 6). Given my comments above on Figure 1, it would also be nice to illustrate/quantify the effects of these compounds on cells by immunofluorescence.

We appreciate the reviewer’s suggestion. We have previously published immunofluorescence data showing increased BMP accumulation in endolysosomes following treatment with bafilomycin A1 Lu A, et al. J Cell Biol. 2009, 184:863-879). However, in the present study, our lipidomics analyses revealed a decrease in both di22:6-BMP and di-18:1-BMP species in cells treated with this compound. As discussed above, this apparent discrepancy likely reflects methodological differences between immunofluorescence, which detects only antibody-accessible BMP pools, and lipidomics, which quantifies total cellular BMP content. 

Moreover, in a recent study (Andreu Z, et al. Nanotheranostics 2023, 7:1-21), BMP levels were analyzed by immunofluorescence in cells treated with spiroepoxide, a potent and selective irreversible inhibitor of nSMase (different from GW4869) known to block EV release. Spiroepoxide-treated cells showed decreased BMP immunostaining; a result that, again, does not align with mass spectrometry data revealing increased cellular BMP levels upon GW4869 treatment. Notably, in that study, spiroepoxide was used instead of GW4869 because the intrinsic autofluorescence of GW4869 could potentially interfere with the immunofluorescence BMP signal.

We therefore consider lipidomics measurements to provide a more reliable and quantitative representation of BMP dynamics under these conditions.

Reviewer #1 (Recommendations for the authors):

Major concerns:

(1) 48 h for MLi2 treatment seems too long. LRRK2 kinase activity is inhibited with much shorter incubation times. The longer the incubation, the more likely off-target effects are. The authors should repeat these experiments with 1-2 h of MLi2.

We thank the reviewer for this valuable comment. We acknowledge that MLi-2 is a potent and selective LRRK2 kinase inhibitor that achieves near-complete target engagement within a few hours of treatment. However, prolonged exposure has been widely used in the field without evidence of significant off-target effects. Several studies, including Fell et al. (2015, J Pharmacol Exp Ther 355:397-409), De Wit et al. (2019, Mol Neurobiol 56:5273-5286), Ho et al. (2022, NPJ Parkinson’s Dis 8:115), Tengberg et al. (2024, Neurobiol Dis 202:106728), and Jaimon et al. (2025, Sci Signal 18:eads5761), have employed long-term (24-48 h) MLi-2 treatments at comparable concentrations without detecting toxicity or off-target alterations, including in MEFs (Ho et al., 2022; Dhekne et al., 2018, eLife 7:e40202).

In our study, 48-hour incubations were necessary to sustain full LRRK2 inhibition throughout the extracellular vesicle (EV) collection period. EV biogenesis, BMP biosynthesis, and packaging into EVs are time-dependent processes; therefore, extended incubation and collection periods (48 h) were required to allow downstream effects of LRRK2 inhibition on BMP production and release to manifest, and to obtain sufficient EV material for biochemical and lipidomic analyses. This experimental design also reflects our and others’ previous observations in humans and non-human primates, where urinary BMP changes are associated with chronic or subchronic LRRK2 inhibitor treatment (Baptista MAS, Merchant K, et al. Sci Transl Med. 2020, 12:eaav0820; Jennings D, et al. Sci Transl Med. 2022, 14:eabj2658; Maloney MT, et al. Mol Neurodegener. 2025, 20:89). Importantly, under these conditions, we did not observe significant changes in cell viability or morphology, supporting that the treatment was well tolerated.

We have clarified this rationale in the revised Methods section to emphasize that the prolonged incubation reflects the experimental design for EV isolation rather than a requirement for achieving LRRK2 inhibition.

(2) Is there a reason why the authors don't include CD81, CD63, and Syntenin-1 in their study as an EV marker? Using solely Flotilin-1 does not seem to be enough to justify their claims.

We actually used not only Flotillin-1 but also LAMP2 as EV markers in our study. While both Flotillin-1 and LAMP2 detection on EVs may vary depending on the cell type, we and others have reported enrichment of Flotillin-1 and LAMP proteins in isolated small EV fractions (Kowal et al., 2016; Lu et al., 2018; Mathieu et al., 2021; Ferreira et al., 2022). In particular, one of these studies reported that “LAMP1-positive subpopulations of EVs represent MVB/lysosome-derived exosomes, which also contain syntenin-1.” Therefore, our choice of EV markers (LAMP2 and Flotillin-1) is consistent with those previously and reliably used to characterize small EVs.

Nevertheless, to further address the reviewer’s concern, we performed additional experiments using a CD63-based fluorescence sensor (CD63-pHluorin), which, combined with TIRF microscopy, enables real-time visualization of CD63-positive exosome release. These experiments were conducted in control and LRRK2-mutant fibroblasts, and the data are presented in new Figure 7 (Panels G-I; Videos 1 and 2). We have also included all relevant references and clarified this point in the revised manuscript.

(3) Indeed, to quantify the amount of certain proteins in EVs, the authors should normalize them by CD63 or CD81.

Protein normalization in isolated EV fractions is indeed challenging. Although tetraspanins such as CD63 and CD81 are commonly enriched in EVs, their abundance can vary considerably across EV subpopulations, cell types, and experimental conditions, making them unreliable as universal normalization markers (Théry et al., J Extracell Vesicles, 2018; Margolis & Sadovsky, Nat Rev Mol Cell Biol, 2019).  Current guidelines from the International Society for Extracellular Vesicles (ISEV), as described in the Minimal Information for Studies of Extracellular Vesicles 2018 (MISEV2018; Théry C, et al. JExtracell Vesicles. 2018, 7:1535750) and updated in MISEV2024 (Welsh JA, et al. J Extracell Vesicles. 2024, 13:e12404), recommend reporting multiple EV markers rather than relying on a single protein for normalization. They also suggest ensuring comparable experimental conditions by using the same number of cells at the start of the experiment and normalizing EV data to cell number or whole-cell lysate protein content at the end of the experiment, among other approaches.

In our study, we normalized EV data to whole-cell lysate (WCL) protein content, as this approach accounts for differences in EV production due to variations in cell number or treatment conditions and is commonly used in the field (Kowal et al., PNAS, 2016; Mathieu et al., Nat Commun, 2021). We also included Flotillin-1 and LAMP2 as EV markers, both of which have been validated as molecular markers of small EV subpopulations.

(4) Hyper normalization in WB quantification in Figure 2E-G is statistically incorrect, as it assumes that one group (in this case, R1441G ctrl) has no variability at all, which is not biologically possible. The authors should repeat the quantification without hypernormalizing one of their groups. This issue is prevalent across the whole manuscript.

We understand the concern regarding “hyper-normalization” (i.e., expressing all values relative to one condition set to 1), which may mask variability in the reference group. However, it is standard practice in immunoblotting analysis to express data relative to a control condition for comparison, as variations in membrane transfer, exposure time, and signal development can differ across blots. In our case, the data are expressed as relative levels (arbitrary units) rather than absolute quantitative values. To facilitate comparison between datasets and account for inter-experimental variation, we continued to express values relative to the mutant LRRK2 MEF condition.

On the other hand, in lipidomics experiments, despite using the same number of seeded cells and identical extraction and analysis protocols, minor biological and technical variability was observed across independent replicates. This variability is inherent to the experimental system and is now explicitly represented in the new table included in Supplemental Figure 1F, which compiles three independent representative lipidomics experiments showing quantitative BMP levels across different conditions.

(5) The authors perform a t-test in Figure 2E-G when comparing more than 2 groups, which is wrong. The authors should use a two-way ANOVA as they are comparing genotype and treatment.

We appreciate the reviewer’s comment and agree with this observation. The MLi-2 and CBE experiments were performed independently and in separate experimental runs; therefore, we have reanalyzed these datasets separately rather than combining them in a two-way ANOVA. To properly compare more than two groups within each dataset, we have now applied a Kruskal-Wallis test followed by an uncorrected Dunn’s post hoc test (Figure 2 D-F and H-J). This non-parametric approach is more appropriate for our data structure, as EV experiments are usually subject to high variability and immunoblot quantifications involving small sample sizes (n≈6) do not always meet the assumptions of normality or equal variance. The Kruskal-Wallis test does not assume normality or equal variances, making it more robust for small, variable biological datasets. The statistical analyses and figure legend have been updated in the revised manuscript accordingly.

In addition, since our CBE treatments yielded statistically non-significant data, we have softened our conclusions throughout the manuscript concerning the contribution of GCase activity to EV-mediated BMP release modulation.

(6) There is a very strong reduction in flotillin-1 in R1441G cells vs WT (Figure 2G) in the EV fraction. That reduction is further exacerbated with MLi2, which likely means it is not kinase activity dependent. Can the authors comment on that?

We agree with the reviewer that Flotillin-1 showed a different behavior compared with LAMP2 in these experiments. As recommended by the MISEV guidelines (Théry C, et al. J Extracell Vesicles. 2018;  7:1535750; Welsh JA, et al. J Extracell Vesicles. 2024, 13:e12404), it is important to analyze more than one EV-associated protein marker. We examined LAMP2, which, together with LAMP1, has been reported to be specifically enriched in EVs of endolysosomal origin (exosomes; Mathieu et al., Nat Commun. 2021, 12:4389 ). In contrast, Flotillin-1 is also associated with small EVs but may represent a distinct EV subpopulation from those positive for LAMP proteins (Kowal J, et al. PNAS 2016, 113:E968-E977).

Nevertheless, the biochemical analysis of isolated EV fractions was complemented by our lipidomics data and, in the revised version, by TIRF microscopy analysis of exosome release in control and G2019S LRRK2 human fibroblasts (new Figure 7, Panels G-I; Videos 1 and 2). In this analysis, we confirmed increased exocytosis of CD63-pHluorin– positive endolysosomes in G2019S LRRK2 human fibroblasts compared to controls, an effect that was reversed by MLi-2 treatment. The CD63-pHluorin–positive compartment of these cells was also largely positive for BMP (new Figure 7G). Collectively, these findings further support the regulatory role of LRRK2 activity in EV-mediated BMP secretion.

(7) In Figure 2C, the authors should express that the LAMP2-EV and flotillin-1 EV fractions from the WB are highly exposed. As presently presented, it is slightly misleading.

We thank the reviewer for this comment. In EV preparations, the amount of protein recovered is typically very low. Therefore, although we loaded all the EV protein obtained from each sample, the immunoblots for LAMP2 and Flotillin-1 in EV fractions required longer exposure times to visualize clear signals across all conditions. We have now indicated in the corresponding figure legend that these EV blots are long-exposure blots to facilitate signal detection and avoid any potential misunderstanding.

(8) If Figure 2C and D are from two different experiments, they should not be plotted together in Figure 2E-G. You cannot compare the effect of MLi2 vs CBE if done in completely different experiments.

We appreciate the reviewer’s comment and agree with this observation. The MLi-2 and CBE experiments were performed independently and in separate experimental runs; therefore, we have reanalyzed these datasets separately rather than combining them in a two-way ANOVA. To properly compare more than two groups within each dataset, we have now applied a Kruskal-Wallis test followed by an uncorrected Dunn’s post hoc test (Figure 2 D-F and H-J). This non-parametric approach is more appropriate for our data structure, as EV experiments are usually subject to high variability and immunoblot quantifications involving small sample sizes (n≈6) do not always meet the assumptions of normality or equal variance. The Kruskal-Wallis test does not assume normality or equal variances, making it more robust for small, variable biological datasets. The revised statistical analyses and figure legends have been updated accordingly in the manuscript.

(9) The authors state that "For the R1441G MEF cells, MLi-2 decreased EV concentration while CBE increased EV particles per ml, in agreement with the effects observed in our biochemical analysis." As Figure S1D shows no statistical significance, the authors don't have sufficient evidence to make this claim.

We apologize for this overstatement. We have revised the text to clarify that, although the differences did not reach statistical significance, a consistent trend toward decreased EV concentration upon MLi-2 treatment and increased EV release following CBE treatment was observed in R1441G MEF cells.

(10) "Altogether, given that BMP is specifically enriched in ILVs (which become exosomes upon release), the data presented above support our biochemical analysis (Figure 2C, D, F) and suggest a role for LRRK2 and GCase in modulating BMP release in association with LAMP2-positive exosomes from MEF cells." As Figure 3E shows no statistical difference of BMP on EVs upon CBE treatment, this sentence is not accurate and should be reframed. Furthermore, the authors claim an increase in EV-LAMP2 in R1441G cells compared to WT, however, the amount of BMP in EVs of R1441G cells vs WT is unchanged with a non-significant reduction. This contradiction does not support the authors' conclusions and really puts into question their whole model.

We thank the reviewer for this observation. After reanalyzing our biochemical data from isolated EV fractions (see new Panels D-F and H-J) using an improved statistical approach, we found that although EV-associated LAMP2 levels were consistently elevated in untreated R1441G LRRK2 MEFs compared to WT cells, CBE treatment only produced a non-significant trend toward increased EV-associated LAMP2 compared to untreated R1441G LRRK2 cells. Accordingly, we have revised the sentence to read as follows:

“Altogether, given that BMP is specifically enriched in ILVs (which become exosomes upon release), the data presented above support our biochemical analysis (Figure 2C, E, G, I) and suggest that LRRK2 activity regulates BMP release in association with LAMP2positive exosomes, whereas GCase activity appears to have a more variable effect under the tested conditions.”

We also agree with the reviewer that, in our MEF model, the amount of BMP in EVs of R1441G cells vs WT is unchanged with a non-significant reduction. However, pharmacological modulation supports our conclusion that BMP release is modulated by LRRK2 activity. Specifically, treatment with the LRRK2 inhibitor MLi-2 decreased EVassociated BMP and LAMP2 levels in R1441G LRRK2 MEFs, and our new data (new Figure 7, Panel G-I; Videos 1 and 2) show increased exocytosis of CD63-pHluorin– positive endolysosomes in G2019S LRRK2 human fibroblasts compared to controls, an effect that was reversed by MLi-2 treatment. The CD63-pHluorin–positive compartment of these cells was also largely positive for BMP (new Figure 7G).

In light of the reviewer’s comment about CBE treatment, we have softened our conclusions throughout the manuscript concerning the contribution of GCase activity in this model.

(11) In Figure 5, 16 h of MLi2 treatment is too long and can lead to off-target effects. I would advise reducing it to 1-4 h.

Prolonged MLi-2 treatments have been extensively used in the field without evidence of significant off-target effects. Several studies, including Fell et al. (2015, J Pharmacol Exp Ther 355:397-409), De Wit et al. (2019, Mol Neurobiol 56:5273-5286), Ho et al. (2022, NPJ Parkinson’s Dis 8:115), Tengberg et al. (2024, Neurobiol Dis 202:106728), and Jaimon et al. (2025, Sci Signal 18:eads5761), have applied long-term (24-48 h) MLi-2 treatments at comparable concentrations without detecting toxicity or off-target alterations, including in MEFs (Ho et al., 2022; Dhekne et al., 2018, eLife 7:e40202). Moreover, the data presented in Figure 5 demonstrate a reduction in CLN5 protein levels in both MEFs and human fibroblasts following MLi-2 treatment, confirming the specificity of the observed effects in LRRK2 mutant cells.

(12) "Our data suggest that BMP is exocytosed in association with EVs and that LRRK2 and GCase activities modulate BMP secretion." Again, cells carrying the R1441G mutation have the same amount of BMP in EVs than WT. This sentence is not factually accurate. Accordingly, CBE did not change the amount of BMP in EVs.

We thank the reviewer for this observation and agree that, in our MEF model, the amount of BMP in EVs from R1441G LRRK2 cells is comparable to that observed in WT cells. However, pharmacological modulation supports our conclusion that BMP release is modulated by LRRK2 activity. Specifically, treatment with the LRRK2 inhibitor MLi-2 decreased EV-associated BMP levels in R1441G LRRK2 MEFs, and our new data (new Figure 7G-I; Videos 1 and 2) show increased exocytosis of CD63-pHluorin–positive endolysosomes in G2019S LRRK2 human fibroblasts compared to controls, an effect that was reversed by MLi-2 treatment. The CD63-pHluorin–positive compartment of these cells was also largely positive for BMP (new Figure 7G). These findings further support the regulatory role of LRRK2 activity in EV-mediated BMP secretion. In addition, in light of the reviewer’s comment about CBE treatment, we have softened our conclusions throughout the paper concerning the contribution of GCase activity in this model.

(13) Figure 6; EV release should have been monitored by more accurate markers such as CD63 and CD81.

We thank the reviewer for this comment. We and others (Kowal et al., 2016; Lu et al., 2018; Mathieu et al., 2021; Ferreira et al., 2022) have reported enrichment of Flotillin-1 and LAMP proteins in isolated small EV fractions. In particular, one of these studies (Mathieu et al., Nat Commun. 2021), in which bafilomycin A1 was also used (to boost exosome release), reported that “LAMP1-positive subpopulations of EVs represent MVB/lysosome-derived exosomes, which also contain syntenin-1.” Altogether, our choice of EV markers (LAMP2 and Flotillin-1) is consistent with those previously and accurately used to characterize EVs. We have now included all relevant references in the revised manuscript to further clarify this point.

(14) Figure 6 suggests that exosomal BMP is controlled by EV release. I would think that is rather obvious.

We agree that the finding that exosomal BMP release is influenced by EV secretion may appear “obvious.” However, our intention in Figure 6 was to provide direct experimental evidence confirming this relationship using pharmacological modulators of EV release. Specifically, inhibition of EV secretion with GW4869 reduced exosomal BMP levels, whereas stimulation with bafilomycin A1 increased them. These data were important to establish a causal link between EV trafficking and BMP export, thereby validating our model and supporting the interpretation that LRRK2 regulates BMP homeostasis through EV-mediated exocytosis, which is further modulated, to some extent, by GCase activity. 

Minor concerns:

(1) Figure 1: Change colors to be color blind friendly.

We thank the reviewer for this helpful suggestion. We have adjusted the colors in Figure 1 to be color-blind friendly. In addition, we have applied the same color-blind friendly palette to the new immunofluorescence data presented in new Figure 7, Panel A and D.

(2) More consistency on "Xmin" vs "X min" would be appreciated.

We thank the reviewer for this observation. We have revised the manuscript to ensure consistent formatting of time indications throughout the text and figures, using the standardized format “X min.”

Reviewer #2 (Recommendations for the authors):

(1)  Figure 2C-D. Were equal amounts of protein loaded in each lane?

Equal protein amounts were loaded in lanes corresponding to whole-cell lysate (WCL) fractions and normalized based on α-Tubulin levels.

For the extracellular vesicle (EV) fractions, all protein recovered from EV pellets after isolation was loaded. In all EV-related experiments, we seeded the same number of EVproducing cells per condition, and the resulting EV-derived data (from both immunoblotting and lipidomics analyses) were normalized to the corresponding whole cell lysate (WCL) protein content to ensure comparability across conditions.

All these technical details have been included in the Materials section of our revised manuscript.

(2) The authors refer to the papers of Medoh et al (ref 43) and Singh et al. (44) for the key role of CLN5 in the BMP biosynthetic pathway. However, Medoh et al reported that CLN5 is the lysosomal BMP synthase. In contrast, Singh et al. reported that PLD3 and PLD4 mediate the synthesis of SS-BMP, and did not find any role for CLN5. 

To avoid any confusion or misinterpretation of our findings regarding CLN5 and given that we do not analyze PLD3 or PLD4 in our study, we have decided to replace the reference to Singh et al. with Bulfon D. et al. (Nat. Commun. 2024, 15:9937) instead. This last work, conducted by an independent group distinct from the one that originally described CLN5, also validated CLN5 as the sole BMP synthase in cells.

Also, authors mention that bafilomycin A1 (B-A1) dramatically boosts EV exocytosis, referring to Kowal et al., 2016 (ref 35) and Lu et al., 2018 (ref 45). However, this is not shown in Kowal et al.

We thank the reviewer for pointing out this mistake. We apologize for the incorrect citation and have now corrected the reference. The statement regarding the effect of bafilomycin A1 on EV exocytosis now appropriately refers to Mathieu et al., 2021 and Lu et al., 2018.

(3) Page 7, it is stated that "No statistically significant differences in intracellular BMP levels were observed in WT LRRK2 MEFs upon LRRK2 or GCase inhibition(Supplemental Figure 1D, E)". The authors probably mean "Supplemental Figure 1F, G"

We thank the reviewer for noting this error. We have corrected the text to refer to panels F and G of Supplemental Figure 1, which correspond to the relevant data. We have also revised the reference to panel I of Supplemental Figure 1 accordingly.

So does it means that if we do not have consciousness thought, we are also monsters? In other words, are these what made us a human?

It is related to the text, when the beautiful aliens come, people welcome, but though the text, we could infer that they could be the one killing all the other aliens and destroy their home-planet.

then can we argue that he is morally wrong, but having a sense of purpose?

Maybe not because he does not understand what kind of crime he was committing.

in other words, it is impossible for him/her to realize that he is committing a crime because he is lack of cognitive learning.

connecting to the argument over "Adopt to the environment or be consistent to the own belief"

In a way, like with energy, which is not spent or generated but continuously transferred, we should not think in close yes-no, action-result. The event is part of a system, it's on the move, and efficiency doesn't emerge from nothing, it requires other work. I am not talking about zero-sum competition, we can most win with tech, but transformations like eye glasses or leg prosthetics need of workers on the other end, but by automatising them, we are just making them less visible, we are moving them from the artisan workshop to the factory or the mine. We'll have to wait a lot until this manual labour gets replaced by robotics, because once again, the trade-off is not "efficient" right now.

Okay, I get it, snake oil vendors are the people who get popular and get government grants to do next to nothing, because the instruction is only a part of the process. Beyond unlearning and learning there must be a change of habits, and no single play session or workshop can achieve that.

Yes, they can nudge toward visibility, but do they re-distribute? They can pinpoint and landmarks to look, and provide ways to not missbehave, but if they then need to be applied, and moral courage is not at its peak, as violence looms on the other side of the spectrum, these products can be almost a self-cleansing sterilisation tool, to merely perform predisposition to change, and to alleviate the cognitive dissonance of not doing so.

Sounds feasible to me, let's see where this goes.

Actually, that's the claim the manosphere makes.

The example is pretty shitty, but it's true that when you work in a commercial game company and you can flip assets and code, Unreal becomes very easy to use with first person shooting and enemies: It's purposelly built, like Fortnite UEFN.

I knew about the scanners, but I hadn't thought about the actual process! Upon lifting the veil, it makes sense that its usage as a visual sensory extension-augmentation is only tangible insofar as we used the data comparatively, like a medic would with Breast cancer or other illness - issue being here data gaps, and non-updating practicioners that do not know about minority illnesses.

What technology lacks politics? A food bowl? A door hinge? Even these have, to a low degree. A paramount example of politics in tech is the printing press, initially barred and then used religiously to spread the cath Bible.

One could argue, like Byung Chul, that phones are religious. Capitalistically religious, therefore, political. Say, they could be sourced differently, yes, and in that sense they are not as political as our context and usage has weaponised them to be, which could be the counterargument.

I love this. <3

Author response:

eLife Assessment

This useful study raises interesting questions but provides inadequate evidence of an association between atovaquone-proguanil use (as well as toxoplasmosis seropositivity) and reduced Alzheimer's dementia risk. The findings are intriguing but they are correlative and hypothesis-generating with the strong possibility of residual confounding.

We thank the editors and reviewers for characterizing our work as useful and for the opportunity to publish a Reviewed Preprint with a corresponding response. However, the statements in the Assessment characterizing the evidence as ‘inadequate’ and asserting a ‘strong possibility of residual confounding’ are factually incorrect as applied to our data and incompatible with the empirical findings presented in the manuscript. We have notified the editors of this factual inaccuracy. As the Assessment will be published as originally written, we provide clarification here to ensure an accurate scientific record for readers of the Reviewed Preprint.

Our study shows that the association between atovaquone–proguanil (A/P) exposure and reduced dementia risk, first identified in a rigorously matched national cohort in Israel, is robustly reproduced across three independently constructed age-stratified cohorts in the U.S. TriNetX network (with exposure at ages 50–59, 60–69, and 70–79). In each cohort, individuals exposed to A/P were compared with rigorously matched individuals who received another medication at the same age and were then followed over a decade for incident dementia. Cases and controls were matched on all major established dementia risk factors: age, sex, race/ethnicity, diabetes, hypertension, obesity, and smoking status.

Across all three strata, each containing more than 10,000 exposed individuals with an equal number of matched controls, we observed substantial and consistent reductions in cumulative dementia incidence (HR 0.34–0.51), extremely low P-values (10^–16 to 10^–40), and continuously widening divergence of Kaplan–Meier curves over the follow-up period. To more rigorously exclude the possibility of unmeasured baseline differences in health status, we additionally performed, for the purpose of this response, comparative analyses of key indicators of frailty and clinical utilization, including emergency and inpatient encounters, as well as the prevalence of mild cognitive impairment prior to medication exposure (values provided below in response to Reviewer #2, Weakness 1). These analyses provide clear evidence showing no pattern suggestive of exposed individuals being medically or cognitively healthier at baseline.

Taken together, these findings constitute a rigorously matched and independently replicated association across two national health systems, using TriNetX, the most widely cited real-world evidence platform in published cohort studies. Replication across three age strata, each with >10,000 exposed individuals, followed for a decade, and matched on all major known risk factors for dementia, meets the accepted epidemiologic definition of strong and reproducible evidence.

Although we disagree with elements of the editorial Assessment that appear inconsistent with the empirical findings, we will proceed with publication of the current manuscript as a Reviewed Preprint in order to ensure timely dissemination of findings with meaningful implications for public health and dementia prevention. In this initial public version, the point-by-point responses below provide concise explanations addressing the critiques underlying the Assessment. A revised manuscript, incorporating expanded baseline comparisons across each TriNetX age stratum, additional stringent exclusions, and an expanded discussion that will address the remarks presented in this review, will be submitted shortly.

Reviewer #1 (Public review):

Summary:

This useful study provides incomplete evidence of an association between atovaquone-proguanil use (as well as toxoplasmosis seropositivity) and reduced Alzheimer's dementia risk. The study reinforces findings that VZ vaccine lowers AD risk and suggests that this vaccine may be an effect modifier of A-P's protective effect. Strengths of the study include two extremely large cohorts, including a massive validation cohort in the US. Statistical analyses are sound, and the effect sizes are significant and meaningful. The CI curves are certainly impressive.

Weaknesses include the inability to control for potentially important confounding variables. In my view, the findings are intriguing but remain correlative / hypothesis generating rather than causative. Significant mechanistic work needs to be done to link interventions which limit the impact of Toxoplasmosis and VZV reactivation on AD.

We thank the reviewer for describing our study as useful and for highlighting several of its strengths, including the very large cohorts, sound statistical analyses, meaningful effect sizes, and the impressive CI curves. We also appreciate the reviewer’s recognition that our findings reinforce prior evidence linking VZV vaccination to reduced AD risk.

Regarding the statement that the evidence remains incomplete due to “inability to control for potentially important confounding variables,” we refer to our introductory explanation above. As noted there, our analyses meet the accepted criteria for reproducible epidemiological evidence, and the assumption of uncontrolled confounding is contradicted by rigorous matching and by additional baseline evaluations. We fully agree that mechanistic work is warranted, and our epidemiologic findings strongly motivate such efforts.

We address the reviewer’s specific comments in detail below.

(1) Most of the individuals in the study received A-P for malaria prophylaxis as it is not first line for Toxo treatment. Many (probably most) of these individuals were likely to be Toxo negative (~15% seropositive in the US), thereby eliminating a potential benefit of the drug in most people in the cohort. Finally, A-P is not a first line treatment for Toxo because of lower efficacy.

We agree that individuals in our cohort received Atovaquone-Proguanil (A-P) for malaria prophylaxis rather than for treatment of toxoplasmosis. However, this does not contradict our interpretation. Because latent CNS colonization by T. gondii is not currently considered clinically actionable, asymptomatic carriers are not offered treatment, and therefore would only receive an anti-Toxoplasma regimen unintentionally, through a medication prescribed for another indication such as malaria prophylaxis. Importantly, atovaquone is an established therapy for toxoplasmosis, including CNS disease, with documented efficacy and CNS penetration in current treatment guidelines. It is therefore reasonable to assume that, during the multi-week course typically administered for malaria prophylaxis, A-P would exert significant anti-Toxoplasma activity in individuals with latent CNS infection, potentially reducing or eliminating parasite burden even though the medication was not prescribed for that purpose.

The reviewer notes that only ~15% of individuals in the U.S. are Toxoplasma-seropositive, based on surveys performed primarily in young adults of reproductive age (serologic testing is most commonly obtained in women during prenatal care). However, seropositivity increases cumulatively over the lifespan, and few reliable estimates exist for the age groups in which Alzheimer’s disease and dementia occur. Even if we accept the lower estimate of ~15% latent colonization in older adults, this proportion is still smaller than the lifetime cumulative incidence of dementia in the general population.

Therefore, if latent toxoplasmosis contributes causally to dementia risk, and A-P is capable of eliminating latent Toxoplasma in the subset of individuals who harbor it, then a multi-week course of treatment—such as the one routinely taken for malaria prophylaxis—would be expected to produce a substantial reduction in dementia incidence at the population level, of the same order of magnitude reported here. A protective effect concentrated in a minority of exposed individuals is fully compatible with, and can mechanistically explain, the large overall reduction in risk that we observe.

Finally, the reviewer notes that A-P is not a first-line treatment for toxoplasmosis due to assumed lower efficacy. This point does not undermine our results. Even a second-line agent, when administered over several weeks—as is routinely done for malaria prophylaxis—is expected to exert substantial anti-Toxoplasma activity. The long duration of exposure in large populations receiving A-P for travel provides a unique natural experiment that does not exist for other anti-Toxoplasma medications, which, when prescribed for their non-Toxoplasma indications, are not taken more than a few days. Thus, the widespread use of A-P for malaria prophylaxis allows a unique opportunity to evaluate long-term outcomes following inadvertent anti-Toxoplasma treatment.

Moreover, “first line” recommendations in clinical guidelines refer to treatment of acute toxoplasmosis in immunosuppressed individuals, where tachyzoites are actively replicating. These guidelines do not consider efficacy against latent CNS colonization, which is dominated by bradyzoites, a biologically distinct form, in immunocompetent individuals. Therefore, the guideline hierarchy is not informative regarding which medication is more effective at clearing latent brain infection, the stage we consider most relevant to dementia risk.

(2) A-P exposure may be a marker of subtle demographic features not captured in the dataset such as wealth allowing for global travel and/or genetic predisposition to AD. This raises my suspicion of correlative rather than casual relationships between A-P exposure and AD reduction. The size of the cohort does not eliminate this issue, but rather narrows confidence intervals around potentially misleading odds ratios which have not been adjusted for the multitude of other variables driving incident AD.

We agree that prior to matching, A-P exposure may be associated with demographic features such as health or to travel internationally. However, this does not apply after matching. In all age-stratified analyses, exposed and control individuals were rigorously matched on all major risk factors known to influence dementia risk, including age, sex, race/ethnicity, smoking status, hypertension, diabetes, and obesity. Owing to the extremely large pool of individuals in TriNetX (~120M), our matching was performed stringently, producing exposed and unexposed cohorts that are near-identical with respect to the established determinants of dementia risk.

The reviewer correctly identifies that large cohorts alone do not eliminate confounding; however, confounding must still be biologically and epidemiologically plausible. Any hypothetical confounder capable of producing a 50–70% reduction in dementia incidence over a decade would need to: (1) produce a very large protective effect against dementia; (2) be strongly associated with A-P exposure; and (3) remain entirely uncorrelated with age, sex, race/ethnicity, smoking, diabetes, hypertension and obesity, which have been rigorously matched. No such factor has been proposed. The suggestion that an unspecified ‘subtle demographic feature’ could produce effects of this magnitude remains hypothetical, and no such factor has been described in the dementia risk literature.

If a specific evidence-supported confounder is proposed that meets these criteria, we would be pleased to test it empirically in our cohorts. In the absence of such a proposal, the interpretation that the association is merely “correlative rather than causal” remains speculative and does not negate the strength of a replicated, rigorously matched, long-term association across large cohorts in two national health systems.

(3) The relationship between herpes virus reactivation and Toxo reactivation seems speculative.

We respectfully disagree with the characterization of the herpesvirus–Toxoplasma interaction as speculative. The mechanism we describe is biologically valid, based on established virology and parasitology literature showing that latent T. gondii infection can reactivate from its bradyzoite state under inflammatory or immune-modifying conditions, including viral triggers. A published clinical report has documented CNS co-reactivation of T. gondii and a herpesvirus, explicitly noting that HHV-6 reactivation can promote Toxoplasma reactivation in neural tissue (Chaupis et al., Int J Infect Dis, 2016).

Moreover, this mechanism is the only currently evidence-supported explanation that simultaneously and parsimoniously accounts for all of the epidemiologic observations in our study:

(1) Substantially higher cumulative incidence of dementia in individuals with positive Toxoplasma serology, indicating that latent infection is a risk factor for subsequent cognitive decline;

(2) Strong protective association following A-P exposure, a medication with established activity against Toxoplasma gondii, including in the CNS;

(3) Independent protection conferred by VZV vaccination, observed consistently for two vaccines with distinct formulations (one live attenuated, one recombinant protein), whose only shared property is suppression of VZV reactivation;

(4) Greater protective effect of A-P among individuals who were not vaccinated against VZV, consistent with a model in which dementia risk requires both herpesvirus reactivation and persistent latent Toxoplasma infection—such that reducing either factor alone (via VZV vaccination or anti-Toxoplasma suppression) substantially lowers risk.

Taken together, these observations are difficult to reconcile under any alternative hypothesis.  

To date, we are unaware of any other biologically coherent mechanism that can explain all four findings simultaneously. We would welcome any alternative explanation capable of accounting for these converging epidemiologic signals, as such a proposal could meaningfully advance the scientific discussion. In the absence of a competing explanation, the interaction between latent toxoplasmosis and herpesvirus reactivation remains the most parsimonious hypothesis supported by current knowledge.

Finally, while observational studies are inherently limited in their ability to provide causal inference, the mechanism we propose is biologically grounded and experimentally testable. Our results provide a strong rationale for mechanistic studies and clinical trials, and warrant publication precisely because they generate a verifiable hypothesis that can now be evaluated directly.

(4) A direct effect on A-P on AD lesions independent on infection is not considered as a hypothesis. Given the limitations above and effects on metabolic pathways, it probably should be. The Toxo hypothesis would be more convincing if the authors could demonstrate an enhanced effect of the drug in Toxo positive individuals without no effect in Toxo negative individuals.

A direct effect of A-P on AD established lesions is indeed possible, and this hypothesis would be of significant therapeutic interest. However, we did not consider it within the scope of our epidemiologic analyses because all cohorts explicitly excluded individuals with existing dementia. Under these conditions, proposing a disease-modifying effect on established Alzheimer’s lesions based on our data would itself be speculative. Evaluating such a mechanism would be better answered by mechanistic or interventional studies rather than inference from populations without baseline disease.

We also agree that demonstrating a stronger protective effect among Toxoplasma-positive individuals would be informative. Unfortunately, this “natural experiment” cannot be performed using the available data: Toxoplasma serology is rarely ordered in older adults, and A-P exposure is itself uncommon, resulting in a cohort overlap far too small to yield valid statistical inference (n≈25 in TriNetX).

Thus, while both proposed hypotheses are scientifically attractive and merit further study, neither can be resolved using currently available real-world clinical data. Our findings provide the rationale to investigate both hypotheses experimentally, and we hope our report will motivate such studies.

Reviewer #2 (Public review):

Summary:

This manuscript examines the association between atovaquone/proguanil use, zoster vaccination, toxoplasmosis serostatus and Alzheimer's Disease, using 2 databases of claims data. The manuscript is well written and concise. The major concerns about the manuscript center around the indications of atovaquone/proguanil use, which would not typically be active against toxoplasmosis at doses given, and the lack of control for potential confounders in the analysis.

Strengths:

(1) Use of 2 databases of claims data.

(2) Unbiased review of medications associated with AD, which identified zoster vaccination associated with decreased risk of AD, replicating findings from other studies.

We thank the reviewer for the thoughtful assessment and for noting key strengths of our work, including (1) the use of two large national databases, and (2) the unbiased discovery approach that replicated the widely reported association between zoster vaccination and reduced Alzheimer’s disease (AD) risk. We agree that these features highlight the validity and reproducibility of the analytic framework.

Below we respond to the reviewer’s perceived weaknesses.

Weaknesses:

(1) Given that atovaquone/proguanil is likely to be given to a healthy population who is able to travel, concern that there are unmeasured confounders driving the association.

We agree that, prior to matching, A-P exposure may correlate with demographic or health-related differences (e.g., ability to travel). However, this potential bias was explicitly controlled for in the study design. Across all three age-stratified TriNetX cohorts, exposed and unexposed individuals were rigorously matched on all major established dementia risk factors: age, sex, race/ethnicity, smoking status, obesity, diabetes mellitus, and hypertension. Comparative analyses confirm that these risk factors are equivalently distributed at baseline.

As noted in our response to Reviewer #1, for any hypothetical unmeasured confounder to explain the results, it would need to satisfy three conditions simultaneously:

(1) Be capable of producing a 50–70% reduction in dementia incidence sustained over a decade and across three distinct age strata (ages 50–79);

(2) Be strongly associated with likelihood of receiving A-P;

(3) Remain entirely uncorrelated with age, sex, race/ethnicity, smoking, diabetes, hypertension, or obesity, all of which were rigorously matched and balanced at baseline.

No such factor has been proposed in the literature or by the reviewer. Thus, the concern remains hypothetical and unsupported by any measurable demographic or biological mechanism.

Importantly, empirical evidence contradicts the notion of a “healthy traveler” bias:

Emergency and inpatient encounter rates prior to exposure were comparable between A-P users and controls. Across the three age-stratified cohorts, emergency visits were similar or slightly higher among A-P users (EMER: 19.6% vs 16.4%, 19.9% vs 14.2%, 22.0% vs 14.8%), and inpatient encounters were effectively equivalent (IMP: 14.8% vs 15.2%, 17.7% vs 17.6%, 22.1% vs 22.2%). These patterns directly contradict the suggestion that A-P users were a healthier or less medically burdened population at baseline.

Prevalence of mild cognitive impairment was not lower among A-P users and was, in fact, slightly higher in the oldest cohort. Across the three age groups, baseline diagnoses of mild cognitive impairment (MCI) were comparable or slightly higher among exposed individuals (0.1% vs 0.1%, 0.3% vs 0.2%, 1.1% vs 0.6%). These data contradict the suggestion that A-P users had superior baseline cognition.

The strongest protective association occurred in the youngest stratum (age 50–59; HR 0.34). At this age, when nearly all individuals are sufficiently healthy to travel internationally, A-P uptake is the least likely to confound health status. A frailty-based “healthy traveler” hypothesis would instead predict the opposite pattern, with older adults showing the greatest apparent benefit, since health limitations are more likely to restrict travel in later life. In contrast, the protective association weakens with increasing age, empirically contradicting any explanation based on differential travel capacity.

In conclusion, the empirical evidence directly contradicts the existence of a ‘healthy traveler’ effect.

(2) The dose of atovaquone in atovaquone/proguanil is unlikely to be adequate suppression of toxo (much less for treatment/elimination of toxo), raising questions about the mechanism.

A few important points should address the reviewer’s concern:

In our cohorts, A-P was prescribed for malaria prophylaxis, as correctly noted. In this setting, it is taken for the entire duration of travel, plus several days before and after, typically resulting in many weeks of continuous exposure. This creates an unintentional but scientifically valuable natural experiment, in which a CNS-penetrating anti-Toxoplasma agent is administered for long durations.

Atovaquone is an established treatment for CNS toxoplasmosis, has strong CNS penetration, and is included in current clinical guidelines for acute toxoplasmosis in immunocompromised patients, although at higher doses. Because latent, asymptomatic CNS colonization is not treated in clinical practice, there are currently no data establishing the dose required to eliminate bradyzoite-stage Toxoplasma in immunocompetent individuals.

Our observations concern atovaquone–proguanil (A-P), a fixed-dose combination of atovaquone with proguanil, a DHFR inhibitor targeting a key metabolic pathway shared by malaria parasites and T. gondii. The combination has well-established synergistic effects in malaria prophylaxis and the same mechanism would be expected to enhance anti-Toxoplasma activity. This fixed-dose regimen has never been formally evaluated for toxoplasmosis treatment at prolonged durations or against latent bradyzoite infection.

Our hypothesis does not require or imply complete eradication of Toxoplasma. A clinically meaningful reduction in latent cyst burden among the subset of colonized individuals may be sufficient to alter long-term disease trajectories. Thus, a population-level decrease in dementia incidence does not require universal clearance of infection, but only partial suppression or reduction of parasite load in susceptible individuals, which is entirely compatible with the known pharmacology and duration of A-P exposure.

(3) Unmeasured bias in the small number of people who had toxoplasma serology in the TriNetX cohort.

The relatively small number of older adults with Toxoplasma serology stems from current clinical practice: serologic testing is mostly performed in women during reproductive years due to risks in pregnancy, whereas in older adults a positive result has no clinical consequence and therefore testing is rarely ordered.

Importantly, the seropositive and seronegative groups were drawn from the same underlying population of individuals who underwent serology testing, and the only difference between groups is the test result itself. Because the decision to order a test is made prior to and independent of the result, there is no plausible rationale by which the serology outcome (positive or negative) would introduce a bias favoring either group beyond the result of the test itself.

Furthermore, the two groups were here also rigorously matched on all major dementia risk factors, including age, sex, race/ethnicity, smoking, diabetes, hypertension, and BMI, and these characteristics are similarly distributed between groups. A small sample size does not imply bias; it simply reduces statistical power. Despite this limitation, the observed association (HR = 2.43, p = 0.001) remains strongly significant.

Finally, this result is consistent with multiple published studies reporting higher rates of Toxoplasma seropositivity among individuals with Alzheimer’s disease, dementia, and even mild cognitive impairment, such that our finding reinforces a broader and independently observed epidemiologic pattern. Importantly, in our cohort the serology testing clearly preceded dementia diagnosis, which supports the plausibility of a causal rather than merely correlative relationship between latent toxoplasmosis and cognitive decline.

To conclude our provisional response, we thank the editor and reviewers for raising points that will be further addressed and expanded upon in the discussion of the forthcoming revision. We welcome transparent scientific dialogue and acknowledge that, as with all observational research, residual confounding cannot be eliminated with absolute certainty. However, we disagree with the overall Assessment and emphasize that our findings—reproduced independently across two national health systems and three age-stratified cohorts, each rigorously matched on all major determinants of dementia risk, meet, and in many respects exceed, current standards for high-quality observational evidence.

Assigning the results to “residual confounding” requires more than speculation: it requires identification of a confounding factor that is (1) anchored in established dementia risk literature, (2) empirically plausible, and (3) quantitatively capable of generating a sustained ~50 percent reduction in dementia incidence over a decade. No such factor has been identified to date. We note that the assertion of “residual confounding” has not been supported by a specific, quantitatively plausible mechanism. A hypothetical bias that is both extremely large in effect and uncorrelated with all major risk factors is not statistically or biologically credible.

The explanation we propose, reduction in dementia risk through elimination of latent Toxoplasma gondii, is biologically grounded, directly supported by independent epidemiologic literature, and uniquely capable of accounting for all convergent observations in our data. No alternative hypothesis has been put forward that can plausibly explain these findings.

A revised version of the manuscript will be submitted shortly, incorporating expanded baseline analyses, with the strictest possible exclusion criteria (including congenital, vascular, chromosomal, and neurodegenerative disorders such as Parkinson’s disease), and complete tabulated comparisons. These data will further reinforce that the observed protective associations are not attributable to any measurable confounding. We also plan to enhance the discussion in order to address the points raised by the reviewers.

In light of the expanded analyses, any reservations expressed in the initial Assessment can now be re-evaluated on the basis of the empirical evidence. The findings reported in our study meet, and in several respects exceed, current epidemiologic standards for high-quality observational research, clearly warrant publication, and provide a robust scientific foundation for future mechanistic and interventional studies to determine whether elimination of latent toxoplasmosis can prevent or treat dementia.

to be changed to:

• 1. Have Content-Type: text/plain in the header.
• 1. Be externally accessible.
• 1. Not be password protected.
• 1. Not be behind a proxy or redirect.

eLife Assessment

This study presents important findings describing the early assembly of vascular basement membrane and how vascular cells switch from responding to cues provided by the external environment to those provided by self-assembled basement membrane. The evidence supporting the claims of the authors is convincing, with state-of-the-art microscopy and several different culture conditions examined. The work will be of interest to cell biologists studying the ECM, vascular development, as well as medical scientists focused on diseases that depend on vascular growth.

Reviewer #1 (Public review):

Summary:

Marchand et al. seek to understand how basement membrane (BM) is initially assembled around developing vasculature (and by extension basement membrane assembly generally progresses). To do this, they use an established cell culture system that is amenable to advanced microscopy techniques, including high-resolution fluorescence imaging and atomic force microscopy. This allows them to produce very high-quality imaging data that includes both protein localization and matrix topography in 3D. They show that fibronectin (FN) is remodeled as collagen IV (Col IV) assembles. Lysyl oxidase-like-2 (LOXL2) is needed for this process, and without it, BM does not form correctly, cells cannot adhere to BM, and cells also don't correctly form junctions with other cells.

Detailed Review:

The authors provide quantitative measures of BM fibril assembly at the earliest timepoints. They show two phases of growth - initial deposition, elongation, and interconnection of short fibers; the second is a significant thickening. As the BM forms, FN is immediately associated with filaments, but laminin and Col IV are not associated with fibers as detected by AFM. LOXL2 is associated with fibers, similar to FN. At a later time point, Col IV becomes associated with fibers, but laminin never does. Likely FN templates LOXL2, which crosslink Col IV into fibrils over time. Could the authors comment on how this data fits with in vivo data from model organisms? Also, I would like to know if they can uncouple LOXL2 from the FN matrix? Could you express a mutated form of LOXL2 that cannot interact with FN but still is able to crosslink Col IV?)

Depletion of LOXL2 supports this mechanism. Without it, Col IV and FN are uncoupled and accumulate as large aggregates rather than a complex fibrous network. Further, long-term thickening/growth of the fibronectin network is inhibited, indicating LOXL2 and/or the Col IV network positively reinforces fibronectin assembly. (Does LOXL2 directly act on FN, or is this effect dependent on Col IV? The nature of the molecular interactions between COL IV, LOXL2, and FN will be an important future research area.)

Next, Marchand et al. ask if failure to produce mature BM (induced by LOXL2 depletion) has consequences for underlying cells. They demonstrate a clear shift in the orientation of actin towards a linear alignment, and similarly, cells change shape from round to very elongated. Cell junctions also shifted to a linear arrangement in LOXL2 depletion. This fits with the known balance between cell-ECM and cell-cell adhesion. The changes in actin network and cell shape/adhesion correlate with a change in B1 integrin localization upon LOXL2 depletion. B1 integrin colocalized with sparse early FN fibers, but was absent from large FN aggregates that occur if LOXL2 is depleted. Similar reorganization of integrin adhesion components (FAK, Vinc, Pax). Clearly, there is feedback between BM assembly and cell junction organization. But I think the authors might emphasize to the reader that this normally reinforces the epithelial fate of these cells. It's less a balance and more like a tipping point. (Related to this section, I could not read Figure 4C graphs unless I enlarged them to 300%.)

Finally, they culture cells on micro groove plates, with or without LOXL2. The grooved substrate can orient the cells, and they show this is superseded by BM once it assembles. Without LOXL2 cells on micro-grooved substrates become disorganized, similar to their observation on flat surfaces (elongated cells, linear actin, etc.). This demonstrates a switch from external topographical cues to self-generated BM. This is consistent with the idea of reorganizing junctions to produce a stable epithelial tube. I was very interested in their 3D culture. The effect of BM assembly on tube diameter makes sense. But how does BM assembly support complex capillary functions like branching? (Can they force branching with targeted mutations that decouple integrin from the BM?) Is this a question of change to cell fate? (Are other remodeling enzymes activated after initial BM assembly that could support growth and/or branching?)

Reviewer #2 (Public review):

Summary:

The manuscript entitled "Adaptation of endothelial cells to microenvironment 1 topographical cues through lysyl oxidase like-2-mediated basement membrane scaffolding" by Marchand et al., aims to determine the impact of LOXL2 on the dynamic formation of vascular basement membranes (BMs).

Strengths:

This manuscript includes a nice combination of different methods and presents the results in an appropriate manner.

Furthermore, the results clearly demonstrate an impact of LOXL2 on collagen IV-fibronectin organization and topography. Finally, the proper arrangement of collagen IV-fibronectin impacts cell alignment.

Weaknesses:

An open question for this reviewer is what the real take-home message of the present study is? Can the authors deliver novel insight into BM formation transferable to the in vivo situation? Why do the authors not see a "real" BM? Could it be that in vivo endothelial cells do not build the vascular BM alone? Thus, are additional cell types needed? And what will happen then if LOXL2 expression is altered?

Major comments:

(1) Can the authors show that LOXL2 cross-links fibronectin and collagen IV?

(2) The authors stated that LOXL2 depletion affects cytoskeleton arrangements and cell shape. Could it be that this is simply a secondary effect mediated primarily through the altered cross-linking of fibronectin and collagen IV?

(3) Can the authors perform cell adhesion studies on CDMs derived from wild-type versus LOXL2-deficient cells?

(4) Line 226-230: Can the authors provide the proliferation data of wildtype and LOXL2-depleted cells supporting their Src and Akt activity findings?

(5) Line 298-299: The authors made a statement about laminin. Can the authors think of a co-culture of wild-type versus LOXL2-depleted endothelial cells in combination with pericytes or fibroblasts, as these cells contribute to the efficient assembly of a functional vascular basement membrane (10.1182/blood-2009-05-222364). Here, you can determine the impact of altered fibronectin-collagen IV cross-linking on laminin network formation. This will affect their conclusion in lines 299-304, as these facts are solely based on endothelial cells.

(6) Suggestion: can the authors supplement recombinant LOXL2 protein in its active version to the LOXL2-depleted endothelial cells to rescue the observed changes? And further compare LOXL2 enzymatic function with LOXL2 protein harbouring Zn instead of Cu, making it enzymatic inactive. Here, the authors might be able to strengthen their hypothesis that LOXL2 might bridge fibronectin and collagen IV or link both proteins.

(7) There are grammatical errors in the manuscript that the authors should work on.

Reviewer #3 (Public review):

This important study shows that basement membrane (BM) generation is a key event mediating cell 3D organization in response to microenvironmental cues. Such a mechanism participates in the endothelial cell capacity to organize into a capillary vessel segment through the shift of interactions with the interstitial ECM to interactions with vascular BM. This is particularly important for the developing, sprouting vasculature. The authors conclusively show, using TIRF and atomic force microscopy substantiated by 3D sprouting assays, that the lysyl oxidase Loxl2 plays a key role herein. With respect to translation into clinical practice, the dysregulation of adherens junctions and barrier properties associated with Loxl2 dysfunction mediated defects in BM supports its involvement in the progression of long-term microvascular diseases.

An outstanding question not answered in the current MS is how Loxl2 integrates into the Dll4-Notch mediated control of tip-stalk-phalanx cell differentiation in the developing (embryonic) vasculature. The authors focused a lot on Loxl2 loss of function; however, in a (patho)physiological context, Loxl2 gain of function would be relevant. Loxl2 is a hypoxia target and Loxl2 accumulates in the ECM upon hypoxic stress (as occurs during ischemic CVD, stroke/heart infarct). It would be interesting to know how Loxl2 gain-of-function impacts BM assembly, endothelial behavior, mechanosensing, and vessel angiogenic remodeling.

eLife Assessment

Amyotrophic lateral sclerosis (ALS) affects nerve cells in the brain and spinal cord. The authors' approach to use genetic code expansion to tag two ALS proteins associated with stress granules has value and should be useful in the ALS field. Parts of the work are well done, but there are concerns that the evidence is incomplete overall, and additional controls would strengthen the study.

Reviewer #1 (Public review):

Summary:

The authors utilize genetic code expansion to tag TDP-43 and G3BP1, and evaluate this protein tagging system (ANAP) compared to antibodies, and evaluate protein trafficking and stress granule formation in response to stress with sodium arsenite treatment. They find similar staining to antibodies in HeLa cells, mouse embryonic stem cells, and primary mouse cortical neurons. This is a useful study that demonstrates the utility of ANAP tagging to evaluate ALS proteins.

Strengths:

Rescue of cell survival by ANAP-tagged TDP-43 is compelling

Weaknesses:

While the ANAP-tagged proteins had similar distributions to antibody staining, there were some discrepancies that may be more explained by the technique than by novel findings, as the authors suggested. The inclusion of additional controls to evaluate this would be helpful.

Reviewer #2 (Public review):

Summary:

In this manuscript, Chen and colleagues describe a novel means of labeling two RNA-binding proteins, G3BP1 and TDP-43, using genetic code expansion. Overexpressed constructs that incorporate the intrinsically-fluorescent non-canonical amino acid Anap redistribute to cytoplasmic granules upon application of external stressors such as sodium arsenite. Similar labeling and redistribution of overexpressed G3BP1 and TDP-43 were observed in cultures of mouse primary neurons.

Strengths:

Genetic code expansion and non-canonical amino acid labeling have quite a few advantages over traditional fusion proteins for tracking protein redistribution in living cells. The authors show that they are able to label exogenous G3BP1 and TDP-43 with the non-canonical amino acid Anap and follow labeled proteins in living cells with and without stress.

Weaknesses:

The authors do not convincingly leverage the advantages of genetic code expansion in the current study. There is no specific question posed by the authors that can be or is answered using this approach, and several of the experiments lack critical controls. This is also not the first example of TDP-43 labeling by genetic code expansion (see PMID: 38290242). As a result, the study as a whole adds little to our understanding of protein trafficking and behavior under stress.

This is interesting because if one place refuse to accept refugees due their worries in failing to provide needed support for refugees, isn't this like an excuse to take responsibility?

This is interesting and connect to "The Wretched and The Beautiful." While in this article it presented an idea that refugees might be facing hostility due to their jobs and increase in competition, the narrator of the story said that the aliens are suppose to work for the things they get. Q: Are there any further clarification on this reason for histility toward refugees?

It recalls me about Asagai and Benethea in ARITS because they are also struggling with their identity

This is relevant to the text because in "The Wretched and The Beautiful", human's different attitudes toward two groups of aliens contribute to different threats that alien met.

what are the causes of having negative views of immigrants?

This remind me some Indian immigrant workers in Russia who could not recognize the size of screws cause some dangerous accidents in industires

eLife Assessment

This study proposes a potentially useful improvement on a popular fMRI method for quantifying representational similarity in brain measurements by focusing on representational strength at the single trial level and adding linear mixed effects modeling for group-level inference. The manuscript provides solid evidence of increased sensitivity with no loss of precision compared to more classic versions of the method. However, several assumptions are insufficiently motivated, and it is unclear to what extent the approach would generalize to other paradigms.

Reviewer #2 (Public review):

This paper proposes two changes to classic RSA, a popular method to probe neural representation in neuroimaging experiments: computing RSA at row/column level of RDM, and using linear mixed modeling to compute second level statistics, using the individual row/columns to estimate a random effect of stimulus. The benefit of the new method is demonstrated using simulations and a re-analysis of a prior fMRI dataset on object perception and memory encoding.

The author's claim that tRSA is a promising approach to perform more complete modeling of cogneuro data, and to conceptualize representation at the single trial/event level (cf Discussion section on P42), is appealing.

In their revised manuscript, the authors have addressed some previous concerns, now referencing more literature aiming to improve RSA and its associated statistical inferences, and providing more guidance on methodological considerations in the Discussion. However, I wish the authors had more extensively edited the Introduction to better contextualize the work and clarify the specific settings in which they see the method as being beneficial over classic RSA. For example, some of the limitations of cRSA mentioned on page 6, e.g. related to presenting the same stimuli to multiple subjects, seem to be quite specific to settings where the researcher expects differential responses across subjects to fundamentally alter the interpretation, rather than something that will just average out by repeatedly offering the same stimulus, or combining data across subjects. It's not clear to me how the switch from 'matrix-level' to 'row-level' analysis in tRSA necessarily addresses this problem. I would be very helpful if the authors would more explicitly outline what problem the row-level aspect of tRSA is solving; what problem statistical inference via LMM is solving; and walk the reader through a very specific use case (perhaps a toy version of the real-data experiment which is now at the end of the paper). Explaining the utility of tRSA for experimental settings in which assessing representational strength for a single-events is crucial would clarify the contribution of this new method better.

A few weaknesses mentioned in my previous review were not adequately addressed. To demonstrate the utility of the method on real neural recordings, only a single dataset is used with a quite complicated experimental design; it's not clear if there is any benefit of using tRSA on a simpler real dataset. Moreover, the cells of an RDM/RSM reflect pairwise comparisons between response patterns. Because the response patterns are repeatedly compared, the cells of this matrix are not independent of one another. While the authors show examples that failure to meet independence assumptions do not affect results in their specific dataset, it does not get acknowledged as a problem at a more fundamental level. Finally, while the paper now states that 'simulations and example tRSA code' are publicly available, the link points to the lab's general github page containing many lab repositories, in which I could not identify a specific repository related to this paper. This is disappointing given that the main goal of this manuscript is to provide a new method that they encourage others to use; a clear pointer to available code is only a minimal requirement to achieve that goal. A dedicated repository, including documentation, READMEs and tutorials/demo's to run simulations, compare methods, etc. would greatly enhance the paper's contribution.

Author response:

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

Reviewer #1 (Public review):

(1) I have to admit that it took a few hours of intense work to understand this paper and to even figure out where the authors were coming from. The problem setting, nomenclature, and simulation methods presented in this paper do not conform to the notation common in the field, are often contradictory, and are usually hard to understand. Most importantly, the problem that the paper is trying to solve seems to me to be quite specific to the particular memory study in question, and is very different from the normal setting of model-comparative RSA that I (and I think other readers) may be more familiar with.

We have revised the paper for clarity at all levels: motivation, application, and parameterization. We clarify that there is a large unmet need for using RSA in a trial-wise manner, and that this approach indeed offers benefits to any team interested in decoding trial-wise representational information linked to a behavioral responses, and as such is not a problem specific to a single memory study.

(2) The definition of "classical RSA" that the authors are using is very narrow. The group around Niko Kriegeskorte has developed RSA over the last 10 years, addressing many of the perceived limitations of the technique. For example, cross-validated distance measures (Walther et al. 2016; Nili et al. 2014; Diedrichsen et al. 2021) effectively deal with an uneven number of trials per condition and unequal amounts of measurement noise across trials. Different RDM comparators (Diedrichsen et al. 2021) and statistical methods for generalization across stimuli (Schütt et al. 2023) have been developed, addressing shortcomings in sensitivity. Finally, both a Bayesian variant of RSA (Pattern component modelling, (Diedrichsen, Yokoi, and Arbuckle 2018) and an encoding model (Naselaris et al. 2011) can effectively deal with continuous variables or features across time points or trials in a framework that is very related to RSA (Diedrichsen and Kriegeskorte 2017). The author may not consider these newer developments to be classical, but they are in common use and certainly provide the solution to the problems raised in this paper in the setting of model-comparative RSA in which there is more than one repetition per stimulus.

We appreciate the summary of relevant literature and have included a revised Introduction to address this bounty of relevant work. While much is owed to these authors, new developments from a diverse array of researchers outside of a single group can aid in new research questions, and should always have a place in our research landscape. We owe much to the work of Kriegeskorte’s group, and in fact, Schutt et al., 2023 served as a very relevant touchpoint in the Discussion and helped to highlight specific needs not addressed by the assessment of the “representational geometry” of an entire presented stimulus set. Principal amongst these needs is the application of trial-wise representational information that can be related to trial-wise behavioral responses and thus used to address specific questions on brain-behavior relationships. We invite the Reviewer to consider the utility of this shift with the following revisions to the Introduction.

Page 3. “Recently, methodological advancements have addressed many known limitations in cRSA. For example, cross-validated distance measures (e.g., Euclidean distance) have improved the reliability of representational dissimilarities in the presence of noise and trial imbalance (Walther et al., 2016; Nili et al., 2014; Diedrichsen et al., 2021). Bayesian approaches such as pattern component modeling (Diedrichsen, Yokoi, & Arbuckle, 2018) have extended representational approaches to accommodate continuous stimulus features or temporal variation. Further, model comparison RSA strategies (Diedrichsen et al., 2021) and generalization techniques across stimuli (Schütt et al., 2023) have improved sensitivity and inference. Nevertheless, a common feature shared across most of improvements is that they require stimuli repetition to examine the representational structure. This requirement limits their ability to probe brain-behavior questions at the level of individual events”.

Page 8. “While several extensions of RSA have addressed key limitations in noise sensitivity, stimulus variance, and modeling (e.g., Diedrichsen et al., 2021; Schütt et al., 2023), our tRSA approach introduces a new methodological step by estimating representational strength at the trial level. This accounts for the multi-level variance structure in the data, affords generalizability beyond the fixed stimulus set, and allows one to test stimulus- or trial-level modulations of neural representations in a straightforward way”.

Page 44. “Despite such prevalent appreciation for the neurocognitive relevance of stimulus properties, cRSA often does not account for the fact that the same stimulus (e.g., “basketball”) is seen by multiple subjects and produces statistically dependent data, an issue addressed by Schütt et al., 2023, who developed cross validation and bootstrap methods that explicitly model dependence across both subjects and stimulus conditions”.

(3) The stated problem of the paper is to estimate "representational strength" in different regions or conditions. With this, the authors define the correlation of the brain RDM with a model RDM. This metric conflates a number of factors, namely the variances of the stimulus-specific patterns, the variance of the noise, the true differences between different dissimilarities, and the match between the assumed model and the data-generating model. It took me a long time to figure out that the authors are trying to solve a quite different problem in a quite different setting from the model-comparative approach to RSA that I would consider "classical" (Diedrichsen et al. 2021; Diedrichsen and Kriegeskorte 2017). In this approach, one is trying to test whether local activity patterns are better explained by representation model A or model B, and to estimate the degree to which the representation can be fully explained. In this framework, it is common practice to measure each stimulus at least 2 times, to be able to estimate the variance of noise patterns and the variance of signal patterns directly. Using this setting, I would define 'representational strength" very differently from the authors. Assume (using LaTeX notation) that the activity patterns $y_j,n$ for stimulus j, measurement n, are composed of a true stimulus-related pattern ($u_j$) and a trial-specific noise pattern ($e_j,n$). As a measure of the strength of representation (or pattern), I would use an unbiased estimate of the variance of the true stimulus-specific patterns across voxels and stimuli ($\sigma^2_{u}$). This estimator can be obtained by correlating patterns of the same stimuli across repeated measures, or equivalently, by averaging the cross-validated Euclidean distances (or with spatial prewhitening, Mahalanobis distances) across all stimulus pairs. In contrast, the current paper addresses a specific problem in a quite specific experimental design in which there is only one repetition per stimulus. This means that the authors have no direct way of distinguishing true stimulus patterns from noise processes. The trick that the authors apply here is to assume that the brain data comes from the assumed model RDM (a somewhat sketchy assumption IMO) and that everything that reduces this correlation must be measurement noise. I can now see why tRSA does make some sense for this particular question in this memory study. However, in the more common model-comparative RSA setting, having only one repetition per stimulus in the experiment would be quite a fatal design flaw. Thus, the paper would do better if the authors could spell the specific problem addressed by their method right in the beginning, rather than trying to set up tRSA as a general alternative to "classical RSA".

At a general level, our approach rests on the premise that there is meaningful information present in a single presentation of a given stimulus. This assumption may have less utility when the research goals are more focused on estimating the fidelity of signal patterns for RSA, as in designs with multiple repetitions. But it is an exaggeration to state that such a trial-wise approach cannot address the difference between “true” stimulus patterns and noise. This trial-wise approach has explicit utility in relating trial-wise brain information to trial-wise behavior, across multiple cognitions (not only memory studies, as applied here). We have added substantial text to the Introduction distinguishing cRSA, which is widely employed, often in cases with a single repetition per stimulus, and model comparative methods that employ multiple repetitions. We clarify that we do not consider tRSA an alternative to the model comparative approach, and discuss that operational definitions of representational strength are constrained by the study design.

Page 3. “In this paper, we present an advancement termed trial-level RSA, or tRSA, which addresses these limitations in cRSA (not model comparison approaches) and may be utilized in paradigms with or without repeated stimuli”.

Page 4. “Representational geometry usually refers to the structure of similarities among repeated presentations of the same stimulus in the neural data (as captured in the brain RSM) and is often estimated utilizing a model comparison approach, whereas representational strength is a derived measure that quantifies how strongly this geometry aligns with a hypothesized model RSM. In other words, geometry characterizes the pattern space itself, while representational strength reflects the degree of correspondence between that space and the theoretical model under test”.

Finally, we clarified that in our simulation methods we assume a true underlying activity pattern and a random error pattern. The model RSM is computed based on the true pattern, whereas the brain RSM comes from the noisy pattern, not the model RSM itself.

Page 9. “Then, we generated two sets of noise patterns, which were controlled by parameters σ_A and σ_B , respectively, one for each condition”.

(4) The notation in the paper is often conflicting and should be clarified. The actual true and measured activity patterns should receive a unique notation that is distinct from the variances of these patterns across voxels. I assume that $\sigma_ijk$ is the noise variances (not standard deviation)? Normally, variances are denoted with $\sigma^2$. Also, if these are variances, they cannot come from a normal distribution as indicated on page 10. Finally, multi-level models are usually defined at the level of means (i.e., patterns) rather than at the level of variances (as they seem to be done here).

We have added notations for true and measured activity patterns to differentiate it from our notation for variance. We agree that multilevel models are usually defined at the level of means rather than at the level of variances and we include a Figure (Fig 1D) that describes the model in terms of the means. We clarify that the σ ($\sigma$) used in the manuscript were not variances/standard deviations themselves; rather, they were meant to denote components of the actual (multilevel) variance parameter. Each component was sampled from normal distributions, and they collectively summed up to comprise the final variance parameter for each trial. We have modified our notation for each component to the lowercase letter s to minimize confusion. We have also made our R code publicly available on our lab github, which should provide more clarity on the exact simulation process.

(5) In the first set of simulations, the authors sampled both model and brain RSM by drawing each cell (similarity) of the matrix from an independent bivariate normal distribution. As the authors note themselves, this way of producing RSMs violates the constraint that correlation matrices need to be positive semi-definite. Likely more seriously, it also ignores the fact that the different elements of the upper triangular part of a correlation matrix are not independent from each other (Diedrichsen et al. 2021). Therefore, it is not clear that this simulation is close enough to reality to provide any valuable insight and should be removed from the paper, along with the extensive discussion about why this simulation setting is plainly wrong (page 21). This would shorten and clarify the paper.

We have added justification of the mixed-effects model given the potential assumption violations. We caution readers to investigate the robustness of their models, and to employ permutation testing that does not make independence assumptions. We have also added checks of the model residuals and an example of permutation testing in the Appendix. Finally, we agree that the first simulation setting does not possess several properties of realistic RDMs/RSMs; however, we believe that there is utility in understanding the mathematical properties of correlations – an essential component of RSA – in a straightforward simulation where the ground truth is known, thus moving the simulation to Appendix 1.

(6) If I understand the second simulation setting correctly, the true pattern for each stimulus was generated as an NxP matrix of i.i.d. standard normal variables. Thus, there is no condition-specific pattern at all, only condition-specific noise/signal variances. It is not clear how the tRSA would be biased if there were a condition-specific pattern (which, in reality, there usually is). Because of the i.i.d. assumption of the true signal, the correlations between all stimulus pairs within conditions are close to zero (and only differ from it by the fact that you are using a finite number of voxels). If you added a condition-specific pattern, the across-condition RSA would lead to much higher "representational strength" estimates than a within-condition RSA, with obvious problems and biases.

The Reviewer is correct that the voxel values in the true pattern are drawn from i.i.d. standard normal distributions. We take the Reviewer’s suggestion of “condition-specific pattern” to mean that there could be a condition-voxel interaction in two non-mutually exclusive ways. The first is additive, essentially some common underlying multi-voxel pattern like [6, 34, -52, …, 8] for all condition A trials, and different one such pattern for condition B trials, etc. The second is multiplicative, essentially a vector of scaling factors [x1.5, x0.5, x0.8, …, x2.7] for all condition A trials, and a different one such vector for condition B trials, etc. Both possibilities could indeed affect tRSA as much as it would cRSA.

Importantly, If such a strong condition-specific pattern is expected, one can build a condition-specific model RDM using one-shot coding of conditions (see example figure; src: https://www.newbi4fmri.com/tutorial-9-mvpa-rsa), to either capture this interesting phenomenon or to remove this out as a confounding factor. This practice has been applied in multiple regression cRSA approaches (e.g., Cichy et al., 2013) and can also be applied to tRSA.

(7) The trial-level brain RDM to model Spearman correlations was analyzed using a mixed effects model. However, given the symmetry of the RDM, the correlations coming from different rows of the matrix are not independent, which is an assumption of the mixed effect model. This does not seem to induce an increase in Type I errors in the conditions studied, but there is no clear justification for this procedure, which needs to be justified.

We appreciate this important warning, and now caution readers to investigate the robustness of their models, and consider employing permutation testing that does not make independence assumptions. We have also added checks of the model residuals and an example of permutation testing in the supplement.

Page 46. “While linear mixed-effects modeling offers a powerful framework for analyzing representational similarity data, it is critical that researchers carefully construct and validate their models. The multilevel structure of RSA data introduces potential dependencies across subjects, stimuli, and trials, which can violate assumptions of independence if not properly modeled. In the present study, we used a model that included random intercepts for both subjects and stimuli, which accounts for variance at these levels and improves the generalizability of fixed-effect estimates. Still, there is a potential for systematic dependence across trials within a subject. To ensure that the model assumptions were satisfied, we conducted a series of diagnostic checks on an exemplar ROI (right LOC; middle occipital gyrus) in the Object Perception dataset, including visual inspection of residual distributions and autocorrelation (Appendix 3, Figure 13). These diagnostics supported the assumptions of normality, homoscedasticity, and conditional independence of residuals. In addition, we conducted permutation-based inference, similar to prior improvements to cRSA (Niliet al. 2014), using a nested model comparison to test whether the mean similarity in this ROI was significantly greater than zero. The observed likelihood ratio test statistic fell in the extreme tail of the null distribution (Appendix 3, Figure 14), providing strong nonparametric evidence for the reliability of the observed effect. We emphasize that this type of model checking and permutation testing is not merely confirmatory but can help validate key assumptions in RSA modeling, especially when applying mixed-effects models to neural similarity data. Researchers are encouraged to adopt similar procedures to ensure the robustness and interpretability of their findings”.

Exemplar Permutation Testing

To test whether the mean representational strength in the ROI right LOC (middle occipital gyrus) was significantly greater than zero, we used a permutation-based likelihood ratio test implemented via the permlmer function. This test compares two nested linear mixed-effects models fit using the lmer function from the lme4 package, both including random intercepts for Participant and Stimulus ID to account for between-subject and between-item variability.

The null model excluded a fixed intercept term, effectively constraining the mean similarity to zero after accounting for random effects:

ROI ~ 0 + (1 | Participant) + (1 | Stimulus)

The full model included the same random effects structure but allowed the intercept to be freely estimated:

ROI ~ 1 + (1 | Participant) + (1 | Stimulus)

By comparing the fit of these two models, we directly tested whether the average similarity in this ROI was significantly different from zero. Permutation testing (1,000 permutations) was used to generate a nonparametric p-value, providing inference without relying on normality assumptions. The full model, which estimated a nonzero mean similarity in the right LOC (middle occipital gyrus), showed a significantly better fit to the data than the null model that fixed the mean at zero (χ²(1) = 17.60, p = 2.72 × 10⁻⁵). The permutation-based p-value obtained from permlmer confirmed this effect as statistically significant (p = 0.0099), indicating that the mean similarity in this ROI was reliably greater than zero. These results support the conclusion that the right LOC contains representational structure consistent with the HMAXc2 RSM. A density plot of the permuted likelihood ratio tests is plotted along with the observed likelihood ratio test in Appendix 3 Figure 14.

(8) For the empirical data, it is not clear to me to what degree the "representational strength" of cRSA and tRSA is actually comparable. In cRSA, the Spearman correlation assesses whether the distances in the data RSM are ranked in the same order as in the model. For tRSA, the comparison is made for every row of the RSM, which introduces a larger degree of flexibility (possibly explaining the higher correlations in the first simulation). Thus, could the gains presented in Figure 7D not simply arise from the fact that you are testing different questions? A clearer theoretical analysis of the difference between the average row-wise Spearman correlation and the matrix-wise Spearman correlation is urgently needed. The behavior will likely vary with the structure of the true model RDM/RSM.

We agree that the comparability between mean row-wise Spearman correlations and the matrix-wise Spearman correlation is needed. We believe that the simulations are the best approach for this comparison, since they are much more robust than the empirical dataset and have the advantage of knowing the true pattern/noise levels. We expand on our comparison of mean tRSA values and matrix-wise Spearman correlations on page 42.

Page 42. “Although tRSA and cRSA both aim to quantify representational strength, they differ in how they operationalize this concept. cRSA summarizes the correspondence between RSMs as a single measure, such as the matrix-wise Spearman correlation. In contrast, tRSA computes such correspondence for each trial, enabling estimates at the level of individual observations. This flexibility allows trial-level variability to be modeled directly, but also introduces subtle differences in what is being measured. Nonetheless, our simulations showed that, although numerical differences occasionally emerged—particularly when comparing between-condition tRSA estimates to within-condition cRSA estimates—the magnitude of divergence was small and did not affect the outcome of downstream statistical tests”.

(9) For the real data, there are a number of additional sources of bias that need to be considered for the analysis. What if there are not only condition-specific differences in noise variance, but also a condition-specific pattern? Given that the stimuli were measured in 3 different imaging runs, you cannot assume that all measurement noise is i.i.d. - stimuli from the same run will likely have a higher correlation with each other.

We recognize the potential of condition-specific patterns and chose to constrain the analyses to those most comparable with cRSA. However, depending on their hypotheses, researchers may consider testing condition RSMs and utilizing a model comparison approach or employ the z-scored approach, as employed in the simulations above. Regarding the potential run confounds, this is always the case in RSA and why we exclude within-run comparisons. We have also added to the Discussion the suggestion to include run as a covariate in their mixed-effects models. However, we do not employ this covariate here as we preferred the most parsimonious model to compare with cRSA.

Page 46 - 47. “Further, while analyses here were largely employed to be comparable with cRSA, researchers should consider taking advantage of the flexibility of the mixed-effects models and include co variates of non-interest (run, trial order etc.)”.

(10) The discussion should be rewritten in light of the fact that the setting considered here is very different from the model-comparative RSA in which one usually has multiple measurements per stimulus per subject. In this setting, existing approaches such as RSA or PCM do indeed allow for the full modelling of differences in the "representational strength" - i.e., pattern variance across subjects, conditions, and stimuli.

We agree that studies advancing designs with multiple repetitions of a given stimulus image are useful in estimating the reliability of concept representations. We would argue however that model comparison in RSA is not restricted to such data. Many extant studies do not in fact have multiple repetitions per stimulus per subject (Wang et al., 2018 https://doi.org/10.1088/1741-2552/abecc3, Gao et al, 2022 https://doi.org/10.1093/cercor/bhac058, Li et al, 2022 https://doi.org/10.1002/hbm.26195, Staples & Graves, 2020 https://doi.org/10.1162/nol_a_00018) that allow for that type of model-comparative approach. While beneficial in terms of noise estimation, having multiple presentations was not a requirement for implementing cRSA (Kriegeskorte, 2008 https://doi.org/10.3389/neuro.06.004.2008). The aim of this manuscript is to introduce the tRSA approach to the broad community of researchers whose research questions and datasets could vary vastly, including but not limited to the number of repeated presentations and the balance of trial counts across conditions.

(11) Cross-validated distances provide a powerful tool to control for differences in measurement noise variances and possible covariances in measurement noise across trials, which has many distinct advantages and is conceptually very different from the approach taken here.

We have added language on the value of cross-validation approaches to RSA in the Discussion:

Page 47. “Additionally, we note that while our proposed tRSA framework provides a flexible and statistically principled approach for modeling trial-level representational strength, we acknowledge that there are alternative methods for addressing trial-level variability in RSA. In particular, the use of cross-validated distance metrics (e.g., crossnobis distance) has become increasingly popular for controlling differences in measurement noise variance and accounting for possible covariance structures across trials (Walther et al., 2016). These metrics offer several advantages, including unbiased estimation of representational dissimilarities under Gaussian noise assumptions and improved generalization to unseen data. However, cross-validated distances are conceptually distinct from the approach taken here: whereas cross-validation aims to correct for noise-related biases in representational dissimilarity matrices, our trial-level RSA method focuses on estimating and modeling the variability in representation strength across individual trials using mixed-effects modeling. Rather than proposing a replacement for cross-validated RSA, tRSA adds a complementary tool to the methodological toolkit—one that supports hypothesis-driven inference about condition effects and trial-level covariates, while leveraging the full structure of the data”.

(12) One of the main limitations of tRSA is the assumption that the model RDM is actually the true brain RDM, which may not be the case. Thus, in theory, there could be a different model RDM, in which representational strength measures would be very different. These differences should be explained more fully, hopefully leading to a more accessible paper.

Indeed, the chosen model RSM may not be the true RSM, but as the noise level increases the correlation between RSMs practically becomes zero. In our simulations we assume this to be true as a straightforward way to manipulate the correspondence between the brain data and the model. However, just like cRSA, tRSA is constrained by the model selections the researchers employ. We encourage researchers to have carefully considered theoretically-motivated models and, if their research questions require, consider multiple and potentially competing models. Furthermore, the trial-wise estimates produced by tRSA encourage testing competing models within the multiple regression framework. We have added this language to the Discussion.

Page 46. ..”choose their model RSMs carefully. In our simulations, we designed our model RSM to be the “true” RSM for demonstration purposes. However, researchers should consider if their models and model alternatives”.

Pages 45-46. “While a number of studies have addressed the validity of measuring representational geometry using designs with multiple repetitions, a conceptual benefit of the tRSA approach is the reliance on a regression framework that engenders the testing of competing conceptual models of stimulus representation (e.g., taxonomic vs. encyclopedic semantic features, as in Davis et al., 2021)”.

Reviewer #2 (Public review):

(1)  While I generally welcome the contribution, I take some issue with the accusatory tone of the manuscript in the Introduction. The text there (using words such as 'ignored variances', 'errouneous inferences', 'one must', 'not well-suited', 'misleading') appears aimed at turning cRSA in a 'straw man' with many limitations that other researchers have not recognized but that the new proposed method supposedly resolves. This can be written in a more nuanced, constructive manner without accusing the numerous users of this popular method of ignorance.

We apologize for the unintended accusatory tone. We have clarified the many robust approaches to RSA and have made our Introduction and Discussion more nuanced throughout (see also 3, 11 and16).

(2) The described limitations are also not entirely correct, in my view: for example, statistical inference in cRSA is not always done using classic parametric statistics such as t-tests (cf Figure 1): the rsatoolbox paper by Nili et al. (2014) outlines non-parametric alternatives based on permutation tests, bootstrapping and sign tests, which are commonly used in the field. Nor has RSA ever been conducted at the row/column level (here referred to by the authors as 'trial level'; cf King et al., 2018).

We agree there are numerous methods that go beyond cRSA addressing these limitations and have added discussion of them into our manuscript as well as an example analysis implementing permutation tests on tRSA data (see response to 7). We thank the reviewer for bringing King et al., 2014 and their temporal generalization method to our attention, we added reference to acknowledge their decoding-based temporal generalization approach.

Page 8. “It is also important to note that some prior work has examined similarly fine-grained representations in time-resolved neuroimaging data, such as the temporal generalization method introduced by King et al. (see King & Dehaene, 2014). Their approach trains classifiers at each time point and tests them across all others, resulting in a temporal generalization matrix that reflects decoding accuracy over time. While such matrices share some structural similarity with RSMs, they do not involve correlating trial-level pattern vectors with model RSMs nor do their second-level models include trial-wise, subject-wise, and item-wise variability simultaneously”.

(3) One of the advantages of cRSA is its simplicity. Adding linear mixed effects modeling to RSA introduces a host of additional 'analysis parameters' pertaining to the choice of the model setup (random effects, fixed effects, interactions, what error terms to use) - how should future users of tRSA navigate this?

We appreciate the opportunity to offer more specific proscriptions for those employing a tRSA technique, and have added them to the Discussion:

Page 46. “While linear mixed-effects modeling offers a powerful framework for analyzing representational similarity data, it is critical that researchers carefully construct and validate their models and choose their model RSMs carefully. In our simulations, we designed our model RSM to be the “true” RSM for demonstration purposes. However, researchers should consider if their models and model alternatives. However, researchers should always consider if their models match the goals of their analysis, including 1) constructing the random effects structure that will converge in their dataset and 2) testing their model fits against alternative structures (Meteyard & Davies, 2020; Park et al., 2020) and 3) considering which effects should be considered random or fixed depending on their research question”.

(4) Here, only a single real fMRI dataset is used with a quite complicated experimental design for the memory part; it's not clear if there is any benefit of using tRSA on a simpler real dataset. What's the benefit of tRSA in classic RSA datasets (e.g., Kriegeskorte et al., 2008), with fixed stimulus conditions and no behavior?

To clarify, our empirical approach uses two different tasks: an Object Perception task more akin to the classic RSA datasets employing passive viewing, and a Conceptual Retrieval task that more directly addresses the benefits of the trialwise approach. We felt that our Object Perception dataset is a simpler empirical fMRI dataset without explicit task conditions or a dichotomous behavioral outcome, whereas the Retrieval dataset is more involved (though old/new recognition is the most common form of memory retrieval testing) and  dependent on behavioral outcomes. However, we recognize the utility of replication from other research groups and do invite researchers to utilize tRSA on their datasets.

(5) The cells of an RDM/RSM reflect pairwise comparisons between response patterns (typically a brain but can be any system; cf Sucholutsky et al., 2023). Because the response patterns are repeatedly compared, the cells of this matrix are not independent of one another. Does this raise issues with the validity of the linear mixed effects model? Does it assume the observations are linearly independent?

We recognize the potential danger for not meeting model assumptions. Though our simulation results and model checks suggest this is not a fatal flaw in the model design, we caution readers to investigate the robustness of their models, and consider employing permutation testing that does not make independence assumptions. We have also added checks of the model residuals and an example of permutation testing in the Appendix. See response to R1.

(6) The manuscript assumes the reader is familiar with technical statistical terms such as Type I/II error, sensitivity, specificity, homoscedasticity assumptions, as well as linear mixed models (fixed effects, random effects, etc). I am concerned that this jargon makes the paper difficult to understand for a broad readership or even researchers currently using cRSA that might be interested in trying tRSA.

We agree this jargon may cause the paper to be difficult to understand. We have expanded/added definitions to these terms throughout the methods and results sections.

Page 12. “Given data generated with 𝑠_{𝑐𝑜𝑛𝑑,𝐴} = 𝑠_{𝑐𝑜𝑛𝑑,B}, the correct inference should be a failure to reject the null hypothesis of ; any significant () result in either direction was considered a false positive (spurious effect, or Type I error). Given data generated with , the inference was considered correct if it rejected the null hypothesis of  and yielded the expected sign of the estimated contrast (b_B-𝐴<0). A significant result with the reverse sign of the estimated contrast (b_B-𝐴<0) was considered a Type I error, and a nonsignificant (𝑝 ≥ 0.05) result was considered a false negative (failure to detect a true effect, or Type II error)”.

Page 2. “Compared to cRSA, the multi-level framework of tRSA was both more theoretically appropriate and significantly sensitive (better able to detect) to true effects”.

Page 25.”The performance of cRSA and tRSA were quantified with their specificity (better avoids false positives, 1 - Type I error rate) and sensitivity (better avoids false negatives 1 - Type II error rate)”.

Page 6. “One of the fundamental assumptions of general linear models (step 4 of cRSA; see Figure 1D) is homoscedasticity or homogeneity of variance — that is, all residuals should have equal variance” .

Page11. “Specifically, a linear mixed-effects model with a fixed effect  of condition (which estimates the average effect across the entire sample, capturing the overall effect of interest) and random effects of both subjects and stimuli (which model variation in responses due to differences between individual subjects and items, allowing generalization beyond the sample) were fitted to tRSA estimates via the `lme4 1.1-35.3` package in R (Bates et al., 2015), and p-values were estimated using Satterthwaites’s method via the `lmerTest 3.1-3` package (Kuznetsova et al., 2017)”.

(7) I could not find any statement on data availability or code availability. Given that the manuscript reuses prior data and proposes a new method, making data and code/tutorials openly available would greatly enhance the potential impact and utility for the community.

We thank the reviewer for raising our oversight here. We have added our code and data availability statements.

Page 9. “Data is available upon request to the corresponding author and our simulations and example tRSA code is available at https://github.com/electricdinolab”.

Reviewer #1 (Recommendations for the authors):

(13) Page 4: The limitations of cRSA seem to be based on the assumption that within each different experimental condition, there are different stimuli, which get combined into the condition. The framework of RSA, however, does not dictate whether you calculate a condition x condition RDM or a larger and more complete stimulus x stimulus RDM. Indeed, in practice we often do the latter? Or are you assuming that each stimulus is only shown once overall? It would be useful at this point to spell out these implicit assumptions.

We agree that stimulus x stimulus RDMs can be constructed and are often used. However, as we mentioned in the Introduction, researchers are often interested in the difference between two (or more) conditions, such as “remembered” vs. “forgotten” (Davis et al., https://doi.org/10.1093/cercor/bhaa269) or “high cognitive load” vs. “low cognitive load” (Beynel et al., https://doi.org/10.1523/JNEUROSCI.0531-20.2020). In those cases, the most common practice with cRSA is to construct condition-specific RDMs, compute cRSA scores separately for each condition, and then compare the scores at the group level. The number of times each stimulus gets presented does not prevent one from creating a model RDM that has the same rows and columns as the brain RDM, either in the same condition (“high load”) or across different conditions.

(14) Page 5: The difference between condition-level and stimulus-level is not clear. Indeed, this definition seems to be a function of the exact experimental design and is certainly up for interpretation. For example, if I conduct a study looking at the activity patterns for 4 different hand actions, each repeated multiple times, are these actions considered stimuli or conditions?

We have added clarifying language about what is considered stimuli vs conditions. Indeed, this will depend on the specific research questions being employed and will affect how researchers construct their models. In this specific example, one would most likely consider each different hand action a condition, treating them as fixed effects rather than random effects, given their very limited number and the lack of need to generalize findings to the broader “hand actions” category.

Page 5. “Critically, the distinction between condition-level and stimulus level is not always clear as researchers may manipulate stimulus-level features themselves. In these cases, what researchers ultimately consider condition-level and stimulus-level will depend on their specific research questions. For example, researchers intending to study generalized object representation may consider object category a stimulus-level feature, while researchers interested in if/how object representation varies by category may consider the same category variable condition-level”.

(15) Page 5: The fact that different numbers of trials / different levels of measurement noise / noise-covariance of different conditions biases non-cross-validated distances is well known and repeatedly expressed in the literature. We have shown that cross-validation of distances effectively removes such biases - of course, it does not remove the increased estimation variability of these distances (for a formal analysis of estimation noise on condition patterns and variance of the cross-nobis estimator, see (Diedrichsen et al. 2021)).

We thank the reviewer for drawing our attention to this literature and have added discussions of these methods.

(16). Page 5: "Most studies present subjects with a fixed set of stimuli, which are supposedly samples representative of some broader category". This may be the case for a certain type of RSA experiments in the visual domain, but it would be unfair to say that this is a feature of RSA studies in general. In most studies I have been involved in, we use a "stimulus" x "stimulus" RDM.

We have edited this sentence to avoid the “most” characterization. We also added substantial text to the introduction and discussion distinguishing cRSA, which is nonetheless widely employed, especially in cases with a single repetition per stimulus (Macklin et al., 2023, Liu et al, 2024) and the model comparative method and explicitly stating that we do not consider tRSA an alternative to the model comparative approach.

(17). Page 5: I agree that "stimuli" should ideally be considered a random effect if "stimuli" can be thought of as sampled from a larger population and one wants to make inferences about that larger population. Sometimes stimuli/conditions are more appropriately considered a fixed effect (for example, when studying the response to stimulation of the 5 fingers of the right hand). Techniques to consider stimuli/conditions as a random effect have been published by the group of Niko Kriegeskorte (Schütt et al. 2023).

Indeed, in some cases what may be thought of as “stimuli” would be more appropriately entered into the model as a fixed effect; such questions are increasingly relevant given the focus on item-wise stimulus properties (Bainbridge et al., Westfall & Yarkoni). We have added text on this issue to the Discussion and caution researchers to employ models that most directly answer their research questions.

Page 46. “However, researchers should always consider if their models match the goals of their analysis, including 1) constructing the random effects structure that will converge in their dataset and 2) testing their model fits against alternative structures (Meteyard & Davies, 2020; Park et al., 2020) and 3) considering which effects should be considered random or fixed depending on their research question. An effect is fixed when the levels represent the specific conditions of theoretical interest (e.g., task condition) and the goal is to estimate and interpret those differences directly. In contrast, an effect is random when the levels are sampled from a broader population (e.g., subjects) and the goal is to account for their variability while generalizing beyond the sample tested. Note that the same variable (e.g., stimuli) may be considered fixed or random depending on the research questions”.

(18) Page 6: It is correct that the "classical" RSA depends on a categorical assignment of different trials to different stimuli/conditions, such that a stimulus x stimulus RDM can be computed. However, both Pattern Component Modelling (PCM) and Encoding models are ideally set up to deal with variables that vary continuously on a trial-by-trial or moment-by-moment basis. tRSA should be compared to these approaches, or - as it should be clarified - that the problem setting is actually quite a different one.

We agree that PCM and encoding models offer a flexible approach and handle continuous trial-by-trial variables. We have clarified the problem setting in cRSA is distinct on page 6, and we have added the robustness of encoding models and their limitations to the Discussion.

Page 6. “While other approaches such as Pattern Component Modeling (PCM) (Diedrichsen et al., 2018) and encoding models (Naselaris et al., 2011) are well-suited to analyzing variables that vary continuously on a trial-by-trial or moment-by-moment basis, these frameworks address different inferential goals. Specifically, PCM and encoding models focus on estimating variance components or predicting activation from features, while cRSA is designed to evaluate representational geometry. Thus, cRSA as well as our proposed approach address a problem setting distinct from PCM and encoding models”.

(19) Page 8: "Then, we generated two noise patterns, which were controlled by parameters 𝜎 𝐴 and 𝜎𝐵, respectively, one for each condition." This makes little sense to me. The noise patterns should be unique to each trial - you should generate n_a + n_b noise patterns, no?

We clarify that the “noise patterns” here are n_voxel x n_trial in size; in other words, all trial-level noise patterns are generated together and each trial has their own unique noise pattern. We have revised our description as “two sets of noise patterns” for clarity starting on page 9.

(20) Page 9: First, I assume if this is supposed to be a hierarchical level model, the "noise parameters" here correspond to variances? Or do these \sigma values mean to signify standard deviations? The latter would make little sense. Or is it the noise pattern itself?

As clarified in 4., the σ values are meant to denote hierarchical components of the composite standard deviation; we have updated our notation to use lower case letter s instead for clarity.

(21) Page 10: your formula states "𝜎_{𝑠𝑢𝑏𝑗}~ 𝙽(0, 0.5^2)". This conflicts with your previous mention that \sigmas are noise "levels" are they the noise patterns themselves now? Variances cannot be normally distributed, as they cannot be negative.

As clarified in 4., the σ values are meant to denote hierarchical components of the composite standard deviation; we have updated our notation to use lower case letter s instead for clarity.

(22) Page 13: What was the task of the subject in the Memory retrieval task? Old/new judgements relative to encoding of object perception?

We apologize for the lack of clarity about the Memory Retrieval task and have added that information and clarified that the old/new judgements were relative to a separate encoding phase, the brain data for which has been reported elsewhere.

Page 14. “Memory Retrieval took place one day after Memory Encoding and involved testing participants’ memory of the objects seen in the Encoding phase. Neural data during the Encoding phase has been reported elsewhere. In the main Memory Retrieval task, participants were presented with 144 labels of real-world objects, of which 114 were labels for previously seen objects and 30 were unrelated novel distractors. Participants performed old/new judgements, as well as their confidence in those judgements on a four-point scale (1 = Definitely New, 2 = Probably New, 3 = Probably Old, 4 = Definitely Old)”.

(23) Page 13: If "Memory Retrieval consisted of three scanning runs", then some of the stimulus x stimulus correlations for the RSM must have been calculated within a run and some between runs, correct? Given that all within-run estimates share a common baseline, they share some dependence. Was there a systematic difference between the within-run and the between-run correlations?

We have clarified in this portion of the methods that within run comparisons were excluded from our analyses. We also double-checked that the within-run exclusion was included in the description of the Neural RSMs.

Page 14. “Retrieval consisted of three scanning runs, each with 38 trials, lasting approximately 9 minutes and 12 seconds (within-run comparisons were later excluded from RSA analyses)”.

Page 18. “This was done by vectorizing the voxel-level activation values within each region and calculating their correlations using Pearson’s r, excluding all within-run comparisons.”

(24) Page 20: It is not clear why the mean estimate of "representational strength" (i.e., model-brain RSM correlations) is important at all. This comes back to Major point #2, namely that you are trying to solve a very different problem from model-comparative RSA.

We have clarified that our approach is not an alternative to model-comparative RSA, and that depending on the task constraints researchers may choose to compare models with tRSA or other approaches requiring stimulus repetition (see 3).

(25) Page 21: I believe the problems of simulating correlation matrices directly in the way that the authors in their first simulation did should be well known and should be moved to an appendix at best. Better yet, the authors could start with the correct simulation right away.

We agree the paper is more concise with these simulations being moved to the appendix and more briefly discussed. We have implemented these changes (Appendix 1). However, we are not certain that this problem is unknown, and have several anecdotes of researchers inquiring about this “alternative” approach in talks with colleagues, thus we do still discuss the issues with this method.

(26) Page 26: Is the "underlying continuous noise variable 𝜎𝑡𝑟𝑖𝑎𝑙 that was measured by 𝑣𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑 " the variance of the noise pattern or the noise pattern itself? What does it mean it was "measured" - how?

𝜎𝑡𝑟𝑖𝑎𝑙 is a vector of standard deviations for different trials, and 𝜎𝑡𝑟𝑖𝑎𝑙 i would be used to generate the noise patterns for trial i. v_measured is a hypothetical measurement of trial-level variability, such as “memorability” or “heartbeat variability”. We have revised our description to clarify our methods.

Reviewer #2 (Recommendations for the authors):

(8) It would be helpful to provide more clarity earlier on in the manuscript on what is a 'trial': in my experience, a row or column of the RDM is usually referred to as 'stimulus condition', which is typically estimated on multiple trials (instances or repeats) of that stimulus condition (or exemplars from that stimulus class) being presented to the subject. Here, a 'trial' is both one measurement (i.e., single, individual presentation of a stimulus) and also an entry in the RDM, but is this the most typical scenario for cRSA? There is a section in the Discussion that discusses repetitions, but I would welcome more clarity on this from the get-go.

We have added discussion of stimulus repetition methods and datasets to the Introduction and clarified our use of the terms.

Page 8. “Critically, in single-presentation designs, a “trial” refers to one stimulus presentation, and corresponds to a row or column in the RSM. In studies with repeated stimuli, these rows are often called “conditions” and may reflect aggregated patterns across trials. tRSA is compatible with both cases: whether rows represent individual trials or averaged trials that create “conditions”, tRSA estimates are computed at the row level”.

(9) The quality of the results figures can be improved. For example, axes labels are hard to read in Figure 3A/B, panels 3C/D are hard to read in general. In Figure 7E, it's not possible to identify the 'dark red' brain regions in addition to the light red ones.

We thank the reviewer for raising these and have edited the figures to be more readable in the manner suggested.

(10) I would be interested to see a comparison between tRSA and cRSA in other fMRI (or other modality) datasets that have been extensively reported in the literature. These could be the original Kriegeskorte 96 stimulus monkey/fMRI datasets, commonly used open datasets in visual perception (e.g., THINGS, NSD), or the above-mentioned King et al. dataset, which has been analyzed in various papers.

We recognize the great utility of replication from other research groups and do invite researchers to utilize tRSA on their datasets.

(11) On P39, the authors suggest 'researchers can confidently replace their existing cRSA analysis with tRSA': Please discuss/comment on how researchers should navigate the choice of modeling parameters in tRSA's linear mixed effects setting.

We have added discussion of the mixed-effects parameters and the various and encourage researchers to follow best practices for their model selection.

Page 46. “However, researchers should always consider if their models match the goals of their analysis, including 1) constructing the random effects structure that will converge in their dataset and 2) testing their model fits against alternative structures (Meteyard & Davies, 2020; Park et al., 2020) and 3) considering which effects should be considered random or fixed depending on their research question”.

(12) The final part of the Results section, demonstrating the tRSA results for the continuous memorability factor in the real fMRI data, could benefit from some substantiation/elaboration. It wasn't clear to me, for example, to what extent the observed significant association between representational strength and item memorability in this dataset is to be 'believed'; the Discussion section (p38). Was there any evidence in the original paper for this association? Or do we just assume this is likely true in the brain, based on prior literature by e.g. Bainbridge et al (who probably did not use tRSA but rather classic methods)?

Indeed, memorability effects have been replicated in the literature, but not using the tRSA method. We have expanded our discussion to clarify the relationship of our findings and the relevant literature and methods it has employed.

Page 38. “Critically, memorability is a robust stimulus property that is consistent across participants and paradigms (Bainbridge, 2022). Moreover, object memorability effects have been replicated using a variety of methods aside from tRSA, including univariate analyses and representational analyses of neural activity patterns where trial-level neural activity pattern estimates are correlated directly with object memorability (Slayton et al, 2025).”

(13) The abstract could benefit from more nuance; I'm not sure if RSA can indeed be said to be 'the principal method', and whether it's about assessing 'quality' of representations (more commonly, the term 'geometry' or 'structure' is used).

We have edited the abstract to reflect the true nuisance in the current approaches.

Abstract. Neural representation refers to the brain activity that stands in for one’s cognitive experience, and in cognitive neuroscience, a prominent method of studying neural representations is representational similarity analysis (RSA). While there are several recent advances in RSA, the classic RSA (cRSA) approach examines the structure of representations across numerous items by assessing the correspondence between two representational similarity matrices (RSMs): usually one based on a theoretical model of stimulus similarity and the other based on similarity in measured neural data.

(14) RSA is also not necessarily about models vs. neural data; it can also be between two neural systems (e.g., monkey vs. human as in Kriegeskorte et al., 2008) or model systems (see Sucholutsky et al., 2023). This statement is also repeated in the Introduction paragraph 1 (later on, it is correctly stated that comparing brain vs. model is most likely the 'most common' approach).

We have added these examples in our introduction to RSA.

Page 3.”One of the central approaches for evaluating information represented in the brain is representational similarity analysis (RSA), an analytical approach that queries the representational geometry of the brain in terms of its alignment with the representational geometry of some cognitive model (Kriegeskorte et al., 2008; Kriegeskorte & Kievit, 2013), or, in some cases, compares the representational geometry of two neural systems (e.g., Kriegeskorte et al., 2008) or two model systems (Sucholutsky et al., 2023)”.

(15) 'theoretically appropriate' is an ambiguous statement, appropriate for what theory?

We apologize for the ambiguous wording, and have corrected the text:

Page 11. “Critically, tRSA estimates were submitted to a mixed-effects model which is statistically appropriate for modeling the hierarchical structure of the data, where observations are nested within both subjects and stimuli (Baayen et al., 2008; Chen et al., 2021)”.

(16) I found the statement that cRSA "cannot model representation at the level of individual trials" confusing, as it made me think, what prohibits one from creating an RDM based on single-trial responses? Later on, I understood that what the authors are trying to say here (I think) is that cRSA cannot weigh the contributions of individual rows/columns to the overall representational strength differently.

We thank the reviewer for their clarifying language and have added it to this section of the manuscript.

“Abstract. However, because cRSA cannot weigh the contributions of individual trials (RSM rows/columns), it is fundamentally limited in its ability to assess subject-, stimulus-, and trial-level variances that all influence representation”.

(17) Why use "RSM" instead of "RDM"? If the pairwise comparison metric is distance-based (e..g, 1-correlation as described by the authors), RDM is more appropriate.

We apologize for the error, and have clarified the Methods text:

Page3-4. First, brain activity responses to a series of N trials are compared against each other (typically using Pearson’s r) to form an N×N representational similarity matrix.

(18) Figure 2: please write 'Correlation estimate' in the y-axis label rather than 'Estimate'.

We have edited the label in Figure 2.

(19) Page 6 'leaving uncertain the directionality of any findings' - I do not follow this argument. Obviously one can generate an RDM or RSM from vector v or vector -v. How does that invalidate drawing conclusions where one e.g., partials out the (dis)similarity in e.g., pleasantness ratings out of another RDM/RSM of interest?

We agree such an approach does not invalidate the partial method; we have clarified what we mean by “directionality”.

Page 8. ”For instance, even though a univariate random variable , such as pleasantness ratings, can be conveniently converted to an RSM using pairwise distance metrics (Weaverdyck et al., 2020), the very same RSM would also be derived from the opposite random variable , leaving uncertain of the directionality (or if representation is strongest for pleasant or unpleasant items) of any findings with the RSM (see also Bainbridge & Rissman, 2018)”.

(20) P7 'sampled 19900 pairs of values from a bi-variate normal distribution', but the rows/columns in an RDM are not independent samples - shouldn't this be included in the simulation? I.e., shouldn't you simulate first the n=200 vectors, and then draw samples from those, as in the next analysis?

This section has been moved to Appendix 1 (see responses to Reviewer 1.13).

(21) Under data acquisition, please state explicitly that the paper is re-using data from prior experiments, rather than collecting data anew for validating tRSA.

We have clarified this in the data acquisition section.

Page 13. “A pre-existing dataset was analyzed to evaluate tRSA. Main study findings have been reported elsewhere (S. Huang, Bogdan, et al., 2024)”.

(22) Figure 4 could benefit from some more explanation in-text. It wasn't clear to me, for example, how to interpret the asterisks depicted in the right part of the figure.

We clarified the meaning of the asterisks in the main text in addition to the existent text in the figure caption.

Page 26. “see Figure 4, off-diagonal cells in blue; asterisks indicate where tRSA was statistically more sensitive then cRSA)”.

(23) Page 38 "the outcome of tRSA's improved characterization can be seen in multiple empirical outcomes:" it seems there is one mention of 'outcomes' too many here.

We have revised this sentence.

Page 41. “tRSA's improved characterization can be seen in multiple empirical outcomes”.

(24) Page 38 "model fits became the strongest" it's not clear what aspect of the reported results in the paragraph before this is referring to - the Appendix?

Yes, the model fits are in the Appendix, we have added this in text citation.

Moreover, model-fits became the strongest when the models also incorporated trial-level variables such as fMRI run and reaction time (Appendix 3, Table 6).

References

Diedrichsen, J., Berlot, E., Mur, M., Schütt, H. H., Shahbazi, M., & Kriegeskorte, N. (2021). Comparing representational geometries using whitened unbiased-distance-matrix similarity. Neurons, Behavior, Data and Theory, 5(3). https://arxiv.org/abs/2007.02789

Diedrichsen, J., & Kriegeskorte, N. (2017). Representational models: A common framework for understanding encoding, pattern-component, and representational-similarity analysis. PLoS Computational Biology, 13(4), e1005508.

Diedrichsen, J., Yokoi, A., & Arbuckle, S. A. (2018). Pattern component modeling: A flexible approach for understanding the representational structure of brain activity patterns. NeuroImage, 180, 119-133.

Naselaris, T., Kay, K. N., Nishimoto, S., & Gallant, J. L. (2011). Encoding and decoding in fMRI. NeuroImage, 56(2), 400-410.

Nili, H., Wingfield, C., Walther, A., Su, L., Marslen-Wilson, W., & Kriegeskorte, N. (2014). A toolbox for representational similarity analysis. PLoS Computational Biology, 10(4), e1003553.

Schütt, H. H., Kipnis, A. D., Diedrichsen, J., & Kriegeskorte, N. (2023). Statistical inference on representational geometries. ELife, 12. https://doi.org/10.7554/eLife.82566

Walther, A., Nili, H., Ejaz, N., Alink, A., Kriegeskorte, N., & Diedrichsen, J. (2016). Reliability of dissimilarity measures for multi-voxel pattern analysis. NeuroImage, 137, 188-200.

King, M. L., Groen, I. I., Steel, A., Kravitz, D. J., & Baker, C. I. (2019). Similarity judgments and cortical visual responses reflect different properties of object and scene categories in naturalistic images. NeuroImage, 197, 368-382.

Kriegeskorte, N., Mur, M., Ruff, D. A., Kiani, R., Bodurka, J., Esteky, H., ... & Bandettini, P. A. (2008). Matching categorical object representations in inferior temporal cortex of man and monkey. Neuron, 60(6), 1126-1141.

Nili, H., Wingfield, C., Walther, A., Su, L., Marslen-Wilson, W., & Kriegeskorte, N. (2014). A toolbox for representational similarity analysis. PLoS computational biology, 10(4), e1003553.

Sucholutsky, I., Muttenthaler, L., Weller, A., Peng, A., Bobu, A., Kim, B., ... & Griffiths, T. L. (2023). Getting aligned on representational alignment. arXiv preprint arXiv:2310.13018.

The fact that nobody react to the murder really attracts people. This actually can show how people are unconcerned when the problem does not harm themselves.

European countries welcoming Ukrainian refugees have previously rejected, demonized, or deported Afghan, Syrian, and African refugees. Now, European media has been normalizing tragedies and also emphasize how Ukrainian worth people's empathy and support only because they look and live similarly with Europeans. Q: Why do people still hold the belief that areas outside of Europe are supposed to be in war?

This reflects how wars in places outside of Europe have been treated as solely background noise in European media.

Humanity should not be constrained by an simple judgement on ethnicity or superficial similarities.

The word civilized used to describe Ukraine implies that war is expected only in uncivilized or impoverished places.

eLife Assessment

This important study uses the delay line axon model in the chick brainstem auditory circuit to examine the interactions between oligodendrocytes and axons in the formation of internodal distances. This is a significant and actively studied topic, and the authors have used this preparation to support the hypothesis that regional heterogeneity in oligodendrocytes underlies the observed variation in internodal length. In a solid series of experiments, the authors have used enhanced tetanus neurotoxin light chains, a genetically encoded silencing tool, to inhibit vesicular release from axons and support the hypothesis that regional heterogeneity among oligodendrocytes may underlie the biased nodal spacing pattern in the sound localization circuit.

[Editors' note: this paper was reviewed by Review Commons.]

Reviewer #2 (Public review):

Summary:

Egawa et al describe the developmental timeline of the assembly of nodes of Ranvier in the chick brainstem auditory circuit. In this unique system, the spacing between nodes varies significantly in different regions of the same axon from early stages, which the authors suggest is critical for accurate sound localization. Egawa et al set out to determine which factors regulate this differential node spacing. They do this by using immunohistological analyses to test the correlation of node spacing with morphological properties of the axons, and properties of oligodendrocytes, glial cells that wrap axons with the myelin sheaths that flank the nodes of Ranvier. They find that axonal structure does not vary significantly, but that oligodendrocyte density and morphology varies in the different regions traversed by these axons, which suggests this is a key determinant of the region-specific differences in node density and myelin sheath length. They also find that differential oligodendrocyte density is partly determined by secreted neuronal signals, as (presumed) blockage of vesicle fusion with tetanus toxin reduced oligodendrocyte density in the region where it is normally higher. Based on these findings, the authors propose that oligodendrocyte morphology, myelin sheath length, and consequently nodal distribution are primarily determined by intrinsic oligodendrocyte properties rather than neuronal factors such as activity.

Significance:

In our view the study tackles a fundamental question likely to be of interest to a specialized audience of cellular neuroscientists. This descriptive study is suggestive that in the studied system, oligodendrocyte density determines the spacing between nodes of Ranvier, but further manipulations of oligodendrocyte density per se are needed to test this convincingly.

Reviewer #3 (Public review):

Summary:

The authors have investigated the myelination pattern along the axons of chick avian cochlear nucleus. It has already been shown that there are regional differences in the internodal length of axons in the nucleus magnocellularis. In the tract region across the midline, internodes are longer than in the nucleus laminaris region. Here the authors suggest that the difference in internodal length is attributed to heterogeneity of oligodendrocytes. In the tract region oligodendrocytes would contribute longer myelin internodes, while oligodendrocytes in the nucleus laminaris region would synthesize shorter myelin internodes. Not only length of myelin internodes differs, but also along the same axon unmyelinated areas between two internodes may vary. This is an interesting contribution since all these differences contribute to differential conduction velocity regulating ipsilateral and contralateral innervation of coincidence detector neurons. However, the demonstration falls rather short of being convincing.

Significance:

The authors suggest that the difference in internodal length is attributed to heterogeneity of oligodendrocytes. In the tract region oligodendrocytes would contribute longer myelin internodes, while oligodendrocytes in the nucleus laminaris region would synthesize shorter myelin internodes. Not only length of myelin internodes differs, but also along the same axon unmyelinated areas between two internodes may vary. This is an interesting contribution since all these differences contribute to differential conduction velocity regulating ipsilateral and contralateral innervation of coincidence detector neurons.

Editors' note: The authors have written an effective rebuttal to the previous round of reviews.

Author response:

The following is the authors’ response to the previous reviews

Reviewer #3:

Comments on revised version:

This revised version is in large improved and the responses to reviewers' comments are generally relevant. However, the response regarding pre-nodes is not satisfactory. I understand that the authors prefer to avoid further experimentations, but I think this is an important point that needs to be clarified. Exploring stages between E12 and E15 are therefore of importance. When carefully examining some of the figures (Fig. 1E or 2D) I think that at E15 they may well be pre-nodes formation prior to myelin deposition, on structure the authors considered to be heminodes. To be convincing they should use double or triple labeling with, in addition to the nodal proteins (ankG and/or Nav pan), a good myelin marker such as antiPLP. The rat monoclonal developed by late Pr Ikenaka would give a sharper staining than the anti MAG they used. (I assume the clone must still be available in Okazaki ).

We appreciate your insightful comment regarding the possible presence of pre-nodal clusters along NM axons and your kind suggestion to use the PLP antibody (clone AA3; Yamamura et al., J Neurochem, 1991). We have obtained this monoclonal antibody from Dr. Kenji Tanaka previously in Okazaki and confirmed that it works well in chicken tissues. However, since this clone recognizes both PLP and DM-20 isoforms, it labels not only myelin-forming oligodendrocytes (MFOLs) but also newly formed oligodendrocytes (NFOLs) (Yokoyama et al., J Neurochem, 2025). Therefore, it is not ideal for determining whether nodal protein clusters are formed before myelin deposition.

Instead, we performed double immunostaining for MAG and AnkG between E12 and E15 to clarify the temporal relationship between myelin maturation and node formation. The results showed that detectable AnkG clusters along NM axons began to appear very sparsely around E13, coinciding with the emergence of MAG signals, and became more prominent with development. This temporal pattern does not match the definition of pre-nodal clusters, which are formed prior to myelination.

Although we cannot completely rule out the possibility of undetectable pre-nodal clusters or those composed of molecules other than AnkG, our results support the view that pre-nodal clusters are unlikely to play a major role in determining the regional difference in nodal spacing along NM axons. These new data have been added as Figure 2—figure supplement 1, and the relevant sections in the Results, Discussion, and Figure legend have been revised accordingly (page 5, line 4; page 10, line 7; page 29, line 1).

I find it's interesting because with the progress of time, even if some people against under authority, it will not change the thinking of major group of people.

I find it's interesting because people may violate their own morality under the influence of authority.

What does a constant degree of compatibility symbolize?

Does this mean that as society and time progresses, humans are gradually losing their sense of self-independence?

Until to nowadays, under the extreme oppression of authority, people still follow their orders and lose their "humanity".

I find it's interesting under the absolute authority, people will continue to follow the rule without considering any factors.

I find it's interesting even children who had no means of defense were also arrested.

This reflects that the US government has not eliminated its prejudice against people of different skin colors until nowadys.

I find it's interesting because the government merely arrests of all "suspects" without making any preparations.

Why doesn't the government take care of some special identities?

I find it's interesting because the government even failed to protect the safety of American citizens during the arrest process.

Why use such a strong military force for just arrest immigrants?

In reality, there are also such people who ignorantly and carelessly disrupts people.Connect to the author mentioned earlier about her aversion to ICE.

If traveling is a part that makes them happy, then why do those officials seem so impatient?

Does the author intend to change this situation?

I find it's interesting because the companies located around are all "illegal and negative".

I find it's interesting because the author employed a irony to express his criticism towards those who applied for ICE.

https://web.archive.org/web/20251210080852/https://www.nature.com/articles/s41467-025-66634-7

Cool! Hotmixing turns out to be the secret behind the durable Roman 'concrete'. A 2023 theory now confirmed by a Pompei construction site find.

Researchers think this new insight may be applied in current building, and impact the climate footprint of modern day concrete.

An unfinished Pompeian construction site reveals ancient Roman building technology in Zotero

eLife Assessment

This is a valuable study that combines biophysical and evolutionary approaches to understand why particular mutations in the SARS-CoV-2 protein N arose during the COVID-19 pandemic. The evidence is solid and supports the conclusions.

Reviewer #1 (Public review):

Summary:

The authors attempted to clarify the impact of N protein mutations on ribonucleoprotein (RNP) assembly and stability using analytical ultracentrifugation (AUC) and mass photometry (MP). These complementary approaches provide a more comprehensive understanding of the underlying processes. Both SV-AUC and MP results consistently showed enhanced RNP assembly and stability due to N protein mutations.<br /> The overall research design appears well planned, and the experiments were carefully executed.

Strengths:

SV-AUC, performed at higher concentrations (3 µM), captured the hydrodynamic properties of bulk assembled complexes, while MP provided crucial information on dissociation rates and complex lifetimes at nanomolar concentrations. Together, the methods offered detailed insights into association states and dissociation kinetics across a broad concentration range. This represents a thorough application of solution physicochemistry.

Weaknesses:

Unlike AUC, MP observes only a part of solution. In MP, bound molecules are accumulated on the glass surface (not dissociated) thus concentration in solution should change as time develops. How does such concentration change impact the result shown here?

Comments on revisions:

The response from the authors is appropriate and reasonable.

Reviewer #2 (Public review):

Summary:

In this manuscript, the authors apply a variety of biophysical and computational techniques to characterize the effects of mutations in the SARS-CoV-2 N protein on the formation of ribonucleoprotein particles (RNPs). They find convergent evolution in multiple repeated independent mutations strengthening binding interfaces, compensating for other mutations that reduce RNP stability but which enhance viral replication.

Strengths:

The authors assay the effects of a variety of mutations found in SARS-CoV-2 variants of concern using a variety of approaches, including biophysical characterization of assembly properties of RNPs, combined with computational prediction of the effects of mutations on molecular structures and interactions. The findings of the paper contribute to our increasing understanding of the principles driving viral self-assembly, and increases the foundation for potential future design of therapeutics such as assembly inhibitors.

Weaknesses:

For the most part, the paper is well-written, the data presented support the claims made, and the arguments made easy to follow. However, I believe that parts of the presentation could be substantially improved. I found portions of the text to be overly long and verbose and likely could be substantially edited; the use of acronyms and initialisms is pervasive, making parts of the exposition laborious to follow; and portions of the figures are too small and difficult to read/understand.

Comments on revisions:

The authors have adequately addressed all of my concerns.

Reviewer #3 (Public review):

Summary:

This manuscript investigates how mutations in the SARS-CoV-2 nucleocapsid protein (N) alter ribonucleoprotein (RNP) assembly, stability, and viral fitness. The authors focus on mutations such as P13L, G214C, G215C combining biophysical assays (SV-AUC, mass photometry, CD spectroscopy, EM), VLP formation, and reverse genetics. They propose that SARS-CoV-2 exploits "fuzzy complex" principles, where distributed weak interfaces in disordered regions allow both stability and plasticity, with measurable consequences for viral replication.

Strengths:

* The paper demonstrates a comprehensive integration of structural biophysics, peptide/protein assays, VLP systems, and reverse genetics.

* Identification of both de novo (P13L) and stabilizing (G214C/G215C) interfaces provides a mechanistic insight into RNP formation.

* Strong application of the "fuzzy complex" framework to viral assembly, showing how weak/disordered interactions support evolvability, is a significant conceptual advance in viral capsid assembly.

* Overall, the study provides a mechanistic context for mutations that have arisen in major SARS-CoV-2 variants (Omicron, Delta, Lambda) and a mechanistic basis for how mutations influence phenotype via altered biomolecular interactions.

Weaknesses:

The weaknesses are shared via detailed comments to follow.

Comments on revisions:

The authors have addressed the criticisms of the original manuscript satisfactorily.

Author response:

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

Reviewer #1 (Public Review):

Summary:

The authors attempted to clarify the impact of N protein mutations on ribonucleoprotein (RNP) assembly and stability using analytical ultracentrifugation (AUC) and mass photometry (MP). These complementary approaches provide a more comprehensive understanding of the underlying processes. Both SV-AUC and MP results consistently showed enhanced RNP assembly and stability due to N protein mutations.

The overall research design appears well planned, and the experiments were carefully executed.

Strengths:

SV-AUC, performed at higher concentrations (3 µM), captured the hydrodynamic properties of bulk assembled complexes, while MP provided crucial information on dissociation rates and complex lifetimes at nanomolar concentrations. Together, the methods offered detailed insights into association states and dissociation kinetics across a broad concentration range. This represents a thorough application of solution physicochemistry.

We thank the Reviewer for this positive assessment. 

Weaknesses:

Unlike AUC, MP observes only a part of the solution. In MP, bound molecules are accumulated on the glass surface (not dissociated), thus the concentration in solution should change as time develops. How does such concentration change impact the result shown here?

We agree with the Reviewer that the concentration in solution above the surface will change with time; however, the impact of surface adsorption turns out to be negligible. To show this we have added a calculation as Supplementary Methods that is based on the number of imaged adsorption events, the fraction of imaged area to total surface area, and the initial sample volume and concentration. Under our experimental conditions the reduction is less than 1%, which is well within the range of experimental concentration errors.

This is in line with the observation that surface adsorption of proteins to glass is critical and needs to be prevented when working at picomolar concentrations (Zhao H, Mayer ML, Schuck P. 2014. Analysis of protein interactions with picomolar binding affinity by fluorescence-detected sedimentation velocity. Anal Chem 86:3181–3187. doi:10.1021/ac500093m), but is ordinarily negligible when working at the mid nanomolar concentration range. The difference in the MP experiments is that where usually the surface adsorption to glass and plastic is invisible, it is being imaged and quantified in MP. The negligible impact of surface adsorption on solution concentration in typical MP experiments is also in line with the results of several studies that have successfully measured dissociation constants of binding equilibria by MP (Young G et al., Science 360 (2018) 432; Wu & Piszczeck, Anal Biochem 592 (2020) 113575; Solterman et al. Angewandte Chemie 59 (2020) 10774) with samples in the 5-50 nM range and similar experimental setup. It should be noted that in the MP experiments no surface functionalization is employed, in contrast to optical biosensors that utilize surface-immobilized ligands and polymeric matrices and thereby enhance the surface binding capacity.

Even though this depletion effect is negligible under ordinary MP conditions, the Reviewer raises a good point and readers may have a similar question with this novel technique. For this reason, we have added in the MP section of the Methods the sentence “In either configuration, the impact of surface binding on the sample concentration is < 1% and negligible, as described in the Supplementary Methods S1.” and added the detailed calculations in the Supplement accordingly. The use of SV as a traditional, orthogonal technique and the observation of consistent results with those of MP should further dispel readers’ methodological concerns in this point.

Reviewer #2 (Public Review):

Summary:

In this manuscript, the authors apply a variety of biophysical and computational techniques to characterize the effects of mutations in the SARS-CoV-2 N protein on the formation of ribonucleoprotein particles (RNPs). They find convergent evolution in multiple repeated independent mutations strengthening binding interfaces, compensating for other mutations that reduce RNP stability but which enhance viral replication.

Strengths:

The authors assay the effects of a variety of mutations found in SARS-CoV-2 variants of concern using a variety of approaches, including biophysical characterization of assembly properties of RNPs, combined with computational prediction of the effects of mutations on molecular structures and interactions. The findings of the paper contribute to our increasing understanding of the principles driving viral self-assembly, and increase the foundation for potential future design of therapeutics such as assembly inhibitors.

Thank you for highlighting the strengths of our paper and the potential impact on future design of therapeutics.

Weaknesses:

For the most part, the paper is well-written, the data presented support the claims made, and the arguments are easy to follow. However, I believe that parts of the presentation could be substantially improved. I found portions of the text to be overly long and verbose and likely could be substantially edited; the use of acronyms and initialisms is pervasive, making parts of the exposition laborious to follow; and portions of the figures are too small and difficult to read/understand.

We are glad the Reviewer concurs the data support our conclusions, and finds the arguments easy to follow.  We appreciate the comment that the work was not optimally presented. To address this point, we have identified multiple opportunities to streamline the text without jeopardizing the clarity. We have also rewritten the end of the Introduction.

As recommended, we have reduced and harmonized the use of acronyms and abbreviations throughout the text to improve readability. Specifically, we have now spelled out nucleic acid (NA), intrinsically disordered regions (IDR), full-length (FL), AlphaFold (AF3), and variants of concern (VOC).

Finally, we have improved the presentation of most figures, adding labels and new panels, and increased the label font sizes to facilitate more detailed inspections of the data.

Reviewer #3 (Public Review):

This manuscript investigates how mutations in the SARS-CoV-2 nucleocapsid protein (N) alter ribonucleoprotein (RNP) assembly, stability, and viral fitness. The authors focus on mutations such as P13L, G214C, and G215C, combining biophysical assays (SV-AUC, mass photometry, CD spectroscopy, EM), VLP formation, and reverse genetics. They propose that SARS-CoV-2 exploits "fuzzy complex" principles, where distributed weak interfaces in disordered regions allow both stability and plasticity, with measurable consequences for viral replication.

Strengths:

(1) The paper demonstrates a comprehensive integration of structural biophysics, peptide/protein assays, VLP systems, and reverse genetics.

(2) Identification of both de novo (P13L) and stabilizing (G214C/G215C) interfaces provides a mechanistic insight into RNP formation.

(3) Strong application of the "fuzzy complex" framework to viral assembly, showing how weak/disordered interactions support evolvability, is a significant conceptual advance in viral capsid assembly.

(4) Overall, the study provides a mechanistic context for mutations that have arisen in major SARS-CoV-2 variants (Omicron, Delta, Lambda) and a mechanistic basis for how mutations influence phenotype via altered biomolecular interactions.

We are grateful for these comments highlighting this work as a significant conceptual advance.

Weaknesses:

(1) The arrangement of N dimers around LRS helices is presented in Figure 1C, but the text concedes that "the arrangement sketched in Figure 1C is not unique" (lines 144-146) and that AF3 modeling attempts yielded "only inconsistent results" (line 149).

The authors should therefore present the models more cautiously as hypotheses instead. Additional alternative arrangements should be included in the Supplementary Information, so the readers do not over-interpret a single schematic model.

We agree that in the absence of high-resolution structures the RNP models are hypothetical, and have now emphasized this in the Results, following the Reviewer’s recommendation. To present alternative arrangements that satisfy the biophysical constraints upfront, we have promoted the previous Supplementary Figure 11 showing different models to the first Supplementary Figure, and expanded it with examples of different oligomers. In this way it is referenced early on in the Results and in the legend to Figure 1C. We agree this strengthens the manuscript, as one of the take-home messages is the inherent polydispersity of the RNPs.

The fact that AF3 can only provide inconsistent results will not come as a surprise, given the substantial disordered regions of the complex, and is a drawback of AF3 rather than our structural model. We slightly emphasized this point so as to clarify that the presentation of the AF3-based RNP structure serves solely as supporting evidence that our hypothetical model is sterically reasonable.

The new Results paragraph reads:

“As suggested in the cartoon of Figure 1C, this supports the hypothesis of a three-dimensional arrangement with a central LRS oligomer with symmetry properties and dimensions similar to low resolution EM images of model RNPs (Carlson et al., 2022, 2020) and cryo-ET of RNPs in virions (Klein et al., 2020; Yao et al., 2020).  It should be noted, however, that the arrangement sketched in Figure 1C is not unique and other subunit orientations could be envisioned that satisfy all constraints from experimentally observed binding interfaces, including different oligomers and anti-parallel subunits as illustrated in Supplementary Figure S1. Extending previous ColabFold structural predictions that show multiple N-protein dimers self-assembled via the LRS coiled-coils (Zhao et al., 2023), we attempted the AlphaFold modeling of RNPs combining multiple N dimers with SL7 RNA ligands, mimicking our biophysical assembly model. Current AlphaFold restrictions limit the prediction to pentamers of N-protein dimers with 10 copies of SL7 RNA. While only inconsistent results were obtained – which is not surprising given the large intrinsically disordered regions exceed the predictive power of AlphaFold – some models did produce an overall RNP organization similar to Figure 1C, suggesting such an arrangement is at least sterically reasonable with regard to possible N-protein subunit orientations in an RNP (Supplementary Figure S2)”

(2) Negative-stained EM fibrils (Figure 2A) and CD spectra (Figure 2B) are presented to argue that P13L promotes β-sheet self-association. However, the claim could benefit from more orthogonal validation of β-sheet self-association. Additional confirmation via FTIR spectra or ThT fluorescence could be used to further distinguish structured β-sheets from amorphous aggregation.

We completely agree that the application of multiple orthogonal biophysical methods can strengthen the conclusions. In addition to EM fibrils and CD spectra (a classical gold standard technique for protein secondary structure in solution), we already have support from ColabFold modeling, as well as NMR results from the Zweckstetter lab showing the potential for for β-sheet-like conformations.

Furthermore, we believe the evidence for the absence of ‘amorphous aggregates’ is very strong, as this would be inconsistent with the long-range order required to create the visibly fibrillar morphology in EM, and amorphous aggregates would be inconsistent with the increased solution viscosity. In this context, it is also highly relevant that the β-sheet-like secondary structure recorded by CD is concentration-dependent and reversible upon dilution. The long-range spatial order of fibrils is consistent with the formation of secondary structure in solution.

In addition, it must be kept in mind that what we see is specific to N-arm peptides carrying the P13L mutation (in EM, CD, and structural prediction) and does not occur in the other two N-arm peptides (ancestral N-arm and N-arm with deletion of 31-33), linker peptides, or C-arm peptides.

Most importantly, as elaborated in more detail below, we do not claim that fibril formation is physiologically relevant. At the heart of this – in the context of the evolution of fuzzy complexes – is that the P13L mutation creates additional weak protein-protein interactions. Indeed, the assembly of fibrils geometrically requires at least two interfaces for each subunit. These weak interactions are at play physiologically in the context of the disordered RNP particles, and in macromolecular condensates, but not in the formation of fibrils. Therefore, while we appreciate the suggestion for FTIR spectra ThT staining, we are afraid further emphasis on the fibril structure might confuse the reader, and therefore we would rather clarify upfront that these fibrillar assemblies are not thought to form in vivo from full-length protein, but merely demonstrate the presence of N-arm self-association interfaces in the model of truncated peptides.

Accordingly, we have amended the Results paragraph reporting the fibrils:

“Thus, the N-arm mutation P13L is responsible for the formation of fibrils in N-arm peptides after prolonged storage. Some of these N-arm fibrils exhibit a twisted morphology with width of »5 nm (Figure 2A), in some instances exhibiting patterns of strand breaks. Such fibrils are frequently encountered in proteins that can stack β-sheets, such as in amyloids (Paravastu et al., 2008). While we have not observed fibril formation in the context of full-length N, and have no evidence such fibrils are physiologically relevant, their occurrence in solutions of truncated N-arm peptide nonetheless demonstrates the introduction of ordered N-arm self-association interfaces in conformations of P13L mutants.”

And more completely summarized experimental evidence prior to describing the ColabFold prediction results (which previously did not include mention of the NMR):

“Finally, confirming the interpretation of the EM images and the CD data, as well as the b-structure propensity reported from NMR data (Zachrdla et al., 2022), the structural prediction of N[10-20]:P13L in ColabFold displayed oligomers with stacking b-sheets …”

(3) In the main text, the authors alternate between emphasizing non-covalent effects ("a major effect of the cysteines already arises in reduced conditions without any covalent bonds," line 576) and highlighting "oxidized tetrameric N-proteins of N:G214C and N:G215C can be incorporated into RNPs". Therefore, the biological relevance of disulfide redox chemistry in viral assembly in vivo remains unclear. Discussing cellular redox plausibility and whether the authors' oxidizing conditions are meant as a mechanistic stress test rather than physiological mimicry could improve the interpretation of these results.

The paper could benefit if the authors provide a summary figure or table contrasting reduced vs. oxidized conditions for G214C/G215C mutants (self-association, oligomerization state, RNP stability). Explicitly discuss whether disulfides are likely to form in infected cells.

We thank the Reviewer for raising this most interesting point.  The reason why the biological relevance of N dilsulfides remains unclear is simply that this is still unknown, unfortunately. Recently, Kubinski et al. have strongly argued for the formation of disulfides in infected cells, but in our view the evidence remains weak since the majority of disulfide bonds in that work presented as post-lysis artifacts, and it appears the non-covalent effects alone could explain the physiological observations. We aimed for a balanced presentation and wrote in the relevant Results section:

“Covalent disulfide bonds in the LRS in non-reducing conditions were found to further promote LRS oligomerization. However, there is no conclusive data yet whether covalent bonds in the LRS occur in vivo, or any G215C effect is entirely non-covalent due to the significant strengthening of LRS helix oligomerization (see Discussion).”

Despite the uncertainty regarding physiological disulfide bond formation, we believe it is useful to ask whether covalently crosslinked N dimers would aid or constrain RNP assembly in our biophysical model. We have now better explained this motivation in the Results section describing the RNP experiments:

“Even though it is still unclear whether disulfide bonds of N cysteine mutants form in vivo, we were curious about the impact of disulfide-linked oligomers of the cysteine mutants on their RNP structure and stability in our biophysical assembly model.”

The referenced paragraph from the Discussion reads:

“Regarding the cysteine mutations that have been repeatedly introduced in the LRS prior to the rise of the Omicron VOCs, it is an open question whether they lead to covalent bonds in vivo or in the VLP assay. While examples of disulfide-linked viral nucleocapsid proteins have been reported (Kubinski et al., 2024; Prokudina et al., 2004; Wootton and Yoo, 2003), a methodological difficulty in their detection is artifactual disulfide bond formation post-lysis of infected cells (Kubinski et al., 2024; Wootton and Yoo, 2003).  However, our results clearly show that a major effect of the cysteines already arises in reduced conditions without any covalent bonds, through extension of the LRS helices, and concomitant redirection of the disordered N-terminal sequence. While oxidized tetrameric N-proteins of N:G214C and N:G215C can be incorporated into RNPs, the covalent bonds provided only marginally improved RNP stability.  Interestingly, the introduction of cysteines imposes preferences of RNP oligomeric states dependent on oxidation state, consistent with our MD simulations highlighting the impact of cysteine orientation of 214C versus 215C relative to the hydrophobic surface of the LRS helices. Overall, considering potentially detrimental structural constraints from covalent bonds on LRS clusters seeding RNPs, energetic penalties on RNP disassembly, as well as the required monomeric state of the LRS helix for interaction with the NSP3 Ubl domain (Bessa et al., 2022), at present it is unclear to what extent the formation of disulfide linkages between LRS helices would be beneficial or detrimental in the viral life cycle.”

We feel that this text addresses the Reviewer’s comment, and that expanding the existing discussion further would conflict with other recommendations to shorten and focus the text.

Finally, we have addressed the valuable suggestion of a new table summarizing the oligomeric state and self-association of the different cysteine mutants by inserting a new column in the existing Table 1 reporting all species’ oligomeric state at low micromolar concentrations. In this way they can be compared at a glance with the other mutants as well. A more detailed comparison of the concentration-dependent size-distribution is provided in Figure 4.

(4) VLP assays (Figure 7) show little enhancement for P13L or G215C alone, whereas Figure 8 shows that P13L provides clear fitness advantages. This discrepancy is acknowledged but not reconciled with any mechanistic or systematic rationale. The authors should consider emphasizing the limitations of VLP assays and the sources of the discrepancy with respect to Figure 8.

We thank the Reviewer for this comment, which highlights a very important point. 

For clarification and to improve the cohesion of the manuscript we have inserted a reference to the Discussion after the presentation of the VLP results, which provides a natural transition to the following description of the reverse genetics experiments:

“As expanded on in the Discussion, the failure to observe enhancement by P13L alone may be related to limitations of the VLP assay in sensitivity, including the restriction to a single round of infection, and protein expression levels.”

This references a paragraph in the Discussion about the limitations of the VLP assay in general and the reasons we believe the enhancement by P13L alone was not picked up:

“…While this assay has been widely used for rapid assessment of spike protein and N variants (Syed et al., 2021), it has limitations due to the addition of non-genomic RNA and the lack of double membrane vesicles from which gRNA emerges through the NSP3/NSP4 pore complex potentially poised for packaging (Bessa et al., 2022; Ke et al., 2024; Ni et al., 2023). It should also be recognized that the results do not directly reflect the relative efficiency of RNP assembly only, since protein expression levels, their localization, and their posttranslational modifications are not controlled for. Susceptibility for such factors might be exacerbated with mutations that modulate weak protein interactions. For example, as shown previously (Syed et al., 2024; Zhao et al., 2024), a GSK3 inhibitor inhibiting N-protein phosphorylation significantly enhances VLP formation and eliminates the advantage provided for by the N:G215C mutation relative to the ancestral N – presumably due to an increase in assembly-competent, non-phosphorylated N-protein erasing an affinity advantage. A similar process may be underlying the absent or marginal improvement in VLP readout from the cysteine LRS mutants and P13L at the achieved transfection level in the present work, and the enhanced signal from R203K/G204R and R203M (the latter being consistent with previous reports (Li et al., 2025; Syed et al., 2021)) modulating protein phosphorylation. Nonetheless, mirroring the results of the biophysical in vitro experiments, the addition of RNP-stabilizing P13L and G214C mutations on top of R203K/G204R led to a significantly larger VLP signal.

The VLP assay may be limited in sensitivity to mutation effects due to its restriction to a single round of infection. To avoid this and other potential limitations of the VLP assay for the study of viral packaging, for the key mutation N:P13L we carried out reverse genetics experiments. These showed the sole N:P13L mutation significantly increases viral fitness (Figure 8).”

(5) Figures 5 and 6 are dense, and the several overlays make it hard to read. The authors should consider picking the most extreme results to make a point in the main Figure 5 and move the other overlays to the Supplementary. Additionally, annotating MP peaks directly with "2×, 4×, 6× subunits" can help non-experts.

We completely agree with the Reviewer – these figures were very dense.  To mitigate this problem without having the reader to switch back-and-forth to the supplement, we subdivided the panels of Figure 5 and showed only a subset of curves in each.  In this way the data are easier to read while still readily compared. It is a large figure, but it contains the key data for the present work and is therefore worthwhile to have in one place. For the MP histogram data we also have inserted the suggested peak labels. Similarly, we have split Figure 6A into two panels for clarity.

(6) The paper has several names and shorthand notations for the mutants, making it hard to keep up. The authors could include a table that contains mutation keys, with each shorthand (Ancestral, Nο/No, Nλ, etc.) mapped onto exact N mutations (P13L, Δ31-33, R203K/G204R, G214C/G215C, etc.). They could then use the same glyphs (Latin vs Greek) consistently in text and figure labels.

Yes, we agree this is a problem and we apologize for the confusion. However, it is not possible to refer exclusively to either Latin or Greek terminology, which we feel would be even more detrimental to readability (the former being exhaustively lengthy and the latter being imprecise). But we have used a rational system: If the complete set of mutations of a variant are present, then its Greek letter will be used as an abbreviation, and otherwise we use Latin amino acid/position indicators for individual mutations or combinations thereof. Unfortunately, previously we inadvertently failed to explicitly mention this, and we are most grateful for the Reviewer to point this out.

We have now rectified this by including upfront the sentence:

“We will adopt a nomenclature where the complete set of defining mutations of a variant will be referred to by its Greek letter, i.e., N:P13L/R203K/G204R/G214C is N_­­λ, and analogously the set of Omicron mutations N:P13L/Δ31-33/R203K/G204R are referred to as N_ο; see Table 1”

This will define the two shorthands N_λ and N_ο used. Furthermore, as suggested and pointed to in the text, Table 1 does provide the keys to mutation and variants, including the information in which variant any of the other mutations studied here occur.

(7) The EM fibrils (Figure 2A) and CD spectra (Figure 2B) were collected at mM peptide concentrations. These are far above physiological levels and may encourage non-specific aggregation. Similarly, the authors mention" ultra-weak binding energies that require mM concentrations to significantly populate oligomers". On the other hand, the experiments with full-length protein were performed at concentrations closer to biologically relevant concentrations in the micromolar range. While I appreciate the need to work at high concentrations to detect weak interactions, this raises questions about physiological relevance.

This is indeed an important point to clarify. We agree that much lower nucleocapsid protein concentrations are present in the cytosol on average, and these were used in our RNP assembly experiments. However, there are at least two important physiologically relevant cases where high local N concentrations do occur:

(1) Once assembled in RNPs, the disordered N-terminal extensions are locally at a very high concentration within the volume they can explore while tethered to the NTD. A back-of-the-envelope calculation assuming 12 N-protein subunits confining 12 N-terminal extensions to the volume of a single RNP (≈14x14x14 nm³ by cryoEM; Klein et al 2020) leads to an effective concentration of 7.4 mM. Obviously the N-arm peptides are not completely free and there will be constraints that would hinder or promote encounter complex probability, but interfaces with mM Kd are clearly strong enough to populate Narm-Narm contacts extending from N-protein in the RNP.

Additionally, any interaction where N-proteins are brought in close proximity could allow weak N-arm interactions to provide additional stability. Besides the RNP, we demonstrate this in our Results for nucleic-acid liganded N tetramers (Figure 4B), but this might similarly occur in complexes with NSP3 or host proteins. Generally, it is quite common that small additional binding energies play important roles in the modulation of multivalent protein complexes.

(2) Within the macromolecular condensate the local concentration will be substantially higher than on average within the infected cell.  While we do not know its precise concentration, it is well-established that the sum of many ultra-weak interactions is driving the formation of this dense liquid phase. In our previous eLife paper (Nguyen et al., 2024) we have shown LLPS is suppressed with the R203K/G204R mutation, but it is ‘rescued’ with the additional P13L/del31-33 mutation of the Omicron variant showing strong LLPS. Similarly, LLPS is suppressed by the LRS mutant L222P, but rescued in conjunction with P13L. This is another biologically relevant scenario where weak interactions are critical.

We have emphasized these points in the revised manuscript as described below.

Specifically:

(a) Could some of the fibril/β-sheet features attributed to P13L (Figure 2A-C) reflect non-specific aggregation at high concentrations rather than bona fide self-association motifs that could play out in biologically relevant scenarios?

We understand this concern from the experience with proteins that often have limited solubility and tendencies to aggregate, sometimes accompanied by unfolding and driven by hydrophobic interactions, or clustering on the path to LLPS. However, we are struggling to reconcile the picture of non-specific aggregation with the context of our P13L N-arm peptides. The term ‘non-specific aggregation’ implies the idea of amorphous aggregates, which we would contend is inconsistent with the observed geometry of fibrils, which exhibit long-range order. In addition, non-specific aggregation does not lead to increased solution viscosity, which we describe, but fibril formation does. Another connotation of ‘aggregates’ is irreversibility.  However, we find the beta-sheet-like conformation seen at 1 mM becomes significantly more disordered when the same sample is diluted to 0.4 mM peptide. This is consistent with a reversible self-association driven by a conformational change toward ordered secondary structure.

To highlight the reversibility, we have clarified the description: “Interestingly, diluting the 1 mM sample (solid) to a concentration of 0.4 mM (dashed) reveals a large shift in the far-UV spectra … both indicative of a significant increase of disorder upon dilution. This is consistent with the stabilization of b-sheets in a reversible, strongly cooperative self-association process with an effective K_D in the high mM to low mM range.”

We have also inserted a concentration conversion to mg/ml units, which shows even 1 mM of peptides is only ~5 mg/ml, i.e. not excessively high. “While the ancestral N-arm at »1 mM (» 4.6 mg/ml) concentrations exhibits CD spectra with a minimum at »200 nm typical of disordered conformations (black)”

With regard to the question of specificity, we have studied similar N-arm peptides without P13L mutations and with the 31-33 deletion under equivalent conditions. But we observe the reversible self-association, conformational change, and fibril formation only for those containing the P13L mutation, consistent with ColabFold predictions. Neither did we observe fibrils with disordered C-arm peptides.

How these weak self-association motifs in the N-arm can be physiologically relevant in the context of full-length protein modulating the stability of multi-molecular complexes and enhancing LLPS was outlined above, and further clarified in the manuscript as detailed below.

(b) How do the authors justify extrapolating from the mM-range peptide behaviors to the crowded but far lower effective concentrations in cells?

As pointed out above, the key to this question is the local preconcentration as the N-arm peptides are tethered to the rest of protein in the context of flexible multi-molecular assemblies. Another mechanism to consider is the formation of condensates. The response to the next comment will expand on this.

The authors should consider adding a dedicated section (either in Methods or Discussion) justifying the use of high concentrations, with estimation of local concentrations in RNPs and how they compare to the in vitro ranges used here. For concentration-dependent phenomena discussed here, it is vital to ensure that the findings are not artefacts of non-physiological peptide aggregation..

The use of high concentration in biophysical experiments is quite common, for example, in NMR or crystallography, insofar as they elucidate molecular properties. We believe this is obvious; the Reviewer will certainly agree with us, and this does not require further elaboration. The property observed in this case is the existence of specific, weak protein self-association interfaces in the N-arm.

Our response to the Reviewer’s point 7(a) addresses the distinction between artefactual aggregation and self-association of N-arm peptides. The relevance of these weak protein self-association interfaces in the context of the full-length protein is the second underlying question.

As we have previously stated in a dedicated Results paragraph:

“In contrast to the modulation of the coiled-coil LRS interfaces, the de novo creation of the N-arm self-association interface through beta-sheet interactions enabled by P13L cannot be readily observed in full-length N-protein at low M concentrations. Similar to the ancestral LRS interface, it provides only ultra-weak binding energies that require mM concentrations to significantly populate oligomers. This is fully consistent with the previous observation by SV-AUC that neither N:P13L,31-33 nor N_o with the full set of Omicron mutations show any significant higher-order self-association at low M concentrations, whereas at high local concentrations – as observed in phase-separated droplets – they can modulate and cooperatively enhance self-association processes (Nguyen et al., 2024). (If fact, P13L can substitute for the LRS promoting LLPS, as observed in the rescue of LLPS by N:P13L,31-33/L222P mutants whereas N:L222P LRS-abrogating mutants are deficient in LLPS.) Another process that increases the local concentration of N-arm chains is the tetramerization of full-length N-protein. As described earlier, occupancy of the NA-binding site in the NTD allosterically promotes self-assembly of the LRS into higher oligomers (Zhao et al., 2021). We hypothesized that these oligomers may be cooperatively stabilized by additional N-arm interactions in P13L mutants.”

To state completely unambiguously why weak interfaces are important, we have followed the Reviewer’s suggestion and added an additional clarification already earlier, at the end of the P13L Results section:

“While this self-association interface in the P13L N-arm is weak and its direct observation in biophysical experiments requires mM concentrations, which far exceed average intracellular concentration of N, such  weak interactions can become highly relevant physiologically when high local concentrations are prevailing, for example, when the disordered extension is preconcentrated while tethered within macromolecular assemblies as in the RNP, or in macromolecular condensates.”

Furthermore, we have added early in the Discussion:

“Even though the solution affinity of the N-arm P13L interface is ultra-weak, the average local concentration of N-arm chains across the RNP volume (in a back-of-the-envelope calculation assuming a ≈14 nm cube (Klein et al., 2020) with a dodecameric N cluster) is ≈7.4 mM, such that disordered N-arm peptides could well create populations of N-arm clusters stabilizing RNPs through this interface.  However, besides the RNP-stabilizing mutants we have also observed unexpected RNP destabilization by the ubiquitous R203K/G204R double mutation, which may be caused by the introduction of additional charges close to the self-association interface in the LRS. In our experiments, this destabilization is more than compensated for by the P13L mutation. (Another scenario where ultra-weak interactions can have a critical impact is in molecular condensates. We previously reported the suppression of LLPS by the R203K/G204R mutation, which is rescued by the additional P13L/Δ31-33 mutation (Nguyen et al., 2024). This is consistent with compensatory weak stabilizing and destabilizing impacts of weak interactions on the RNP observed here.)”

Reviewer #1 (Recommendations for the Authors):

In Figure 1B, it is unclear what the orange lines connecting polypeptides represent, as well as the zig-zag orange lines in the N-arm.

We thank the Reviewer for this comment. We intended this to represent regions of self-association but recognize the patterned background is confusing. We have changed this now to solid-colored backgrounds, and indicated this in the figure legend:

“Regions of self-association are indicated by shaded backgrounds.”

Regarding presentation, in Figure 5 (MP), the relationship between mass and oligomer size should be shown more clearly.

We agree. To this end we have labeled the peaks in the MP histograms in Figure 5 with the oligomeric state of the 2N/2SL7 subunits.

Reviewer #2 (Recommendations for the Authors):

I find the science of the paper to be convincing and compellingly supported.

Thank you for this positive statement.

My primary complaints are with presentation or minor technical questions that, honestly, primarily arise due to my own ignorance and unfamiliarity with some of the techniques employed.

My primary issue is with the figures. I find, generally, the text in axes labels, ticks, and legends to be too small to comfortably read. This is particularly true in the CD spectra and

other data presented in Figures 1D, 2B, 4, 5, 6, and 8.

We agree and have increased the font size of all text and labels of the plots in Figure 1, 2, 4, 5, 6, and 8.

I also found the use of initialisms to be a bit overbearing and inconsistent. For example, the authors repeatedly switch between spelling out "nucleic acid" and the initialism "NA" (which is also never explicitly spelled out in the text). With the already substantial length of the text, my own personal opinion would be to suggest spelling out all initialisms in the interest of making the reading easier.

This is a valid criticism. To improve the readability, we have followed this advice and systematically spelled out “nucleic acid” instead of using “NA”.  Similarly, we have now written out full-length instead of the abbreviation FL, and omitted the abbreviation IDR for intrinsically disordered regions, as well as VOC for variant of concern, and AF3 for AlphaFold.

Regarding the reference to mutants, we have now explained upfront the system of Latin and Greek nomenclature we consistently applied.

“We will adopt a nomenclature where the complete set of defining mutations of a variant will be referred to by its Greek letter, i.e., N:P13L/R203K/G204R/G214C is N­­_l, and analogously the set of Omicron mutations N:P13L/Δ31-33/R203K/G204R are referred to as N_ο; see Table 1”

I found the text to be verbose, bordering on overly so; the Introduction is more than two pages long. The section "Enhanced oligomerization of the leucine-rich sequence through cysteine mutations" has two long paragraphs of introduction before the present results are discussed, et cetera. An (admittedly, very rough) estimation of the length of the paper places it at ~9,000 -10,000 words long, and I think that the presentation might benefit from significant editing and

shortening.

We agree the manuscript is longer than would be desirable, and we generally prefer not to insert mini-introductions into Results sections. On the other hand, in order to make a solid contribution to understanding the big picture of fuzzy complexes in molecular evolution of RNA virus proteins it is indispensable to go into the details of RNP assembly and several of the interfaces. Therefore, we feel the length is in the range that it needs to be without losing clarity. In addition, other Reviewer suggestions to extend the discussion, for example, of limitations of VLP assays and the in vivo state of cysteines, conflict with significant shortening.

In the particular case of the cysteine mutations, cited by the Reviewer, we believe it is important to add detailed background on G215C, because the Results proceed in a comparison of the self-association mode between G215C and G214C. This is of significant interest in the present context not only for the independent introduction of interface-enhancing mutations highlighting the evolution of fuzzy complexes, but also because it illustrates the pleomorphic ability of RNPs.

Nonetheless, we have slightly shortened this text and merged the background into a single paragraph. More generally, we have critically reread the text to remove tangential sentences where possible and to make it more concise.

I have a few more specific comments.

In Figure 1A, I suggest explicitly labeling the location of the LRS, as it comes up repeatedly.

Yes, we thank the Reviewer for this suggestion and have introduced this label in Figure 1A.

In Figure 1B, the legend indicates that the red lines indicate "new inter-dimer interactions." However, these red lines are overlayed on a vertical stripe of red squiggles; it is unclear to me and not explicitly described in the legend what these squiggles are meant to illustrate.

We agree this background was confusing. As mentioned in our Response to Reviewer #1 we have replaced the structured background with a solid background and explained in the figure legend that these areas depict regions of self-association.

On lines 44-45, the authors state, "The IDRs amount to 45%, ..." 45% of what?

Thank you, this was unclear.  We have now clarified “The IDRs amount to ≈45% of total residues”

In lines 244 - 246, the authors compare the sizes of complexes in reducing versus non- reducing conditions as measured by dynamic light scattering, stating, "However, dynamic light scattering (DLS) revealed the presence of N210-246:G214C complexes with hydrodynamic radii 244 ranging from 6 to 40 nm (in comparison to 1-2 nm for N210- 246:G215C(Zhao et al., 2022)) in reducing conditions, and slightly larger in non-reducing conditions (Supplementary Figure S4)." Using this single statistic seems to me to be a less-than-ideal way of characterizing what seems to me to be happening here. In Supplementary Figure 4, it appears to me that what is happening is that in non-reduced conditions, the sample is monodisperse, whereas in reducing conditions, the distribution becomes polydisperse/bimodal, with two clearly separate populations. I feel that this could use a more

thorough description rather than just stating the overall range of particle sizes.

Yes, the Reviewer is correct – it is indeed a good idea to be more precise here. To this end we have carried out cumulant analyses on the autocorrelation functions, as a time-honored method to quantify the polydispersity.  Both samples are polydisperse, but more so in reducing conditions. We have now added “For N210-246:G214C a cumulant analysis results in radii of 8.8 nm and 10.6 nm and polydispersity indices of 0.40 and 0.35 for reducing and non-reducing conditions, respectively”

Finally, I have one remaining comment that is a result of my own inexperience with circular dichroism and interpreting the spectra. For me personally, I would appreciate a more thoroughdescription/illustration of the statistics involved in the CD spectra, but perhaps this is not necessary for people who are more familiar with interpreting these kinds of data. For example, in Figure 1D, it is not clear to me what the error bars/confidence intervals for the CD data look like. I see many squiggles, some of which the authors claim are significant (e.g., the differences between ~215 - 230 nm), and others are not worthy of comment. Let's say, for example, that I fit a smoothed spline through these data and then measure the magnitude of the fluctuations from that spline to define/quantify confidence intervals. What does that distribution look like? Or maybe the confidence intervals are so small that all squiggles are significant?

Thank you, this is a good question. As mentioned in the methods section, the CD spectra shown are averages of triplicate scans. Therefore, it is straightforward to extract the standard deviation at each wavelength from the three measurements (although a spline would probably work just as well). The values are what one would expect for the squiggles to be random noise. In the region 215 – 220 nm characteristic for helical secondary structure the standard deviations are small relative to the separation between curves, which indicates that the differences are highly significant. Naturally, the curves do overlap in other spectral regions, which would make a plot including the wavelength-dependent error bars or confidence bands too crowded. Therefore, we have kept the plot of the averaged triplicate scans, but have now provided the average standard deviations for all species in the figure legend and mentioned their significant separation:

“Triplicate scans yield average standard deviations of 0.13 (N), 0.17 (N+SL7), 0.16 (N_l), and 0.21 (N_l +SL7) 10³ deg cm²/dmol, respectively, with non-overlapping confidence bands for the different species, for example, between 215-220 nm.”

Reviewer #3 (Recommendations for the Authors):

(1) The Discussion reiterates much of the background (mutational tolerance, fuzziness, SLiMs) already covered in the Introduction, diluting focus on the key new findings. The authors should consider shortening and refocusing the discussion on the main contributions in light of existing knowledge of viral assembly.

In the Introduction we have provided background on intrinsically disordered proteins in general and their mutational tolerance, as well as the concept of fuzzy complexes. The first several paragraphs of the Discussion have a different focus, which is protein binding interfaces between viral proteins (obviously key in fuzzy complexes), specifically their modulation and the remarkable de novo introduction of binding interfaces. We believe this deserves emphasis, since this highlights a novel aspect of fuzziness, for the mutant spectrum of RNA viruses to encode a range and of assembly stabilities and architectures. 

To reduce redundancy between the end of the Introduction and the beginning of the Discussion, we have shortened the last paragraph of the Introduction and removed its preview of the conclusions, as described in the response to the next comment of the Reviewer (see below).

Unfortunately, the length of the Discussion is dictated in part also by the need to discuss methodological aspects, among them the limitations of VLP assays, and the redox state of the cysteine in the LRS mutants, which were important points recommended by other suggestions of the Reviewers. Similarly, we believe the discussion of other potential functions of Omicron N-arm mutations is warranted, as well as the background of the R203K/G204R double mutation that has attracted significant attention in the field due to its effects on phosphorylation and expression of truncated N species that also form RNPs. Our goal was to integrate the results by us and other laboratories regarding specific mutation effects into a comprehensive picture of molecular evolution of N, which we believe the framework of fuzzy complexes can provide.

(2) The Abstract and early Introduction set a broad stage (IDPs, fuzziness), but don't explicitly state the concrete hypotheses that the experiments test. Please add 2-3 sentences in the Introduction that enumerate testable hypotheses, e.g.:

(a) P13L creates a new N-arm interface that increases RNP stability.

(b) G214C/G215C strengthens LRS oligomerization to stabilize higher-order N assemblies.

We agree the introduction can be improved.  However, it seems to us that it cannot be neatly framed in the hypothesis – answer dichotomy, without losing a lot of nuances and without requiring an even longer and more detailed introduction.

One of the main questions is to test whether the framework of fuzzy complexes can be applied to understand molecular evolution of N, and we feel the introduction is already flowing well towards this:

“ … In fuzzy complexes the total binding energy is distributed into multiple distinct ultra-weak interaction sites (Olsen et al., 2017). Similar to individual RNA virus proteins with loose or absent structure, maintaining disorder and a spatial distribution of low-energy interactions in the protein complexes may increase the tolerance for mutations and improve evolvability of protein complexes.\

The unprecedented worldwide sequencing effort of SARS-CoV-2 genomes during its rapid evolution in humans provides a unique opportunity to examine these concepts. ...”

To bring this to a more concrete set of questions in the end, we have shortened and rewritten the last paragraph in the Introduction:

“To examine how architecture and energetics of RNP assemblies can be impacted by N-protein mutations we study a panel of N-proteins derived from ancestral Wuhan-Hu-1 and different VOCs, including Alpha, Delta, Lambda, and Omicron (see Table 1), in biophysical experiments, VLP assays, and mutant virus. Specifically, we ask how the RNP size distribution and life-time is modulated by: (1) the novel binding interface created by the P13L mutation of Omicron; (2) enhancements of other weak self-association interfaces through G215C of Delta and G214C of Lambda; (3) the ubiquitous R203K/G204R double mutation of Alpha, Lambda, and Omicron.  We also test whether the P13L mutation improves viral fitness, similar to G215C and R203K/G204R. The results are discussed in the framework of fuzzy complexes and molecular evolution of N in the course of viral adaptation to the human host. Understanding the salient features of the binding interfaces in viral assembly and their evolution expands our foundation for the design of therapeutics such as assembly inhibitors.”

eLife Assessment

Glioblastoma is among the most aggressive cancers without a cure, and its cells are characterized by high mitochondrial membrane potential. This manuscript provides convincing evidence that glioblastoma tumorigenesis is closely linked to mitochondrial stress. The study makes a valuable contribution to the field by advancing our understanding of the metabolic mechanisms driving glioblastoma and highlighting potential therapeutic targets.

Reviewer #1 (Public review):

Summary:

Cai et al have investigated the role of msiCAT-tailed mitochondrial proteins that frequently exist in glioblastoma stem cells. Overexpression of msiCAT-tailed mitochondrial ATP synthase F1 subunit alpha (ATP5) protein increases the mitochondrial membrane potential and blocks mitochondrial permeability transition pore formation/opening. These changes in mitochondrial properties provide resistance to staurosporine (STS)-induced apoptosis in GBM cells. Therefore, msiCAT-tailing can promote cell survival and migration, while genetic and pharmacological inhibition of msiCAT-tailing can prevent the overgrowth of GBM cells.

Strengths:

The CATailing concept has not been explored in cancer settings. Therefore, the present provides new insights for widening the therapeutic avenue.

Reviewer #2 (Public Review):

This work explores the connection between glioblastoma, mito-RQC, and msiCAT-tailing. They build upon previous work concluding that ATP5alpha is CAT-tailed and explore how CAT-tailing may affect cell physiology and sensitivity to chemotherapy. The authors conclude that when ATP5alpha is CAT-tailed, it either incorporates into the proton pump or aggregates and that these events dysregulate MPTP opening and mitochondrial membrane potential and that this regulates drug sensitivity. This work includes several intriguing and novel observations connecting cell physiology, RQC, and drug sensitivity. This is also the first time this reviewer has seen an investigation of how a CAT tail may specifically affect the function of a protein.

Comment from the Reviewing Editor:

The revisions made the work more valuable and convincing. The authors adequately made point-by-point response to the reviewers comments by providing new data. Image acquisition and data analysis were further clarified. NEMF knockdown experiments and additional control data for ATP5α featuring a poly-glycine-serine (GS) tail support their conclusion.

Author response:

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

eLife Assessment:

Glioblastoma is one of the most aggressive cancers without a cure. Glioblastoma cells are known to have high mitochondrial potential. This useful study demonstrates the critical role of the ribosome-associated quality control (RQC) pathway in regulating mitochondrial membrane potential and glioblastoma growth. Some assays are incomplete; further revision will improve the significance of this study.

For clarity, we propose revising the second sentence to: "It is well-established that certain cancer cells, such as glioblastoma cells, exhibit elevated mitochondrial membrane potential."

Reviewer #1 (Public Review):

Summary:

Cai et al have investigated the role of msiCAT-tailed mitochondrial proteins that frequently exist in glioblastoma stem cells. Overexpression of msiCAT-tailed mitochondrial ATP synthase F1 subunit alpha (ATP5) protein increases the mitochondrial membrane potential and blocks mitochondrial permeability transition pore formation/opening. These changes in mitochondrial properties provide resistance to staurosporine (STS)-induced apoptosis in GBM cells. Therefore, msiCAT-tailing can promote cell survival and migration, while genetic and pharmacological inhibition of msiCAT-tailing can prevent the overgrowth of GBM cells.

Strengths:

The CAT-tailing concept has not been explored in cancer settings. Therefore, the present provides new insights for widening the therapeutic avenue. 

Your acknowledgment of our study's pioneering elements is greatly appreciated.

Weaknesses:

Although the paper does have strengths in principle, the weaknesses of the paper are that these strengths are not directly demonstrated. The conclusions of this paper are mostly well-supported by data, but some aspects of image acquisition and data analysis need to be clarified and extended.

We are grateful for your acknowledgment of our study’s innovative approach and its possible influence on cancer therapy. We sincerely appreciate your valuable feedback. In response, this updated manuscript presents substantial new findings that reinforce our central argument. Moreover, we have broadened our data analysis and interpretation, as well as refined our methodological descriptions.

Reviewer #2 (Public Review):

This work explores the connection between glioblastoma, mito-RQC, and msiCAT-tailing. They build upon previous work concluding that ATP5alpha is CAT-tailed and explore how CAT-tailing may affect cell physiology and sensitivity to chemotherapy. The authors conclude that when ATP5alpha is CAT-tailed, it either incorporates into the proton pump or aggregates and that these events dysregulate MPTP opening and mitochondrial membrane potential and that this regulates drug sensitivity. This work includes several intriguing and novel observations connecting cell physiology, RQC, and drug sensitivity. This is also the first time this reviewer has seen an investigation of how a CAT tail may specifically affect the function of a protein. However, some of the conclusions in this work are not well supported. This significantly weakens the work but can be addressed through further experiments or by weakening the text.

We appreciate the recognition of our study's novelty. To address your concerns about our conclusions, we have revised the manuscript. This revision includes new data and corrections of identified issues. Our detailed responses to your specific points are outlined below.

Reviewer #1 (Recommendations For The Authors):

(1) In Figure 1B, please replace the high-exposure blots of ATP5 and COX with representative results. The current results are difficult to interpret clearly. Additionally, it would be helpful if the author could explain the nature of the two different bands in NEMF and ANKZF1. Did the authors also examine other RQC factors and mitochondrial ETC proteins? I'm also curious to understand why CAT-tailing is specific to C-I30, ATP5, and COX-V, and why the authors did not show the significance of COX-V.

We appreciate your inquiry regarding the data.  Additional attempts were made using new patient-derived samples; however, these results did not improve upon the existing ATP5⍺, (NDUS3)C-I30, and COX4 signals presented in the figure.  This is possibly due to the fact that CAT-tail modified mitochondrial proteins represent only a small fraction of the total proteins in these cells.  It is acknowledged that the small tails visible above the prominent main bands are not particularly distinct. To address this, the revised version includes updated images to better illustrate the differences. We believe the assertion that GBM/GSCs possess CAT-tailed proteins is substantiated by a combination of subsequent experimental findings. The figure (refer to new Fig. 1B) serves primarily as an introduction. It is important to note that the CAT-tailed ATP5⍺ plays a vital role in modulating mitochondrial potential and glioma phenotypes, a function which has been demonstrated through subsequent experiments.

It is acknowledged that the CAT-tail modification is not exclusive to the ATP5⍺protein.  ATP5⍺ was selected as the primary focus of this study due to its prevalence in mitochondria and its specific involvement in cancer development, as noted by Chang YW et al.  Future research will explore the possibility of CAT tails on other mitochondrial ETC proteins. Currently, NDUS3 (C-I30), ATP5⍺, and COX4 serve as examples confirming the existence of these modifications. It remains challenging to detect endogenous CAT-tailing, and bulk proteomics is not yet feasible for this purpose. COX4 is considered significant.  We hypothesize that CAT-tailed COX4 may function similarly to the previously studied C-I30 (Wu Z, et al), potentially causing substantial mitochondrial proteostasis stress.  

Concerning RQC proteins, our blotting analysis of GBM cell lines now includes additional RQC-related factors. The primary, more prominent bands (indicated by arrowheads) are, in our assessment, the intended bands for NEMF and ANKZF1.  Subsequent blotting analyses showed only single bands for both ANKZF1 and NEMF, respectively. The additional, larger molecular weight band of NEMF, which was initially considered for property analysis (phosphorylation, ubiquitination, etc.), was not examined further as it did not appear in subsequent experiments (refer to new Fig. S1C).

References:

Chang YW, et al. Spatial and temporal dynamics of ATP synthase from mitochondria toward the cell surface. Communications biology. 2023;6(1).

Wu Z, et al. MISTERMINATE Mechanistically Links Mitochondrial Dysfunction With Proteostasis Failure. Molecular cell. 2019;75(4).

(2) In addition to Figure 1B, it would be interesting to explore CAT-tailed mETC proteins in cancer tissue samples.

This is an excellent point, and we appreciate the question. We conducted staining for ATP5⍺ and key RQC proteins in both tumor and normal mouse tissues. Notably, ATP5⍺ in GBM exhibited a greater tendency to form clustered punctate patterns compared to normal brain tissue, and not all of it co-localized with the mitochondrial marker TOM20 (refer to new Fig. S3C-E). Crucially, we observed a significant increase in NEMF expression within mouse xenograft tumor tissues, alongside a decrease in ANKZF1 expression (refer to new Fig. S1A, B). These findings align with our observations in human samples.

(3) Please knock down ATP5 in the patient's cells and check whether both the upper band and lower band of ATP5 have disappeared or not.

This control was essential and has been executed now. To validate the antibody's specificity, siRNA knockdown was performed. The simultaneous elimination of both upper and lower bands upon siRNA treatment (refer to new Fig. S2A) confirms they represent genuine signals recognized by the antibody.

(4) In Figure 1C and ID, add long exposure to spot aggregation and oligomer. Figure 1D, please add the blots where control and ATP5 are also shown in NHA and SF (similar to SVG and GSC827).

New data are included in the revised manuscript to address the queries. Specifically, the new Fig 1D now displays the full queue as requested, featuring blots for Control, ATP5α, AT3, and AT20. Our analysis reveals that AT20 aggregates exhibit higher expression and accumulation rates in GSC and SF cells.

Fig. 1C has been updated to include experimental groups treated with cycloheximide and sgNEMF. Our results show that sgNEMF effectively inhibits CAT-tailing in GBM cell lines, whereas cycloheximide has no impact. After consulting with the Reporter's original creator and optimizing expression conditions, we observed no significant aggregates with β-globin-non-stop protein, potentially due to the length of endogenous CAT-tail formation (as noted by Inada, 2020, in Cell Reports). Our analysis focused on the ratio of CAT-tailed (red box blots) and non-CAT-tailed proteins (green box blots). Comparing these ratios revealed that both anisomycin treatment and sgNEMF effectively hinder the CAT-tailing process, while cycloheximide has no effect.

(5) In Figure 1E, please double-check the results with the figure legend. ATP5A aggregated should be shown endogenously. The number of aggregates shown in the bar graph is not represented in micrographs. Please replace the images. For Figure 1E, to confirm the ATP5-specific aggregates, it would be better if the authors would show endogenous immunostaining of C-130 and Cox-IV.

Labels in Fig. 1E were corrected to reflect that the bar graph in Fig. 1F indicates the number of cells with aggregates, not the quantity of aggregates per cell. The presence

(6) Figure 3A. Please add representative images in the anisomycin sections. It is difficult to address the difference.

We appreciate your feedback. Upon re-examining the Calcein fluorescence intensity data in Fig. 3A, we believe the images accurately represent the statistical variations presented in Fig. 3B. To address your concerns more effectively, please specify which signals in Fig. 3A you find potentially misleading. We are prepared to revise or substitute those images accordingly.

(7) Figure 3D. If NEMF is overexpressed, is the CAT-tailing of ATP 5 reversed?

Thank you. Your prediction aligns with our findings. We've added data to the revised Fig. S6A, B, which demonstrates that both NEMF overexpression and ANKZF1 knockdown lead to elevated levels of CRC. This increase, however, was not statistically significant in GSC cells. A plausible explanation for this discrepancy is that the MPTP of GSC cells is already closed, thus any additional increase in CAT-tailing activity does not result in further amplification.

(8) Figure 3G. Why on the BN page are AT20 aggregates not the same as shown in Figure 2E?

We appreciate your inquiry regarding the ATP5⍺ blots, specifically those in the original Fig. 3G (left) and 2E (right). Careful observation of the ATP5⍺ band placement in these figures reveals a high degree of similarity. Notably, there are aggregates present at the top, and the diffuse signals extend downwards. Given that this is a gradient polyacrylamide native PAGE, the concentration diminishes towards the top. Consequently, the non-rigid nature of the Blue Native PAGE gel may lead to slight variations in the aggregate signals; however, the overall patterns are very much alike. To mitigate potential misinterpretations, we have rearranged the blot order in the new Fig. 3M.

(9) Figure 4D. The amount of aggregation mediated by AT20 is more compared to AT3. Why are there no such drastic effects observed between AT3 and AT20 in the Tunnel assay?

The previous Figure 4D presents the quantification of cell migration from the experiment depicted in Figure 4C. But this is a good point. TUNEL staining results are directly influenced by mitochondrial membrane potential and the state of mitochondrial permeability transition pores

(MPTP), not by the degree of protein aggregation. Our previous experiments showed comparable effects of AT3 and AT20 on mitochondria (Fig. 2E, 3K), which aligns with the expected similar outcomes on TUNEL staining. As for its biological nature, this could be very complicated. We hope to explore it in future studies.

(10) Figure 5C: The role of NEMF and ANKZF1 can be further clarified by conducting Annexin-PI assays using FACS. The inclusion of these additional data points will provide more robust evidence for CAT-tailing's role in cancer cells.

In response to your suggestion, we have incorporated additional data into the revised version.Using the Annexin-PI kit, we labeled apoptotic cells and detected them using flow cytometry (FACS). Our findings indicate that anisomycin pretreatment, NEMF knockdown (sgNEMF), and ANZKF1 upregulation (oeANKZF1) significantly increase the rate of STS-induced apoptosis compared to the control group (refer to new Fig. S9D-G).

(11) Figure 5F: STS is a known apoptosis inhibitor. Why it is not showing PARP cleavage? Also, cell death analysis would be more pronounced, if it could be shown at a later time point. What is the STS and Anisomycin at 24h or 48h time-point? Since PARP is cleaved, it would also be better if the authors could include caspase blots.

I guess what you meant to say here is "Staurosporine is a protein kinase inhibitor that can induce apoptosis in multiple mammalian cell lines." Our study observed PARP cleavage even in GSCs, which are typically more resistant to staurosporine-induced apoptosis (C-PARP in Fig. S9B). The ratio of C-PARP to total PARP increased. We selected a 180-minute treatment duration because longer treatments with STS + anisomycin led to a late stage of apoptosis and non-specific protein degradation (e.g., at 24 or 48 hours), making PARP comparisons less meaningful. Following your suggestion, we also examined caspase 3/7 activity in GSC cells treated with DMSO, CHX, and anisomycin. We found that anisomycin treatment also activated caspases (Fig. S9A).

(12) In Figure 5, the addition of an explanation, how CAT-tailing can induce cell death, would add more information such as BAX-BCL2 ratio, and cytochrome-c release from the mitochondria.

Thank you for your suggestion. In this study, we state that specific CAT-tails inhibit GSC cell death/apoptosis rather than inducing it. Therefore, we do not expect that examining BAX-BCL2 and mitochondrial cytochrome c release would offer additional insights.

(13) To confirm the STS resistance, it would be better if the author could do the experiments in the STS-resistant cell line and then perform the Anisomycin experiments.

Thank you. We should emphasize that our data primarily originates from GSC cells. These cells already exhibit STS-resistance when compared to the control cells (Fig. S8A-C).

(14) It would be more advantageous if the author could show ATP5 CATailed status under standard chemotherapy conditions in either cell lines or in vivo conditions.

This is an interesting question. It's worth exploring this question; however, GSC cells exhibit strong resistance to standard chemotherapy treatments like temozolomide (TMZ).

Additionally, we couldn't detect changes in CAT-tailed ATP5⍺ and thus did not include that data.

(15) In vivo (cancer mouse model or cancer fly model) data will add more weight to the story.

We appreciate your intriguing question. An effective approach would be to test the RQC pathway's function using the Drosophila Notch overexpression-induced brain tumor model. However, Khaket et al. have conducted similar studies, stating, "The RNAi of Clbn, VCP, and Listerin (Ltn), homologs of key components of the yeast RQC machinery, all attenuated NSC over-proliferation induced by Notch OE (Figs. 5A and S5A–D, G)." This data supports our theory, and we have incorporated it into the Discussion. While the mouse model more closely resembles the clinical setting, it is not covered by our current IACUC proposal. We intend to verify this hypothesis in a future study.

Reference:

Khaket TP, Rimal S, Wang X, Bhurtel S, Wu YC, Lu B. Ribosome stalling during c-myc translation presents actionable cancer cell vulnerability. PNAS Nexus. 2024 Aug 13;3(8):pgae321.

Reviewer #2 (Recommendations For The Authors):

Figure 1B, C: To demonstrate that Globin, ATP5alpha, and C-130 are CAT-tailed, it is necessary to show that the high mobility band disappears after NEMF deletion or mutagenesis of the NFACT domain of NEMF. This can be done in a cell line. The anisomycin experiment is not convincing because the intensity of the bands drops and because no control is done to show that the effects are not due to translation inhibition (e.g. cycloheximide, which inhibits translation but not CAT tailing). Establishing ATP5alpha as a bonafide RQC substrate and CAT-tailed protein is critical to the relevance of the rest of the paper.

Thank you for suggesting this crucial control experiment. To confirm the observed signal is indeed a bona fide CAT-tail, it's essential to demonstrate that NEMF is necessary for the CAT-tailing process. We have incorporated data from NEMF knockdown (sgNEMF) and cycloheximide treatment into the revised manuscript. Our findings show that both sgNEMF and anisomycin treatment effectively inhibit the formation of CAT-tailing signals on the reporter protein (Fig. 1C). Similarly, NEMF knockdown in a GSC cell line also effectively eliminated CAT-tails on overexpressed ATP5⍺ (Fig. S2B).

In general, the text should be weakened to reflect that conclusions were largely gleaned from artificial CAT tails made of AT repeats rather than endogenously CAT-tailed ATP5alpha. CAT tails could have other sequences or be made of pure alanine, as has been suggested by some studies.

Thank you for your reminder. We have reviewed the recent studies by Khan et al. and Chang et al., and we found their analysis of CAT tail components to be highly insightful. We concur with your suggestion regarding the design of the CAT tail sequence. We aimed to design a tail that maintained stability and resisted rapid degradation, regardless of its length. In the revised version, we clarify that our conclusions are based on artificial CAT tails, specifically those composed of AT repeat sequences (p. 9). We acknowledge that the presence of other sequence components may lead to different outcomes (p. 19).

Reference:

Khan D, Vinayak AA, Sitron CS, Brandman O. Mechanochemical forces regulate the composition and fate of stalled nascent chains. bioRxiv [Preprint]. 2024 Oct 14:2024.08.02.606406. Chang WD, Yoon MJ, Yeo KH, Choe YJ. Threonine-rich carboxyl-terminal extension drives aggregation of stalled polypeptides. Mol Cell. 2024 Nov 21;84(22):4334-4349.e7. 

Throughout the work (e.g. 3B, C), anisomycin effects should be compared to those with cycloheximide to observe if the effects are specific to a CAT tail inhibitor rather than a translation inhibitor.

We agree that including cycloheximide control experiments is crucial. The revised version now incorporates new data, as depicted in Fig. S5A, B, illustrating alterations in the on/off state of MPTP following cycloheximide treatment. Furthermore, Fig. S6A, B present changes in Calcium Retention Capacity (CRC) under cycloheximide treatment. The consistency of results across these experiments, despite cycloheximide treatment, suggests that anisomycin's role is specifically as a CAT tail inhibitor, rather than a translation inhibitor.

Line 110, it is unclear what "short-tailed ATP5" is. Do you mean ATP5alpha-AT3? If so this needs to be introduced properly. Line 132: should say "may indicate accumulation of CAT-tailed protein" rather than "imply".

We acknowledge your points. We have clarified that the "short-tailed ATP5α" refers to ATP5α-AT3 and incorporated the requested changes into the revised manuscript.

Figure 1C: how big are those potential CAT-tails (need to be verified as mentioned earlier)?They look gigantic. Include a ladder.

In the revised Fig. 1D, molecular weight markers have been included to denote signal sizes. The aggregates in the previous Fig. 1C, also present in the control plasmid, are likely a result of signal overexposure. The CAT-tailed protein is observed just above the intended band in these blots. These aggregates have been re-presented in the updated figures, and their signal intensities quantified.

Line 170: "indicating that GBM cells have more capability to deal with protein aggregation". This logic is unclear. Please explain.

We appreciate your question and have thoroughly re-evaluated our conclusion. We offer several potential explanations for the data presented in Fig. 1D: (1) ATP5α-AT20 may demonstrate superior stability. (2) GSC (GBM) cells might lack adequate mechanisms to monitor protein accumulation. (3) GSC (GBM) cells could possess an increased adaptive capacity to the toxicity arising from protein accumulation. This discussion has been incorporated into the revised manuscript (lines 166-169).

Line 177: how do you know the endogenous ATP5alpha forms aggregates due to CAT-tailing? Need to measure in a NEMF hypomorph.

We understand your concern and have addressed it. Revised Fig. 3G, H demonstrates that a reduction in NEMF levels, achieved through sgNEMF in GSC cells, significantly diminishes ATP5α aggregation. This, in conjunction with the Anisomycin treatment data presented in revised Fig. 3E, F, confirms the substantial impact of the CAT-tailing process on this aggregation.

Line 218: really need a cycloheximide or NEMF hypomorph control to show this specific to CAT-tailing.

We have revised the manuscript to include data from sgNEMF and cycloheximide treatments, specifically Fig. 3G, H, and Fig. S5C, D, as detailed in our response above.

Lines 249,266, Figure 5A: The mentioned experiments would benefit from controls including an extension of ATP5alpha that was not alanine and threonine, perhaps a gly-ser linker, as well as an NEMF hypomorph.

We sincerely appreciate your insightful comments. In response, the revised manuscript now incorporates control data for ATP5α featuring a poly-glycine-serine (GS) tail. This data is specifically presented in Figs. S2E-G, S4E, S7A, D, E, and S8F, G. Our experimental findings consistently demonstrate that the overexpression of ATP5α, when modified with GS tails, had no discernible impact on protein aggregation, mitochondrial membrane potential, GSC cell mobility, or any other indicators assessed in our study.

Figure S5A should be part of the main figures and not in the supplement.

This has been moved to the main figure (Fig. 5C).

This requires an Adult type character in the room, sad thing is we don't have that. There are no "Adult" gen Z characters. I haven't even seen any Millenial "Adults".

When I say Adult imagine a competent school principal that everyone respects and listens to. The kids are sort of scared from him but they know they can talk to him in a friendly way. When there is a dispute he is the arbitrator that people trust.

There is no "Arbitrator people trust" in organizations any more. There is just HR.

There's also a race problem here, not just gender. The Indian in group preference is very noticeable.

Quote from the Article

Many people think wokeness is over, slain by the vibe shift, but if wokeness is the result of demographic feminization, then it will never be over as long as the demographics remain unchanged.

Most important quote of the article

Wow that's a strong statement

Nice to see "The Longhouse" mentioned in here

The Lawyers, and the way law is interpreted, of the 2050's are going to be very different from the 1950's

I feel like this would be used to bully people, like if "Mean Girls" politics resonates with real life this is a dangerous president.

Wow there buddy, I may need to come back and read this later, that's an intense statement

Is there are presidence for this?

There is a meme for this I saw on Twitter, "But not all X are Y"

A knowledge garden of contested NYT facts and biases would be interesting, I wonder if an AI could do it

Possibly like fighting over a mate

That's a Heuristic I plan on using in the future

This is very well articulated, I think about this all the time but this really get's to the point and makes it clear

I would like to know more about the attributes of this 29% of men verses the 71% of men. Do they go to the gym? What do they eat? What was their father like growing up?

The Political Parties are Gender Sex Based now,

I like the idea of a Man get's a vote, if he get's married he get's two, and if he has over 3+ children then has three votes. Get divorced, only gets one now

Those are some good dates to remember

Okay let me process this,

So if women attain positions of power, then power starts to operate in a feminized way

Wokeness is just the byproduct of people wielding Feminized Power

Ah gotcha, makes sense now

Reminds me of the theme in Wicked: For Good where "The Wizard" of says if Alphaba tells the people of Oz the truth they will not believe her because they will not want to

Ideology producing a physiological response is fascinating

I would like to see some definitive sources defining what Masculine and Feminine power look like

A web search for "Feminine Power" returns this -> True Feminine Power & Why Modern Female Empowerment Has Got It All Wrong | by Anna Rova | Medium

Google's Ideological Echo Chamber - Wikipedia

The gender dynamic in "Woke", "Leftist" culture is facinating. Like what does leadership look like in those communities, how much mob mentality is there, how does one attain power (Social Reputation + Audience) in those communities

eLife Assessment

Weindel et al examine behavioural and EEG data in an innovative contrast comparison paradigm where they vary mean contrast widely while keeping contrast difference constant. As intended, this allowed an elegant decomposition of processing stages: while sensory encoding shortened with increasing contrast in keeping with Pieron's law, the period of decision formation lengthened, in keeping with Fechner's law, which was applied to drift rates in a diffusion model of that period. This is an important demonstration of how these two laws apply in concert, to two distinct processing levels, and the multivariate topography parsing, mixed effect models and diffusion models are convincing.

Reviewer #1 (Public review):

This study uses a new 'hidden multivariate pattern method' to parse in time and space the neural events intervening between stimulus and response in an immediately-reported perceptual decision, and use the resultant neural event timing information to show quite convincingly that Pieron's and Fechner's laws can apply in concert at distinct processing levels.

They designed a clever contrast comparison paradigm in which the contrast difference is kept constant while widely manipulating mean contrast, so that sensory encoding of the overall stimulus would be boosted with increasing mean contrast, whereas decision difficulty and hence duration would increase. With this, they found that the time intervening between early sensory-evoked components, up to an 'N200'-type component associated with launching the decision process, varies inversely with contrast according to Pieron's law. Meanwhile, the time intervals running up to neural events peaking near the time of response, consistent with decision termination, increases with contrast, fitting Fechner's law. Further, a diffusion model whose drift rates are scaled by Fechner's law, fit to RT, predicts the observed proportion of correct responses very well.

In the process of review and revision it was highlighted that presumably the full sequence of neural events intervening between stimulus and response is massively task dependent, but;

(1) The method is intended to capture all key components that specifically covary with RT, as opposed to each and every component in general, and

(2) The main conclusions of the study mentioned above do not change whether the method is set up to track three neural events, or five, as was done in the final analysis.

The propensity for topographic parsing algorithms to potentially lump-together distinct processes that partially co-evolve was acknowledged, but a key clarification in review was that even though the method entails a specification of neural event duration - which was changed from 50 to 25 ms - the success of the method is not strongly contingent on the actual underlying neural events in question having that very duration - indeed, the components extracted using that short template duration can be observed to evolve over a longer time frame associated with the Fechner diffusion process.

Notably, standard average event-related potential analysis was able to show expected amplitude effects - where sensory signals increased with contrast but decision signals decreased - but assessment of the by-trial distribution of their timings was grealy aided by the HMP method.

One of the stages of processing implicated in the parsing analysis was linked to attention orientation, and the authors speculate on whether this might reflect a spatially-selective deployment of attention or a resource allocation, but sensibly refrain from speculating too far since the focus here was on the sensory and decision process durations and their respective adherence to Pieron and Fechner's laws.

Reviewer #2 (Public review):

Summary:

The authors decomposed response times into component processes and manipulated the duration of these processes in opposing directions by varying contrast, and overall by manipulating speed-accuracy tradeoffs. They identify different processes and their durations by identifying neural states in time and validate their functional significance by showing that their properties vary selectively as expected with predicted effects of the contrast manipulation. They identify 4 processes: stimulus encoding, attention orienting, decision and motor execution. These map onto 5 classical event related potentials. The decision-making component matched the CPP and its properties varied with contrast and predicted decision-accuracy.

Strengths:

The design of the experiment is remarkable and offers crucial insights. The analyses techniques are beyond-state-of-the art and the analyses are well motivated and offer clear insights.

Weaknesses:

The number of identified events depends on the parameter setting of the analysis. While the authors discuss weaknesses of the approach this needs to be made explicit as well. It is also unclear to what extent topographies map onto processes since e.g., different combinations of sources can lead to the same scalp topography.

Reviewer #3 (Public review):

Summary:

In this manuscript the authors examine the processing stages involved in perceptual decision-making using a new approach to analysing EEG data, combined with a critical stimulus manipulation. This new EEG analysis method enables single-trial estimates of the timing and amplitude of transient changes in EEG time-series recurrent across trials in a behavioural task. The authors find evidence for five events between stimulus onset and the response in a two-spatial-interval visual discrimination task. By analysing the timing and amplitude of these events in relation to behaviour and the stimulus manipulation, the authors interpret these events as related to separable processing stages for stimulus encoding (first two events), attention orientation (second event), motor planning (fourth event) and decision (deliberation, final event). This is largely consistent with previous findings from both event-related potentials (across trials) and single-trial estimates using decoding techniques and neural network approaches. However, by taking a data-driven approach (as opposed to theory-driven decoding analyses) a more nuanced picture emerges: there are several stimulus encoding steps which may contribute differently to behaviour, and decision processes extend beyond the planning of the motor response.

Strengths:

This work is not only important for the conceptual advance, but also in promoting this new analysis technique, which will likely prove useful in future research. For the broader picture, this work is an excellent example of the utility of neural measures for mental chronometry.

Weaknesses:

Though beyond the scope of this manuscript, these results should be considered within the broader decision-making literature, where task or domain-specific processes may not generalise (for example, in value-based decision-making).

Author response:

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

Reviewer #1 (Public review):  

From my reading, this study aimed to achieve two things:  

(1) A neurally-informed account of how Pieron's and Fechner's laws can apply in concert at distinct processing levels.  

(2) A comprehensive map in time and space of all neural events intervening between stimulus and response in an immediately-reported perceptual decision.  

I believe that the authors achieved the first point, mainly owing to a clever contrast comparison paradigm, but with good help also from a new topographic parsing algorithm they created. With this, they found that the time intervening between an early initial sensory evoked potential and an "N2" type process associated with launching the decision process varies inversely with contrast according to Pieron's law. Meanwhile, the interval from that second event up to a neural event peaking just before response increases with contrast, fitting Fechner's law, and a very nice finding is that a diffusion model whose drift rates are scaled by Fechner's law, fit to RT, predicts the observed proportion of correct responses very well. These are all strengths of the study.   

We thank the reviewer for their comments that added context to the events we detected in relation to previous findings. We also believe that the change in the HMP algorithm suggested by the reviewer improved the precision of our analyses and the manuscript. We respond to the reviewer’s specific comments below.

(1) The second, generally stated aim above is, in the opinion of this reviewer, unconvincing and ill-defined. Presumably, the full sequence of neural events is massively task-dependent, and surely it is more in number than just three. Even the sensory evoked potential typically observed for average ERPs, even for passive viewing, would include a series of 3 or more components - C1, P1, N1, etc. So are some events being missed? Perhaps the authors are identifying key events that impressively demarcate Pieron- and Fechner-adherent sections of the RT, but they might want to temper the claim that they are finding ALL events. In addition, the propensity for topographic parsing algorithms to potentially lump together distinct processes that partially co-evolve should be acknowledged.  

We agree with the reviewer that the topographical solutions found by HMP will be dependent on the task and the quality and type of data. We address this point in the last section of the discussion (see also response to R3.5). We would also like to add that the events detected by HMP are, by construction, those that contribute to the RT and not necessarily all ERPs elicited by a stimulus.

In addition to the new last section of the discussion we also make these points clear in the revised manuscript at the discussion start: 

“By modeling the recorded single-trial EEG signal between stimulus onset and response as a sequence of multivariate events with varying by-trial peak times, we  aimed to detect recurrent events that contribute to the duration of the reaction time in the present perceptual decision-making task”.

Regarding the typical visual ERPs, in response to this comment but also comments R1.2, R1.3 and R2.1, we aimed for a more precise description of the topographies and thus reduced the width of the HMP expected events to 25ms. This ensures that we do not miss events shorter than the initial expectations of 50ms (see Appendix B of Weindel et al., 2024 and also response to  R1.3). This new estimation provides evidence for at least two of the visual ERPs that, based on their timings and topographies (in relation with the spatial frequency of the stimulus), we interpret as the N40 and the P100 (see response to R1.5 for the justification of this categorization). We provide a description and justification of the interpretations in the result section “Five trial-recurrent sequential events occur in the EEG during decisions” and the discussion section “Visual encoding time”.

(2) To take a salient example, the last neural event seems to blend the centroparietal positivity with a more frontal midline negativity, some of which would capture the CNV and some motor-execution related components that are more tightly time-locked to, of course, the response. If the authors plotted the traditional single-electrode ERP at the frontal focus and centroparietal focus separately, they are likely to see very different dynamics and contrast- and SAT-dependency. What does this mean for the validity of the multivariate method? If two or more components are being lumped into one neural event, wouldn't it mean that properties of one (e.g., frontal burstiness at response) are being misattributed to the other (centroparietal signal that also peaks but less sharply at response)?

Using the new HMP parameterization described above we show that the reviewer's intuition was correct. Using an expected pattern duration of 25ms the last event in the original manuscript splits in two events. The before-last event, now referred to the lateralized readiness potential (LRP) presents a strong lateralization (Figure 3) with an increased negativity over the motor cortex contralateral to the right hand. The effect of contrast is mostly on the last event that we interpret as the CPP (Figure 5). Despite the improved precision of the topographies of the identified events, it is however to be noted that some components will overlap. If the LRP is generated when a certain amount of evidence is accumulated (e.g. that the CPP crosses a certain value) then a time-based topography will necessarily include that CPP activity in addition to the lateralized potential. We discuss this in the section “Motor execution” of the discussion:

“Adding the abrupt onset of this potential, we believe that this event is the start of motor execution, engaged after a certain amount of evidence. The evidence for this interpretation is manifest in the fact that the event's topography shares some activity with the CPP event that follows, an expected result if the LRP is triggered at a certain amount of evidence, indexed by the CPP”.

(3) Also related to the method, why must the neural events all be 50 ms wide, and what happens if that is changed? Is it realistic that these neural events would be the same duration on every trial, even if their duration was a free parameter? This might be reasonable for sensory and motor components, but unlikely for cognitive.  

The HMP method is sensitive to the event's duration as shown in the manuscript about the method (Appendix B of Weindel et al., 2024). Nevertheless as long as the topography in the real data is longer than the expected one it shouldn't be missed (i.e. same goes for by-trial variations in the event width). For this reason we halved the expected event width of 50ms (introduced by the original HsMM-MVPA paper by Anderson and colleagues) in the revision. This new estimation with 25ms thus is much less likely to miss events as evidenced by the new visual and motor events. In the revised manuscript this is addressed at the start of the Results section:

“Contrary to previous applications (Anderson et al.,2016; Berberyan et al., 2021; Zhang et al., 2018; Krause et al., 2024) we assumed that the multivariate pattern was represented by a 25ms half-sine as our previous research showed that a shorter expected pattern width increases the likelihood of detecting cognitive events (see Appendix B of Weindel et al., 2024)”.

Regarding the event width as a free parameter this is both technically and statistically difficult to implement as the amount of computing capacity, flexibility and trade-offs among the HMP parameters would, given the current implementation, render the model unfit for most computers and statistically unidentifiable.

(4) In general, I wonder about the analytic advantage of the parsing method - the paradigm itself is so well-designed that the story may be clear from standard average event-related potential analysis, and this might sidestep the doubts around whether the algorithm is correctly parsing all neural events.  

Average ERP analysis suffers from an impossibility to differentiate between an effect of an experimental factor on the amplitude vs. on the timing of the underlying components (Luck, 2005). Furthermore the overlap of components across trials bluries the distinction between them. For both reasons we would not be able to reach the same level of certainty and precision using ERP analyses. Furthermore the relatively low number of trials per experimental cell (contrast level X SAT X participant = 6 trials) makes the analyses hard to perform on ERP which typically require more trials per modality. From the reviewer’s comment we understand that this point was not clear. We therefore discuss this in the revision, Section “Functional interpretation of the events” of the results:

“Nevertheless identifying neural dynamics on these ERPs centered on stimulus is complicated by the time variation of the underlying single-trial events (see probabilities displayed in Figure 3 for an illustration and Burle et al., 2008, for a discussion). The likely impact of contrast on both amplitude and time on the underlying single-trial event does not allow one to interpret the average ERP traces as showing an effect in one or the other dimension without strong assumptions (Luck, 2005)”.

(5) In particular, would the authors consider plotting CPP waveforms in the traditional way, across contrast levels? The elegant design is such that the C1 component (which has similar topography) will show up negative and early, giving way to the CPP, and these two components will show opposite amplitude variations (not just temporal intervals as is this paper's main focus), because the brighter the two gratings, the stronger the aggregate early sensory response but the weaker the decision evidence due to Fechner. I believe this would provide a simple, helpful corroborating analysis to back up the main functional interpretation in the paper.  

We agree with the suggestion and have introduced the representation on top of Figure 5 for sets of three electrodes in the occipital, posterior and frontal regions. The new panels clearly show an inversion of the contrast effect dependent on the time and locus of the electrodes. We discuss this in Section “Functional interpretation of the events” of the results:

“This representation shows that there is an inversion of the contrast effect with higher contrasts having a higher amplitude on the electrodes associated with visual potentials in the first couple of deciseconds (left panel of Figure 5A) while parietal and frontal electrodes shows a higher amplitude for lower contrasts in later portions of the ERPs (middle and right panel of Figure 5A)”.

To us, this crucially shows that we cannot achieve the same decomposition using traditional ERP analyses. In these plots it appears that while, as described by the reviewer, there is an inversion, the timing and amplitude of the changes due to contrast can hardly be interpreted.

(6) The first component is picking up on the C1 component (which is negative for these stimulus locations), not a "P100". Please consult any visual evoked potential study (e.g., Luck, Hillyard, etc). It is unexpected that this does not vary in latency with contrast - see, for example. Gebodh et al (2017, Brain Topography) - and there is little discussion of this. Could it be that nonlinear trends were not correctly tested for?  

We disagree with the reviewer on the interpretation of the ERP. The timing of the detected component is later than the one usually associated with a C1. Furthermore the central display does not create optimal conditions to detect a C1

We do agree that the topography raises the confusion but we believe that this is due to the spatial frequency of the stimulus that generates a high posterior positivity (see references in the following extract). The new HMP solution also now happens to show an effect of contrast on the P100 latencies, we believe this is due to the increased precision in the time location of the component. We discuss this in the “Visual encoding time” section of the discussion:

“The following event, the P100, is expressed around 70ms after the N40, its topography is congruent with reports for stimuli with low spatial frequencies as used in the current study (Kenemans et al., 2002, 2000; Proverbio et al., 1996). The timing of this P100 component is changed by the contrast of the stimulus in the direction expected by the Piéron law (Figure 4A)”. 

(7) There is very little analysis or discussion of the second stage linked to attention orientation - what would the role of attention orientation be in this task? Is it spatial attention directed to the higher contrast grating (and if so, should it lateralise accordingly?), or is it more of an alerting function the authors have in mind here?  

We agree that we were not specific enough on the interpretation of this attention stage. We now discuss our hypothesis in the section “Attention orientation” of the discussion:  

“We do however observe an asymmetry in the topographical map Figure 3. This asymmetry might point to an attentional bias with participants (or at least some participants) allocating attention to one side over the other in the same way as the N2pc component (Luck and Hillyard, 1994, Luck et al., 1997). Based on this collection of observations, we conclude that this third event represents an attention orientation process. In line with the finding of Philiastides et al. (2006), this attention orientation event might also relate to the allocation of resources. Other designs varying the expected cognitive load or spatial attention could help in further interpreting the functional role of this third event”.

We would like to add that it is unlikely that the asymmetry we mention in the discussion cannot stem from the redirection towards higher contrast as the experimental design balanced the side of presentation. We therefore believe that this is a behavioral bias rather than a bias toward the highest contrast stimulus as suggested by the reviewer. We hope that, while more could be tested and discussed, this discussion is sufficient given the current manuscript's goal.

Reviewer #2 (Public review):  

Summary:  

The authors decomposed response times into component processes and manipulated the duration of these processes in opposing directions by varying contrast, and overall by manipulating speed-accuracy tradeoffs. They identify different processes and their durations by identifying neural states in time and validate their functional significance by showing that their properties vary selectively as expected with the predicted effects of the contrast manipulation. They identify 3 processes: stimulus encoding, attention orienting, and decision. These map onto classical event-related potentials. The decision-making component matched the CPP, and its properties varied with contrast and predicted decision-accuracy, while also exhibiting a burst not characteristic of evidence accumulation.  

Strengths:  

The design of the experiment is remarkable and offers crucial insights. The analysis techniques are beyond state-of-the-art, and the analyses are well motivated and offer clear insights.  

Weaknesses:  

It is not clear to me that the results confirm that there are only 3 processes, since e.g., motor preparation and execution were not captured. While the authors discuss this, this is a clear weakness of the approach, as other components may also have been missed. It is also unclear to what extent topographies map onto processes, since, e.g., different combinations of sources can lead to the same scalp topography.  

We thank the reviewer for their kind words and for the attention they brought on the question of the missing motor preparation event. In light of this comment (and also R1.1, R3.3) the revised manuscript uses a finer grained approach for the multivariate event detection. This preciser estimation comes from the use of a shorter expected pattern in which the initial expectation of a 50ms half-sine was halved, therefore ensuring that we do not miss events shorter than the initial expectations (see Appendix B of Weindel et al., 2024 and also response to  R1.3). In the new solution the motor component that the reviewer expected is found as evidenced by the topography of the event, its lateralization and a time-to-response congruent with a response execution event. This is now described in the section “Motor execution” of the revised manuscript: 

“The before last event, identified as the LRP, shows a strong hemispheric asymmetry congruent with a right hand response. The peak of this event is approximately 100 ms before the response which is congruent with reports that the LRP peaks at the onset of electromyographical activity in the effector muscle (Burle et al., 2004), typically happening 100ms before the response in such decision-making tasks (Weindel et al., 2021). Furthermore, while its peak time is dependent on contrast, its expression in the EEG is less clearly related to the contrast manipulation than the following CPP event”.

Reviewer #3 (Public review):  

Summary:  

In this manuscript, the authors examine the processing stages involved in perceptual decision-making using a new approach to analysing EEG data, combined with a critical stimulus manipulation. This new EEG analysis method enables single-trial estimates of the timing and amplitude of transient changes in EEG time-series, recurrent across trials in a behavioural task. The authors find evidence for three events between stimulus onset and the response in a two-spatial-interval visual discrimination task. By analysing the timing and amplitude of these events in relation to behaviour and the stimulus manipulation, the authors interpret these events as related to separable processing stages for stimulus encoding, attention orientation, and decision (deliberation). This is largely consistent with previous findings from both event-related potentials (across trials) and single-trial estimates using decoding techniques and neural network approaches.  

Strengths:  

This work is not only important for the conceptual advance, but also in promoting this new analysis technique, which will likely prove useful in future research. For the broader picture, this work is an excellent example of the utility of neural measures for mental chronometry.  

We appreciate the very positive review and thank the reviewer for pointing out important weaknesses in our original manuscript and also providing resources to address them in the recommendations to authors. Below we comment on each identified weakness and how we addressed them.   

Weaknesses:  

(1) The manuscript would benefit from some conceptual clarifications, which are important for readers to understand this manuscript as a stand-alone work. This includes clearer definitions of Piéron's and Fechner's laws, and a fuller description of the EEG analysis technique.

We agree that the description of both laws were insufficient, we therefore added the following text in the last paragraph of the introduction:

“Piéron’s law predicts that the time to perceive the two stimuli (and thus the choice situation) should follow a negative power law with the stimulus intensity (Figure 1, green curve). In contradistinction, Fechner’s law states that the perceived difference between the two patches follows the logarithm of the absolute contrast of the two patches (Figure 1, yellow curve). As the task of our participants is to judge the contrast difference, Piéron’s law should predict the time at which the comparison starts (i.e. the stimuli become perceptible), while Fechner’s law should implement the comparison, and thus decision, difficulty”.

Regarding the EEG analysis technique we added a few elements at the start of the result:

“The hidden multivariate pattern model (HMP) implemented assumed that a task-related multivariate pattern event is represented by a half-sine whose timing varies from trial to trial based on a gamma distribution with a shape parameter of 2 and a scale, controlling the average latency of the event, free-to-vary per event (Weindel et al., 2024)”.

We also made the technique clearer at the start of the discussion:

“By modeling the recorded single-trial EEG signal between stimulus onset and response as a sequence of multivariate events with varying by-trial peak times, we aimed to detect recurrent events that contribute to the duration of the reaction time in the present perceptual decision-making task. In addition to the number of events, using this hidden multivariate pattern approach (Weindel et al., 2024) we estimated the trial-by-trial probability of each event’s peak, therefore accessing at which time sample each event was the most likely to occur”.

Additionally, we added a proper description in the method section (see the new first paragraph of the “Hidden multivariate pattern” subsection). 

(2) The manuscript, broadly, but the introduction especially, may be improved by clearly delineating the multiple aims of this project: examining the processes for decision-making, obtaining single-trial estimates of meaningful EEG-events, and whether central parietal positivity reflects ramping activity or steps averaged across trials.

For the sake of clarity we removed the question of the ramping activity vs steps in the introduction and focused on the processes in decision-making and their single-trial measurement as this is the main topic of the paper. Furthermore the references provided by the reviewer allowed us to write a more comprehensive review of previous studies and how the current study is in line with those. These changes are mainly manifested in these new sentences:

“As an example Philiastides et al. (2006) used a classifier on the EEG activity of several conditions to show that the strength of an early EEG component was proportional to the strength of the stimulus while a later component was related to decision difficulty and behavioral performance (see also Salvador et al., 2022; Philiastides and Sajda, 2006). Furthermore the authors interpreted that a third EEG component was indicative of the resource allocated to the upcoming decision given the perceived decision difficulty. In their study, they showed that it is possible to use single-trial information to separate cognitive processes within decision-making. Nevertheless, their method requires a decoding approach, which requires separate classifiers for each component of interest and restrains the detection of the components to those with decodable discriminating features (e.g. stimuli with strong neural generators such as face stimuli, see Philiastides et al., 2006)”.

(3) A fuller discussion of the limitations of the work, in particular, the absence of motor contributions to reaction time, would also be appreciated. 

As laid out in responses to comments R1.1 and R2 the new estimates now include evidence for a motor preparation component. We discuss this in the new “motor execution” paragraph in the discussion section. Additionally we discuss the limitation of the study and the method in the two last paragraphs of the discussion (in the new Section “Generalization and limitation”).

(4) At times, the novelty of the work is perhaps overstated. Rather, readers may appreciate a more comprehensive discussion of the distinctions between the current work and previous techniques to gauge single-trial estimates of decision-related activity, as well as previous findings concerning distinct processing stages in decision-making. Moreover, a discussion of how the events described in this study might generalise to different decision-making tasks in different contexts (for example, in auditory perception, or even value-based decision-making) would also be appreciated.  

We agree that the original text could be read as overstating. In addition to the changes linked to R3.2 we also now discuss the link with the previous studies in the before-last paragraph of the discussion before the conclusion in the new “Generalization and limitations” section:

“The present study showed what cognitive processes are contributing to the reaction time and estimated single-trial times of these processes for this specific perceptual decision-making task. The identified processes and topographies ought to be dependent on the task and even the stimuli (e.g. sensory events will change with the sensory modality). More complex designs might generate a higher number of cognitive processes (e.g. memory retrieval from a cue, Anderson et al., 2016) and so could more natural stimuli which might trigger other processes in the EEG (e.g. appraisal vs. choice as shown by Frömer et al., 2024). Nevertheless, the observation of early sensory vs. late decision EEG components is likely to generalize across many stimuli and tasks as it has been observed in other designs and methods (Philiastides et al., 2006; Salvador et al., 2022). To these studies we add that we can evaluate the trial-level contribution, as already done for specific processes (e.g. Si et al., 2020; Sturm et al., 2016), for the collection of events detected in the current study”.

Reviewing Editor Comments:  

As you will see, all three reviewers agree that the paper makes a valuable contribution and has many strengths. You will also see that they have provided a range of constructive comments highlighting potential issues with the interpretation of the outcomes of your signal decomposition method. In particular, all three reviewers point out that your results do not identify separate motor preparation signals, which we know must be operating on this type of task. The reviewers suggest further discussion of this issue and the potential limitations of your analysis approach, as well as suggesting some additional analyses that could be run to explore this further. While making these changes would undoubtedly enhance the paper and the final public reviews, I should note that my sense is that they are unlikely to change the reviewers' ratings of the significance of the findings and the strength of evidence in the final eLife assessment  

Reviewer #1 (Recommendations for the authors):  

(1) Abstract: "choice onset" is ill-defined and not the label most would give the start of the RT interval. Do you mean stimulus onset?  

We replaced with "choice onset" with "stimulus onset" in the abstract

(2) Similarly "choice elements" in the introduction seem to refer to sensory attributes/objects being decided about?  

We replaced "choice-elements" with "choice-relevant features of the stimuli"

(3) "how the RT emerges from these putative components" - it would be helpful to specify more what level of answer you're looking for, as one could simply answer "when they're done."  

We replaced with "how the variability in RTs emerges from these putative components"

(4) Line 61-62: I'm not sure this is a fully correct characterisation of Frömer et al. It was not similar in invoking a step function - it did not invoke any particular mechanism or function, and in that respect does not compare well to Latimer et al. Also, I believe it was the overlap of stimulus-locked components, not response-locked, that they argued could falsely generate accumulator-like buildup in the response-locked ERP.  

We indeed wrongly described Frömer et al. The sentence is now "In human EEG data, the classical observation of a slowly evolving centro-parietal positivity, scaling with evidence accumulation, was suggested to result from the overlap of time-varying stimulus-related activity in the response-locked event related potential"

(5) Line 78: Should this be single-trial *latency*?  

This referred to location in time but we agree that the term is confusing and thus replaced it with latencies.

(6) The caption of Figure 1 should state what is meant by the y-axis "time"  

We added the sentence "The y-axis refers the time predicted by each law given a contrast value (x-axis) and the chosen set of parameters." in the caption of Figure 1

(7) Line 107: Is this the correct description of Fechner's law? If the perceived difference follows the log of the physical difference, then a constant physical difference should mean a constant perceived difference. Perhaps a typo here.  

This was indeed a typo we replaced the corresponding part of the sentence with "the perceived difference between the two patches follows the logarithm of the absolute contrast of the two patches"

(8) Line 128: By scale, do you mean magnitude/amplitude?  

No, this refers to the parameter of a gamma distribution. To clarify we edited the sentence:  "based on a gamma distribution with a shape parameter of 2 and a scale parameter, controlling the average latency of the event, free-to-vary per event"

(9) The caption of Figure 3 is insufficient to make sense of the top panel. What does the inter-event interval mean, and why is it important to show? What is the "response" event?  

We agree that the top panel was insufficiently described. To keep the length of the paper short and because of the relatively low amount of information provided by these panels we replaced them for a figure only showing the average topographies as well as the asymmetry tests for each event.

(10) Figure 4: caption should say what the top vs bottom row represents (presumably, accuracy vs speed emphasis?), and what the individual dots represent, given the caption says these are "trial and participant averaged". A legend should be provided for the rightmost panels.  

We agree and therefore edited Figure 4. The beginning of the caption mentioned by the reviewer now reads: “A) The panels represent the average duration between events for each contrast level, averaged across participants and trials (stimulus and response respectively as first and last events) for accuracy (top) and speed instructions (bottom).”. Additionally we added legends for the SAT instructions and the model fits.

(11) Line 189: argued for a decision-making role of what?  

Stafford and Gurney (2004) proposed that Pieron’s law could reflect a non-linear transformation from sensory input to action outcomes, which they argued reflected a response mechanism. We (Van Maanen et al., 2012) specified this result by showing that a Bayesian Observer Model in which evidence for two alternative options was accumulated following Bayes Rule indeed predicted a power relation between the difference in sensory input of the two alternatives, and mean RT. However, the current data suggest that such an explanation cannot be the full story, as also noted by R3. To clarify this point we replaced the comment by the following sentence:

“Note that this observation is not necessarily incongruent with theoretical work that argued that Piéron’s law could also be a result of a response selection mechanism (Stafford and Gurney, 2004; Van Maanen et al., 2012; Palmer et al., 2005). It could be that differences in stimulus intensity between the two options also contribute to a Piéron-like relationship in the later intervals, that is convoluted with Fechner’s law (see Donkin and Van Maanen, 2014 for a similar argument). Unfortunately, our data do not allow us to discriminate between a pure logarithmic growth function and one that is mediated by a decreasing power function”.

(12) Table 2: There is an SAT effect even on the first interval, which is quite remarkable and could be discussed more - does this mean that the C1 component occurs earlier under speed pressure? This would be the first such finding.  

The original event we qualified as a P100 was sensitive to SAT but the earliest event is now the N40 and isn’t statistically sensitive to speed pressure in this data. We believe that the fact that the P100 is still sensitive to SAT is not a surprise and therefore do not outline it.

(13) Line 221: "decrease of activation when contrast (and thus difficulty) increases" - is this shown somewhere in the paper?  

The whole section for this analysis was rewritten (see comment below)

(14) I find the analysis of Figure 5 interesting, but the interpretation odd. What is found is that the peak of the decision signal aligns with the response, consistent with previous work, but the authors choose to interpret this as the decision signal "occurring as a short-lived burst." Where is the quantitative analysis of its duration across trials? It can at least be visually appraised in the surface plot, and this shows that the signal has a stimulus-locked onset and, apart from the slowest RTs, remains present and for the most part building, until response. What about this is burst-like? A peak is not a burst.  

This was the residue of a previous version of the paper where an analysis reported that no evidence accumulation trace was found. But after proper simulations this analysis turned out to be false because of a poor statistical test. Thus we removed this paragraph in the revised manuscript and Figure 5 has now been extended to include surface plots for all the events.

Reviewer #2 (Recommendations for the authors):  

Overall, I really enjoyed reading this paper. However, in some places the approach is a bit opaque or the results are difficult to follow. As I read the paper, I noted:  

Did you do a simple DDM, or did you do a collapsing bound for speed?  

The fitted DDM was an adaptation of the proportional rate diffusion model. We make this clearer at the end of the introduction: "Given that Fechner’s law is expected to capture decision difficulty we connected this law to the classical diffusion decision models by replacing the rate of accumulation with Fechner’s law in the proportional rate diffusion model of Palmer et al.(2005).”

It is confusing that the order of intervals in the text doesn't match the order in the table. It might be better to say what events the interval is between rather than assuming that the reader reconstructs.  

We agree and adapted the order in both the text and the table. The table is now also more explicit (e.g. RT instead of S-R)

Otherwise, I do wonder to what extent the method is able to differentiate processes that yield similar scalp topographies and find it a bit concerning that no motor component was identified.  

We believe that the new version with the LRP/CPP is a demonstration that the method can handle similar topographies. The method can handle events with close topographies as long as they are separate in time, however if they are not sequential to one another the method cannot capture both events. We now discuss this, in relation with the C1/P100 overlap, in the discussion section “Visual encoding time”:

“Nevertheless this event, seemingly overlapping with the P100 even at the trial level (Figure 5C), cannot be recovered by the method we applied. The fact that the P100 was recovered instead of the C1 could indicate that only the timing of the P100 contributes to the RT (see Section 3 of Weindel et al., 2024)”.

And we more generally address the question of overlap in the new section “Generalization and limitation”.

Reviewer #3 (Recommendations for the authors):  

Major Comments:  

(1) If we agree on one thing, it is that motor processes contribute to response time. Line 364: "In the case of decision-making, these discrete neural events are visual encoding, attention-orientation, and decision commitment, and their latency make up the reaction time." Does the third event, "decision commitment", capture both central parietal positivity (decision deliberation) and motor components? If so, how can the authors attribute the effects to decision deliberation as opposed to motor preparation?  

Thanks to the suggestions also in the public part. This main problem is now addressed as we do capture both a motor component and a decision commitment.

Line 351 suggests that the third event may contain two components.  

This was indeed our initial, badly written, hypothesis. Nevertheless the new solution again addresses this problem.

The time series in Figure 6 shows an additional peak that is not evident in the simulated ramp of Appendix 1.  

This was probably due to the overlap of both the CPP and the LRP. It is now much clearer that the CPP looks mostly like a ramp while the LRP looks much more like a burst-like/peaked activity. We make this clear in the “Decision event” paragraph of the discussion section:

“Regarding the build-up of this component, the CPP is seen as originating from single-trial ramping EEG activities but other work (Latimer et al., 2015; Zoltowski et al., 2019) have found support for a discrete event at the trial-level. The ERPs on the trial-by-trial centered event in Figure 5 show support for both accounts. As outlined above, the LRP is indeed a short burst-like activity but the build-up of the CPP between high vs low contrast diverges much earlier than its peak”.

Previous analyses (Weindel et al., 2024) found motor-related activity from central parietal topographies close to the response by comparing the difference in single-trial events on left- vs right-hand response trials. The authors suggest at line 315 that the use of only the right hand for responding prevented them from identifying a motor event.  

The use of only the right hand should have made the event more identifiable because the topography would be consistent across trials (rather than inverting on left vs right hand response trials).  

The reviewer is correct, in the original manuscript we didn’t test for lateralization, but the comment of the reviewer gave us the idea to explicitly test for the asymmetry (Figure 3). This test now clearly shows what would be expected for a motor event with a strong negativity over the left motor cortex.

The authors state on line 422 that the EEG data were truncated at the time of the response.  

Could this have prevented the authors from identifying a motor event that might overlap with the timing of the response?  

We thank the reviewer for this suggestion. This would have been a possibility but the problem is that adding samples after the response also adds the post-response processes (error monitoring, button release, stimulus disappearance, etc.). While increasing the samples after the response is definitely something that we need to inspect, we think that the separation we achieved in this revision doesn’t call for this supplementary analysis.

The largest effects of contrast on the third event amplitude appear around the peak as opposed to the ramp. If the peak is caused by the motor component, how does this affect the conclusions that this third event shows a decision-deliberation parietal processes as opposed to a motor process (a number of studies suggest a causal role for motor processes in decision-making e.g. Purcell et al., 2010 Psych Rev; Jun et al., 2021 Nat Neuro; Donner et al., 2009 Curr Bio).  

This result now changed and it does look like the peak capturing most of the effect is no longer true. We do however think that there might be some link to theories of motor-related accumulation. We therefore added this to the discussion in the Motor execution section:

“Based on all these observations, it is therefore very likely that this LRP event signs the first passage of a two-step decision process as suggested by recent decision-making models (Servant et al., 2021; Verdonck et al., 2021; Balsdon et al., 2023)”.

I would suggest further investigation into the motor component (perhaps by extending the time window of analysed EEG to a few hundred ms after the response) and at least some discussion of the potential contribution of motor processes, in relation to the previous literature.  

We believe that the absence of a motor component is sufficiently addressed in the revised manuscript and in the responses to the other comments.    

(2) What do we learn from this work? Readers would appreciate more attention to previous findings and a clearer outline of how this work differs. Two points stand out, outlined below. I believe the authors can address these potential complaints in the introduction and discussion, and perhaps provide some clarification in the presentation of the results.  

In the introduction, the authors state that "... to date, no study has been able to provide single-trial evidence of multiple EEG components involved in decision-making..." (line 64). Many readers would disagree with this. For example, Philiastides, Ratcliff, & Sadja (2006) use a single-trial analysis to unravel early and late EEG components relating to decision difficulty and accuracy (across different perceptual decisions), which could be related to the components in the current work. Other, network-based single-trial EEG analyses (e.g., Si et al., 2020, NeuroImage, Sturn et al., 2016 J Neurosci Methods) could also be related to the current component approach. Yet other approaches have used inverse encoding models to examine EEG components related to separable decision processes within trials (e.g., Salvador et al., 2022, Nat Comms). The results of the current work are consistent with this previous work - the two components from Philiastides et al., 2006 can be mapped onto the components in the current work, and Salvador et al., 2022 also uncover stimulus- and decision-deliberation related components.  

We completely agree with the reviewer that the link to previous work was insufficient. We now include all references that the reviewer points out both in the introduction (see response R3.2) and in the discussion (see response R3.4). We wish to thank the reviewer for bringing these papers to our attention as they are important for the manuscript.

The authors relate their components to ERPs. This prompts the question of whether we would get the same results with ERP analyses (and, on the whole, the results of the current work are consistent with conclusions based on ERP analyses, with the exception of the missing motor component). It's nice that this analysis is single-trial, but many of the follow-up analyses are based on grouping by condition anyway. Even the single-trial analysis presented in Figure 4 could be obtained by median splits (given the hypotheses propose opposite directions of effects, except for the linear model). 

We do not agree with the reviewer in the sense that classical ERP analyses would require much more data-points. The performance of the method is here to use the information shared across all contrast levels to be able to model the processing time of a single contrast level (6 trials per participant). Furthermore, as stated in the response to R1.4 and R1.5, the aim of the paper is to have the time of information processing components which cannot be achieved with classical ERPs without strong, and likely false, assumptions.

Medium Comments:  

(1) The presentation of Piéron's law for the behavioural analysis is confusing. First, both laws should be clearly defined for readers who may be unfamiliar with this work. I found the proposal that Piéron's law predicts decreasing RT for increasing pedestal contrast in a contrast discrimination paradigm task surprising, especially given the last author's previous work. For example, Donkin and van Maanen (2014) write "However, the commonality ofPiéron's Law across so many paradigms has lead researchers (e.g., Stafford & Gurney, 2004; Van Maanen et al., 2012) to propose that Piéron's Law is unrelated to stimulus scaling, but is a result of the architecture of the response selection (or decision making) process." The pedestal contrast is unrelated to the difficulty of the contrast discrimination task (except for the consideration of Fechner's law). Instead, Piéron's law would apply to the subjective difference in contrast in this task, as opposed to the pedestal contrast. The EEG results are consistent with these intuitions about Piéron's law (or more generally, that contrast is accumulated over time, so a later EEG component for lower pedestal contrast makes sense): pedestal contrast should lead to faster detection, but not necessarily faster discrimination. Perhaps, given the complexity of the manuscript as a whole, the predictions for the behavioural results could be simplified?  

We agree that the initial version was confusing. We now clarified the presentation of Piéron's law at the end of the introduction (see also response to R2).

Once Fechner's law is applied, decision difficulty increases with increasing contrast, so Piéron's law on the decision-relevant intensity (perceived difference in contrast) would also predict increasing RT with increasing pedestal contrast. It is unlikely that the data are of sufficient resolution to distinguish a log function from a power of a log function, but perhaps the claim on line 189 could be weakened (the EEG results demonstrate Piéron's law for detection, but do not provide evidence against Piéron's law in discrimination decisions).  

This is an excellent observation, thank you for bringing it to our attention. Indeed, the data support the notion that Pieron’s law is related to detection, but do not rule out that it is also related to decision or discrimination. In earlier work, we (Donkin & Van Maanen, 2014) addressed this question as well, and reached a similar conclusion. After fitting evidence accumulation models to data, we found no linear relationship between drift rates and stimulus difficulty, as would have been the case if Pieron's law could be fully explained by the decision process (as -indirectly- argued by Stafford & Gurney, 2004; Van Maanen et al., 2012). The fact that we observed evidence for a non-linear relationship between drift rates and stimulus difficulty led us to the same conclusion, that Pieron’s law could be reflected in both discrimination and decision processes. We added the following comment to the discussion about the functional locus of Pieron's law to clarify this point:

“Note that this observation is not necessarily incongruent with theoretical work that argued that Piéron’s law could also be a result of a response selection mechanism (Stafford and Gurney, 2004; Van Maanen et al., 2012; Palmer et al., 2005). It could be that differences in stimulus intensity between the two options also contribute to a Piéron like relationship in the later intervals, that is convoluted with Fechner’s law (see Donkin and Van Maanen, 2014, for a similar argument). Unfortunately, our data do not allow us to discriminate between a pure logarithmic growth function and one that is mediated by a decreasing power function”.

(2) Appendix 1 shows that the event detection of the HMP method will also pick up on ramping activity. The description of the problem in the introduction is that event-like activity could look like ramping when averaged across trials. To address this problem, the authors should simulate events (with some reasonable dispersion in timing such that they look like ramping when averaged) and show that the HMP method would not pull out something that looked like ramping. In other words, the evidence for ramping in this work is not affected by the previously identified confounds.  

We agree that this demonstration was necessary and thus added the suggested simulation to Appendix 1. As can be seen in the Figure 1 of the appendix, when we simulate a half-sine the average ERP based on the timing of the event looks like a half-sine.

(3) Some readers may be interested in a fuller discussion of the failure of the Fechner diffusion model in the speed condition.  

We are unsure which failure the reviewer refers to but assumed it was in relation to the behavioral results and thus added: 

It is unlikely that neither Piéron nor Fechner law impact the RT in the speed condition. Instead this result is likely due to the composite nature of the RT where both laws co-exist in the RT but cancel each other out due to their opposite prediction.

Minor Comments:  

(1) "By-trial" is used throughout. Normally, it is "trial-by-trial" or "single-trial" or "trial-wise".

We replaced all occurrences of “by-trial”  with the three terms suggested were appropriate.

(2) Line 22: "The sum of the times required for the completion of each of these precessing steps is the reaction time (RT)." The total time required. Processing.  

Corrected for both.

(3) Line 26/27: "Despite being an almost two century old problem (von Helmholtz, 2021)." Perhaps the citation with the original year would make this point clearer.  

We agree and replaced the citation.

(4) Line 73: "accounted by estimating". Accounted for by estimating.  

Corrected.

(5) Line 77 "provides an estimation on the." Of the.  

Corrected.

(6) Line 86: "The task of the participants was to answer which of two sinusoidal gratings." The picture looks like Gabor's? Is there a 2d Gaussian filter on top of the grating? Clarify in the methods, too.  

We incorrectly described the stimuli as those were indeed just Gabor’s. This is now corrected both in the main text and the method section.

(7) Figure 1 legend: "The Fechner diffusion law" Fechner's law or your Fechner diffusion model?  

Law was incorrect so we changed to model as suggested.

(8) Line 115: "further allows to connects the..." Allows connecting the.  

Corrected.

(9) Line 123: "lower than 100 ms or higher than..." Faster/slower.  

Corrected.

(10) Line 131: "To test what law." Which law.?  

Corrected to model.

(11) Figure 2 legend: "Left: Mean RT (dot) and average fit (line) over trials and participants for each contrast level used." The fit is over trials and participants? Each dot is? Average trials for each contrast level in each participant?  

This sentence was corrected to “Mean RT (dot) for each contrast level and averaged predictions of the individual fits (line) with Accuracy (Top) and Speed (Bottom) instructions.”.

(12) Line 231: "A comprehensive analysis of contrast effect on". The effect of contrast on.  

This title was changed to “functional interpretation of the events”.

(13) Line 23: "the three HMP event with". Three HMP events.

The sentence no longer exists in the revised manuscript.

(14) Line 270: "Secondly, we computed the Pearson correlation coefficient between the contrast averaged proportion of correct." Pearson is for continuous variables. Proportion correct is not continuous. Use Spearman, Kendall, or compute d'.  

The reviewer rightly pointed out our error, we corrected this by computing Spearman correlation.

(15)  Line 377: "trial 𝑛 + 1 was randomly sampled from a uniform distribution between 0.5 and 1.25 seconds." It's just confusing why post-response activity in Figure 5 does look so consistent. Throughout methods: "model was fitted" should be "was fit", and line 448, "were split".  

We do not have a specific hypothesis of why the post-response activity in the previous Figure 5 was so consistent. Maybe the Gaussian window (same as in other manuscripts with a similar figure, e.g. O’Connell et al. 2012) generated this consistency. We also corrected the errors mentioned in the methods.

(16) The linear mixed models paragraph is a bit confusing. Can it clearly state which data/ table is being referred to and then explain the model? "The general linear mixed model on proportion of correct responses was performed using a logit link. The linear mixed models were performed on the raw milliseconds scale for the interval durations and on the standardized values for the electrode match." We go directly from proportion correct to raw milliseconds...  

The confusion was indeed due to the initial inclusion of a general linear mixed model on proportion correct which was removed as it was not very informative. The new revision should be clearer on the linear mixed models (see first sentence of subsection ‘linear mixed models' in the method section).

(17) A fuller description of the HMP model would be appreciated.  

We agree that this was necessary and added the description of the HMP model in the corresponding method section “Hidden multivariate pattern” in addition to a more comprehensive presentation of HMP in the first paragraph of the Result and Discussion sections.

(18) Line 458: "Fechner's law (Fechner, 1860) states that the perceived difference (𝑝) between the two patches follows the logarithm of the difference in physical intensity between..." ratio of physical intensity.  

Corrected.

(19) P is defined in equations 2 and 4. I would include the beta in equation 4, like in equation 2, then remove the beta from equations 3 and 5 (makes it more readable). I would also just include the delta in equation 2, state that in this case, c1 = c+delta/2 or whatever.  

This indeed makes the equation more readable so we applied the suggestions for equations 2, 3, 4 and 5. The delta was not added in equation 2 but instead in the text that follows:

“Where 𝐶1 = 𝐶0 + 𝛿, again with a modality and individual specific adjustment slope (𝛽).” 

(20) The appendix suggests comparing the amplitudes with those in Figure 3, but the colour bar legend is missing, so the reader can only assume the same scale is used?  

We added the color bar as it was indeed missing. Note though that the previous version displayed the estimation for the simulated data while this plot in the revised manuscript shows the solution on real data obtained after downsampling the data (and therefore look for a larger pattern as in the main text). We believe that this representation is more useful given that the solution for the downsampled data is no longer the same as the one in the main text (due to the difference in pattern width).

The checklist feels too extensive for what we would ask a student to do. Suggest pairing it back alot or sticking with the 3 S style of short evaluation. Just Part B comparison for example.

My thoughts on the article,

The difference in appearance and body type between that of toys of Ariana Grande and the literal Aria Grande is notable.

A part of me asks why Aria Grande wants to be so skinny. Like she's not even attractive anymore. It may have something to do with Women who are attractive interpreting their beauty as a curse because is attracts attention they would no longer have.

I have never heard of "Stan Culture" before. It's basically fan accounts for people and fandoms. I wonder if the Aria Grande "Stan Accounts" interface with the memes I posted in this other part of the article

https://hyp.is/OY-gdNWMEfCXyP8m6RghMA/spitfirenews.com/p/ariana-grande-eating-disorder-wicked-cynthia-erivo

I believe that "Skinny" as a beauty standard is just a leftover effect of Gay Men of Power in the Designer Fashion industry choosing women that look like the men they want to fuck

I never thought about how Ariana Grande looks so skinny that she is sick in real life yet there are toys and lego of here. I wonder how they contrast

Twitter was pretty big on that issue last week,

https://x.com/hippojuicefilm/status/1996215993094270997

https://x.com/tragicworldapp/status/1998555521343959464

https://x.com/MmisterNobody/status/1998469984687714523

Transparent Peer Review

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

I feel like the title discussing transparency doesn't completely match the content. Should it be discussing and acknowledging.

I think this is a good opportunity to have an activity that relates to bias in written articles and the need to check facts. https://ecu.au.libguides.com/generative-ai/limitations

This correlates with Sarah Ahmed' feminist killjoy, the homework to unlearn everything, including the sciences; this is the works of becoming a feminist (Ahmed 2016).

I came back to this after reading the disability and paragraph and the authors' prospective at the end. I realize that scientists like this are ignorant on understanding plants. He clearly lacked the initiative to study plants and instead plastered his perception of plants based on societal expectations.

I notice a common pattern from sex to disability, where plants are always humanized when being studied.

I was always very confused about this as well. How do plants and animals without mammal genital get involved in "sex"? Why is their reproduction always sexualized?

This reminds me of the invasive species of the European Pine trees placed in Palestine during the Nakba. The Israeli forces planted pine trees while occupying Palestinian villages to replicate the European infrastructures, this is a biological warfare of colonialism (Josephson 2025). https://origins.osu.edu/read/environmental-nakba-israel-palestine-water

With the rise of anti-South Asian sentiment, South Asian countries are posted on social media with negative criticism of the polluted rivers, and littered streets; media consumers use this as an excuse to dehumanize South Asian people. But South Asian countries are lack the studies and sources for environmental work, as well as the politics involved, lobbyed by the BJP right wing Indian party and the Trump administration.

The author explains that botany has only been properly studied at Europe and North American institutions, compared to the rest of the world such as Africa and Asia where specimens are barely discovered.

Banu's introduction is very metaphorical to the term Labyrinth, making her book very enaging for readers, especially non-stem students like myself.

Science is often very exclusive to stem students, or its western epistemologies as Harding explains is only comprehensible from a anglo saxon male perspective. Banu encurages botany to be accessible to everyone, making learning and findings unlimted, and creating that change of feminist science.

Banu reassures to her readers that botany and this book is for everyone to read, which is why she is inclusive with these scientific terms.

This challenges Linnaeus notion of the male and female genitalia of plants, and his concept and time of the sex reproduction of plants correlating with the nuclear family tradition. If scientists like Linnaeus could not comprehend the difference of time with the growth of plants, they would be labeled as weird, or the way human beings are labelled, "queer", a term now reclaimed by the queer community.

I appreciate how Banu correlates the growth and sciences of a plant to queer theory. She pulls apart the definition of queer as not only a homosexual term but something does not align the strict labels and frameworks that human beings apply. She explains in a way that nature and plants are queer in itself if humans wanted to label it.

The author explains decolonization to not be a simple process, and they previously mention how colonization was a huge project. For that reason, we need to approach colonization with variaety of other resources also affected, creating a bigger alternative project.

This is the author's thesis to challenging the colonial science and overturning it with feminist science and botany.

This reminds me of a conversation I had in my sociology class about hate crime and racial discrimination in the healthcare systems. How the demographics of nurses are BIPOC and women, and have faced tremendous racism and sexism at workplace during the COVID-19 pandemic.

How the colonial mindset truly affected the kinship and families in South Asia, leaving countries like India truly displaced and underesourced. How this broke families apart by not believing in disabilities and making children feel less valuable in a competitized society, catching up to the first world countries.

These are the multitude of genres the author speaks on, to taking epistemology at a radical stance.

The author encourages to expand epistemology across variety of metholodologies to find more findings, implications and overall increase the studies of the botany field.

The author challenges Linnaeus concept of labeling human beings and living things from a negative perception, the way that hethinks. Lorde is used here to explain that studying plants can involve celebrating their differences.

The author answers their question so beautifully at the very end of this paragraph. Instead decentering the human out of plants, the author visualized a space where humans, living things and plants exist without a hierarchy, an eco-cosmipolitanism, the idea that all humans, animals, and living things are members of a single community.

In what way nature is considered feminine as mother nature? What are "the maternal instincts" of nature that are constructed by the patriarchy to call something mother nature?

This is the kind of feminist epistemology that Hardings encourages in her reading about the feminist research method.

I find it quite pathetic how easily people sexualize objects and living things and I cannot understand how that works, but I see the influence of scientists like Linnaeus encouraging this type of objectification in scientific studies. This reminds me of Paasnonen's concept of objectification, where people and things simply exist to be objectified, and that is due to the cultural dynamics and social constructions of a society.

The author beings with a strong evidence of the sexism uprooted in the plant biology of the classification of species. This correlates with Mulvey's concept of phallocentrism, where the attraction of a woman is centred by the male genital. In the sense of this reading. Linnaeus has labeled plants based on human anatomy, aligned with social contructions of rigid gender roles.

The author's main point of this chapter and this book is to highlight the colonial epistemology and influence on plant biology and how it lacks accuracy on the study of plants. The author recommends different epistemologies, especially the field of botany and how it is beneficial for the study of plants, also encouraging social justice. (p.1-2).

This is your interpretation. It is up to you to support it with evidence provided to you by the author. This is your time to take a stance on your argument.

make sure to place your thesis at the bottom of your first paragraph as well as restate your thesis near the end. Me personally this helps me keep track of the argument as I am writing my thesis.

A thesis statement is an argument that YOU must support with evidence was well as an explanation for how this evidence is relevant to your argument. This is a claim made by you with the help of an author. Make sure that your argument is valid as well as relevant to the prompt of your essay.

It is cool how they either translated or distributed the books in Greek. Also, for the ones who could not afford those books they had less expensive options for them.

This is also very interesting. Even at this time, they had could make ships that were over 400 feet long. It is also intriguing Zheng left with over 200 small ships and 28,000 soldiers. This is a lot!

This is interesting that Timur stripped the city of all jewels, gold, masons, etc. I feel like this is a huge act to do.

It's more likely students will ask AI a question on a research topic and within the answer false references will emerge. Student then need to verify them. It's also unlikely students would look at a list of references and then not recognise a few. Likely they would not recognise any if they haven't done any searching of journals. All of this practice is dishonest, not just potentially. Add link to Library evaluating outputs page. https://ecu.au.libguides.com/generative-ai/critical-assessment

The Last Samurai: A Conversation with Takeshi Kitano

url:: https://mubi.com/es/notebook/posts/the-last-samurai-a-conversation-with-takeshi-kitano

date:: 2024-01-11

accessed:: 2025-12-09

The revered director describes his latest film, "Kubi": a samurai film by turns gruesome and ravishing set in 16th-century Japan. Leonardo Goi

This seems like a poor caricature of how the brain works...

I suspect/hope you did these ancestral state reconstructions on time-calibrated phylogenies? If so, I would just clarify this here given these methods are only described in the supplement, and this methodological detail will have strong impacts on their outcome, as it does not make sense to conduct ASR on non-time-calibrated trees.

the reflection writing process lets you get better understanding of your strengths in writing and weaknesses.

reflection lets you look at your writing and better get a full picture of what you wrote.

Your enhancer-insertion library is a useful tool for probing carotenoid regulation, and the CMRP1 follow-up is both convincing and compelling. The long-range activity of the enhancer in your top hit is intriguing and raises a few questions about how ERGO should be interpreted.

If an insertion can influence genes across ~2 Mb, then many nearby loci are plausible targets. How confident are you that CMRP1 is the primary driver rather than one member of a broader set of co-activated genes? More generally, because NHEJ insertions favor open and insertion-tolerant regions, regulators positioned in less permissive chromatin may never be sampled. Insertions that disrupt essential genes, or essential neighboring genes, would also eliminate the corresponding clones before screening, also impacting sampling. Do you have a sense of how much of the genome is protected in this way? Along those lines, have you looked at whether enhancer effectiveness varies with chromatin context, and whether some genomic regions tend to dampen or block enhancer activity?

Did you characterize the expression of neighboring genes at all to distinguish between CMRP1-driven changes and insertion-related ones? Given that many insertions are tandem or structurally complex, did you assess whether enhancer copy number, truncation, or orientation contributed to the expression patterns or phenotypes you observed?

Finally, the use of ERGO here implements a pigment phenotype in the yellow-in-the-dark background. Do you envision pairing the enhancer library with non-colorimetric reporters or selectable screens to expand beyond carotenoid metabolism in the future?

Jolla Mind2 runs venho.ai. Unclear if this product actually ships at the moment. 'backorder' which means 8 week delay. Originally Oct 25, so end of year

venho.ai, Finnish, only to be available in EU/EFTA desk top based AI. There's a 600 Euro Jolla device that runs it that can be ordered. Comes with a subscription it seems, and has cloud connection, but it seems not for the AI stuff / data.

This passage explains that when we watch certain media styles over and over, we start using them to understand real life. If a news show always shows loud, heated arguments, viewers may think that’s the “normal” way to handle disagreements. Media formats can quietly shape how people act and communicate.

This says that when people with minority opinions stay quiet, it starts to look like everyone agrees with the majority. That makes even more people stay silent, creating a cycle where only one viewpoint is visible. The example of Germany during World War II shows how fear and pressure can stop people from speaking up, even when many disagree.

This violates standard English conventions in the narration rather than in dialogue, which is unusual for the book.

"understood" => I like this framing of centering that a purpose of all of this is to make sure that the learning can be understood (and thereby valued) by the audiences who need to understand it. In order for credentials to have currency, the brokers of currency need to know what the assets are and how to value them.

Great quote on the potential with industry partnerships in particular

A huge value proposition and promise is not yet realized bc the key beneficiary doesn't (yet) get it

Even before they were actually portable, a key benefit was rooted in portability and verifiability (and, to what purpose: "flexible and accessible way to recognize and communicate skills"). That is the WHY. And that ethos has been lost (or never acquired) by too many. There could be opportunity to refocus and zoom way in on the why and the key benefits.

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

Learn more at Review Commons

Reply to the reviewers

We are grateful to the reviewers for their thoughtful and constructive evaluations of our manuscript. Their comments helped us clarify key aspects of the study and strengthen both the presentation and interpretation of our findings. The central goal of this work is to dissect how the opposing activities of GATA4 and CTCF coordinate chromatin topology and transcriptional timing during human cardiomyogenesis. The reviewers’ feedback has allowed us to refine this message and better contextualize our results within the broader framework of chromatin regulation and cardiac development.

In response to the reviews, in our preliminary revision we have already implemented substantial improvements to the manuscript, including additional analyses, clearer data visualization, and revisions to the text to avoid overinterpretation. These refinements enhance the robustness of our conclusions without altering the overall scope of the study. A small number of additional analyses and experiments are ongoing and will be added to the full revision, as detailed below.

We believe that the revised manuscript, together with the planned updates, fully addresses the reviewers’ concerns and substantially strengthens the contribution of this work to the field.

Reviewer 1 – Point 1:

In the datasets you are examining, what are the relative percentages in each of the four groups relating compartmentalization change to expression change (A→B, expression up; A→B, down; B→A, up; B→A, down)?

We quantified compartment–expression relationships using Hi-C and bulk RNA-seq from H9 ESCs and CMs. The percentages for each category are shown below and incorporated into updated Figure S2H.

Group

Downregulated in CM

Upregulated in CM

A-to-A

11.92%

8.44%

A-to-B

18.20%

2.79%

B-to-A

7.96%

18.07%

B-to-B

14.36%

6.44%

A chi-squared test comparing observed vs. expected distributions (based on gene density across bins) confirmed a strong association between compartment dynamics and transcriptional behavior. B-to-A genes are significantly enriched among genes upregulated in CMs, while A-to-B genes are enriched among those downregulated (updated Figure S2H).

We next assessed with GSEA how these gene classes respond to GATA4 and CTCF knockdown. In 2D CMs, GATA4 knockdown reduces expression of CM-upregulated B-to-A genes and increases expression of CM-downregulated A-to-B genes, whereas CTCF knockdown produces the opposite pattern (updated Figure 2F).

Applying the same analysis to cardioid bulk RNA-seq (updated Figure 4E) revealed the strongest effects in SHF-RV organoids, consistent with monolayer data. In SHF-A organoids, only GATA4 knockdown had a measurable impact on CM-upregulated B-to-A and CM-downregulated A-to-B genes. Because the subsets of CM-downregulated B-to-A and CM-upregulated A-to-B genes were very small and showed no consistent trends, Figure 4 focuses on the two informative categories only. The full classification is provided in Reviewer Figure 1 below.

(The figure cannot be rendered in this text-only format)

Reviewer Figure 1. GSEA for CM-upregulated B-to-A and CM-downregulated A-to-B genes. p-values by Adaptive Monte-Carlo Permutation test.

Reviewer 1 – Point 2

This phrase in the abstract is imprecise: ‘whereas premature CTCF depletion accelerates yet confounds cardiomyocyte maturation.’

---

The abstract has been revised to: “whereas premature CTCF depletion accelerates yet alters cardiomyocyte maturation.” (lines 29-30).

Reviewer 1 – Point 3

Regarding this statement: "Disruption of [3D chromatin architecture] has been linked to genetic dilated cardiomyopathy (DCM) caused by lamin A/C mutations8,9, and mutations in chromatin regulators are strongly enriched in de novo congenital heart defects (CHD)10, underscoring their pathogenic relevance11." The first studies to implicate chromatin structural changes in heart disease, including the role of CTCF in that process, were PMID: 28802249, a model of acquired, rather than genetic, disease.

We added the following sentence to the paragraph introducing CTCF: “Moreover, depletion of CTCF in the adult cardiomyocytes leads to heart failure28,29.” (line 72)

Reviewer 1 – Point 4

Can you quantify this statement: ‘the compartment switch coincided with progressive reduction of promoter–gene body interactions’?

We quantified promoter–gene body contacts by calculating the area under the curve (AUC) of the virtual 4C signal derived from H9 Hi-C data across differentiation. As a result of this analysis we added the following sentence: “Quantitatively, interactions between the TTN promoter and its gene body decreased by ~55% from the pluripotent stage to day 80 cardiomyocytes.” (lines 89-91).

Reviewer 1 – Point 5

Regarding this statement: "six regions became less accessible in CMs, correlating with ChIP-seq signal for the ubiquitous architectural protein CTCF." I don't see 6 ATAC peaks in either TTN trace in Figure 1A.

We corrected the text as it follows: “TTN experienced clear changes in chromatin accessibility during CM differentiation: ATAC-seq identified two CM-specific peaks that correlated with ChIP-seq signal for the cardiac pioneer TF GATA4 at the two promoters, one driving full length titin and the other the shorter cronos isoform. In contrast, two regions became less accessible in CMs, correlating with two of the six ChIP-seq peaks for the ubiquitous architectural protein CTCF” (lines 93-97). We attribute the differences between ChIP-seq and ATAC-seq profiles to methodological sensitivity and/or biological variability between datasets generated in different laboratories and cell batches.

Reviewer 1 – Point 6

Western blots need molecular weight markers.

We edited the relevant panels accordingly (updated Figures 1E and 2B).

Reviewer 1 – Point 7

Regarding this statement: "The decrease in CTCF protein levels may explain its selective detachment from TTN during cardiomyogenesis." At face value, these findings suggest the opposite: i.e. that a massive downregulation of CTCF at protein level should affect its binding across the genome, which is not tested and is hard to evaluate between ChIP-seq studies from different groups and from different developmental timeframes.

We revised the text to avoid implying selective detachment and performed a genome-wide analysis of CTCF occupancy using ENCODE ChIP-seq datasets generated by the same laboratory with matched protocols in hESCs and hESC-derived CMs. This analysis shows that 43.2% of CTCF sites present in ESCs are lost in CMs, whereas only 5.7% are gained, confirming a broad reduction in CTCF binding during differentiation. These results are now included in__ updated Figure 1B__.

Reviewer 1 – Point 8a

A couple thoughts on the FISH experiments in Figure 2. A claim of 'impaired B-A transition' would be more convincing if you show, by FISH, that the relative distance of TTN from lamin B increases with differentiation.

Although prior work from us and others has established that TTN transitions from the nuclear periphery in hESCs to a more internal position during cardiomyogenesis (Poleshko et al. 2017; Bertero et al. 2019a), we are reproducing this trajectory in WTC11 hiPSCs as part of the FISH experiments for the full revision.

__Reviewer 1 – Point 8b __

In the [FISH] images: are you showing a total projection of all z planes? One assumes the quantitation is relative to a 3D reconstruction in which the lamin B signal is restricted to the periphery. Have you shown this? __

Quantification was performed on full 3D reconstructions from Z-stacks, as detailed in the Methods (lines 721-727). While the original submission displayed maximum-intensity projections, updated Figure 2D and Figure S2E now show representative single optical sections, which more clearly highlight the spatial relationship between the TTN locus and the nuclear lamina.

Reviewer 1 – Point 8c

Lastly, these data are very interesting and important, provoking reexamination of your interpretation of the results in Figure 1. Figure 1 was interpreted to show that less CTCF binding led to decreased lamina (and thus B compartment) association during development. Figure 2 shows that depleting CTCF does not change association of TTN with lamina.

Our interpretation is that by day 25 of hiPSC-CM differentiation the TTN locus may have reached its maximal radial repositioning even in control cells, limiting the ability to detect earlier effects of CTCF depletion. To test whether CTCF knockdown accelerates lamina detachment at earlier stages, we are repeating the FISH analysis for the inducible CTCF knockdown line at multiple time points during differentiation.

Reviewer 1 – Point 9

A thought about this statement: "Altogether, these results suggest that GATA4 and CTCF function as positive and negative regulators of B-to-A compartment switching, likely acting through global and local chromatin remodeling, respectively." GATA4 induces TTN expression and its knockdown prevents TTN expression-the evidence that GATA4 affects compartmentalization is unclear. By activating the gene, GATA4 may shift TTN to B classification.

Our current data do not allow us to disentangle whether GATA4-driven transcriptional activation precedes or follows the B-to-A compartment shift. We have therefore removed the mechanistic speculation from this sentence to avoid overinterpretation. Nevertheless, the analyses in updated Figure 2F, discussed in the response to Reviewer 1 - Point 1, show that GATA4 knockdown preferentially reduces expression of CM-upregulated B-to-A genes, while CTCF knockdown has the opposite effect, supporting the conclusion that both factors influence the transcriptional programs associated with B-to-A transitions.

Reviewer 1 – Point 10

__I'm not sure what I am looking at in Figure 3C. Are those traces integration of interactions over a defined window? "Each [mutant is] clearly different from WT" is not obvious from the presentation. The histograms are plotting AUC of what? Interactions of those peaks with the mutated region? I genuinely appreciate how laborious this experiment must have been and encourage you to explain better what you are showing. __

We revised the main text to avoid overstating the differences (“clearly” “in a similar manner”, line 192) and expanded the l__egends of updated Figures 3C–D__ to clarify what is being shown: “(C) 4C-seq in hiPSCs using the promoter-proximal region of TTN as viewpoint. The top panel shows raw interaction profiles. The lower panels plot pairwise differences between conditions to reveal subtle changes. A schematic indicating the 4C viewpoint is included for clarity. Right inset: zoom of the CBS4–5 region. Mean of n = 3 cultures. (D) AUC of the differential 4C-seq signal for defined intervals (panel C). p-values by one-sample t-test against μ = 0.”. We also added a visual cue in updated Figure 3C indicating the 4C viewpoint to facilitate interpretation.

Reviewer 1 – Point 11

Again acknowledging how challenging these experiments are: when you mutant a locus, you change CTCF binding but you also change the DNA. Thus, attributing the changes in interactions to presence/absence of CTCF binding is difficult, because the DNA substrate itself has changed. Perhaps you are presenting all of this as a negative result, given the modest effect on transcription, which is as important as a positive result, given the assumptions usually made about such things. But the results are not clearly described and your interpretation seems to go between implying the structural change causative and being agnostic.

We recognize that deleting a genomic region can affect both CTCF binding and the DNA substrate itself. For this reason, we implemented two parallel genome-editing strategies:

(1) a straightforward Cas9-mediated deletion of ~100 bp centered on each CBS, and

(2) a more precise HDR approach replacing only the 20 bp core CTCF motif.

Because the HDR strategy succeeded, all downstream analyses were carried out on these minimal edits, which substantially limit disruption of other transcription factor motifs and reduce the likelihood of sequence-dependent polymer effects unrelated to CTCF.

Nevertheless, to avoid implying unwarranted causality in the absence of more conclusive evidence, we added a paragraph to the Discussion outlining these limitations, including the sentence: “Our study also reflects general challenges in separating chromatin-architectural and transcriptional mechanisms. Although the CBS edits were restricted to the core CTCF motifs, additional sequence-dependent effects cannot be fully excluded, and we therefore interpret the resulting changes as consistent with—but not exclusively due to—loss of CTCF binding.” (lines 365-368)

Reviewer 1 - Point 12.

Figure 4C: since you have RNA-seq data, a much more objective way to present these data would be to show all data (again, A-B, up; A-B, down; B-A, up; B-A, down) and the effects of CTCF or GATA4. Regardless, you can still focus on the cardiac specific genes. But my guess is if you examine all genes, the pattern you show in panel C will not be present in the majority of cases. Furthermore, if this hypothesis is wrong, such an analysis will allow you to identify other genes affected by the mechanisms you describe and your analysis will test whether these mechanisms are in fact conserved at different loci.

As outlined in our response to Point 1, we extended the analysis to all genes undergoing compartment changes and incorporated this into the cardioid RNA-seq dataset. This revealed a clear and consistent relationship between GATA4 or CTCF knockdown and the expression of B-to-A and A-to-B gene classes (updated Figure 4E).

Reviewer 2 - Point 1.1

1. CTCF regulation at TTN locus:

(1) Figure 1A: The claim of the authors about convergent CTCF sites and transcriptional activation of TTN is quite simplistic. This claim is only valid when we know where cohesin is loaded. If cohesin is loaded at then intragenic GATA4 binding site, then the only important CTCF sites is at the promoter of TTN. I suggest that the authors read few more publications which may help the authors to better understand how cohesin and CTCF team up to regulate transcription, such as Hsieh et al., Nature Genetics, 2022; Liu et al., Nature Genetics, 2021; Rinzema et al., Nature Structural and Molecular Biology, 2022.

__Suggestion: The authors should add cohesin (RAD21/SMC1A) and NIPBL ChIP-seq for better interpretation. __

In line with the reviewer’s insightful suggestion, we integrated cohesin ChIP-seq data into updated Figure 1A. Specifically, we added a RAD21 ChIP-seq track from hESCs, which provides direct evidence of cohesin occupancy across the TTN locus. RAD21 binding closely parallels CTCF binding at five sites within the gene body, supporting a model in which promoter-proximal CTCF anchors cohesin to stabilize repressive loops at this locus. This analysis substantially strengthens the mechanistic framework and is consistent with the studies recommended by the reviewer, which we have now cited (lines 68 and 104).

Reviewer 2 - Point 1.2. (2) Figure 3B: If delta2CBS only has heterozygenous deletion of CBS6, why we would expect the binding will be weaken to 50%. However, the CTCF binding is reduced to around 1/10 in the ChIP-qPCR. How do the authors explain this?

Sequencing of the Δ2CBS line shows that one CBS6 allele carries the intended EcoRI replacement, while the second allele contains a 2-bp deletion within the core CTCF motif (Figure S3C). Remarkably, this small deletion is sufficient to abolish CTCF binding, resulting in complete loss of occupancy at CBS6 despite heterozygosity. We clarified this in the text as follows: “CTCF ChIP-qPCR in hiPSCs confirmed complete loss of CTCF binding at the targeted sites, including CBS6 in the Δ2CBS line, indicating that the 2-bp deletion sufficed to disrupt CTCF binding while occupancy at other CBSs remained unaffected.” (lines 187–189).

Reviewer 2 - Point 1.3a (3) Figure 3C: There are two problems with the 4C experiments: (a) The changes are really mild. In fact, none of the p-values in Figure 3D are significant.

The effect of deleting CBS1 is indeed modest, consistent with reports that individual CTCF binding sites often show functional redundancy (i.e., Rodríguez-Carballo et al. 2017; Barutcu et al. 2018; Kang et al. 2021). Nevertheless, our 4C-seq experiments have reproducibly shown the same directional trend across biological replicates. To increase statistical power and more rigorously assess the robustness of this effect, we are generating additional 4C replicates as part of the full revision.

Reviewer 2 - Point 1.3b [In the 4C experiments] (b) The authors should also consider a model that CTCF directly serves as a repressor. In this way, 3D genome may not be involved. B-A switch is simply caused by the activation of the locus.

We now explicitly acknowledge this possibility in the Discussion. The revised text states: “Moreover, our data cannot unambiguously separate CTCF’s architectural role from potential direct repressive activity. Both mechanisms could contribute to the observed effects, and our findings likely reflect the combined influence of CTCF on chromatin topology and gene regulation.” (lines 368–371).

Reviewer 2 - Point 2.1a 2. __(CTCF) detachment: The authors mentioned few times "detachment". In the context of this manuscript, the authors indicate detachment from nuclear lamina. However, the authors haven't provide convincing evidence about this. __

In the two instances where we used the term “detachment,” we intended it to refer exclusively to reduced CTCF binding to DNA, not to lamina repositioning. To avoid ambiguity, we have replaced “detachment” with “reduced binding” in both locations (lines 123 and 329). We do not use this term to describe TTN–lamina positioning.

Reviewer 2 - Point 2.1b (1) Figure 1D: I doubt whether such changes of CTCF protein abundance will lead to LAD detachment. Suggest the authors read van Schaik et al., Genome Biology, 2022. With the full depletion of CTCF, the effects on LADs are still very restricted.

We agree that the observed correlation between reduced CTCF levels and the relocation of TTN away from a LAD does not establish causality. As outlined in our response to Reviewer 1 – Point 8c, we are performing additional FISH experiments at earlier differentiation stages in the CTCF inducible knockdown line to directly assess whether partial CTCF depletion is sufficient to alter the timing of TTN–lamina separation.

Reviewer 2 - Point 2.2 (2) Figure 2D: Lamin B1 should be mostly at nuclear periphery. I have few questions: (1) is the antibody specific? (2) do these cells carry mutation in LMNB1 gene? (3) is the staining actually LMNA?

As also clarified in response to Reviewer 1 – Point 8b, the original images displayed maximum-intensity projections of Z-stacks, which obscured the peripheral distribution of LMNB1. We have updated Figure 2D and Figure S2E to show representative individual optical sections, which more clearly display the expected peripheral LMNB1 signal. We also confirm that the antibody used is specific for LMNB1 and previously validated (Bertero et al. 2019b), and that the WTC11-derived lines used in this study carry no mutation in LMNB1.

Reviewer 2 - Point 3

3. Opposite functions of GATA4 and CTCF: These data in Figure 5E-H argues the opposite role of GATA4 and CTCF in transcriptional regulation. Would it be that CTCF KD just affected cell proliferation, which is actually known for many cell types, rather than affect CM differentiation process? If this is the reason, inversed correlation between CTCF KD and GATA4 KD in Figure 4D could also be explained by opposite effects on cell cycle.

We directly evaluated this possibility. In FHF–LV cardioids, cell cycle profiling in Figure 6C and Figure S6C (now S7C) showed that CTCF knockdown does not alter the distribution of CMs across G1/S/G2–M phases, in contrast to the marked increase in proliferation observed with GATA4 knockdown.

Because this comment referred specifically to the SHF data, we also analyzed mitotic gene expression in the SHF–RV bulk RNA-seq dataset using GSEA. CTCF knockdown did not significantly enrich any cell cycle–related gene sets, whereas GATA4 knockdown produced a strong enrichment for mitotic cell cycle terms, in line with FHF-LV data (Reviewer Figure 2).

These results are summarized in updated Figure S5C, reporting also the results of the broader GSEA analysis, and together indicate that the transcriptional divergence between CTCF and GATA4 knockdown is not simply explained by opposing effects on proliferation.

(The figure cannot be rendered in this text-only format)

Reviewer Figure 2. GSEA for mitotic cell cycle in SHF-RV after inducible knockdown of CTCF (left) or GATA4 (right). p-values by Adaptive Monte-Carlo Permutation test.

Reviewer 2 - Point 4 4. In discussion, the authors suggested that CTCF is a local chromatin remodeller. In my view, association with local chromatin compaction doesn't qualify CTCF as a chromatin remodeler. To my knowledge, CTCF does not have an enzymatic domain, then how does it remodel chromatin?

Our intended meaning was that CTCF shapes 3D chromatin architecture through its role in organizing intergenic looping, not that it remodels chromatin enzymatically. To avoid confusion, we have removed the original sentence from the Discussion.

Reviewer 2 - Point 5. 5. Some conclusions are drawn based on insignificant p-values, e.g. Figure 2F, Figure 3D, etc. The authors should be careful about their conclusion, and tone down their statement for the observations have borderline significance.

The conclusions based on bulk RNA-seq have been revised in response to Reviewer 1 – Point 1 (updated Figure 2F). By subsetting B-to-A and A-to-B genes according to their expression dynamics, this analysis now yields clearer and statistically significant differences between conditions.

Regarding the 4C-seq data, as acknowledged in Reviewer 2 – Point 3a, the observed effects are modest. We are generating additional biological replicates to increase statistical power. In the meantime, we have adjusted the text to avoid overstating these findings. The revised manuscript now states: “While the difference did not reach significance, these trends suggest …” (lines 199–200).

Reviewer 2 - Minor comment 1. Minor comments: 1. Figure 1A: (1) I suggest to label two promoters in the gene model. It's unclear in the figure in the current version; (2) I was a bit confused with the way how the authors labeled CTCF directionality. I thought there are a lot of promoters. Why didn't they use triangles?

We updated Figure 1A to label both TTN promoters and indicate their orientation. For CTCF sites, we now clearly display the motif direction and core binding region as determined by FIMO analysis of the CTCF ChIP-seq peaks, improving consistency and interpretability.

Reviewer 2 - Minor comment 2. 2. Figure 2C: I think the drastical reduction of titin-mEGFP levels is only due to the way how the authors analyze their FACS data. Can the author quantify on median fluorescence intensity?

The gating strategy for titin-mEGFP⁺ cells was defined using a reporter-negative control, and cells lacking TNNT2 expression showed no detectable titin-mEGFP signal, confirming the specificity of the gate. To complement this analysis, we also quantified the median fluorescence intensity (MFI) of titin-mEGFP⁺ cells. The MFI analysis corroborates the original findings, showing a significant decrease in GATA4 knockdown and an increase in CTCF knockdown (updated Figure S2D).

__Reviewer 2 - Minor comment 3. 3. Figure S2G: P value should be -log10, I assume. Please label it accurately. __

We appreciate the reviewer pointing out this labeling error. In the revised manuscript, this panel has been removed to accommodate the updated compartment–expression analysis now presented in updated Figure 2H (see response to Reviewer 1 – Point 1), and the issue is no longer applicable.

References

Barutcu AR, Maass PG, Lewandowski JP, Weiner CL, Rinn JL. 2018. A TAD boundary is preserved upon deletion of the CTCF-rich Firre locus. Nat Commun 9: 1444.

Bertero A, Fields PA, Ramani V, Bonora G, Yardımcı GG, Reinecke H, Pabon L, Noble WS, Shendure J, Murry CE. 2019a. Dynamics of genome reorganization during human cardiogenesis reveal an RBM20-dependent splicing factory. Nature communications 10: 1538.

Bertero A, Fields PA, Smith AS, Leonard A, Beussman K, Sniadecki NJ, Kim D-H, Tse H-F, Pabon L, Shendure J, et al. 2019b. Chromatin compartment dynamics in a haploinsufficient model of cardiac laminopathy. Journal of Cell Biology 218: 2919–44.

Kang J, Kim YW, Park S, Kang Y, Kim A. 2021. Multiple CTCF sites cooperate with each other to maintain a TAD for enhancer–promoter interaction in the β-globin locus. The FASEB Journal 35: e21768.

Poleshko A, Shah PP, Gupta M, Babu A, Morley MP, Manderfield LJ, Ifkovits JL, Calderon D, Aghajanian H, Sierra-Pagán JE, et al. 2017. Genome-Nuclear Lamina Interactions Regulate Cardiac Stem Cell Lineage Restriction. Cell 171: 573–587.

Rodríguez-Carballo E, Lopez-Delisle L, Zhan Y, Fabre PJ, Beccari L, El-Idrissi I, Huynh THN, Ozadam H, Dekker J, Duboule D. 2017. The HoxD cluster is a dynamic and resilient TAD boundary controlling the segregation of antagonistic regulatory landscapes. Genes Dev 31: 2264–2281.

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

Learn more at Review Commons

Referee #2

Evidence, reproducibility and clarity

Becca et al. characterized the functions of GATA4 and CTCF in the context of cardiomyogenesis. The authors aim to establish a link between 3D genome changes (A/B compartment and long-range chromatin interactions) and activation of cardiac specific genes such as TTN. They showed opposite effects of GATA4 and CTCF in regulating these genes as well as phenotypical traits. I have the following suggestions and questions:

Major comments:

1. CTCF regulation at TTN locus:

(1) Figure 1A: The claim of the authors about convergent CTCF sites and transcriptional activation of TTN is quite simplistic. This claim is only valid when we know where cohesin is loaded. If cohesin is loaded at then intragenic GATA4 binding site, then the only important CTCF sites is at the promoter of TTN. I suggest that the authors read few more publications which may help the authors to better understand how cohesin and CTCF team up to regulate transcription, such as Hsieh et al., Nature Genetics, 2022; Liu et al., Nature Genetics, 2021; Rinzema et al., Nature Structural and Molecular Biology, 2022.

Suggestion: The authors should add cohesin (RAD21/SMC1A) and NIPBL ChIP-seq for better interpretation. (2) Figure 3B: If delta2CBS only has heterozygenous deletion of CBS6, why we would expect the binding will be weaken to 50%. However, the CTCF binding is reduced to around 1/10 in the ChIP-qPCR. How do the authors explain this?

(3) Figure 3C: There are two problems with the 4C experiments: (a) The changes are really mild. In fact, none of the p-values in Figure 3D are significant; (b) The authors should also consider a model that CTCF directly serves as a repressor. In this way, 3D genome may not be involved. B-A switch is simply caused by the activation of the locus. 2. (CTCF) detachment: The authors mentioned few times "detachment". In the context of this manuscript, the authors indicate detachment from nuclear lamina. However, the authors haven't provide convincing evidence about this.

(1) Figure 1D: I doubt whether such changes of CTCF protein abundance will lead to LAD detachment. Suggest the authors read van Schaik et al., Genome Biology, 2022. With the full depletion of CTCF, the effects on LADs are still very restricted.

(2) Figure 2D: Lamin B1 should be mostly at nuclear periphery. I have few questions: (1) is the antibody specific? (2) do these cells carry mutation in LMNB1 gene? (3) is the staining actually LMNA? 3. Opposite functions of GATA4 and CTCF: These data in Figure 5E-H argues the opposite role of GATA4 and CTCF in transcriptional regulation. Would it be that CTCF KD just affected cell proliferation, which is actually known for many cell types, rather than affect CM differentiation process? If this is the reason, inversed correlation between CTCF KD and GATA4 KD in Figure 4D could also be explained by opposite effects on cell cycle. 4. In discussion, the authors suggested that CTCF is a local chromatin remodeller. In my view, association with local chromatin compaction doesn't qualify CTCF as a chromatin remodeler. To my knowledge, CTCF does not have an enzymatic domain, then how does it remodel chromatin? 5. Some conclusions are drawn based on insignificant p-values, e.g. Figure 2F, Figure 3D, etc. The authors should be careful about their conclusion, and tone down their statement for the observations have borderline significance.

Minor comments:

1. Figure 1A: (1) I suggest to label two promoters in the gene model. It's unclear in the figure in the current version; (2) I was a bit confused with the way how the authors labeled CTCF directionality. I thought there are a lot of promoters. Why didn't they use triangles?
2. Figure 2C: I think the drastical reduction of titin-mEGFP levels is only due to the way how the authors analyze their FACS data. Can the author quantify on median fluorescence intensity?
3. Figure S2G: P value should be -log10, I assume. Please label it accurately.

Significance

Strengths and limitations:

I feel that single-cell analysis and functional analysis of GATA4 and CTCF using cardiac organoid model are elegant. However, the weak part of the manuscript is the link between 3D genome and activation of TTN. I also think the authors should include more possible explanations for the interpretation of some genome organization data (CTCF site deletion, 4C, etc).

Advance: The study does provide useful information to understand transcriptional regulation during cardiac lineage specification. The link between 3D genome and cardiac lineage specification is conceptually nice but needs more data to support.

Audience: developmental biologists who is interested in heart development and molecular biologists with specific interests in gene regulation.

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

Learn more at Review Commons

Referee #1

Evidence, reproducibility and clarity

This manuscript by Becca and others examines the relationship between GATA4 and CTCF in chromatin organization and cardiac maturation. There are several very interesting observations that lead to potentially new insights into the relationship between genome folding, gene expression and the relationship between transcription factors and chromatin structural proteins. To better justify their interpretations and provide a more objective analysis of the data, the authors may consider the following:

In the datasets you are examining, what are the relative percentages in each of the four groups relating compartmentalization change to expression change (A to B, expression up; A-B, down; B-A, up; B-A, down)?

This phrase in the abstract is imprecise: "whereas premature CTCF depletion accelerates yet confounds cardiomyocyte maturation."

Regarding this statement: "Disruption of [3D chromatin architecture] has been linked to genetic dilated cardiomyopathy (DCM) caused by lamin A/C mutations8,9, and mutations in chromatin regulators are strongly enriched in de novo congenital heart defects (CHD)10, underscoring their pathogenic relevance11." The first studies to implicate chromatin structural changes in heart disease, including the role of CTCF in that process, were PMID: 28802249, a model of acquired, rather than genetic, disease.

Can you quantify this statement: "the compartment switch coincided with progressive reduction of promoter-gene body interactions"?

Regarding this statement: "six regions became less accessible in CMs, correlating with ChIP-seq signal for the ubiquitous architectural protein CTCF." I don't see 6 ATAC peaks in either TTN trace in Figure 1A.

Western blots need molecular weight markers.

Regarding this statement: "The decrease in CTCF protein levels may explain its selective detachment from TTN during cardiomyogenesis." At face value, these findings suggest the opposite: i.e. that a massive downregulation of CTCF at protein level should affect its binding across the genome, which is not tested and is hard to evaluate between ChIP-seq studies from different groups and from different developmental timeframes.

A couple thoughts on the FISH experiments in Figure 2. A claim of 'impaired B-A transition' would be more convincing if you show, by FISH, that the relative distance of TTN from lamin B increases with differentiation. In the images: are you showing a total projection of all z planes? One assumes the quantitation is relative to a 3D reconstruction in which the lamin B signal is restricted to the periphery. Have you shown this? Lastly, these data are very interesting and important, provoking reexamination of your interpretation of the results in Figure 1. Figure 1 was interpreted to show that less CTCF binding led to decreased lamina (and thus B compartment) association during development. Figure 2 shows that depleting CTCF does not change association of TTN with lamina.

A thought about this statement: "Altogether, these results suggest that GATA4 and CTCF function as positive and negative regulators of B-to-A compartment switching, likely acting through global and local chromatin remodeling, respectively." GATA4 induces TTN expression and its knockdown prevents TTN expression-the evidence that GATA4 affects compartmentalization is unclear. By activating the gene, GATA4 may shift TTN to B classification.

I'm not sure what I am looking at in Figure 3C. Are those traces integration of interactions over a defined window? "Each [mutant is] clearly different from WT" is not obvious from the presentation. The histograms are plotting AUC of what? Interactions of those peaks with the mutated region? I genuinely appreciate how laborious this experiment must have been and encourage you to explain better what you are showing.

Again acknowledging how challenging these experiments are: when you mutant a locus, you change CTCF binding but you also change the DNA. Thus, attributing the changes in interactions to presence/absence of CTCF binding is difficult, because the DNA substrate itself has changed. Perhaps you are presenting all of this as a negative result, given the modest effect on transcription, which is as important as a positive result, given the assumptions usually made about such things. But the results are not clearly described and your interpretation seems to go between implying the structural change causative and being agnostic.

Figure 4C: since you have RNA-seq data, a much more objective way to present these data would be to show all data (again, A-B, up; A-B, down; B-A, up; B-A, down) and the effects of CTCF or GATA4. Regardless, you can still focus on the cardiac specific genes. But my guess is if you examine all genes, the pattern you show in panel C will not be present in the majority of cases. Furthermore, if this hypothesis is wrong, such an analysis will allow you to identify other genes affected by the mechanisms you describe and your analysis will test whether these mechanisms are in fact conserved at different loci.

Significance

This manuscript by Becca and others examines the relationship between GATA4 and CTCF in chromatin organization and cardiac maturation. There are several very interesting observations that lead to potentially new insights into the relationship between genome folding, gene expression and the relationship between transcription factors and chromatin structural proteins.

Reinforces the instability of reputation in Salem: even the innocent may be misrepresented. Reflects real community anxieties during the trials.

Direct warning: improper participation in church rituals could enhance the devil’s influence. Reinforces community-wide fear and moral control.

Implied threat: Christ already knows the false believers. Encourages the listeners to assume divine surveillance over every action.

Hypocrisy is morally worse than ordinary sin. Links moral failure with spiritual damnation and eternal consequences.

Foundation of Parris’s message: the church is a mixed body. This dualism legitimizes suspicion and moral policing within the congregation.

Repetition intensifies the theme: even those closest to God can harbor corruption. For his listeners, this implies that anyone among them could be a hidden threat.

Parris draws a biblical parallel between the apostles and his own congregation. The “Devil among the twelve” allows him to accuse indirectly without naming individuals. He emphasizes the presence of evil within the church and warns against hypocrisy.

Shows direct involvement of church members in witchcraft accusations. Illustrates how fear and suspicion infiltrated the community. Consecrates Parris as moral guardian.

Indicates the immediate social fear of witchcraft in Salem, which originated the sermon. Parris frames the witchcraft events as a divine or moral crisis requiring immediate attention.

Parris delivers this sermon during the height of the Salem witchcraft panic. Several church members are under suspicion of witchcraft. The purpose of the sermon is to warn the congregation about “Devils” within the church, encourage vigilance, and discourage hypocrisy.

Marks the religious setting: the sermon is delivered during a sacred ritual, intensifying its emotional impact. Parris uses the solemnity of the sacrament to lend authority and urgency to his message.

Links sin (especially greed) directly to spiritual damnation. Greed is portrayed as a moral failing that can turn believers into “Devils,” reflecting Puritan values and fear of moral corruption.