On 2026-04-24 03:59:28, user Alizée Malnoë wrote:

The manuscript by Cebrero et al. investigates the mechanism by which the Streptococcus pneumoniae teichoic acid flippase, TacF, recognizes the phosphocholine content of teichoic acids to ensure only fully modified subunits are transported. To determine the mechanistic basis of teichoic acid recognition, the authors utilize cryo-electron microscopy to determine the structure of TacF reconstituted in lipid nanodiscs mimicking the S. pneumoniae cell membrane, perform evolutionary coupling analysis, and utilize molecular dynamic simulations coupled with in vivo functional characterization to identify residues important for teichoic acid binding. Finally, the authors contextualize their findings by comparing TacF to other members of the multidrug/ oligosaccharide-lipid/polysaccharide (MOP) superfamily, suggesting other flippases within this family may recognize substrates via a similar mechanism. Overall, we feel that the data presented within this manuscript is well done and largely support the claims made by the authors. The consistent color-coding throughout the figures increased audience accessibility, and the model summarized the proposed mechanism well. We outline major and minor adjustments aimed at strengthening the data provided and improving clarity for a broader audience.

Major comments <br /> -While we found the data compelling, the main focus of this study appears to be a structural, mutational, and conservation analysis of the TacF protein suggesting a conserved flippase mechanism as opposed to solving the mechanistic basis of the teichoic acid transport. We recommend altering the title of the manuscript to better reflect the data presented. One such alteration would be ?Structural, evolutionary, and mutational analysis of TacF suggests a conserved flippase mechanism.? Additionally, we recommend changing the title of Figure 6 to ?Proposed mechanistic model of teichoic acid flipping by TacF? because binding and/or transport was not directly experimentally tested in this study. <br /> -The titles of each figure and section were not summarizing the conclusions drawn from the data presented. We recommend rewriting the titles of each section and figure to improve paper flow and guide readers naturally through the text. <br /> -The schematic provided in Figure 4B is excellent and greatly improves the clarity of the figure. We suggest including all residues which are mutated in 4C-E in the schematic provided in 4B. For example, R152, F227, and Y38 are mutated but not shown in 4B. <br /> - While the data shown in Figure 4C-D, indicating certain double and triple alanine substitutions of residues identified in the molecular dynamics simulations shown in Figure 3 is convincing, this data could be strengthened by confirming that the proteins harboring these point mutations are stable. The authors pose the hypothesis that only double and triple substitutions impacted growth due to cooperation of multiple residues needed to recognize the large teichoic acid precursors. It is also possible that double and triple substitutions of alanine for charged and/or aromatic residues destabilize the protein, which would give the same growth readout. This could be directly addressed in the text. <br /> - Certain combinatorial point mutants could further strengthen the finding that disrupting the binding of phosphate groups of phosphocholine and GalNac units compromise cell growth (lines 250-252). For example, including a protein harboring a double R230A and R333A mutation, rather than with the R250A mutation as is currently shown in Figure 4C, would inform solely on the phosphocholine interaction. Another example would be including mutations in R15 and R269 alone and combinatorially to inform on the diphosphate moiety interaction, particularly because these residues are discussed extensively throughout the manuscript. The study could incorporate the growth phenotype of the aforementioned mutants to adequately support the conclusion that these residues are involved in recognition of the teichoic acid, or the text could be revised to ensure the conclusions tightly align to the findings currently included.<br /> - We were surprised to see growth of the VL4012 strain when no inducer was present (Figure 4A, yellow line), along with the growth of the strains in Supplementary Figure 7. It had been previously stated that deletion of TacF was lethal in S. pneumoniae. Is the growth observed due to suppressor mutations, such as in the Plac promoter which would allow constitutive TacF expression? Or do TacF mutants often recover following an initial growth defect? The growth defect from hours 0-8 is striking, but the growth following this period could be addressed when first introducing the inducible system shown in 4A. <br /> -Figure 3 specifies that the role of residues R15 and R269 is to coordinate the diphosphate linker during recognition of the teichoic acid. However, Figure 3C shows that residue 15 is equally likely to be an asparagine or an arginine in the TacF homologs. This variation seems like it would affect the ability of TacF to properly align and recognize the molecule since the two amino acids vary in their physical properties (R is positively charged and larger, N is amidic and smaller). It is also unclear where the sequences portrayed in the sequence logo analysis originated from; are these sequences including proteins that are not true TacF homologs? Please include a more thorough description of how sequence logos were generated in the figure legend, and address the R and N dichotomy occurring at the R15 residue and how it may potentially impact the ability of TacF to recognize the teichoic acid.

Minor comments<br /> -In the text, figures are referred to with capital letters while on the figures themselves the sections are denoted with lowercase letters.<br /> - The LicB protein depicted within the model shown in Figure 1A is not mentioned in the text. A brief description of the protein?s function is needed to contextualize it with the other proteins shown. Alternatively, the LicB protein could be removed from the model and replaced with example choline-binding proteins (CBPs) which are specifically mentioned in the introduction and discussion sections.<br /> -Figure 1A, line 858: Define the P-C in the diagram as being the phosphocholine modifications.<br /> -Figure 1F-H: Including an arrow to show how the orientation of the protein has changed between panels (as done in Figure 5B/C) would increase clarity. Additionally, the boxes shown in Figure 1H do not align exactly with the boxes shown in the cytoplasmic view of the protein shown in Figure 1F directly above 1H. The cartoon model of the 14 transmembrane domains in 1F is very helpful in understanding how the protein is oriented in the cellular membrane, but the cartoon is flipped when compared with the protein structure directly below it. <br /> -The representative homologues used in Figure 2 should be explicitly named and PDB identifiers provided within the figure. <br /> -Figure 4A: As written in the figure caption on line 920, it is not immediately clear that D39V is a true WT strain, while VL4012 is the WT strain with the double expression system. Please consider more directly stating this in the figure caption.<br /> -Figure 4C-E: The x-axis label font size could be increased to improve legibility. The figure legend should also state how graphs C, D, and E differ from each other. Should nine datapoints be shown for the three technical replicates per three biological replicates? Some samples appear to have fewer data points than others. <br /> -Supplemental Figure 5D, which depicts and numbers the specific amino acids that are referenced over the course of the manuscript, was incredibly helpful for understanding where these amino acids are relative to each other and where they are relative to the distal and proximal site and the groove domain. Adding this panel to Figure 4 would serve as a good reference for readers unfamiliar with this protein to quickly reference which amino acids are being discussed and where these amino acids are on the protein. <br /> -Figure 5A: Should red letters be GalNac rather than GlcNac? It appears this nomenclature may have been switched elsewhere in the manuscript as well (Line 246, 251, 258, etc.).<br /> -Figure 5B-C: At first glance, it is not clear which orientation is the inward ?facing and which is the outward-facing conformation. Adding lines with labels indicating the two sides of the membrane similar to Figure 1F would help improve reader comprehension. <br /> -Lines 95-96 describe ?BRIL?, a fiducial marker used to aid in structural characterization. It would be helpful to further explain why this specific marker was chosen, both to contextualize how these markers function and to highlight the fact that such markers are widely used in this type of work. <br /> -Lines 101-121 detail the process of determining which construct to proceed with for further analysis using cryo-EM, but the process was difficult to follow. It would improve the flow of this passage to explain which tests were included and why before explaining the results of each screen, as listing them out in a chronological order makes it confusing as to why models that seem to be incompatible with this analysis were not ruled out and were instead included in subsequent screens.<br /> -Lines 125-126 could state the native ratio of phosphatidylglycerol to glycolipids in the S. pneumoniae membrane (if known) and whether the nanodiscs were representative of this ratio.<br /> -Line 232 states ?the first one...? when describing the two constructs introduced into the tacF knock-out strain. Replacing ?...while the other...? on line 233 with ?...the second one? would improve reader comprehension and emphasize the identity of the two constructs. <br /> -Lines 261-262 contain a sentence fragment; changing the period to a comma would fix this.

Carter Collins and Camy Guenther (Indiana University Bloomington) - not prompted by a journal; this review was written within a Peer Review in Life Sciences graduate course led by Aliz?e Malno? with input from group discussion including Clay Fuqua, Josy Joseph, Lily Pumphrey and Tahreem Zaheer. We are part of the Dept. of Biology where Malcolm E. Winkler?s group is located. Malcolm is a coauthor on a recent publication with one of the corresponding authors (Jan-Willem Veening); this prior interaction did not influence the choice of this preprint for our class.

On 2025-10-31 06:25:41, user xiaojun_ wrote:

I’m curious, how do you control for sampling bias in geography or collection time, given the uneven GISAID data? And do you think this SHAP-based framework would still hold up for viruses with strong recombination signals like SARS-CoV-2?

On 2023-07-14 23:30:00, user Zach Hensel wrote:

The revised manuscript overlooks the dispositive analysis first suggested to the authors, to my knowledge, in the first week of September 2022. The manuscript’s hypothesis of an endonuclease “fingerprint” of a synthetic origin in the SARS2 genome makes a testable claim: if regions around the sites composing the “fingerprint” are sampled in nature, engineered nucleotides will stick out like a sore thumb.

Authors were told about this test in the first week of September 2022 when people independently noted the recombinant evolutionary history and that almost all elements in the “fingerprint” are sampled in a handful of the most closely related genomes. Others rephrased essentially the same test, with Francois Balloux commenting to Alex Washburne on September 5, 2022:

Assuming we wished to follow up on this, the next step would be to test if high homology can be found to different Sarbecoviruses for (some of) the 6 fragments defined by the restrictions site (ie. there's no reason to expect natural breakpoints to match restriction sites).

This step was not taken. And it was not a difficult step. Shortly after the manuscript’s publication, Crits-Cristoph and colleagues rigorously showed that the hypothesis fails this test: https://github.com/alexcritschristoph/ancestral_reconstruction_endonucleases – the conclusion is noteworthy considering the public record, which demonstrates bias in site selection and post hoc selection of statistical tests. In fact, this manuscript’s hypothesis gained attention only after Justin Kinney, who is acknowledged for his assistance on the manuscripted, prompted the discussion by suggesting a different hypothesis about a different restriction endonuclease, BsaXI.

In the comments section of V1 of this manuscript, Alex Washburne proposed a second test of his hypothesis, claiming that “the rapid loss of this pattern is indicative of its evolutionary instability, suggesting what we observe in the SARS-CoV-2 ancestral state is not a stable pattern resulting from recombination, but a transient, unstable pattern that perhaps went against selection and reverted back once the infectious clone was subjected to selection from considerable onward transmission.” While this statement makes some dubious claims and another test is not needed, this comment shows that Washburne considers fitness changes in mutations at these sites to be another test of his hypothesis. This is a test that Washburne can conduct based upon published analysis of the fitness impacts of mutations: https://github.com/jbloomlab/SARS2-mut-fitness – as Washburne and co-authors have not published the results of this test, I will briefly do so here.

The mean, median, maximum, and minimum fitness change estimated for point mutations in the “fingerprint” of the 5 BsmBI or BsaI sites in SARS2 are -1.7, -1.4, 2.2, and -6.5. The same calculations for 1000 random samples of 30 nucleotides give -1.7, -1.5, 1.8, and -6.4 (see link above on interpreting these numbers, or simply note their similarity). A search on https://cov-spectrum.org/ shows that point mutations or deletions for one or more of these 30 nucleotides have been reported in 0.75% of sequences sampled in the most recent 3 months. Point mutations or deletions for one or more of 30 random nucleotides (a single random sample; results will vary) have been reported in 0.96% of sequences in the same period. All in all, the main point of interest in these 30 nucleotides is the attention given a hypothesis of a “fingerprint” of synthetic origin that was effectively disproven before this manuscript was published.

Finally, considering the countless number of equivalent hypotheses, I suggest that a better effort would be immune to these tests (and I can think of at least one example myself). It is critical that a manuscript of this type demonstrate that there is an unbiased rationale behind the hypotheses tested and that is plainly not the case here. One simply needs to observe that “longest fragment” is referred to 20 times in the manuscript, while “shortest fragment” goes unmentioned.

On 2025-10-31 09:55:08, user Zach Hensel wrote:

https://arxiv.org/abs/2510.23833

I have published a preprint concluding:<br /> 1. All sites contested by Bruttel et al -- those differing between SARS-CoV-2 and one or more of BANAL-20-247, BANAL-20-52, and RaTG13 (Fig 3A) -- are found in closely related genomes. The opposite is expected if these sites were engineered.<br /> 2. Equivalent fingerprints can be found for both of the non-bat, SARS2-like, sarbecoviruses (pangolin viruses MP789/Guangdong and P2V/Guangxi). These are both natural genomes.<br /> 3. BsaI/BsmBI sites in SARS2 are not anomalously "evenly spaced" when correcting errors made by Bruttel et al and when using a more appropriate metric (coefficient of variance of fragment lengths). One obvious error is dividing 1 by 1429 to obtain 0.07%; 1429 is equal to 71*(6 + choose(6,2)) i.e., the total number of restriction maps generated and not the number "within the ideal range of 5-7 fragments."

I also note that "5-8" fragments turns into "5-7 fragments" half way through the manuscript.

Things I do not address in my new preprint that are worth noting here:<br /> 1. I did not reproduce the "all enzymes" analysis because this is irrelevant to the "IVGA fingerprint" criteria.<br /> 2. Like Bruttel et al, I did not investigate whether all sticky ends were unique when calculating significance of the site distribution.<br /> 3. I did not reproduce the analysis of mutation rates and types (last two rows of Table S2). First, this is irrelevant because it's very likely that none of these sites are mutated from the most recent common ancestor with recently sampled bat coronaviruses. Second, this is inappropriately circular analysis since the sites were already identified as being different from RaTG13 and BANAL52; of course they will have a excess concentration of differences. Third, these are not "mutations" because SARS-CoV-2 did not evolve from RaTG13 or BANAL52 by "mutation". The null hypothesis is nonsensical.

I also did not address point "f" in the "IVGA fingerprint" -- "Two unique recognition sites may flank regions meant to be further manipulated." -- because it is based on false information. Bruttel et al wrote, falsely, that a 2017 paper reported a method "enabling efficient manipulations of the flanking region without having to reassemble the entire viral backbone for each variant." This is false because:<br /> 1. The substituted fragment in the 2017 paper was not flanked by BsaI sites in the resulting construct (pBAC-CMV-rWIV1); BsaI sites were not retained in the assembly.<br /> 2. There is a BsaI site in the BAC backbone (pBeloBAC11) that is retained.<br /> 3. There are 5 BsaI sites in WIV1 that were retained. There was no need to remove them because they occur in fragments that were not digested with BsaI.

Of course, one-pot Golden Gate assembly protocols require removing backbone sites to combine restriction digestion and ligation into a single step; but this is not one of them since it is a modification of a previous system, Zeng et al 2016, that used BglI, SacII, and AscI.

On 2026-04-29 16:29:42, user Samuel Ogunsola wrote:

This preprint is now published in Genome Biology. you can access it using this link - https://link.springer.com/article/10.1186/s13059-026-04094-y

On 2026-04-29 16:26:09, user Tania Lupoli wrote:

From Bioorganic Sp 2026, 2:

In this article, the authors develop an RNA-mediated activity-based protein profiling method (RNABPP-PS) to identify RNA-modifying enzymes in bacteria. The study introduces fluorinated pyrimidine nucleoside probes (5-FUrd and 5-FCyd) that are metabolically incorporated into RNA, enabling trapping of active RNA-modifying enzymes during catalysis. Notably, the method enables the discovery of a third methylation site on a previously characterized enzyme, YfjO (renamed RlmS), which is shown through knockout, rescue, and site-specific LC-MS mapping to validate m5U formation at U620 in 23S rRNA. Additionally, the authors investigate the mechanism of MnmG, providing evidence for a cysteine-dependent covalent RNA intermediate using probe trapping and mutagenesis.<br /> The study presents a compelling chemical biology approach that integrates probe design, proteomics, and genetics to enable activity-based profiling of RNA-modifying enzymes in bacteria. The conclusions are generally well-supported by the data, particularly the use of metabolic controls, quantitative incorporation measurements, and rigorous validation of the RlmS enzyme. However, several aspects of the study limit the generality of the conclusions and raise questions about the interpretation of certain datasets.<br /> Major:<br /> The scope of the method is chemically constrained. RNABPP-PS relies on fluorinated pyrimidine probes and selectively captures enzymes that act on pyrimidines and form covalent intermediates. This misses entire classes of RNA-modifying enzymes, particularly those acting on purines or using non-covalent mechanisms, and should be pointed out as a limitation to this specific method in the Introduction.<br /> Interpretation of the MnmG Western blot data, particularly for the cysteine-to-serine mutants, does demonstrate covalent probe?enzyme adduct formation, but does not fully address which cysteine is covalently bound to the substrate. Additional evidence of covalent adducts by trypsin digestion and mass spectrometry will help distinguish which cysteine is responsible for covalent binding.<br /> The interpretation of pseudouridine synthase (PUS) enrichment is not clear. The authors say that high-affinity RNA binding may explain this enrichment, but this undermines that RNABPP selectively captures catalytic intermediates.Adding additional comments would be helpful to clarify this. <br /> The mechanistic link between m5U620 and ribosome function is not established. Structural or ribosome profiling experiments would strengthen the biological conclusions.<br /> Minor: <br /> The dependence on salvage enzymes could pose an issue if certain bacteria are deficient in these enzymes; a clearer explanation of probe metabolism across species would be beneficial.<br /> The discussion of RNABPP-PS relative to prior RNABPP approaches could be more concise and better highlight the conceptual advance. <br /> Use of YfjO vs. RmlS is confusing at times. <br /> Figures 1D, 2C, and 2G lacked mechanistic clarity.<br /> In the SI, additional clarification of LC-QQQ-MS calibration and normalization procedures would improve transparency and reproducibility.<br /> The addition of RNA-sequencing can provide more information on the location of RNA sequence modifications after LC-MS/MS<br /> The presentation of proteomics data (Table S5) could be streamlined to emphasize statistically significant hits and reduce potential ambiguity in interpretation.

On 2026-04-29 16:25:27, user Tania Lupoli wrote:

From Bioorganic, Spring 2026:

In this paper, Zaber et al. successfully implement a new LC-QQQ-MS technique with OOPP enrichment and fluorinated nucleotides as chemical probes to identify a novel methyltransferase in B. subtilis. They demonstrated uptake and enrichment of fluorinated uracil in several prokaryotic species, as well as fluorinated cytosine in some of these organisms. They demonstrated that this method correctly identifies many of the known RNA-modifying proteins in E. coli and identifies YfjO as a methyltransferase in B. subtilis. Although previously annotated, they are the first to prove and characterise its activity in B. subtilis. They propose a new name for YfjO accordingly, RlmS, and demonstrate through LC-QQQ-MS the specific site of modification, and that it resides in the 23S rRNA. They further demonstrate that this is specific to B. subtilis, YfjO/RlmS, and demonstrate the phenotypic effects of its knockout.

We feel that they successfully identify a novel methyltransferase and give a good initial proof of concept for the method they describe. However, we feel that the goals of this paper are not clearly identified in the introduction, and they do not fully prove what they claim. Although the initial data of fluorinated nucleotide incorporation is promising for its effectiveness across prokaryotic organisms, similar LC-QQQ-MS would have to be done to make this claim, as the probe is only one part of the stated method.

Major: <br /> In figure 1b/c, they show the difference in FU/FC incorporation. They theorise that the large differences in activity are due to deaminase activity, through an E. coli deaminase knockout. However, they do not include the normal E. coli activity, which would be an important comparison, and they should attempt to reason why there is such a variance in deaminase activity. Deaminase knockout for some more of the organisms could be beneficial, especially B. subtilis. <br /> When reviewing the volcano plot data for E. coli they explain that PUS proteins often don't work with this type of probe. However, 8 do, and only one does not. Furthermore, dusB also does not, and this is not addressed fully. Perhaps a better explanation could be offered. <br /> In Western blots for a proposed mechanism for MnmG, they show that modifications of cysteine residues lead to a decrease in MnmG bound to RNA. They suggest this proves that both could be catalytically important, but do not really address that it could lead to changes in tertiary structure, and thus activity. Perhaps they could attempt to find previously identified X-ray structures.

Minor:<br /> In the beginning of methods/end of introduction, they briefly describe the method, and the choice of enrichment novel to this paper. This should be expanded upon to give the reader a better sense of the theory behind it, and how it might work, and the rationale behind its choice. <br /> Figures 4 and 5, although important, seem somewhat redundant, especially the LCMS data. At least one of these could be put into the SI or otherwise summarized. <br /> Figure 6f is not commented on in the main text

On 2026-04-29 16:23:46, user Tania Lupoli wrote:

From Bioorganic, Spring 2026 class 2:

The paper describes a two-step cPTM strategy for generating covalent genetically encoded libraries (cGELs) of peptide-derived macrocycles. Current approaches for installing electrophilic warheads require neutral-to-basic conditions, which promote undesired side reactions with nucleophilic side chain residues. To address this limitation, the authors propose separating the macrocyclization from warhead installation using a DKL followed by Knorr?pyrazole ligation under acidic conditions. They successfully found micromolar covalent inhibitors of PKM2 and confirmed it by functional validation assays, LCMS adduct analysis,and molecular docking experiments. This is potentially a valuable strategy for identifying new covalent ligands, but the target choice and strength of the validations should be further explored.

While the two-step synthesis of cGELs is chemically innovative and addresses prior reactivity limitations, the practical, rational, and clinical advantage of this platform remain poorly defined. Specifically, compared to existing small-molecule PKM2 inhibitors (PKM2-IN-12), PKM2-IN-12 exhibit an IC50 of 10?0.9 nM in cell-free assays. (DOI: 10.1021/acs.jmedchem.5c02169)

Furthermore, there are lingering questions regarding the ?covalent? claim; since peptides 15d, 18d, and 21d maintain low micromolar activity when the warhead is replaced with PEA, the inhibitory effect appears to rely on reversible non-covalent binding rather than the intended covalent mechanism. Finally, the observation of 1:2 adducts in LC-MS may suggest a lack of sufficient specificity for the target, which should be better explained.

The manuscript would be improved if the following concerns were addressed:<br /> Major:<br /> Existing characterization of the covalent adduct made by 27(c) and PKM2 is limited. We recommend a trypsin digest and mass spectrometry analysis of the resulting peptides as a means of identifying the specific cysteine residue that is labeled and providing conclusive evidence that the covalent adduct is formed.<br /> An explanation is needed for why the mass of PKM2 consistently differs by 12 Da from the expected mass while the mass of the adduct is accurate to 1 Da.<br /> SDS-Page gels and/or analytical FPLC needed to support protein purification given unassigned peaks in MS.<br /> Location of labeled cysteine residues is proposed to ?[modulate] the equilibrium between the dimeric and tetrameric states?. A brief explanation of the proposed mechanism of this modulation, which could be supported by docking with the tetramer, would support this claim.<br /> A comparison with existing small-molecule PKM2 inhibitors is necessary to explain the strengths and weaknesses of the method<br /> Clearer comparison of existing one-step vs two-step covalent library methods.<br /> Fig 5 caption doesn?t match figure<br /> Several LCMS traces of purified macrocycles could be improved for confidence and S2.48 is not convincing.<br /> Minor:<br /> Number abbreviations for the peptides should be provided in the main text alongside the sequences in fig 4a.<br /> Location of labeled cysteine residues is proposed to ?[modulate] the equilibrium between the dimeric and tetrameric states?. A brief explanation of the proposed mechanism of this modulation, which could be supported by docking with the tetramer, would support this claim.<br /> Should consider reorganizing several figures for clarity.<br /> Fig 1 A,E could be its own figure highlighting their goals.<br /> Fig 2, could simplify the phage for clarity.<br /> Fig 3, eliminate repetitive figures for clarity.<br /> Fig 4 perhaps eliminated entirely and put simply in text.

On 2026-04-29 16:22:14, user Tania Lupoli wrote:

From Bioorganic, Spring 2026 class:

In this article, Walker et al. address a major limitation in the construction of covalent genetically encoded libraries of peptide-derived macrocycles (cGELs). Traditional methods of strategies often rely on electrophile installation under basic conditions, which can promote off-target reactions with nucleophilic peptide side chains, creating undesired products, which neutralize electrophilic warheads, reduce the effective concentration of active binders, and increase the risk of false negatives during screening. The authors demonstrate these problems experimentally, showing how conventional electrophile addition can compromise library quality, then introduce their improved strategy, which decouples macrocyclization from electrophilic addition under different pH conditions. To prove their concept, they apply the method to generating inhibitors of Pyruvate Kinase M2 (PKM2). Ultimately, six lead macrocyclic peptides with IC50 values in the low micromolar range are identified. Overall, the study presents a more controlled and chemically selective platform for display-based covalent inhibitor discovery, directly addressing the challenge of unwanted electrophile?peptide side-chain reactivity.<br /> The shortcomings of the field are clearly outlined, however, the response to these shortcomings is not decisive. While the goal of the PKM2 inhibitor target identification is to prove the proposed two-step method, it appears to be split between both a proof of methods and a drug discovery narrative. To have both in one, both arguments need to be strengthened. The manuscript would be improved if the following comments were fully addressed:

Major Comments:<br /> Further characterization of the drug properties: Binding kinetics and competition assays could strengthen the argument supporting the identified hit compounds.<br /> Why aren?t the experiments described in Figs 1B-1D performed with the proposed two-step method to have a direct, vis a vis comparison? This would make the argument in favor of the two-step method more convincing. <br /> Although the authors report 6 PKM2-targeting macrocyclic peptides with low IC50 values, comparing these values directly to known PKM2 inhibitors reported in literature could better contextualize how effective the 6 lead compounds are relative to known inhibitors.

Minor Comments:<br /> The authors discuss their choice of peptide size as balancing diversity of ligands and cell permeability. Despite this, it was not addressed again in the paper. Indication of future studies to show the ability of the macrocycles to work in vivo would be helpful to demonstrate permeability<br /> Figures<br /> Figure 3 is overpopulated and difficult to interpret at a glance<br /> Figure 1 contains a lot of information (LC and MS data) that could be moved to the SI. 1D contained important comparison values for the traditional strategy of electrophile installation, the table should be formatted to be more easily interpretable at a glance.<br /> Figs 5B, 5D need explanation why PKM2 mass is off by 12 Da, but PKM2 + inhibitor mass is an exact match<br /> Supporting Information<br /> In SI2, where HPLC and LCMS data is presented, some of the starting material and product masses are missing. For example, ASCLFNCPL-DKL.<br /> The retention times of the LCMS peaks should be indicated.<br /> Overall Aesthetic<br /> Spelling errors<br /> ?...interact with PKM2 thought non-covalent interactions? (page 5)<br /> ?...suggesting multiple binding sights.? (page 6)

On 2026-04-29 03:57:22, user Sam Olechnowicz wrote:

Now published: https://www.nature.com/articles/s42004-025-01665-1

On 2026-04-28 21:15:11, user Fiona Clark wrote:

Hello! I am a student at UCLA and recently read this paper as part of a journal club. I wanted to say that this is a very impressive study overall, especially in how deeply it characterizes the biology of a-syn aggregation in human neurons. The PTM mapping in particular stood out to me as incredibly thorough. Tracking multiple modifications like pS19, pS87, and nitration across several aggregate morphologies really highlights how Lewy Bodies are chemically evolving structures and able to be replicated using this model. I also thought the functional link to dopaminergic dysfunction was very compelling, particularly the data showing TH sequestration into inclusions. On top of that, the overall technical rigor was very strong throughout (the CLEM revealing detailed ultrastructural features like MLBs was impressive!).

That said, there are a few areas where I think the study could be strengthened. The proposed stepwise maturation model is really interesting, but it relies on static timepoints (like D21 and D56), which makes it hard to definitively conclude that one aggregate form transitions into another within the same cell. Incorporating live, longitudinal tracking would make this model much more convincing. Additionally, while the use of PFFs is a clear improvement over overexpression models, introducing synthetic seeds at relatively high concentrations still creates a somewhat artificial starting point compared to the slow, progressive nature of aggregation in the human brain. It would be interesting to see whether lower concentrations over longer timescales produce different aggregation dynamics.

Overall though, this paper provides a really valuable and physiologically relevant system for studying PD. It feels like a versatile and well-validated model that could be extremely useful and impactful for future mechanistic studies and therapeutic testing.

On 2026-04-28 16:41:22, user willard wrote:

I believe the github repo [https://github.com/MGI-EU/PanVariants](https%3A%2F%2Fgithub.com%2FMGI-EU%2FPanVariants%3Al2r-gUOirQQbTqx0BtP6SwluE3k&cuid=2634513 "https://github.com/MGI-EU/PanVariants") is not yet public?

On 2026-04-28 08:22:27, user Sauvage Thomas wrote:

Hello,

Fig1 caption (boxplot) is to be clearly labelled as 16S data. Overall, should be more clearly and explicitely indicated results based on 16S and those on 18S.

Also there is no information on where to ASVs will be deposited. Since some ASV are clearly listed in the text as abundant etc, they should be made available. This will help molecualr taxonomists do further work and eventually increase reference databases for classification purposes and evolutionnary studies -Thanks

On 2026-04-27 18:02:55, user Mrunal Natu wrote:

Summary of findings

Elhaw and colleagues propose RHOV as a potentiator for anchorage independence, anoikis resistance, migration, invasion, and cytoskeletal remodeling in response to anchorage independence in ovarian cancer. The authors link RHOV expression with the aforementioned phenotypes. They suggest that anchorage independence is through RHOV mediated<br /> transcriptional changes that influence key processes such as migration, adhesion, and apoptosis that shape the invasive behavior and collagen-binding capacity. Additionally, they observed that RHOV-KO increased susceptibility to anoikis, an important precursor for anchorage independent growth and subsequent transcoelomic metastasis. They analyzed RHOV?s role in promoting migration, invasion, compaction, and mesothelial clearance. Their findings suggest that RHOV promotes an increase in these phenotypes. Finally, they suggest RHOV?s activation of cJun as a possible mechanism mediating this. This work is appealing, given the unfortunately low survival rate of epithelial ovarian cancer due to the fact that the majority of patients are diagnosed at an advanced stage. By presenting evidence elucidating novel mechanisms by which ovarian cancer cells migrate and invade the omentum it can inform strategies to improve clinical outcome. This paper will be impactful in the field as RHOV has not previously been shown to be involved in ovarian cancer metastasis. However, some concerns exist that are detailed below.

Major concerns

In the in vivo model, there is no control for tumor size. It looks as though RHOV plays a role in tumor size which, in mouse models, is directly correlated with metastasis. Not controlling for size could preclude evidence of metastasis unless both the control tumors and the RHOV-KO tumors are stopped at a set size. Additionally, there is a lack of orthogonal testing in figure 3, that could be used to verify the same results and build confidence in the data leading to the ultimate conclusions of the paper. The connection between RHOV and cJun wasn?t clearly tested, while the rationale for looking at cytoskeletal remodeling makes sense, the authors should refrain from making claims about RHOV regulating cJun because RHOV is also not the only factor that is regulating C-Jun, so the direct connection is still not clear and additional testing would need to be done in future studies to clearly establish their claims. Finally, the RHOV constructs in Figure 6 should be confirmed to be working as expected.

Minor concerns

A few limitations raise several concerns in the confidence of the overall conclusions drawn from this paper. Statistical analyses in panels 1E-F are missing. The figure alone doesn't appear convincing, and a lack of statistical significance doesn?t build confidence for the overall conclusions of figure 1. There are also minor concerns in data across panels in figure 2H and 2D that raise concern in the interpretation of RHOV?s role in tumor growth. In panel 2H, it looks as though panels for day 29 and 35 might have been switched for the RHOV-KO. In the way figures are currently presented, it looks as though there was a decrease in tumor size between days 29 and 35, if accurate, this should be explained. In panel 2D, the statistical test should be clarified as a student?s t-test cannot be used since one group has all 0 values. Additionally, there are inconsistencies in legend orientation, specifically in panel 2M, that creates confusion in data interpretation.

On 2026-04-27 17:43:39, user Rachael Shaver wrote:

The authors conveyed crucial background information on how neutrophils are components of innate immunity and defend against bacterial and fungal pathogens, as well as how recent studies have identified lipids as playing a significant role in neutrophil biology. They identified a clear gap in knowledge in lack of understanding how neutrophils acquire lipids from the environment, as well as an additional gap in knowledge regarding what the impact of lipid uptake is on neutrophil mobility. They also conveyed a gap in knowledge concerning if neutrophil lipid uptake contributes to atherosclerosis progression. One of the main findings of the paper is that plaque progression is driven by neutrophil lipid uptake. Second, lipid uptake is TLR receptor specific. Finally, that LDL uptake negatively affects cell migration creating a decrease in movement. These main findings lay the groundwork of the research done in the paper and the experiments conducted. This paper had major strengths which helped to support these main findings. One major strength was that each sequential assay was easy to follow and was logical according to the main points that the paper was trying to address. This paper first identified an interest in neutrophils and how they acquire lipids from the environment, followed by receptor uptake of lipids using TLRs, and then looking at how this uptake alters atherosclerosis. This order of events helped to facilitate understanding of why each assay was conducted. A second major strength was that the paper adequately defined how the active pathway requires actin polymerization for LDL uptake. Moreover, they included a strong positive control with FSL activating phagocytosis. Finally, I believe they provided clear emphasis on how TLR 2/6 acts in FSL-mediated uptake. While these results are of high interest to this research and build on knowledge surrounding neutrophils, there are some concerns within the paper that should be addressed.<br /> One major concern of this paper is found in figure 3, which surrounds the use of FACS when studying cell morphology and structure. FACS is very stressful for cells and disrupts cell structure following cell sorting, therefore, it is crucial to note when researching the cytoskeleton, ECM and cell-cell interactions. Specifically, in figure 3.a-d, it would be helpful to know when the cells were used following FACS as well as when the cells were imaged for immunofluorescence as cells require at least 24 hours to rebound from FACS. This also introduces a problem of temporality as we do not know whether LDL uptake is causing actin polymerization or if actin polymerization is required for LDL uptake of the cells, which is integral to understanding the progression of this process in neutrophils. Another major concern is in figure 1, as this figure did not include control images in panels F and G, which would allow for stronger comparison of results. Without a control group, we cannot accurately discern how strong phalloidin staining was in oxLDL-positive neutrophils.<br /> A minor concern of this paper is how the data presents in figure 4 with regards to the IF experiment the authors performed. The authors normalized area covered by macrophages to plaque area, however, the representative IF images chosen do not reflect the finding of macrophage recruitment by neutrophils. Alternate representative images could be used here to address this, however, given that there is not a significant difference shown between the cre+/cre- groups, claims of macrophage recruitment lack subsistence without additional data.<br /> Another minor concern of this paper lies in figures 1 through 5. While these figures indicate significance, these figures are missing specific significance values between groups. Adding significance will strengthen the results shown and provide clarification to the analysis being performed as some graphs have a decent amount of overlap in their data points.<br /> Final minor concerns in this paper include that the author?s takeaway from the data presented in figure 5 that TLR2 is required for FSL-mediated uptake and active process is not answered by the experiment they performed. The authors used FSL1, which is an agonist of TLR2, but other ligands could activate other receptors and produce this result. Additionally, the authors did not provide sufficient data to discuss the significance of monocyte recruitment that was not cohesively drawn together at the conclusion of results and discussion. To address this, the working model could be altered to remove the monocyte recruitment aspect. <br /> Overall, the authors found sufficient evidence to show that targeting neutrophils can decrease plaque formation in mice with atherosclerosis. This highlights a need to research other potential cell types that may contribute to plaque formation, such as macrophages or monocytes, which may be novel targets for treatment as well.

On 2026-04-23 06:43:54, user David Patrick Kateete wrote:

Dear technical team, our preprint article was peer-reviewed and is now published in the African Health Sciences Journal, see here https://www.ajol.info/index.php/ahs/article/view/322161

On 2026-04-22 19:16:13, user Ashley wrote:

Great preprint demonstrating how AI can be leveraged to identify novel disease targets effectively. I was wondering if you could provide supplementary figure and table files.

On 2026-04-22 17:56:16, user Williams Lab wrote:

This paper was published in Nature on April 22, 2026 as an open access at

Our paper is openly accessible to anyone today here

https://www.nature.com/articles/s41586-026-10407-9

On 2026-04-22 10:15:50, user Layne Sadler wrote:

Other EMT markers to consider would be Cdh1 (Ecad) and Snai2 (Slug)

On 2026-04-22 01:29:29, user Layne Sadler wrote:

Supplemental table 8 (cancer sphere in response to ADCP + v.s. -) shows Twist2 isn't significantly upregulated while Sox9 and Vegfa are downregulated. Isn't that at odds with the cluster analysis and EMT claims?

On 2026-04-22 00:25:28, user Luis Eduardo Figueroa Rivera wrote:

Amazing work!! Now I can run some low-medium sized structures on 8 GB. Thank you.

On 2026-04-21 18:07:12, user Logan Suits wrote:

The figure 5 figure legend says that these scale bars are 50 um and 100 um, but this should be 50 nm and 100 mM, right?

On 2026-04-21 13:29:28, user jean r wrote:

Very interesting method to monitor AAV expression. I believe this recent article would be nice to cite in the discussion or introduction: https://doi.org/10.1093/brain/awag026

On 2026-04-20 09:13:04, user Bruno Ghirotto wrote:

Now published in Nature Communications: https://www.nature.com/articles/s41467-026-71317-y

On 2026-04-20 07:54:22, user bagnaninchi wrote:

Please also consider these references from our group that were instrumental in establishing the field, and their application in monitoring drug toxicity in liver spheroids.

Bagnaninchi, P. O., Holmes, C., Drummond, N., Daoud, J., & Tabrizian, M. (2011). "Two-dimensional and three-dimensional viability measurements of adult stem cells with optical coherence phase microscopy." Journal of Biomedical Optics, 16(8), 086003

Holmes, C., Tabrizian, M., & Bagnaninchi, P. O. (2013). "Motility imaging via optical coherence phase microscopy enables label-free monitoring of tissue growth and viability in 3D tissue-engineering scaffolds." J Tissue Eng Regen Med, 9(5), 641-645.

Martucci, N. J., & Bagnaninchi, P. O., et al. (2018). "Nondestructive Optical Toxicity Assays of 3D Liver Spheroids with Optical Coherence Tomography." Advanced Biosystems, 2(3), 1700212.

On 2026-04-18 14:53:49, user Z. Brown wrote:

This manuscript makes a selective, and therefore potentially misleading, case for H3.3-Q5H. The authors pool Q5 mutations across all H3 genes, note that Q-H is the most frequent Q5 substitution, and then move directly into HA-tagged H3.3-Q5H follow-up in A549 cells, without the most obvious controls: side-by-side testing of canonical H3.1/2 Q5 transgenes. Under those conditions, observing a chromatin and growth phenotype from forced H3.3 expression is not especially persuasive. Given that the authors also acknowledge reliance on lentiviral cDNA overexpression rather than endogenous editing, the claim that H3-Q5H is a bona fide human oncohistone feels overstated relative to the evidence presented.

On 2026-04-18 10:15:30, user Prof. Emad M. Abdallah wrote:

I have some notes on the feasibility of such a combined approach to control or "eradicate" multidrug-resistant Serratia marcescens biofilms.

Using several antibiotics, phages, and antimicrobial peptides together seems more like a trial-and-error combination than a carefully designed treatment strategy!.

In addition, the study does not examine possible toxicity or the behavior of these "cocktail" agents in the body. or how they may interact with each other, which limits its clinical value???

Therefore, although the claimed in vitro results are promising, this approach is not yet ready for clinical use and should be presented as an early-stage strategy that still needs careful optimization and safety testing.

Regards,

Prof. Emad M. Abdallah

On 2026-04-18 02:05:56, user Alizée Malnoë wrote:

The manuscript by Boyeldieu et al. explores the cell-wall poisoning mechanism of the RumC1 bacteriocin, produced by Ruminococcus gnavus E1. The authors developed an ordered genome-wide mutagenesis strategy in Streptococcus pneumoniae to identify RumC1-resistant mutants. These studies helped confirm RumC1's inhibition of transpeptidase activity, thereby blocking peptidoglycan (PG) synthesis and cell growth. The selective binding of RumC1 to the nascent PG had not been reported before. They also identified the immunity protein RumIc1 from the RumC1 biosynthetic gene cluster as sufficient to counteract RumC1 toxicity in S. pneumoniae. The findings reported in this study have exciting implications for antimicrobial resistance, potentially shifting the paradigm for treating infections caused by multidrug-resistant pathogens. Overall, the data in this manuscript is well presented and supports the authors' claims. We enjoyed reading this manuscript and outline major and minor adjustments to improve clarity in reporting and presentation, as well as providing additional context for a broader audience.

Major comments<br /> - Figure 1D shows the morphology and HADA incorporation in WT and Rum1C-resistant mutants. In lines 206-208, you mention that Rum1C-resistant mutants show reduced HADA labeling; it would be helpful to include fluorescence quantification, as represented in Figure 6. Similarly, for Figures 2C and D, consider incorporating the quantification for cell width. It will also help support the results mentioned in lines 251-258. For instance, quantification of cellular morphology would provide clear evidence of “effects of RumC1 on cell morphology and PG synthesis differ from those caused by vancomycin”.<br /> - To strengthen the conclusion that variants of RumIc1 do not confer immunity, consider including evidence that the protein variants are stable by showing an immunoblot. Along this line, please provide immunoblots for the strains used in Figure 4 and onwards, in which genes are expressed from an ectopic promoter site.

Minor comments<br /> -All figures: Consider color coding the name of the strain and applying this color-coding across figures to improve reader comprehension. <br /> -Ensure all supplementary figures are referred to in the main text. <br /> -Statistic data representation: Consider changing “,” to “.” for the p values.<br /> -The introduction is well written. A description of WalR and WalK and how they play a role in PG homeostasis would benefit non-initiated readers.<br /> -RumC1 exhibits antibacterial efficacy against monoderm and LPS-poor strains, and an unencapsulated strain of Spn was used for this study. We understand how difficult it would have been to use an encapsulated strain for this work as they are more genetically recalcitrant, but clinically all pathogenic strains have a capsule. Please consider including discussion on the potential impact of RumC1 on an encapsulated strain.<br /> -Figure 1F: The controls were appropriate and nicely contextualizes RumC1 stress induction; Bacitracin targets undecaprenyl pyrophosphate (membrane carrier for cell wall precursors) which serves as a positive control and kanamycin, which targets protein synthesis, serves as a negative control. Please consider including this information in the legend.<br /> -Line 272: A little more background is needed about peptidoglycan (PG) synthesis, especially before mentioning the nascent PG.<br /> -Figure 4: At 0.4 µM RumC1, explain why rumIc2 is having a lag? Why are MIC and growth curves not corresponding?<br /> -Figure S9: How was the highest immunity achieved when all the immunity proteins were co-expressed? How did RumIc2 become neutral?<br /> -Figure 5C: Comment on the trend of growth of the strains at 0.4 µM RumC1 after 16 h, it appears that strains have started to grow again, likely because of suppressors. Consider including the growth curve for 24 h for clarity.<br /> -Figure 6F: Please explain how the cleavage sites from MD simulation were predicted. <br /> -Adding a few sentences in the discussion of how much RumC1 may be produced in the gut would be of interest, as well as speculation to what its effect on the gut microbiome may be.<br /> - Figure 1C: Consider modifying the color of the control (WT) so it’s clearly apparent that 0.6 µM RumC1 results in no growth.<br /> -Figure 1D: WT has debris; consider picking another field or enhancing the brightness.<br /> -Figure 1E: spr1875 appears to be a little stretched in the X axis.<br /> -Figure S3D: Please consider using in-text citations of Figure S3C and Figure S3D. Moreover, please provide an explanation of Figure S3D, in which the WT strain, when incubated with RumC1, shows lower expression of pcsB and spr1875 or is it a mislabeling?<br /> -Figure 6A: Consider color coding the last 2 D-Alanine in the model where the cleavage is happening.<br /> -Line 290: Consider defining sacculus and protoplast. <br /> -Lines 121, 141, 252: Missing references.<br /> -Line 200: Fluorescent microscopy should be changed to fluorescence microscopy.<br /> -Line 410: Points mutants should be changed to point mutants.<br /> -Figure S4: Please consider including the time of treatment on the graph for clarity.<br /> -Line 231: Refer to Figure S5C to improve clarity.<br /> -Figure 6C: This panel is not mentioned in the main text.

Josy Joseph and Tahreem Zaheer (Indiana University Bloomington) - not prompted by a journal; this review was written within a Peer Review in Life Sciences graduate course led by Alizée Malnoë with input from group discussion including Carter Collins, Camy Guenther and Lily Pumphrey. We are part of the Dept. of Biology where Malcolm E. Winkler’s group is located. Malcolm is a co-author on a publication in which one of the authors (Cécile Morlot) provided antibodies; this prior interaction did not influence the choice of this preprint for our class.

On 2026-04-17 11:28:23, user Roman Perez-Fernandez wrote:

Regarding the article titled “Reprogramming breast cancer cells to breast cancer stem cell-like by the POU1F1 transcription factor” published in BioRivx (doi: https://doi.org/10.1101/2025.04.30.651395 ), I would like to inform that it has been published in the journal: npj Breast Cancer. doi: 10.1038/s41523-026-00929-w. (Avila L, Seoane S, Rodriguez-Gonzalez S, Gois M, Arias ME, Martinez-Delgado D, Gomez-Lado N, Garcia-Caballero T, Aguiar P, Perez-Fernandez R. POU1F1 induces cancer stem cell-like traits in breast cancer cells by IL-6/JAK2/STAT3 activation and enrichment of ALDH. npj Breast Cancer. 2026 Mar 19) . Although the title of the article is not exactly the same, the content is practically identical.<br /> Thank you.

Prof. Roman Perez-Fernandez, MD, PhD<br /> roman.perez.fernandez@usc.es

On 2026-04-17 07:29:57, user alex W wrote:

Think there is a typo in the method section (?):

quote: "B.2.12. Host tropism assay...50 ????L of phage stock at a concentration of<br /> ∼105 PFU/mL was then added to each well..."

I assume it should be "10^5 PFU/ ml" since "∼105 PFU/mL (around 10^2 PFU/ ml)" would be too little?

On 2026-04-16 23:36:11, user Paul wrote:

Link to the book chapter based on this pre-print:<br /> https://doi.org/10.1007/978-1-0716-2617-7_6

On 2026-04-16 18:43:38, user Dan Eastwood wrote:

The relationship shown in the paper is entirely to be expected. Nucleotides that are overrepresented in the distribution (greater than expected frequency under the assumption of randomness) are more likely to mutate to a nucleotide that is underrepresented than the opposite. This makes the distribution more uniform, decreasing the Shannon Information.

EDIT: I misspoke: - the information should increase towards the maximum as the distribution becomes more uniform, yet the author shows it decreasing (table 2).<br /> I am also unable to replicate the values of IE shown in the IE Mutated column of table 2.

On 2026-04-16 14:20:08, user Prof. T. K. Wood wrote:

Calling these stationary-phase bacteria 'dormant' is absurd after 24 hr. Replace with better term, "non-dividing". Takes E. coli 2 weeks after starving to be dormant (doi:10.1111/1462-2920.14075).

On 2026-04-16 13:35:50, user Brett wrote:

Hi, I really enjoyed your recent paper on cerebellar-focused ultrasound stimulation. I had one quick question: did your experiments control for possible auditory artifacts induced by the pulsed ultrasound (e.g., audible sounds from the transducer or pulse–skull interaction), which might have indirectly influenced the neural responses in the hippocampus or cerebellum? Thanks!

On 2026-04-15 19:16:56, user AM Burroughs wrote:

Our paper (Nucleic Acids Research, 2015; https://doi.org/10.1093/nar/gkv1267) specifically predicted the systems described here and highlights NA37/YejK-family proteins in defense-linked nucleotide-signaling architectures. It also predicted their role in “sensing DNA, perhaps with unusual structural or compositional features...” Given that, its absence from the reference list here is a rather inexplicable omission of the published literature on the topic.

On 2026-04-07 15:15:54, user Prof. T. K. Wood wrote:

Line 69: The first and therefore the seminal report that (i) linked TA systems to phage defense for the first time and (ii) the first report of shutoff of transcription is Hok/Sok (doi: 10.1128/jb.178.7.2044-2050.1996), not ToxIN and RnlAB.

On 2026-04-15 18:48:44, user Shahriar Kamal wrote:

Dear Authors, I think this work is very significant since the binding pocket of CDN1163 has always been elusive. <br /> However, I see that to visualize the interaction of CDN1163, you used MD simulation. Why did you not use the cryoEM structure for visualization?

On 2026-04-15 14:55:36, user Kristin Jonasson wrote:

Jonasson et al. (2025) present experimental evidence for North American bats, not for European ones.

On 2026-04-15 03:57:38, user Lois wrote:

Acriflavine is a nonspecific nucleic acid stain that binds both RNA and DNA (reference: Interaction of acriflavine with DNA and RNA, L.M. Chan and Q. Van Winkle, 1969), making it unsuitable for genome size estimation as conducted by the authors. The gold standard for flow cytometric genome size estimation is the use of DNA-specific fluorochromes, such as DAPI or PI.

On 2026-04-14 20:33:09, user Dan Walker wrote:

Spelling: under Pseudovirus deep mutational scanning, maybe "log enrichment ration" should be "log enrichment ratio"?

On 2026-04-14 20:32:39, user Sergey Leikin wrote:

Manuscript has been published online (Cell Rep Methods, 6, 101383):<br /> https://doi.org/10.1016/j.crmeth.2026.101383

On 2026-04-14 16:09:26, user molviro wrote:

Title: Methodological Discrepancies in Viral Inoculum Dosing and Challenges in the Mechanistic Interpretation of the Orexin-NeuN Axis

The study by Yoon et al. identifies a correlation between SARS-CoV-2-induced hypothalamic orexin suppression and persistent cortical neuronal pathology.<br /> While the study provides a novel framework for understanding Long COVID, several methodological and interpretative aspects warrant further clarification to ensure the robustness of the proposed mechanism.

1. Discrepancies in Viral Inoculum Titers

The experimental design utilizes significantly different viral loads across the various infection models, which complicates the comparative analysis of neuropathological signatures.

Intra-model Variation: The K18-hACE2 model was inoculated with 50 PFU for longitudinal assessment , whereas variant re-analyses (Beta and KP.3) used 2*10^3 PFU , and the vaccine challenge employeed 2*10^4 PFU.

Inter-model Disparity: The mouse-adapted (MA10) model used a titer of 4*10^5 PFU. In contrast, the Influenza A (IAV) model—intended to serve as a comparative control for neurotropic effects—received only 2*10^2 PFU.<br /> This thousand-fold difference between the SARS-CoV-2 MA10 titer and the IAV titer suggests that the absence of pathology in the IAV group may be a consequence of the lower inoculum dose rather than a virus-specific signature.

2. Supra-physiological NeuN Induction and "Rescue" Interpretation<br /> The study defines the exogenous administration of orexin as a "restoration" of neuronal integrity. However, the protein expression data suggest a different biological phenomenon:

Quantitative Baseline: While focal NeuN loss is observed via immunohistochemistry, the degree of loss relative to the whole-brain baseline is not fully quantified prior to treatment.

Supra-physiological Levels: In both in vitro and in vivo models, treatment with recombinant orexin-A/B induced NeuN levels approximately 1.8 to 2.8 times higher than those of the healthy (Mock) controls.<br /> Such high levels of induction suggest that orexin signaling may be acting as a potent stimulator of NeuN protein synthesis rather than simply returning the neurons to a pre-infection homeostatic state. This distinction is critical for defining a true "rescue" effect.

3. Generalized Hypothalamic Circuit Suppression<br /> The focus on the orexin system as the primary driver of pathology may be an over-simplification given the broader transcriptomic evidence presented in the study.

Concurrent Neuropeptide Suppression: Transcriptomic and network analyses indicate that Hcrt was part of a wider module of suppressed neuropeptides, including Ucn (urocortin), Avp (vasopressin), and Npvf (neuropeptide VF).

Systemic Failure: The authors acknowledge that this concurrent downregulation suggests a coordinated collapse of hypothalamic homeostatic circuits.<br /> Attributing the observed cortical pathology exclusively to disrupted orexin signaling potentially overlooks the synergistic impact of the simultaneous loss of these other critical homeostatic regulators.

Conclusion<br /> While the identification of the orexin system as a target for SARS-CoV-2 is compelling, the methodological discrepancies in viral dosing and the supra-physiological induction of NeuN pose challenges to the proposed mechanistic link. A more standardized dosing approach and a holistic analysis of the broader hypothalamic circuitry are necessary to validate the "Orexin-NeuN axis" as a definitive pathway for Long COVID pathogenesis.

On 2026-04-14 13:37:31, user gneifd wrote:

I found the discussion could do a better job at acknowledging one important limitation of the approach. The article does make a good job concerning ecology. Although somehow bizzare, the ecology is realistic to the level that there are independent agents freely interacting. The limitation is that there is no realistically complex genotype-phenotype map nor any process of development or protein folding or any similarly complex process leading from the genotype to the phenotype. Simply there is some sort of genotype that directly determines the phenotype by a set of non-physical rules chosen by the researchers. In nature genes do not determine phenotypes directly, genes determine amino acid sequences and these fold in complex non-linear ways that lead to a somehow finite (and slightly finite and partially predictable) set of phenotypes. The same happens with development, genes do not determine the shape of the nose but the sequences of the genes that need to interact, between themselves and the already existing mechanical properties of cells, to lead to the construction of a complex morphological phenotype. These rules are known to lead to some predictability of the possible phenotypes (not all phenotypes are possible) and this to some predictability. This is explained in here and in other places

https://pubmed.ncbi.nlm.nih.gov/36739577/

On 2026-04-14 13:26:09, user gneifd wrote:

There is a mistake on the interpretation of one of the references. The article on Milocco and Salazar-Ciudad does not claim that evolution is unpredictable but that it is not predictable from quantitative genetics. In fact the same authors have an article proposing that some aspects of evolution are in fact predictable: https://pubmed.ncbi.nlm.nih.gov/36739577/

On 2026-04-14 08:38:59, user Edward Chesney wrote:

Excellent work. You may be interested in our study which found that CBD enhances THC's effects in people with schizophrenia: https://www.nature.com/articles/s41386-025-02175-3

On 2026-04-13 12:46:01, user Grant McConachie wrote:

This paper has been published at ICLR 2026. Review discussion and final version can be located here: https://openreview.net/forum?id=BUUfUcIcfE

On 2026-04-13 06:49:19, user Grégory Hammad wrote:

Dear BioRxiv,

This manuscript has been published here: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1009514 <br /> Could you let me know how to link these two records?

Thank you.

Greg

On 2026-04-13 05:56:05, user Andrew Kinghorn wrote:

This is a potentially useful paper, but there is a major issue in that previous methods are misrepresented, being described as missing functionality compared to the proposed method, when in fact this is not the case.

• Page 1 Line 20: "Building on this, Kinghorn(2011) optimized contributions and mating simultaneously (optimal mate allocation, OMA)"

This is not just OMA. Optimising contributions is optimising selection, so this is Optimal Mate Selection.

• Page 7 Line 10: "diversity was controlled through the covariance between parental pairs, obtained from G relationship matrices (Kinghorn2011; Peixotoetal.2024a)"

This could be misleading: covariance between parental pairs might be taken as, for each pair, 'covariance between the male and female crossed. This relates to inbreeding of pairs, not population diversity. Diversity was controlled through the covariance among all selected individuals. For more up-to-date information, see https://mateseldocs.com/

• Page 9 Line 67:"instead of electing the top-performing individuals, we should find those who balance expected gains with variance (Figure1). This aspect was not considered by previous studies aimed at genetic diversity preservation (Kinghorn2011; AkdemirandSánchez2016; Gorjancetal. 2018; GorjancandHickey2018; DanguydesDésertsetal.2023; Sakuraietal.2024; Peixotoetal.2024b,a; Endelman2025)"

These authors did in fact consider both genetic merit and genetic variance (via coancestries). They did not just use truncation selection of "top-performing individuals", as implied here. They used optimal contributions involving both "performance" (some measure of genetic merit) and mean coancestry to the rest of the population.

On 2026-04-12 13:33:11, user JonNir wrote:

This has been published as an open-access article in Behavior Research Methods: https://doi.org/10.3758/s13428-026-02983-5

On 2026-04-11 07:50:08, user REMeD wrote:

Peer-reviewed version is available at: https://www.nature.com/articles/s41467-026-71493-x

On 2026-04-10 23:19:39, user Tomas wrote:

Article is now published in https://doi.org/10.1016/j.ymben.2026.03.018

On 2026-04-10 18:53:20, user Devanshi Khokhani wrote:

This paper is now published. https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2026.1714069/full

On 2026-04-10 16:27:14, user Hao Wu wrote:

This preprint now published at the peer-reviewed journal: https://link.springer.com/article/10.1186/s13059-026-04061-7

On 2026-04-10 10:05:47, user Martin Lochner wrote:

Thanks for sharing this interesting manuscript as preprint. Please note that the shown structure of NBA is wrong (e.g. in Figure 5b and Extended Data Figures 9a and 9c).

NBA structure shown in your manuscript (SMILES): ClC1=CC=C(C(O)=O)C(NC(NC2=CC(C=CC=C3)=C3C=C2)=O)=C1

Correct NBA structure (SMILES): ClC1=CC=C(C(O)=O)C(NC(COC2=CC=CC3=C2C=CC=C3)=O)=C1

It is most obvious in Extended Data Figure 9c that the NBA this work (yellow) and NBA PDB: 8RD9 (cyan) are two differnt compounds.

Thank you for considering this correction.

On 2026-04-09 21:38:21, user Alizée Malnoë wrote:

The manuscript by Fridman et al. explores the unexpected finding that Aeromonas jandaei antagonistically employs a Type VI secretion system (T6SS) in a liquid environment. While researching the effector protein Awe1, which forms part of the T6SS apparatus, the authors observed T6SS-dependent intoxication of susceptible bacteria. Using a novel fluorescence-based screening method (named LiQuoR for liquid quantification of rivalry), the authors further determine that this intoxication is contact-dependent, and that contact between kin and non-kin Aeromonas bacteria in liquid is mediated by specific adhesins. Fridman et al. also identify additional marine bacteria capable of inflicting T6SS-mediated intoxication in liquid media, suggesting a mechanism for specific and contact-dependent bacterial competition and positing that such competition in liquid media may be more common in marine bacteria than previously documented. These findings have exciting implications for bacterial antagonism, potentially shifting the paradigm of how we view bacterial interactions in marine environments. We found this study to be well-written, containing high-quality data. Overall, the data presented in this manuscript are done well and support the claims made by the authors. We outline some major and minor adjustments aimed at aiding the clarity of reporting and presentation, strengthening the findings, as well as providing additional context for a broader audience.

Major Comments<br /> - We are interested in the broader implications of the LiQuoR assay, particularly pertaining to this workflow’s application to different bacteria. The observation that the amount of prey luminescence in WT on solid media grew/increased after 4 h seemed counterintuitive to us (Figure 1E). It seems as if this result could make the workflow less sensitive for experiments done solely on solid media, further explanation of this finding would clarify on the workflows applicability to other solid surface experiments. Is this related to surface area? While this does not change the findings that inhibition is occurring in both liquid and in solid, it would enhance the clarity of these results to provide speculation on why this was seen.<br /> - We are curious about your perspective on the observation that kin-kin aggregation facilitated by CaCl2 supplementation does not increase kin intoxication but does increase non-kin intoxication (Figure 2A). Please speculate on this result in the discussion. Is the concentration used physiological? <br /> - While the images shown in Figure 2B make it clear that aggregates are forming in liquid media, we have a suggestion to improve the strength of these results and account for the images not shown. For instance, quantification of the % of prey cells displaying Sytox staining would more strongly demonstrate the presence of permeabilized E. coli in multiple aggregates. This quantification could substitute Figure 2C (which can be moved into the supplemental): it was not totally clear to us why an orthogonal view was included here. If this is significant for the findings, it would increase clarity to include an explanation for an audience less familiar with this system.<br /> -Lines 192-214: From a genomics perspective, we think further explaining how potential adhesins were identified would be helpful to increase the clarity and reproducibility of the experimental design. Please explain how you narrowed down these adhesins and located them in the genome, and why adhesins were targeted for this analysis over other proteins that could facilitate a physical interaction between predator and prey species. Define the acronyms and provide rationale for naming. <br /> -Figure 6B nicely demonstrates that intoxication takes place in liquid between certain marine bacteria but not in Vpara. However, please include a control showing that V. para does intoxicate prey in solid media to strengthen these findings and confirm that this strain of V. para is capable of intoxicating prey under typical conditions.<br /> -Given the significance of the TssB deletion for the core message of this work that type VI intoxication occurs in liquid media, please consider including data that confirm the TssB deletion e.g. sanger sequencing in supplemental or as source data. A complementation assay of TssB to show that regaining TssB restores the awe1 toxicity would be valuable.<br /> - Lines 224-225/Figure 5: We are curious and excited about the implications of the balance between kin-aggregation and non-kin aggregation and how this may aid our understanding of bacterial interactions in marine environments. Based on our understanding of these results, the observation that deletion of CraAj (responsible for kin-kin aggregation) increased non-kin intoxication (mediated by LapAj) could suggest that aggregation between two kin cells, who both contain the needed immunity proteins, could dampen the intoxication of nearby non-kin cells. This result is implied by the data but not specifically speculated on or addressed. Though it may not be within the scope of this experimental design, our group was intrigued by these findings. Given your expertise in this area, consider discussing how these bacterial interactions may play out and/or include these observations as part of Figure 5.

Minor Comments<br /> -All figures: In the legends, it is stated “these experiments were repeated three times with similar results”. Please define what is meant by an experiment e.g. technical or biological replicate.<br /> -All figures: We felt that having the exact p-values indicating statistical significance is not necessary. For instance, in Figure 3B and 3D, we found it distracting that all of the values were significant by a factor of <1E-4, even when they appear different from each other. If this is simply a cutoff value, it would be helpful to keep that consistent between figures. Also, Figure 6A/B: The p-values presented, specifically the comparison between WT and T6SS – supplemented with 1 mM CaCl2 (6A) and the two left hand panels of 6B, do not appear to match the differences shown between the experimental groups. By eye, these groups do not appear different from one another but are shown to be either highly statistically significant or not statistically significant at all.<br /> - Figure 1A: To increase readability, we suggest that the colors could be more intuitive here- put WT in grey and then mix colors for double mutants. Bringing the light pink line (Δawei1 ΔtssB + pAwe1) to the front of the graph would further increase clarity.<br /> -Figure 1B/F: Making color scheme consistent between 1B and 1F would increase clarity.<br /> -LiQuoR assay: As there is often some level of variation in expression levels when working with a transformed population, confirmation that all prey strains luminesce to a similar level would provide further validation of this novel assay (similarly to what is done in FigS3B). <br /> -Figure 2A: The colored box legends showing whether CaCl2 is present or absent are inverted relative to one another, which we found to be confusing. To increase readability, please make them on the same side.<br /> -Figure 3B,C,D,E: To help guide the eye on the graphs, we suggest adding dashed lines between each new mutation group (+/- TssB).<br /> - Figure S1: Please include a loading control to verify assay input. <br /> - Table S1: Clarify the gene and strain for each mutation.<br /> - Line 112-113: It serves as an excellent control that the action of the T6SS apparatus is required for intoxication, however, since the T6SS apparatus is contained within the bacterium, would spent media contain free-floating T6SS proteins, or are these proteins only ejected from the bacterium in the presence of prey species? Please clarify. Direct evidence, such as immunoblotting, that effectors are present in the spent media from WT would make this claim more compelling.<br /> - Line 35: While this part of the introduction provides excellent background regarding the role of T6SS in interactions with eukaryotic cells, it would be helpful to also specifically mention the role of T6SS in prokaryotic communities, as much of the later work focuses on competition between bacteria.<br /> -Lines 70-71: A more thorough background on Aeromonas (lifestyle, importance, etc) is warranted.<br /> -Line 84: Please provide the exact genotype when first introducing this mutant, it would improve clarity for the reader to explicitly state that this is a double mutant.<br /> -Line 97: Clarify here that “Aj prey” in this paragraph refers to Aj which do not possess the cognate immunity protein, as the current phrasing could be interpreted to mean “prey of Aj”.<br /> -Line 138: “Desired conditions for competition” is vague. Is solid media also incubated with shaking or is it static?<br /> -Lines 156-157: The statement that all three effectors are injected into prey cells is broad and not necessarily supported within these findings. The injection of one effector could be favored, but other effectors could compensate in its absence.<br /> -Line 189: Describes Aj as stably binding to other competing bacteria. To this point, imaged aggregates have been fixed so stability of aggregates may not be known.<br /> -Line 248: Here, it is mentioned that there was a switch from using the Lux operon to using the RFP mCherry for improved cell detection. It might be helpful to clarify which fluorescent tag was used for each assay, as multiple different fluorescent tags are used.<br /> -Line 317: As the choice to test CaCl2 and the biological relevance of calcium for Aeromonas hosts is explained earlier in the manuscript, it would be interesting to include a brief explanation about the choice to include sodium chloride when assessing Vibrio intoxication rates. Presumably, sodium chloride was picked because Vibrio is commonly found in brackish water, but someone from outside the field may not be familiar with this biology. Additionally, since Aeromonas can be found in both fresh and brackish water, an interesting follow-up experiment would be to test the Aeromonas strains under different salinities.<br /> -Line 375-377: Needs citation.<br /> -Line 385: Clarify “under specific conditions not addressed within the scope of this study”.

Carter Collins and Lily Pumphrey (Indiana University Bloomington) - not prompted by a journal; this review was written within a Peer Review in Life Sciences graduate course led by Alizée Malnoë with input from group discussion including Camy Guenther, Josy Joseph and Tahreem Zaheer. We are part of the Dept. of Biology where Julia Van Kessel’s group is located, Julia is a collaborator of the corresponding author and did not influence the choice of this preprint for our class.

On 2026-04-09 13:41:09, user Michael Lichten wrote:

Figures are missing

On 2026-04-09 11:43:46, user Mohieddin Jafari wrote:

It is now published at Nat Comm: https://doi.org/10.1038/s41467-026-70394-3

On 2026-04-09 11:08:10, user Selwin Wu wrote:

Correction to Fig. 5p: A blue divider separates the gel images https://figshare.com/articles/figure/Fig_5p/31971354?file=63580170 .

On 2026-04-09 08:53:08, user Stefan Kusch wrote:

The preprint has been published in a peer-reviewed journal in the meantime: https://www.nature.com/articles/s41467-025-63889-y

On 2026-04-08 22:15:29, user Kevin J. Black, M.D. wrote:

Very cool work!

On 2026-04-08 20:58:39, user Adam Pavlinek wrote:

Link to published paper: https://www.cell.com/cell-reports-methods/fulltext/S2667-2375(26)00071-8

On 2026-04-08 13:17:35, user Hanna wrote:

The Table 2 does not correspond to its description (There is a combination of three assigned and PIK2CA E453K is two-color combination).

On 2026-04-08 09:56:56, user Mahmoud Alyosify wrote:

An exceptional achievement for Egypt and a long-overdue step in global genomics.

What makes this work truly impactful is not only the scale—1,024 whole genomes—but the fact that it begins to correct a fundamental bias in our field: the underrepresentation of MENA populations in genomic reference datasets.<br /> As someone working at the intersection of AI and Bioinformatics, this resonates deeply.<br /> For years, many predictive models—especially polygenic risk scores—have been trained on predominantly European data, leading to systematic miscalibration when applied to populations like ours.

EGP1K is more than a dataset;<br /> it is a recalibration of how we define “normal” in genomics.

It shifts Egypt from being a consumer of generalized models to a contributor of population-specific knowledge—enabling more accurate, fair, and clinically relevant AI-driven insights.

Proud to see science, data, and national identity intersect at this level✌️

On 2026-04-07 16:33:27, user Medhat Mahmoud wrote:

Congratulations to the team for this important achievement in population genetics research in Egypt. I have one question, and I hope you can help me better understand it:

We often hear that certain groups of genes are described as “Egyptian,” “Mediterranean,” or linked to some other geographic region. But I always wonder: how was that conclusion actually made?

If these labels are based mainly on samples taken from people who are living in those regions today, that does not necessarily prove that the same genetic patterns truly originated there or belonged to ancient populations from that area.

In other words, finding a gene frequently in people living in Egypt in 2025 does not automatically mean that this gene was also characteristic of people living in Egypt in the year 1000 AD, or in 2000 BC. Populations move, mix, and change over time. So attaching a modern geographic label to a gene can be misleading unless there is strong historical and ancient DNA evidence to support it.

On 2026-04-08 09:16:04, user Chris Rinke wrote:

published paper: <br /> A complete domain-to-species taxonomy for Bacteria and Archaea<br /> https://www.nature.com/articles/s41587-020-0501-8

On 2026-04-08 00:46:00, user Amanda Curtis wrote:

This article is now published online in Environmental DNA at https://doi.org/10.1002/edn3.70251

On 2026-04-07 23:24:09, user Gabriel Lander wrote:

We acknowledge a typo in this manuscript: all instances of (v/v) should be (w/v). apologies for the confusion!

On 2026-04-07 20:49:08, user John Dominy wrote:

Although supplementary data/figures are referenced in this preprint, there is no supplementary information posted

On 2026-04-07 14:45:59, user Yujia Huang wrote:

The manuscript is now published in Magnetic Resonance in Medicine<br /> https://doi.org/10.1002/mrm.70366

On 2026-04-07 08:39:09, user Guest wrote:

I must confess, on first reading I found the manuscript quite exciting, but having gone through the earlier comments, I now see rather more clearly the gulf between what the data actually show and what the authors claim.

One thing I would add to what has already been said: there is, quite remarkably, no protein localisation of KCNT1: not by GFP tag, not by antibody, in the multiciliated epidermis of Xenopus, mouse, or indeed human tissue. That is a rather glaring omission, to put it mildly. I would also agree that the proposed connection between KCNT1 and Piezo is tenuous at best.

On 2026-04-04 12:13:58, user Reader wrote:

1. The GOF vs. LOF problem is fundamental, not a caveat.<br /> All pathogenic KCNT1 variants are GOF missense mutations increasing channel conductance. This study models LOF exclusively (morpholino, pharmacological inhibition). A GOF channel hyperpolarizing the membrane and a LOF knockdown reducing potassium flux could have opposite effects on membrane potential, Piezo mechanosensitivity, and FOXJ1 regulation. The translational and therapeutic claims require demonstration that GOF variants also disrupt MCC development, e.g., overexpression of patient alleles in Xenopus epidermis. Without this, the disease relevance is speculative. The authors should also consider CRISPR-based LOF as a parallel genetic approach that would bypass the well-known limitations of morpholinos and strengthen the loss-of-function claims independently.

2. Respiratory morbidity is multifactorial; the MCC interpretation is overstated.<br /> The authors' own cohort undermines a simple MCC-defect explanation. With 59% hypotonia and 43% joint contractures, impaired cough mechanics, aspiration from oropharyngeal dysfunction, and neuromotor respiratory failure are obvious and likely dominant contributors to ventilator dependence common in severe encephalopathies regardless of cilia status. These alternatives are not discussed.

3. Human airway expression of KCNT1 is not supported.<br /> HPA classifies KCNT1 as "Tissue enhanced (Brain, Lymphoid tissue, Skeletal muscle)." Its expression cluster is "Cerebellum – Synaptic function" and all 15 nearest neighbors are neuronal genes (SCN2A, SNAP25, GABRD). Bronchus and nasopharynx show no RNA data; lung expression is minimal. At single-cell resolution, KCNT1 peaks in brain inhibitory neurons with no enrichment in respiratory ciliated cells or respiratory tissue. The authors should query human airway scRNA-seq datasets before building an airway-cilia narrative from Xenopus epidermis.

4. No gene-specific rescue for the morpholino phenotype.<br /> A kcnt1 mRNA rescue experiment is standard and missing. Pharmacological phenocopy does not rule out off-target morpholino effects. The Centrin-CFP observation (lost in morphants, present in controls) was abandoned rather than quantified a missed opportunity. CRISPR F0 knockouts would provide an independent genetic approach.

5. Quantification is insufficient for the claims.<br /> All functional data rely on categorical embryo scoring (mild/moderate/severe). This is acceptable for initial characterization but not for mechanistic or therapeutic conclusions. No MCC counts, cilia density/length measurements, basal body quantification, foxj1 qPCR, or flow assays are provided. The distinction between "fewer cilia per MCC" and "fewer MCCs" critical for mechanism cannot be made from these data.

6. Figure 4 does not establish a KCNT1–Piezo interaction.<br /> The KCNT1 antagonist-alone bars differ markedly between Fig. 4B and 4D, indicating high inter-experiment variability that undermines interpretation of co-treatment shifts. More critically, the representative embryo images are inconsistent with the stated conclusions: the KCNT1 antagonist condition in 4C and the KCNT1 antagonist + Piezo antagonist condition in 4A look essentially indistinguishable, yet the quantification claims a dramatic enhancement. The "rescue" in 4D is marginal: the majority of embryos remain moderate-to-severe. Without genetic validation of even one arm (e.g., Piezo1 morpholino + kcnt1 depletion), this is pharmacological correlation at best, compounding off-target effects at worst.

7. The therapeutic framing is unjustified.<br /> The abstract and discussion frame Piezo1 as a therapeutic avenue for KCNT1-related disease. This rests on a LOF model of a GOF disease, a weak pharmacological rescue with high variability, no evidence of KCNT1 expression in human airway MCCs, and no molecular mechanism connecting KCNT1 to Piezo signaling. The conclusions should be restricted to what the data show: KCNT1 perturbation in Xenopus disrupts epidermal MCC development, and this can be modulated by Piezo pharmacology under specific conditions.

On 2026-04-07 08:34:38, user Yue Jia-Xing wrote:

Author comment:<br /> This work has been formally published in PNAS recently under a different title:<br /> "Insights into cephalochordate genome and gene evolution from the early-diverging amphioxus Asymmetron lucayanum". <br /> Advanced online on 2026.03.20, formal publication in 2026.03.24<br /> DOI: 10.1073/pnas.2521280123<br /> Link: https://www.pnas.org/doi/10.1073/pnas.2521280123

On 2026-04-07 08:29:17, user Philip wrote:

Great work! This computational approach offers a smart, efficient path toward identifying key antigen–TCR interactions and could really accelerate malaria vaccine design.

On 2026-04-07 07:07:48, user Ayat Taheri wrote:

This preprint has now been published in Plant Communications.

Published article DOI: https://doi.org/10.1016/j.xplc.2026.101826

On 2026-04-07 03:53:05, user Prof. T. K. Wood wrote:

6 referrals to abortive infection yet it doesn't exist as Sorek defined it as cell suicide and certainly does not exist for TA systems (see doi: 10.3389/fmicb.2018.00814), since the claim of Abi is based on non-physiological over production of any toxin).

Not necessary, either, since first report of TA systems blocking phage found the mechanism is simply transcription shutoff in 1996 (doi: 10.1128/jb.178.7.2044-2050.1996), which should be cited here, for relevance of DarTG.

On 2026-04-05 06:07:37, user ari wrote:

phenomenal photos, i'm doing an art project about changing shapes in a vernal pool and all these timelines are so perfect its like a treasure chest. thank you newt scientists very much

On 2026-04-04 05:08:23, user hannah97 wrote:

Interesting work!

On 2026-03-23 11:37:45, user bbbli wrote:

This preprint provides compelling evidence that STAT1-GOF promotes Tfh1 differentiation and drives autoimmunity, with IFN-γ neutralization showing therapeutic potential.

On 2026-04-03 20:53:12, user Mengyi Cao wrote:

This manuscript is now accepted for publication in Environmental Microbiology Reports (Article DOI: 10.1111/1758-2229.70326)

On 2026-04-03 20:00:51, user Amado Rosendo wrote:

Pyroptosis is a lytic form of programmed cell death mediated by members of the gasdermin family. However, the regulation of gasdermin activity, particularly after the cleavage event that releases the proteolytic N-terminal domain, remains poorly understood. The first identified palmitoylation was discovered in bacterial gasdermins by Johnson et al. (2022). Now, Du et al. (2024) have shown that Gasdermin D (GSDMD) is also palmitoylated during pyroptosis, suggesting that lipid modification may be an important, more broadly conserved mechanism for regulating gasdermin function. These findings shed light on events that occur after gasdermin cleavage and suggest that modifications to the liberated N-terminal domain may influence pyroptosis. In this research manuscript, Tao et al. examine whether Gasdermin A (GSDMA) is regulated by S-acylation, as is the case for other gasdermins. The authors used the known S-acylation mechanism, in which fatty acyl chains are attached to a cysteine via a thioester bond, and employed this system to determine which fatty acid length is preferred. They focus on S-palmitoylation, a specific type of S-acylation in which the attached fatty acid is palmitate, a 16-carbon saturated fatty acid. In this paper, the authors ask whether GSDMA is regulated by S-acylation at conserved N-terminal cysteines and whether dynamic control of this acylation, including deacylation by ABDHD17A, provides a new mechanism for regulating GSDMA-mediated pyroptosis. <br /> To assess GSDMA S-acylation, Tao et al. performed a click reaction with biotin-azide to test which acyl chains preferentially labeled GSDMA3 and found that the Alk-C16 and Alk-C18 probes gave the strongest labeling. They then tested whether endogenous GSDMA could be acylated with Alk-C16, a biologically relevant acyl chain, and showed that the protein was labeled using a second click-labeling approach. The authors further showed that GSDMA3 S-acylation is a dynamic process. After 1 hour of Alk-C16 labeling, GSDMA3 acylation was detectable and increased up to 12 hours. In transfected HEK293 cells, pulse-chase experiments showed that the turnover rate of GSDMA3 acylation was approximately 1.8 hours. This indicates that the lipid modification undergoes active cycling rather than remaining permanently attached. These data conclude that biologically relevant fatty acyl chains modify GSDMA3 and suggest that its S-acylation is a dynamic process regulated by the addition or removal of fatty acyl chains. <br /> The authors compared GSDMA sequences across different species and identified three conserved cysteine sites: C37, C205, and C225 in the N-terminal domain of GSDMA. They mutated each cysteine residue to alanine in GSDMA3 and found that each mutation reduced acylation, but a stable triple-cysteine mutant completely abolished GSDMA3 acylation. They sought to determine whether recombinant human GSDMA can be acylated independently of palmitoyl transferases via direct interaction with palmitoyl-CoA. They concluded that GSDMA can be S-palmitoylated in vitro and that this modification is influenced by both temperature and GSDMA concentration. The authors transfected HEK293 cells with the GSDMA N-terminal domain carrying a leucine 184 to aspartic acid mutation (NT-L) or the previously described triple cysteine-to-alanine mutant (NT-L-3CA). Using immunofluorescence imaging, they observed that GSDMA3-NT-L-3CA remained cytosolic and did not form oligomers. In contrast, GSDMA3-NT-L localized to the membrane and showed oligomer formation by Western blot. To test these effects in a more relevant cellular context, they performed cell viability and LDH release assays comparing GSDMA3-NT-L and GSDMA3-NT-L-3CA in an immortalized keratinocyte cell line. Tao et al. found that the triple cysteine mutant increased cell viability and decreased LDH release. Based on these results, they conclude that S-acylation of GSDMA3 is crucial for membrane localization, pore formation, and pyroptosis.<br /> In the section labeled “ABHD17A regulates GSDMA3 deacylation and pyroptosis,” the authors identified several deacylating enzymes that interact with GSDMA3. They proceed with ABHD17A, which removes the 16-carbon fatty acyl chain and demonstrates the strongest deacylating activity. Their data in Figures 4A and 4B show that ABHD17A, along with other deacylating enzymes, can reduce the acylation signal of GSDMA3. The quantitative data in Figures 4D and E show only a partial rescue of cell viability and LDH release in cells co-transfected with ABHD17A compared to GSDMA NT-L alone. While the authors report statistically significant differences, cells co-transfected with ABHD17A remain closer to the NT-L group than to the control. The microscopy data in Figure 4C are not representative of the quantitative data, as they show confluence levels more similar to those in the control. Therefore, it would be warranted to include representative images to maintain consistency with the quantitative data. As a result, Figure 4C appears to overemphasize the biological significance of ABHD17A overexpression, and the authors should consider scaling their conclusion accordingly, treating ABHD17A as a partial inhibitor rather than a pyroptosis suppressor. This concern follows with the latter half of Figure 4, as the knockdown experiments share a similar critique to the overexpression data. In the reviewer’s opinion, the microscopy image (Figure 4G) and the quantitative data (Figure 4H and I) are not in line with the author’s conclusion, “ABHD17A knockdown increased GSDMA3 acylation and exacerbated GSDMA3 NT-L-induced pyroptosis, measured by LDH release and cell viability.” Thus, the reviewer suggests including images that align with the author's conclusion. Additionally, single-cell analysis (e.g., SYTOX Green for flow cytometry, DAPI, or PI staining) would be warranted to quantify the fraction of cells undergoing cell death. If the authors continue to use microscopy, a quantitative, blinded assessment of multiple micrographs with a clearly stated n would be necessary to strengthen their work. <br /> “Dynamic S-acylation of GSDMA regulates pyroptosis” should be altered to “...GSDMA3 regulates pyroptosis” to specifically tailor to the authors’ findings. The original title gives readers the impression that the research is solely focused on human GSDMA; in fact, the authors perform experiments using a mouse homolog, GSDMA3, to transfect human cell lines. Hence, the paper's structure misleads readers and fails to encapsulate all aspects of the study.<br /> In conclusion, this preprint is well written and offers important insight into S-acylation and its role in regulating GSDMA3-mediated pyroptosis. The study presents a coherent mechanistic framework, multiple complementary assays, and has the potential to advance our understanding of gasdermin regulation. Overall, the authors did an excellent job of presenting most figures clearly and helped guide the reader in understanding this new pathway.

Minor critiques: <br /> -The preprint contains several spelling errors that should be corrected throughout. <br /> -Figure headings should be consistently aligned with the labeled lanes. <br /> References to figures in the main text should match the panel labels exactly. (e.g., “HEK293 cells were transiently transfected with GSDMA N-terminal domain carrying…cysteine residues were mutated to alanines (NT-L-3CA)” (pg. 9). However, Figure 3E, which describes the mutation above, is labeled as GSDMA3, not GSDMA.)<br /> -Figure panels should be labeled consistently throughout the figures (e.g., by numbering them A, B, C, etc). <br /> -Unclear data where Figure 2F is established. Supplemental Figure S2 shows a single time point; multiple time points at different concentrations of GSDMA are needed to justify Figure 2F and must be clearly organized. <br /> -The relevance of Figure 2A is unclear and unexplained. <br /> -Figure 4 cell viability (D and H) and LDH release (E and I) experiments compare adjacent conditions, where they should compare each condition to the Control of their systems.

Sincerely, <br /> Amado Rosendo, Undergraduate Researcher <br /> Special Topics 390 with Prof. Dovey <br /> Amherst College <br /> 4/2/26

On 2026-04-03 18:23:31, user איתי לוק wrote:

I learned so much from this article. It gave me new perspectives and deepened my understanding of the topic.<br /> Truly inspiring and makes me want to keep exploring and learning more.

On 2026-04-03 16:17:01, user Hesham Tawfeek wrote:

The findings here are very similar to those of a PKA transgenic mouse model that expresses a constitutively active PKA catalytic subunit in the osteoblast lineage cells. It is very surprising that the authors never discussed or referenced this paper:

Endocrinology<br /> . 2016 Jan;157(1):112-26. doi: 10.1210/en.2015-1614. Epub 2015 Oct 21.<br /> Activation of Protein Kinase A in Mature Osteoblasts Promotes a Major Bone Anabolic Response<br /> Liana Tascau 1, Thomas Gardner 1, Hussein Anan 1, Charlie Yongpravat 1, Christopher P Cardozo 1, William A Bauman 1, Francis Y Lee 1, Daniel S Oh 1, Hesham A Tawfeek 1

On 2026-04-02 18:58:32, user Flavia Costa wrote:

This is really interesting and lovely work! I was not sure if you were aware of a previously published paper on this locus ( https://doi.org/10.3390/ijms23126443 ) so I wanted to bring that to your attention in case it should be acknowledged and discussed in the context of your compelling data. Furthermore, I am curious if you had any thoughts about why the sequence conservation is lower and more variable in length between amino acids ~150-300 within the staphylococci. Great work and I hope the review process goes smoothly!

On 2026-04-02 17:52:40, user Abhishek Agarwal wrote:

Very nice ????

On 2026-04-02 17:04:46, user Claes Gustafsson wrote:

Co-authors Surya Karunakaran and Ferenc Boldog discuss this BioRxiv paper in a broader bio-manufacturing perspective in the linked Thought Leadership review in GEN https://reader.emagazines.com/issue/gen-magazine/march-2026/genome-editing-for-biopharmaceutical-manufacturing#p20

On 2026-04-02 15:35:26, user M Ganesh Kumar wrote:

Now published in the Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:31892-31929, 2025. URL at https://proceedings.mlr.press/v267/kumar25a.html

On 2026-04-02 05:26:18, user Maninder wrote:

1) What was the rationale for administering Clozapine prior to immunization? Was post-immunization administration of clozapine considered to assess its therapeutic efficacy?<br /> 2) Could this paradigm be adapted to a milder or more chronic model in which delayed clozapine administration would be feasible?<br /> 3) On what basis were the observed behavioral features interpreted as psychosis rather than catatonia?

On 2026-04-01 16:40:03, user Prof. T. K. Wood wrote:

I note that we have recently shown phage infection itself induces persistence (<br /> DOI: 10.1111/1751-7915.145432024) and that the toxin/antitoxin system MqsR/MqsA/MqsC creates dormant cells and eliminates T2 phage while the cell is dormant (with the restriction system McrBC, doi 10.1128/spectrum.03388-23). Since there are few papers about phage and persisters, I would expect these two to be included in your work.

In addition, we have shown that for E. coli it takes two weeks for starvation to induce persistence in E. coli (doi:10.1111/1462-2920.14075, 2018).

Your paper middle page 2: my group was the first to show acid and oxidative stress induce persistence by 12,000-fold (doi:10.1111/j.1751-7915.2011.00327.x<br /> © 2012, 2012). This 3 years prior to your ref 27.

Page 3 end of Introduction: the most prevalent phage defense is not CRISPR-Cas but toxin/antitoxin systems, as we discovered in 1996 (doi: 10.1128/jb.178.7.2044-2050.1996) and the role of TA systems in persistence with phage should be referenced as indicated for MqsR/MqsA/MqsC above.

Discussion: there is ONLY one mechanistic model for persister cell formation and resuscitation ( https://doi.org/10.1016/j.bbrc.2020.01.102 for ppGpp leading to ribosome dimerization, https://www.sciencedirect.com/science/article/pii/S2589004219305371?via%3Dihub ) and it fits well with your ‘hibernation’ text.

Results Section 7: not solely CRISPR shown to employ dormancy as we have shown, as indicated in previous e-mail, that toxin/antitoxin system induce persistence to fight phage (MqsR/MqsA/MqsC with T2 and E. coli).

On 2026-04-01 14:58:01, user Giuliano Zanchetta wrote:

The paper has now been published in NAR as "Decoding nucleic acid contributions to phase separation and ordering in biomolecular condensates":<br /> https://doi.org/10.1093/nar/gkag253

On 2026-04-01 14:11:04, user Willard Ford wrote:

The linked GitHub repository does not contain the code to run this tool. Is there a target date to actually publish the code?

On 2026-04-01 10:56:26, user Juan David Gutiérrez wrote:

Was the source of epidemiological information the national surveillance system (SIVIGILA)?

Did the authors include all dengue cases, or did they apply additional criteria (e.g., laboratory-confirmed cases, clinical confirmation)?

Did the authors control for allochthonous cases? This question is particularly relevant in large cities, where reported cases may represent a combination of infections acquired locally and infections acquired in other municipalities.

On 2026-03-31 23:39:18, user Kamal Md Mostafa wrote:

Please read and cite the updated and peer-reviewed version of the article. <br /> Titled "Environment-dependent mutualism–parasitism transitions in the incipient symbiosis between Tetrahymena utriculariae and Micractinium tetrahymenae." <br /> Now published in The ISME Journal

On 2026-03-31 16:00:42, user Vasan wrote:

https://link.springer.com/article/10.1186/s12913-026-14420-6

On 2026-03-30 14:23:59, user M. Flores wrote:

Very interesting dataset with translational implications! I wonder if similar molecular signatures exist during collateral remodeling in other tissues/organ types?

On 2026-03-29 16:38:24, user Masha Niv wrote:

now published in Nature Communications https://www.nature.com/articles/s41467-024-54157-6

On 2026-03-29 15:08:20, user REMeD wrote:

I think this preprint has been published here: 10.1016/j.devcel.2024.06.011

On 2026-03-27 21:42:54, user Priya Banerjee wrote:

Author's note (Priya Banerjee): We would like to note that an assembly error in Figure 3B is present in this preprint version of the manuscript due to an error during Figure 3 preparation. This issue has been corrected in the published version of the paper in Nature Communications (link below). As the article is now formally published, the BioRxiv version has not been updated. The correction does not affect the interpretation of the data or the conclusions of the study.

https://www.nature.com/articles/s41467-026-69244-z

On 2026-03-27 17:27:32, user Markus Kalkum wrote:

Now available as a peer reviewed article on:<br /> https://doi.org/10.3389/fcimb.2025.1671048

On 2026-03-27 15:07:21, user David Cooke wrote:

Great use of an enclosure to quantify spread of eDNA. Wondered in what form DNA is being released and detected. Faeces, fur, skin cells? Realise it would be speculation but some comment could be warranted. BTW legend text in Fig 4 confused me for a while as colour coding did not match image.

On 2026-03-27 07:33:01, user Clashing_titans wrote:

Interesting short report prvodong some specific insight! An exciting update to the growing KD-Th17 story writing itself in the literature.

On 2026-03-27 01:23:51, user Riss Kell wrote:

Please note this article has been published in Frontiers in Immunology as of 3/18/26: https://doi.org/10.3389/fimmu.2026.1787989

On 2026-03-26 20:12:55, user Elisha Krieg wrote:

This article has been published with peer review:

Gupta, K. & Krieg, E. Y-switch: a spring-loaded synthetic gene switch for robust DNA/RNA signal amplification and detection. Nucleic Acids Research 52, e80 (2024). https://doi.org/10.1093/nar/gkae680

On 2026-03-26 20:08:08, user Elisha Krieg wrote:

This article has now been published:

Hsiao, S.-K., Mukenhirn, M., Werner, C., Honigmann, A. & Krieg, E. Reconfigurable Microenvironments Uncover Mechano-Sensing Timescales and Direct Cell Polarity. Advanced Functional Materials 35, e13028 (2025). https://doi.org/10.1002/adfm.202513028

On 2026-03-26 20:05:55, user Elisha Krieg wrote:

This article has now been published:

Speed, S. K., Peng, Y.-H., Atabay, A., Gupta, K., Müller, T., Fischer, C., Hermes, I. M., Sommer, J.-U., Lang, M. & Krieg, E. Assembling a True “Olympic Gel” From over 16 000 Combinatorial DNA Rings. Advanced Materials 2026, e20549. https://doi.org/10.1002/adma.202520549

On 2026-03-26 14:45:44, user Virginia G. de Yebenes wrote:

Accepted for publication in Leukemia (March 2026)

On 2026-03-26 13:17:18, user Javier Ruiz wrote:

This work has now been published in Cell Reports ( https://doi.org/10.1016/j.celrep.2026.117163 <br /> ) under the title: Ecological specialization and admixture drive genomic and phenotypic diversity in Yarrowia lipolytica.

On 2026-03-26 09:26:07, user TGS wrote:

Is the analysis code publicly available?

On 2026-03-26 00:03:05, user Martin Kekchidis wrote:

This is an insightful study. Some minor feedback:<br /> Discussion (p.12): "Similarly, selective deletion of Nos1 from Sst cells results in deficits in delta rhythms." <-- would be helpful to clarify what "deletion of Nos1" means.<br /> Duplicate citations: 24 & 103; 60 & 104<br /> p.6: "subsequent"->"subsequently"<br /> It would be interesting to explore if mPFC SST-Chodl have unique projections (including to hypothalamus), due to its role in regulating sleep. Tossel et al. 2023 suggests this may be the case: SST neurons from mPFC were found to synapse in LPO and LH. Although they found that chemogenetically activating mPFC Nos1+ neurons didn't promote sleep (except for a significant increase in NREM), I wonder if activating specifically SST-Chodl (a subset of Nos1 neurons) would have shown positive effects.

On 2026-03-25 21:43:59, user Misha Koksharov wrote:

Actually I hope that at least Alphafold-like tools well predict when proteins DO NOT dimerize. This could be very useful, for example, when looking at a transition from monomeric to dimeric versions in protein families when no experimental data is yet available.

There is also an important point that these tools don't use thermodynamics, so they are unable to predict, for example, monomerizing mutations: they still predict that most of the monomizerized protein variants (of fluorescent proteins, for example) form the original complexes perfectly fine, since the difference on the interface is very small.

On 2026-03-25 13:57:58, user Scott Horowitz wrote:

Curious if you've seen this paper- seems relevant: https://www.jbc.org/article/S0021-9258(20)33435-9/fulltext

On 2026-03-25 07:30:27, user Kathryn Wicht wrote:

Please see link to published article: https://www.nature.com/articles/s42003-025-08991-z

On 2026-03-24 00:02:50, user Rui Han wrote:

Hi ~ thanks for great job! <br /> I have a few questions regarding the Figure 6, in the same binding pocket, predicted amino acids involved in ligand binding are quite different when cellotriose, cellotetraose and XXXG were used as ligands? which amino acids will be your candidates when site-directed mutagenesis is performed? <br /> Have you tried to co-crystallize XXXG and this SBP? how was it? <br /> Based on the genome and transcriptome information, other potential proteins including intracellular enzymes for XyG complete degradation might be figured out to complete the proposed model in Figure 7.

some tips: Anaerocellum bescii and Caldicellulosiruptor bescii are the same strain, could you clarify in the manuscript? it is a little confusing that protein sequence conservation was carried out in two genera if readers dont know this strain is also classify into another genus.

Thank you!

Good luck!

Best wishes,<br /> Rui

On 2026-03-23 15:06:57, user Mathias Witte Paz wrote:

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0326299

On 2026-03-23 06:54:50, user Tomas Ros wrote:

Walkthrough video of the paper can be found here:

https://www.youtube.com/watch?v=qSM5narynzc

On 2026-03-22 18:30:41, user Wei Wang wrote:

A coding error leads to inflated performance and the uploaded dataset is inconsistent with the published preprint.

We have identified two significant concerns regarding this preprint:<br /> 1) Dataset does not match the preprint: The dataset uploaded by the authors to Zenodo https://zenodo.org/records/18019781 is missing approximately 8 million datapoints compared to what is reported in the preprint, and the preprint figures cannot be precisely reproduced from the shared code and data.<br /> 2) Author claimed ultra-resolution is partially due to a coding error: The authors introduce an F-to-D ratio to support their claim of “ultra-resolution” and “unprecedented” spatiotemporal resolution relative to prior work. However, this claim appears to stem from a bug in their code that artificially inflates their performance compared to earlier studies.<br /> Further details on each point are provided below.

1. Dataset issues<br /> A central claim of the preprint is the achievement of ultra-resolution and unprecedented spatiotemporal resolution and data volume. Main Figure 2 is largely devoted to this point. Specifically, Fig. 2K and 2L emphasize the quantity of data collected, and Fig. 2L states that the authors obtained 18 million 3D distance observations. The main text further refers to “>18 million datapoints.”

However, the dataset deposited on Zenodo ( https://zenodo.org/records/18019781) contains only 10,450,620 measurements, not >18 million. Thus, nearly half of the reported data appears to be missing. This discrepancy has also been raised on PubPeer ( https://pubpeer.com/publications/5E2C872645F18730F49DCA98D54026) but, to our knowledge, has not received a response from the authors.

A fundamental principle is that the preprint should accurately reflect the deposited data, and the deposited data should correspond to the preprint. We therefore respectfully request that the authors either revise the preprint to align with the dataset currently available on Zenodo, or update the Zenodo dataset so that it fully matches the published claims.

2. The claimed ultra-resolution appears to result from a bug in the code<br /> Main Figure 2 is devoted to demonstrating improved spatiotemporal resolution relative to prior work. To support this, the authors introduce a new metric termed the F-to-D ratio, defined in Section 6.2 (page 22) of the Supplementary Information. Here, F represents frame-to-frame movement, and D represents the interquartile range (25th to 75th percentile) of 3D distances.<br /> Figure 2 emphasizes that the F-to-D ratio in this study (approximately 0.15, see Fig. 2I) is substantially lower than in prior work (approximate 0.5), supporting the claim of vastly superior resolution.

However, the reported F-to-D ratio of approximately 0.15 does not appear to be correct and instead seems to result from a bug in the published analysis code. Specifically, the interquartile range (IQR) for the authors’ own data is calculated incorrectly. The code implements:

IQR(X) = Q75(X) − min(X),

with X representing 3D distance values. This corresponds to the 0th to 75th percentile range, rather than the correct interquartile range (25th to 75th percentile). Importantly, this issue affects only the authors’ own data. For previously published public datasets, the IQR is computed correctly as:

IQR(X) = Q75(X) − Q25(X).

This discrepancy leads to artificially reduced F-to-D ratios (approx. 0.15) for the authors’ data, while prior datasets yield ratios in the range of approximately 0.5. When the IQR is computed consistently (25th to 75th percentile) for all datasets, the F-to-D ratio for the authors’ data falls within approximately 0.25–0.5. Because the F-to-D ratio is central to establishing the preprint’s claim of “unprecedented spatiotemporal resolution,” this coding error substantially undermines a key conclusion of the work.

This issue has also been discussed on PubPeer. In Comment #4 ( https://pubpeer.com/publications/5E2C872645F18730F49DCA98D54026) , Chromohalobacter israelensis attempted to reproduce the published F-to-D ratio but obtained values of 0.25–0.5 instead of those reported in the preprint.

The bug itself was identified in Comment #6 on PubPeer ( https://pubpeer.com/publications/5E2C872645F18730F49DCA98D54026) by Eontia ponderosa. As shown on GitHub ( https://github.com/BoettigerLab/TRACK-IT/blob/master/FigureCode/Fig2_Ultraresolution_and_Supp1.m , commit 9fb3d2f), line 54 uses:

iqr_h(c) = quantile(trac,mx,2) - min(trac)

for the authors’ own data, effectively computing the 0th to 75th percentile range and thereby inflating their apparent performance.

By contrast, GitHub line 119 uses:

iqr_h(c) = quantile(trac,mx,2) - quantile(trac,mn,2);

for public datasets, correctly computing the 25th to 75th percentile interquartile range.

Thus, the claimed superiority is at least partly attributable to applying a more favorable (and incorrect) formula to the authors’ own data, while applying the correct formula to the public data used for comparison.

Summary<br /> The dataset deposited at https://zenodo.org/records/18019781 does not appear to match the dataset described in the published preprint. In addition, the central claim of “ultra-resolution,” as presented in the title and supported by the F-to-D ratio analysis in Figure 2, appears to be largely due to a coding error that applies different formulas to the authors’ own data and to public comparison datasets, thereby artificially inflating the reported performance. These have been raised on PubPeer, but not received a response. A response from the authors appears necessary as does an update to this preprint.

On 2026-03-21 15:33:47, user Ben Scott wrote:

Really interesting paper. I had some questions/suggestions for investigation that I hope you find useful. Particularly in relation to Figure 4

• Does the location of the variant in the primary amino acid sequence have any connection to concordance? I would assume that variants at N and C termini have more accurate predictions of function (generally not impacting protein function), and therefore methods may agree more.

• Do the number of known protein-protein interaction partners for each protein analysed have an effect on model concordance, and also on the predicted severity of variants? Since many models use this information to compute variant effect, it may reveal concordance. This would of course be highly dependent on the accuracy of such protein-protein interaction data, which is biased towards a fraction of proteins

On 2026-03-20 22:52:59, user Jayson Zhao wrote:

scikit-bio paper is out, you can cite it here: https://www.nature.com/articles/s41592-025-02981-z

On 2026-03-20 21:27:49, user disqus_2UwkiK23H2 wrote:

Table 1 says the Alpha Amylase one to many set has 412 train/77 val/3217 test, but the downloaded csv from the website has 3218 entries with "train" in the column and 488 entries with "test" in the column. Are these flipped?

On 2026-03-20 19:21:30, user Misha Koksharov wrote:

I hope that at least when Alphafold-like tools predict that proteins do not dimerize - it is probably indeed so. <br /> This could be very useful, for example, when looking at a transition from monomeric to dimeric versions in protein families when no experimental data is yet available.

There is also an important point that these tools don't use thermodynamics, so they are unable to predict, for example, monomerizing mutations and still predict that most of the monomizerized protein variants (of fluorescent proteins, for example) form the original complexes perfectly fine, since the difference on the interface is very small.

On 2026-03-20 16:01:10, user Leyla Ovchinnikova wrote:

Thank you for this valuable work. Databases focused on autoimmune diseases are critically important, and we have been waiting for a resource like this for a long time. Bringing MS BCR-sequencing data together in a uniform and accessible format will be highly useful for the community. I would like to share a brief comment on this research.

The two cited papers — Lomakin (Multiple Sclerosis Is Associated with Immunoglobulin Germline Gene Variation of Transitional B Cells) and Zvyagin (Deconvolution of B cell receptor repertoire in multiple sclerosis patients revealed a delay in tBreg maturation. 2022) — come from the same research group, use the same patient cohort, and rely on the same BCR sequencing data (ArrayExpress: E-MTAB-10859). As a result, the number of sequences reported in Lomakin and Zvyagin should be identical.

If the sequence counts shown in Figure 1B are taken directly from the articles, the observed discrepancy likely stems from different filtering criteria applied in each study. Nevertheless, because the underlying raw sequences are identical, the sequence numbers should match.

Additionally, the patients in these studies can be categorized under RRMS (blood) as well, and not exclusively under the total MS group. It is important to note that Highly Active Multiple Sclerosis (HAMS) and Benign Multiple Sclerosis are not distinct forms of the disease; instead, they describe, respectively, a severe and treatment-resistant subgroup or a mild clinical course within Relapsing-Remitting Multiple Sclerosis (RRMS).

It is indicated in the article (Table S1) that the BCR datasets from Lomakin and Zvyagin partially overlap, and different SRA references are provided – this should also be corrected. Furthermore, as stated in ArrayExpress (accession E-MTAB-10859) and the study "Deconvolution of B cell receptor repertoire in multiple sclerosis patients revealed a delay in tBreg maturation," different subpopulations of peripheral B cells were analyzed: total peripheral B cells and transitional peripheral B cells with the phenotype CD19(+)CD24(hi)CD38(hi). It is not entirely clear from the article whether both of these populations were included in the database or only one of them?

On 2026-03-20 15:51:25, user Michael Hothorn wrote:

Dear Dr. Wang, thank you for your comments on our preprint. We have decided to experimentally address the different points, the outcome of which you can find in our updated preprint. Please find a point-by-point response below:

1. BSL/PPKLs are not plant orthologs of human PP1s,. Plant genomes encode many PP1s that are more similar to human PP1s than BSLs to human PP1s in the catalytic domain.
2. BSLs/PPKLs are conserved in green algae, plants, and Apicomplexa, but are absent in fungi and animals (10.1128/mbio.02254-23).

Our response: Our structural (Fig. 1 H) and sequence analyses (SI Appendix Fig. S1) reveal that the phosphatase domain of BSU1 shares very strong structural homology with yeast, animal and plant (TOPP) PP1 enzymes. However, we agree that this does not imply that BSU1/BSLs are functional orthologs of PP1 in plants. We have added a comparison with the canonical TOPP family of PP1 phosphatases to our revised manuscript (revised Appendix Fig. S1), and the following statement to our discussion section (lines 450-451) “Whether Kelch phosphatases share redundant functions with type-one serine/threonine protein phosphatases (TOPPs) in plants remains to be characterized (102).”

1. Human PP1-gamma displayed tyrosine phosphatase activity (FEBS Letters 397 (1996) 235-238).<br /> AND
2. The peptide (RKLRRKYGKRGSY) synthesized and tested in this study is based on a substrate sequence of animal PP1, and is distinct from the two reported BSL substrates, BIN2 (anisyicsrfy, Nat Cell Biol 11 (2009):1254-60) and CDKB (grgtygkvyk, Nat Plants, 2025 Nov;11(11):2395-2408), which share some sequence similarity (Maybe BSLs/PPKLs are sequence-specific, not residue-specific phosphatases? Both pT14 and pY15 of CDKB seem to be dephosphorylated – see Fig. 5d in Nat Plants, 2025 Nov;11(11):2395-2408).

Our response: We have synthesized the suggested substrate peptides for the reported natural substrates AtBIN2 and CrCDKB1 suggested by you, and performed enzyme kinetic phosphatase assays. The new data are shown in revised Fig. 1J and K. Fully consistent with our enzyme assays using designed synthetic substrate peptides, we found no detectable activity against the suggested BIN2 peptide containing pTyr200 (revised Fig. 1J and K). The corresponding statement in the text reads (lines 159-161): “In agreement with the dephosphorylation experiments using full-length BIN2 as substrate (Fig. 1G), BSU1 showed no detectable activity against a short BIN2 substrate peptide containing pTyr200 (193-KGEANIS(pY)ICSRFYR-207) (Fig. 1 J and K).”<br /> This experiment is also consistent with our dephosphorylation assay of full-length BIN2 shown in Fig.1G. We have added another enzyme assays to revised Fig. 1G, which demonstrates that also full-length BSU1 is not able to dephosphorylate pTyr200 in BIN2, or in other words a putative tyrosine phosphatase activity of BSU1 is not encoded in its N-terminal Kelch domain. The corresponding statement in the result section now reads (lines 137-139):”No dephosphorylation of pTyr200 was observed when BIN2 was incubated with either the isolated BSU1 phosphatase domain, or with full-length BSU1, even at a 4:1 substrate-to-enzyme ratio (Fig. 1G).”

For CrCDKB1, two neighboring amino-acids have been reported to be phosphorylated in planta (doi: 10.1038/s41477-025-02145-z), Thr14 and Tyr15. BSU1 could readily dephosphorylate pThr14 but not pTyr15 in CrCDKB1 substrate peptides, again confirming that BSU1 is a Ser/Thr, not a tyrosine phosphatase. The revised statement in the result section reads (lines 161-167): “Recently, the CYCLIN-DEPENDENT KINASE B1 (CDKB1) was identified as an in vivo substrate for the Chlamydomonas reinhardtii Kelch phosphatase BSL1 (58) (SI Appendix, Fig. S1). Specifically, Thr14 and neighboring Tyr15 in CrCDKB1 were identified to be hyperphosphorylated in a bsl1-1 mutant phosophoproteome (58). We found that BSU1 was able to de-phosphorylate a synthetic CrCDKB1 peptide containing pThr14 (kcat/KM ~1 × 103 M-1 s-1), but not the pTyr15-containing substrate peptide (kcat/KM ~0.04 × 103 M-1 s-1) (Fig. 1 J and K).”

Together, the newly added enzyme assays further support and substantiate our finding that BSU1 is a serine/threonine phosphatase, not a tyrosine phosphatase targeting BIN2 pTyr200. Our revised discussion statement reads (lines 371-374): “However, based on our structural and enzyme kinetic analyses, the BSU1 phosphatase domain consistently behaves as PP1-like Ser/Thr phosphatase and we could not detect dephosphorylation of BIN2 pTyr200, either in the full-length kinase or in a synthetic peptide containing this site (Fig. 1G, J–L).”<br /> The revised abstract reads (lines 52-53): “Taken together, our experiments suggest that plant Kelch phosphatases act as PP1-like cell cycle regulators, rather than as tyrosine phosphatases in brassinosteroid signaling.”

1. PP1 substrate specificity varies with metals (Journal of Inorganic Biochemistry 149 (2015) 1–5), and metal loading to PP1 depends on chaperones (FEBS Letters 598 (2024) 2876–2885).

Our response: That is indeed an important point to consider, which we have addressed experimentally. We chose the preferred BSU1 pThr-containing substrate peptide and its corresponding pTyr variant and assayed the ability of BSU1 to dephosphorylate these peptides in the presence of different metal salts. These assays confirm that that BSU1 is a Mn2+-dependent Ser/Thr phosphatase. Importantly, no significant enzyme activity is detected for the pTyr peptide, independent of the catalytic metal chosen. The new results are shown in newly added Fig. 1L. The corresponding statement in the result section reads (lines 167-172): “It has been previously reported that the substrate specificity and enzyme activity of animal PP1 are in part determined by the type of metal ions present in the catalytic center (Fig. 1I) (70). We therefore assayed BSU1-mediated dephosphorylation of the synthetic pThr- and pTyr-containing substrate peptides in the presence of different metal ions (Fig. 1 K and L). These assays confirmed BSU1 to represent a Mn2+-dependent Ser/Thr protein phosphatase (Fig. 1L).”

1. The authors acknowledged that "It remains formally possible that a single BSL2 or BSL3 allele can support BR signaling to wild type-like levels (Fig. 2 D and E).”. I was confused by other statements inconsistent with this.

Our response: Our structural and biochemical analyses of BSU1 show that this enzyme is not mediating BIN2 dephosphorylation at pTyr200. The reanalysis of the bsu1-D allele in addition suggests that BSU1 acts as a neomorph in bri1-5. However, we agree with you that this does not exclude the possibility the BSU1 and BSLs redundantly regulate other aspects of BR signaling, such as the direct dephosphorylation of BZR1 (as shown in Fig. 2G and H). To accurately reflect the outcome of our reverse genetic analysis, we have therefore rephrased our discussion section. The revised text reads (lines 417-419): “However, our genetic experiments and previous studies (37) do not provide compelling evidence for the involvement of these broadly acting Ser/Thr phosphatases in BR signaling.” This leaves the possibility that other bsu1 bsl alleles in Arabidopsis and/or in other plant species, or other assay conditions may still reveal functions for BSU1 or BSLs in BR signaling.

Thank you for critically reading our manuscript and for your suggestions. On behalf of the authors, Michael Hothorn, corresponding author, University of Geneva, CH

On 2026-03-20 14:31:34, user Gauri Paduthol wrote:

Recently published in Nature communications - https://doi.org/10.1038/s41467-026-68573-3

On 2026-03-19 22:55:00, user Rachel Thayer wrote:

This article is now published in Molecular Biology and Evolution: https://doi.org/10.1093/molbev/msaf210

On 2026-03-19 20:13:03, user FAISAL MADANAT wrote:

This preprint presents a convincing genetic case that UHRF1 is required for RB progression and contributes to a tumor-promoting inflammatory microenvironment.<br /> In the Results section describing Fig. 2C, the authors state that “there were no significant differences” in the expression of multiple retinal markers between Uhrf1 cKO and Cre-negative controls. However, in the Fig. 2 legend, there are reported significant changes (e.g., increases in Chx10 and Prkca and a decrease in Rho in EGFP+ cells). This creates confusion on whether the claim is meant to apply to whole retina qPCR or recombined (EGFP+) populations specifically. It would help to clarify in the text what population was analyzed for the “no significant differences” statement (whole retina vs sorted EGFP+ cells), and to align the wording with the legend.

Figure 6 and the model in 6F suggest that loss of UHRF1 reduces NF-κB pathway activity and chemokine secretion, which then reduces microglial infiltration. While the downstream observations are convincing, the upstream mechanism remains unclear: how does UHRF1 increase NF-κB signaling in retinoblastoma cells? A useful follow-up would be genome-wide mapping of UHRF1 occupancy (e.g., CUT&RUN or ChIP-seq) combined with RNA-seq integration to identify direct targets, especially any NF-κB regulators that could explain the pathway shift.

Because most experiments are in mouse genetics and mouse-derived tumor cells, it would strengthen the translational argument to include at least one human system (e.g., patient-derived xenografts, primary retinoblastoma samples, or human retinoblastoma cell lines with UHRF1 alterations, especially for the NF-κB/chemokine/microglia axis. Even limited validation that UHRF1-high tumors show correlation to NF-κB/chemokine signatures in human data would help.

Overall, I found the study interesting and the genetic result (tumor suppression upon Uhrf1 loss) particularly strong.

On 2026-03-05 18:17:45, user Pearl Macias wrote:

In your results (page 6), the authors state that "There were no significant differences in the expression of Arr3, Calb1, Chx10, Crx, GS, Hes1, Lhx1, Nr2e3, Pax6, Prkca, Prox1, Rcvrn, and Rho between Uhrf1 cKO and Cre-negative controls" for figure 2C. But when you go to the figure, we do see a difference in Chx10, Prkca, and Rho and this difference is even mentioned in the description section below figure 2C, so this is a bit confusing.

For figure 6F the authors propose a mechanism of UHRF1's role in retinoblastoma, however it is still unclear how UHRF1 turns on the NF-kB inflammatory pathway. I recommend for a future direction to perform a ChIP-seq assay to map where UHRF1 binds to the genome in retinoblastoma tumor cells. This may show us how UHRF1 blocks an inhibitor of the NF-kB, to constantly promote this pathway.

Since you only test experiments using mouse models I would recommend performing these tests on patient derived xenografts for patients that have retinoblastoma.

For some of the western blots like figure 6E, I still see a band for the UHRF1 knockout and it is not explained why we still see expression. I recommend addressing this result in your discussion because it is one of your more important assays that is used to confirm that removing UHRF1 decreases tumor progression.

Overall, I liked your paper and I found it interesting learning about some of the epigenetics behind what contributes to retinoblastoma.

On 2026-03-19 16:06:31, user ROBIN MATHIEU wrote:

Paper has now been published in Frontiers in Aging.<br /> DOI:<br /> https://doi.org/10.3389/fragi.2026.1715723

On 2026-03-19 14:03:34, user Jonathan Legrand wrote:

That's a much needed initiative, congratulations! Will you release the list of the ImmPort study IDs that you used for pretraining?

On 2026-03-17 21:38:34, user markus wrote:

This is a review of the preprint “Mental calculation speeds surpass known limits for high-level cognition”, by Wiederhold, B., M. Stemmler, and A.V.M. Herz. 2026. bioRxiv.<br /> https://www.biorxiv.org/content/10.1101/2025.05.28.656137v2

This article regards the information rate of brain processing in humans. It is organized in large part in reference to a recent review article (ref 1), which claims that the information throughput of a human being is limited to about 10 bits per second. We are the authors of that article and so have a special interest in the topic. The present report represents a valuable and substantive addition to the field, including new experimental data, new analysis, and even a mathematical theorem.

We summarize briefly the claims of our article (ref 1) to the extent that they relate to the present report. The literature review from much of the past century suggests that the total information throughput of a human being – from sensory input to motor output – is limited to about 10 bits per second. Remarkably, this is independent of the precise nature of the task. It applies equally for simple motor challenges, like moving a ballpoint pen from one small circle on the page to another, as to complex tasks, like solving a Rubik's cube while blindfolded. We further argued that this throughput is the result of many stages of lossy information processing, starting with the visual image sampled by the cones of the retina. Along the way one finds very high rates of information processing, for example the transformation from cone signals to retinal ganglion cell signals proceeds at megabits per second. Each stage discards much of the information received from the previous one, leading eventually to an abstract mental representation of the digits on the screen, irrespective of their size, color, or position. All those intermediate processing stages with high information rates are subconscious.

The present report considers specifically information processing by mental calculators: individuals who compete in contests to calculate sums or products or square roots in the shortest possible amount of time. The authors explain the special techniques these human calculators employ. Then they compute the Shannon information needed in order to complete all these intermediate steps that lead to the final result. They argue that these steps are a form of conscious mental processing, and therefore the total information rate they obtain should be considered a lower bound on the conscious throughput rate of a human being. In fact, the rate they compute of 93 bits per second (or perhaps 215 bits per second, see questions below) is much greater than what the literature cites as the throughput rate of a human being. This result would extend in interesting ways our understanding of how information processing is limited in the human brain. In particular the article recommends a reliable method for estimating the entropy of an algorithmic step during cognition.

Main recommendations

1. Definitions of entropy<br /> The authors define a number of different entropy measures related to the mental calculation task: input entropy, extended input entropy, algorithmic entropy, and output entropy. They each show some systematic relationship to the calculation time. Among these, only the algorithm entropy and the output entropy seem related to Shannon information, defined as the reduction in entropy through the communication process.

The output entropy divided by the calculation time is the throughput rate, to be compared to the values obtained on other tasks, as reviewed in ref 1.

The algorithmic entropy is the sum of the output entropies of intermediate calculations performed by the mental calculator. This total divided by the calculator time represents the information rate of those mental calculations. This number can be legitimately compared to throughput rates or other mutual information rates in the literature. The highest algorithmic rate of 93 bits/s (Table 1) is in fact considerably higher than the typical throughput rate of 10 bits/s. The power law relationship between calculation time and algorithmic entropy (Fig 1B) is impressive.

We do not understand why one should sum the various entropy measures (line 96, Fig 2), leading to the “peak rate” of 215 bits/s. This sum does not seem to represent mutual information, in which case it cannot be compared to the numbers in ref 1. This summing step should be explained better, if it is in fact justified.

2. Conscious vs unconscious processing<br /> The distinction between unconscious and conscious processing deserves some elaboration. As noted above, the preconscious processing rates in the early visual system are enormously higher than anything discussed here. The rate of processing grows systematically as one reaches further and further back into the system from the final output of the mental calculation to the sensory signals. As the authors tease apart the steps of mental calculation preceding the result, is it obvious that all those steps are “willful” or “conscious”? For example, should we apply that label to lookup in the memorized product tables (line 138), or do those precomputed results present themselves to the subject automatically in a subconscious way? Some introspection from these performers might be useful here.

As explained in the text, the mental calculators lose track of those intermediate results as soon as they are used, and cannot recall those numbers even one or a few seconds later. One could argue that such a fleeting internal result is not really experienced consciously. Instead the mental calculator has trained a reflex-like subconscious procedure that delivers these intermediate results and uses them for further calculations. This would be more akin to the rapid reflexes trained by a competitive tennis player returning a serve, which also do not involve conscious deliberation in the loop.

If, in fact, these intermediate sums are subconscious events, then they simply form part of the long subconscious processing chain starting in cone photoreceptors, that we already know involves much higher processing rates. More discussion of this issue would help the report.

Other suggestions<br /> Line 88: “The resulting ratios are conservative measures, as they assume all processes run in parallel, instead of sequentially”. That seems like an unreasonable assumption, because some steps of mental calculation depend on the results of preceding steps. On the other hand, the total information rate one computes is the same whether the sub-processes are carried out in series or parallel.

The algorithmic entropy estimates are obtained based on an error-free performance in the shortest time of many attempts (see line 147). In practice the mental calculations come with some error rate and the performance times scatter. How large would the correction be if one takes finite error rates and typical calculation times into account, for example using results during training sessions, rather than only the medal-winning performance.

The output rates are almost all below 10 bits/s, and as the authors note (line 95) “exhibit almost no correlation with calculation time.” This seems to extend the observation of ref 1 – namely that the throughput rate is limited to 10 bits/s – to a whole new range of cognitive tasks. This seems worth mentioning. In table 1, please list the output rates in all cases, to allow this comparison.

Line 270: It is unclear why the algorithmic entropy of one lookup step should equal the size of the lookup table. The Shannon information between the inputs to the lookup (previous intermediate result and new number) and the output is simply the entropy of the output (the result of the operation), not the combined entropy of input and output. The subsequent section “Algorithm” (line 316ff) seems to only count the entropy of the output.

Small things<br /> The main text would benefit from more precise references to Method subsections.

Line 136 “as competitors must attend to each digit shown, the estimated visual processing rate should be reliable.” Meaning unclear.

Line 174 “eights”

Line 248: maybe add “where $H_i$ is the entropy in bits.”

Line 302 “An outline of the entropy of mental calculations has previously been given by Timms.” This reference is a blog post. The bit counting there is somewhat ad hoc and poorly justified and doesn’t necessarily match the authors’ approach.

Line 346 “The output entropy can be considered a lower bound on the “perceived” entropy”. Meaning unclear.

Line 402: “first” should be “last”?

Line 440: “the multiplicities above reduced the number of distinct products by a factor of roughly four to six.” Meaning reduced the entropies by about 2 bits?

Line 462: “ · log2(7)” should be “3 · log2(7)”?

Line 556: “multipliation”

- Markus Meister and Jieyu Zheng, Caltech

On 2026-03-17 19:17:16, user Sarah Lindsey wrote:

Great paper! Would the E2-induced decrease in stiffness then be enhanced in an OVEX SMC-ERa-KO? Did you look at whether SMC-ERa-KO enhanced the expression of any other hormone receptors? Which of these is most likely - that ERa is constitutively active, is responding to local E2 synthesis, or is being activated by other factors such as IGF-I?

On 2026-03-17 18:12:11, user markus wrote:

This is a review of the preprint Connectome of a human foveal retina, by Kim, Y.J., O. Packer, T. Macrina, A. Pollreisz, C.A. Curcio, K. Lee, N. Kemnitz, D. Ih, T. Nguyen, R. Lu, S. Popovych, A. Halageri, J.A. Bae, J. Strout, S. Gerhard, R.G. Smith, P.R. Martin, U. Grünert, and D.M. Dacey. 2026. https://www.biorxiv.org/content/10.1101/2025.04.05.647403v5 , version of Jan 28, 2026.

This report describes a remarkable new data set, which is being made available for public use: The connectomic reconstruction of the neuronal network in a small volume of a human fovea, based on 3D electron microscopy.

The fovea is the very central portion of our retina, responsible for all our high-acuity vision at the center of gaze. In some sense, this is the part of the retina most precious to us, and an inordinately large fraction of the visual brain is busy processing signals from the fovea. Much earlier research has focused on retinas of model animal species, in particular the closely related macaque retina. This is the first time a full sample of neurons, processes, and synapses has been assembled from the human fovea. I expect that many interesting discoveries will come from exploring this data set.

In addition to describing the process used in the connectome assembly, this article also reports some early results of the analysis. Topics include the electrical coupling between cones of different spectral sensitivity, the connections of horizontal cells to different cone types, the connectivity of the AII amacrine cells, and the number of ganglion cell pathways that encode the foveal output. My review and suggestions here focus on this part of the article reporting scientific claims.

The data set<br /> Is this the same data set as used for the reports in Kim 2023 (ref 44) and Kim 2024 (ref 106)? If so, please state that early on, and explain what are the methods enhancements since then.

Some of the results overlap with the Kim 2023 report: Fig 2 has a close relative in Kim 2023 Figs S2 and S3. Fig 7 top shows some of the same things as Kim 2023 Fig S8.

Line 124 "including those related to processes extending into or through the volume but not linked to cell bodies": It would be useful to state, for each cell type, how many complete neurons are in the 180 x 180 um EM volume vs those with only partial dendritic fields (e.g. Fig 6H).

Line 330 "Moreover, all cell types in the HFseg1 volume are identified": Meaning? Presumably not every process is identified.

Coupling of S to LM cones<br /> Line 127: "17 cone pedicles were identified as short wavelength-sensitive (S) cones (Fig. 2A)(44)." Please state how this was achieved. Ref 44 says "A clear population of likely S cone pedicles was not conspicuous in the human retina". Identifying these is essential for the subsequent claims of cone coupling.

Line 221 "We found abundant S-LM cone contacts": The authors already reported this in ref 44. Please emphasize what is new here.

Model of the S-cone spectral sensitivity<br /> The coupling between S and LM cones is a puzzle. In other retinas, biology goes to some trouble to avoid it. The principle extends to amphibians, in fact it was first reported in turtle (Detwiler 1979). The unexpected departure in human retina has been reported and debated for some time. One would hope that the modeling effort here would shed additional light on the issue.

Line 412: "A network model of this connectivity showed that ... mixing S and LM cone signals can alter the action spectrum of the S cone light response." This much was obvious from the outset and didn't require any modeling. *How much* does it alter the action spectrum, and does that explain any puzzles from human color vision? Given all the quantitative detail that went into the model, one would like to see a quantitative interpretation of the result.

Line 421: "Coupling among the S, L, and M cones prior to this connection may serve to preserve the smaller S cone signal within the H1 cell network." The meaning here is unclear. How does mixing in other signals serve to preserve the S cone signal?

Constraints on the number of ganglion cell pathways<br /> The authors report finding "only 11 visual pathways, with 5 high-density RGC pathways accounting for over 95% of foveal output to the brain: a dramatic contrast to the 40+ ganglion cell types recognized in mouse retina." (line 50). This difference between human fovea and mouse may well be correct, but the present study does little to support the claim. The volume included 599 RGCs (line 141). If we subtract the obvious midgets and parasols, that leaves only 37 RGCs. In the mouse, many of the RGC types have a prevalence of less than 1%. Under a Poisson distribution, there is therefore a good probability of missing such rare types within the relatively small sample here.

The complementary analysis of the "vertical connectome" (line 294ff) rests on the assumption that every cone must ultimately have an excitatory pathway to every ganglion cell. That is an assumption, not a natural law. If the probability of connection is less than 1, then tracing the output of just 3 LM cones may well miss some ganglion cells (line 319).

That said, the transcriptomic analysis of human fovea in Yan 2020 poses much more stringent limits on the existence of rare RGC types, because it included 11404 RGCs. The present work is consistent with the idea that there are only 11 or 12 pathways in human fovea, but doesn't strengthen the evidence much.

Synaptic circuitry of AII cells<br /> The textbook picture is that the AII cell feeds signals from rod bipolars into cone on and off bipolars, with opposite sign, via gap junctions and glycinergic synapses respectively. When rod signals aren't available, e.g. at very high light levels, the cone on bipolar excites the AII through the gap junctions, which leads to crossover inhibition of the cone off bipolar cells.

The current report (line 282ff) adds at least two non-canonical connections: (1) ribbon synapses from cone off bipolars and (2) ribbon synapses from cone on bipolars, both presumably glutamatergic and excitatory. Both these components have been described before in the mammalian retina. However, in the presence of the overwhelming input from rod bipolars, they were seen as a minor distraction; for example Strettoi 2018 say that the large majority of inputs in the On sublamina is from rod BCs, with only 23% from cone BCs. This raises a few questions about the present data set from human fovea:

What fraction of the ribbon synapses in the on sublamina are from rod bipolars? Clearly there are some rod bipolars in the data set (Fig 4B), but no results are presented about them. One would like to hear more.

The AII in Fig 9 gets more synapses from Off bipolars than from On bipolars. So should one expect the human AII cell to have a net Off response, unlike those in other retinas? Or is it conceivable that the postsynaptic response at Off bipolar synapses is sign-inverting?

Line 290 "to provide an inhibitory pathway from ON to OFF bipolar cells under photopic conditions". But also an excitatory pathway from OFF to ON bipolars? How can one reconcile that with the idea of "crossover inhibition"?<br /> Seeing how intensely the AII cell has been studied across species, it would be remarkable if it plays a very different functional role in the human fovea. More detail on this would be useful.

Abstract<br /> As it stands, I feel the abstract does not optimally highlight the unique contributions of the report. The biophysical model is new, but in the present form doesn't add new insight (see above). The small number of visual pathways compared to the mouse retina has been reported elsewhere with stronger numerical evidence (see above). There is reference to "novel synaptic pathways" and "distinctive features of human neural circuitry"; perhaps these could be spelled out? For example the cone connectivity of H1 vs H2 cells seems to be a new result (line 255ff). Similarly the observation of the "monopolar bipolar" cells previously seen in mouse retina (Fig 4B, line 270ff).

Other things<br /> Where is "SI Appendix" and Tables 1-4?<br /> Line 121 "We were therefore able to use the original segmentation": Meaning unclear.<br /> Line 233 "We used a fixed ratio of L to M cones of ~1.7:1": What motivated that choice? Empirically that ratio varies a lot across humans. How much does it matter, i.e. what would be the effect on the final modeling results of different ratios?<br /> Line 291 "while at the same time suggesting specialized synaptic alterations for the AII cells in the human fovea in relation to the midget circuit": Unclear to what this refers.<br /> Line 335 "The HFseg1 volume permits a comparison of the two approaches": Which two?<br /> Line 422 "This in turn may be critical for the unique requirements of human color vision": Unclear. What are those unique requirements? Something imposed by the ecology of humans vs macaques?<br /> Line 521 "White arrowhead shows branching processes of unknown significance arising from the cell body": These look remarkable. Other RGC types don't do that? Is there functional circuitry within the RGC layer? What is at the end of those processes? Gap junctions? This seems worth following up.<br /> Line 198 "Vesicle clouds were detected with high sensitivity": Please explain "vesicle cloud"? What is its functional meaning?<br /> Line 740 "Vesicle cloud detection and conventional synapse assignment in the IPL." These methods seem rather ad hoc. What is the ground truth here? How did the human annotators distinguish "vesicle clouds"?<br /> Line 620 "as previously described1.": Meaning of the superscript 1?<br /> Line 839 "La rètine des vertèbrès": Please check the direction of those accents.<br /> Line 846 "et al.": Now that we're not killing trees anymore for publications, there's no cost to listing all the authors of a cited paper.

• Markus Meister, meister@caltech.edu